Changes in extratropical storm track activity and their

Institut für Physik und Astronomie Arbeitsgruppe Prof. Dr. Anders Levermann und Arbeitsgruppe Dr. Dim Coumou

Changes in extratropical storm track activity and their implications for extreme weather events

Kumulative Dissertation zur Erlangung des akademischen Grades “doctor rerum naturalium” (Dr. rer. nat.) in der Wissenschaftsdisziplin “Klimaphysik” eingereicht an der Mathematisch-Naturwissenschaftlichen Fakultät der Universität Potsdam

angefertigt am Potsdam-Institut für Klimafolgenforschung

von

Jascha In-su Lehmann Potsdam, im März 2016

Abstract The climate has always been changing but the current speed under human influences greatly exceeds past warming events. This has consequences for thermodynamic and dynamic processes in the climate system and implications for the hydrological cycle. Extreme weather events, like heat waves and heavy rainfall, have increased in intensity and frequency in recent decades. Thermodynamic drivers behind changes in extremes are reasonably well understood but dynamic mechanisms much less so. There is an urgent need for an improved understanding of their relative importance. Successful adaptation strategies will strongly depend on better information on how climate variability and in particular extremes events will change in the future. Extratropical storm tracks play a central role in understanding the Earth’s climate and its variability. They are a major component of the large-scale atmospheric circulation and their position and strength account for much of the day-to-day weather variability in the mid-latitudes. Moreover, extreme rainfall is often associated with strong storm activity. However, changes in simulated extratropical storm activity still reveal large uncertainties and knowledge about their influence on weather extremes is limited. This thesis aims at improving the understanding of these processes. It is based on two different approaches including record-statistics of rainfall extremes and the analysis of the storm track activity. First, I show that worldwide the number of record-breaking daily rainfall events from 1981 to 2010 is 12 % higher then expected in a stationary climate. The long-term increase has reached 26 % by 2010. I show that this increase is consistent with the thermodynamically expected increase in the atmosphere’s water holding capacity. Regional changes in rainfall extremes differ markedly across the globe with a tendency of wet regions seeing an over proportional increase and dry regions less so. This pattern is also reflected at monthly timescales where my analyses reveal substantial drying over Central Africa and significant increases in observed record-wet months in tropical monsoon climates as well as in the northern mid- to high latitudes. In the second part of my thesis I show that in the mid-latitudes such monthly rainfall extremes are strongly coupled to extratropical storm track activity as measured by the eddy kinetic energy (EKE). Moreover, EKE modulates continental temperatures, because storms bring relatively moist and moderate temperatures from the ocean to the continents. I show that summer EKE has declined over 1979-2014, which likely favors the buildup of heat and drought conditions over the continents. Likewise, cold spells in winter are associated with low EKE over parts of North America, Europe, and central- to eastern Asia. The observed weakening of summer EKE is associated with a weakening of the zonal wind related to a reduction in the equator-to-pole temperature gradient. This gradient reduces due to Arctic amplification. Consistently, I show that climate models project a robust weakening in summer EKE for the 21st century under a high-emission scenario which will further increase the risk of prolonged heat and dry periods over mid-latitude land regions.

Zusammenfassung Das Erdklima war schon immer Veränderungen unterworfen. Aber die derzeitige durch den Menschen beeinflusste Geschwindikeit übertrifft die letzten Erwärmungsphasen deutlich. Das hat Folgen für thermodynamische und dynamische Prozesse im Klimasystem und Auswirkungen auf den Wasserkreislauf. Extreme Wetterereignisse, wie Hitzewellen und Starkregen, haben in den letzten Dekaden an Intensität und Häufigkeit zugenommen. Ein besseres Verständnis dieser Ereignisse und der zugrundeliegenden, thermodynamischen und dynamischen Mechanismen ist dringend notwendig. Erfolgreiche Strategien zur Klimaanpassung werden davon abhängen, wie gut wir zukünftige Änderungen in der Klimavariabilität uns insbesondere in den Extremereignissen verstehen und vorhersagen können. Extratropische Sturmbahnen spielen eine zentrale Rolle für das Verständnis des Erdklimas und seiner Variabilität. Sie sind ein wesentlicher Bestandteil der großräumigen, atmosphärischen Zirkulation und ihre Lage und Intensität machen einen Großteil der täglichen Wetterumschwünge in den mittleren Breitengraden aus. So ist Starkregen oft mit einer erhöhten Sturmaktivität verknüpft. Änderungen in der simulierten Sturmaktivität sind jedoch mit großen Unsicherheiten verknüpft und ihr Einfluss auf Wetterextreme ist unklar. Das Ziel dieser Dissertation ist es, das Verständnis dieser Prozesse zu verbessern. Die Arbeit basiert auf zwei unterschiedlichen Ansätzen, der statistischen Analyse von Rekord-Ereignissen und der Analyse der Sturmaktivität. Zunächst wird gezeigt, dass die Anzahl von rekordebrechenden Starkregenfällen in den Jahren 1981 bis 2010 um 12 % höher lag, als in einem stationären Klima zu erwarten wäre. Der ansteigende Trend erreicht 26 % im Jahr 2010. Ich zeige, dass dieser Anstieg übereinstimmend ist mit der thermodynamisch erhöhten Wasserspeicherfähigkeit der Atmosphäre. Regionale Veränderungen im Starkregen variieren weltweit deutlich. Dabei zeigt sich die Tendenz, dass Starkregenfälle insbesondere in onehin sehr feuchten Regionen zunehmen. Dieses Muster deutet sich auch auf monatlicher Zeitskala an. Hier zeigt mein Analyse von Beobachtungsdaten einen signifikanten Anstieg besonders trockener Monate in Zentralafrika und besonders regenreicher Monate in den tropischen Monsoonregionen sowie in den nördlichen mittleren bis hohen Breitengraden. Im zweiten Teil meiner Arbeit zeige ich, dass in den mittleren Breitengraden extreme Regenfälle eng mit der extratropischen Sturmaktivität verknüpft sind. Diese lässt sich über die kinetische Energie der Eddies berechnen (Engl. Eddy Kinetik Energy (EKE)). Die EKE hat einen ausgleichenden Effekt auf die Landtemperaturen, da Stürme Meeresluft von den Ozeanen zu den Kontinenten tragen. Ich zeige, dass in den Sommern von 1979 bis 2014 die EKE abgenommen hat, was vermutlich die Entstehung von Hitze und Dürreperioden begünstigt hat. Gleichermaßen waren Kälteeinbrüche im Winter mit sehr geringer EKE in Teilen von Nordamerika, Europa und Zentralund Ostasien verknüpft. Die beobachtete Abnahme der Sommer EKE ist verbunden mit einer Abschwächung des zonalen Windes, die wiederum im Zusammenhang steht mit einer Reduktion des Temperaturgefälles zwischen Äquator und Nordpol. Dieser Temperaturunterschied wird durch die beschleunigte Erwärmung der Arktis reduziert. Als Konsequenz prognostizieren Klimamodelle eine weitere Abnahme der Sommer EKE für das 21. Jahrhundert in einem Szenarium mit stark ansteigenden Treibhausgasemissionen. Dies würde das Risiko für lang anhaltende Hitzewellen und Dürreperioden in den mittleren Breitengraden weiter verstärken.

Contents

1. Introduction 1.1. Motivation . . . . . . . . . . . . . . . 1.2. The general circulation . . . . . . . . 1.3. Extratropical storm tracks . . . . . . 1.4. Climate change and weather extremes 1.5. Scope and contents of the thesis . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

2. Original Manuscripts 2.1. Increased record-breaking precipitation events under global warming . . 2.2. Changes in record-wet and record-dry months in global land observations 2.3. The weakening summer circulation in the Northern Hemisphere midlatitudes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4. The influence of mid-latitude storm tracks on hot, cold, dry, and wet extremes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5. Future changes in extratropical storm tracks and baroclinicity under climate change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. Discussion and Conclusions

1 1 4 5 7 11 15 17 47 75 101 129 147

Appendix

155

Bibliography

217

Danksagung

223

1 Introduction

2

1.1

1. Introduction

Motivation

Climate change is one of the biggest global challenges of the 21st century for society, politics, and science. Fifteen of the sixteen warmest years on record have occurred since the year 2000 [World Meteorological Organization, 2015]. Ice caps are melting, oceans are acidifying, and tipping points have likely been crossed with uncontrollable consequences for the Earth system [IPCC, 2013; Joughin et al., 2014; Rignot et al., 2014]. Anthropogenic forcing has been identified as the main cause for rising global mean temperatures [IPCC, 2013]. Consequently, one of the new scientific challenges is to analyze how climate change may alter the intensity and frequency of surface weather extremes. The latter is important, because climate extremes generally impose the greatest impact on society and nature [Seneviratne et al., 2012]. Thus, public recognition of human-made climate change strongly depends on their perception of recent local extreme events [Hansen et al., 2012]. We need a better understanding of the underlying physical drivers within the climate system and their interactions in causing observed weather extremes. Only if this is understood we can address future scenarios and the impacts of climate change which will be of great importance for governments and decision-makers involved in risk management. The importance of extremes and large-scale dynamics is reflected by the Grand Challenges defined by the World Climate Research Program, which include “What do we understand about the interactions between large-scale drivers and regional-scale land-surface feedbacks that affect extremes and how can these processes be improved in models?”, “How has drought changed in the past and what were the causes, and how will it change in the future?”, and “How will clouds and circulation respond to global warming or other forcings?” [World Climate Research Programme, 2015]. In the last decades, there has been a worldwide accumulation of some types of weather extremes [Rahmstorf and Coumou, 2011; Coumou and Rahmstorf, 2012; Munich Re, 2015]. Robust evidence exists for a significant increase in the global occurrence of heat waves and heavy rainfall events [Coumou et al., 2013; Lehmann et al., 2015]. This has been reflected in an exceptional number of unprecedented climate extremes over the last years. Examples include the global-mean temperature of last year, which has by far been the warmest since weather records began with 2015 temperatures about 1°K above pre-industrial levels [World Meteorological Organization, 2015]. California has been suffering a severe drought lasting for several years now [Wang et al., 2014; Diffenbaugh et al., 2015] whereas intensive storms brought heavy rain to the UK leading to massive floodings in 2013/14 winter [Stephens and Cloke, 2014]. Part of the observed changes can be explained by

1.1. Motivation

3

Fig. 1: Illustration of the polar and subtropical jet stream. Represented by a snap shot of wind speeds in summer 2010 (source NASA [2014]). simple thermodynamic arguments. If the world is warming heat extremes are expected to occur more often and heavy rainfall extremes will likely increase in intensity, because a warmer atmosphere can hold more moisture which in turn fuels the intensity of heavy rainfall. However, pronounced regional differences in temperature and precipitation pattern do not seem to fit into this simple thermodynamic picture. To a large extend, such regional patterns are due to dynamical processes in the Earth system, such as the general atmospheric circulation. In the last years, great progress has been made in linking atmospheric circulation patterns to specific regional weather extremes including hot, cold, dry and wet extremes [Screen and Simmonds, 2014; Horton et al., 2015]. For example, the 2013/14 UK floodings were due to an unusual clustering and persistence of extratropical storms [Stephens and Cloke, 2014] while at the same time the Californian drought was associated with a lack of storm activity [Wang et al., 2014]. Whereas thermodynamic aspects of the climate are well understood, dynamical processes and feedbacks are still attached with large uncertainties [Shepherd, 2014]. Thus, how climate change alters the atmospheric circulation and what this implies for regional weather extremes is the central topic of this thesis. In the following, I will give a short introduction into the key dynamical features of the general circulation with a focus on the mid-latitudes (35° 75° North/South). Though I will discuss circulation phenomena like jet streams, extratropical storms and Rossby waves separately, it will become clear that all these processes strongly interact

4

1. Introduction

with each other. Afterwards, I briefly describe how climate change and particularly changes in the general atmospheric circulation affect surface weather extremes. I highlight present scientific caveats which lead to three research questions framing my thesis.

1.2

The general circulation

The general circulation in the atmosphere is in the first place caused by the rotation of the Earth and the meridional contrast in solar heating [Peixoto and H., 1992; Holton, 2004]. Strong winds develop in order to balance the temperature difference between the warm equator and the colder poles. Thereby they transport moisture, heat and momentum and determine local climate, day-to-day weather variability and weather extremes. A major component of the general circulation in the mid-latitudes is the jet stream. The jet stream describes regions of strong westerly winds at the upper troposphere with wind speeds of up to 160 m/s (Fig. 1). There are two types of jet streams, the subtropical jet stream and the eddy-driven jet stream (or polar front jet stream). The former is a consequence of the poleward movement of air masses at the upper branch of the Hadley circulation (Fig. 2). Here, warm air in the tropics rises and then moves poleward to balance the horizontal temperature gradient. At higher latitudes the air is cooled and sinks down to the surface before it returns equatorward forming the circulation loop of the Hadley cell. The air masses that travel poleward at the upper branch of the Hadley cell are diverted to the east (in both hemispheres) creating the subtropical jet stream. The additional zonal wind component is a result of conserving the angular momentum of air masses which travel from lower to higher latitudes and thus from larger to smaller rotation radii of the Earth. The underlying process is well known as the Coriolis effect. The subtropical jet is located at the boundary between warm air masses of the tropics and the colder polar air. The eddy-driven jet is located more poleward between the Ferrel cell and the Polar cell (Fig. 2). These westerly winds are fed by synoptic-scale eddies and thus - in contrast to the subtropical jet can extend down through the depth of the troposphere. Likewise, the eddy-driven jet can itself reinfluence eddies creating a positive feedback loop [Hoskins and Valdes, 1990]. The jet stream does not only have a zonal wind component but also shows a characteristic NorthSouth extension (Fig. 1). These wave-like features are so called Rossby waves. They can span around the whole globe exhibiting typically four to six large-scale meanders. When waviness becomes very pronounced warm air masses in the upward swing of the Rossby wave can form anticyclones (high pressure systems) and cold air masses in the downward swing can develop into cyclones (low pressure systems). Under certain conditions waves can get trapped and become quasi-

1.3. Extratropical storm tracks

5

Fig. 2 Schematic illustration of circulation cells in the troposphere. stationary in time [Petoukhov et al., 2013]. In such situations, high and low pressure systems persist longer than usual over particular regions in the mid-latitudes. This can lead to prolonged weather conditions and hence extremes on weekly to monthly timescales. For example, one or two days with summer surface temperatures in Europe above 30 °C might actually be favorable for some people, but continuing over weeks this might develop into a devastating heat wave.

1.3

Extratropical storm tracks

In the mid-latitudes (35°-70°N/S), the large-scale atmospheric circulation is generally characterized by (i) fast-traveling free Rossby waves (the so called synoptic transients) and (ii) quasi-stationary Rossby waves with smaller wave numbers in response to quasi-stationary diabatic and orographic forcing [Fraedrich and Böttger, 1978; Boer and Shepherd, 1983; Petoukhov et al., 2013]. In this thesis I focus on fast-traveling transient waves. They are associated with cyclones (extratropical storms) and anticyclones, which typically extend over a region of around 1000 km in diameter (synoptic-scale). Therefore, they are also referred to as synoptic-scale eddies. Synoptic-scale eddies show variability on 2 - 6 day timescales and dominate the day-to-day weather variability in the midlatitudes. In both hemispheres, they travel from west to east on preferable paths forming the storm track regions. The interplay between frequency and intensity of cyclones and anticyclones is described by the Eddy Kinetic Energy (EKE) which can thus be interpreted as a measure for the storm track activity. A common tool to extract the EKE is to apply a 2.5-6 day band pass filter to high-resolution wind field data [Blackmon, 1976; Yin, 2005; Ulbrich et al., 2008; Harvey et al., 2013]. EKE is a key quantity of the atmosphere and is used throughout most of the presented papers as a quantification of the storm track activity.

6

1. Introduction

Fig. 3: Storm track activity as shown by the climatology (1979-2014) of annual EKE at 500 mbar. Extratropical storms draw their energy from the temperature gradient between the warm equator and the cold poles. Thus, storms in the Northern Hemisphere mid-latitudes primarily form and develop over the oceans where the temperature gradient is strongest [Gulev and Zolina, 2001]. These regions are indicated by high baroclinic instabilities as represented by the maximum Eady growth rate [Lindzen and Farrell, 1980; Hoskins and Valdes, 1990]. To avoid possible confusion, please note that the Eady growth rate describes the growth rate of eddies but it is named after the British meteorologist Eric Eady who introduced this model in 1949 [Eady, 1949]. The magnitude of the Eady growth rate is determined by the static stability of the atmosphere and the vertical wind shear. Static stability depends on the vertical potential temperature gradient whereas the vertical wind shear is closely related to the horizontal temperature gradient via the thermal wind equation. In the Northern Hemisphere, this leads to two main regions of storm track activity, one over the North Atlantic sector and a second over the North Pacific (Fig. 3). Southern Hemisphere storm tracks are characterized by a rather zonal uniform distribution due to almost no orographic friction. Generally, storm track activity is stronger in winter when the horizontal temperature gradient and land-ocean temperature contrasts are larger. Extratropical storms transport energy, momentum and water vapor towards the poles, thereby reducing the temperature gradient that drives the storms.

1.4. Climate change and weather extremes

7

Climate change can affect the position and strength of extratropical storm tracks [Zappa et al., 2013; Lehmann et al., 2014]. Among other effects, global warming leads to a reduction in the horizontal temperature gradient in the lower troposphere in response to an over proportional warming of the Arctic. In contrast, warming in the tropics leads to increased vertical water vapor flux and latent heat release which increases static stability and the temperature gradient in the upper troposphere. This has competing effects on storm track development. Further, a surplus of moisture in a warmer atmosphere implies that a single storm can transport more energy towards the poles. This will affect the hydrological cycle with large implications for rainfall extremes. In the midlatitudes, where rainfall is strongly coupled to storm activity, an increased amount of atmospheric water vapor could lead to severe floodings, whereas enhanced evaporation in the subtropics could extend and amplify the intensity of droughts. In addition to the described circulation phenomena like jet streams, Rossby waves and extratropical storms, there exist large-scale teleconnection patterns which are able to describe important parts of the climate variability, especially in the Northern Hemisphere. A prominent example is the North Atlantic Oscillation (NAO) which reflects the pressure difference between the Azores High and the Icelandic Low. Both the negative and positive phase of the NAO are associated with basin-wide changes in the position and intensity of the jet stream and the storm tracks [Marshall et al., 2001]. During positive NAO index winters, the North Atlantic storm track shifts northward associated with mild and wet conditions over northern Europe and colder and drier conditions over southern Europe and the Mediterranean region [Hurrell, 1995; Thompson and Wallace, 2001; Bladé et al., 2012]. During negative winter NAO phases, the westerly winds weaken and cold Arctic air can dip farther south leading to anomalous cold winter temperature over northern Europe [Luterbacher et al., 2010; Bladé et al., 2012]. It should be noted, that there are various other teleconnection patterns covering other regions, seasons of the year or different aspects of the climate variability [Trenberth et al., 1998].

1.4

Climate change and weather extremes

Changes in mean climate states like, for example, in global mean temperature or atmospheric CO2 concentration are well documented and are important indicators of climate change [IPCC, 2013]. It is disproportionately more difficult to detect changes in the most extreme events. However, in general, it is these events that impose the greatest challenges and impacts on the environment and society [Seneviratne et al., 2012]. For example, if you are building a dike you will not care so much about the mean change in sea level rise but rather be interested in the maximal water level possible over the next few decades. Weather extremes also strongly affect health conditions [World Health

8

1. Introduction

Organization, 2010; Luber et al., 2014]. Heat waves, floods, droughts and cold spells are often associated with high mortality rates [World Health Organization, 2010]. Syria has suffered severe droughts in three consecutive years (2007-2010) forcing about 1.5 million people to migrate from rural farming areas to the peripheries of urban centers. This likely contributed to destabilization of the country. Present climate change has more than doubled the likelihood of such severe and persistent droughts [Kelley et al., 2015]. Further, natural hazards cause high economic losses representing a large threat to economic development in developing and emerging economies [Munich Re, 2013]. Climate science is thus coming under increasing pressure to provide robust projections of possible near-term changes in extreme weather events that can be used as a basis for mitigation and adaptation activities and policies. A growing body of research has quantified changes in weather extremes and tested if they can be attributed to anthropogenic climate change. There is emerging evidence that human-made climate change has significantly contributed to observed temperature and rainfall extremes [Min et al., 2011; Rahmstorf and Coumou, 2011; Coumou and Rahmstorf, 2012; Delworth and Zeng, 2014; Fischer and Knutti, 2015]. However, station based measurements have relatively short time lengths and some remote regions are only well-covered since 1979, when first satellites started to provide a global assessment. These limitations of the available data and the fact that extreme events are by definition rare make it challenging to detect and separate long-term changes in extremes from natural variability. Multi-decadal natural variability further complicates the detection of anthropogenic fingerprints on observed trends of similar timescales. Conceptual models based on thermodynamics only, have been used to explain historically observed weather extremes. Under such highly simplified models the probability density function of, for example monthly-mean local temperatures, would simply shift towards warmer temperatures without changing its shape [Coumou and Robinson, 2013]. Such a shift results in an increased frequency of heat extremes as observed in recent years. There is also a direct thermal effect of global warming on the hydrological cycle [Trenberth, 2011]. Increasing temperatures lead to higher evaporation rates and thus surface drying. On the other hand, the water holding capacity of the atmosphere rises by around 7 % per degree of warming fueling comparable increases in rainfall events driven by moisture convergence. Thus, climate change might amplify the hydrological cycle, a paradigm which is generally referred to as “the wet gets wetter and the dry gets drier”.

1.4. Climate change and weather extremes

9

Fig. 4: Trends and natural variability. a, The global-mean annual surface temperature anomaly [GISTEMP, 2015] shows a clear climate change signal on top of natural year-to-year variability. b, The annual North Atlantic Oscillation (NAO) index [Hurrell, 2015] is dominated by internal multi-decadal variability.

There are, however, other observations that cannot be understood by thermodynamic arguments alone. These include a frequent occurrence of cold spells (particularly after 1990) over some regions, the anomalous long-lasting drought in California, the documented drying trend in the Mediterranean or the disproportionate magnitude and duration of heat waves in Europe over last years. These examples indicate that atmospheric dynamics play an important role in driving regional climate and weather variability. Extreme weather conditions in the winter of 2013/14 in the U.S. have received a lot of attention both in the media and the scientific debate [Cohen et al., 2014; Tollefson, 2014; Wallace et al., 2014]. This winter was characterized by cold bursts and snow storms in the north-east of the U.S., also referred to as polar vortex break down. An upper tropospheric trough dragged cold arctic air towards lower latitudes causing record-breaking cold spells in New York and Boston [Wallace et al., 2014]. At the same time, the ongoing drought in California has been associated with an exceptionally persistent high pressure system located just offshore the U.S. west coast deflecting rain-bringing storms to the North [Wang et al., 2014]. In contrast, the UK floodings in 2013/14 were associated with an increased frequency of extratropical storms [Stephens and Cloke, 2014]. The British Islands belong to the most affected storm track regions in Europe and one of the few regions where climate models project robust increases in storm activity under rising greenhouse gas emissions [Zappa et al., 2013]. Thus, in this region, severe floodings might occur even more often in the future.

10

1. Introduction

Whereas thermodynamic processes (like surface temperature changes) are well understood, dynamic aspects of the climate (like storm track changes or precipitation) are still attached with large uncertainties [Shepherd, 2014]. This difference partly results from the fact that changes in thermodynamic-related quantities are directly affected by radiative forcing. In contrast, circulationrelated fields are chaotic in nature because of the nonlinear internal dynamics involved in the atmosphere. Thus, dynamics are always indirectly driven by radiative forcing [Shepherd, 2014]. The nonlinear internal feedbacks occur on different timescales making it difficult to remove them from the climate change signal through for example time averages. Therefore, it is often easier to detect a climate change fingerprint on thermodynamic-related quantities such as the global mean temperature (Fig. 4a) than it is to do so for dynamically driven quantities like the North Atlantic Oscillation (NAO) index with its natural internal variability on multi-decadal timescales (Fig. 4b). Consequently, there is a considerable model spread in projections of circulation-related aspects of the climate, including regional changes in storm track activity [Ulbrich et al., 2009; Zappa et al., 2013] and precipitation [Knutti and Sedláček, 2012]. The latter is influenced by both thermodynamics and dynamics suggesting that uncertainty comes from the dynamical contribution. Other examples include the ongoing debate whether Arctic amplification, i.e. the enhanced warming of the Arctic, might be the reason for some observed changes in atmospheric circulation [Cohen et al., 2014; Kintisch, 2014; Wallace et al., 2014]. Improving the knowledge of circulation processes and reducing the uncertainty in projected dynamical changes are thus important research topics. Progress in this field is needed if we want to understand the deviations in the changes of observed weather patterns from thermodynamic expectations of global warming. The scientific evidence of anthropogenic climate change has shifted the focus of climate research from detecting and attributing global climate change to predicting and quantifying its impacts at the regional scale. Thereby an increasing attention has been given to the question of whether and how the frequency and intensity of extreme weather events is affected by climate change. Assessing this question will be highly relevant for the planning and implementation of efficient mitigation and adaptation strategies to climate change.

1.5. Scope and contents of the thesis

1.5

11

Scope and contents of the thesis

In my thesis I address the following questions: (1) How has extreme rainfall on daily and monthly timescales changed, and to what extend can changes be explained by thermodynamics? (2) How do changes in extratropical storm tracks influence surface weather extremes and what are the underlying physical mechanisms? (3) Have extratropical storm tracks changed in the past and will they in the future? The first question is addressed in the first part of this thesis (Sect. 2.1 and 2.2) whereas the second part (Sect. 2.3 – 2.5) deals with the other two questions. Precipitation is strongly influenced by thermodynamics. There is a growing body of evidence that the intensity of heavy daily rainfall events increase with the ability of a warmer atmosphere to hold more moisture as described by the Clausius-Clapeyron equation, i.e. by a rate of about 7 % per 1 °K increase in temperature [Pall et al., 2007]. On the other hand, total rainfall at monthly timescales largely depends on the evaporation rate which is limited by the available energy that is needed for this phase transition and thus increases at a lower rate of about 2 – 4 % [Allen and Ingram, 2002; Held and Soden, 2006; Frieler et al., 2011]. There are many factors that have to add up to create an extreme rainfall event. Due to the complex nature of precipitation and the various factors involved, internal variability is large leading to only medium confidence about changes in extreme rainfall patterns [Seneviratne et al., 2012]. The first part of my thesis (Sect. 2.1 and 2.2) addresses this knowledge gap by assessing changes in record-breaking rainfall events at the global and regional scale. Thereby, the first article focuses on daily rainfall extremes and the second article analyzes record-breaking wet and dry months. Whereas most studies define rainfall extremes in terms of particular threshold exceedances, analyzing record-breaking events has the advantage that no assumption on the underlying probability distribution function has to be made [Benestad, 2003]. Moreover, it is often the most extreme (record-breaking) events that have most-severe impacts and are prominently placed in the media and hence it is useful and necessary to analyze these events in a comprehensive scientific framework. To this end, I further developed a new statistical method which can handle temporally and spatially inhomogeneous data sets including missing values and applied it for the first time to rainfall observations. In the first article of my thesis (Sect. 2.1, Lehmann et al., 2015), I report a significant upward trend in the global occurrence of observed record-breaking daily rainfall events from 1981 to 2010. During this time period my results show a significant surplus of these events by 12 % for the global

12

1. Introduction

long-term mean compared to what would be expected in a climate with no long-term change. I show that the global upward trend is consistent with thermodynamic expectations based on the Clausius-Clapeyron equation. As the trend goes upwards, the long-term increase in record-breaking rainfall events reaches 26 % in 2010 implying that in this year one out of five record-breaking events can only be explained if the long-term warming is taken into account. Over the time period from 1901 to 1980 the long-term trend in record-breaking rainfall events is characterized by multidecadal variability associated with the El Niño Southern Oscillation. Further, I find that at the regional scale the record-breaking anomaly has distinct patterns across the Earth’s continents with generally wet regions seeing an over proportional increase and drier regions less so. The most pronounced increases are found in South East Asia, Europe and in Central US with observed increases of respectively 56 %, 31 %, and 24 %. In contrast, some regions experienced a significant decrease of heavy daily rainfall events. In the Mediterranean, the reduction is 27 %, and in the western US 21 %. Both regions are at risk of severe droughts. These regional drying trends deviate from those projected by a simple thermodynamic model indicating that dynamics play an important role. Increased evaporation of surface moisture in response to global warming alters the hydrological cycle. Because changes in the hydrological cycle cannot keep up with the faster increase in daily rainfall intensities, it will take more time for evaporation to refill the atmospheric moisture after an extreme rainfall event. In other words, the mean residence time of water vapor in the atmosphere increases in a warmer climate and hence ‘it never rains but it pours’ [Trenberth, 2011]. This might lead to prolonged dry periods between storm bursts in summer and consequently increase the number of extreme low rainfall events on monthly timescales. In the second article of this thesis (Sect. 2.2, Lehmann et al., 2016), I address this question by analyzing changes in record-breaking dry and wet months in global gridded rainfall observations covering the time period 1901-2013. I find that between 1980 and 2013 the number of these events significantly deviates from that expected in a stationary climate. The main outcome of the study is that over the northern mid- to high latitudes, i.e. over large parts of central and eastern US, Europe, and Russia, the number of record-breaking wet months has increased between +15% to +35%. In contrast, Central Africa experienced significant drying leading to an increased occurrence of record-breaking dry months of up to 46% compared to a climate with no long-term warming. Tropical monsoon climates in India and South East Asia have seen a pronounced wetting over the last three decades consistent with findings presented in the previous article. In the first part of my thesis I show that climate change has a considerable influence on the intensity and frequency of heavy rainfall events and that some of the observed changes can be explained by

1.5. Scope and contents of the thesis

13

thermodynamics. However, my results also suggest that dynamical contributions play an important role in driving precipitation changes particularly at the regional scale. Such dynamical forcings might derive from circulation-related processes such as changes in extratropical storm activity. Those are in the focus of the following second part of my thesis. In the third article (Sect. 2.3, Coumou et al., 2015) significant weakening in three key dynamical features of the large-scale atmospheric circulation is reported over the last three decades: (i) the zonal-mean zonal wind, (ii) the amplitude of fast-moving Rossby waves, and (iii) the EKE of extratropical storm tracks. For the study, two re-analysis data sets are analyzed covering the time period 1979-2014 where satellite measurements provide reliable observations for daily wind fields. Weakening of the zonal wind is explained by a reduction in the equator-to-pole temperature gradient in response to Arctic amplification. As expected, Arctic amplification is strongest in late autumn or early winter when air temperatures drop below sea surface temperatures and thus heat from the oceans is released into the atmosphere [Cohen et al., 2014]. Consequently, most studies analyzing the effect of Arctic change on mid-latitude weather have focused on the winter season. However, here, it is shown that averaged over the mid-latitudes (0-360 °E and 35-70 °N), the strongest reductions in the meridional temperature gradient are observed in summer. Consistently, the zonal-mean zonal wind has weakened in summer associated with a more pronounced decline in storm track activity as measured by the eddy kinetic energy (EKE). Using regression analysis it is shown that low EKE is associated with monthly heat extremes over the storm track affected regions. The results provide first indications that the observed weakening in summer storm track activity contributed to more-persistent weather, and thus favoring the occurrence of prolonged heat waves. In the fourth article of this thesis (Sect. 2.4, Lehmann and Coumou, 2015), I further analyze the link between EKE and temperature and rainfall extremes in both summer and winter season. One of the main outcomes of this study is that extratropical storms have a moderating effect on continental temperatures. I find that whereas in summer low EKE is associated with high temperature extremes over essentially all storm track affected regions, low wintertime EKE creates favorable conditions for cold spells over parts of eastern North America, Europe, and central- to eastern Asia. Further, monthly rainfall extremes are associated with strong storm track activity throughout the year and dry spells with a lack thereof. This is consistent with previous work showing that summers tend to be either wet and cold or dry and hot, but not any other combination [Trenberth and Shea, 2005]. Trenberth [2011] explains the negative correlation between summer temperatures and precipitation by the availability of moisture and associated soil-moisture feedbacks. Interestingly, this relationship reverses over Northern Hemisphere storm track regions in winter. Here, warmer winters are generally wet and cold winters are rather dry. My results suggest that this is due to the

14

1. Introduction

moderating effect of extratropical storms on continental temperatures by bringing wet and relatively warm conditions from the oceans to the continents during winter. Trend analyses between 1979 and 2014 reveal significant reductions in wintertime EKE over large parts of the US, Europe, Russia, and China and over as much as 80 % of total mid-latitude land during summer. I further show that EKE is significantly anti-correlated with geopotential height anomaly fields over essentially all storm track affected land regions. Low synoptic activity is thus associated with predominantly high pressure conditions which can act as blocking systems deflecting storms away from their usual path. Altogether this study highlights how sensitive regional weather conditions are to any changes in large-scale atmosphere dynamics. The main results of section 2.3 and 2.4 have been summarized and discussed in the broader context of general atmospheric circulation phenomena including resonance of quasi-stationary waves and published as a peer reviewed book chapter in “Patterns of Climate Extremes: Trends and Mechanisms” published by the American Geophysical Union (article A1 in appendix). The fifth article of my thesis (Sect. 2.5, Lehmann et al., 2014) aims at a better understanding of the observed and projected storm track response to climate change. I show that under a high emission scenario climate models project little changes for mid-latitude mean EKE in Northern Hemisphere winter but a substantial weakening in summer. In the Southern Hemisphere, EKE significantly increases in winter whereas in summer a trend is absent. The results imply a stronger seasonality in both hemispheres in a future climate. Further, I show that changes in EKE are explained by changes in baroclinicity, i.e. through compound changes in vertical wind shear and static stability. Moreover, I demonstrate that storm track variability is dominated by changes in the vertical wind shear, and that this quantity alone can statistically explain the changes in EKE in some seasons. The latest state-of-the-art climate models included in the Coupled Model Intercomparison Project 5 (CMIP5) were used for the provided analysis. Whereas most studies on Northern Hemisphere extratropical storms have focused on the winter season, when storms are stronger, the most prominent result of the first paper is the consistent and pronounced decline of EKE and associated storm tracks in summer.

2 Original Manuscripts

This thesis is organized around five scientific articles, four of which are published and one which has been submitted. Each article provides its own introductory and concluding remarks as well as references. Associated supplementary materials have been attached to the main manuscripts. In the following section, a brief overview is given of the titles, contents, and author contributions of the individual articles. I was also involved in writing a peer reviewed book chapter related to the work presented in the third and fourth article of this thesis. In addition, I participated in two side projects which resulted in two further publications. The book chapter and the two papers are provided in the Appendix.

2.1. Increased record-breaking precipitation events under global warming Jascha Lehmann, Dim Coumou, and Katja Frieler. Based on two re-analysis data sets of historical rainfall measurement the occurrence of record-breaking daily rainfall events is analyzed and compared to that expected in a stationary climate. Significant increases in observed record-breaking rainfall events are explained by a thermodynamic model. Jascha Lehmann developed the idea with support from Dim Coumou. He developed the statistical method to assess confidence intervals, developed the thermodynamic Clausius-Clapeyron model, performed all analyses, analyzed the data and wrote the text. All authors participated in the interpretation of the results and helped to improve the manuscript. Published in Climatic Change, 2015, doi:10.1007/s10584-015-1434-y.

2.1. Increased record-breaking precipitation events under global warming

Climatic Change DOI 10.1007/s10584-015-1434-y

Increased record-breaking precipitation events under global warming Jascha Lehmann 1,2 & Dim Coumou 1 & Katja Frieler 1

Received: 26 September 2014 / Accepted: 17 May 2015 # Springer Science+Business Media Dordrecht 2015

Abstract In the last decade record-breaking rainfall events have occurred in many places around the world causing severe impacts to human society and the environment including agricultural losses and floodings. There is now medium confidence that human-induced greenhouse gases have contributed to changes in heavy precipitation events at the global scale. Here, we present the first analysis of record-breaking daily rainfall events using observational data. We show that over the last three decades the number of record-breaking events has significantly increased in the global mean. Globally, this increase has led to 12 % more record-breaking rainfall events over 1981–2010 compared to those expected in stationary time series. The number of record-breaking rainfall events peaked in 2010 with an estimated 26 % chance that a new rainfall record is due to long-term climate change. This increase in record-breaking rainfall is explained by a statistical model which accounts for the warming of air and associated increasing water holding capacity only. Our results suggest that whilst the number of rainfall record-breaking events can be related to natural multi-decadal variability over the period from 1901 to 1980, observed record-breaking rainfall events significantly increased afterwards consistent with rising temperatures.

1 Introduction The last decade has produced a large number of extreme weather events worldwide, including record-breaking rainfall events (Coumou and Rahmstorf 2012). The year 2010 has so far been the wettest year on record over land in terms of total precipitation (NOAA National Climatic Data Center 2010), setting new record-breaking rainfall events on different time scales over

Electronic supplementary material The online version of this article (doi:10.1007/s10584-015-1434-y) contains supplementary material, which is available to authorized users.

* Jascha Lehmann [email protected] 1

Potsdam Institute for Climate Impact Research, Telegrafenberg A26, 14473 Potsdam, Germany

2

University of Potsdam, Potsdam, Germany

19

20

2. Original Manuscripts

many parts of the world (Trenberth 2012). This seeming accumulation of weather extremes in an exceptionally warm decade has raised the question of whether these events are related to climatic change. A number of studies have addressed this issue using observational data sets and climate models (Pall et al. 2007; Zhang et al. 2007, 2013; Min et al. 2011; Shiu et al. 2012; Benestad 2013; Berg et al. 2013; Singh and O’Gorman 2014). The main finding is that over the recent past trends towards stronger precipitation extremes can be found over a larger fraction of the land area than trends towards weaker precipitation extremes (e.g. see references in Seneviratne et al. 2012). For example, Westra et al. (2013) found a significant increase in annual maximum daily precipitation extremes on a global scale. Climate models suggest that the thermodynamic change in saturation vapor pressure as described by the Clausius-Clapeyron relation is a good predictor for changes in extreme rainfall intensities (Pall et al. 2007). This relationship predicts an increase in water vapor of typically 7 % per degree of warming assuming constant relative humidity. However, dynamical changes can also influence the frequency and intensity of precipitation and thus disrupt the Clausius-Clapeyron expected change (Trenberth 2011). For example, changes in extratropical storm tracks will affect rainfall in mid-latitude regions (Scaife et al. 2011; Hawcroft et al. 2012). The response of convective precipitation to warming can exceed the ClausiusClapeyron rate (Berg et al. 2013) which will have the strongest effects in the tropics. Most studies used extreme value theory to analyze changes in threshold events, i.e. those exceeding a specified threshold of the climatological precipitation distribution (e.g., Kharin et al. 2007; Trenberth et al. 2007; Min et al. 2011; Westra et al. 2013). This usually involves fitting an extreme value distribution to the tail of the observed distribution. However, small sample sizes in the tail result in unstable fits which can have strong affects on the results (Frei and Schär 2001). Here, we present the first global analysis of observed record-breaking daily precipitation events between 1901 and 2010 and how their frequency differs from that expected in a stationary climate. Analyzing record-breaking events has the advantage that no assumption on the underlying probability distribution function has to be made. This also implies that with this method no statements can be made about the exact changes in the underlying probability distribution in terms of shifts in the mean or higher order moments. However, here, we are interested in whether record-breaking rainfall events have increased or not, irrespective of the exact underlying changes in probability distribution. Thereby, the number of observed recordbreaking rainfall events can be compared to the number expected in a climate with no longterm trend. This approach has been proven insightful for understanding the increase of heat extremes in a warming world (Benestad 2003, 2004; Redner and Petersen 2005; Meehl et al. 2009; Anderson and Kostinski 2011; Coumou et al. 2013).

2 Data and method 2.1 Data We use monthly maximum 1-day precipitation data (Rx1day) from HadEX2 (Donat et al. 2013b), a 3.75×2.5° gridded data set covering 1901–2010 (Fig. S1 in Supplementary Information (SI)). Globally aggregated quantities over land are dominated by the northern extratropics which account for roughly 2/3 of the total available data (Fig. 1, Fig. S2b). Spatial coverage varies over time with best coverage between 1960 and 2000 (Fig. S2a).


21

N. extratropics

N. Subtropics

Tropics

S. Subtropics

30

110 number of precipitation values

Fig. 1 Time series length of monthly maximum 1-day precipitation data covered at each grid point with a value of 110 indicating full coverage over 1901–2010. The magnitude is represented by the color and size of the dots

We analyze record-breaking events in each monthly Rx1day time series and subsequently take annual (12 calendar months) and boreal winter (Nov-Dec-Jan-Feb-Mar (NDJFM)) and summer (May-Jun-Jul-Aug-Sep (MJJAS)) averages. To ensure feasible statistics we restrict analysis to (1) time series with at least 30 years of data and (2) regions and time periods which had at least 100 non-missing values at each time slice. The maximum time period was found for which these criteria hold. To test the robustness of our findings, we applied the analysis also to a second data set, the Global Historical Climatology Network (GHCNDEX) (Donat et al. 2013a), which has a spatial resolution of 2.5×2.5° and covers 1951–2014. The spatial coverage is similar to that for HadEX2 (Fig. S3) and thus also the relative share of climatic zones to the global aggregate is roughly the same (Fig. S4). We repeated the analysis with HadEX2 for the time period 1951– 2010 to allow for a direct comparison of both data sets for the overlapping time period. In general, results are very similar between both data sets and thus confirm the robustness of our findings (see detailed description in SI, Fig. S5).

2.2 Observed versus iid-expected record-breaking events A rainfall value (in mm) is defined as record-breaking if it exceeds all previous values in the given time series. Due to the sparseness of record-breaking events, it is difficult to make statements about climate change for a particular location only. We therefore aggregate the number of recordbreaking events over seasons and regions as defined in Fig. 4 (see also table S1 in SI). We assume that time series in a stationary climate can be described by independent and identically distributed (iid) values. For iid time series the number of expected record-breaking N

events at time N is equal to ∑ 1=n. We normalize the number of observed record-breaking n¼1

events with the analytical solution by defining the record-breaking anomaly: Robs −R . Ranom ¼

1

R. 1

n

n

⋅100 ð%Þ;

ð1Þ

22


where Robs is the sum of all observed record-breaking events in time series within a given region for a given time period and all calendar months of the given season. Analogously, R1/n is the analytically expected number of record-breaking events summed over the same region and time span. Thereby, missing values in the original time series are also accounted for in the calculation of R1/n, respectively (see Fig. S6 for a schematic illustration). The null-hypothesis of a stationary climate thus accounts for the same spatial and temporal inhomogeneity as in the observations and hence for vacancies and the systematic increase of available data over time. An increase in observing sites will thus increase Robs but also R1/n in the same way and therefore this will not lead to a bias in Ranom. By definition, the first value of each time series is always a record-breaking event and hence equal to the iid-expected value. To avoid this artificial start value of Ranom =0 we start counting record-breaking events at the second time step. To test whether the observed number of record-breaking events is significantly different from those in a stationary climate we create a distribution of simulated record-breaking events derived from iid time series. The iid assumption for the stationary model is justified because the detrended observational time series are close to iid (see S1: ‘Testing iid assumption’ in SI). The original month specific observational time series are shuffled (in the process of which any trend, change in variance, and autocorrelation is removed) to create a set of iid time series which are based on the original observational data. This method has the advantage that no assumption on the parametric form of the underlying distribution is made. The spatial and temporal inhomogeneity of the observations had to be treated carefully. The number of observing sites systematically increased over time with some regions only providing observations after a certain time (Fig. 1 and Fig. S2). A limitation of our method is that for a set of time series one can either account for these inhomogeneities in the data availability over time (by Bfreezing^ missing values during resampling) or for spatial correlation (by synchronous re-sampling) but not both at the same time. To overcome this, our method accounts for spatial correlation within regions and for data inhomogeneities between regions. The latter tends to be small within regions but can be large between regions (Fig. S8–S10), i.e. between for example Europe and Central Africa. The method thus accounts for these large-scale data inhomogeneities as shown by a similar increasing trend in data coverage in the shuffled data set as compared to observations (Fig. S8–S10). Spatial correlation on the other hand can be pronounced for nearby grid points and hence primarily within regions. Our method is a balance to account for both effects when estimating confidence intervals, and it can be considered a conservative estimate (see sensitivity analysis later). Time series within one region are shuffled in time using exactly the same re-sampling order such that the existing spatial correlation is maintained. Also to calculate seasonal aggregates the same re-sampling order is used for each monthly time series of the given season to account for possible correlations from 1 month to the next. The shuffling is repeated 10.000 times to create a distribution of record-breaking events under the Null hypothesis of the iid-model. The mean of this distribution follows the analytically expected number of record-breaking events in N iid time series ∑ 1=n and thus normalizing according to Eq. (1) provides a distribution n¼1

centered around 0. We define the observed record-breaking anomaly to be statistically significant if it is outside the 95 % confidence range of this distribution of record-breaking anomalies calculated based on the shuffled time series.


23

To compute global mean statistics all grid points are separated into 21 smaller regions (see table S1) to which the shuffling is applied. Subsequently, all regional results are combined to come up with a global aggregate. Thus, the shuffling method accounts for changes in data coverage between regions which results in a suitable representation of temporal changes in global data coverage (Fig. S8–10). Other studies which analyzed precipitation extremes have used a similar approach but shuffled fixed-sized blocks of 2–3 years to account for autocorrelation (Kiktev et al. 2003, 2007; Alexander et al. 2006). The block size is derived from the variance inflation factor (V) which depends on the autocorrelation at all time lags and can thus be interpreted as a measure for the Btime between effectively independent samples^ (Wilks 1997). Similar to Westra et al. (2013), we find that due to little autocorrelation in our time series (we find a mean V of ~1.1) a block size greater than one is not necessary. We tested the sensitivity of our confidence intervals to different block sizes and found essentially no changes in confidence intervals (see Fig. S11–12). Further, we find that the confidence intervals for regional analyses decrease if we account for data inhomogeneity rather than spatial correlation within regions (compare Fig. 2 and Fig. S7). Thus, our confidence intervals can be considered as conservative and robust estimates. Long-term non-linear trends in record-breaking anomaly time series are computed using singular spectrum analysis (ssa) with a window length of 15 years. This method uses eigenvalue decomposition to separate non-linear trends from white noise which gives similar results as a 30-year moving average (Allen 1997; Golyandina et al. 2001).

2.3 Statistical Clausius-Clapeyron model We further compare the number of observed record-breaking events to the expected number of record-breaking events assuming that the intensity of maximum daily precipitation increases with saturation vapor pressure according to the Clausius-Clapeyron equation. The ClausiusClapeyron model consists of ensembles of precipitation time series which are composed of (1) a thermally induced long-term non-linear trend (prtrend therm) where precipitation changes are deterministically based on temperature using the Clausius-Clapeyron equation with (2) added stochastic year-to-year variability (Δpr) based upon the non-linearly detrended original precipitation time series: pr ¼ prtrend therm þ Δpr:

ð2Þ

The thermally induced non-linear trend is calculated for each grid point and each month using prtrend therm ¼ pr⋅δprtherm ;

ð3Þ

where pr is the climatological mean of Rx1day time series over the full time period and es T trend −es T ⋅100 ð4Þ δprtherm ¼ es T

24


60 20 60

200 100

20

40 20

0

grid points: 766

(h) Northern subtropics grid points: 243

(m) Northern subtropics 50

100

(c) Northern subtropics

50

0

20

0 60

(l) Northern extratropics grid points: 596

grid points: 766

grid points: 243

grid points: 187

50

(i) Tropics

50

50

grid points: 138

100

(d) Tropics

grid points: 138

0

(j) Southern subtropics

50 100 150

150 100

grid points: 124

grid points: 83

1920

1940

1960

1980

2000

grid points: 124

0

50

50 1900

(o) Southern subtropics

50

0

0

50

50

100

50

50

50

0

0

grid points: 87

(e) Southern subtropics

(n) Tropics

50

100

40

100 150 50

20

0

0

0

20

40

grid points: 974

0

50 0

0

grid points: 1107

(g) Northern extratropics

60

80

(b) Northern extratropics

(k) Global

40

40 20

grid points: 1107

(f) Global

20

100 150 200

(a) Global

MJJAS

40

NDJFM

60

80

annual

1900

1920

1940

1960

1980

2000

1900

1920

1940

1960

1980

2000

Fig. 2 Annual record-breaking anomaly (grey bars) shown (a) globally and for (b) northern extratropics, (c) northern subtropics, (d) tropics, and (e) southern subtropics. The long-term non-linear trend in record-breaking anomaly (black line) is calculated using singular spectrum analysis with window length of 15 years. The shaded areas reflect the 90 % (light blue shading) and 95 % (dark blue shading) confidence interval for equally computed long-term non-linear trends of the iid-model. The solid black line is thus directly comparable to the shaded confidence intervals. (f)–(j) and (k)–(o) are the same as (a)–(e), respectively, but for seasonal recordbreaking anomalies representing NDJFM (middle panel) and MJJAS (right panel)

is a time series with changes in precipitation due to the changes in temperature, which arise from the difference between the temperature averaged over the full time period T and the non-linear trend in temperature (Ttrend). The change in precipitation is estimated by an approximation of the Clausius-Clapeyron equation 17:625 T : ð5Þ es ðT Þ ¼ 6:1094⋅exp T þ 243:04 This statistical model thus assumes that extreme precipitation changes with temperature according to the potential of the atmosphere to hold more moisture at higher temperature. We used the CRU TS3.21 monthly temperature data (Harris et al. 2014) taken from the Climate Research Unit (CRU), which provides absolute surface temperatures on a 0.5×0.5° grid. The record-breaking anomaly for the Clausius-Clapeyron is normalized using Eq. (1) and its confidence intervals are determined using the same shuffling method as for the iid model.


3 Record-breaking anomaly over time 3.1 Comparison between observed and iid-expected record-breaking events On a global scale, the most prominent feature is a strong and consistent increase in the annual record-breaking anomaly since the 1980s as indicated by the long-term non-linear trend shown in Fig. 2a (black line). The record-breaking anomaly peaks in 2010, which saw +88 % more record-breaking events (grey bars) than expected by the iid case. The long-term non-linear trend of the global record-breaking anomaly significantly increases from 1980 onward reaching +26 % in 2010. A significant increase in the long-term non-linear trend between 1980 and 2010 is also seen over the northern extratropics (+31 % in 2010) and in the tropics (+31 % in 2010). The northern subtropics have also seen an upward trend but it is not statistically significant (+13 % in 2010). The long-term non-linear trend of the record-breaking anomaly shows multi-decadal variability which is most pronounced in the northern extratropics but also seen globally. Over the first 80 years the observed non-linear trend varies within the 95 % confidence interval of the iid-model with the only exception of a short negative excursion around 1930 in the northern subtropics and the tropics. This coincides with a relatively warm period between 1920 and 1940 over the Northern Hemisphere (Rogers 1985). Note that the northern subtropics and the tropics were only coarsely sampled in the 1930s (Fig. 2a) so the negative excursion is likely a local phenomenon only. Due to the applied data requirements the sparse data coverage over the tropics only allows to compute the record-breaking anomaly for the 1901–1940 period if all calendar months are included but not for individual seasons. During NDJFM, the evolution of the record-breaking anomaly is very similar to that for the annual results (Fig. 2f–j). However, during boreal winter, the year-to-year variability in the number of record-breaking events in the northern extratropics and subtropics is generally larger compared to annual results. In addition, the increase in record-breaking anomaly over the Northern Hemisphere towards the end of the time series is stronger in NDJFM than in the annual mean. Specifically, the global recordbreaking anomaly peaks in the 2010/11 winter with a value of +230 % (Fig. 2f). Also the observed long-term non-linear trend of the record-breaking anomaly during NDJFM is large reaching +30 % (globally), +37 % (northern extratropics), and + 18 % (northern subtropics) by 2010. In the tropics and southern subtropics the nonlinear trend in record-breaking anomaly is similar in both seasons and in annual analysis. During MJJAS (Fig. 2k–o), the year-to-year variability and the long-term non-linear trend of the record-breaking anomaly are similar to that of the full year (Fig. 2k and l). Nevertheless, whereas globally and over the northern extratropics the time series of the annual and boreal winter record-breaking anomaly show generally positive values between 1950 and 1980, the record-breaking anomaly during MJJAS remains negative. Over the northern subtropics this pattern inverts with negative anomalies in NDJFM and positive anomalies in MJJAS during 1950–1980. Thus, globally and over the northern extratropics the increase in the long-term non-linear trend of record-breaking events towards the end of the 20th century is significant at the 5 % confidence level in both summer and winter season (Fig. 2f, g, k and l).

25

26


3.2 Comparing trends in observed, iid-expected, and thermal induced record-breaking anomalies The Clausius-Clapeyron model predicts an increase in annual record-breaking rainfall anomaly starting around 1970 for the northern extratropics, northern subtropics, and globally but little change in the tropics and southern subtropics (see Fig. 3a–e). The model is able to capture the statistically significant increase of the observed long-term non-linear trend since the 1980s detected globally and over the northern extratropics and subtropics. In the tropics, the Clausius-Clapeyron model predicts no change in contrast to a consistent (but not significant) increase in the observed record-breaking anomaly (Fig. 3d, i and n). In time series with a linear trend, the number of record-breaking events scales with the ratio of the magnitude of the linear trend to the short term variability (Rahmstorf and Coumou 2011). In the warm tropics, the absolute increase in thermal induced precipitation per degree of warming as described by Eq. (5) is larger compared to cooler regions. On the other hand, the short term variability in tropical time series is about four times larger than in the extratropics leading to a annual

NDJFM 40

40

20

20

0 −20

−20

80 40

40

20

20

0

0

−20

−20

40 0

0

−20 60 60 −40 −20

0

20

40 0 1960

1980

2000

20 −40

0

20 1940

(o) Southern subtropics

40

60 −40 −20 40

(j) Southern subtropics

−40 1920

(n) Tropics

40

60

(i) Tropics

0

20 0 −40 1900

(m) Northern subtropics

20

20

40

(h) Northern subtropics

20

20 0

(e) Southern subtropics

(l) Northern extratropics

60

60

80

(g) Northern extratropics

−20

0 −20

(d) Tropics

40

60

(k) Global

0

20 0 −20

(c) Northern subtropics

20

40

−20

0

20

40

60

80

(b) Northern extratropics

40

60 −40 −20

MJJAS

(f) Global

40

(a) Global

1900

1920

1940

1960

1980

2000

1900

1920

1940

1960

1980

2000

Fig. 3 Long-term non-linear trends in annual record-breaking anomaly (black line) shown (a) globally and for (b) northern extratropics, (c) northern subtropics, (d) tropics, and (e) southern subtropics. The shaded areas reflect the 90 % (light shading) and 95 % (dark shading) confidence interval derived from long-term non-linear trends of the iidmodel (blue color) and Clausius-Clapeyron model (red color). (f)–(j) and (k)–(o) are the same as (a)–(e), respectively, but for seasonal record-breaking anomalies representing NDJFM (middle panel) and MJJAS (right panel)


relatively small trend-to-variability ratio. As a consequence, the Clausius-Clapeyron model projects an increase in record-breaking anomaly in the northern extratropics but only little change in the tropics. In the tropics we instead find indications for super Clausius-Clapeyron scaling, i.e. that record-breaking rainfall increases at a higher rate than expected by the Clausius-Clapeyron relation. Accordingly, the observed record-breaking anomaly shows an increase that reaches the upper end of the 95 % confidence level of the Clausius-Clapeyron model. Over the southern subtropics the Clausius-Clapeyron model shows little trend consistent with the observations. The Clausius-Clapeyron model predicts larger trends for boreal winter (Fig. 3f, g and h) compared to boreal summer (Fig. 3k, l and m). This is explained by the trend-to-variability ratio and follows the same argumentation as given for the comparison between the northern extratropics and the tropics: The larger thermal induced trend in boreal summer is counter acted by the up-to three times larger year-to-year variability in its Rx1day time series.

4 Regional analysis of recent past (1981–2010) The most distinct feature of the global record-breaking anomaly is a robust increase over the last 30 years. However, we showed that this trend is differently expressed across the latitudes. We therefore analyze the time period 1981–2010 in more detail on a smaller regional scale using a spatial division of the land area similar to that in Field et al. (2012) (Table S1 in SI). Time-averaged record-breaking anomalies between 1981 and 2010 show distinct regional patterns. While the record-breaking anomaly is positive globally and over all latitudinal belts (see bottom panels of Fig. 4), it is more diverse regionally with values ranging from −27 % (Mediterranean) to +56 % (South East Asia). The box panels in the map of Fig. 4 show the regional mean record-breaking anomaly (+ symbol) with confidence intervals of the iid-model (blue bars) and Clausius-Clapeyron model (red bars). On a global scale, the mean recordbreaking anomaly has significantly increased to +12 % more rainfall extremes compared to iidexpected in 1980–2010. Significant increases are also found over the northern extra-/subtropics and tropics (see bottom panels of Fig. 4). The magnitude of the increase is largest for the tropics (+18 %) and moderate for the northern extra-/subtropics (+12, +9 %). Consistent with this, most subcontinental regions also show an increased record-breaking anomaly over the last 30 years. Significant increases can be found over Central North America (+24 %), Europe (+ 31 %), Northern Asia (+21 %), the Tibetan Plateau (+31 %), and South East Asia (+56 %). Some regions show exceptionally high increases in record-breaking anomalies (e.g. South-East Asia) which reach the upper 95 % confidence limits of the Clausius-Clapeyron model. Conversely, also significant negative record-breaking anomalies are found, notably in the Mediterranean region (−27 %) and Western North America (−21 %). The Clausius-Clapeyron model projects an increase in record-breaking anomaly compared to the iid-model for all regions as all regions have warmed. Thus, significant decreases as found in Western North America and the Mediterranean region cannot be explained by this model. However, for all regions with significant increases in record-breaking anomaly the Clausius-Clapeyron model is able to capture this increase. Regional record-breaking anomalies for 1980–2010 during NDJFM are similar to those for the full year. The largest exception exists for Southern Africa, which exhibits a record-breaking anomaly of +30 % during NDJFM compared to −7 % for the full year. In addition, the decrease in record-breaking

27

28


Fig. 4 Annual observed record-breaking anomaly between 1981 and 2010. The magnitude is indicated by different colors at grid cells which contributed to the regional record-breaking anomaly. For each region a separate diagram is shown which includes the observed record-breaking anomaly (+ symbol) and the 90 and 95 % confidence interval estimates from the iid-model (blue bars) and the Clausius-Clapeyron model (red bars). Lower panels show the same results for the global mean and the four latitudinal belts (same regions as in Figs. 2 and 3)

anomaly over the Mediterranean region is more pronounced during NDJFM (−36 %) than for the full year (−27 %). South East Asia experienced a higher record-breaking anomaly during NDJFM (+80 %) than during the full year (+56 %). Some regions in the tropics do not provide results for the seasonal record-breaking anomaly due to lack of data (Fig. 5a and b). Results for MJJAS are qualitatively similar. Notable differences only exist for Southern Asia, Australia, and the Mediterranean region. Over Southern Asia, the increase in recordbreaking anomaly is significant and about three times larger in MJJAS (+45 %) compared to NDJFM (+15 %, not significant). Over Australia the decrease in record-breaking anomaly is much larger during winter (MJJAS, −24 %) than during summer (NDJFM, −2 %). Similarly, in the Mediterranean the decrease in record-breaking anomaly is only significant during winter (NDJFM, −36 %), but not during summer (MJJAS, −9 %). On a global scale, the 1981–2010 record-breaking anomaly increases significantly in both seasons, i.e. +15 % during NDJFM and +11 % during MJJAS.

5 Relation with ENSO One major contributor to natural variability in precipitation patterns is the interplay between El Niño and La Niña events (Hurrell 1995; Dai and Wigley 2000; Trenberth et al. 2003). In the Southern Hemisphere, a clear spatial pattern of the correlation between the year-to-year variability of the record-breaking anomaly and ENSO (represented by the nino3.4 index) is observed (Fig. 6). Over South East Asia and Australia we see a significant anti-correlation during all seasons implying that these regions experience significantly more record-breaking


(a)

Record-breaking anomaly during NDJFM

(b) Record-breaking anomaly during MJJAS

Fig. 5 Same as in Fig. 4, but for record-breaking anomalies during (a) NDJFM and (b) MJJAS

rainfall during La Niña events than during El Niño. The opposite pattern can be seen over Central and South America. Here, El Niño results in more record-breaking rainfall compared to La Niña events. In the Northern Hemisphere a separation can be seen over North America with the western areas (Central and Western North America and Alaska) exhibiting more record-breaking rainfall during El Niño events and the eastern regions (East North America, East Canada, and Greenland) experiencing more record-breaking events during La Niña. However,

29

30


0 0

0

+

+ + +

-0 0

+

0

+

--

+

0

0

--

0

+ +

+ 0

+

0 na na

+

+

+

+ + --

--na

----

--

--

+ na +

na na

0

0 0 0

na na

+ +

+

+

-- +

+

+

Legend slope < 0 p-val < 0.05 0.05 < p-val < 0.50 0.50 < p-val

--0

Global N. extratropics N. subtropics

slope > 0

+ +

NDJFM

0

MJJAS

Annual

+ + +

+ 0

+

+ + +

Tropics

0

-0

S. subtropics

--

-+

Fig. 6 Correlation between the year-to-year variability of the record-breaking anomaly and ENSO represented by the nino3.4 index. For each region and season symbols indicate the sign and strength of the correlation with the B+^ symbol indicating more records during El Niño compared to La Niña years and vice versa for the B−^ symbol (see legend for details)

regressions are not significant in most cases. Significant results are found over central west Asia and over the Tibetan Plateau, both of which show a positive correlation between ENSO and record-breaking anomaly. Over the latitudinal belts, correlations between record-breaking anomaly and ESNO are less obvious (bottom box in Fig. 6). This is due to the fact that regressions with opposite signs in different regions can cancel each other out, as e.g. in the southern subtropics. Nevertheless, rainfall in the Northern Hemisphere (and especially in the subtropics) tends to be intensified during El Niño years leading to a surplus of record-breaking rainfall events.

6 Summary and discussion The number of record-breaking rainfall events between 1901 and 2010 is investigated using observations from the HadEX2 data set. We find an increase of +12 % in the globally aggregated number of record-breaking rainfall events compared to that expected in a stationary climate over the time period 1981 to 2010. This implies that over the last 30 years, roughly one in ten record-breaking events would not have occurred without climate change. The increase in record-breaking anomaly peaks in 2010 with +88 % more record-breaking events than expected in a climate with no long-term change. We show that the long-term increase in record-breaking anomaly cannot be explained by (multi-decadal) natural variability alone but that it is consistent with what would be expected from rising temperatures. A large and consistent increase in the long-term trend of the record-breaking anomaly since the 1980s is found over the northern extratropics (+37/+27 % in 2010), the tropics (+25/+30 % in 2010), and partly over the northern subtropics (+19/+11 % in 2010), independent of the season (NDJFM/MJJAS). Over the southern subtropics the long-term record-breaking anomaly shows no trend. On a regional scale the mean record-breaking anomalies between 1981 and 2010 are more diverse and in some cases in the opposite direction. For example, Australia experiences a


decline in the number of record-breaking rainfall of −24 % (during winter season) at the same time as the record-breaking anomaly over South East Asia increases by +80 % (during summer season). Our results support previous findings (Seneviratne et al. 2012), but some new insights are also obtained. For example, we find a significant decrease in the winter record-breaking anomaly over the Mediterranean region. So far, there has only been low confidence about changes in extreme precipitation in this region due to inconsistent trends within domains and across studies (Kiktev et al. 2003; Caballero 2005; Alexander et al. 2006; García et al. 2007; Pavan et al. 2008). Confidence has also been low for the Asian continent in general except for Western Asia where medium confidence exists for an increase in extreme precipitation confirming our results (Kwarteng et al. 2009; Rahimzadeh et al. 2009). However, we also find significant results for other Asian regions, i.e. significant increases in record-breaking rainfall over Northern Asia, the Tibetan Plateau, India and South East Asia. Further, we show that thermally driven moisture increase has significantly contributed to the intensification of extreme rainfalls since the 1980s. In particular, the number of recordbreaking events in the last three decades is quantitatively consistent with those projected by the Clausius-Clapeyron model. This model assumes that changes in extreme rainfall intensities scale with temperature changes as given by the Clausius-Clapeyron equation, implying that the maximum moisture in the atmosphere limits the intensity of rainfall extremes. Observations and model results agree best over the northern mid-latitudes and northern subtropics. Conversely, over the tropics, the observed record-breaking anomaly is at the upper end of the 95 % range for the Clausius-Clapeyron model, indicating that, here, a super Clausius-Clapeyron scaling is required to explain the observed changes in record-breaking rainfall. This is not unreasonable given that the composition of precipitation types is different between the tropics and sub-/extratropics. Whilst in the tropics daily precipitation is largely convective it can be a mixture of convective and stratiform in the sub-/extratropics. Stratiform precipitation extremes increase with temperature at approximately the Clausius-Clapeyron rate whereas the intensity of convective rainfall tends to be more sensitive to temperature changes and can thus exceed the Clausius-Clapeyron rate (Berg et al. 2013). Our results indicate that thermodynamics are able to explain much of the observed increase in the record-breaking rainfall which cannot be related to natural climate variability. However, other factors such as changes in dynamics which were not addressed in this study likely also play a role. On a regional scale we find examples of decreasing record-breaking anomalies which cannot be attributed to either natural climate variability or to changes in atmospheric moisture content. One example is the negative Mediterranean record-breaking anomaly in winter between 1981 and 2010 which is significant at the 5 % confidence level and also well outside the confidence range of the Clausius-Clapeyron model. Hoerling et al. (2012) analyzed Mediterranean rainfall in detail showing that the drying is likely related to changes in sea surface temperature through external radiative forcing. Results from the linear regression analysis suggest that over some regions the number of record-breaking events is strongly influenced by the ENSO cycle. In particular, we find that South East Asia and Australia exhibit significantly more record-breaking rainfall events during La Niña years, whilst the correlation is reversed for South America, the Tibetan Plateau and the adjacent region of Western Asia. This is in good agreement with results for global patterns of ENSO-induced precipitation shown by Dai and Wigley (2000). We argue that the multi-decadal variability of the record-breaking anomaly can partly be explained by the multi-decadal variability in the ENSO cycle. In particular, the drop in the global record-breaking anomaly between 1920 and 1950 as well as the increase during 1950–

31

32


1980 coincide with years of strong La Niña and El Niño years, respectively (Fig. S14). This is consistent with the positive correlation we find for the global record-breaking anomaly and ENSO. However, over the last 30 years natural climate variability (and in particular ENSO) cannot explain the large and consistent increase in record-breaking anomaly. Instead, over this period, changes in temperature seem to have favored the increased number of record-breaking precipitation events globally. Acknowledgments We thank the Met Office Hadley Center, GHCN, and CRU for making their data available. The work was supported by the German Federal Ministry for the Environment, Nature Conservation and Nuclear Safety (11 II 093 Global A SIDS and LDCs), by the German research Foundation (CO994/2-1), and the German Federal Ministry of Education and Research (01LN1304A).

References Alexander LV et al (2006) Global observed changes in daily climate extremes of temperature and precipitation. J Geophys Res 111(D5):D05109. doi:10.1029/2005JD006290 Allen M (1997) Optimal filtering in singular spectrum analysis. Phys Lett A 234:419–428. doi:10.1016/S03759601(97)00559-8 Anderson A, Kostinski A (2011) Evolution and distribution of record-breaking high and low monthly mean temperatures. J Appl Meteorol Climatol 50(9):1859–1871. doi:10.1175/JAMC-D-10-05025.1 Benestad R (2003) How often can we expect a record event? Clim Res 25:3–13 Benestad RE (2004) Record-values, non-stationarity tests and extreme value distributions. Glob Planet Chang 44(1–4):11–26 Benestad RE (2013) Association between trends in daily rainfall percentiles and the global mean temperature. J Geophys Res Atmos 118(19):10,802–10,810. doi:10.1002/jgrd.50814 Berg P, Moseley C, Haerter JO (2013) Strong increase in convective precipitation in response to higher temperatures. Nat Geosci 6(3):181–185. doi:10.1038/ngeo1731 Caballero R (2005) The dynamic range of poleward energy transport in an atmospheric general circulation model. Geophys Res Lett 32(2):L02705. doi:10.1029/2004GL021581 Coumou D, Rahmstorf S (2012) A decade of weather extremes. Nat Clim Chang 2:491–496. doi:10.1038/ nclimate1452 Coumou D, Robinson A, Rahmstorf S (2013) Global increase in record-breaking monthly-mean temperatures. Clim Chang 118(3–4):771–782. doi:10.1007/s10584-012-0668-1 Dai A, Wigley T (2000) Global patterns of ENSO‐induced precipitation. Geophys Res Lett 27(9):1283–1286 Donat MG, Alexander LV, Yang H, Durre I, Vose R, Caesar J (2013a) Global land-based datasets for monitoring climatic extremes. Bull Am Meteorol Soc 94(7):997–1006. doi:10.1175/BAMS-D-12-00109.1 Donat MG et al (2013b) Updated analyses of temperature and precipitation extreme indices since the beginning of the twentieth century: the HadEX2 dataset. J Geophys Res Atmos 118(5):2098–2118. doi:10.1002/jgrd.50150 Field CB, Barros V, Stocker TF, Dahe Q (2012) In: Field CB, Barros V, Stocker TF, Dahe Q (eds) Managing the risks of extreme events and disasters to advance climate change adaptation. Cambridge University Press, Cambridge Frei C, Schär C (2001) Detection probability of trends in rare events: theory and application to heavy precipitation in the alpine region. J Clim 14(7):1568–1584. doi:10.1175/1520-0442(2001) 0142.0.CO;2 García J, Gallego MC, Serrano A, Vaquero J (2007) Trends in block-seasonal extreme rainfall over the Iberian Peninsula in the second half of the twentieth century. J Clim 20(1):113–130. doi:10.1175/JCLI3995.1 Golyandina N, Nekrutkin VV, Zhigljavsky AA (2001) Analysis of time series structure: SSA and related techniques Harris I, Jones PD, Osborn TJ, Lister DH (2014) Updated high-resolution grids of monthly climatic observations - the CRU TS3.10 Dataset. Int J Climatol 34(3):623–642. doi:10.1002/joc.3711 Hawcroft MK, Shaffrey LC, Hodges KI, Dacre HF (2012) How much Northern Hemisphere precipitation is associated with extratropical cyclones? Geophys Res Lett 39(24):L24809. doi:10.1029/2012GL053866 Hoerling M, Eischeid J, Perlwitz J, Quan X, Zhang T, Pegion P (2012) On the increased frequency of Mediterranean drought. J Clim 25(6):2146–2161. doi:10.1175/JCLI-D-11-00296.1 Hurrell JW (1995) Decadal trends in the north atlantic oscillation: regional temperatures and precipitation. Science 269(24):676–679. doi:10.1126/science.269.5224.676


Kharin VV, Zwiers FW, Zhang X, Hegerl GC (2007) Changes in temperature and precipitation extremes in the IPCC ensemble of global coupled model simulations. J Clim 20(8):1419–1444. doi:10.1175/JCLI4066.1 Kiktev D, Sexton DMH, Alexander L, Folland CK (2003) Comparison of modeled and observed trends in indices of daily climate extremes. J Clim 16(22):3560–3571. doi:10.1175/1520-0442(2003) 0162.0.CO;2 Kiktev D, Caesar J, Alexander LV, Shiogama H, Collier M (2007) Comparison of observed and multimodeled trends in annual extremes of temperature and precipitation. Geophys Res Lett 34(April):2–6. doi:10.1029/ 2007GL029539 Kwarteng AY, Dorvlo S, Vijaya GT (2009) Analysis of a 27-year rainfall data (1977–2003) in the Sultanate of Oman. 617:605–617. doi:10.1002/joc Meehl GA, Tebaldi C, Walton G, Easterling D, McDaniel L (2009) Relative increase of record high maximum temperatures compared to record low minimum temperatures in the U.S. Geophys Res Lett 36:1–5. doi:10. 1029/2009GL040736 Min S-K, Zhang X, Zwiers FW, Hegerl GC (2011) Human contribution to more-intense precipitation extremes. Nature 470(7334):378–381. doi:10.1038/nature09763 NOAA National Climatic Data Center (2010), State of the climate: global analysis for annual 2010, Publ. online December 2010, (July 2009), http://www.ncdc.noaa.gov/sotc/global/2010/13 Pall P, Allen MR, Stone DA (2007) Testing the Clausius–Clapeyron constraint on changes in extreme precipitation under CO2 warming. Clim Dyn 28(4):351–363. doi:10.1007/s00382-006-0180-2 Pavan V, Tomozeiu R, Cacciamani C, Di Lorenzo M (2008) Daily precipitation observations over EmiliaRomagna: mean values and extremes. Int J Climatol 28(15):2065–2079. doi:10.1002/joc.1694 Rahimzadeh F, Asgari A, Fattahi E (2009) Variability of extreme temperature and precipitation in Iran during recent decades. 343:329–343. doi:10.1002/joc Rahmstorf S, Coumou D (2011) Increase of extreme events in a warming world. Proc Natl Acad Sci U S A 108(44):17905–17909. doi:10.1073/pnas.1101766108 Redner S, Petersen MR (2005) On the role of global warming on the statistics of record-breaking temperatures. Phys Rev E 11. doi:10.1103/PhysRevE.74.061114 Rogers JC (1985) Atmospheric circulation changes associated with the warming over the Northern North Atlantic in the 1920s. J Clim Appl Meteorol 24(12):1303–1310. doi:10.1175/1520-0450(1985) 0242.0.CO;2 Scaife AA et al (2011) Climate change projections and stratosphere–troposphere interaction. Clim Dyn 38(9–10): 2089–2097. doi:10.1007/s00382-011-1080-7 Seneviratne SI, Nicholls N, Easterling D et al (2012) Changes in climate extremes and their impacts on the natural physical environment. In: Field CB et al (eds) Managing the risks of extreme events and disasters to advance climate change adaptation. Cambridge University Press, Cambridge, pp 109–230 Shiu C-J, Liu SC, Fu C, Dai A, Sun Y (2012) How much do precipitation extremes change in a warming climate? Geophys Res Lett 39(17). doi:10.1029/2012GL052762 Singh MS, O’Gorman PA (2014) Influence of microphysics on the scaling of precipitation extremes with temperature. Geophys Res Lett. doi:10.1002/2014GL061222 Trenberth K (2011) Changes in precipitation with climate change. Clim Res 47(1):123–138. doi:10.3354/ cr00953 Trenberth KE (2012) Framing the way to relate climate extremes to climate change. Clim Chang. doi:10.1007/ s10584-012-0441-5 Trenberth KE, Dai A, Rasmussen RM, Parsons DB D2003] The changing character of precipitation. Bull Am Meteorol Soc 84:1205–1217. doi:10.1175/BAMS-84-9-1205, +1161 Trenberth KE, Jones PD, Ambenje P et al (2007) Obervations: surface and atmospheric climate change. In: Solomon S, Dahe Q, Manning MR, Chen Z, Marquis M, Averyt KB, Tignor M, Miller HL (eds) Climate change 2007: the physical science basis. Cambridge University Press, Cambridge Westra S, Alexander LV, Zwiers FW (2013) Global increasing trends in annual maximum daily precipitation. J Clim 26(11):3904–3918. doi:10.1175/JCLI-D-12-00502.1 Wilks DS (1997) Resampling hypothesis tests for autocorrelated fields. J Clim 10:65–82. doi:10.1175/15200442(1997)0102.0.CO;2 Zhang X, Zwiers FW, Hegerl GC, Lambert FH, Gillett NP, Solomon S, Stott PA, Nozawa T (2007) Detection of human influence on twentieth-century precipitation trends. Nature 448(7152):461–465. doi:10.1038/ nature06025 Zhang X, Wan H, Zwiers FW, Hegerl GC, Min S-K (2013) Attributing intensification of precipitation extremes to human influence. Geophys Res Lett 40(19):5252–5257. doi:10.1002/grl.51010

33

34


Supplementary Information (SI) for:

Increased record-breaking precipitation events under global warming Jascha Lehmann*, Dim Coumou, Katja Frieler *Corresponding author. Email: [email protected]

S1: ‘Testing the iid assumption’ The iid assumption for the stationary model is justified because the detrended observational time series are close to iid. To show this, we first detrend the original Rx1day time series by subtracting the smoothed mean value calculated using singular spectrum analysis with window length of 15 years. The residuals contain the year-to-year variability for a specific month. We then test whether the residuals are temporally independent by calculating the serial correlation in the detrended Rx1day time series for each calendar month. We find that the correlation values are randomly distributed over all land areas and are generally small and within -0.2 and 0.2 for all months (Fig. S13). Some outliers reach values of -0.6 and +0.5, but for such relatively small values of serial correlation the 1/n solution holds1.

1

Coumou, D., A. Robinson, and S. Rahmstorf (2013), Global increase in record-breaking monthly-mean temperatures, Clim. Change, 118(3-4), 771–782, doi:10.1007/s10584-012-0668-1.


S2: Additional figures

Fig. S1 Location of the 11 391 observing weather stations used to create the HadEX2 data set which is given on a 3.75° x 2.5° grid.

Fig. S2 Number of grid points with monthly maximum 1-day precipitation data for each point in time given in (a) absolute numbers and (b) relative to the total number of grid points with data.

35

36


Fig. S3 Same as Fig. 1, but for the GHCNDEX data set.

Fig. S4 Same as Fig. 2, but for the GHCNDEX data set.


Fig. S5 Time series of the annual record-breaking anomaly calculated from HadEX2 (grey bars) and GHCNDEX (pink bars) shown for (a) Global, (b) northern extratropics, (c) northern subtropics, (d) tropics, and (e) southern subtropics. The long-term non-linear trend of the record-breaking anomaly (solid line) is calculated using singular spectrum analysis with window length of 15 years. (f)-(j) and (k)-(o) are the same as (a)-(e), respectively, but for seasonal record-breaking anomalies representing NDJFM (middle panel) and MJJAS (right panel). To ensure comparability between both data sets record-breaking anomalies were only calculated for the period 1951-2010, where both data sets provide data. For each region and season, we computed the Pearson correlation coefficient ( ρ XY ) between the record-breaking anomaly time series of both data sets, which is shown in the corresponding panels in Fig. S5. In general, results are in good agreement between the two data sets indicated by high positive Pearson correlation coefficients implying that the variables are positively linearly related. In the tropics, correlation coefficients are in a range of 0.070.20 indicating a positive but weaker linear relationship. This could be due to larger uncertainties due to sparse data coverage in this region. A large and consistent increase in record-breaking anomaly can be found in both data sets over the northern extratropics, northern subtropics, and on the global scale. However, the increase is slightly stronger in GHCNDEX compared to HadEX2. Over the southern subtropics both data sets show no trend in record-breaking anomaly.

37

38


Fig. S6 Schematic illustration of (Step 1) finding region specific time boundaries for the shuffling process, (Step 2) computing the observed regional record-breaking anomaly, and (Step 3) computing a set of modeled record-breaking anomalies based on the iid-model. In the first step, for each month, the Rx1day data is organized in a p x n matrix where the number of rows p equals the number of grid cells and thus denotes the location and the number of columns n refers to the number of years. For each region a time period is defined for which this regions provides data (see colored rectangles). The following steps are applied to each individual region in the limits of these time boundaries. To compute the observed record-breaking anomaly (“Step 2”), record-breaking events are counted in each row, i.e. for each grid cell in the given region with a value of 1 denoting that this value has set a new record and a value of 0 that this particular value was not a record-breaking event (see upper matrix in middle panel of Step 2). We subsequently sum up all values of this matrix along the p grid cells which leaves a vector of length n giving the total number of record-breaking events per year in the given region. This vector is normalized with the number of expected record-breaking events (lower matrix in middle panel) using eq. [1] to come up with a time series of the regional record-breaking anomaly (right panel). The black dashed vertical lines in the middle panel of “Step 2” denote the time period which fulfills the applied data requirements. “Step 3” explains how the iid-model is computed. First, the n columns are randomly shuffled in which process the order in time is lost, but the spatial correlation within the given region is kept. From the shuffled matrix a time series of simulated regional record-breaking anomaly is computed in the same way as described for the observational data. The resulting record-breaking anomaly refers to one realization of the iid-model. The full procedure described in “Step 3” is repeated 10.000 times to create a set of possible record-breaking anomalies under the Null hypothesis of the iid assumption. From this set of time series the 90th and 95th confidence intervals are determined.


Fig. S7 Same as Fig. 2 in main manuscript but, here, confidence ranges are estimated using a shuffling process which does not account for spatial correlation but therefore keeps missing values fixed in space and time and thus conserves trends in the number of data points per year. This leads to generally smaller confidence ranges compared to Fig. 2, where spatial correlation within each region is taken into account at the expense of neglecting changes in the number of data points per year.

39

40


Fig. S8 Temporal heterogeneity in monthly data coverage, exemplarily shown for January. In the left panel (a), for each grid point, years with data are colored corresponding to the region the grid point belongs to. Only those values are colored for which the data requirements for the full year are fulfilled, i.e. minimum time series length of 30 years and minimum 100 time series per year. For each region, grid points are sorted by the start year of the given time series to illustrate temporal heterogeneity within individual regions. For each year we sum up all grid points with data which results in a time series with the total (global) number of available data as depicted by black circles in panel (b). Randomly shuffling years of each time series in each region within its individual time boundaries leads – on average – to a nearly equal distribution of data coverage in the given region. This is shown by red circles in panel (b). This time series is characterized by steplike increases in years where new regions start to supply data as indicated by the vertical dashed lines. The shuffling method is thus able to reproduce a similar curve of changes in the amount of data over time.


Fig. S9 same as Fig. S8 but with data requirements applied to winter season.

Fig. S10 Same as Fig. S8 but shown for June and with data requirements applied to summer season.

41

42


Fig. S11 Same as Fig. 2 in main manuscript but, here, confidence ranges are estimated using a blockshuffling method with a fixed block size of 2 years.


Fig. S12 Same as Fig. S11 but with a fixed block size of 3 years.

43

44


Fig. S13 Serial correlation in the non-linear detrended HadEX2 Rx1day time series for each calendar month.


Fig. S14 Time series of annual record-breaking anomaly shown for the global mean (black line). Colored bars represent the ENSO time series (nino3.4 index) with positive values indicating El Niño years (blue bars) and negative values corresponding to La Niña years (red bars).

45

46


S3: Additional tables # 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Region Label Alaska Australia Central Africa Central America CGI Central North America Central West Asia Eeastern Asia Eastern North America Europe Mediterranean Northern Asia Northern South America Southern Africa Sahara Southern Asia Southern South America South East Asia Tibetan Tableau Western South America Western North America Northern Extratropics Northern Subtropics Tropics Southern Subtropics

lat1 lon1 lat2 lon2 lat3 lon3 lat4 lon4 lat5 lon5 lat6 lon6 lat7 lon7 60 105 60 168 73 169 73 105 -------50 110 -10 110 -10 155 -30 155 -30 180 -50 180 ---11 -20 15 -20 15 52 -11 52 ------11 50

-68 -10

-1 50

-80 105

29 85

118 105

29 85

-90 -10

---

---

---

---

---

---

50

-85

29

-85

29

105

50

105

--

--

--

--

--

--

15 20

40 100

50 50

40 100

50 50

75 145

30 20

75 145

30 --

60 --

15 --

60 --

---

---

25 45 30 50

-60 -10 -10 40

25 75 45 70

-85 -10 -10 40

50 75 45 70

-85 40 40 180

50 45 30 50

-60 40 40 180

-----

-----

-----

-----

-----

-----

-20

-66

-1

-80

11

-69

11

-50

0

-50

0

-34

-20

-34

-35 15 5

-10 -20 60

-11 30 30

-10 -20 60

-11 30 30

52 40 100

-35 15 20

52 40 100

--20

--95

--5

--95

----

----

-20

-39

-57

-39

-57

-67

-50

-72

-20

-66

--

--

--

--

-10

95

20

95

20

155

-10

155

--

--

--

--

--

--

30

75

50

75

50

100

30

100

--

--

--

--

--

--

-1

-80

-20

-66

-50

-72

-57

-67

-57

-82

1

-82

--

--

29

105

29

130

60

130

60

105

--

--

--

--

--

--

40

180

90

180

90

180

40

180

--

--

--

--

--

--

20 -20

180 180

40 20

180 180

40 20

180 180

20 -20

180 180

---

---

---

---

---

---

-40

180

-20

180

-20

180

-40

180

--

--

--

--

--

--

Table S1 Coordinates of corners of regions displayed in Fig. 4 and Fig. 5. Values are given in degrees North (for latitudes) and in degrees East (for longitudes). Regions 1-21 are used to compute the global aggregate.

2.2. Changes in record-wet and record-dry months in global land observations Jascha Lehmann, Finn Mempel, and Dim Coumou. In this paper, record-breaking dry and wet months are analyzed in global-gridded rainfall observations. Some of the main findings include significant wetting of northern mid- to high latitudes and contrasting trends in the tropics indicated by substantial drying in Central Africa and increases in record-wet months in monsoon regions including India and South East Asia. Jascha Lehmann initiated the research, performed all analyses, analyzed the data and wrote the text. Finn Mempel contributed to the trend analysis and in assessing record-breaking dry months. Dim Coumou participated in the interpretation of the results and helped to improve the manuscript. Submitted to Environmental Research Letters.

2.2. Changes in record-wet and record-dry months in global land observations

1

Changes in record-wet and record-dry months in global land

2

observations

3

Jascha Lehmann*1,2, Finn Mempel3, and Dim Coumou1

4 5 6 7 8

1

Potsdam Institute for Climate Impact Research, Germany

2

University of Potsdam, Germany

3

Universidade do Algarve, Portugal

*Corresponding author: J. Lehmann, [email protected], Telegrafenberg A62, 14473 Potsdam, Germany

9 10 11 12

Abstract.

13

Extreme rainfall events can lead to severe floodings whereas consecutive months of low rainfall can

14

strongly affect the occurrence of droughts. Climate change alters the hydrological cycle whereby both

15

thermodynamic and dynamic processes are considered important. Generally, it is expected that

16

relatively dry regions will get drier and relatively wet regions will get wetter. Limitations in

17

observational records have made it difficult to detect such changes in either mean or extreme rainfall.

18

Here, we analyze record-breaking wet and dry months in global gridded rainfall observations over

19

land. We find that between 1980 and 2013 the number of these events significantly deviates from that

20

expected in a stationary climate. The key finding is a pronounced increase in record-breaking wet

21

months by up to 37% in the northern mid- to high latitudes compared to a climate without long-term

22

trends. In the tropics we find opposing trends indicated by increased record-dry months over relatively

23

dry Central Africa and substantial wetting over monsoon regions including India and South East Asia.

24

These changes in record-dry and record-wet months are broadly consistent with trends detected using

25

Mann-Kendall tests and imply severe consequences for water resource management.

1

49

50


26

Introduction:

27

Changes in monthly precipitation pattern have large impacts on the environment and on society with

28

too much rain leading to floods and too little rain leading to droughts and reduced water supply [Field

29

et al., 2014a, 2014b]. Both can impose severe impacts on agriculture and thus food production. The

30

recent decade has experienced a seemingly large number of record-breaking rainfall events on both

31

sides – extreme wet and dry months. In 2014, the UK was affected by severe floodings [Stephens and

32

Cloke, 2014] and May 2015 was the wettest month ever recorded in the US with precipitation setting

33

new records over many regions, locally up to 5 times the monthly climatology [NOAA, 2015]. At the

34

same time, some subtropical regions have experienced long-lasting droughts including the middle

35

East, Australia and Southwestern US [Heberger, 2011; Wang et al., 2014; Barlow et al., 2015; Kelley

36

et al., 2015]. Drought in California lead to crop losses and the implementation of an emergency

37

regulation enforcing residents to reduce potable urban water usage by 25% [Wang et al., 2014]. Some

38

of these hydrological extremes have been attributed to anthropogenic forcing of the climate [e.g., Pall

39

et al., 2011; Hoerling et al., 2012; Kelley et al., 2015].

40

Climate change is expected to alter the intensity and frequency of rainfall extremes [Pall et al., 2007;

41

Zhang et al., 2007, 2013; Min et al., 2011; Shiu et al., 2012; Benestad, 2013; Berg et al., 2013; Singh

42

and O’Gorman, 2014]. However, the sign of change strongly depends on the timescale, season, and

43

location at which rainfall occurs. During the heaviest daily rainfall events all the moisture in the air is

44

precipitated out and hence those short lived extremes scale with the water-holding capacity of air

45

which increases by ~7% per degree of warming following the Clausius-Clapeyron equation.

46

Significant upward trends, in agreement with Clausius-Clapeyron scaling, have been detected in

47

annual maximum daily precipitation extremes as well as for daily maxima in individual seasons

48

[Groisman et al., 2005; Alexander et al., 2006; Westra et al., 2013; Lehmann et al., 2015]. Global-

49

mean monthly precipitation and evaporation are primarily constrained by the global energy budget

50

which increases at a lower rate of 2 – 4% per degree warming [Allen and Ingram, 2002; Held and

51

Soden, 2006; Frieler et al., 2011]. In a warmer climate it will thus take longer for evaporation to refill

52

the atmospheric moisture after an extreme rainfall event which might lead to prolonged dry periods.

53

This might thus alter the variability of rainfall on monthly timescales. Still, global-mean monthly

54

precipitation shows a near-zero trend [Sun et al., 2012] though pronounced trends are found at the

55

regional level. Over the 20th century, monthly-mean precipitation increased in the northern mid- to

56

high latitudes and in the southern subtropics and tropics while it decreased in the northern subtropics

57

and tropics [Zhang et al., 2007] in agreement with the ‘dry gets drier, wet gets wetter’ paradigm

58

[Trenberth, 2011]. Further, seasonality is important as precipitation is driven by different mechanisms

59

in winter and summer, at least in the mid-latitudes [Schönwiese et al., 2003; Zheng et al., 2015]. Here,

60

winter precipitation is largely controlled by extratropical synoptic-scale storms [Raible et al., 2007; 2


61

Hawcroft et al., 2012; Pfahl and Wernli, 2012] while summer precipitation is mainly driven by short-

62

lived and smaller meso-scale storm bursts [Zheng et al., 2015].

63

Changes in precipitation extremes, including both prolonged dry and prolonged wet periods, generally

64

impose a stronger impact on society and ecosystems compared to changes in mean rainfall. The

65

frequency of extremes can be altered by shifts in the mean and by changes in variance, whereby the

66

latter can have the strongest influence especially on the most-extreme events [Katz and Brown, 1992].

67

To quantify changes in extremes, we analyze record-breaking wet and dry months using worldwide

68

rainfall observations covering the time period 1901-2013. The observed changes are compared to

69

those expected in a climate with no long-term trends.

70

Data and methods

71

We use the Global Precipitation Climatology Center (GPCC, [Schneider et al., 2015]) full data

72

reanalysis version 7 provided on a 0.5° x 0.5° grid. It contains quality-controlled monthly rainfall

73

observations on land covering the time period 1901-2013 (Fig. 1). Several data requirements are

74

applied before the analysis of record-breaking rainfall events. To minimize interpolation problems, we

75

only consider those monthly rainfall values from a given grid cell which have at least one

76

measurement station. All other rainfall values are set to ‘missing’. This approach is similar but

77

somewhat stricter than in previous studies [Sun et al., 2012; Tett et al., 2013; Simmons et al., 2014].

78

For example, Sun et al. [2012] analyzed global land precipitation variability in different observational

79

datasets using only those grid cells having at least one measurement site for 90% of the months over

80

the analyzed time period. Our data requirement particularly removes rainfall observations in South

81

America, Africa and East Asia during the first half of the 20th century (see Fig. 8 in [Becker et al.,

82

2013]). Further, rainfall time series with less than 30 years of data or with zero rainfall are excluded

83

from the analysis (Fig. S1 in Supplementary Information). The latter requirement is needed for

84

assessing record-dry months since a value of zero cannot be undercut and thus would prevent the

85

occurrence of any further record-dry month. In leap years, the absolute monthly rainfall value of

86

February is modified by subtracting the mean daily rainfall value of this month from the absolute

87

value. This is necessary in order to compare February values in leap years with February values in

88

other years. Ultimately, we only report results for regions providing at least 100 non-missing values at

89

each time slice.

90 91

A rainfall value (in mm) is defined as a record-wet (dry) month if it exceeds (or is lower than) all

92

previous values in the given time series of an individual grid cell. Record-statistics have the advantage

93

that no assumption on the underlying probability density distribution is made [Coumou et al., 2013].

94

Record-wet and record-dry months are analyzed for each calendar month individually and then 3

51

52


95

aggregated to annual (12 calendar months) or boreal winter (November to March (NDJFM)) and

96

summer (May to August (MJJA)) averages.

97

To assess how climate change has affected the occurrence and frequency of record-rainfall events, the

98

number of observed record-rainfall events is compared to that expected in a stationary climate with no

99

long-term changes. As in Lehmann et al. [2015], we assume that in a stationary climate rainfall

100

observations can be described as independent and identically distributed (iid) for which the number of

101

expected record-events after N time steps is 𝑅𝑁 = ∑𝑁 𝑛=1 1/𝑛. We define the record-anomaly as 𝑅𝑎𝑛𝑜𝑚,𝑖 =

102 103 104

𝑅𝑜𝑏𝑠,𝑖 − 𝑅𝑁,𝑖 ∙ 100(%) 𝑅𝑁,𝑖

which describes how much the observed number of record-events (𝑅𝑜𝑏𝑠 ) at grid point i deviates from

that expected in a climate with no long-term trends (𝑅𝑁 ). Regional aggregates of record-anomalies are calculated using

𝑅𝑎𝑛𝑜𝑚,𝑅 =

105

∑𝑖 𝑅𝑜𝑏𝑠,𝑖−∑𝑖 𝑅𝑁,𝑖 ∑𝑖 𝑅𝑁,𝑖

∙ 100(%),

106

with the sum including all grid points i in region R. We would like to stress that each monthly rainfall

107

time series of a given grid cell is compared to its individual 1/𝑛 time series thus accounting for

108

missing values and hence any spatial and temporal inhomogeneities in the observations.

109

Statistical significance is determined using the shuffling method as described in detail in Lehmann et

110

al. [2015]. Thereby, each time series is shuffled 10,000 times – in which process any trend, change in

111

variance and autocorrelation is removed – to create a set of iid time series under the null hypothesis of

112

a stationary climate. The method takes care of spatial correlation within a given region by using the

113

same re-sampling order for all shuffled time series available for this region. This way, possible trends

114

in regional data coverage are lost which, however, only has a minor effect on the analysis [Lehmann et

115

al., 2015]. We define the observed record-anomaly to be statistically significant if it is outside of the

116

95% confidence range which is computed from the distribution of sampled record-anomalies based on

117

the shuffled time series.

118

We use a τ-based Mann-Kendall test to examine the direction of change in monthly precipitation time

119

series [Chandler and Scott, 2011; Westra et al., 2013]. This method does not make any assumption on

120

the underlying distribution of the data or on the particular form of the trend. The Mann-Kendall

121

parameter τ statistically assess whether there is a monotonic upward or downward trend in the given

122

data with a positive value implying that observations later in the time series tend to be larger than

123

earlier observations and vice versa for negative τ values. The parameter τ can be as high as 1 in which

124

case the time series is monotonically increasing or as low as -1 in case of a monotonically decreasing

125

time series. We calculate τ for each calendar month and grid point and then average over the same 4


126

regions and seasons as for the analysis of the record-anomaly. We test whether the observed trends are

127

statistically different from the null hypothesis of no trends using the same shuffling method as

128

described above. Hence, a distribution of 10,000 shuffled time series is created from which the 95%

129

confidence range is extracted to define statistically significant trends in the observations.

130

We repeated all analyses for the GPCC data sets provided on a 1° x 1° grid and a 2.5° x 2.5° grid to

131

test the effect of different numbers of measurement stations per grid cell. Based on these datasets the

132

same conclusions hold (Fig. S3-S8).

133

Changes in the occurrence of record-breaking wet and dry months

134

The annual long-term trend in record-wet anomaly shows a significant upward trend for the global

135

mean starting around 1980 (blue line in Fig. 2a). This increase is also seen in boreal winter where the

136

long-term trend reaches a value of +18% in 2013, implying 18% more observed record-wet months

137

than expected in a stationary climate (Fig. 2f). The long-term trend during boreal summer shows a

138

similar increase (Fig. 2k). The northern extratropics reveal a similar pattern but with generally larger

139

magnitudes (Fig. 2b, g, l). In particular, record-wet months significantly increased by up to 30% in the

140

long-term trend over the last three decades both for the annual mean and in boreal summer. In the

141

northern subtropics, the tropics, and the southern subtropics the long-term trend of record-wet and

142

record-dry months mostly varies within the 95% confidence range of the stationary climate. A notable

143

exception is a short period of increased record-dry months in the tropics from the 1980s to the end of

144

the 1990s (Fig. 2d, i, n). In the boreal winter season this dry period is followed by a steep increase in

145

record-wet months over the last decade (Fig. 2i).

146

Record-anomalies averaged between 1980 and 2013 show distinct zonally-averaged patterns which

147

can roughly be grouped into four latitudinal belts: The northern mid- to high latitudes (30°-75° N), the

148

tropics (15° S – 15° N), and two latitudinal bands from 15° to 30° N and 15° to 30° S. Though record-

149

anomalies are presented for the period 1980-2013, the definition of whether a given rainfall value in

150

this period represents a new record-breaking event is based on all previous values of the full time

151

series starting before 1980 (start year depends on the grid cell). Thus, zonal-mean record-anomalies

152

are compared to Mann-Kendall trends computed for the full time series.

153

In the northern mid- to high latitudes (30°-75° N) the number of annual record-wet months

154

significantly increased by up to 28% with the magnitude amplifying towards higher latitudes (Fig. 3a).

155

Changes in record-dry months over the same region are absent. This indicates a broadening of the

156

rainfall distribution at the wet-tail consistent with upward trends in mean monthly precipitation (Fig.

157

3b). Consequently, we find significantly more land areas with upward monthly rainfall trends (74%

158

±10%) than land areas with downward trends over this latitudinal belt (Fig. 3c). This annual pattern is

159

also seen in winter and to a slightly less extend in summer (see second and third column in Fig. 3). In 5

53

54


160

winter, an up to +37% increase in record-wet months is detected at high latitudes associated with

161

significant and more-pronounced upward trends in monthly precipitation (Fig. 3d, e, f). Summer

162

changes in record-wet months show a similar increase between 30° and 60° N and a significant but

163

smaller increase at 60°-75° N (Fig. 3). These changes are associated with significantly more grid cells

164

showing upward than downward trends (Fig. 3i). The most prominent feature at 15°-30° N is a

165

significant increase in record-wet months during boreal winter (Fig. 3d). At 15°-30° S a significant

166

increase in record-dry months in seen in austral winter (Fig. 3g). In both cases, the change is not

167

captured by trends in mean precipitation. Changes over tropical land between 15° S and 15° N are

168

dominated by a drying tendency independent of the season. This is reflected by an increased number

169

of record-dry months for the annual mean (24% ±3%) as well as for the boreal winter (22% ±2%) and

170

summer season (26% ±5%). Changes in mean tropical precipitation consistently show negative (albeit

171

generally insignificant) trends.

172

Changes in record-wet and record-dry months at the regional scale reflect the distinct latitudinal

173

patterns reported above but give additional insights (Fig. 4). The significant increase in record-wet

174

months in the latitudinal band from 30° to 75° N is in agreement with regional results. Here, strongest

175

increases are found over central (22% ±5%) and eastern North America (27% ±6%), Europe (32%

176

±2%), and Russia (22% ±7%) (Fig. 4a, d, g). Among these regions significant upward trends in mean

177

precipitation are only observed in Russia (Fig. 4c, f). The increase in boreal summer record-wet

178

months at 15°-30° N is primarily due to the Indian monsoon region (+20%). Likewise, South East

179

Asia experienced a strong increase in record-wet months during boreal winter (+17%) and for the

180

annual mean (+11%). Record-dry months significantly increased over Central Africa in all seasons

181

(42% ±3%) (Fig. 4b, e, h). Moreover, the number of record-wet months has declined over the same

182

region suggesting a shift in the distribution towards less rainfall consistent with pronounced downward

183

trends (Fig. c, f, i). Together with pronounced but insignificant increases in record-dry months over

184

northern South America this contributes to the overall drying signal shown for the zonally-averaged

185

tropical belt from 15° S to 15° N (Fig. 3). The southern subtropics are characterized by increased

186

record-wet months over southern South America in austral summer (+36%) and in the annual mean

187

(+31%) and an increased occurrence of annual record-dry months over southern Africa (+24%).

188

Summary and discussion

189

We report significant changes in the occurrence of observed record-breaking wet and dry months over

190

the period 1980-2013 in global land observations. These changes have distinct regional patterns and

191

are generally consistent with computed trends. The mid- to high latitudes in the northern hemisphere

192

have seen a wetting trend and associated increases in wet-records. The tropics, on the other hand, are

193

characterized by a significant increase in record-dry months over Central Africa in contrast to 6


194

pronounced increases in record-wet months over monsoon regions including India and South East

195

Asia.

196

Based on theory, it has long been expected that in response to climate change relatively dry regions

197

will get drier and relatively wet regions will get wetter [Held and Soden, 2006]. Observational studies

198

assessing these changes have largely focused on annual trends in mean precipitation which show

199

indications for subtropical drying and high latitude wetting over the latter half of the 20th century

200

consistent with model projections [Zhang et al., 2007; Noake et al., 2012; Allan et al., 2014;

201

Karnauskas and Ummenhofer, 2014]. Generally, however, poor data coverage and low signal-to-noise

202

rations, caused by the complex interplay of thermodynamic and dynamic drivers of precipitation [Stott

203

et al., 2010; Shepherd, 2014] are hampering the detection of regional rainfall changes. Our results

204

show that while trends in mean rainfall are often difficult to detect, record-statistics reveal significant

205

changes in both tails of the rainfall distribution, i.e. in heavy rainfall events and long dry spells.

206

Precipitation in the mid-latitudes is strongly connected to the position and strength of extratropical

207

storm tracks [Raible et al., 2007; Hawcroft et al., 2012; Pfahl and Wernli, 2012]. Thus changes in

208

rainfall pattern are likely driven by changes in storm track activity. In particular, the observed wetting

209

in the northern high latitudes is consistent with the reported poleward shift of the northern

210

extratropical storm tracks [Harnik and Chang, 2003; Trigo, 2006; Wang and Swail, 2006].

211

Besides dynamical drivers of rainfall changes (like extratropical storm activity), changes in the amount

212

of rainfall are also influenced by thermodynamic processes and the timescales at which both drivers

213

occur and interact. Consistently, the number of record-breaking wet days has increased over the period

214

1981-2010 in agreement with thermodynamic expectations based on the Clausius-Clapeyron equation

215

[Lehmann et al., 2015]. Here, we find a similar increase in record-wet months. This is reasonable since

216

synoptic storms can last for several days leading to extreme rainfall at both daily and monthly time

217

scales. The pronounced signal of northern high-latitude wetting is hence likely influences by both an

218

amplification of extratropical storm activity (due to a shift of storm tracks towards higher latitudes)

219

and the surplus of available water vapor in the atmosphere under rising temperatures.

220

In the tropics, we find contrasting changes in rainfall extremes between Central Africa and tropical

221

monsoon climates in India and South East Asia. In Central Africa, record-dry months increased by up

222

to 46% in 1980-2013 implying that approximately one out of three record-dry months would not have

223

occurred without long-term climate change. Drying trends are also observed over Brazil and the

224

Amazon, albeit with insignificant changes, leading to an overall drying signal over tropical land

225

between 15° S and 15° N during the last three decades. Conversely, India and South East Asia

226

experienced pronounced increases in record-wet months over the same period. This is consistent with

227

previous findings showing worldwide highest increases in daily record-breaking rainfall events over

228

these two regions [Lehmann et al., 2015]. The wetting trends in India and South East Asia started 7

55

56


229

around 1980 and 1990 (Fig. S9-S10), respectively, and thus at peak time of increased record-dry

230

months over Central Africa (Fig. S11). These two contrary trends likely explain the latitude band time

231

series plots over tropical land. Here, the reported dry period from 1980 to 2000 is replaced by a

232

wetting trend thereafter as indicated by a strong increase in record-wet months. This is in agreement

233

with trends in tropical precipitation based on different observations [Fig. 2.28 in IPCC, 2013]. Our

234

results are broadly consistent with previous studies who report positive precipitation trends in the wet

235

regions and negative trends in the dry regions of the tropics [Allan et al., 2010; Liu and Allan, 2013].

236

Time series with zero rainfall were removed from the analysis and thus we are not able to make

237

statements about changes in record-dry months in the driest regions, i.e. for large parts of southwestern

238

North America, Australia, and the Sahara (Fig. S1c, d). However, we find an increased occurrence of

239

record-dry months in southern Africa during the rainy season. Indeed, this region was struck by severe

240

droughts in the past leading to decreases in crop and stock production [Richard et al., 2001]. CMIP5

241

models project a decrease in soil-moisture content during the 21st century over southern Africa which

242

would further enhance the risk of severe droughts [Dai, 2012].

243

Whereas some regions are thus facing the risk of prolonged dry periods, the reported wetting at

244

northern mid- to high latitudes favors the occurrence of floodings. The observed increase in the

245

number of record-wet months is especially pronounced over central and eastern US, Europe and

246

Russia showing increases between +15% and +35%. These regions are strongly affected by

247

extratropical storm tracks and have experienced extreme rainfall events in the recent past leading to

248

severe floodings. Climate change will likely continue to alter the occurrence of record-breaking wet

249

and dry months in the future under increasing CO2 emissions with severe consequences for

250

agricultural production and food security.

251 252

Acknowledgments. We thank the Global Precipitation Climatology Centre for making their data

253

available. The work was supported by the German research Foundation (CO994/2-1) and the German

254

Federal Ministry of Education and Research (01LN1304A).

255 256

References

257 258

Alexander, L. V. et al. (2006), Global observed changes in daily climate extremes of temperature and precipitation, J. Geophys. Res., 111(D5), D05109, doi:10.1029/2005JD006290.

259 260

Allan, R. P., B. J. Soden, V. O. John, W. Ingram, and P. Good (2010), Current changes in tropical precipitation, Environ. Res. Lett., 5(2), 025205, doi:10.1088/1748-9326/5/2/025205.

261 262

Allan, R. P., C. Liu, M. Zahn, D. a. Lavers, E. Koukouvagias, and A. Bodas-Salcedo (2014), Physically Consistent Responses of the Global Atmospheric Hydrological Cycle in Models and 8


263

Observations, Surv. Geophys., 35(3), 533–552, doi:10.1007/s10712-012-9213-z.

264 265

Allen, M. R., and W. J. Ingram (2002), Constraints on future changes in climate and the hydrologic cycle., Nature, 419(6903), 224–32, doi:10.1038/nature01092.

266 267

Barlow, M., B. Zaitchik, S. Paz, E. Black, J. Evans, and A. Hoell (2015), A Review of Drought in the Middle East and Southwest Asia, J. Clim., 150729114230005, doi:10.1175/JCLI-D-13-00692.1.

268 269 270 271

Becker, a., P. Finger, A. Meyer-Christoffer, B. Rudolf, K. Schamm, U. Schneider, and M. Ziese (2013), A description of the global land-surface precipitation data products of the Global Precipitation Climatology Centre with sample applications including centennial (trend) analysis from 1901present, Earth Syst. Sci. Data, 5(1), 71–99, doi:10.5194/essd-5-71-2013.

272 273

Benestad, R. E. (2013), Association between trends in daily rainfall percentiles and the global mean temperature, J. Geophys. Res. Atmos., 118(19), 10,802–10,810, doi:10.1002/jgrd.50814.

274 275

Berg, P., C. Moseley, and J. O. Haerter (2013), Strong increase in convective precipitation in response to higher temperatures, Nat. Geosci., 6(3), 181–185, doi:10.1038/ngeo1731.

276 277

Chandler, R., and M. Scott (2011), Statistical Methods for Trend Detection and Analysis in the Environmental Sciences, John Wiley & Sons.

278 279

Coumou, D., A. Robinson, and S. Rahmstorf (2013), Global increase in record-breaking monthly-mean temperatures, Clim. Change, 118(3-4), 771–782, doi:10.1007/s10584-012-0668-1.

280 281

Dai, A. (2012), Increasing drought under global warming in observations and models, Nat. Clim. Chang., 3(1), 52–58, doi:10.1038/nclimate1633.

282 283 284 285

Field, C. B. et al. (2014a), Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press, Cambridge, UK, and New York, NY.

286 287 288 289

Field, C. B. et al. (2014b), Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part B: Regional Aspects Working Group II Contribution to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press, Cambridge, UK, and New York, NY.

290 291 292

Frieler, K., M. Meinshausen, T. Schneider Von Deimling, T. Andrews, and P. Forster (2011), Changes in global-mean precipitation in response to warming, greenhouse gas forcing and black carbon, Geophys. Res. Lett., 38(4), 1–5, doi:10.1029/2010GL045953.

293 294 295

Groisman, P. Y., R. W. Knight, D. R. Easterling, T. R. Karl, G. C. Hegerl, and V. N. Razuvaev (2005), Trends in Intense Precipitation in the Climate Record, J. Clim., 18(9), 1326–1350, doi:10.1175/JCLI3339.1.

296 297 298

Harnik, N., and E. K. M. Chang (2003), Storm track variations as seen in radiosonde observations and reanalysis data, J. Clim., 16(3), 480–495, doi:10.1175/15200442(2003)0162.0.CO;2.

299 300 301

Hawcroft, M. K., L. C. Shaffrey, K. I. Hodges, and H. F. Dacre (2012), How much Northern Hemisphere precipitation is associated with extratropical cyclones?, Geophys. Res. Lett., 39(24), L24809, doi:10.1029/2012GL053866.

302 303

Heberger, M. (2011), Australia’s Millennium Drought: Impacts and Responses, in The World’s Water, pp. 97–125.

304 305

Held, I. M., and B. J. Soden (2006), Robust responses of the hydrological cycle to global warming, J. Clim., 19(21), 5686–5699, doi:10.1175/JCLI3990.1.

306

Hoerling, M., J. Eischeid, J. Perlwitz, X. Quan, T. Zhang, and P. Pegion (2012), On the Increased 9

57

58


307 308

Frequency of Mediterranean Drought, J. Clim., 25(6), 2146–2161, doi:10.1175/JCLI-D-1100296.1.

309 310 311 312

IPCC (2013), Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, edited by T. F. Stocker, D. Qin, G.-K. Plattner, M. Tignor, S. K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex, and P. M. Midgley, Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

313 314

Karnauskas, K. B., and C. C. Ummenhofer (2014), On the dynamics of the Hadley circulation and subtropical drying, Clim. Dyn., 42(9-10), 2259–2269, doi:10.1007/s00382-014-2129-1.

315 316

Katz, R. W., and B. G. Brown (1992), Extreme events in a changing climate: Variability is more important than averages, Clim. Change, 21(3), 289–302, doi:10.1007/BF00139728.

317 318 319

Kelley, C. P., S. Mohtadi, M. A. Cane, R. Seager, and Y. Kushnir (2015), Climate change in the Fertile Crescent and implications of the recent Syrian drought, Proc. Natl. Acad. Sci., 112(11), 3241– 3246, doi:10.1073/pnas.1421533112.

320 321

Lehmann, J., D. Coumou, and K. Frieler (2015), Increased record-breaking precipitation events under global warming, Clim. Change, doi:10.1007/s10584-015-1434-y.

322 323

Liu, C., and R. P. Allan (2013), Observed and simulated precipitation responses in wet and dry regions 1850–2100, Environ. Res. Lett., 8(3), 034002, doi:10.1088/1748-9326/8/3/034002.

324 325

Min, S.-K., X. Zhang, F. W. Zwiers, and G. C. Hegerl (2011), Human contribution to more-intense precipitation extremes., Nature, 470(7334), 378–81, doi:10.1038/nature09763.

326 327 328

NOAA (2015), May 2015 was wettest month ever recorded in U.S., Publ. online June 2015. Available from: https://www.climate.gov/news-features/featured-images/may-2015-was-wettestmonth-ever-recorded-us

329 330

Noake, K., D. Polson, G. Hegerl, and X. Zhang (2012), Changes in seasonal land precipitation during the latter twentieth-century, Geophys. Res. Lett., 39(3), 1–5, doi:10.1029/2011GL050405.

331 332 333

Pall, P., M. R. Allen, and D. a. Stone (2007), Testing the Clausius–Clapeyron constraint on changes in extreme precipitation under CO2 warming, Clim. Dyn., 28(4), 351–363, doi:10.1007/s00382006-0180-2.

334 335 336

Pall, P., T. Aina, D. a Stone, P. a Stott, T. Nozawa, A. G. J. Hilberts, D. Lohmann, and M. R. Allen (2011), Anthropogenic greenhouse gas contribution to flood risk in England and Wales in autumn 2000., Nature, 470(7334), 382–5, doi:10.1038/nature09762.

337 338

Pfahl, S., and H. Wernli (2012), Quantifying the Relevance of Cyclones for Precipitation Extremes, J. Clim., 25(19), 6770–6780, doi:10.1175/JCLI-D-11-00705.1.

339 340 341 342

Raible, C. C., M. Yoshimori, T. F. Stocker, and C. Casty (2007), Extreme midlatitude cyclones and their implications for precipitation and wind speed extremes in simulations of the Maunder Minimum versus present day conditions, Clim. Dyn., 28, 409–423, doi:10.1007/s00382-0060188-7.

343 344 345

Richard, Y., N. Fauchereau, I. Poccard, M. Rouault, and S. Trzaska (2001), 20th century droughts in southern Africa: spatial and temporal variability, teleconnections with oceanic and atmospheric conditions, Int. J. Climatol., 21(7), 873–885, doi:10.1002/joc.656.

346 347 348

Schneider, U., A. Becker, P. Finger, A. Meyer-Christoffer, B. Rudolf, and M. Ziese (2015), GPCC Full Data Reanalysis Version 7.0 at 0.5°: Monthly Land-Surface Precipitation from Rain- Gauges built on GTS-based and Historic Data.

349 350

Schönwiese, C.-D., J. Grieser, and S. Trömel (2003), Secular change of extreme monthly precipitation in Europe, Theor. Appl. Climatol., 75(3-4), 245–250, doi:10.1007/s00704-003-0728-6. 10


351 352

Shepherd, T. G. (2014), Atmospheric circulation as a source of uncertainty in climate change projections, Nat. Geosci., (September), 6–11, doi:10.1038/ngeo2253.

353 354

Shiu, C.-J., S. C. Liu, C. Fu, A. Dai, and Y. Sun (2012), How much do precipitation extremes change in a warming climate?, Geophys. Res. Lett., 39(17), doi:10.1029/2012GL052762.

355 356 357

Simmons, a. J., P. Poli, D. P. Dee, P. Berrisford, H. Hersbach, S. Kobayashi, and C. Peubey (2014), Estimating low-frequency variability and trends in atmospheric temperature using ERA-Interim, Q. J. R. Meteorol. Soc., 140(679), 329–353, doi:10.1002/qj.2317.

358 359

Singh, M. S., and P. a. O’Gorman (2014), Influence of microphysics on the scaling of precipitation extremes with temperature, Geophys. Res. Lett., doi:10.1002/2014GL061222.

360 361

Stephens, E., and H. Cloke (2014), Improving flood forecasts for better flood preparedness in the UK (and beyond), Geogr. J., 180(4), 310–316, doi:10.1111/geoj.12103.

362 363 364

Stott, P. a., N. P. Gillett, G. C. Hegerl, D. J. Karoly, D. a. Stone, X. Zhang, and F. Zwiers (2010), Detection and attribution of climate change: A regional perspective, Wiley Interdiscip. Rev. Clim. Chang., 1(2), 192–211, doi:10.1002/wcc.34.

365 366

Sun, F., M. L. Roderick, and G. D. Farquhar (2012), Changes in the variability of global land precipitation, Geophys. Res. Lett., 39(19), L19402, doi:10.1029/2012GL053369.

367 368 369

Tett, S. F. B., K. Deans, E. Mazza, and J. Mollard (2013), Are recent wet northwestern european sumemrs a response to sea ice retreat? [in “Explaining Extremes of 2012 from a Climate Perspective”], Bull. Am. Meteorol. Soc., 94(9), S32–S35.

370 371

Trenberth, K. (2011), Changes in precipitation with climate change, Clim. Res., 47(1), 123–138, doi:10.3354/cr00953.

372 373 374

Trigo, I. F. (2006), Climatology and interannual variability of storm-tracks in the Euro-Atlantic sector: a comparison between ERA-40 and NCEP/NCAR reanalyses, Clim. Dyn., 26(2-3), 127–143, doi:10.1007/s00382-005-0065-9.

375 376 377

Wang, S.-Y., L. Hipps, R. R. Gillies, and J.-H. Yoon (2014), Probable causes of the abnormal ridge accompanying the 2013-2014 California drought: ENSO precursor and anthropogenic warming footprint, Geophys. Res. Lett., 41(9), 3220–3226, doi:10.1002/2014GL059748.

378 379

Wang, X., and V. Swail (2006), Climatology and changes of extratropical cyclone activity: Comparison of ERA-40 with NCEP-NCAR reanalysis for 1958-2001, J. Clim., 19(13), 3145–3166.

380 381

Westra, S., L. V. Alexander, and F. W. Zwiers (2013), Global Increasing Trends in Annual Maximum Daily Precipitation, J. Clim., 26(11), 3904–3918, doi:10.1175/JCLI-D-12-00502.1.

382 383 384

Zhang, X., F. W. Zwiers, G. C. Hegerl, F. H. Lambert, N. P. Gillett, S. Solomon, P. a Stott, and T. Nozawa (2007), Detection of human influence on twentieth-century precipitation trends., Nature, 448(7152), 461–5, doi:10.1038/nature06025.

385 386 387

Zhang, X., H. Wan, F. W. Zwiers, G. C. Hegerl, and S.-K. Min (2013), Attributing intensification of precipitation extremes to human influence, Geophys. Res. Lett., 40(19), 5252–5257, doi:10.1002/grl.51010.

388 389

Zheng, F., S. Westra, and M. Leonard (2015), Opposing local precipitation extremes, Nat. Clim. Chang., 5(5), 389–390, doi:10.1038/nclimate2579.

390

11

59

60

391


Figures

392

393 394

Fig. 1 Locations of GPCC monthly rainfall observations on a 0.5° x 0.5° grid shown for January.

395

Colors indicate the number of years providing rainfall observations after applying the data

396

requirements. Spatial and temporal coverage differs between different calendar months (see Fig. S2).

397

Dotted lines mark the latitudinal boundaries for the regions used in Fig. 2.

12


398 399

Fig. 2 Record-wet (blue vertical bars) and record-dry (brown vertical bars) anomalies and their long-

400

term nonlinear trend based on singular spectrum analysis (solid lines) for (a) globally, (b) northern

401

extratropics (40°-90° N), (c) northern subtropics (20°-40° N), (d) tropics (20°S-20°N), and (e)

402

southern subtropics (40°-20° S). The shaded areas reflect the 95% confidence interval of the long-term

403

changes in the stationary model and are thus directly comparable to the solid-lines. (f)-(j) and (k)-(o)

404

are the same as (a)-(e), respectively, but for seasonal record-anomalies representing NDJFM (middle

405

panel) and MJJAS (right panel).

13

61

62


406 407

Fig. 3 Zonal-mean record-anomalies and Mann-Kendall trends. The upper panel shows the 1980-2013

408

record-wet (blue) and dry (brown) anomaly for (a) annual and (d, g) seasonal averages. The middle

409

panel (b, e, h) shows zonal-means of τ derived from a Mann-Kendall test based on the full time series

410

and the lower panel (c, f, i) indicates the ratio of land areas with positive τ compared to the total land

411

area. In all figures, crosses indicate zonal-mean values and the shaded area reflects the 95%

412

confidence range derived from 10,000 Monte-Carlo simulations.

14

and (c) boreal summer averages. Similarly, the middle panel (b, e, h) shows regional-mean record-dry anomalies. In the lower panel, the regional-mean Mann-

Kendall parameter τ calculated for the full time series is shown. Boxes indicate the regions for which the quantities are calculated with significant changes

marked with a black cross. Horizontal dotted lines indicate the boundaries for the latitudinal bands used in Fig. 3.

415

416

417

15

Fig. 4 Global maps of record-anomalies and Mann-Kendall trends. The upper panel shows the 1980-2013 record-wet anomaly for (a) annual, (b) boreal winter

414

413

2.2. Changes in record-wet and record-dry months in global land observations 63

64


Supplementary Materials for: Changes in record-wet and record-dry months in global land observations Jascha Lehmann*, Finn Mempel, and Dim Coumou *Corresponding author. E-mail: [email protected]

Figures S3-S5 and S6-S8 are based on GPCC datasets with grid resolution of 1° x 1° and 2.5° x 2.5°, respectively. From these figures the same conclusions can be drawn as from Figures 2-4 in the main manuscript. As expected, the magnitudes of record-anomaly changes slightly increase with larger grid cells. The reason for this is that in a first-order approximation the number of record-events scales with μ/σ, where μ is the trend in the time series and σ the standard deviation [Coumou et al., 2013]. A given trend in precipitation will thus lead to more record-events if internal variability is smaller which is likely the case for larger grid cells. Varying numbers of measurement stations contribute to this effect. In the main manuscript we present “conservative” results for smallest grid cells and thus smallest changes in station numbers.


65

Additional figures

Fig. S1 Data requirements. Shown are those grid cells which were removed from the analysis because of too short time series (left column) or zero rainfall in the time series (right column). Maps are exemplarily shown for January (upper panel) and June (lower panel).

66


Fig. S2 Locations of GPCC monthly rainfall observations on a 0.5° x 0.5° grid as shown in Fig. 1 of the main manuscript but for (a) January and (b) June.


67

Fig. S3 Time series plots of record-wet and record-dry months as shown in Fig. 2 of the main manuscript but based on GPCC at 1° x 1° spatial resolution.

68


Fig. S4 Zonal-mean record-anomalies and Mann-Kendall trends as shown in Fig. 3 of the main manuscript but based on GPCC at 1° x 1° spatial resolution.

Fig. S5 Regional-mean record-anomalies and Mann-Kendall trends as shown in Fig. 4 of the main manuscript but based on GPCC at 1° x 1° spatial resolution.

2.2. Changes in record-wet and record-dry months in global land observations 69

70


Fig. S6 Time series plots of record-wet and record-dry months as shown in Fig. 2 of the main manuscript but based on GPCC at 2.5° x 2.5° spatial resolution.


71

Fig. S7 Zonal-mean record-anomalies and Mann-Kendall trends as shown in Fig. 3 of the main manuscript but based on GPCC at 2.5° x 2.5° spatial resolution.

Fig. S8 Regional-mean record-anomalies and Mann-Kendall trends as shown in Fig. 4 of the main manuscript but based on GPCC at 2.5° x 2.5° spatial resolution.

72 2. Original Manuscripts


Fig. S9 Time series plots of record-wet anomaly for India.

Fig. S10 Time series plots of record-wet anomaly for South East Asia.

73

74


Fig. S11 Time series plots of record-dry anomaly for Central Africa.

References Coumou, D., A. Robinson, and S. Rahmstorf (2013), Global increase in record-breaking monthly-mean temperatures, Clim. Change, 118(3-4), 771–782, doi:10.1007/s10584-0120668-1.

2.3. The weakening summer circulation in the Northern Hemisphere mid-latitudes Dim Coumou, Jascha Lehmann, and Johanna Beckmann. Significant weakening of summer circulation is reported for three key dynamical quantities: (i) the zonal-mean zonal wind, (ii) the eddy kinetic energy, and (iii) the amplitude of fast-moving Rossby waves. It is shown that a reduced equator-to-pole temperature gradient associated with Arctic amplification contributed to the observed weakening of atmospheric circulation which in turn likely favored more persistent heat waves over continental land in summer. Jascha Lehmann processed and analyzed the CMIP5 data. He was responsible for the analysis of the zonal wind component and the EKE. Dim Coumou developed the idea, performed analyses and wrote the text with the help of Jascha Lehmann. All authors participated in the interpretation of the results and helped to improve the manuscript. Published in Science 348, 324-327 (2015), DOI 10.1126/science.126176.

2.3. The weakening summer circulation in the Northern Hemisphere mid-latitudes

77

R ES E A RC H | R E PO R TS

CLIMATE CHANGE

The weakening summer circulation in the Northern Hemisphere mid-latitudes Dim Coumou,1* Jascha Lehmann,1,2 Johanna Beckmann1,2 Rapid warming in the Arctic could influence mid-latitude circulation by reducing the poleward temperature gradient. The largest changes are generally expected in autumn or winter, but whether significant changes have occurred is debated. Here we report significant weakening of summer circulation detected in three key dynamical quantities: (i) the zonal-mean zonal wind, (ii) the eddy kinetic energy (EKE), and (iii) the amplitude of fast-moving Rossby waves. Weakening of the zonal wind is explained by a reduction in the poleward temperature gradient. Changes in Rossby waves and EKE are consistent with regression analyses of climate model projections and changes over the seasonal cycle. Monthly heat extremes are associated with low EKE, and thus the observed weakening might have contributed to more persistent heat waves in recent summers.

E

nhanced warming in the Arctic could change circulation patterns in the mid-latitudes by reducing the pole–to–mid-latitude thermal gradient (1–3). This hypothesis, which was first proposed in the 1970s based on model experiments (4, 5), has recently received considerable attention due to rapid observed warming in the Arctic (6–9), associated with a decline in sea ice and other factors (1, 3, 10). Most studies addressing the link between Arctic change and mid-latitude weather have focused on winter circulation. The extra heat stored in the ocean owing to sea-ice loss is released into the atmosphere by late autumn or early winter, when air temperatures drop below sea surface temperatures. Consequently, the largest absolute increases in Arctic geopotential height have been detected in autumn and winter (6), consistent with climate model simulations (11). Autumn has, at least in the western half of the hemisphere, 1

Potsdam Institute for Climate Impact Research, Earth System Analysis, 14412 Potsdam, Germany. 2University of Potsdam, Potsdam, Germany.

7.6

1980 1985 1990 1995 2000 2005 2010

pval 0 slope −0.24 (ms−1 35yr)

1980 1985 1990 1995 2000 2005 2010

7.4 7.2 7.0

UT (m s)

7.8 pval 0 slope −0.91 (m2s−2 35yr)

7.2

7.4

7.6

U (m s)

9.0 8.5 8.0 7.5

E K E (m2 s2)

9.5 10.0

*Corresponding author. E-mail: [email protected]

also seen a reduction in the zonal-mean flow (6, 12). This might cause a slowdown in wave propagation (6), but the results are sensitive to the exact metrics used to describe waves (12, 13). Thus, whether observed changes in geopotential height have affected mid-latitude Rossby waves remains disputed (6, 12–14). We studied changes in mid-latitude circulation in boreal summer instead. Although the oceanic heat flux is smaller in this season (11), Arctic amplification has reduced the pole–to–mid-latitude temperature gradient (1), and Arctic geopotential heights have increased (6). These changes are likely to be related to the earlier loss of snow cover over land and increased Arctic sea surface temperatures where sea ice has been lost (7). In recent summers, mid-latitude circulation has been dominated by a negative Arctic Oscillation index; i.e., anomalously small pressure differences between mid- and high-latitudes (7, 15–17). Moreover, several recent heat waves, such as in Russia in 2010, were associated with persistent hemispheric circulation patterns (15, 16, 18). Generally, the large-scale mid-latitude atmosphere dynamics [supplementary materials (SM)

text S4] are characterized by (i) fast-traveling free Rossby waves (the so-called synoptic transients) with zonal wave numbers typically larger than 6, and (ii) quasi-stationary Rossby waves with normally smaller wave numbers as a response to quasi-stationary diabatic and orographic forcing (15, 19–21). We focus on the first. These waves are associated with synoptic-scale cyclones (storms) and anticyclones (high-pressure systems), which form the storm track regions in the mid-latitudes. They have a relatively fast phase velocity (i.e., eastward propagation) and cause weather variability on time scales of less than a week. Typically, the intensity of synoptic-scale wave (or eddy) activity is estimated by applying a 2.5- to 6-day bandpass filter to high-resolution wind field data (22–24). This way, the total eddy kinetic energy (EKE) is extracted, which is a measure of the interplay between the intensity and frequency of high- and low-pressure systems associated with fast-traveling Rossby waves. Due to the quasi-stationary nature of thermally and orographically forced waves, as analyzed in related studies (15, 16, 25, 26), they have lower frequencies and are thus excluded from our EKE computations (SM text S4). We calculated EKE in the Northern Hemisphere from daily ERA-Interim wind fields (27), using a 2.5- to 6-day bandpass filter [see (23, 24, 28)]. We limited our analysis to the satellite-covered period (after 1979) to minimize the effects of changes in the observing system (SM text S2). The 1979–2013 period has seen a steady decline in summertime EKE (Fig. 1A). This decline is statistically significant at the 1% level and observed at all pressure levels, with the strongest relative changes in the lower to mid-troposphere (figs. S1 to S4 and table S1). Moreover, it occurs over the full hemisphere and over all relevant latitudes (fig. S5). The observed changes are thus not due to a north-south shift of storm tracks but due to a spatially homogeneous weakening. Other reanalysis products give similar results (figs. S6 to S11). For the other seasons, trends in EKE are also downward, but none are significant (figs. S12 to S17). The decline in summer EKE is accompanied by a long-term decline in the zonal-mean zonal wind (U, Fig. 1B). Again, this weakening of the

pval 0 slope −0.21 (ms−1 35yr)

1980 1985 1990 1995 2000 2005 2010

Fig. 1. Weakening summer circulation in the mid-latitudes. Absolute changes in (A) EKE, (B) zonal wind U, and (C) thermal wind UT over 1979–2013 in summer (June, July, and August). Variables are calculated at 500 mb and averaged over 35°N to 70°N and all longitudes, with gray lines plotting observations, solid black lines the linear trend, and dashed black lines the T1 residual SE range. Slope and P values for the trend estimates are given in the panels.

324

17 APRIL 2015 • VOL 348 ISSUE 6232

sciencemag.org SCIENCE

78


RE S EAR CH | R E P O R T S

0 −5 −10 −15

∆ EKE (%)

−20 −25 −30

slope = 1.4 p−val = 3.5e−05 −10

−5

tive changes in U and EKE over the past 35 years map reasonably well on the regression of projected future changes (Fig. 2). The pronounced weakening in EKE should also be reflected in changed wave characteristics. To test this, we applied spectral analysis to the north-south wind component v in daily wind field data and calculated amplitude (Av), phase speed, and period for wave numbers 1 to 15 (SM text S1.1). This way, fast-moving and quasi-stationary waves are not explicitly separated (as in EKE by using bandpass filtering), but because we used daily data, the mean wave amplitudes are dominated by fast-moving waves (SM text S4). The results are therefore comparable with EKE. We determined the wave quantities for the northsouth wind component v averaged over 35°N to 70°N, and Av thus reflects wind speeds with units of meters per second. The amplitudes of all wave numbers except 7 have declined, with significant reductions in waves 1, 3, 4, 6, and 10 and in the mean amplitude of all waves. These changes are robust, detected in ERA-Interim and NCEP-NCAR (National Centers for Environmental Prediction– National Center for Atmospheric Research) data and for different pressure levels, with the strongest changes in the mid-troposphere (Fig. 3A and fig. S19). This vertical pattern is consistent with the more pronounced changes in EKE, U (table S1), and poleward temperature gradient (1) in the lower troposphere. The mean amplitude declined by –5% over 1979–2013 (Fig. 3A), similar to the relative reduction in U (fig. S1B). This is consistent with the seasonal correlation of these quantities (Fig. 3B), which shows that, to a first order, daily anomalies in mean Av scale with those in U. This positive correlation is expected as daily wind fields, and thus Av, will be dominated by transient eddies, because their kinetic energy is nearly an order of magnitude larger than that of quasi-stationary waves (30). Transient eddies are not only forced by the zonal flow via vertical shear (and thus baroclinicity) but can also accelerate it via the eddy-driven jet (1, 31), explaining the positive correlation between U and Av (SM text S4). The reduction of –5% in mean wind speed (Av) implies a –10% reduction

declines primarily because of decreased vertical wind shear associated with weakening of the zonal-mean flow (24). This projected reduction in EKE is spatially homogeneous, similar to the observed changes. Regression analysis of future changes in EKE and in zonal flow for individual CMIP5 models reveals a significant linear relationship (Fig. 2). The regression slope of ~1.4 indicates that a reduction in U is associated with a more pronounced reduction in EKE. This is seen at all pressure levels, with the regression slope increasing with altitude (fig. S18). Increased static stability plays a role as well (24, 29), which explains why the linear regression crosses the y axis at negative values: Even for zero change in U, increased static stability in a warmer climate causes EKE to decline. The observed rela-

zonal flow is seen at all altitudes and in different reanalysis products (figs. S1 to S4 and S6 to S10). The long-term weakening of the zonal flow is consistent with the decline in the pole– to–mid-latitude thermal gradient. This is shown by the downward trend of similar magnitude in thermal wind UT (Fig. 1C), which depends on changes in the temperature gradient only (eq. S2). Although the relative decrease in EKE has been by 8 to 15% (depending on pressure level) over the 35-year period, the zonal flow weakened by only 4 to 6% (table S1). A similar relationship between changes in EKE and zonal flow is seen in future projections of CMIP5 (Coupled Model Intercomparison Project Phase 5) climate models. Under a high-emission scenario, summer EKE

0

ERA−Interim multi−model mean ACCESS1−0 CMCC−CESM CMCC−CM CMCC−CMS CNRM−CM5 CanESM2 FGOALS−g2 GFDL−CM3 GFDL−ESM2G GFDL−ESM2M HadGEM2−CC IPSL−CM5A−LR IPSL−CM5A−MR IPSL−CM5B−LR MIROC−ESM MIROC−ESM−CHEM MIROC5 MPI−ESM−LR MPI−ESM−MR MRI−CGCM3 NorESM1−M inmcm4

5

∆ U (%)

SCIENCE sciencemag.org

40

1e−04

3e−04

0

∆ A v (%)

20

10

3e−0

−20

0

4

−40

−10

Trend (%/35yr)

2e−04

−20

Fig. 3. (A) Trends in planetary wave amplitudes (Av, red) and phase speed (black) at 500 mb in summer for wave numbers 1 to 15 and for the mean of all waves (M) in units of percentage change per 35 years; i.e., the period 1979–2013. Solid circles indicate 5% statistical significance, gray-filled circles indicate 10% statistical significance, and open circles are not significant. (B) Two-dimensional probability density distribution of daily deviations (in percentage change of their annual mean climatological values) of the zonal flow and the mean amplitude of waves 1 to 15. The bisecting line is shown in black.

20

Fig. 2. Relationship between relative changes (D) in EKE and U in climate model projections. The percentage change in the future (2081–2100, under scenario RCP8.5) relative to 1981–2000 for individual CMIP5 climate models is shown. Both quantities are averaged over 35°N to 70°N, all longitudes, and over 850 to 250 mb (mass-weighted). The solid black line shows the linear fit, with slope and P value given at lower right. Relative changes in EKE and U in the ERA-Interim data are given for the 1979–2013 period.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 M

Wave number

−40

−20

0

20

40

∆ U (%)

17 APRIL 2015 • VOL 348 ISSUE 6232

325


79

R ES E A RC H | R E PO R TS

30 20 10 0

EKE (m 2 s-2 )

40

Moscow 55.5N37.5E

−4 2 -2 o

−2

0

2

4

6 o

Regression slope (m s / C)

Temperature anomaly ( C)

Fig. 4. Regression analysis between EKE and near-surface temperature during the summer months (June, July, and August). (A) Slope of the regression analysis. Both variables were linearly detrended, and stippling indicates significance at the 5% level. (B) EKE plotted against near-surface temperature anomaly for Moscow [red dot in (A)] for individual summer months, showing that the 10% coldest summer months (left of the vertical blue dotted line) have a substantially higher EKE (14 m2/s2; i.e., the horizontal blue dashed line) and the 10% warmest months (right of the vertical red dotted line) a substantially lower EKE (5.7 m2/s2; i.e., the horizontal red dashed line). Red and purple circles indicate, respectively, July and August 2010.

in kinetic energy, which is in good agreement with the bandpass-filtered results. Wave 10 has seen a significant reduction in both amplitude and phase speed (Fig. 3A), which dropped respectively by –11% and –20% over the 1979–2013 period, with both negative trends acting to reduce EKE. A reduction in amplitude means lower wind speeds associated with weaker high- and low-pressure synoptic weather systems and thus lower EKE. A reduced phase speed implies more-persistent synoptic weather systems and fewer of them over the full season. The probability-density distribution of the wave period shows that wave periods in the EKE-relevant range (2.5 to 6 days) are dominated by wave 10 (fig. S21). During roughly half of all summer days, wave 10 had a wave period within this range. This suggests that the reduction in amplitude and phase speed of wave 10 contributed substantially to the reduction in EKE. Summer EKE declined by 8 to 15% over the past decades, whereas the CMIP5 models project similar changes only by the end of the 21st century under a high-emission scenario (24). Either the climate models underpredict dynamical changes, or multidecadal variability played a role in the observed changes. In the other seasons, dynamics weakened as well, but here significant changes are only detected for the zonal-mean flow in autumn (SM text S3). Although the Arctic has warmed most in winter (1), the strongest changes in the meridional temperature gradient within the mid-latitudes occurred in summer, followed by autumn (fig. S17). Therefore, UT and U itself weakened most in those seasons (fig. S17). The smaller year-to-year variability in those seasons (and especially in summer), as compared to winter, improves signal-to-noise ratios, making trend detection possible at an earlier stage (fig. S15). Likewise, variability in summer EKE is only half that of the other seasons (fig. S16B), and hence the signal-to-noise ratio is much larger for summer. In fact, summer EKE has weakened by more than two standard deviations over 1979–2013 (fig. S16C). Therefore, contrary to previous sug326

17 APRIL 2015 • VOL 348 ISSUE 6232

gestions (1–3), the influence of Arctic amplification on mid-latitude weather is unlikely to be limited to autumn and winter only. In summer, synoptic storms transport moist and cool air from the oceans to the continents, bringing relief during periods of oppressive heat. Low cyclone activity over Europe in recent years has led to more-persistent weather (32, 33) contributing to prolonged heat waves. Regression analysis between EKE and near-surface temperature for summer months reveals that over mid-latitude continental regions, these quantities are negatively correlated (Fig. 4A). Thus, hot summer months are associated with low EKE (SM text S5). Over most of Eurasia and the United States, the negative regression slope is significant at the 5% level. In these storm track–affected regions, EKE in the 10% hottest months was only about half its summer climatological value (Fig. 4B and figs. S22 to S24). Low cyclone activity (and thus low EKE) imply that cool maritime air masses become less frequent, creating favorable conditions for the buildup of heat and drought over continents. This probably prolongs the duration of blocking weather systems, as, for example, during the Russian heat wave of 2010 (18, 34). In particular, the record-breaking July temperatures over Moscow were associated with extremely low EKE (Fig. 4B). Recent studies have emphasized the importance of quasi-stationary waves for summer heat extremes (15, 16, 25, 35), showing that the frequency of wave-resonance events associated with high-amplitude quasi-stationary waves has increased since the onset of rapid Arctic amplification in 2000 (16). Here we show that the amplitude of fast-moving waves has steadily decreased, and also that the rate in this weakening seems to have increased since 2000 (fig. S25). Both of these observations are consistent with more-persistent summer weather (SM text S6). Low monthly EKE implies low weather variability within that month, indicating persistent weather conditions, consistent with quasi-stationary waves. The long-term reduction in EKE should lead

to a reduction in weather variability on short time scales (less than a week), in agreement with the reduced intraseasonal daily temperature variance observed (36) and theoretical arguments (37). However, our results show that low EKE is associated with heat extremes on monthly time scales. Therefore, on such longer time scales, variability might actually increase due to a reduction in EKE. This seems consistent with Huntingford et al. (38), who report that the largest increase in interannual seasonal temperature variance occurred in the mid-latitude boreal summer. To test this hypothesis, studies are needed that quantify both interannual and intraannual variability on all relevant subseasonal time scales. This study shows that boreal summer circulation has weakened, together with a reduction in the pole–to–mid-latitude temperature gradient. This has made weather more persistent and hence favored the occurrence of prolonged heat extremes. REFERENCES AND NOTES

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16.

J. Cohen et al., Nat. Geosci. 7, 627–637 (2014). D. Budikova, Global Planet. Change 68, 149–163 (2009). J. E. Walsh, Global Planet. Change 117, 52–63 (2014). R. L. Newson, Nature 241, 39–40 (1973). M. Warshaw, R. R. Rapp, J. Appl. Meteorol. 12, 43–49 (1973). J. A. Francis, S. J. Vavrus, Geophys. Res. Lett. 39, L06801 (2012). J. E. Overland, J. A. Francis, E. Hanna, M. Wang, Geophys. Res. Lett. 39, L19804 (2012). R. Jaiser, K. Dethloff, D. Handorf, A. Rinke, J. Cohen, Tellus Ser. A Dyn. Meterol. Oceanogr. 64, 1–11 (2012). J. L. Cohen, J. C. Furtado, M. A. Barlow, V. A. Alexeev, J. E. Cherry, Environ. Res. Lett. 7, 014007 (2012). F. Pithan, T. Mauritsen, Nat. Geosci. 7, 181–184 (2014). J. A. Screen, I. Simmonds, C. Deser, R. Tomas, J. Clim. 26, 1230–1248 (2013). E. A. Barnes, Geophys. Res. Lett. 40, 4734–4739 (2013). J. A. Screen, I. Simmonds, Geophys. Res. Lett. 40, 959–964 (2013). E. Kintisch, Science 344, 250–253 (2014). V. Petoukhov, S. Rahmstorf, S. Petri, H. J. Schellnhuber, Proc. Natl. Acad. Sci. U.S.A. 110, 5336–5341 (2013). D. Coumou, V. Petoukhov, S. Rahmstorf, S. Petri, H. J. Schellnhuber, Proc. Natl. Acad. Sci. U.S.A. 111, 12331–12336 (2014).

sciencemag.org SCIENCE

80


RE S EAR CH | R E P O R T S

17. J. A. Screen, Environ. Res. Lett. 8, 044015 (2013). 18. S. Schubert, H. Wang, M. Suarez, J. Clim. 24, 4773–4792 (2011). 19. J. Pedlosky, Geophysical Fluid Dynamics (Springer, New York, 1979). 20. K. Fraedrich, H. Böttger, J. Atmos. Sci. 35, 745–750 (1978). 21. G. J. Boer, T. G. Shepherd, J. Atmos. Sci. 40, 164–184 (1983). 22. M. Blackmon, J. Atmos. Sci. 33, 1607–1623 (1976). 23. D. Coumou, V. Petoukhov, A. V. Eliseev, Nonlinear Process. Geophys. 18, 807–827 (2011). 24. J. Lehmann, D. Coumou, K. Frieler, A. V. Eliseev, A. Levermann, Environ. Res. Lett. 9, 084002 (2014). 25. H. Teng, G. Branstator, H. Wang, G. Meehl, W. M. Washington, Nat. Geosci. 6, 1–6 (2013). 26. K. E. Trenberth, J. T. Fasullo, G. Branstator, A. S. Phillips, Nat. Clim. Change 4, 911–916 (2014). 27. D. P. Dee et al., Q. J. R. Meteorol. Soc. 137, 553–597 (2011). 28. M. Murakami, Mon. Weather Rev. 107, 994–1013 (1979).

29. J. Lu, G. Chen, D. M. W. Frierson, J. Clim. 21, 5835–5851 (2008). 30. J. P. Peixoto, A. H. Oort, Physics of Climate (American Institute of Physics, New York, 1992). 31. T. Woollings, M. Blackburn, J. Clim. 25, 886–902 (2012). 32. J. Kyselý, R. Huth, Theor. Appl. Climatol. 85, 19–36 (2005). 33. J. Kyselý, Global Planet. Change 62, 147–163 (2008). 34. R. Dole et al., Geophys. Res. Lett. 38, L06702 (2011). 35. J. A. Screen, I. Simmonds, Nat. Clim. Chang. 4, 704–709 (2014). 36. J. A. Screen, Nat. Clim. Chang. 4, 577–582 (2014). 37. T. Schneider, T. Bischoff, H. Plotka, J. Clim. 28, 2312–2331 (2014). 38. C. Huntingford, P. D. Jones, V. N. Livina, T. M. Lenton, P. M. Cox, Nature 500, 327–330 (2013). ACKN OWLED GMEN TS

We thank the CMIP5 climate modeling groups and the European Centre for Medium-Range Weather Forecasts and NCEP-NCAR for making their model and reanalysis data available. Comments

26 September 2014; accepted 26 February 2015 Published online 12 March 2015; 10.1126/science.1261768

Volume loss from Antarctic ice shelves is accelerating Fernando S. Paolo,1* Helen A. Fricker,1 Laurie Padman2 The floating ice shelves surrounding the Antarctic Ice Sheet restrain the grounded ice-sheet flow. Thinning of an ice shelf reduces this effect, leading to an increase in ice discharge to the ocean. Using 18 years of continuous satellite radar altimeter observations, we have computed decadal-scale changes in ice-shelf thickness around the Antarctic continent. Overall, average ice-shelf volume change accelerated from negligible loss at + 64 cubic kilometers per year for 1994–2003 to rapid loss of 310 – + 74 cubic kilometers 25 – per year for 2003–2012. West Antarctic losses increased by ~70% in the past decade, and earlier volume gain by East Antarctic ice shelves ceased. In the Amundsen and Bellingshausen regions, some ice shelves have lost up to 18% of their thickness in less than two decades.

T 1

Scripps Institution of Oceanography, University of California, San Diego, CA, USA. 2Earth & Space Research, Corvallis, OR, USA. *Corresponding author. E-mail: [email protected]

SCIENCE sciencemag.org

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/348/6232/324/suppl/DC1 Text S1 to S6 Figs. S1 to S25 Table S1 References Data Deposition

ICE SHEETS

he Antarctic Ice Sheet gains mass through snowfall and loses mass at its margin through submarine melting and iceberg calving. These losses occur primarily from ice shelves, the floating extensions of the ice sheet. Antarctica’s grounded-ice loss has increased over the past two decades (1, 2), with the most rapid losses being along the Amundsen Sea coast (3) concurrent with substantial thinning of adjoining ice shelves (4, 5) and along the Antarctic Peninsula after ice-shelf disintegration events (6). Ice shelves restrain (“buttress”) the flow of the grounded ice through drag forces at the icerock boundary, including lateral stresses at sidewalls and basal stresses where the ice shelf rests on topographic highs (7, 8). Reductions in iceshelf thickness reduce these stresses, leading to a speed-up of ice discharge. If the boundary between the floating ice shelf and the grounded ice (the grounding line) is situated on a retro-

by three anonymous reviewers, S. Rahmstorf, and P. Eickemeier have considerably improved the manuscript. Data presented in this manuscript will be archived for at least 10 years by the Potsdam Institute for Climate Impact Research. The work was supported by the German Research Foundation (grant no. CO994/2-1) and the German Federal Ministry of Education and Research (grant no. 01LN1304A). D.C. designed the research; D.C., J.L., and J.B. performed the analysis; and D.C., J.L., and J.B. wrote the manuscript.

grade bed (sloping downwards inland), this process leads to faster rates of ice flow, with potential for a self-sustaining retreat (7, 9, 10). Changes in ice-shelf thickness and extent have primarily been attributed to varying atmospheric and oceanic conditions (11, 12). Observing iceshelf thickness variability can help identify the principal processes influencing how changing large-scale climate affects global sea level through the effects of buttressing on the Antarctic Ice Sheet. The only practical way to map and monitor ice-shelf thickness for this vast and remote ice sheet at the known space and time scales of ice-shelf variability is with satellite altimetry. Previous studies have reported trends based on simple line fits to time series of ice-shelf thickness (or height) averaged over entire ice shelves or broad regions (4, 13) or for short (~5-year) time intervals (5, 14, 15). Here, we present a record of ice-shelf thickness that is highly resolved in time (~3 months) and space (~30 km), using the longest available record from three consecutive overlapping satellite radar altimeter missions (ERS-1, 1992–1996; ERS-2, 1995–2003; and Envisat, 2002–2012) spanning 18 years from 1994 to 2012.

Our technique for ice-shelf thickness change detection is based on crossover analysis of satellite radar altimeter data, in which time-separated height estimates are differenced at orbit intersections (13, 16, 17). To cross-calibrate measurements from the different satellite altimeters, we used the roughly 1-year overlap between consecutive missions. The signal-to-noise ratio of altimeterderived height differences for floating ice in hydrostatic equilibrium is roughly an order of magnitude smaller than over grounded ice, requiring additional data averaging to obtain comparable statistical significance. We aggregated observations in time (3-month bins) and space (~30-km cells). Because the spatial distribution of crossovers changes with time (due, for example, to nonexact repeat tracks and nadir mispointing), we constructed several records at each cell location and stacked them in order to produce a mean time series with reduced statistical error (18). We converted our height-change time series and rates to thickness changes by assuming that observed losses occurred predominantly at the density of solid ice (basal melting) (4, 5, 17). This is further justified by the relative insensitivity of radar measurements to fluctuations in surface mass balance (18). For volume changes, we tracked the minimum (fixed) area of each ice shelf (18). We assessed uncertainties for all estimates using the bootstrap approach (resampling with replacement of the residuals of the fit) (19), which allows estimation of formal confidence intervals. All our uncertainties are stated at the 95% confidence level [discussion of uncertainties are provided in (18) and the several corrections applied are stated in (20)]. We estimated 18-year trends in ice-shelf thickness by fitting low-order polynomials (degree n ≤ 3) to the data using a combination of lasso regularized-regression (21) and cross-validation for model-parameter selection (the shape of the fit is determined by the data). This combined approach allowed us to minimize the effect of shortterm variability on the 18-year trends. Relative to previous studies (4, 5, 13, 22), we have improved estimations by (i) using 18-year continuous records, (ii) implementing a time series averaging 17 APRIL 2015 • VOL 348 ISSUE 6232

327


www.sciencemag.org/cgi/content/full/science.1261768/DC1

Supplementary Materials for The weakening summer circulation in the Northern Hemisphere midlatitudes Dim Coumou, Jascha Lehmann, Johanna Beckmann *Corresponding author. E-mail: [email protected] Published 12 March 2015 on Science Express DOI: 10.1126/science.1261768 This PDF file includes: Text S1 to S6 Figs. S1 to S25 Table S1 References Data Deposition

81

82


S1: Materials and Methods In several figures in the manuscript and Supplementary Online Material (SOM), changes in dynamic variables are normalized and presented in terms of relative percentage-changes compared to the climatological mean: (S1) Where θ can be any variable with θclim its climatological mean. Compared to presenting absolute values, this has the advantage that relative changes between variables can be understood and compared to relative changes during the seasonal cycle and in climate model projections. This way, in Fig. S1-S5, S7-S10, 2 and 3a, the observed variables are normalized to their summer climatology (model data shown in Fig. 2 is the difference between future (2081-2100) and late 20th century (1981-2000) summer climatology). In Fig. 3b which shows daily values of the full year, variables are normalized to their annual climatology such that the seasonal cycle centers around zero. In Fig. 1, S6, and S12-17, variables are provided in absolute units. We calculate thermal wind UT by solving the geostrophic equation

(S2) assuming that sea-level pressure (psl) differences can be neglected. Thus only the second term on the right hand side is used for which g is acceleration due to gravity, a is the radius of the Earth, T*=273.16K and f = 2Ωsin(ϕ) with ϕ representing latitude (in radians) and z altitude (in meters).


83

S1.1: Spectral Analysis We apply spectral analysis as in Coumou et al. (1) to daily wind field data from ERA-Interim reanalysis (2) for the months of June, July and August over the period 1979-2013. For each day, we determine the amplitude and phase for each wave number by taking a Fast Fourier Transform (FFT) of the meridional wind at given pressure level averaged from 35oN to 70oN. The wave quantities are thus determined for the meridional (North-South) wind component v and wave amplitude Av thus reflects meridional wind speeds and has units of m/s. We calculate the phase speed (eastward propagation) of each wave by taking a fourth-order accurate numerical approximation of the transient derivative of its phase. We tried different numerical methods to calculate the derivative and found the estimate of the phase speed to be robust. We also found the results to be insensitive to the exact choice of latitudinal boundaries. The wave period can directly be determined from the phase speed and the wave number. Probability-density distributions (Fig. 3b of main manuscript and Fig. S21) are obtained by applying a nonparametric kernel density estimation to the daily spectral data. S2: Using reanalyses data to estimate wind field trends Reanalyses data are the most accurate product available for this type of study but caution is needed in analyzing and interpreting them. In particular, the analysis of climate trends in reanalysis data has to be done with care since artificial shifts in magnitude of observed variables may arise caused by changes in the observing system in 1979 when satellite data was introduced (3). We therefore focus our analysis entirely on the post-1979 time period. In our study we use two of the most widely used and tested reanalyses, i.e ERA-Interim and NCEP-NCAR. Both of which have been shown to adequately reproduce observations and which have been successfully used in a large number of meteorological and climatic studies (4–7). In particular, Decker et al. (4) find that ERA-Interim compared to other reanalysis products is most accurate in projecting wind speeds. Nevertheless, there can be biases or inhomogeneities in absolute values of near-surface wind (see references in Della-Marta et al. (8)). However, these constraints are primarily of systematic nature and thus relative changes in wind speed will be much more reliable. Also, in our study we focus on wind fields in the free troposphere only. More importantly, we exclusively study dynamics in the Northern Hemisphere, which is very well sampled due to the many meteorological observing stations here. In conclusion, despite the known caveats in reanalysis products, the choice for ERA-Interim and NCEP-NCAR (which are well tested and have shown to be most accurate in reproducing actual observations), the focus on the post-1979 period in the Northern Hemisphere and the analysis of relative trends will give reliable and robust results. S3: Signal-to-noise analysis Trends in zonal-mean U are negative for all seasons over the 1979-2013 period (see Fig. S15a and Fig. S12-14) with the strongest trends in autumn (about -12 10-3 ms-1 yr-1) and summer (about -7.5 10-3 ms-1 yr-1). Trends in winter and spring are also downwards but smaller. The standard

84


deviation (σ) of year-to-year variability is smallest in summer when it is about half the autumn value (see Fig. S15b). Therefore the total shift in U over the 35 year period in units of standard deviation is comparable between summer and autumn. The mean has shifted by roughly -1.5σ in both seasons over 1979-2013 (Fig. S15c): In autumn due to a strong downward trend and a moderate standard deviation and in summer due to a more-moderate trend in combination with a small standard deviation. As a result, the long-term trends in both seasons are detectable at the 5%-level (manuscript Fig. 1b and Fig. S14b). For the other two seasons no significant changes are detected because the trends are small (Fig. S15a) and the standard deviations relatively large (Fig. S15b), resulting in very small signal-to-noise ratios (Fig. S15c). Francis and Vavrus (9) argued that the decrease in zonal-mean zonal wind is most pronounced in autumn and winter, but they analyzed the North American sector only (140W-0W) whereas we analyze the full hemisphere. Still it is somewhat surprising why the absolute trend in winter zonal-mean flow is small, although, as shown by Cohen et al. (10), the strongest warming in the Arctic (i.e. north of 70oN) occurred in that season (see figure 1 in Cohen et al. (10)). This figure, however, also shows that wintertime warming-rates have been essentially flat in the mid-latitudes and thus the equator-topole thermal gradient between 35oN-70oN has also seen little change. In contrast, summer and autumn have seen a more spatially-uniform warming with warming-rates smoothly increasing with latitude. For this reason, the mid-latitude thermal gradient (between 35oN and 70oN) has seen the most-pronounced changes in summer followed by autumn, as shown in Fig. S17. Therefore, also the strongest changes in mid-latitude thermal wind UT are seen in these seasons. This thus explains why the strongest trends in U in the mid-latitudes are seen in summer and autumn and not in winter. Similarly to U, all seasons show downward trends in EKE (see Fig. S16a). For winter and spring, the two reanalysis products show substantial disagreements on the magnitude of the downward trend, but this is not the case for autumn and summer. In absolute terms, the downward trend in summer (about -35 10-3 m2 s-2/yr) is about double that of autumn (about -17 10-3 m2s-2/yr). On top of that, the standard deviation due to year-to-year variability is about half as large in summer as compared to autumn (see Fig. S16b). Therefore, the total shift in EKE over the 35 year period in units of standard deviation is much larger in summer than in any of the other seasons (see Fig. S16c). In summer, EKE has shifted by more than two standard deviations towards weaker EKE in both reanalyses which implies a significant downward trend. In the other seasons this shift has been much smaller. S4: Mid-latitude circulation Generally, the large-scale mid-latitude atmospheric dynamics are characterized by 1) fast traveling free Rossby waves (the so called synoptic transients) with zonal wave numbers typically larger than 6, which propagate predominantly eastwards with relatively large phase speeds (~6-12m/s), and 2) quasi-stationary Rossby waves with typically smaller zonal wave


85

numbers as a response to quasi-stationary diabatic and orographic forcing (11–13). This characterization follows for example from the work of Boer (14) and Boer and Shepherd (15). The theoretical study by Boer (14) shows that for homogeneous and isotropic turbulence on a sphere the kinetic energy is roughly equally partitioned between the east-west (u) and northsouth (v) components. Based on data-analysis, Boer and Shepherd (15) show that this equipartitioning is indeed observed in the atmosphere for the larger wavenumbers (>6). Thus this high wave number regime is dominated by transient eddies and exhibits at least approximately several of the necessary conditions for homogeneous and isotropic flow. Thus, EKE is approximately equally partitioned between the u and v components. The low wave number regime of the flow is quite different from the high wave number regime. The former is dominated by the stationary component of the flow and exhibits marked anisotropy as well as a lack of equipartitioning of energy. In our analysis we focus on the fast-moving free Rossby waves. By applying a 2.5-6 day bandpass filter to calculate EKE, we explicitly extract only the kinetic energy associated with fast traveling free Rossby waves. Thermally and orographically forced waves have much longer wave periods (or lower frequencies), with timescales from weeks to months, and are thus excluded from EKE. In our spectral analysis, no explicit separation between fast traveling and quasi-stationary waves is done but since we use daily data, the mean wave amplitude (Av) is dominated by high-frequencies, i.e. fast-traveling waves. (Note that Av reflects the strength of the North-South wind component - see section S1.1). The reason for the dominance of highfrequencies is that in the mid-latitudes the kinetic energy associated with transient eddies is estimated to be nearly an order of magnitude larger than that of quasi-stationary waves (16). Only when averaging over longer time-spans, like 15 days to monthly means, as done in other studies (13, 17), one effectively removes the transient waves and the quasi-stationary components remain. Av is thus dominated by fast-traveling waves and the equipartitioning of kinetic energy in east-west and north-south components for such waves underlies the positive linear regression between ∆U and mean ∆Av as seen in figure 3b of the manuscript. The observed reduction in both EKE and Av in daily wind-field data as a consequence of weaker zonal-mean zonal wind is also in agreement with the comprehensive literature on storm tracks. In general, transient eddies (associated with storm tracks) are both forced by the zonal flow via vertical shear (and thus baroclinicity) but can also accelerate it via the eddy-driven jet (10, 18, 19). Further, Pinto et al. (20) show that changes in the Northern Hemisphere storm tracks are associated with changes in the zonal wind at 250 hPa. In particular, they found that an intensification and downstream extension of the storm track (and thus an increase in EKE and Av) is accompanied by an intensification of the eddy-driven jet. Further, scaling analysis based on theoretical arguments show that a reduction in the meridional temperature gradient leads to a reduction in synoptic temperature variability (21, 22), which implies a reduction EKE and Av. A reduction in meridional temperature gradient is also related to a reduction in zonal wind through the thermal wind equation. These scaling analyses are confirmed by numerical experiments using

86


an idealized GCM and the CMIP5 archive (21). Using both CMIP5 and CMIP3, Mizuta et al. (23, 24) find high correlations between the mean growth rate of cyclones and the upper-level zonal wind over regions of increasing growth rates in the Northern Hemisphere. They propose that the projected changes in mean growth rate can largely be explained by changes in the zonal wind. Using NCEP-NCAR reanalysis data, Graham et al. (25) found that an increase in winter storm activity in the North Pacific resulted from increasing zonal winds. Similar results were found by Chang et al. (26) by applying an empirical orthogonal function analysis on high-passfiltered 300 hPa meridional wind variance in NCEP-NCAR reanalysis data. Their results indicate that variations in Northern Hemisphere winter storm tracks (and thus EKE and Av) are highly correlated with the zonal wind at 300 hPa. In conclusion, stronger zonal mean winds give rise to stronger vertical shear and hence baroclinicity (19). Therefore during positive phases of the (N)AO, characterized by a stronger eastward flow, more or stronger baroclinic waves are likely to develop (27). These transient waves can again drive the zonal flow via the eddy-driven jets (10, 18). This, and keeping in mind that Av is dominated by transient waves, explains the positive linear relationship between U and mean Av seen in Fig. 3b. S5: EKE and monthly heat extremes Figure 4 of the manuscript shows the regression analysis between EKE and near-surface temperature (T2M) of summer months (June, July, August) in the ERA-Interim dataset (19792013). Before regressing, both datasets were linearly detrended to remove the effects of longterm trends in the correlation. However, we found that regression of the original, non-detrended data, gave essentially the same result (Fig. S23). Over all mid-latitude continental regions the regression analysis reveals a negative correlation between EKE and T2M. Thus, warm summer months are associated with low EKE. Over most of Eurasia as well as the U.S. this negative correlation is significant at the 5% level (see stippling in Fig. 4 of manuscript and Fig. S23). As a typical example, figure 4b of the manuscript shows the regression between EKE and T2M (p-value < 0.01) of summer months over Moscow (red dot in Fig. 4a). This figure shows that the 10% coldest summer months (left of the vertical dashed blue line) have a substantially higher EKE of 14m2/s2 (horizontal dashed blue line) and the 10% warmest months (right of the vertical dashed red line) a substantially lower EKE of 5.7m2/s2 (horizontal dashed red line). The two months affected by the Russian heat wave in 2010 (July and August, plotted respectively with red and purple circles) also saw extremely low EKE. Notably July 2010, the hottest months in the data with a temperature anomaly of 6.3oC, had an extremely low EKE of 2.3 m2/s2 (or only ~25% of the summer climatology). This can be understood by the fact that the 2010 Moscow heat wave was associated with a strong and long-lived blocking event (28), i.e. extremely persistent weather conditions. This high-pressure system diverted any synoptic disturbances coming from the Atlantic region (which could bring cool maritime air) towards the North. Figure 4b shows that EKE during the 10% hottest months in Moscow was about 40% below the mean summer climatology of 9.2 m2/s2. Fig. S24 shows this percentage reduction in EKE during the 10%


87

hottest summer months for the full hemisphere. Only those regions are plotted for which a statistically significant regression exists with p-values smaller than 0.1 (which roughly coincides with those regions with pronounced EKE, i.e. the storm track regions, Fig. S22). The figure shows that over most of Eurasia and over large areas in North-America – i.e. those land regions influenced by the storm tracks – EKE was down by typically 50% during the 10% hottest months. S6: Heat extremes: Quasi-stationary versus fast-moving waves Previous studies have highlighted the importance of quasi-stationary waves in causing summer heat and rainfall extremes (1, 13, 17, 29, 30). These studies analyzed typically monthly wind field data and showed that high-amplitude waves (i.e. strong north-south flow) are associated with extreme weather at the surface and especially with heat waves (1, 13, 17, 29, 30). Thus, high-amplitude quasi-stationary waves can cause persistent weather and therefore often extreme weather conditions. This also intuitively makes sense since blocking high-pressure systems represent quasi-stationary Rossby waves and these are often associated with extreme heat waves as for example over Russia in summer 2010 (28, 31). In the regression analysis presented in figure 4 of the manuscript (and SOM section “S5: EKE and heat extremes”) we show that monthly heat extremes in the mid-latitudes (and especially in those regions influenced by the storm tracks) are associated with low EKE. Low EKE implies low amplitudes of fast moving synoptic waves. Thus, if for given month and location, EKE is low then weather variability within that month was low, or in other words, weather conditions were persistent. This is thus consistent with blocking high-pressure conditions associated with quasi-stationary waves. From a synoptic point of view, well-developed blocking high-pressure systems can divert fast moving waves from their normal path. When the transient waves are diverted, this means that cyclones which could bring moderate weather conditions (i.e. cool and humid air from the oceans) do not reach the hot continental regions anymore. This creates favorable conditions for the build-up of heat and drought over the continents, possibly amplifying local feedback mechanisms (such as soil-moisture feedbacks) (32, 33), which can further enhance the anticyclonic flow over the continents by inhibiting cloud formation (34). Thus an increased frequency of high-amplitude quasi-stationary waves or a decrease in EKE are both expected to result in more persistent weather conditions on timescales beyond the synoptic timescale.

88


S7: Figures and Tables

Fig. S1. As Fig. 1 of manuscript (ERA-Interim, 500mb) but in units of relative percentagechanges compared to the climatological mean (SOM section S1).

Fig. S2. As Fig. S1 but at 250mb.




Fig. S5. Trend analysis of zonal-mean eddy kinetic energy EKE in summer, giving trends in percentage-per-year over the 1979-2013 period for different pressure levels for all Northern Hemisphere latitudes in the ERA-Interim reanalysis. For all mid- to high-latitudes (which have non-zero EKE, see Fig. S22) the trend is downward, with the strongest changes in the lower (700mb) to mid-troposphere (500mb).

Fig. S6. As Fig. 1 of manuscript but for NCEP-NCAR (500mb) in absolute units.

Fig. S7. As Fig. S1 but at 250mb and for the NCEP-NCAR reanalysis.

89

90





Fig. S11. Summer climatology of the thermal wind UT as calculated by equation 2 for ERAInterim (left) and NCEP-NCAR (right) for pressure levels 850mb (red), 700mb (green), 500mb (blue) and 250mb (black).


Fig. S12. As Fig. 1 of manuscript (ERA-Interim, 500mb) but for winter (December-JanuaryFebruary).

Fig. S13. As Fig. 1 of manuscript (ERA-Interim, 500mb) but for spring (March-April-May).

Fig. S14. As Fig. 1 of manuscript (ERA-Interim, 500mb) but for autumn (September-OctoberNovember).

91

92


Fig. S15. Signal-to-noise analysis of zonal-mean zonal wind U for each season. a) Linear trend over 1979-2013 in ms-1/yr (note that all trends are negative), b) standard deviation due to year-toyear variability (ms-1) and c) the total shift in the mean over the 35 year period in units of standard deviation. Red dots indicate ERA-Interim and black dots NCEP-NCAR.

Fig. S16. Signal-to-noise analysis of zonal-mean EKE for each season. a) Linear trend over 1979-2013 (m2s-2/yr), b) standard deviation due to year-to-year variability (m2s-2) and c) the total shift in the mean over the 35 year period in units of standard deviation. Red dots indicate ERAInterim and black dots indicate NCEP-NCAR.

Fig. S17. Trend in (left) pole-to-equator thermal gradient and (right) thermal wind UT, at 500mb averaged over 35oN-70oN for ERA-Interim (red) and NCEP-NCAR (black) for each season. In this mid-latitude range, summer and autumn have seen the most pronounced changes (see SOM section S3). For summer-trends all p-values are smaller than 0.01, except for the NCEP-NCAR thermal gradient (p-value = 0.07).


Fig. S18. As Fig. 2 of manuscript but for individual pressure levels: a) 850mb, b) 700mb, c) 500mb and d) 250mb.

93

94


A)

B)

C)

D)

Fig. S19. As Fig. 3a of manuscript but for a) ERA-Interim reanalysis at 500mb, b) ERA-Interim at 300mb, c) NCEP-NCAR at 500mb and d) NCEP-NCAR at 300mb.


A)

B)

C)

D)

Fig. S20. As Fig. 3a of manuscript but for winter (DJF) for a) ERA-Interim reanalysis at 500mb, b) ERA-Interim at 300mb, c) NCEP-NCAR at 500mb and d) NCEP-NCAR at 300mb.

95

96


Fig. S21. Contribution of individual wave numbers to 2.5-6 day weather variability. Probabilitydensity distribution of wave period in summer days in 1979-2013 for wave numbers 8-15. It shows that wave periods in the EKE relevant range of 2.5-6 days (shaded region) are dominated by wave 10. During roughly half of all summer days, wave 10 had a wave period within this range. Wave numbers 8, 9, 11, 12 and 13 also contribute to 2.5-6 day wave periods, but substantially less so.

Fig. S22. Summer (JJA) climatology of EKE (m2/s2) in the ERA-Interim data.

Fig. S23. Regression analysis between EKE and T2M in summer months (similar to Fig.4a of manuscript but without detrending EKE and T2M before regressing).


97

Fig. S24. Composite plot of EKE during the 10% hottest summer months (June, July, August). EKE is given as percentage change compared to local climatology (Fig. S22) and only those regions are shown for which the p-value of the regression analysis between monthly EKE and near-surface temperature (Fig. 4a of manuscript) is smaller than 0.1. Over most of Eurasia and large areas in Northern America, EKE was reduced by about 50% during these extremely warm months.

Fig. S25. Equivalent to Fig. 1 (ERA-Interim at 500mb) but with linear trend lines added for the 1979-2000 period (blue). Slope and p-values in the panels refer to the linear trend of the 19792000 period. Trends in U and UT were thus more than a factor 2 smaller over 1979-2000 compared to the full period 1979-2013. For EKE this factor is about 1.5. This thus shows that the rate of weakening in mid-latitude circulation seems to have amplified since 2000.

Table S1 250mb 500mb 700mb 850mb

U - 3.9% - 4.4% - 4.9% - 5.6%

ERA-Interim UT -3.2% -4.0% -4.3% -4.2%

EKE - 7.6% - 14% - 12% - 9.8%

U - 2.5% - 2.6% - 3.2% - 5.0%

NCEP-NCAR UT -2.2% -2.1% -2.0% -3.2%

EKE - 0.32% - 14% - 13% - 9.1%

Table S1. Linear trends in U, UT and EKE in summer for the ERA-Interim and NCEP-NCAR reanalysis for different pressure levels. Units are in percentage-change over the full 35 year period 1979-2013. Bold red and bold black numbers indicate respectively 5% and 10% statistical significance.

98


S8: References 1. D. Coumou, V. Petoukhov, S. Rahmstorf, S. Petri, H. J. Schellnhuber, Proc. Natl. Acad. Sci. U. S. A. 111, 12331–12336 (2014). 2. D. P. Dee et al., Q. J. R. Meteorol. Soc. 137, 553–597 (2011). 3. L. Bengtsson, J. Geophys. Res. 109, D11111 (2004). 4. M. Decker et al., J. Clim. 25, 1916–1944 (2012). 5. R. C. Cornes, P. D. Jones, J. Geophys. Res. Atmos. 118, 10,262–10,276 (2013). 6. M. Menendez et al., Clim. Dyn. 42, 1857–1872 (2013). 7. A. J. Simmons et al., Q. J. R. Meteorol. Soc. 140, 329–353 (2014). 8. P. M. Della-marta, H. Mathis, C. Frei, M. A. Liniger, J. Kleinn, 459, 437–459 (2009). 9. J. A. Francis, S. J. Vavrus, Geophys. Res. Lett. 39, L06801 (2012). 10. J. Cohen et al., Nat. Geosci. 7, 627–637 (2014). 11. J. Pedlosky, Geophysical Fluid Dynamics (Springer, New York, USA, 1979). 12. K. Fraedrich, H. Böttger, J. Atmos. Sci. 35, 745–750 (1978). 13. V. Petoukhov, S. Rahmstorf, S. Petri, H. J. Schellnhuber, Proc. Natl. Acad. Sci. U. S. A. 110, 5336–5341 (2013). 14. G. J. Boer, J. Atmos. Sci. 40, 154–163 (1983). 15. G. J. Boer, T. G. Shepherd, J. Atmos. Sci. 40, 164–184 (1983). 16. J. P. Peixoto, A. H. Oort, Physics of Climate (American Institute of Physics, New York, 1992), p. 520. 17. J. A. Screen, I. Simmonds, Nat. Clim. Chang. 4, 704–709 (2014). 18. T. Woollings, M. Blackburn, J. Clim. 25, 886–902 (2012). 19. J. Lehmann, D. Coumou, K. Frieler, A. V Eliseev, A. Levermann, Environ. Res. Lett. 9, 084002 (2014). 20. J. G. Pinto et al., Clim. Dyn. 29, 195–210 (2007).


99

21. T. Schneider, T. Bischoff, H. Plotka, J. Clim. (2014). 22. D. Coumou, V. Petoukhov, A. V. Eliseev, Nonlinear Process. Geophys. 18, 807–827 (2011). 23. R. Mizuta, M. Matsueda, H. Endo, S. Yukimoto, J. Clim. 24, 6456–6470 (2011). 24. R. Mizuta, Geophys. Res. Lett. 39, L19707 (2012). 25. N. E. Graham, H. F. Diaz, BAMS 82, 1869–1893 (2001). 26. E. K. M. Chang, Y. Fu, J. Clim. 15, 642–658 (2002). 27. J. Nie, P. Wang, W. Yang, B. Tan, 159, 153–159 (2008). 28. R. Dole et al., Geophys. Res. Lett. 38 (2011). 29. H. Teng, G. Branstator, H. Wang, G. a. Meehl, W. M. Washington, Nat. Geosci. 6, 1–6 (2013). 30. S. Schubert, H. Wang, M. Suarez, J. Clim. 24, 4773–4792 (2011). 31. S. Rahmstorf, D. Coumou, Proc. Natl. Acad. Sci. U. S. A. 108, 17905–9 (2011). 32. C. Schär et al., Nature 427, 332–336 (2004). 33. B. Mueller, S. I. Seneviratne, Proc. Natl. Acad. Sci. U. S. A. 109, 12398–403 (2012). 34. L. Alexander, Nat. Publ. Gr. 4, 12–13 (2010).

S9: Data Deposition Raw input data (daily wind fields) were downloaded from publicly accessible databases maintained by World Climate Research Programme's Working Group on Coupled Modelling (CMIP5), European Center for Medium Range Weather Forecast (ECMWF, ERA-Interim reanalysis) and National Oceanic and Atmospheric Administration (NOAA, NCEP-NCAR reanalysis). Processed data and details on which raw data was downloaded are available at: www.pik-potsdam.de/research/earth-system-analysis/projects/project-pagesrd1/sacrex/publications/database-for-the-weakening-of-summer-circulation-in-the-northernhemisphere-mid-latitudes-science-2015

2.4. The influence of mid-latitude storm tracks on hot, cold, dry, and wet extremes Jascha Lehmann and Dim Coumou. In this paper the influence of storm track activity on mid-latitude temperature and rainfall extremes is analyzed. Results show that storms have a moderating affect on continental temperatures with low storm activity favoring more persistent weather conditions which can lead to prolonged heat waves in summer and cold spells in winter. Further, it is demonstrated that rainfall extremes are associated with strong storm track activity and dry spells with a lack thereof. The study was designed by Jascha Lehmann who performed all analyses, analyzed the data and wrote the text. Dim Coumou helped in interpreting the results and improving the manuscript. Published in Scientific Reports (2015), 5, 17491, doi:10.1038/srep17491.

2.4. The influence of mid-latitude storm tracks on hot, cold, dry, and wet extremes

103

www.nature.com/scientificreports

OPEN

The influence of mid-latitude storm tracks on hot, cold, dry and wet extremes

received: 16 June 2015 accepted: 13 October 2015 Published: 11 December 2015

Jascha Lehmann1,2 & Dim Coumou1 Changes in mid-latitude circulation can strongly affect the number and intensity of extreme weather events. In particular, high-amplitude quasi-stationary planetary waves have been linked to prolonged weather extremes at the surface. In contrast, analyses of fast-traveling synoptic-scale waves and their direct influence on heat and cold extremes are scarce though changes in such waves have been detected and are projected for the 21st century. Here we apply regression analyses of synoptic activity with surface temperature and precipitation in monthly gridded observational data. We show that over large parts of mid-latitude continental regions, summer heat extremes are associated with low storm track activity. In winter, the occurrence of cold spells is related to low storm track activity over parts of eastern North America, Europe, and central- to eastern Asia. Storm tracks thus have a moderating effect on continental temperatures. Pronounced storm track activity favors monthly rainfall extremes throughout the year, whereas dry spells are associated with a lack thereof. Trend analyses reveal significant regional changes in recent decades favoring the occurrence of cold spells in the eastern US, droughts in California and heat extremes over Eurasia.

In 2014, California suffered a severe drought1 whereas the northeastern US experienced record-breaking Arctic cold in winter2. In the same year, droughts in the northern province of Liaoning – China’s northern bread basket – threatened crop yields while the UK was affected by severe floodings3. This seeming accumulation of weather extremes demonstrates the importance of understanding the physical mechanisms driving climate variability in the Northern Hemisphere mid-latitudes at interannual to multi-decadal time scales4,5. Anomalous large-scale atmospheric circulations are often directly linked to the occurrence of regional temperature and precipitation extremes6–11. It is thus critically important to understand how different types of waves in the mid-latitudes influence surface weather and extremes2,8,12–14. There is substantial evidence that amplified quasi-stationary planetary waves favor particular regional weather extremes14,15 including hot, cold, dry and wet extremes. Such waves are associated with high-pressure blocking systems which cause persistent and therefore often extreme weather like for example during the Russian heat wave in summer 201016–18. Storm track variability also has a direct influence on regional weather conditions, and their strength and position might change under future climate change19–21. Recent studies have shown that the Californian drought was linked to a lack of storm activity1 whereas the UK flooding was due to an unusual clustering and persistence of storms3. Studies on synoptic-scale wave activity and its association with blocking have mainly focused on wintertime circulation and the influence of shifting storm tracks on blocking frequencies22,23. Notably, Dong et al.24 report less atmospheric blocking (defined by a blocking index) in the UK and northwestern Europe when more summer storms travel across these regions. High- and low frequency waves are linked, with large-scale quasi-stationary waves influencing the location and activity of storm tracks which in turn help to maintain the large-scale flow via eddy transports22,25,26. Thus, findings derived from storm tracks will also relate to the larger scale flow including quasi-stationary waves. 1

Potsdam Institute for Climate Impact Research, 14412 Potsdam, Germany. 2University of Potsdam, Am Neuen Palais 10, 14469 Potsdam, Germany. Correspondence and requests for materials should be addressed to J.L. (email: [email protected])

Scientific Reports | 5:17491 | DOI: 10.1038/srep17491

1

104


www.nature.com/scientificreports/ In the mid-latitudes, synoptic storms bring moist air from the ocean to the land and in regions like Europe and North America they account for over 70% of total precipitation27. Extreme rainfall is thus often associated with strong storm activity28,29. Storm activity is also likely to affect temperature extremes30. In summer – when near-surface temperatures are lower over oceans than over land – storms transport cool air from the oceans to continental regions. A decrease in summer storm activity could thus lead to the build-up of hot and dry conditions over the continents31. During winter this effect might reverse as sea surface temperatures are higher than land temperatures. Storm activity is thus likely to have a moderating effect on continental temperatures and hence changes in storm tracks could affect not only precipitation and wind extremes but also heat waves and cold spells. The position and strength of storm track activity is strongly coupled to large-scale teleconnection patterns like the North Atlantic Oscillation (NAO)9,10,32. The NAO index is often defined as the mean winter (December to March) standardized air pressure difference between the Azores High and the Icelandic Low with positive index values indicating that air pressure is anomalously high over the Azores and anomalously low over Iceland and vice versa for negative index values. During positive NAO phases the North Atlantic storm track shifts northwards associated with mild and wet conditions over northern Europe and colder and drier conditions over southern Europe and the Mediterranean7,8,33. During negative NAO phases, the westerly winds weaken associated with more frequent blocking conditions over Greenland32,34. In these situations cold air from the Arctic can be dragged to northern Europe leading to anomalously cold winter temperatures over the UK and Scandinavia35. There exists a counterpart for the NAO in summer called the summer North Atlantic Oscillation (SNAO)24,33,36. In its positive phase, this pattern is characterized by low air pressures over Greenland and high pressures over northern Europe33. The SNAO strongly influences summertime variability of temperature, precipitation and cloudiness over Europe and parts of North America and the Sahel zone through changes in the position of the North Atlantic storm track36. During a negative phase of SNAO (with relatively high pressure conditions over Greenland and low pressure anomalies over the British Isles and Scandinavia) the North Atlantic storm track shifts southwards and extends downstream into central Europe associated with more storms and thus more precipitation over the UK and northwest Europe and less rain over southern Europe24. In contrast, high-index SNAO summers are related to warm, dry and cloud-free conditions over northwest Europe and much of southern Scandinavia and anomalously cool, wet and cloudier conditions over southern Europe and the Mediterranean36,33. The (S)NAO index is usually determined through first empirical orthogonal function analysis of sea level pressure in the North Atlantic sector and thus represents a single number that captures the dominant variability pattern on a near-hemispheric scale. Composite analysis of positive and negative (S) NAO phases thus illustrates the regional changes associated with the dominant variability pattern only and might thus provide little information on local extremes. Most studies on NAO simulations in climate change scenarios focus on winter only9. Based on these studies, no clear statements about NAO changes under CO2 doubling can be made since some models project an upward trend and others a downward trend. In contrast, robust changes in summer storm track activity have been observed31 and are projected for the 21st century under continued global warming21,31. In particular, climate models project a substantial weakening of summer storm track activity over essentially all of the mid-latitudes21 and a poleward shift and downstream extension into Europe during winter20,21,37,38 which will likely alter the intensity and frequency of surface weather extremes in these regions. Here we take an alternative approach and study the link between storm track activity and surface weather extremes directly. We analyze extremes in land surface temperature and precipitation using monthly ECWMF ERA-Interim data39. We thus consider extremes associated with prolonged heat waves, cold spells and wet or dry periods. To analyze the effect of storm track activity on temperature extremes over continents, we separate between summer (May-September) and winter (November-February) season since their effect is expected to be opposite (see Methods). Storms are associated with moist air independent of season and hence the analysis of precipitation extremes is applied to the full year. Storm track activity is represented by monthly eddy kinetic energy (EKE), which is computed by bandpass filtering daily wind field data thereby extracting wind speed variability on 2.5–6 days associated with synoptic-scale (eddy) activity21,40. Since quasi-stationary waves have much lower frequencies they are excluded from the computation of EKE. Likewise, short-lived thunderstorms which typically form in summer have much higher wave frequencies and are thus not considered in this study. We apply quantile regressions41 between EKE and temperature or precipitation at each individual grid point. This method allows us to analyze how the tails of the EKE distribution (10th and 90th-percentile) as well as the median (50th-percentile) are linked to regional surface weather extremes, rather than just estimating the effect of changes in the mean. Linear regression analyses are applied between EKE and geopotential height anomaly fields at individual grid points to assess how synoptic eddy activity is related to blocking anticyclonic flow. All analyses presented in the main manuscript are shown for the 850 mb pressure level. Sensitivity analyses performed at the 500 mb pressure level lead to similar results (see Methods).

Results

Quantile regression analysis between EKE and temperature and precipitation. Significant

negative correlations between anomalies in EKE and land surface temperature in summer can be found over storm track relevant land regions, i.e. North America and Eurasia (Fig. 1a, for EKE climatology see


2


105

www.nature.com/scientificreports/

Figure 1. Slope of quantile regressions between anomalies of EKE and temperature in summer. Regression slopes are shown for the 90th-percentile (a), 50th-percentile (b), and 10th-percentile (c) with the middle panel being similar to Fig. 4 from Coumou et al.31. Stippling indicates significance at the 5%-level and grey contour lines denote EKE climatology in summer. Land regions higher than 1 km have been masked. All maps are created using the open source software R.

Figure 2. Slope of quantile regressions between anomalies of EKE and temperature in winter. Regression slopes are shown for the 90th-percentile (a), 50th-percentile (b), and 10th-percentile (c). Stippling indicates significance at the 5%-level and grey contour lines denote EKE climatology in winter. Land regions higher than 1 km have been masked. All maps are created using the open source software R.

contours in Fig. 1a and Figs S6 and S7 in Supplementary Information (SI)). In these regions, EKE in the 10% hottest summer months is reduced by 20–40% compared to summer climatology (Fig. S8a). Most regression slopes are significant and steepest for the 90th-percentile and smaller in magnitude for the 50th and 10th-percentile, respectively (Fig. 1, S10a). These different magnitudes imply that the hottest summer months are always associated with low EKE, whereas anomalously cool summer months are associated with larger EKE and a broader range in possible EKE values. Significant regression slopes are evident in all three quantile regressions emphasizing the strong inverse link between EKE and temperatures in summer. In winter, significant positive regression slopes in the 90th-percentile are found over large parts of North America and Eurasia where mean EKE in the 10% coldest months decreased by 20–40% compared to local winter climatology (Fig. 2a, Fig. S8b). Over North America, we see a dipole pattern with significant negative correlations west of the Rocky Mountains and significant positive correlations to the east. In Scandinavia and most regions in central and east Asia low EKE during winter is significantly Scientific Reports | 5:17491 | DOI: 10.1038/srep17491

3

106



Figure 3. Slope of quantile regressions between anomalies of EKE and precipitation in all calendar months. Regression slopes are shown for the 90th-percentile (a), 50th-percentile (b), and 10th-percentile (c). Stippling indicates significance at the 5%-level and grey contour lines denote annual EKE climatology. Land regions higher than 1 km have been masked. All maps are created using the open source software R.

Figure 4. Slope of linear regression between anomalies of EKE and GPH. Regression slopes are shown for summer (a) and winter (b). Stippling indicates significance at the 5%-level and grey contour lines denote EKE climatology in the corresponding season. Land regions higher than 1 km have been masked. All maps are created using the open source software R.

correlated with low temperatures. Locally, this relationship is opposite over the UK and parts of western Russia. Regression slopes are again largest for the high quantiles, whereas slopes for the 10th-percentile become almost flat (Fig. 2c, Fig. S10b). This implies that in regions with positive correlations the coldest winter months are always associated with low EKE but warm winter months can be associated with a broad range of EKE values. Monthly precipitation extremes are related to significantly higher EKE in essentially all mid-latitude continental regions (Fig. 3). In particular, we find significant positive correlations between EKE and precipitation with mean EKE significantly increasing by 20–50% in the 10% wettest months compared to annual climatology (Fig. S8c). Consistently, the 10% driest months were associated with an approximately 20–30% drop in EKE (Fig. S8d). Regression slopes are generally larger for the upper quantiles (Fig. S10c) but with significant correlations in all quantiles. We repeated the analysis with precipitation records derived from rain gauge and satellite data taken from the Global Precipitation Climatology Project (GPCP)42. We find quantitatively similar regression patterns which confirm the robust link between EKE and precipitation (Figs S14 and S15).

Linear regression analysis between EKE and geopotential height anomaly fields. EKE is significantly anti-correlated with monthly geopotential height (GPH) over essentially all storm track relevant land regions in both summer and winter season (Fig. 4). This implies that low EKE is associated with positive geopotential height anomalies and hence high pressure systems. In winter, an inverse Scientific Reports | 5:17491 | DOI: 10.1038/srep17491

4


107

www.nature.com/scientificreports/ link with low EKE being associated with negative GPH anomalies (i.e. low pressure systems) is shown over Alaska, eastern North America and the coastal regions of eastern Asia (Fig. 4b). However, also in winter, significant negative correlations between EKE and GPH are dominant over most continental land regions.

Discussion. Our findings show that monthly rainfall extremes are associated with strong storm track activity and dry extremes with a lack thereof. Storms bring moderate weather conditions to the continents and therefore also modulate temperature extremes. This implies that in summer wet spells are associated with cool months whereas dry periods come along with warm months. In winter, we find an inverse link with higher storm track activity being associated with warmer and wetter conditions and less storm track activity being associated with colder and drier months. In agreement with our results, Trenberth and Shea30 show that summers over the continents in both hemispheres tend to be either cold and wet or hot and dry, but not any other combination. In winter they find a positive link between temperature and precipitation at high latitudes, implying that here warm winters are generally wet, whereas cold winters are rather dry. In the Northern Hemisphere, these high latitude regions largely overlap with storm track affected regions in agreement with our findings. The (S)NAO largely influences the position of storm tracks over the North Atlantic affecting weather conditions particularly over Europe and the Mediterranean9,10,32. In winter, a positive NAO results in a poleward shift of the North Atlantic storm track associated with anomalously warm and wet conditions over northern Europe and colder and drier conditions over southern Europe and the Mediterranean7,8,33 consistent with negative regression slopes between temperature and EKE and positive regression slopes between precipitation and EKE seen over these regions. Composite plots of negative and positive SNAO phases reveal that warm and dry conditions in summer are located over regions characterized by blocking anticyclonic flow24,36,33 and thus low storm track activity. This is also seen in our regression maps which indicate that low summertime EKE is significantly linked to positive geopotential height anomalies and thus warm summer temperatures and low precipitation over Europe. The intensified precipitation over the Mediterranean region in positive SNAO phases cannot be explained by SNAO itself as its characteristic pressure anomalies are too far north to directly influence the inflow of maritime air into southern Europe33. Over this region, we find significant correlations between EKE and both temperature and precipitation which explain the observed precipitation pattern by changes in storm track activity. The moderating affect of storm track activity on near-surface temperature is more pronounced in summer than in winter. This could be related to soil moisture-temperature feedbacks which are important in summer43. Low summer EKE implies low rainfall and high temperatures which both have a drying effect on the soil. Once the soil has dried out, no more heat can be converted into latent heat by evaporation and temperatures can spike44,45. There is thus a direct and indirect way how EKE can influence summer temperatures. Moreover, we have not addressed the effect of wind direction on temperature anomalies which is likely more important in winter46 and thus could be a reason for weaker correlations found between EKE and winter temperatures. Further, we show that low EKE is significantly correlated with anomalously large geopotential heights over storm track relevant land regions. We thus argue that low EKE creates favorable conditions for atmospheric blocking and hence persistent weather over continental land at least in summer. In agreement, Dong et al.24 report an increased storm activity over northern Europe associated with less blocking in this region. The presented correlations between storm track activity and surface weather conditions in terms of temperature and precipitation do not allow for direct conclusions about causality. However, extratropical storms mostly form and develop over the oceans (with exception of the US Rockies)47 and therefore it is reasonable to assume that over land storm track activity is influencing surface conditions and not the other way around7,32. Nevertheless, causal interpretations for the cyclogenesis region to the lee of the Rockies have to be treated with caution. Regionally, significant trends in EKE are detected (Fig. 5) which thus might have contributed to observed weather extremes. Notably ~78% of the mid-latitude continental land regions (35° N–65° N) have experienced downward trends in summertime EKE (Fig. 5a). Our results suggest that weakening of summertime EKE over Eurasia and eastern North America created favorable conditions for observed heat extremes like the European heat wave in 2003 or 2010 in Russia and likewise the high temperature anomalies over central to eastern North America in 2012 setting new all-time record high temperatures in multiple states along the US east coast48. These trend analyses are in agreement with Horton et al.12 who report upward trends in the frequency and persistence of summertime anticyclonic circulation patterns since 1979 over eastern US, Europe and western Asia. They show that the observed trends in circulation contributed significantly to high temperature extremes in these regions. Consistently, these regions also show pronounced downward trends in summertime EKE (Fig. 5a or Fig. S16 for better comparison). Coumou et al.31 proposed a physical mechanism to explain the observed weakening in summer circulation over recent decades. The observed hemispheric-mean reduction in summer EKE is associated with a decline in the zonal flow, in a similar way as projected by CMIP5 climate models. They argue that the downward trend has likely been influenced by the reduction in the equator-to-pole temperature gradient in response to Arctic amplification. Under a high emission scenario, CMIP5 models project further weakening in the summertime EKE which can be explained by a decrease in the vertical wind shear21. Here we report that EKE mostly has Scientific Reports | 5:17491 | DOI: 10.1038/srep17491

5

108



Figure 5. Trends in EKE. Trends are calculated for the time period 1979–2014 in summer (a), winter (b), and annually (c). Grey contour lines denote EKE climatology in the corresponding season and land regions higher than 1 km have been masked. All maps are created using the open source software R.

a moderating effect on surface weather conditions, notably in summer. Regional weakening of storm activity in recent years thus likely favored the occurrence of summer heat extremes over storm track relevant land regions. In winter, our results suggest that regional downward trends in EKE likely favored the occurrence of observed cold extremes. Eastern North America has been affected by severe cold spells in recent years2,49,50 which is in agreement with a pronounced reduction in storm track activity found for this region. California also experienced a significant reduction in EKE in winter which likely contributed to the observed periods of extreme droughts. Note, that in California the rainy season falls into the months between October and April whereas the summer months are generally dry. In contrast, central North America has seen an increase in wintertime EKE. Upward trends in EKE are also found over the western North Pacific and the North Atlantic with pronounced regions of negative trends in between. These wave-like patterns over the North American sector, which are also seen in the mid-troposphere (Fig. S18), might reflect changes in the position of the jet. Wang et al.1 showed that in winter large-scale wave energy in the western North Pacific intensified the high-pressure system offshore California which lead to extreme dry conditions over California in 2013/14. Consistently, we find that downward trends in wintertime EKE in this region are associated with an intensification of high-pressure systems and hence persistent anticyclonic flow (Figs 4b and 5b). Cyclones traveling eastwards across the North Pacific would be deflected and curl around this high-pressure system consistent with the pronounced increase in EKE observed to the northwest and northeast of the ridge. Another explanation for the observed correlation between synoptic-scale activity and surface weather extremes could be a common (third) driver which influences both transient waves and surface conditions. Large-scale quasi stationary waves might act as such a common driver. They are closely connected to fast-moving transients22,25,26 and amplification of these waves have been shown to favor heat waves in western North America and central Asia, cold spells in eastern North America, wet spells in western Asia and droughts in central North America, Europe, and central Asia14. Consistently, we find strong correlations between EKE and temperature or rainfall extremes over similar regions. Especially, we see a similar sensitivity of eastern North America to cold spells and a general link between higher than normal summer temperatures and low EKE. Synoptic eddies bring weather variability on 2–6 day timescales. We find that changes in the synoptic eddy activity have a moderating affect on near-surface temperatures and are positively correlated with rainfall over mid-latitude continental regions. Low EKE is associated with anticyclonic flow and persistent weather conditions which can lead to extremes on monthly timescales. This is shown for cold spells but is especially pronounced for summertime heat extremes and droughts. We argue that substantial trends in EKE over recent years created favorable conditions for the occurrence of observed extreme weather events in the US and Eurasia and that projected robust changes in EKE under future global Scientific Reports | 5:17491 | DOI: 10.1038/srep17491

6


109

www.nature.com/scientificreports/ warming20,21,37,38 will alter the occurrence of both temperature and rainfall extremes over storm track relevant continental regions.

Method

Data. Monthly-mean near-surface temperature (2 m above surface), precipitation and geopotential

height for the time period 1979–2014 were taken from the ECMWF ERA-Interim data sets39. In addition, monthly-mean precipitation derived from a combination of rain gauges and satellites were taken from the GPCP v2.2 data set42. Monthly-mean EKE was computed from bandpass filtered21,40,51 daily zonal (u’) and meridional wind speeds (v’) at 850 mb with EKE = 0.5 × (u’2 + v’2). The v’ component is very well linked to storm tracks and the u’ component less so. However, applying the analysis to v’2 instead of EKE leads to similar results (Figs S3 to S5). The analysis was also repeated with EKE computed at mid-troposphere (500 mb) which resulted in similar observed correlations with temperature but weaker correlations with precipitation (SI text S2), indicating that precipitation is sensitive to surface-near circulation changes. Similarly, correlations between EKE and geopotential height anomaly fields are stronger at the lower troposphere than at 500 mb (Fig. S17). The daily wind field data was taken from the ECMWF ERA-Interim data set, for the same time period and with the same grid resolution of 1.5° × 1.5°. For each calendar month and grid point the local climatological mean was subtracted from the grid-box value to compute time series of anomalies for EKE, temperature, precipitation, and geopotential height. Subsequently, all time series were linearly detrended over the full time period, except for the precipitation data set because of its non-Gaussian distribution. However, analysis of linearly detrended precipitation time series gave essentially the same results.

Seasonality. Mid-latitude storms bring maritime air from the ocean to continental regions. Depending

on the season this air can have a cooling (in summer) or warming (in winter) effect. For the analysis of temperature extremes we thus define two seasons; a summer season (May-June-July-August-September) – where ocean surface temperatures are lower than land surface temperatures – and a winter season (November-December-January-February) – where ocean surface temperatures are higher than on land. For the analysis of precipitation extremes we tested and showed that the seasonal cycle has no significant effect (SI text S1). Hence, in the paper results are shown for the full year.

Quantile regression. Quantile regressions were applied between detrended EKE anomalies and

detrended temperature anomalies or precipitation anomalies at each individual grid point. We computed regression slopes for the 10th, 50th, and 90th-percentiles in order to analyze how changes in the tails of the EKE distribution are related to changes in temperature or precipitation. The applied method is described in detail in Koenker and d’Orey41. Significant regression slopes were defined at the 5%-level based on confidence intervals computed from a rank test52.

Linear regression. Linear regression analyses were applied at each grid point between detrended EKE anomalies and detrended anomalies of geopotential height with both variables taken at 850 mb. Significant regression slopes were defined at the 5%-level.

Trend analysis. Linear trends were computed for seasonal-mean EKE at each grid point with significant trends defined at the 5%-level. EKE change in hot, cold, dry and wet months. For each grid point and month, the 10% most positive (or most negative) detrended temperature anomalies or precipitation anomalies were chosen as a representation of hot, cold, dry and wet months, respectively. This corresponds to approximately 14–43 data points depending on the number of months in the given season. Hence, ∆EKE = EK E extr. − EK E clim. × 100 EK E clim. is the change in percent between the mean EKE in these “extreme” months (EKEextr.) and the climatology of EKE (EKEclim.) in a particular month at a certain grid point. To test whether EKEextr. is significantly different from EKEclim. we applied a two-sample t-test as well as a Kolmogorow-Smirnow test. We defined significance at the 5%-level with both methods giving similar results (Figs S8 and S9). Note, that we focus on meteorological extremes, i.e. largest deviations from climatology. This definition of extremes could in principle be different to specifying extremes based on absolute values. However, since we analyze heat extremes in summer and cold extremes in winter we expect this difference to be minor.

References

1. Wang, S.-Y., Hipps, L., Gillies, R. R. & Yoon, J.-H. Probable causes of the abnormal ridge accompanying the 2013-2014 California drought: ENSO precursor and anthropogenic warming footprint. Geophys. Res. Lett. 41, 3220–3226 (2014). 2. Wallace, J. M., Held, I. M., Thompson, D. W. J., Trenberth, K. E. & Walsh, J. E. Global Warming and Winter Weather. Science 343, 729–730 (2014). 3. Stephens, E. & Cloke, H. Improving flood forecasts for better flood preparedness in the UK (and beyond). Geogr. J. 180, 310–316 (2014). 4. Coumou, D. & Rahmstorf, S. A decade of weather extremes. Nat. Clim. Chang. 2, 491–496 (2012). 5. Seneviratne, S. I. et al. in Manag. Risks Extrem. Events Disasters to Adv. Clim. Chang. Adapt. (Field, C. B. et al.) 109–230 (Cambridge University Press, 2012).


7

110


www.nature.com/scientificreports/ 6. Trenberth, K. E. Recent Observed Interdecadal Climate Changes in the Northern Hemisphere. Bull. Am. Meteorol. Soc. 71, 988–993 (1990). 7. Hurrell, J. W. Decadal trends in the north atlantic oscillation: regional temperatures and precipitation. Science 269, 676–679 (1995). 8. Thompson, D. W. & Wallace, J. M. Regional climate impacts of the Northern Hemisphere annular mode. Science 293, 85–89 (2001). 9. Wanner, H. et al. North Atlantic oscillation - Concepts and studies. Surv. Geophys. 22, 321–382 (2001). 10. Marshall, J. et al. North Atlantic climate variability: phenomena, impacts and mechanisms. Int. J. Climatol. 21, 1863–1898 (2001). 11. Trenberth, K. E., Fasullo, J. T. & Shepherd, T. G. Attribution of climate extreme events. Nat. Clim. Chang. 5, 725–730 (2015). 12. Horton, D. E. et al. Contribution of changes in atmospheric circulation patterns to extreme temperature trends. Nature 522, 465–469 (2015). 13. Hoskins, B. & Woollings, T. Persistent Extratropical Regimes and Climate Extremes. Curr. Clim. Chang. Reports (2015), doi: 10.1007/s40641-015-0020-8. 14. Screen, J. A. & Simmonds, I. Amplified mid-latitude planetary waves favour particular regional weather extremes. Nat. Clim. Chang. 4, 704–709 (2014). 15. Coumou, D., Petoukhov, V., Rahmstorf, S., Petri, S. & Schellnhuber, H. J. Quasi-resonant circulation regimes and hemispheric synchronization of extreme weather in boreal summer. Proc. Natl. Acad. Sci. 1412797111– (2014). doi: 10.1073/pnas.1412797111. 16. Petoukhov, V., Rahmstorf, S., Petri, S. & Schellnhuber, H. J. Quasiresonant amplification of planetary waves and recent Northern Hemisphere weather extremes. Proc. Natl. Acad. Sci. USA 110, 5336–41 (2013). 17. Schubert, S., Wang, H. & Suarez, M. Warm season subseasonal variability and climate extremes in the northern hemisphere: The role of stationary Rossby waves. J. Clim. 24, 4773–4792 (2011). 18. Dole, R. et al. Was there a basis for anticipating the 2010 Russian heat wave? Geophys. Res. Lett. 38, 1–5 (2011). 19. Bengtsson, L., Hodges, K. I. & Roeckner, E. Storm tracks and climate change. J. Clim. 19, 3518–3543 (2006). 20. Harvey, B. J., Shaffrey, L. C., Woollings, T. J., Zappa, G. & Hodges, K. I. How large are projected 21st century storm track changes? Geophys. Res. Lett. 39, 1–5 (2012). 21. Lehmann, J., Coumou, D., Frieler, K., Eliseev, A. V & Levermann, A. Future changes in extratropical storm tracks and baroclinicity under climate change. Environ. Res. Lett. 9, 084002 (2014). 22. Chang, E. K. M., Lee, S. & Swanson, K. L. Storm track dynamics. J. Clim. 15, 2163–2183 (2002). 23. Pinto, J. G. et al. Changes in storm track and cyclone activity in three SRES ensemble experiments with the ECHAM5/MPI-OM1 GCM. Clim. Dyn. 29, 195–210 (2007). 24. Dong, B., Sutton, R. T., Woollings, T. & Hodges, K. Variability of the North Atlantic summer storm track: mechanisms and impacts on European climate. Environ. Res. Lett. 8, 034037 (2013). 25. Hoskins, B. & Valdes, P. On the existence of storm-tracks. J. Atmos. Sci. 47, 1854–1864 (1990). 26. Branstator, G. Organization of Storm Track Anomalies by Recurring Low-Frequency Circulation Anomalies. J. Atmos. Sci. 52, 207–226 (1995). 27. Hawcroft, M. K., Shaffrey, L. C., Hodges, K. I. & Dacre, H. F. How much Northern Hemisphere precipitation is associated with extratropical cyclones? Geophys. Res. Lett. 39, L24809 (2012). 28. Raible, C. C., Yoshimori, M., Stocker, T. F. & Casty, C. Extreme midlatitude cyclones and their implications for precipitation and wind speed extremes in simulations of the Maunder Minimum versus present day conditions. Clim. Dyn. 28, 409–423 (2007). 29. Pfahl, S. & Wernli, H. Quantifying the Relevance of Cyclones for Precipitation Extremes. J. Clim. 25, 6770–6780 (2012). 30. Trenberth, K. E. & Shea, D. J. Relationships between precipitation and surface temperature. Geophys. Res. Lett. 32, 1–4 (2005). 31. Coumou, D., Lehmann, J. & Beckmann, J. The weakening summer circulation in the Northern Hemisphere mid-latitudes. Science 348, 324–327 (2015). 32. Luterbacher, J. et al. Circulation dynamics and its influence on European and Mediterranean January-April climate over the past half millennium: Results and insights from instrumental data, documentary evidence and coupled climate models. Clim. Change 101, 201–234 (2010). 33. Bladé, I., Liebmann, B., Fortuny, D. & van Oldenborgh, G. J. Observed and simulated impacts of the summer NAO in Europe: Implications for projected drying in the Mediterranean region. Clim. Dyn. 39, 709–727 (2012). 34. Nakamura, H. Year-to-Year and interdecadal variability in the activity of intraseasonal fluctuations in the Northern Hemisphere wintertime circulation. Theor. Appl. Climatol. 55, 19–32 (1996). 35. Overland, J. E., Francis, J. a., Hanna, E. & Wang, M. The recent shift in early summer Arctic atmospheric circulation. Geophys. Res. Lett. 39, 1–6 (2012). 36. Folland, C. K. et al. The summer North Atlantic oscillation: Past, present, and future. J. Clim. 22, 1082–1103 (2009). 37. Zappa, G., Shaffrey, L. C., Hodges, K. I., Sansom, P. G. & Stephenson, D. B. A Multimodel Assessment of Future Projections of North Atlantic and European Extratropical Cyclones in the CMIP5 Climate Models*. J. Clim. 26, 5846–5862 (2013). 38. Ulbrich, U., Leckebusch, G. C. & Pinto, J. G. Extra-tropical cyclones in the present and future climate: a review. Theor. Appl. Climatol. 96, 117–131 (2009). 39. Dee, D. P. et al. The ERA-Interim reanalysis: configuration and performance of the data assimilation system. Q. J. R. Meteorol. Soc. 137, 553–597 (2011). 40. Blackmon, M. A climatological spectral study of the 500 mb geopotential height of the Northern Hemisphere. J. Atmos. Sci. 33, 1607–1623 (1976). 41. Koenker, R. & D’Orey, V. Computing Regression Quantiles. Appl. Stat. 36, 383–393 (1987). 42. Adler, R. F. et al. The Version-2 Global Precipitation Climatology Project (GPCP) Monthly Precipitation Analysis (1979–Present). J. Hydrometeorol 4, 1147–1167 (2003). 43. Schär, C. et al. The role of increasing temperature variability in European summer heatwaves. Nature 427, 332–336 (2004). 44. Schär, C., Lüthi, D., Beyerle, U. & Heise, E. The soil-precipitation feedback: A process study with a regional climate model. J. Clim. 12, 722–741 (1999). 45. Mueller, B. & Seneviratne, S. I. Hot days induced by precipitation deficits at the global scale. Proc. Natl. Acad. Sci. 109, 12398–12403 (2012). 46. Screen, J. A. Arctic amplification decreases temperature variance in northern mid- to high-latitudes. Nat. Clim. Chang. 4, 577–582 (2014). 47. Gulev, S. & Zolina, O. Extratropical cyclone variability in the Northern Hemisphere winter from the NCEP/NCAR reanalysis data. Clim. Dyn 17, 795–809 (2001). 48. Diffenbaugh, N. S. & Scherer, M. Likelihood of July 2012 U.S. temperatures in pre-industrial and current forcing regimes. in Explain. Extrem. Events 2012 from a Clim. Perspect. Bull. Amer. Meteor. Soc. 94(9) S6–S9 (2013). 49. Vavrus, S., Walsh, J. E., Chapman, W. L. & Portis, D. The behavior of extreme cold air outbreaks under greenhouse warming. Int. J. Climatol. 26, 1133–1147 (2006). 50. Cohen, J. L., Furtado, J. C., Barlow, M. a, Alexeev, V. A. & Cherry, J. E. Arctic warming, increasing snow cover and widespread boreal winter cooling. Environ. Res. Lett. 7, 014007 (2012).


8


111

www.nature.com/scientificreports/ 51. Murakami, M. Large-Scale Aspects of Deep Convective Activity over the GATE Area. Mon. Weather Rev. 107, 994–1013 (1979). 52. Koenker, R. in Asymptot. Stat. 349–359 (1994). doi: 10.1007/978-3-642-57984-4_29.

Acknowledgements

We thank ECMWF and NOAA for making their data available. The work was supported by the German research Foundation (CO994/2-1) and the German Federal Ministry of Education and Research (01LN1304A).

Author Contributions

D.C. and J.L. conceived and designed the research, analyzed the data and co-wrote the manuscript. J.L. performed the data analysis.

Additional Information

Supplementary information accompanies this paper at http://www.nature.com/srep Competing financial interests: The authors declare no competing financial interests. How to cite this article: Lehmann, J. and Coumou, D. The influence of mid-latitude storm tracks on hot, cold, dry and wet extremes. Sci. Rep. 5, 17491; doi: 10.1038/srep17491 (2015). This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/


9

112


Supplementary Materials for:

The influence of mid-latitude storm track activity on hot, cold, dry and wet extremes Jascha Lehmann* and Dim Coumou *Corresponding author. E-mail: [email protected]

S1: Seasonal analysis of extremes The analysis of temperature extremes requires a seasonal approach, since the effect of storms on land surface temperatures will be different between summer and winter season. This is demonstrated in Fig. S1a,b which show the seasonal cycles of mid-latitude mean land surface and sea surface temperature. In summer, the oceans warm slower than the continents, but they also keep the heat longer than land when it gets colder in winter. Consequently, the seasonal cycle is less pronounced over the oceans than over land with sea surface temperature lagging behind land surface temperature by about one month. Two distinct seasons can be derived from this mechanism: the summer season (May-September) where land surface temperature is above sea surface temperature and the winter season (November-February) with the opposite behavior. More precisely, mid-latitude storms bring maritime conditions from the Atlantic to Europe and from the Pacific to North America. Seasonal surface temperatures over the Atlantic and Europe (Fig. S1c) and over the Pacific and North America (Fig. S1d) reveal the same delay between sea and land surface temperatures as observed for the mid-latitudes in general. The influence of mid-latitude storms on precipitation does not show a significant dependence on the season. Consistently, regression patterns between EKE and precipitation anomalies look essentially identical between summer and winter season (Fig. S2). Only the magnitude of the regression slopes differs as precipitation is generally stronger in winter. In our study we thus use all calendar months for the analysis of precipitation extremes.

S2: Sensitivity analysis of altitude To test the sensitivity of our regression analysis to EKE at different pressure levels, we repeated the regression of temperature and precipitation anomalies with EKE at 500 mb. Results show that in summer EKE at 500 mb is represented by a much more zonal flow compared to EKE at 850 mb (see Fig. S6-S7). This is due to less orographic friction at higher altitudes. However,


113

resulting slopes of quantile regression analyses reveal similar patterns at both pressure levels, i.e. negative correlations over essentially all mid-latitude continental regions with most slopes being statistically significant at the 5%-level (Fig. S11). Similarly, we find consistent regression patterns between both pressure levels in winter (Fig. S12). Correlation analysis between precipitation and EKE seems to be more sensitive to the pressure level at which we compute EKE. Whereas the correlation between precipitation and EKE is generally positive and statistically significant at the lower troposphere, regression slopes are very small in magnitude and mostly non-significant at mid-troposphere (Fig. S13).

S3: Additional figures

Fig. S1. Seasonal cycle of surface temperatures taken from the ERA-Interim data set at 1000 mb. Temperature curves are shown for land (solid line) and ocean (dashed line) averages in different regions: mid-latitudes in 1979-1982 (a), mid-latitudes in 2010-2013 (b), Atlantic and Europe (c), and Pacific and North America (d) in 1979-1982. We define a summer season (May-

114


September, pink bars) where land surface temperatures are above sea surface temperatures and vice versa for the winter season (November-February, blue bars).

Fig. S2. Slope of 90th-percentile regressions between anomalies in EKE and precipitation in different seasons. Results are similar between summer (a) and winter season (b). Stippling indicates significance at the 5%-level. Grey contour lines indicate EKE climatology of the given season. Land regions higher than 1 km have been masked. All maps are created using the open source software R.


115

Fig. S3 . Same as Fig. 1 in main manuscript but for quantile regression of v’2 with summer temperatures. All maps are created using the open source software R.

Fig. S4 . Same as Fig. 2 in main manuscript but for quantile regression of v’2 with winter temperatures. All maps are created using the open source software R.

116


Fig. S5 . Same as Fig. 3 in main manuscript but for quantile regression of v’2 with precipitation. All maps are created using the open source software R.

Fig. S6. EKE climatology at 850 mb. Contour plots of seasonally averaged EKE climatology at 850 mb computed for the full time period considered (1979-2014) and shown for May-June-JulyAugust-September (a) and November-December-January-February (b). All maps are created using the open source software R.


117

Fig. S7. EKE climatology at 500 mb. Contour plots of seasonally averaged EKE climatology at 500 mb computed for the full time period considered (1979-2014) and shown for May-June-JulyAugust-September (a) and November-December-January-February (b). All maps are created using the open source software R.

118


Fig. S8. Change in EKE in extreme months compared to local climatology. a-d, EKE in 10% hottest summer months (a), 10% coldest winter months (b), 10% wettest months (c), and 10% driest months (d) given as percentage change compared to local EKE climatology of given season. Stippling indicates significance at the 5%-level derived from a two-sample t-test. Land regions higher than 1 km have been masked. All maps are created using the open source software R.


119

Fig. S9. Change in EKE in extreme months compared to climatology. Same as Fig. S6 but using Kolmogorow-Smirnow test to estimate significance at the 5%-level. Significant changes are found over similar regions as determined using a two-sample t-test (Fig. S6). All maps are created using the open source software R.

120


Fig. S10. Difference in absolute regression slopes between 90th and 10th-percentile. Results are shown for regressing EKE with temperatures in summer (a) and winter (b), and for regression analysis between annual EKE and precipitation (c). The difference was computed from absolute values of the regression slopes so that positive values in close to all regions in all three panels imply that regression slopes are steeper for higher quantiles. Land regions higher than 1 km have been masked. All maps are created using the open source software R.


121

Fig. S11. Slope of quantile regressions between anomalies of EKE at 500 mb and temperature in summer. Same as Fig. 1 in main manuscript but for EKE at 500 mb. All maps are created using the open source software R.

122


Fig. S12. Slope of quantile regressions between anomalies of EKE at 500 mb and temperature in winter. Same as Fig. 2 in main manuscript but for EKE at 500 mb. All maps are created using the open source software R.


123

Fig. S13. Slope of quantile regressions between anomalies of EKE at 500 mb and precipitation in all calendar months. Same as Fig. 3 in main manuscript but for EKE at 500 mb. All maps are created using the open source software R.

124


Fig. S14. Slope of quantile regressions between anomalies of EKE (at 850 mb) and precipitation (GPCP) in all calendar months. Same as Fig. 3 in main manuscript but with precipitation time series taken from the GPCP data set. All maps are created using the open source software R.

Fig. S15. Slope of 90th-percentile regressions between anomalies in EKE (at 850 mb) and precipitation (GPCP) in different seasons. Same as Fig. S2 but with precipitation time series taken from the GPCP data set. All maps are created using the open source software R.


125

Fig. S16. Trends in EKE at 500 mb. Same as Fig. 5a, b in main manuscript but at 500 mb and seasonal averages are computed for June-July-August (a) and December-January-February (b) to make the figures comparable to results from Horton et al.1. Black boxes indicate the following regions used in Horton et al.1: (from left to right) eastern US, central US, eastern US, Europe, western Asia, central Asia, eastern Asia. All maps are created using the open source software R.

126


Fig. S17. Slope of linear regression between anomalies of EKE and GPH at 500 mb. Same as Fig. 4 in main manuscript but at 500 mb and seasonal averages are computed for June-JulyAugust (a) and December-January-February (b) to make the figures comparable to results from Horton et al.1. Black boxes indicate the following regions used in Horton et al.1: (from left to right) eastern US, central US, eastern US, Europe, western Asia, central Asia, eastern Asia. All maps are created using the open source software R.


127

Fig. S18. Trends in EKE at 500 mb. Same as Fig. 5 in main manuscript but for EKE at 500 mb. All maps are created using the open source software R.

References used in SI 1.

Horton, D. E. et al. Contribution of changes in atmospheric circulation patterns to extreme temperature trends. Nature 522, 465–469 (2015).

2.5. Future changes in extratropical storm tracks and baroclinicity under climate change Jascha Lehmann, Dim Coumou, Katja Frieler, Alexey V. Eliseev, and Anders Levermann. This article provides a comprehensive analysis of extratropical storm track changes under a high-emission scenario as projected by an ensemble of CMIP5 climate models. Among other results, it is shown that strongest changes are expected in Northern Hemisphere summer where the pronounced weakening of storm track activity is consistent with changes in the Eady growth rate and in particular in the vertical wind shear. Jascha Lehmann performed all analyses, analyzed the data and wrote the text. All authors participated in the interpretation of the results and helped to improve the manuscript. Alexey V. Eliseev pre-designed the R-Code for the applied bandpass filter which was adapted and improved by Jascha Lehmann. Published in Environmental Research Letters 9, 084002 (2014), DOI 10.1088/17489326/9/8/084002.

2.5. Future changes in extratropical storm tracks and baroclinicity under climate change

131

Environmental Research Letters Environ. Res. Lett. 9 (2014) 084002 (8pp)

doi:10.1088/1748-9326/9/8/084002

Future changes in extratropical storm tracks and baroclinicity under climate change Jascha Lehmann1,2, Dim Coumou1, Katja Frieler1, Alexey V Eliseev1,3,4 and Anders Levermann1 1

Potsdam Institute for Climate Impact Research, Germany University of Potsdam, Germany 3 A. M. Obukhov Institute of Atmospheric Physics RAS, Russia 4 Kazan (Volga Region) Federal University, Russia 2

E-mail: [email protected] Received 8 May 2014, revised 7 July 2014 Accepted for publication 8 July 2014 Published 5 August 2014 Abstract

The weather in Eurasia, Australia, and North and South America is largely controlled by the strength and position of extratropical storm tracks. Future climate change will likely affect these storm tracks and the associated transport of energy, momentum, and water vapour. Many recent studies have analyzed how storm tracks will change under climate change, and how these changes are related to atmospheric dynamics. However, there are still discrepancies between different studies on how storm tracks will change under future climate scenarios. Here, we show that under global warming the CMIP5 ensemble of coupled climate models projects only little relative changes in vertically averaged mid-latitude mean storm track activity during the northern winter, but agree in projecting a substantial decrease during summer. Seasonal changes in the Southern Hemisphere show the opposite behaviour, with an intensification in winter and no change during summer. These distinct seasonal changes in northern summer and southern winter storm tracks lead to an amplified seasonal cycle in a future climate. Similar changes are seen in the mid-latitude mean Eady growth rate maximum, a measure that combines changes in vertical shear and static stability based on baroclinic instability theory. Regression analysis between changes in the storm tracks and changes in the maximum Eady growth rate reveal that most models agree in a positive association between the two quantities over mid-latitude regions. S Online supplementary data available from stacks.iop.org/ERL/9/084002/mmedia Keywords: storm tracks, baroclinicity, climate change Introduction

Ulbrich 2004, Pinto et al 2007, Schwierz et al 2009). Thus, the question of how extratropical storm tracks will change under global warming has been intensively analyzed in recent studies (Yin 2005, Bengtsson et al 2009, Ulbrich et al 2009, Catto et al 2011) with an emerging attention given to the analyses of multi-model ensembles (Ulbrich et al 2008, O’Gorman 2010, Chang et al 2012, Harvey et al 2012, 2013, Zappa et al 2013). Some authors have identified and analyzed individual cyclones and have shown that in the Northern Hemisphere (NH) the total number of cyclones are projected to decrease under climate change, whereas a potential increase exists for the number of extreme cyclones (Ulbrich et al 2009, Mizuta 2012, Zappa et al 2013). Other studies have used bandpass filtered measures of the storm tracks to

The day-to-day variability of weather in the mid-latitude regions is strongly affected by extratropical storm tracks. Storms in the northern mid-latitudes account for much of the total and extreme precipitation climatology (Hawcroft et al 2012, Pfahl and Wernli 2012) and the strong winds and potentially associated storm surges are among the major natural hazards in these regions (Leckebusch and Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. 1748-9326/14/084002+08$33.00

1

© 2014 IOP Publishing Ltd

132


J Lehmann et al

Environ. Res. Lett. 9 (2014) 084002

gain information about their changes on a global scale. Here, the largest consensus exists for a poleward and upward shift of the storm tracks in both hemispheres (Yin 2005, Chang et al 2012). Both diagnostic tools have furthermore been used to analyze the influence of different fields on the observed storm track changes, such as the horizontal temperature gradient, the upper-level zonal wind, the Eady growth rate, or ocean circulation changes (e.g. O’Gorman 2010, Mizuta 2012, Woollings et al 2012, Harvey et al 2013). However, there are still substantial differences in the projected storm track responses to climate change between different state-of-the-art climate models (Harvey et al 2012) and even larger uncertainties when it comes to the underlying mechanisms causing these storm track changes. In particular, the magnitude and sign of local storm track changes are in weak agreement between many individual model projections and the multi-model mean response (Ulbrich et al 2008, 2009, Harvey et al 2012). This letter aims to contribute to the understanding of the storm track responses to climate change. Therefore, the latest generation of climate model projections from the Coupled Model Intercomparison Project Phase 5 (CMIP5) has been used to analyze the seasonal responses of extratropical storm tracks to future warming. We present historic and future changes (based on the emission scenario RCP8.5 (Moss et al 2010)) in winter and summer storm track activity of 22 CMIP5 models for both hemispheres. After a general assessment of the pattern of storm track changes, we analyze their influence on the seasonal cycle before we examine how the simulated storm track changes are associated with changes in the large-scale baroclinicity.

Petoukhov et al 2008, Ren et al 2010). The difference between the applied 2.5–6 day bandpass filtering in this letter and the original 2–6 day bandpass range is likely to be minor, because only a relatively small amount of energy is in the 2–2.5 range (Randel and Held 1991). The EKE per unit volume is thereby simply calculated from EKE = 0.5 ⋅ u′2 + v′2

(

)

where u′ and ν′ are the band pass filtered zonal and meridional wind speeds. Subsequently, the EKE of each model is interpolated onto a common 2.5° × 2.5° grid and a massweighted vertical average between 250 hPa and 850 hPa is applied. We define the seasonality in EKE as the difference in magnitude of mid-latitude mean EKE between winter and summer. Mid-latitude means are calculated by averaging EKE over all longitudes and between 35°–65°N for the NH and between 35°–65°S for the Southern Hemisphere (SH). The changes in EKE and the seasonality over time are given as relative changes with respect to the historical base period (1981–2000), where we focus on anomalies by the end of the 21st century (2081–2100). The 95% confidence interval and the statistical significance of the mean change in seasonality are derived for each model and the multi-model mean from a simple two-sample t-test. In the second part of this study we analyze the relation between changes in EKE and changes in large-scale baroclinicity, represented by the maximum Eady growth rate (Lindzen and Farrell 1980, Hoskins and Valdes 1990). Changes in Eady growth rate are determined by changes in static stability and changes in vertical wind shear. Whilst static stability depends on the vertical potential temperature gradient, the vertical shear is closely related to the horizontal temperature gradient via the thermal wind equation. To analyze the influence of vertical shear on storm track changes compared to contributions from static stability, we also analyze the relation between EKE and vertical shear. In this letter, we only present results for the analysis of the Eady growth rate. However, differences to the analysis of the vertical shear, calculated between 250–850 hPa, are discussed in the text and equivalent figures are given in the SI. The maximum Eady growth rate is defined as

Data and methods Daily-mean zonal and meridional wind speed data are used from all CMIP5 models for which the appropriate data were available at the time of writing. A list of the models is given in table S1 in the supplementary information (SI), available at stacks.iop.org/ ERL/9/084002/mmedia. The analysis is based on the time period 1950–2100, where 1950–2005 is based on historical forcing, and concatenated with 2006–2100 based on the high emissions scenario RCP8.5. We chose the scenario with the highest emission pathway, because we are interested in storm track changes under large global warming effects. In order to ensure comparability between models, only a single realization (the r1i1p1 run) from each model is used. In this letter, storm tracks are estimated by the eddy kinetic energy (EKE) which is calculated for each individual month by applying a 2.5–6 day bandpass filter to the described daily wind field data. A similar approach was suggested by Blackmon (1976) for a 2–6 day bandpass range, and has been followed by several studies (e.g. Yin 2005, Ulbrich et al 2008, Harvey et al 2013). The EKE can hence be used as a measure for the interplay between the intensity and frequency of high and low pressure systems. The applied filter was developed by Murakami (Murakami 1979) and has been shown to produce accurate results (Christoph et al 1995,

σ BI = 0.31 ⋅

f dV ⋅ N dz

where f is the Coriolis parameter, N the Brunt-Väisälä frequency, V the horizontal wind vector and z the vertical height. We calculate this quantity between 250–850 hPa. To assess the relationship between storm tracks and Eady growth rate, a linear regression is applied between both quantities at each grid point and for each model and season. To gain information about the correlation of the year-to-year variability of the storm tracks and the Eady growth rate, the time series are detrended before the regression analysis by subtracting the smoothed mean value with a half-width of 30 years. The same calculation is then repeated, but this time including all twelve months and without the detrending and the seasonal averaging process. The latter regression analysis 2


133

J Lehmann et al


2

2

Figure 1. Projected storm track changes under future climate conditions represented by the difference in multi-model mean EKE (m /s )

between the end of the 21st century (2081–2100) and the 20th century (1981–2000) for (a) DJF season and (b) JJA season. Contours of the 20th century EKE are shown in black, and regions of land higher than 1 km have been masked. Four regions of large EKE changes have been framed with black rectangles.

Storm track seasonality

therefore describes the association between the seasonal cycle of EKE and the seasonal cycle of Eady growth rate.

After this general assessment of seasonal EKE responses, the mid-latitude means of EKEs are used to analyze changes in the seasonality in both hemispheres. Figure 2 shows the evolution of changes in mid-latitude mean EKE and seasonality. In NH winter, the multi-model median EKE increases by 5% until the end of the 21st century (figure 2(a), blue line). In contrast, the multi-model median EKE weakens substantially during JJA with a peak reduction of −14% in the year 2100 (figure 2(a), black line). The SH exhibits a strong increase in multi-model median EKE during JJA (14%) and almost no change during DJF (2%). These changes in EKE result in an amplified seasonal cycle in both hemispheres, with the amplification increasing at a rate of roughly 2% per decade in the NH and 6% per decade in the SH (figures 2(a) and (b), red line). The enhanced seasonality in both hemispheres is a robust projection across all models (figure 3). For the NH, the relative change in seasonality predicted by individual models is close to the value of the multi-model mean, except for one model (GFDL-CM3). For this model, the historical seasonality is exceptionally small compared to the other models (see figure S8 in SI). In the SH, the inter-model spread is generally larger. This is mainly due to the chosen representation of the change in seasonality relative to the historical period, which is about three times smaller in magnitude in the SH than in the NH. Therefore small absolute changes in SH seasonality can lead to large relative changes and large confidence intervals, as can be seen in figure 3. However, an enhanced seasonality is evident in both hemispheres in all models, leading to a statistically significant increase in seasonality at the 95% confidence interval for the multi-model mean. The amplification at the end of the century is about four times stronger in the SH (92%, figure 3(b)) than in the NH (23%, figure 3(a)). However, the seasonal cycle of EKE is about three times

Spatial pattern of storm track changes Figure 1 shows the response of the multi-model mean EKE to climate change. During the northern winter (December-January-February (DJF)), the storm tracks shift polewards in the SH indicated by a reduction in EKE over the subtropics and an intensification southwards of the peak in the historical EKE. In the NH, the most prominent reductions in EKE are confined to the subtropical Atlantic and Pacific region and the strongest intensification can be seen over the North-East Atlantic, Eurasia, the North Pacific, and North America. In contrast, EKE changes during June-July-August (JJA) are rather uniform in each hemisphere. Here, the NH exhibits a general decrease in EKE in the mid-latitudes, with the greatest reduction over the Atlantic and Pacific Ocean basin, and a strong increase over the SH mid-latitudes. The pattern of EKE changes are similar across different altitudes of the troposphere, with the only exception of the northern winter. Here, the strongest differences can be seen at high latitudes between changes in the vertically averaged EKE and changes in EKE at 500 hPa (figure S1 in SI). In particular, EKE at 500 hPa exhibits negative changes in response to climate change over the North Pacific and North America. Also, changes in EKE during JJA are more consistent between the models than during DJF (figure S15 in SI). Our results are qualitatively similar to findings from other studies which use similar diagnostic tools (O’Gorman 2010, Chang et al 2012, Harvey et al 2013). However, notable differences in the pattern of EKE changes can again be seen during DJF over North America and Eastern Europe. Here, the response of EKE seems to be sensitive to the applied diagnostic tool and the vertical height. 3

134


J Lehmann et al


Figure 2. Changes in mid-latitude mean EKE shown separately for the (a) NH and (b) SH. The upper panel of each figure shows the time evolution of mid-latitude mean EKE changes during DJF (blue line) and JJA (black line). The evolution is given relative to the historical base period (1981–2000). Both lower panels show the change in seasonality (SNL), again given relative to the magnitude of the historical seasonal cycle. In all figures the shaded area reflects the interquartile range of the model spread and the thick solid line depicts the smoothed median value with a half-width of 8 years.

(5%, figure 2(a)) is associated with an analogous increase in the multi-model median Eady growth rate (7%, figure 4(a)). During JJA, both quantities show a decrease over the 21st century (Eady growth rate, −4%). In the SH, the multi-model median Eady growth rate exhibits a weak amplification during DJF (3%, figure 4(b)), but increases substantially during JJA (7%, figure 4(b)), consistent with projected changes in EKE. We also find similar changes in the mid-latitude mean vertical shear (figure S9 in SI). However, the strongest differences can be seen during southern summer, where the multi-model median vertical shear increases by 7% until the 21st century, whereas both EKE and Eady growth rate show almost no change. This difference can be explained by the influence of static stability on baroclinicity. Whereas vertical shear increases more or less homogeneously in the SH midlatitudes, changes in Eady growth rate show a dipole pattern with a decrease at the equatorward side and an increase at the poleward side (similar to EKE, figure 1). Thus, an increase in static stability counterbalances the increase in vertical shear

larger in the NH than in the SH, implying that in absolute terms the projected changes in seasonality are comparable between the two hemispheres. We tested the sensitivity of our results to the vertical pressure level, which revealed that notable differences in the trends of mid-latitude mean EKE changes are only evident in NH winter and this is consistent with results from the spatial pattern of EKE changes at different altitudes (see figures S2–S7 in SI). However, the amplification of the seasonal cycle in both hemispheres is a robust feature across the troposphere.

Atmospheric dynamics Figure 4 shows the seasonal changes of the mid-latitude mean Eady growth rate as presented in figure 3 for EKE. Similar trends between both quantities can be seen in each hemisphere and season. In the NH, the increase in EKE during DJF 4


135

J Lehmann et al


In both hemispheres, large regions with positive regression slopes are evident during both seasons. These regions are primarily found over the mid-latitudes, i.e. regions of high EKE values (see contours in figure 1). In the NH mid-latitudes, the year-to-year variability of EKE during JJA (figure 5(b)) is especially well correlated with the year-to-year variability of Eady growth rate. Here, almost all models show a positive correlation (red shading) and most models have a statistically significant correlation at 95% confidence (stippling). During DJF, most models agree in a significant positive relation between EKE and Eady growth rate over the North Pacific. Model agreement is somewhat weaker over parts of North America and the western North Atlantic. This is probably related to the uncertainty in the projected EKE changes during northern winter over this region (see figure 1 and figure S1 in SI). Strictly speaking, based on baroclinic theory one would not necessarily expect a point-to-point correlation between EKE and baroclinicity. Baroclinic theory tells that small disturbances grow to large amplitudes along sufficiently (1000 s of km) long baroclinic zones. Thus, downstream of such baroclinic zones EKE can still be large although baroclinicity weakens, which might be a reason for the weak agreement over Eurasia. Over the (sub-) tropics there are large regions that indicate a negative relation between EKE and Eady growth rate. This can be explained by very low magnitudes of EKE and Eady growth rate in these latitudes, which lead to high uncertainties in the estimated regression slopes, that vary around zero. This is confirmed by the multi-model mean regression slope which is close to zero in most parts of the (sub-) tropics (see figure S13 in SI). In the SH summer (figure 5(a)), a positive and, for most models, significant correlation between EKE and Eady growth rate exists over an area around New Zealand and over a zonal band between 50°–65°S. During JJA, the regression pattern looks similar. For both seasons, there is weaker agreement between EKE and Eady growth rate around 45°S, which is exactly at the core of the storm track. Speculatively, non-linear effects due to saturation of EKE play a role, such that an increase in Eady growth rate does not lead to a linear increase in EKE. Regression analysis of EKE and vertical shear (figure S14 in SI) shows similar patterns of model agreement for both hemispheres and both seasons. This implies that Eady growth rate, but also vertical shear by itself, can explain much of the year-to-year changes in EKE over the mid-latitudes. Strong changes in EKE and Eady growth rate are not only seen in the year-to-year variability but also during the course of a year (figure S12 in SI). We therefore also regressed the changes in EKE and Eady growth rate due to their seasonal cycle against each other. Figure 6 shows the spatially averaged regression slopes for four chosen regions with especially strong changes in EKE over the 21st century. The region borders are shown in figures 1 and 4, and are defined for Europe (Europe: 40°–60°N, 30°W–50°E), the North Pacific (NH.Pacific: 40°–60°N, 140°E–120°W), and two zonal bands in the SH (SH.35°–45°: 35°–45°S, 180° W–180°E and SH.45°–65°: 45°–65°S, 180°W–180°E). For each region, the multi-model mean and the inter-model spread

Figure 3. Change in seasonality of mid-latitude mean EKE under

global warming. This is shown for the (a) NH and (b) SH. For each model and the multi-model mean, the mean value of the seasonality change is shown as a thick vertical line and the 95% confidence level is shown as a box. Both quantities were derived from a two-sample ttest, based on the last 20 years of the 20th and 21st centuries.

and therefore the mid-latitude mean EKE is almost unaffected. Consistent with findings for mid-latitude mean EKE, the projected trends in mid-latitude mean vertical shear are qualitatively similar between the upper and lower troposphere, except for northern winter. Here, the multi-model median vertical shear increases in the upper troposphere (between 250 hPa and 500 hPa) but decreases in the lower troposphere (between 500 hPa and 850 hPa, figures S10–S11 in SI). This agrees well with changes in the horizontal temperature gradient, which are projected to increase in the upper troposphere, but decrease in the lower troposphere during northern winter (e.g. Harvey et al 2013). For the full troposphere this therefore yields only small relative changes in mid-latitude mean vertical shear, Eady growth rate, and EKE during northern winter, with the latter being consistent with findings from O’Gorman (2010). Our results suggest that mid-latitude mean changes in Eady growth rate drive changes in EKE. To quantitatively analyze this relationship, we assess how the year-to-year variabilities of both quantities are correlated by applying a regression analysis between seasonally averaged and detrended time series of EKE and Eady growth rate. Figure 5 shows the ratio between the number of models which exhibit a positive and those which exhibit a negative regression slope. Thus, a ratio of one implies that all models exhibit a positive correlation between EKE and Eady growth rate and a ratio of zero indicates that all models show a negative correlation. To highlight the storm track relevant regions, areas outside of the mid-latitudinal belt between 35°–65°N and 35°–65°S are shown semi-transparently. 5

136


J Lehmann et al


Figure 4. Changes in mid-latitude mean Eady growth rate separately shown for the (a) NH and (b) SH as in figure 3 for EKE.

filtered daily-mean zonal wind from 22 different CMIP5 models. In the first section, the change in seasonality, i.e. the difference in magnitude between winter and summer EKE, has been studied in detail. Afterwards, a linear regression model has been applied to EKE and Eady growth rate to analyze how changes in EKE are related to changes in largescale baroclinicity. The pattern of multi-model mean storm track responses to climate change is different in each hemisphere, and between winter and summer seasons. Whereas most studies focus on the stronger winter time storm tracks, here we show that for the mid-latitude mean storm tracks, CMIP5 models project a large and consistent change in EKE in both hemispheres during JJA, which implies a larger seasonality in a future climate. The latter is in agreement with previous analysis based on CMIP3 models (O’Gorman 2010). The amplified seasonal cycle is a robust feature across all models, and leads to a significant increase in seasonality of the multimodel mean at the 95% confidence interval in both hemispheres. This implies that whereas the SH exhibits an amplification of the stronger winter storm tracks, the largest relative changes in the NH are expected during summer, where CMIP5 models project a robust weakening of the storm tracks. Similar trends are also evident for changes in the midlatitude mean Eady growth rate. This suggests, that midlatitude mean changes in Eady growth rates drive changes in EKE. The Eady growth rate is a suitable measure of baroclinicity and describes the potential of small perturbations to develop into larger storms. It is thus a good predictor for EKE at the early stage of storm evolution, but might be less applicable for later stages. Nevertheless, regression analysis reveals that in both seasons there are large regions where models agree on a positive correlation between the year-to-

of the regionally averaged regression slope is shown for regression of the year-to-year variability during DJF (dark blue boxes) and JJA (light blue boxes) and regression of the seasonal cycle (yellow boxes). Within all four regions, the magnitudes of the regression slopes are in general agreement across the different regression methods, as indicated by the overlapping ranges of the full inter-model spread. Regression slopes from regressing the seasonal cycle are qualitatively similar to regression slopes from the year-to-year variability, but generally larger in magnitude. We suggest that these differences in the magnitude of the regression slopes between different regression methods can at least partly be explained by differences in the variability of Eady growth rate changes. Changes in Eady growth rate are much larger over the seasonal cycle (i.e. between summer and winter) than from year-to-year (i.e. from one summer to the next). Therefore, a regression using the full seasonal cycle covers a larger range in the input variable (i.e. Eady growth rate) and is thus less affected by regression dilution, which causes a bias in the regression slope towards zero. An estimate of the magnitude in variability is the standard deviation, which is shown for the multi-model mean Eady growth rate in figure 6 above each box and whisker symbol. These numbers suggest that without the bias caused by the regression dilution we could expect to see even stronger agreements in the regression slopes between the methods.

Conclusion and discussion This letter has analyzed storm track changes under the RCP8.5 greenhouse gas emission scenario. The storm tracks are represented by EKE which we calculated from bandpass 6


137

J Lehmann et al


Figure 5. Model agreement on positive or negative regression coefficients. Colours indicate the ratio between the number of models showing positive (red shading) or negative (blue shading) regression coefficients, derived from linear regression analysis between the seasonally averaged and detrended time series of EKE and Eady growth rate in (a) DJF and (b) JJA. Stippling indicates regions where more than 50% of the models exhibit regression slopes that are significant at the 95% confidence level, as indicated by a corresponding p-value < 0.05. Regions outside of the zonal bands between 35°–65°N and 35°–65°S are shown semi-transparently and regions with topography above 1 km have been masked in grey. White boxes illustrate the regions used for the detailed regression analysis (see figure 6).

variability of EKE and the Eady growth rate. Regression analysis of the year-to-year variability and the variability due to changes associated with the seasonal cycle give similar regression slopes. This suggests that the relation between changes in Eady growth rate and EKE in different years is equivalent to the relation between changes driven by the seasonal cycle. Regression slopes are slightly larger for regressions of the seasonal cycle than for regressions of the year-to-year variability. We argue that this can at least partly be explained by the larger intra-annual variability in Eady growth rate, as compared to the inter-annual variability. Therefore, a regression using the full seasonal cycle is less affected by regression dilution than a regression using the year-to-year changes. Our results are in general agreement with other studies using similar metrics to analyze contributing factors to the projected storm track changes under future climate conditions (Yin 2005, Pinto et al 2008, O’Gorman 2010, Ren et al 2010, Wu et al 2010, Mizuta 2012, Harvey et al 2013). However, comparability is difficult, as most studies focus on the association of climatological changes due to global warming. This letter, on the other hand, analyzes the correlation of the yearto-year variability and the correlation of the seasonal cycle. Vertical shear and static stability mainly determine the baroclinicity in the atmosphere. Our results show that over the storm track relevant regions, Eady growth rate can explain much of the projected storm track changes. In addition, we find that storm track variability is dominated by changes in shear, and that the shear alone can statistically explain the changes in EKE in some seasons. This presumption is supported by results from Ren et al (2010) who find that during winter and summer, baroclinicity is mainly determined by the vertical shear over two regions confined to the North Pacific and Central Asia.

Figure 6. Regression slopes of different regions for both regression

methods. Results from regressing the year-to-year variability between EKE and vertical shear are coloured in light blue (JJA) and dark blue (DJF) and results from the regression analysis of the seasonal cycle are coloured in yellow. In each case, the box and whisker symbols indicate the median, the interquartile range and the extreme values of the inter-model spread. Above each box and whisker symbol the standard deviation (104 1/s) of the corresponding time series of the Eady growth rate is shown.

year variability in EKE and Eady growth rate. These regions are primarily found over the storm track relevant mid-latitudes, where more than half of the models exhibit significant correlations. Regression of the year-to-year variability between EKE and vertical shear yields similar results. For the four chosen regions of large storm track changes (Europe, NH.Pacific, SH.35°–45°, SH.45°–65°), models agree in a strong correlation between the year-to-year 7

138


J Lehmann et al


We showed that CMIP5 models project a robust weakening of EKE and associated storm tracks in boreal summer. Storms bring moist and cool air from the oceans to the continents and thus a weakening of storm tracks could possibly lead to more prolonged heat waves or droughts in the midlatitudes. We will study this possible relation in future research.

Leckebusch G C and Ulbrich U 2004 On the relationship between cyclones and extreme windstorm events over Europe under climate change Glob. Planet. Change 44 181–93 Lindzen R and Farrell B 1980 A simple approximate result for the maximum growth rate of baroclinic instabilities J. Atmos. Sci. 37 1648–54 Mizuta R 2012 Intensification of extratropical cyclones associated with the polar jet change in the CMIP5 global warming projections Geophys. Res. Lett. 39 L19707 Moss R H et al 2010 The next generation of scenarios for climate change research and assessment Nature 463 747–56 Murakami M 1979 Large-scale aspects of deep convective activity over the GATE area Mon. Weather Rev. 107 994–1013 O’Gorman P A 2010 Understanding the varied response of the extratropical storm tracks to climate change Proc. Natl. Acad. Sci. USA 107 19176–80 Petoukhov V, Eliseev A V, Klein R and Oesterle H 2008 On statistics of the free-troposphere synoptic component: an evaluation of skewnesses and mixed third-order moments contribution to the synoptic-scale dynamics and fluxes of heat and humidity Tellus A 60 11–31 Pfahl S and Wernli H 2012 Quantifying the relevance of cyclones for precipitation extremes J. Clim. 25 6770–80 Pinto J G, Fröhlich E L, Leckebusch G C and Ulbrich U 2007 Changing European storm loss potentials under modified climate conditions according to ensemble simulations of the ECHAM5/MPI-OM1 GCM Nat. Hazards Earth Syst. Sci. 7 165–75 Pinto J G, Zacharias S, Fink A H, Leckebusch G C and Ulbrich U 2008 Factors contributing to the development of extreme north atlantic cyclones and their relationship with the NAO Clim. Dyn. 32 711–37 Randel W and Held I 1991 Phase speed spectra of transient eddy fluxes and critical layer absorption J. Atmos. Sci. 48 688–97 Ren X, Yang X and Chu C 2010 Seasonal variations of the synopticscale transient eddy activity and polar front jet over East Asia J. Clim. 23 3222–33 Schwierz C, Köllner-Heck P, Zenklusen Mutter E, Bresch D N, Vidale P-L, Wild M and Schär C 2009 Modelling european winter wind storm losses in current and future climate Clim. Change 101 485–514 Ulbrich U, Pinto J G, Kupfer H, Leckebusch G C, Spangehl T and Reyers M 2008 Changing northern hemisphere storm tracks in an ensemble of IPCC climate change simulations J. Clim. 21 1669–79 Ulbrich U, Leckebusch G C and Pinto J G 2009 Extra-tropical cyclones in the present and future climate: a review Theor. Appl. Climatol. 96 117–31 Woollings T, Gregory J M, Pinto J G, Reyers M and Brayshaw D J 2012 Response of the North Atlantic storm track to climate change shaped by ocean–atmosphere coupling Nat. Geosci. 5 313–7 Wu Y, Ting M, Seager R, Huang H-P and Cane M. a. 2010 Changes in storm tracks and energy transports in a warmer climate simulated by the GFDL CM2.1 model Clim. Dyn. 37 53–72 Yin J H 2005 A consistent poleward shift of the storm tracks in simulations of 21st century climate Geophys. Res. Lett. 32 2–5 Zappa G, Shaffrey L C, Hodges K I, Sansom P G and Stephenson D B 2013 A multi-model assessment of future projections of north atlantic and european extratropical cyclones in the CMIP5 climate models J. Clim. 26 5846–62

Acknowledgments We acknowledge the World Climate Research Programme’s Working Group on Coupled Modelling, which is responsible for CMIP, and thank the climate modeling groups (listed in table S1 in SI) for producing and making available their model output. For CMIP, the US Department of Energy’s Program for Climate Model Diagnosis and Intercomparison provides coordinating support and led development of the software infrastructure in partnership with the Global Organization for Earth System Science Portals. The work was supported by the German Federal Ministry for the Environment, Nature Conservation and Nuclear Safety (11 II 093 Global A SIDS and LDCs), the German Federal Ministry of Education and Research (03IS2191B), the Russian Foundation for Basic Research, and the Programs of the Russian Academy of Sciences (programs of the Presidium RAS and programs by the Department of Earth Sciences RAS).

References Bengtsson L, Hodges K I and Keenlyside N 2009 Will extratropical storms intensify in a warmer climate? J. Clim. 22 2276–301 Blackmon M 1976 A climatological spectral study of the 500 mb geopotential height of the Northern Hemisphere J. Atmos. Sci. 33 1607–23 Catto J L, Shaffrey L C and Hodges K I 2011 Northern hemisphere extratropical cyclones in a warming climate in the HiGEM high resolution climate model J. Clim. 24 5336–52 Chang E K M, Guo Y and Xia X 2012 CMIP5 multimodel ensemble projection of storm track change under global warming J. Geophys. Res. 117 (D23)D23118 Christoph M, Ulbrich U and Haak U 1995 Faster determination of the intraseasonal variability of storm tracks using Murakami’s recursive filter Mon. Weather Rev. 123 578–81 Harvey B J, Shaffrey L C, Woollings T J, Zappa G and Hodges K I 2012 How large are projected 21st century storm track changes? Geophys. Res. Lett. 39 1–5 Harvey B J, Shaffrey L C and Woollings T J 2013 Equator-to-pole temperature differences and the extra-tropical storm track responses of the CMIP5 climate models Clim. Dyn. doi:10.1007/s00382-013-1883-9 Hawcroft M K, Shaffrey L C, Hodges K I and Dacre H F 2012 How much Northern Hemisphere precipitation is associated with extratropical cyclones? Geophys. Res. Lett. 39 L24809 Hoskins B and Valdes P 1990 On the existence of storm-tracks J. Atmos. Sci. 47 1854–64

8


139

Supplementary Information Tables Table S1 CMIP5 models used in this study with their spatial resolution of the atmosphere.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22

Model ACCESS1-0 CMCC-CESM CMCC-CM CMCC-CMS CNRM-CM5 CanESM2 FGOALS-g2 GFDL-CM3 GFDL-ESM2G GFDL-ESM2M HadGEM2-CC IPSL-CM5A-LR IPSL-CM5A-MR IPSL-CM5B-LR MIROC-ESM MIROC-ESM-CHEM MIROC5 MPI-ESM-LR MPI-ESM-MR MRI-CGCM3 NorESM1-M Inmcm4

Spatial resolution 192 x 144 x 8 96 x 48 x 11 480 x 240 x 8 192 x 96 x 11 256 x 128 x 8 128 x 64 x 8 128 x 60 x 8 144 x 90 x 8 144 x 90 x 8 144 x 90 x 8 192 x 140 x 8 96 x 96 x 8 144 x 143 x 8 96 x 96 x 8 128 x 64 x 8 128 x 64 x 8 256 x 128 x 8 192 x 96 x 8 192 x 96 x 15 320 x 160 x 8 144 x 96 x 8 180 x 120 x 8

Additional Figures

Figure S1 Same as Fig. 1, but for EKE at 500 hPa. Differences between EKE changes at different pressure levels and changes from vertically averaged EKE are largest for EKE at 500 hPa, and confined to the NH during DJF.

140

Figure S2 Same as Fig. 2, but for EKE at 250 hPa.






141

142





143

Figure S8 Historical seasonality, i.e. the difference between vertically averaged mid-latitude mean EKE during winter and summer.

Figure S9 Same as Fig. 4 but for the mid-latitude mean vertical shear calculated between 250-850 hPa.

144




Figure S12 Seasonal cycle of vertically averaged mid-latitude mean EKE in the NH (blue) and SH (black) at the end of the 20th century (1981-2000). The solid line depicts the multi-model median and the shaded areas reflect the interquartile range of the model spread.


145

Figure S13 The multi-model mean regression slope from regressing the year-to-year variability of Eady growth rate against the year-to-year variability of EKE during (a) DJF and (b) JJA. The region around the equator has been removed from the figure, because here Eady growth rate is zero and thus artificial spikes in the regression slope arise.

Figure S14 Same as Fig. 5, but for regression between EKE and vertical shear calculated between 250-850 hPa.

146


Figure S15 Same as Fig. 1, but with stippling which indicates regions where more than 90% of the models agree on the sign of EKE change.

Figure S16 Same as Fig. 1, but for five different realizations from the CanESM2 model. Stippling indicates regions where 4 out of 5 realizations show the same sign in EKE change.

3 Discussion and Conclusions

148

3. Discussion and Conclusions

With ongoing climate change and increasing occurrences of certain type of weather extremes, it is of fundamental interest to better understand the role atmospheric circulation changes play for climate variability in general and for weather extremes in particular. An improved understanding of the involved processes could help to reduce prevailing uncertainties in projected dynamical aspects of the atmosphere. This would in turn improve the assessment of climate impacts at the regional scale. In my thesis, these issues are addressed by analyzing extratropical storm track changes and their impacts on surface weather extremes. Identifying changes in extreme weather events is challenging because such events are, by definition, poorly sampled in observation records and because chaotic internal variability of the climate adds a considerable amount of noise to the climate change signal [Shepherd, 2015]. In the first part of my thesis, I show that the frequency of record-breaking daily rainfall events has significantly risen in the global-mean since around 1980 with the upward trend being consistent with the thermodynamic expectation of increasing available moisture in a warming atmosphere. A global increase in observed daily rainfall extremes has also been reported by other studies [e.g., Alexander et al., 2006; Westra et al., 2013]. However, using a statistical framework based on observations only, it is the first study that quantitatively documents that record-breaking rainfall is occurring more frequent than it would in a climate with no long-term climate change. This finding contributes to the growing evidence of anthropogenic-induced intensification of heavy rainfall events [Zhang et al., 2007; Fischer and Knutti, 2015]. Moreover, I detected significant changes in regional rainfall extremes which differ markedly across the globe. Some of these regional results support previous findings [Seneviratne et al., 2012]. However, I also provide evidence for robust changes in regions with so far only low or medium confidence in extreme rainfall changes [Seneviratne et al., 2012] including significant drying in the Mediterranean region and pronounced increases in wet extremes over large parts of the Asian continent. Some of the regional changes, like the decrease of rainfall extremes over western North America or the Mediterranean, cannot be explained by the applied thermodynamic model indicating that dynamic contributions play an important role in driving regional precipitation changes. Both regions mentioned above are at risk of severe droughts [Herring et al., 2015]. This fits the image of ‘dry regions getting drier and wet regions getting wetter’, a simplified paradigm which nevertheless is supported by the latitudinal redistribution of monthly rainfall extremes with increasing rainfall at high latitudes and drying at (sub-) tropical latitudes as shown by model simulations [Zhang et al., 2007]. Moreover, the same pattern of rainfall changes is suggested among relatively dry and wet regions within the tropics [Allan et al., 2010; Liu and Allan, 2013]. This is consistent with results from the second article of this thesis where I report a substantial increase in record-dry months over

149

Central Africa in contrast to significant wetting over tropical monsoon regions including India and South East Asia. The phenomenon of an intensified hydrological cycle has been intensively discussed [Allen and Ingram, 2002; Held and Soden, 2006; Trenberth, 2011] and builds on the idea that under global warming an increased amount of available moisture in the atmosphere can be transported from the ‘sources’ in the subtropical dry regions to the ‘sinks’ in the wet extratropics, while assuming constant atmospheric circulation. However, as shown in the second part of my thesis, atmospheric circulation in terms of extratropical storm activity has significantly changed in the past. Thereby, the observed poleward shift of extratropical storm tracks is consistent with reported wetting of northern high latitudes. Changes in extratropical storm track activity thus might help to explain the rainfall response to climate change in the mid- to high latitudes. Extratropical storm track activity also strongly influences continental temperature extremes including prolonged heat waves and cold spells on monthly timescales as shown in the second part of the thesis. While a connection between fast-traveling synoptic waves and changes in surface weather extremes is in principle not new [e.g., Hawcroft et al., 2012; Mahlstein et al., 2012], a thorough analysis at the global scale covering winter and summer season has never been done in a comprehensive way. Other studies have shown that temperature and precipitation extremes are linked to amplified quasi-stationary waves or to large-scale teleconnection patterns like the (summer) North Atlantic Oscillation (NAO) [e.g., Marshall et al., 2001; Thompson and Wallace, 2001; Folland et al., 2009; Dong et al., 2013]. However, analyzing both planetary waves or (S)NAO is by nature a hemispheric approach and circulation changes are thus difficult to be linked to local weather patterns. With the approach presented in my thesis, changes in storm track activity can be calculated locally and thus be related to regional changes in temperature or precipitation. Moreover, I found good agreements between the different approaches. For example, Horton et al. [Horton et al., 2015] showed that since 1979 high temperature extremes over the US, Europe and western Asia were significantly driven by reported upward trends in the frequency and persistence of summertime anticyclonic circulation patterns over these regions. Anticyclonic flow can act as a blocking high and strongly correlates with low storm activity (Sect. 2.4). Consistently, I found pronounced downward trends in summertime storm activity over the same regions. More importantly, results from section 2.3 indicate a physical mechanism explaining the observed weakening. It is argued that the downward trend in storm activity over recent decades, associated with a weakening of the jet stream, has likely been influenced by a reduced equator-to-pole temperature gradient due to Arctic amplification. A different approach has been developed by a group around Petoukhov [Petoukhov et al., 2013; Coumou et al., 2014]. Based on theoretical

150


considerations they propose a mechanism explaining shifts in extreme weather regimes by quasiresonant amplification of planetary waves in the atmosphere. In simplified terms, they suggest that mid-troposphere planetary waves – with zonal wave numbers six, seven and eight – can get ‘trapped’ and become quasi-stationary with amplified meridional wind speeds favoring weather extremes like droughts, heat waves, heavy precipitation and floods at the flanks of the planetary wave. In conclusion, my thesis underlines the importance of analyzing dynamical changes in the atmosphere because they are able to explain changes in weather extremes in a physically consistent way [Jacobeit et al., 2009]. The question of how surface extremes will change in the future is thus also a question of how extratropical storm tracks will change. Whereas most studies have focused on the more intense winter storms, I have show that strongest changes are expected during summer with a projected weakening of mid-latitude mean storm track activity of -14 % until the end of the 21st century under a high-emission scenario. My findings implicate that this could increase the risk of severe droughts and prolonged heat waves over storm track affected regions like the US and large parts of Europe and western Asia. While providing new findings my thesis also identifies open questions that are worthwhile to address in future research. I have shown that surface weather extremes are linked to extratropical storm activity providing a physically based line of argumentation for the occurrence of weather extremes on monthly timescales. However, despite the fact that it is reasonable to assume that storm track activity influences continental surface temperatures and not the other way around [Lehmann and Coumou, 2015], my findings are limited in making statements about causality in the reported links. For example, I cannot exclude the possibility of a common third driver like jet stream changes which influences both the storm tracks and surfaced weather extremes. One way to address the question of causality would be to test the robustness of my findings using simulations from a climate model where anthropogenic forcings and single effects like Artic amplification can be regulated. Another promising approach is a novel method known as “Causal Effect Networks (CEN)”. CEN distinguishes direct from indirect relationships and has been used in previous studies, for example, to quantify the role of the Atlantic meridional overturning circulation for global-mean temperature variability [Schleussner et al., 2014], to identify regions important for spreading and mediating perturbations in the atmosphere [Runge et al., 2015] or to analyze different Arctic drivers of mid-latitude winter circulation [Kretschmer et al., 2016]. CEN are calculated for a set of time series representing different quantities, so called actors, and quantify the causal relations between them. It would be very interesting to use CEN to test the causal relationship of actors calculated from (1) mid-latitude mean EKE, (2) mid-latitude mean zonal wind velocity at 500 mbar, (3) Arctic

151

sea ice extend, and (4) mid-latitude mean Eady growth rate. This would give further insights into the dominant drivers of extratropical storm track changes and extreme weather events. There is a natural desire to know whether a certain extreme event was caused by anthropogenic climate change or not. This is by nature a challenging task – if not impossible to be answered. Studies addressing this question are generally based on two different approaches. The first approach relates the observed extreme event to a specific weather pattern, like, for example, the severe drought in California which has likely been ‘caused’ by an anomalous persistent ridge located over the eastern North Pacific [Wang et al., 2014]. However, these types of studies rather give an explanation for the observed extreme event than a physical cause in terms of why the ridge occurred and which external factors contributed to the existence and persistence of the ridge. The second approach tries to quantify the influence of an external (anthropogenic) forcing on the likelihood or intensity of an observed extreme event. One realization of this approach consists of simulating two possible climates under the climate conditions of the given extreme event. Thereby, one ensemble of simulations includes all natural and anthropogenic forcings and the other only natural climate forcings. The anthropogenic fingerprint is then calculated from the difference between both simulations [e.g., Min et al., 2011; Bergaoui et al., 2015]. Another possible realization compares the probability to observe an extreme weather event during a pre-industrial control period (𝑃𝑜 ) to the

probability of observing it in simulated projections under, for example, a certain degree of global warming (𝑃1 ). Both probabilities are used to calculate the fraction of attributable risk (𝐹𝐴𝑅 = 1 − (𝑃0 /𝑃1 )). This method has shown to be insightful in addressing thermodynamically driven extreme

events, like, for example, the Russian heat wave in 2010 [Otto et al., 2012], Australia’s record summer temperatures in 2013 [Lewis and Karoly, 2013] or the drought of 2014 in the southern Levant region [Bergaoui et al., 2015]. It should be noted, that FAR may not be an appropriate tool to assess single extreme events that are primarily driven by circulation-related processes including local rainfall changes [Trenberth et al., 2015]. The work presented in my thesis is not a classical attribution study in this sense. Instead, I have used a statistical model, based upon observations, to detect changes in heavy rainfall events that cannot be explained by natural variability. My analysis does not provide a direct cause for the reported changes, but shows that the global upward trend in record-breaking rainfall since 1980 is consistent with what would be expected from the thermodynamic increase in atmospheric moisture under global warming. The applied approach allows to make probabilistic statements for rainfall extremes in considerably large regions and time periods. The advantage of my statistical model is that it is based on observed rainfall records from rain gauges and satellite measurement. Such data sets are continuously being improved through more sophisticated interpolation methods and longer

152


time series. Modeled precipitation data, on the other hand, is attached with large uncertainties because of deficient implementation of circulation processes within the models [IPCC, 2013]. Importantly, both approaches, based on observations or models, tend to give quantitatively similar results when it comes to attribution of general changes in climate extremes in contrast to single events. For example, Fischer and Knutti [2015] have used FAR to analyze how climate change has changed the odds of observing precipitation extreme on global land. They showed that at the present-day warming about one in five daily precipitation extremes over land is attributable to the observed temperature increase. In comparison, I found an increase in record-breaking daily rainfall events of 26 % for the global-mean in 2010 which implies the same ratio of one in five that a record-breaking rainfall event cannot be explained by natural variability. Further, the humaninduced latitudinal redistribution of simulated precipitation changes including decreasing precipitation in the subtropics and increasing precipitation in the northern high latitudes [Zhang et al., 2007] is something also found in observations. Another question that has arisen through my studies is why both low storm track activity and amplified quasi-stationary waves favor surface weather extremes. The reported link between low storm track activity and anomalous large geopotential heights already provides first indications for a possible relationship as discussed above. It should be further investigated how the meandering of the jet stream interacts with extratropical storm activity. CEN analyses could be used to study this interplay. This would require a spatial dimension reduction in the analyzed quantities as shown, for example, by Runge et al. [2015]. However, it remains challenging to capture jet stream meandering with a one-dimensional index, as results are very sensitive to the methodology [Barnes, 2013]. In my thesis, I argue that low storm track activity implies more persistent weather favoring hot or cold extremes on monthly timescales. In a next step, an improved (quantifiable) definition of persistence should be developed and tested to verify this relationship. A possible indicator may consist of some kind of “inertia”-index calculated from the probability transition between warm and cold days which have to be defined in an appropriate way. A low index-value would thus indicate a high probability of hot or cold temperatures to persist over time. My thesis reveals the benefits of using EKE as a regional indicator for storm track activity. Moreover, it has shown to be a suitable tool to assess regional weather extremes on monthly timescales. However, heavy rainfall events leading to severe floodings usually occur on much smaller timescales of hours to days. The signal of such rainfall events likely disappears in monthlymean or monthly-total rainfall aggregates and thus cannot be assessed by monthly EKE. Here, algorithms for tracking individual cyclone properties on daily timescales are preferable methods [Neu et al., 2013] and could be compared and, at best, combined with the analysis of EKE.

153

Further, EKE measures the interplay between the frequency and intensity of high and low pressure systems. Thus, an increase in EKE could be caused by an intensification of individual cyclones or an increase in the frequency of cyclones or both. Likewise, a decrease in frequency of moderate cyclones together with an increase in intensity of the strongest cyclones as projected by CMIP5 models for the 21st century under a high-emission scenario in winter might cancel each other out in the calculation of monthly EKE and thus lead to a zero trend as shown in section 2.5. Here, further analysis is needed to separate both affects. Cyclone tracking algorithms (as described above) should be combined with EKE in order to tackle this issue and broaden the applicable field of research questions regarding the link to climate extreme events. The climate is changing and “the abnormal is the new normal” as stated by the UN SecretaryGeneral Ban Ki-moon [2016]. In this thesis, I present a comprehensive analysis of extratropical storm track changes in both observations and climate simulations. It is shown that these changes are strongly coupled to changes in surface weather extremes. Thereby, I provide insights into the physical mechanisms (dynamic and thermodynamic) involved in driving climate variability in general and surface weather extremes in particular. Moreover, I show that changes in the hydrological cycle in response to climate change have had strong effects on the intensity and occurrence of record-breaking wet and dry extremes. Two useful frameworks are provided: (i) a band pass filter to quickly extract a measure of extratropical storm track activity on a longitudelatitude grid and (ii) a statistical method which is able to analyze record-breaking events in observations with inhomogeneous spatial and temporal data coverage. I hope that my thesis helps to initiate further research leading to a better understanding of the complex system of atmospheric circulation changes and its implications for daily weather in the mid-latitudes.

Appendix

A1. Weakened Flow, Persistent Circulation and Prolonged Extreme Weather Events in Boreal Summer Dim Coumou, Kai Kornhuber, Jascha Lehmann, and Vladimir Petoukhov. In this book chapter the link between historic circulation changes and the occurrence of long-lasting heat waves in the Northern Hemisphere summer is addressed. It is argued that the reduction in amplitude of fast-moving transient waves and the more-frequent occurrence of high-amplitude quasi-stationary waves both favor more persistent weather conditions and thus prolonged heat waves in summer. Published as peer reviewed book chapter in ’Patterns of Climate Extremes: Trends and Mechanisms’ American Geophysical Union (2015).

A1. Weakened Flow, Persistent Circulation and Prolonged Extreme Weather Events in Boreal 159 Summer

Weakened Flow, Persistent Circulation and Prolonged Extreme Weather Events in Boreal Summer

Dim Coumoua, 1, Kai Kornhubera,b, Jascha Lehmanna,b, Vladimir Petoukhova

a

Potsdam Institute for Climate Impact Research, D-14412 Potsdam, Germany b

1

University of Potsdam, Faculty of Science

To whom correspondence may be addressed. E-mail: [email protected]

Abstract Changes in atmospheric circulation can strongly alter the frequency and/or magnitude of high-impact extreme weather events. Here we address the link between circulation changes and the occurrence of long-lasting heat waves in the Northern Hemisphere summer. We show that boreal summer circulation has seen pronounced changes in circulation over the last decades, possibly related to rapid warming of the Arctic. Generally, the mid-latitude zonal mean flow has weakened and also the kinetic energy associated with transient synoptic eddies has reduced. At the same time, for some wave numbers, we see an increased occurrence-frequency of high-amplitude quasi-stationary waves. We argue that this increase in frequency is associated with a recent cluster of resonance events which can create such highamplitude waves. The reduction in amplitude of fast-moving transient waves and the more-frequent occurrence of high-amplitude quasi-stationary waves both favor more persistent weather conditions. We present statistical evidence of links between such persistent upper level flow and the occurrence of heat extremes at the surface.

160

Appendix

Detected changes in mid-latitude summer circulation The Northern Hemisphere mid-latitudes have seen significant changes in the large-scale summer circulation over the last decades (Overland et al 2012, Francis and Vavrus 2015, Coumou et al 2015). The zonal mean zonal wind (or “jet”) has weakened by about 5% (see Figure 1), a signal which is seen in different reanalysis products and at different pressure levels (Coumou et al 2015). The driver behind this weakening is likely to be the much faster warming in the Arctic as compared to the rest of the Hemisphere, a phenomenon known as Arctic Amplification (Figure 2). This reduces the pole-to-equator temperature gradient and thereby the thermally driven jet (top panel Figure 1). However, a strict direction of causality has not been shown: In principle it could be the other way around, too, whereby a change in mid-latitude circulation alters the poleward heat transport which would give rise to more rapid warming in the Arctic (Cohen et al 2014, Walsh 2014). For example, an alternative hypothesis has been proposed where the warmer Arctic is a consequence of mid-latitude circulation changes which would increase the exchanges of cold air from the Arctic with warmer air from lower latitudes, triggered by teleconnections originating from the tropical Pacific (Trenberth et al 2014, Palmer 2014). Irrespective of the underlying drivers, changes in mid-latitude circulation are expected to affect day-to-day weather variability and possibly extreme weather events. In conjunction with the summer jet, the kinetic energy associated with transient synoptic-scale weather systems has seen a significant weakening as well (Coumou et al 2015). This kinetic energy can for example be extracted by bandpass filtering daily wind field data, because synoptic activity has typical timescales of 2 to 5 days, bringing weather variability on sub-weekly timescales. Typical wave lengths of synoptic transients are of the order of 2000-3500km, or wave numbers from 6 to 10, with relatively fast phase speeds (i.e. eastward propagation) of the order of 6-12m/s. These waves are free waves in the sense that they do not require any forcing but are a direct product of the atmospheric instability in the mid-latitudes. Thus, in the absence of any temperature or orographic forcing, as for example on a so called “aquaplanet”, these waves would still emerge (Schneider et al 2014). Based on linear Rossby wave theory, one would expect the phase speed of these waves to become smaller under a weakened back ground flow. Starting from the linearized non-divergent barotropic vorticity equation (Pedlosky 1979) without any thermal or orographic forcing (i.e., an equation describing adiabatic free atmospheric waves), an equation for the phase speed c can be derived: (1)


where U denotes the zonal mean zonal wind, β the Rossby parameter and k and l respectively the zonal and meridional wave numbers. Thus, these free waves essentially flow on top of the background flow (U) and a weakening of the latter results in slower phase speeds. However, so far, no robust changes in phase speed have been detected in the observations. In contrast, significant trends have been detected in the amplitude of these waves, defined as the magnitude of the North-South flow (meridional wind in units of m/s). Amplitudes are expected to decline when the jet weakens, which can be derived from theoretical considerations (Coumou et al 2011). This behavior is also seen in numerical experiments of both idealized general circulation models (GCMs) and comprehensive GCMs (Schneider et al 2014). In general, transient eddies are forced by the jet via vertical shear (and thus baroclinicity) but can also accelerate it via the eddy-driven jet (Woollings and Blackburn 2012, Cohen et al 2014, Lehmann et al 2014). Thus, a weakening of the jet is expected to weaken synoptic eddy activity, in agreement with the observed long-term trends. As shown in figure 1, statistically significant reductions are seen in boththe mean amplitude of wave numbers 6-10 as well as in the total kinetic energy associated with synoptic eddies (“Eddy Kinetic Energy” or EKE). This summertime weakening of both the jet and of EKE is also a very robust signal in future projections of CMIP5 climate models (O’Gorman 2010, Lehmann et al 2014). In addition to free waves, orography and thermal anomalies (like land-ocean temperature contrasts) generate forced waves in the mid-latitudes. Since the forcing patterns change on substantially longer timescales of several weeks, these waves can be considered quasi-stationary. Due to the large-scale nature of the forcing patterns, forced waves tend to have smaller wave numbers (< 6). The forcedcomponent (and therefore also the quasi-stationary component) of wave numbers 6-10 is normally small. Note that free waves can also be quasi-stationary when the second term in equation 1 is similar in magnitude as the zonal mean jet (U), which is normally the case for wave 4. Thus, in general, the low wave number regime of the flow is quite different from the high wave number regime, with the latter dominated by fast moving waves. Only when averaging over longer time-spans, like 15 days to monthly means, one effectively removes this transient wave component and the quasi-stationary component remains. For waves 6-10, this quasi-stationary component has generally a low magnitude and also little significant trends can be detected. Nevertheless, the trend in the highest amplitudes of the quasistationary component of wave 7 is upward. The lowest panel of Figure 1 plots how often the 70th percentile wave-amplitude in 15-day running mean wind field data is exceeded, showing a moderate upward trend. This is also consistent with the reported increase in frequency of high-amplitude, quasistationary wave 7 (Coumou et al 2014). Though this signal is seen in different reanalyses and at different

162

Appendix

pressure levels (Coumou et al 2014), it should be noted that statistically it is only marginally significant. Nevertheless, it is consistent with the recent cluster of resonance events involving wave 7 as discussed later. In summary we can conclude that boreal summer circulation has changed in three important ways over the last decades: 1) A weakening of the zonal mean jet, 2) a reduction in fast-moving synoptic wave activity and 3) a (marginally significant) increase in frequency of the occurrence of high-amplitude quasistationary wave 7. These three detectable changes all point at more persistent circulation patterns. Mid-latitude Rossby waves and extreme weather In general, persistent circulation patterns can lead to extreme weather and in particular to prolonged heat waves in summer. Slow wave propagation prolongs the prevailing weather conditions at the surface and can therefore lead to extremes on timescales of weeks: One day with temperatures over 30oC in Western Europe is not unusual, but 10 or 20 days in a row will be. It has been demonstrated that high-amplitude quasi-stationary waves in the atmosphere are statistically linked to extreme weather at the surface (Screen and Simmonds 2014, Coumou et al 2014). Especially regions at the western boundary of the continents like the western U.S. or Europe show the strongest association between surface extremes and upper-level waves. Here, moderate temperatures tend to be associated with reduced quasi-stationary wave amplitudes and extreme temperatures with amplified quasi-stationary waves (Screen and Simmonds 2014). In contrast, strong transient wave activity, as captured by EKE, is linked to moderate surface temperatures and vice versa (Coumou et al 2015). Over most continental regions affected by storm tracks, there is a significant negative correlation between monthly EKE and surface temperature (Figure 3). Thus, the hottest summers are associated with extremely low EKE, while mild summers are associated with more pronounced EKE. Again the western boundaries of the continents, notably Europe, are especially sensitive but also the eastern US and western Asia. This makes sense as these regions are most directly influenced by the storm tracks. If EKE is low, this implies that weather variability on 2-5 day timescales is low, which is consistent with blocking anticyclones associated with quasi-stationary waves. In summer, transient waves transport moist and cool air from the oceans to the continents bringing relief during periods of oppressive heat (Kyselý and Huth 2005, Kyselý 2008). Thus, low transient wave activity implies that cool maritime air masses become less frequent creating favorable conditions for blocking anti-cyclonic flow over continents and the buildup of heat and drought conditions. Amplifying


local feedback mechanisms, such as soil-moisture feedbacks, can further drive temperatures upward, inhibit cloud formation and thereby enhance the anticyclonic flow over the continents (Schär et al 2004, Alexander 2010, Mueller and Seneviratne 2012). Thus, an absence of substantial transient wave activity likely prolongs the duration of blocking weather systems, as for example during the Russian heat wave of 2010. The record-breaking July temperatures over Moscow in that year were associated with extremely low EKE (Coumou et al 2015). In summary, the observed trends in boreal summer circulation (Figure 1) favor more persistent weather in summer and hence heat extremes on timescales of several weeks. Resonant Circulation Regimes The observed reduction in transient wave activity seems to be a direct response of the weakening jet. Climate models project both summer jet and EKE to weaken under future high-emission scenarios and the projected relative changes in these quantities are consistent with those observed (Coumou et al 2015). However, the observed increase in frequency of high-amplitude quasi-stationary wave 7 cannot be directly linked to the weakening jet. To explain this increased frequency, a highly non-linear, dynamical mechanism is invoked: Resonance. This mechanism, which was first proposed by Petoukhov et al (2013), can amplify the quasi-stationary component of waves 6, 7 or 8. As stated above, these wave numbers normally have a fast eastward propagation and their quasi-stationary component tends to be weak. Forcing of these waves is normally weak in the mid-latitudes and their energy is effectively dispersed towards the poles and equator. However, under specific conditions, their wave energy can be trapped in mid-latitudinal waveguides, with only weak dispersion at the lateral boundaries. If this happens and if the forcing for the trapped wave is large enough, then resonance can strongly magnify the wave amplitude. There are thus two important criteria which need to be fulfilled for resonance to occur: 1) waveguide formation which traps the energy of wave 6, 7 or 8 in the mid-latitudes and 2) reasonably large forcing of the trapped free wave. Waveguide The jet stream can guide waves in the zonal direction and thereby trap them inside the mid-latitudes. Whether a particular wave is trapped depends on the shape of the zonal mean zonal wind (U) and the wave number itself (k). Based on the linearized stationary barotropic vorticity equation at the equivalent

164

Appendix

barotropic level (in the mid/upper troposphere at about 300-500hPa), an expression for the square of the meridional wavenumber of the trapped wave (𝑙 2 ) can be derived (Petoukhov et al 2013): 𝑙 2 (𝑈, 𝑘, 𝜑) =

2Ω 𝑐𝑜𝑠 3 𝜑 𝑎𝑈

−

𝑐𝑜𝑠 2 𝜑 𝑑2 𝑈 𝑎 2𝑈

𝑑𝜑 2

+

𝑠𝑖𝑛𝜑 𝑐𝑜𝑠𝜑 𝑑𝑈 𝑎2 𝑈

1

𝑘 2

+ 𝑎 2 − �𝑎 � 𝑑𝜑

(2)

Here 𝑎 is the Earth’s radius, Ω its angular velocity and 𝜑 latitude. Note that 𝑙 may be a real or purely

imaginary number and thus 𝑙 2 can change sign. For a waveguide to form for wave number k, two turning points are needed at two sufficiently distant latitudinal locations. At these turning points, the

energy of wave k gets reflected back towards the mid-latitudes. For this to occur, 𝑙 2 has to be positive

inside the waveguide and change sign at the turning points themselves while U has to be positive (i.e. eastward) inside and directly outside the waveguide. The latter is normally the case outside the tropics and thus it is especially the change in sign of 𝑙 2 which is important. Finally, the turning points have to be

located at a sufficiently large distance (several degrees latitude) so that the waveguide exceeds the characteristic scale of the relevant Airy function, which governs the latitudinal boundary conditions of the waveguide (Petoukhov et al 2013). Wave forcing Trapping of a quasi-stationary free wave in a mid-latitude waveguide not necessary leads to amplification of that wave: Forcing is needed as well. When a waveguide exists for a synoptic wave with wavenumber 𝑘 , the following approximat equation for the amplitude of the forced quasi-stationary

component is valid (Petoukhov et al 2013): 𝐴𝑚 =

𝐴𝑚,𝐹

2

�[(𝑘/𝑎)2−(𝑚/𝑎)2]2 +�𝐿/𝑎 2+𝑅02/𝐿� (𝑚/𝑎)2

(3)

Here R0 is the Rossby number for eddies and L is the characteristic Rossby radius which are both taken as constants. Am,F is the amplitude of the forcing at the equivalent barotropic level (in units of m/s) for wavenumber m, and is a function of temperature and orography. Equation 3 shows that when k is close to m (i.e. when the forcing pattern is close to that of the trapped free wave), the denominator becomes small and thus Am large. This is the essence of resonant amplification. Note that the forcing Am,F can thus be moderate (although not zero) for resonance and hence high wave amplitudes to occur.


Resonance Occurrence Based upon the above described criteria, resonance events in the Northern Hemisphere have been identified to occur primarily in summer and also spring when the background flow is relatively weak. In these seasons, so called double jet flow configurations, characterized by two peaks in the zonal mean zonal wind, are much more likely to occur. Such double jet states favor waveguide formation due to the unusual shape of the zonal mean zonal wind across latitudes which primarily determines 𝑙 2 . From the

thermal wind equation it follows that in order to generate double jet states, sharp poleward temperature gradients are needed at mid (~45°N) and sub-polar (~70°N) latitudes with weaker gradients in between. Weak thermal gradients around 50°N-65°N are unlikely to occur in winter and autumn when the equator to pole temperature difference is much larger. This is likely the underlying reason that double jet states and hence resonance events are limited to summer and spring, when the overall equator to pole temperature drop is much more moderate. Likewise, the cluster of resonance events in recent summers (see section “Cluster of resonance events”) might thus be related to the significant reduction in the meridional temperature gradient around 50oN over the last decades (Figure 1). Case study: Summer 2010 The summer of 2010 saw anomalous large-scale flow patterns over the Eurasian continent which has been identified as a key factor behind both the Russian heatwave and Pakistan flooding that summer (Schubert et al 2011, Schneidereit et al 2012, Lau and Kim 2012, Tachibana et al 2010, Petoukhov et al 2013, Schubert et al 2014). These extreme events illustrated how destructive extreme weather can be to societies: The death toll in Moscow has been estimated at 11,000 and drought caused grain-harvest losses of 30%, leading to a Russian export ban on wheat (Coumou and Rahmstorf 2012). At the same time Pakistan was hit by the worst flooding in its history, which affected approximately one-fifth of its total land area and 20 million people (Hong et al 2011, Coumou and Rahmstorf 2012). The heatwave over western Russia was due to a blocking high which resulted in record temperatures exceeding the extreme European heat wave of 2003 in amplitude and spatial extent (Dole et al 2011, Rahmstorf and Coumou 2011, Barriopedro et al 2011). In fact, based upon a global heat wave index, the event was quantified as the strongest heat wave ever (Russo et al 2014). Especially striking was the long lifetime of the blocking event: Roughly 30% of summer days in western Russia were considered as blocking days where normally this is only about 10% (Schneidereit et al 2012). The anticyclonic flow anomaly was associated with a stationary Rossby wave train extending over the full Eurasian continent and beyond. In fact, several studies have highlighted the hemispheric nature of these flow anomalies (Tachibana et al

166

Appendix

2010, Schubert et al 2011, Petoukhov et al 2013). The trough downstream of the western Russia anticyclone was instrumental in triggering anomalously heavy rainfall over northern Pakistan leading to flooding of the Indus River (Lau and Kim 2012). Figure 4 shows Hovmöller diagrams of a set of zonal mean dynamic variables during the summer of 2010 over the Northern Hemisphere. In early July, a pronounced second peak in the zonal mean zonal wind emerges at about 70oN creating a clear double jet flow configuration which persists till half of August (Figure 4a). This shift in the zonal mean flow regime is associated with a pronounced warming of regions around 60oN, i.e. the approximate latitude of Moscow. In fact, in the single jet regime (June and late August), pronounced positive temperature anomalies are seen over the Arctic Ocean (North of ~70oN) and further south (between 40o-50oN), while in the double jet regime they are concentrated around 60oN (Figure 4b). The zonal mean flow changes result in the formation of a waveguide by mid-July which becomes pronounced in late July and early August (Figure 4c). Two turning points can be distinguished in Figure 4c: one at 40oN and one at about 50oN (solid black lines), i.e. placed roughly in between the latitudes of the heat wave in Moscow and the flooding in Pakistan. Following the criteria defined above, 𝑙 2 is positive between those latitudes (red) and changes sign at the turning points themselves. Further,

the zonal mean flow is clearly positive over all latitudes north of about 30oN including the position of the

waveguide. In addition, the turning points are located at sufficiently large distance and thus all criteria for a waveguide trapping waves with wavenumbers close to 6 in the mid-latitudes were fulfilled. Since also the forcing AF for wave 6 was sufficiently large (see Eq.3), the trapped wave underwent resonant amplification (Figure 4d). This amplification of wave 6 is also observed: It starts by mid-July when the waveguide is first detected and maximum amplitudes are reached in early August (red in Figure 4e). This panel also shows that in mid-June similarly high wave amplitudes are observed which were not related to the resonance mechanism, highlighting that of course other mechanisms can also cause highamplitude waves. This analysis, and more detailed analyses described in (Petoukhov et al 2013), illustrate that resonance played an important role in creating high-amplitude quasi-stationary Rossby waves from mid-July to early August 2010, which was the most severe period of the Russian heat wave and which includes the period of flooding in Pakistan.


Cluster of resonance events Based upon the criteria discussed above and analyzing monthly data for July and August, (Petoukhov et al 2013) identified 19 resonance months since 1980 (Figure 5). Many of these months were associated with memorable extreme weather events, including severe heatwaves and flooding events in Europe and heatwaves and droughts in the United States. Coinciding with the onset of rapid Arctic Amplification in 2000, there appears to be a cluster of resonance events though its significance cannot be assessed. Nevertheless, in the post-2000 period, the frequency of July or August resonance months has almost doubled compared to the pre-2000 period. Especially, resonance involving waves 7 and 8 seems to have occurred more frequently recently. Spectral analysis of daily data of July and August days, (Coumou et al 2014) showed that there indeed have been significant changes with more frequent high-amplitude quasi-stationary waves with these wavenumbers. In the post-2000 period, the number of days with a quasi-stationary wave 7 with amplitudes larger than 3 m/s increased by 30%, and, with amplitudes larger than 5 m/s, have more than doubled (Coumou et al 2014). These findings are also consistent with the independent analysis as plotted in the lowest panel of Figure 1. This shows a moderate upward trend in the occurrence-frequency of a quasi-stationary wave 7 with an amplitude exceeding the 70th percentile. This threshold is somewhat subjective and the upward trend is only marginally significant and thus this finding should be treated with caution. It nevertheless is consistent with the cluster analysis presented in Coumou et al (2014). We thus argue that the recent cluster of resonance events has led to a detectable increase in the overall frequency of high-amplitude quasi-stationary wave 7 since 2000. This occurred while transient wave activity, best captured by EKE, declined pronouncedly.

Statistics of resonance events Based on the resonance periods identified by Petoukhov et al. (2013), Coumou et al. (2014) performed statistical analysis to quantify how anomalous upper level wind fields and surface weather were during resonance events. To determine upper level wave activity, spectral analysis of the 500mb and 300mb daily meridional wind field over a mid-latitudinal belt stretching from 35oN to 65oN were performed using both the ERA-Interim (Dee et al 2011) and NCEP-NCAR (Kalnay et al 1996) reanalyses. Since the meridional wind field was analyzed, wave amplitudes reflect actual North-South wind speeds. Further, also phase-speed and the daily zonal mean zonal wind in this mid-latitudinal belt were extracted (Coumou et al 2014). Figure 6 summarizes the key findings of these analyses presented as 2D probability density plots of phase-speed with respect to wave amplitude (top panels) and zonal mean wind (bottom

168

Appendix

panels) for waves 6 (left panels), 7 (middle panels) and 8 (right panels). The July-August climatological mean (solid curves) shows that the wave spectrum is dominated by eastward traveling waves (positive phase speed) with the speed increasing with wave number, as predicted by linear Rossby wave theory (Eq. 1). Still, especially for waves 6 and 7, a sizeable proportion of the probability density distribution is associated with quasi-stationary waves (with an absolute phase speed c less than ∼2m/s) or with

westward propagating waves. However, these slow moving waves generally have low amplitudes. Color

coding in figure 6 shows the anomaly in the distributions during resonance periods, showing distinct patterns. First of all, a pronounced increase in occurrence-frequency of high-amplitude quasi-stationary waves is observed, notably for waves 6 and 7. Next, the occurrence-frequency of fast-moving waves (i.e., faster than the July−August climatological mean phase speed) is substantially reduced (blue), again most pronounced for waves 6 and 7. Further, Coumou et al (2014) report that the mean phase speed during resonance periods reduces by more than a factor of two for wave 7 and by more than a factor of three for wave 6. Also distinct patterns are observed in the anomaly for the zonal-mean zonal wind spectra (Figure 6, bottom panels). Generally, a weakening of the zonal flow is observed, which is most pronounced for wave 7 (by roughly 5% in the mean). This weakening can thus explain some of the observed slowdown in wave propagation (following Eq. 1) but the change is too small to fully account for the strong reduction in phase speed observed. Moreover, also the high-amplitude character of the quasi-stationary waves requires an amplification mechanism, not just a slowdown. In a next step, it is analyzed whether surface weather conditions were also more extreme during these resonance months. To quantify this a simple mid-latitudinal extreme (MEX) index is defined (Coumou et al 2014):

(4) o

Here x can in principle refer to any meteorological variable defined in the mid-latitudes (35 N-65oN) but here we use near-surface temperature, with Δxi(t) its anomaly at time step t and σ(xi) its standard deviation at location i. The MEX index is normalized by subtracting its time-averaged mean (μMEX) and division by its standard deviation (σMEX) such that the climatological distribution centers around 0 and is defined in units of standard deviation (Coumou et al 2014). If the MEX index has large positive values this indicates that extreme temperatures (hot or cold) occur simultaneously in many locations throughout the mid-latitudes. Combining hot and cold anomalies (by means of the square in Eq. 4)


makes sense as strong wave activity is expected to bring hot conditions in some regions and cold conditions in others (Screen and Simmonds 2014). Figure 7 shows probability density distributions for the MEX index based on daily and monthly mean surface temperatures. The July/August climatological distributions (black) are clearly asymmetric with a fat extreme tail and center around 0 (by definition). The daily heat index during resonance periods shows a statistically significant shift (p-value < 0.05) to more extremes with MEX indices beyond 2 standard deviations becoming much more likely. Moreover, the distribution of the monthly MEX index shows even more pronounced changes during resonance, with a significant shift (p-value < 0.05) towards more extremes. The very pronounced changes seen in monthly data, as compared to more moderate changes in daily data, indicates that it was especially the persistence in weather that led to extreme heat on longer, i.e. monthly, timescales. These statistical analyses thus reveal that, in support of the theoretical insights described above, resonance periods are characterized by (1) high-amplitude, slowly propagating waves and (2) persistent, and therefore extreme, surface weather. Conclusions & Discussions Boreal summer circulation has seen pronounced changes over the last decades, trends which seem to have amplified since the onset of rapid Arctic Amplification around 2000. Especially the reduction in EKE, but also in zonal mean flow, have created conditions favorable for the buildup of heat and drought conditions over the continents. Moreover, a cluster of resonance events is observed since 2000, which has increased the occurrence-frequency of high-amplitude quasi-stationary waves with wavenumbers close to 7. This generally implies a weakening of transient synoptic eddy activity and more-frequent states of quasi-stationary flow. The intimate links between quasi-stationary forced waves and transient synoptic eddies (Chang et al 2002) are not addressed here, but it seems reasonable to assume that a reduction in fast-wave activity will favor the occurrence of quasi-stationary flow. In fact, Horton et al. report significant increasing trends in anticyclonic summer circulation over eastern US, Europe and western Asia since 1979 (Horton et al 2015). Blocking high-pressure anticyclones are occurring more often and also persist longer in these regions (Horton et al 2015). As stated above, these regions are directly influenced by the North Atlantic storm track and thus weakening EKE is likely to favor anticyclonic flow here. In any case, these different detected changes in observed large-scale summer circulation strongly point towards more-persistent flow patterns and therefore more extreme surface

170

Appendix

weather. This is also consistent with the dramatic increase in heat extremes in Europe and some other mid-latitude regions (Coumou et al 2013, Coumou and Robinson 2013, Russo et al 2014, Christidis et al 2014). Also climate models project these trends in boreal summer circulation to continue under future high-emission scenarios (Lehmann et al 2014). The underlying drivers behind these circulation changes are currently less well understood though they are generally consistent with theories and numerical model experiments involving Arctic change (Francis and Vavrus 2012, Cvijanovic and Caldeira 2015, Schneider et al 2014, Petrie et al 2015). The classical way to attribute drivers behind observed changes involves numerical modeling experiments with either in- or excluding certain forcing components (Screen et al 2013, Trenberth et al 2014, Cvijanovic and Caldeira 2015). Though certainly a lot can be learned this way, it has been shown that current GCMs appear deficient in simulating near stationary Rossby wave patterns during summer (Schubert et al 2011, Overland et al 2012). A supplementary approach would be to apply graph-theoretical statistical models (Runge et al 2012, 2014) to disentangle the contributions from different drivers in observational and climate model data. Such methods can detect and quantify causal interactions and they have successfully been applied to disentangle causality in ocean circulation (Schleussner et al 2014) and atmospheric circulation in the tropics (Runge et al 2014). Most recent studies on trends in persistent circulation and extreme events have focused on boreal winter circulation (Cohen et al 2014, Walsh 2014). In this season, recent Arctic amplification has been strongest (Cohen et al 2014) and also possible tropical drivers are thought to be much more important (Trenberth et al 2014). Irrespective of these possible far-away drivers, the equator-to-pole temperature gradient in the mid-latitudes has seen the most pronounced changes in boreal summer (Coumou et al 2015). Further, due to its smaller year-to-year variability as compared to the other seasons, long-term trends in summer circulation are detectable at an earlier stage. But apart from that, the important physical mechanisms are likely to be quite different between warm and cold season. For example, in winter the variability in the polar night jet and sudden stratospheric warming events are thought to be important in forcing tropospheric circulation (Baldwin and Dunkerton 2001, Kim et al 2014). In summer such stratospheric interactions are unlikely to play a role. Clearly more research is needed to understand the differences between cold- and warm-season circulations including the drivers behind their variability.Much can be learned by analyzing possible resonance events in non-summer seasons and also in the Southern Hemisphere. Due to its almost opposite nature in terms of land-ocean distribution, the Southern Hemisphere can be used to test different theories. For example, in boreal summer double jet


circulation patterns occur relatively frequently whereby the poleward jet forms around 70oN, i.e. roughly at the boundary of the Arctic Ocean where a strong thermal contrast forms. Therefore, double jet patterns are somewhat unlikely in the Southern Hemisphere (although this remains to be tested) and this could thus provide insights in the relative role of double jets in triggering resonance events if the latter appear to exist in the Southern Hemisphere. Also long-term reanalysis, like the 20th century reanalysis (Compo et al 2011), and climate models from the Coupled Climate Model Intercomparison Project (Covey et al 2003) should be analyzed for resonance events in both hemispheres. Key questions to address are whether climate models can actually reproduce such non-linear wave dynamics and what the links are with the typical modes of variability like the Arctic Oscillation or ENSO. Atmosphere dynamical changes associated with long-term climate change are a key contributor to uncertainty in future climate model projections (Shepherd 2014, Bony et al 2015). Here we have highlighted the importance of atmospheric circulation in causing devastating extreme surface weather events throughout the Northern Hemisphere mid-latitudes. To reduce uncertainty in climate model projections and assess future impacts from extreme weather, the underlying processes and drivers will need to be understood in much more detail. Acknowledgments: JL was supported by the German Research Foundation (DFG, CO994/2-1) and DC and KK were supported by the German Federal Ministry of Education and Research (BMBF, 01LN1304A).

172

Appendix

References Alexander L 2010 Extreme heat rooted in dry soils Nat. Publ. Gr. 4 12–3 Online: http://dx.doi.org/10.1038/ngeo1045 Baldwin M P and Dunkerton T J 2001 Stratospheric harbingers of anomalous weather regimes. Science 294 581–4 Online: http://www.ncbi.nlm.nih.gov/pubmed/11641495 Barriopedro D, Fischer E M, Luterbacher J, Trigo R M and García-Herrera R 2011 The Hot Summer of 2010: Redrawing the Temperature Record Map of Europe Science (80-. ). 332 220–4 Bony S, Stevens B, Frierson D M W, Jakob C, Kageyama M, Pincus R, Shepherd T G, Sherwood S C, Siebesma a. P, Sobel A H, Watanabe M and Webb M J 2015 Clouds, circulation and climate sensitivity Nat. Geosci. 8 261–8 Online: http://www.nature.com/doifinder/10.1038/ngeo2398 Chang E K M, Lee S and Swanson K 2002 Storm Track Dynamics J. Clim. 15 2163–83 Christidis N, Jones G S and Stott P a. 2014 Dramatically increasing chance of extremely hot summers since the 2003 European heatwave Nat. Clim. Chang. 3–7 Online: http://www.nature.com/doifinder/10.1038/nclimate2468 Cohen J, Screen J a., Furtado J C, Barlow M, Whittleston D, Coumou D, Francis J, Dethloff K, Entekhabi D, Overland J and Jones J 2014 Recent Arctic amplification and extreme mid-latitude weather Nat. Geosci. 7 627–37 Compo G P, Whitaker J S, Sardeshmukh P D, Matsui N, Allan R J, Yin X, Gleason B E, Vose R S, Rutledge G, Bessemoulin P, Brönnimann S, Brunet M, Crouthamel R I, Grant a. N, Groisman P Y, Jones P D, Kruk M C, Kruger a. C, Marshall G J, Maugeri M, Mok H Y, Nordli Ø, Ross T F, Trigo R M, Wang X L, Woodruff S D and Worley S J 2011 The Twentieth Century Reanalysis Project Q. J. R. Meteorol. Soc. 137 1–28 Online: http://doi.wiley.com/10.1002/qj.776 Coumou D, Lehmann J and Beckmann J 2015 The weakening summer circulation in the Northern Hemisphere mid-latitudes Science (80-. ). 348 324–7 Coumou D, Petoukhov V and Eliseev A V. 2011 Three-dimensional parameterizations of the synoptic scale kinetic energy and momentum flux in the Earth’s atmosphere Nonlinear Process. Geophys. 18 807–27 Online: http://www.nonlin-processes-geophys.net/18/807/2011/ Coumou D, Petoukhov V, Rahmstorf S, Petri S and Schellnhuber H J 2014 Quasi-resonant circulation regimes and hemispheric synchronization of extreme weather in boreal summer Proc. Natl. Acad. Sci. U. S. A. 111 12331–6 Coumou D and Rahmstorf S 2012 A decade of weather extremes Nat. Clim. Chang. 2 491–6 Coumou D and Robinson A 2013 Historic and future increase in the global land area affected by monthly heat extremes Environ. Res. Lett. 8 0–6


Coumou D, Robinson A and Rahmstorf S 2013 Global increase in record-breaking monthly-mean temperatures Clim. Chang. 118 771–82 Covey C, Achutarao K M, Cubasch U, Jones P, Lambert S J, Mann M E, Phillips T J and Taylor K E 2003 An overview of results from the Coupled Model Intercomparison Project 37 103–33 Cowtan K and Way R G 2013 Coverage bias in the HadCRUT4 temperature series and its impact on recent temperature trends Q. J. R. Meteorol. Soc. 133 459–77 Online: http://doi.wiley.com/10.1002/qj.22 Cvijanovic I and Caldeira K 2015 Atmospheric impacts of sea ice decline in CO2 induced global warming Clim. Dyn. Online: http://link.springer.com/10.1007/s00382-015-2489-1 Dee D P, Uppala S M, Simmons A J, Berrisford P, Poli P, Kobayashi S, Andrae U, Balmaseda M A, Balsamo G, Bauer P, Bechtold P, Beljaars A C M, Berg L van de, Bidlot J, Bormann N, Delsol C, Dragani R, Fuentes M, Geer A J, Haimberger L, Healy S B, Hersbach H, Hólm E V., Isaksen L, Kållberg P, Köhler M, Matricardi M, McNally A P, Monge-Sanz B M, Morcrette J-J, Park B-K, Peubey C, Rosnay P de, Tavolato C, Thépaut J-N and Vitart F 2011 The ERA-Interim reanalysis: configuration and performance of the data assimilation system Q. J. R. Meteorol. Soc. 137 553–97 Dole R, Hoerling M, Perlwitz J, Eischeid J, Pegion P, Zhang T, Quan X W, Xu T and Murray D 2011 Was there a basis for anticipating the 2010 Russian heat wave? Geophys. Res. Lett. 38 Francis J a and Vavrus S J 2015 Evidence for a wavier jet stream in response to rapid Arctic warming Environ. Res. Lett. 10 Online: http://stacks.iop.org/17489326/10/i=1/a=014005?key=crossref.74581076f734b2377ec8042d3aebe25d Francis J A and Vavrus S J 2012 Evidence linking Arctic amplification to extreme weather in the midlatitudes Geophys. Res. Lett. 39 L06801 Hong C, Hsu H, Lin N and Chiu H 2011 Roles of European blocking and tropical-extratropical interaction in the 2010 Pakistan flooding Geophys. Res. Let. 38 6 Horton D E, Johnson N C, Singh D, Swain D L, Rajaratnam B and Diffenbaugh N S 2015 Contribution of changes in atmospheric circulation patterns to extreme temperature trends. Nature 522 465–9 Online: http://www.ncbi.nlm.nih.gov/pubmed/26108856 Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D, Gandin L, Iredell M, Saha S, White G, Woollen J, Zhu Y, Chelliah M, Ebisuzaki W, Higgins W, Janowiak J, Mo K C, Ropelewski C, Wang J, Leetma A, Reynolds R, Jenne R and Joseph D 1996 The NCEP/NCAR 40-year reanalysis project Bull. Am. Meteorol. Soc. 437–71 Kim B-M, Son S-W, Min S-K, Jeong J-H, Kim S-J, Zhang X, Shim T and Yoon J-H 2014 Weakening of the stratospheric polar vortex by Arctic sea-ice loss. Nat. Commun. 5 4646 Online: http://www.ncbi.nlm.nih.gov/pubmed/25181390

174

Appendix

Kyselý J 2008 Influence of the persistence of circulation patterns on warm and cold temperature anomalies in Europe: Analysis over the 20th century Glob. Planet. Change 62 147–63 Online: http://linkinghub.elsevier.com/retrieve/pii/S0921818108000118 Kyselý J and Huth R 2005 Changes in atmospheric circulation over Europe detected by objective and subjective methods Theor. Appl. Climatol. 85 19–36 Online: http://link.springer.com/10.1007/s00704-005-0164-x Lau W K M and Kim K-M 2012 The 2010 Pakistan Flood and Russian Heat Wave: Teleconnection of Hydrometeorological Extremes J. Hydrometeorol. 13 392–403 Online: http://journals.ametsoc.org/doi/abs/10.1175/JHM-D-11-016.1 Lehmann J, Coumou D, Frieler K, Eliseev A V and Levermann A 2014 Future changes in extratropical storm tracks and baroclinicity under climate change Environ. Res. Lett. 9 084002 Online: http://stacks.iop.org/17489326/9/i=8/a=084002?key=crossref.b87e6badff3d8c6dcae24acefd0aedb1 Mueller B and Seneviratne S I 2012 Hot days induced by precipitation deficits at the global scale. Proc. Natl. Acad. Sci. U. S. A. 109 12398–403 Online: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3411978&tool=pmcentrez&rendertyp e=abstract O’Gorman P A 2010 Understanding the varied response of the extratropical storm tracks to climate change Proc. Natl. Acad. Sci. USA 107 19176–80 Overland J E, Francis J A, Hanna E and Wang M 2012 The recent shift in early summer Arctic atmospheric circulation Geophys. Res. Lett. 39 L19804 Palmer T 2014 Record-breaking winters and global climate change Science (80-. ). 344 803–4 Online: http://www.sciencemag.org/cgi/doi/10.1126/science.1255147 Pedlosky J 1979 Geophysical Fluid Dynamics (New York, USA: Springer) Petoukhov V, Rahmstorf S, Petri S and Schellnhuber H J 2013 Quasiresonant amplification of planetary waves and recent Northern Hemisphere weather extremes. Proc. Natl. Acad. Sci. U. S. A. 110 5336– 41 Petrie R E, Shaffrey L C and Sutton R T 2015 Atmospheric response in summer linked to recent Arctic sea ice loss Q. J. R. Meteorol. Soc. n/a – n/a Online: http://doi.wiley.com/10.1002/qj.2502 Rahmstorf S and Coumou D 2011 Increase of extreme events in a warming world. Proc. Natl. Acad. Sci. U. S. A. 108 17905–9 Online: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3207670&tool=pmcentrez&rendertyp e=abstract


Runge J, Heitzig J, Marwan N and Kurths J 2012 Quantifying causal coupling strength: A lag-specific measure for multivariate time series related to transfer entropy Phys. Rev. E 86 061121 Online: http://link.aps.org/doi/10.1103/PhysRevE.86.061121 Runge J, Petoukhov V and Kurths J 2014 Quantifying the Strength and Delay of Climatic Interactions: The Ambiguities of Cross Correlation and a Novel Measure Based on Graphical Models J. Clim. 27 720– 39 Online: http://journals.ametsoc.org/doi/abs/10.1175/JCLI-D-13-00159.1 Russo S, Dosio A, Graversen R G, Sillmann J, Carrao H, Dunbar M B, Singleton A, Montagna P, Barbosa P and Vogt J V. 2014 Magnitude of extreme heat waves in present climate and their projection in a warming world J. Geophys. Res. Atmos. n/a – n/a Online: http://doi.wiley.com/10.1002/2014JD022098 Schär C, Vidale P L, Lüthi D, Frei C, Häberli C, Liniger M A and Appenzeller C 2004 The role of increasing temperature variability in European summer heatwaves Nature 427 332–6 Schleussner C F, Runge J, Lehmann J and Levermann a. 2014 The role of the North Atlantic overturning and deep ocean for multi-decadal global-mean-temperature variability Earth Syst. Dyn. 5 103–15 Online: http://www.earth-syst-dynam.net/5/103/2014/ Schneider T, Bischoff T and Plotka H 2014 Physics of changes in synoptic midlatitude temperature variability J. Clim. Schneidereit A, Schubert S, Vargin P, Lunkeit F, Zhu X, Peters D H W and Fraedrich K 2012 Large-Scale Flow and the Long-Lasting Blocking High over Russia: Summer 2010 Mon. Weather Rev. 140 2967– 81 Schubert S D, Wang H, Koster R D, Suarez M J and Groisman P Y 2014 Northern Eurasian Heat Waves and Droughts J. Clim. 27 3169–207 Online: http://journals.ametsoc.org/doi/abs/10.1175/JCLI-D-1300360.1 Schubert S, Wang H and Suarez M 2011 Warm season subseasonal variability and climate extremes in the Northern Hemisphere: The role of stationary Rossby waves J. Clim. 24 4773–92 Screen J A and Simmonds I 2014 Amplified mid-latitude planetary waves favour particular regional weather extremes Nat. Clim. Chang. 4 704–9 Screen J A, Simmonds I, Deser C and Tomas R 2013 The Atmospheric Response to Three Decades of Observed Arctic Sea Ice Loss J. Clim. 26 1230–48 Online: http://journals.ametsoc.org/doi/abs/10.1175/JCLI-D-12-00063.1 Shepherd T G 2014 Atmospheric circulation as a source of uncertainty in climate change projections Nat. Geosci. 7 703–8 Tachibana Y, Nakamura T, Komiya H and Takahashi M 2010 Abrupt evolution of the summer Northern Hemisphere annular mode and its association with blocking J. Geophys. Res. 115 13

176

Appendix

Trenberth K E, Fasullo J T, Branstator G and Phillips A S 2014 Seasonal aspects of the recent pause in surface warming Nat. Clim. Chang. 4 911–6 Walsh J E 2014 Intensified warming of the Arctic: Causes and impacts on middle latitudes Glob. Planet. Change 117 52–63 Online: http://linkinghub.elsevier.com/retrieve/pii/S0921818114000575 Woollings T and Blackburn M 2012 The North Atlantic Jet Stream under Climate Change and Its Relation to the NAO and EA Patterns J. Clim. 25 886–902 Online: http://journals.ametsoc.org/doi/abs/10.1175/JCLI-D-11-00087.1


Figures

Figure 1. Weakening summer circulation in the Northern Hemisphere mid-latitudes over 19792014.

Absolute

changes

in

meridional

temperature gradient, zonal mean wind, eddy kinetic energy (EKE), daily wave amplitudes for waves 6-10 and frequency of high-amplitude quasi-stationary wave 7. Variables are based on ERA-Interim reanalysis, calculated at 500mb, averaged over 35oN-65oN and all longitudes, for June-July-August.

Black

lines

plot

actual

observations and red lines their linear trend, which are all significant at 95% confidence except for the linear trend in the bottom panel which is significant at 90% confidence only. Similar trends are found at other pressure levels and in other reanalyses (Coumou et al 2014, 2015).

178

Appendix

Figure 2. Arctic amplification of temperature trends over 1979–2014 for each season. Trends in zonal mean temperature for a) winter (December–February), b) spring (March–May), c) summer (JuneAugust) and d) autumn (September–November) based on the ERA-Interim reanalysis. Black contours indicate regions with significance at the 99% (solid lines) and 95% (dotted lines) confidence levels. The line graphs show trends averaged over the lower atmosphere (950–1,000 hPa; solid lines) and over the entire atmospheric column (300–1,000 hPa; dotted lines). Source: (Cohen et al 2014).


Figure 3. Regression slope between EKE and near-surface temperature in summer months (June-August) in units of m2s-2/K. Both variables were calculated from ERA-Interim data and linearly detrended at each grid point. Significant negative correlation (stippling indicating 95% confidence) exists over most storm track affected regions. Source: (Coumou et al 2015).

180

Appendix

Figure 4. Double jet and resonant amplification during summer of 2010. Northern Hemisphere Hovmöller diagrams during summer 2010 for a) zonal wind, b) temperature anomaly and c) 𝑙 2 (𝑘 = 6).

Variables are calculated from the ERA-Interim reanalysis using a 15day running mean low-frequency

bandpass filter and averaged over all longitudes. Temperature anomalies from climatology (panel b) were calculated at each grid point before averaging over all longitudes. The black lines in panel c) indicate the emergence of the 2 turning points of the waveguide for wave 6. Panel d) plots the number of consecutive days for which resonance conditions were fulfilled for waves 6-8 and panel e) plots the observed amplitude of wave 6.


Figure 5. Number of identified July and August resonance events (Petoukhov et al 2013) binned together in 4-year periods from 1980 to 2011. Text in the grey bars indicates the actual months and wave numbers involved (in brackets) and the table on the left lists the associated extreme weather events. The red line plots the difference in surface warming in the Arctic (north of 65°N) and in the rest of the Northern Hemisphere (south of 65°N), based upon the temperature dataset by Cowtan and Way (Cowtan and Way 2013). Source: (Coumou et al 2014).

182

Appendix

Figure 6. 2D probability density distributions for daily values of phase speed (horizontal axis) against respectively (top) wave amplitude and (bottom) zonal mean zonal wind for (left) wave 6, (middle) wave 7 and (right) wave 8. Variables are based on ERA-Interim reanalysis at 500mb, averaged over 35oN-65oN for days in July−August (solid lines). Color shading indicates anomalies from July-August climatology during resonance events showing increased frequencies of high-amplitude quasi-stationary waves (red in top panels) Adapted from: (Coumou et al 2014).


Figure 7. Mid-latitudinal extreme index MEX (Eq. 4) in units of standard deviation for a) monthly heat extremes and b) daily heat extremes for July−August climatology (black) and resonance events (red). Adapted from: (Coumou et al 2014).

A2. Consistent increase in Indian monsoon rainfall and its variability across CMIP-5 models Arathy Menon, Anders Levermann, Jacob Schewe, Jascha Lehmann, and Katja Frieler. This paper analyzes changes in projected seasonal monsoon rainfall from an ensemble of 20 different CMIP5 climate models. Robust results across most models include (i) a consistent increase in seasonal mean rainfall during the summer monsoon periods, (ii) a northward shift in monsoon circulation by the end of the 21st century, and (iii) a positive trend in the inter-annual variability of the Indian monsoon rainfall. Published in Earth System Dynamics, 2013, 4(2), 287-300, doi:10.5194/esd-4-287-2013.

Open Access

Biogeosciences

Climate of the Past

Open Access

Earth System Dynamics

Open Access

Geoscientific Instrumentation Methods and Data Systems

B

Open Access

A2. Consistent increase in Indian monsoon rainfall and its variability across CMIP-5 models 187

Earth Syst. Dynam., 4, 287–300, 2013 www.earth-syst-dynam.net/4/287/2013/ doi:10.5194/esd-4-287-2013 © Author(s) 2013. CC Attribution 3.0 License.

Model Development

A. Menon1,2 , A. Levermann1,2 , J. Schewe1 , J. Lehmann1,2 , and K. Frieler1 1 Potsdam

Mode

Open Access

Institute for Climate Impact Research, 14473 Potsdam, Germany of Physics, University of Potsdam, 14476 Potsdam, Germany

Hydrology and Earth System Correspondence to: A. Menon ([email protected]) Sciences Received: 3 December 2012 – Published in Earth Syst. Dynam. Discuss.: 7 January 2013 2 Institute

Open Access

Consistent increase in Indian monsoon rainfall and its variability Geoscientific across CMIP-5 models

Revised: 27 May 2013 – Accepted: 24 July 2013 – Published: 28 August 2013

Indian summer monsoon rainfall affects the lives of the large population of India by determining its water availability as well as food security (Parthasarathy et al., 1988; Auffhammer et al., 2006). About 80 % of the annual precipitation over India occurs during the monsoon months from June to

Published by Copernicus Publications on behalf of the European Geosciences Union.

Open Access

Introduction

et al., 2003; Turner and Annamalai, September (KripalaniOcean Science 2012), and the released latent heat plays an important role in the atmospheric circulations as well as the radiative heat budget of the region (Webster, 1972). Even after achieving growth in service and industrial sectors, agriculture plays a vital role in the economy of the country as it is the predominant occupation in the ruralSolid regions Earth of India. Extreme rainfall events and crop failure have adverse effects on the millions of inhabitants as well as the national economy. Hence it is of critical importance to understand how the monsoon will change under future warming, in order to take sufficient adaptation measures. Changes in the The Indian Cryosphere monsoon under global warming are still a matter of intense scientific debate (e.g., Sabade et al., 2011; Turner and Annamalai, 2012). An analysis of observational data for the past 130 yr shows no clear evidence of the effect of global warming on Indian monsoon rainfall strength and its interannual variability (Mooley and Parthasarathy, 1984; Kripalani et al., 2003; Guhathakurta and Rajeevan, 2008). While no clear trend can be found for monsoon rainfall averaged over India as a whole (Mooley and Parthasarathy, 1984; Guhathakurta and Rajeevan, 2008), observations show significant trends in rainfall over several smaller regions of the country (Jagannathan and Parthasarathy, 1973; Kumar et al., 1992; Guhathakurta and Rajeevan, 2008). Some subdivisions of India show a positive trend in monsoon rainfall, while some show a significant negative trend, whereas there are some small regions that do not show any significant trends (Kumar et al., 1992; Guhathakurta and Rajeevan, 2008). Observations based on a 1◦ × 1◦ gridded daily dataset suggest that

O

Open Access

1

Open Access

Abstract. The possibility of an impact of global warming on the Indian monsoon is of critical importance for the large population of this region. Future projections within the Coupled Model Intercomparison Project Phase 3 (CMIP-3) showed a wide range of trends with varying magnitude and sign across models. Here the Indian summer monsoon rainfall is evaluated in 20 CMIP-5 models for the period 1850 to 2100. In the new generation of climate models, a consistent increase in seasonal mean rainfall during the summer monsoon periods arises. All models simulate stronger seasonal mean rainfall in the future compared to the historic period under the strongest warming scenario RCP-8.5. Increase in seasonal mean rainfall is the largest for the RCP-8.5 scenario compared to other RCPs. Most of the models show a northward shift in monsoon circulation by the end of the 21st century compared to the historic period under the RCP-8.5 scenario. The interannual variability of the Indian monsoon rainfall also shows a consistent positive trend under unabated global warming. Since both the long-term increase in monsoon rainfall as well as the increase in interannual variability in the future is robust across a wide range of models, some confidence can be attributed to these projected trends.

Th

188

288

Appendix

A. Menon et al.: Consistent increase in Indian monsoon rainfall

monsoon rainfall in a homogeneous region over central India shows no significant long-term trend during the past few decades (Goswami et al., 2006). Even though the frequency and magnitude of extreme events show a rising trend over central India (Rajeevan et al., 2006; Goswami et al., 2006), a significant trend is absent in seasonal mean rainfall. This is attributed to a decrease in the frequency of moderate events (Goswami et al., 2006). A study by Fu et al. (1999) shows an increase in the Indian monsoon rainfall in relation to an abrupt warming around the year 1920. By contrast several ice core records and speleothem records show a decreasing trend in the Indian summer monsoon (ISM) rainfall in the last century. ISM rainfall intensity measured from the Dasuopu ice core shows a decreasing trend during the past century (Thompson et al., 2000; Duan et al., 2004). ISM intensity from speleothem record over southern Arabia also shows a decreasing trend over the past century, which is attributed to the increase in sea surface temperature over the Indian Ocean (Burns et al., 2002). The ISM intensity reconstructed from a tree ring record over Tibetan Plateau also shows a decreasing trend from 1860 to present (Xu et al., 2012). Models have their own limitations in capturing the regional rainfall accurately (Turner and Annamalai, 2012). While some model studies find very little impact on the all India monsoon rainfall in transient and time-slice climate change experiments (Mahfouf et al., 1994; Lal et al., 1994; Timbal et al., 1995; Lal et al., 1995), some others suggest an increase in the mean Indian monsoon precipitation as well as the interannual variability under enhanced warming (Meehl and Washington, 1993; Kitoh et al., 1997; Hu et al., 2000; Lal et al., 2001; Cubasch et al., 2001; May, 2002; Fan et al., 2012). A study (Ashfaq et al., 2009) based on a high-resolution nested model suggests a suppression in the Indian monsoon rainfall in the future, which is attributed to a weakening of the monsoon circulation and a suppression of the intraseasonal modes. CMIP-5 models consistently project a significant increase in Indian summer monsoon rainfall sub-seasonal variability under unmitigated climate change (Menon et al., 2013). The Hamburg COSMOS model shows a complex behavior with changing skewness of the rainfall distribution and an associated increase in monsoon failure events (Schewe and Levermann, 2012). A subset of the IPCC AR4 models suggest an increase in the strength of the monsoon precipitation (Kripalani et al., 2007), whereas the monsoon circulation is projected to weaken (Tanaka et al., 2005; Ueda et al., 2006), while earlier studies using slab ocean models suggest a strengthening of monsoon precipitation as well as the circulation (Zhao and Kellogg, 1988). The projected precipitation from few CMIP-3 models, which are considered more realistic, shows a range of trends including negative trends in monsoon rainfall by 2100 (Turner and Annamalai, 2012) under the SRES A1B scenario. Cherchi et al. (2011) analyze global monsoons based on a fully coupled atmosphere–ocean general circulation model and suggest Earth Syst. Dynam., 4, 287–300, 2013

that the Indian summer monsoon intensifies in the future, mainly in response to the increased moisture content under various CO2 forcings. Studies based on CMIP-5 models under RCP-4.5 scenario project an increase in global mean precipitation of around 3.2 % K−1 (Hsu et al., 2013) and a larger increase in annual mean precipitation over the entire Asian monsoon region with less uncertainty as compared to CMIP-3 models (Lee and Wang, 2012). Compared to CMIP3 models, CMIP-5 models have higher horizontal and vertical resolution in the atmosphere and ocean, and they have a more detailed representation of aerosols. Some of the CMIP5 models have a more complete representation of the carbon cycle compared to CMIP-3 models. Sperber et al. (2012) suggest that, because of the higher spatial resolution, CMIP5 models have a better representation of rainfall compared to CMIP-3 models, especially in the vicinity of steep topography (like Western Ghats). CMIP-5 models outperform CMIP-3 models in simulating the monsoon annual cycle, the onset of monsoon as well as the time of the monsoon peak. The spatial extent of monsoon in CMIP-5 multi-model mean (MMM) is more realistic than in the CMIP-3 MMM. The magnitude of intra-seasonal variance is also more realistic in CMIP-5 MMM compared to CMIP-3 MMM. Both CMIP5 and CMIP-3 MMM capture low-level monsoon circulation quite well with pattern correlations of 0.98 and 0.97 respectively, compared to ERA40 observational dataset. Even though the time–mean rainfall error has a consistent pattern between CMIP-5 and CMIP-3 MMM, the amplitude of error is less for CMIP-5 MMM compared to CMIP-3 MMM (Sperber et al., 2012). Aerosols play an important role in shaping monsoons over South Asia. IPCC Fourth Assessment Report (Meehl et al., 2007) examined the role of scattering aerosols like sulfate aerosols in monsoons. The presence of sulfate aerosols over South Asia masks the effect of increased temperature gradient by greenhouse gases by reflecting the solar radiation and thereby reducing the land warming and hence the thermal contrast. Recent studies consider the effect of black carbon aerosols on South Asian monsoons (Ramanathan et al., 2005; Meehl et al., 2008) and suggest that a “business as usual” black carbon scenario can result in about 25 % decrease in mean monsoon rainfall by the mid-21st century. Some of the models also show a projected increase in the rainfall interannual variability. However, the seasonal projection of interannual variability of South Asian monsoon rainfall is a major challenge (Sperber and Palmer, 1996; Goswami, 1998). Multi-decadal fluctuations are also present in the Indian summer monsoon rainfall and are forced mainly by the tropical sea surface temperatures, and partly by extra-tropical oscillations like Atlantic multidecadal oscillation (Kucharski et al., 2009). Indian monsoon rainfall shows considerable decrease from late 1950s to 1990s. Kucharski et al. (2009) use CMIP-3 models to show that the increase in greenhouse gases (GHGs) in the 20th century has not contributed significantly to the observed Indian summer monsoon decadal variability. www.earth-syst-dynam.net/4/287/2013/



289

Rainfall (mm/d) 1

2

3

4

5

6

7

8

9

MIROC−ESM MIROC−ESM−CHEM CCSM4 NorESM1−M GFDL−CM3 GFDL−ESM2G GFDL−ESM2M INM−CM4 ACCESS1.0 MPI−ESM−LR FGOALS−s2 CNRM−CM5 CanESM2 BCC−CSM1.1 HadGEM2−CC HadGEM2−ES IPSL−CM5A−MR IPSL−CM5A−LR CSIRO−Mk3.6.0 MRI−CGCM3

Fig. 1. JJAS mean rainfall over all India region from different models for the historic period. The black vertical line shows the all India mean monsoon rainfall from observations for the period 1871–2004, and the dashed lines show mean plus/minus twice the standard deviation of all India mean rain. Circles with error bars represent mean and mean plus/minus one standard deviation for the 20 comprehensive models from 1871 to 2004.

In order to capture the full range of possible future scenarios, including mitigation strategies, the Representative Concentration Pathways (RCPs) have been developed as a basis for the IPCC fifth assessment report. There are four RCPs categorized according to their approximate radiative forcing in the year 2100. We use data from RCP model simulations in order to study the projected changes in the mean and variability of ISM rainfall in the future. In this study, we examine the mid-19th century to the end of the 21st century variability of ISM rainfall simulated by 20 of the models that participated in the CMIP-5. Section 2.1 shows a brief model evaluation of the Indian summer monsoon mean rainfall. Section 2.2 gives the trend in all India mean monsoon rainfall and Sect. 2.3 its interannual variability in the RCP-based simulations. Section 2.4 deals with the changes in monsoon circulation in the future. The results are discussed in Sect. 3.

2 2.1

Results Model evaluation

In this study, we use simulated rainfall obtained from 20 of the models that participated in the CMIP-5. Models are chosen according to the availability of the data: only those models are analyzed for which data for historic period (1850– 2005), RCP-8.5 and at least one more scenario were available at the time of the study. The model information is summarized in Table 1. The range in global mean temperature www.earth-syst-dynam.net/4/287/2013/

as constrained by past climate observations allows for a wide range of responses within an RCP (Schewe et al., 2011). Historical simulations are based on solar and volcanic forcing, land use, observed concentrations of greenhouse gases, and reconstructed aerosol emissions. Future projections are based on the four Representative Concentration Pathways (RCPs) (Moss et al., 2010). RCP-8.5 is the pathway for which radiative forcing reaches 8.5 W m−2 by 2100. Similarly RCP4.5 and RCP-6.0 represent the pathways for which radiative forcing reaches 4.5 and 6 W m−2 in 2100. RCP-2.6 peaks in radiative forcing at 3 W m−2 before 2100 and declines afterwards reaching 2.6 W m−2 in 2100. India as a whole (allIndia) is selected for the study, and data are masked over all-India region. Mean rainfall is obtained by averaging the June–September (JJAS) rainfall over the all-India land region and denoted as all-India summer monsoon rainfall (AISMR). The all-India rainfall dataset from Parthasarathy et al. (1994) is used to compare the seasonal mean rainfall from models during historical periods with observations. The observational data cover a period from 1871 to 2004. In order to identify models with a potentially realistic representation of the monsoon rainfall, we compare their long-term seasonal mean with the observed precipitation (Parthasarathy et al., 1994) for the period 1871 to 2004 (Fig. 1). The climatological mean rainfall from observations is 7.1 mm day−1 , with a standard deviation of 0.7 mm day−1 . About half of the models capture seasonal mean rainfall within twice the standard deviation (vertical dashed lines in

Earth Syst. Dynam., 4, 287–300, 2013

190

Appendix

290


Table 1. Details of the data availability for the 20 comprehensive models used in the study. Only those models are selected for which data for historic period, RCP-8.5 and at least one more scenario are available at the time of the study. Modeling Center (Group)

Model

RCP-8.5

RCP-6.0

RCP-4.5

RCP-2.6

Commonwealth Scientific and Industrial Research Organization (CSIRO) and Bureau of Meteorology (BOM), Australia

ACCESS1.0

Y

N

Y

N

Beijing Climate Center, China Meteorological Administration

BCC-CSM1.1

Y

Y

Y

Y

Canadian Centre for Climate Modelling and Analysis

CanESM2

Y

N

Y

Y

National Center for Atmospheric Research

CCSM4

Y

Y

Y

Y

Centre National de Recherches Météorologiques/Centre Européen de Recherche et Formation Avancées en Calcul Scientifique

CNRM-CM5

Y

N

Y

Y

Commonwealth Scientific and Industrial Research Organization in collaboration with Queensland Climate Change Centre of Excellence

CSIRO-Mk3.6.0

Y

Y

Y

Y

LASG, Institute of Atmospheric Physics, Chinese Academy of Sciences

FGOALS-s2

Y

Y

Y

Y

NOAA Geophysical Fluid Dynamics Laboratory

GFDL-CM3

Y

Y

N

Y

GFDL-ESM2G GFDL-ESM2M

Y Y

Y Y

Y Y

Y Y

Met Office Hadley Centre

HadGEM2-CC HadGEM2-ES

Y Y

N Y

Y Y

N Y

Institute for Numerical Mathematics

INM-CM4

Y

N

Y

N

Institut Pierre-Simon Laplace

IPSL-CM5A-LR IPSL-CM5A-MR

Y Y

Y N

Y Y

Y Y

Japan Agency for Marine-Earth Atmosphere and Ocean Research Science and Technology, Institute (The University of Tokyo), and National Institute for Environmental Studies

MIROC-ESM

Y

Y

Y

Y

MIROC-ESM-CHEM

Y

Y

Y

Y

Max Planck Institute for Meteorology

MPI-ESM-LR

Y

N

Y

Y

Meteorological Research Institute

MRI-CGCM3

Y

Y

Y

Y

Norwegian Climate Centre

NorESM1-M

Y

Y

Y

Y

Earth Syst. Dynam., 4, 287–300, 2013

www.earth-syst-dynam.net/4/287/2013/


A. Menon et al.: Consistent increase in Indian monsoon rainfall Fig. 1) of the observed mean for the period 1871 to 2004. Models like MIROC-ESM and MIROC-ESM-CHEM show a slight overestimation of seasonal mean rainfall, while models like CSIRO-Mk3 and MRI-CGCM3 show an underestimation. The error bars in Fig. 1 represent long-term standard deviations for each of the models under consideration, and the values vary from 0.4 to 0.7 for various models. NorESM1M and GFDL-CM3 capture the mean rainfall closest to the observed mean. The spatial pattern of JJAS rainfall climatology over India (Fig. 2), from the India Meteorological Department observational dataset (Rajeevan et al., 2006), shows that the mean precipitation is highest over the south-west coast, central India and the north-east India. The spatial patterns of rainfall during the monsoon season as simulated by all models are shown in Fig. 3. The models that underestimate the climatological rainfall do not capture the spatial pattern of monsoon well. CSIRO-Mk3.6.0 and MRI-CGCM3 model more rainfall over the east coast of Bay of Bengal and the tropical Indian Ocean. They show very low rainfall over the all-India region. Similarly, the Hadley Centre models (HadGEM2-CC and HadGEM2-ES) and the Institute Pierre Simon Laplace models (IPSL-CM5A-LR and IPSLCM5A-MR) capture very little rainfall over the all-India region with comparatively higher rainfall over the Himalayan mountains and the Bay of Bengal. As discussed by Levermann et al. (2009), the monsoon region can enter a climatic regime in which latent heat transport towards land is insufficient to sustain a monsoon circulation, which may lead to abrupt monsoon transition (Zickfeld et al., 2005) as observed in the past (Schewe et al., 2012; Cook et al., 2010; Sinha et al., 2011). While observations clearly show that the ISM is currently within the active monsoon regime, it is possible that the CMIP-5 models that exhibit a very weak ISM are outside this regime. In this study, we decide to interpret the results of the future monsoon evolution from models with historical mean precipitation below the observed mean minus twice its standard deviation (5.7 mm day−1 ) as well as the ones with an unrealistic spatial pattern, with care as they are less likely to provide a good approximation of the real evolution. Full information is, however, provided for all models. 2.2

Long-term trend in Indian monsoon rainfall under various RCP warming scenarios

AISMR is analyzed for the four RCPs (Figs. 4 and 5). AISMR shows a clear positive long-term trend in all models under the RCP-8.5 scenario, whereas the long-term trend is small under RCP-2.6 scenario (Fig. 4). And even for the lowest concentration scenario RCP-2.6, only 3 out of 20 models show a small decreasing trend in rainfall. Under RCP-8.5 scenario, the majority of the models simulate rainfall response outside the envelope of the baseline variability (black horizontal lines) towards the end of the 21st century. FGOALS-s2 shows a rainfall response beyond the baseline www.earth-syst-dynam.net/4/287/2013/

291

Fig. 2. June–September (JJAS) rainfall climatology (mm/day) during the period 1970–2000 from the India Meteorological Department (IMD) daily gridded observational dataset. Mean precipitation is highest over south-west peninsular India, north-east India and central India.

variability from the beginning of the 21st century onwards under all RCP scenarios. The percentage changes in the AISMR (δmean) by the end of the 21st century (2070–2100) with respect to the pre-industrial period (1870–1900) under all RCPs are summarized in Fig. 6. Models listed in the upper panel of Fig. 6 are those that capture the AISMR well with mean rainfall for the historic period (1871–2004) falling within twice the standard deviation (0.7 mm day−1 ) of the observed mean (7.1 mm day−1 ). The relative increase in mean monsoon rainfall is less (up to < 15 %) for these models compared to the ones with a much lower historic mean. The significance of δmean values are obtained from a Student’s t test, and it shows that 19 out of the 20 models show a significant increase in δmean under the RCP-8.5 scenario at a 95 % confidence level. MPI-ESM-LR shows a slight increase in the AISMR during the end of the 21st century compared to the pre-industrial period under RCP-8.5 scenario, which is not significant at a 95 % confidence level. MRICGCM3 shows the maximum increase in AISMR of about 60 % by the end of the 21st century compared to the end of the 19th century for RCP-8.5 and RCP-6.0 scenarios. But as shown in Fig. 3, MRI-CGCM3 does not capture the spatial pattern of AISMR realistically. All models show a consistent increase in δmean at 95 % confidence level under all scenarios. None of the negative values of δmean are significant at a 95 % confidence level. In summary, a consistent picture of an increasing seasonal mean rainfall under global warming arises from the CMIP-5 intercomparison. Due to the relatively fast adjustment time of the atmosphere, most models show little path dependence of the ISM change, in the sense that changes are very similar for the same increase in global mean temperature compared to preindustrial period independent of which scenario was followed. Therefore, it is possible to provide the percentage Earth Syst. Dynam., 4, 287–300, 2013

192

292

Appendix


Fig. 3. June–September (JJAS) rainfall climatology (mm day−1 ) during the period 1871–2004 for all 20 models. Models with lowest seasonal mean rainfall do not capture the spatial pattern realistically. The models are shown in the same order as in Fig. 1.

Earth Syst. Dynam., 4, 287–300, 2013




293 10

10 NorESM1−M

8

8 MIROC−ESM

MIROC−ESM−CHEM

CCSM4

8

8 INM−CM4

JJAS Rainfall (mm/d)

6

6 GFDL−ESM2G

GFDL−CM3

GFDL−ESM2M

8

8 ACCESS1.0

MPI−ESM−LR

FGOALS−s2

CNRM−CM5

6

6

8

8 BCC−CSM1.1

CanESM2

HadGEM2−CC

HadGEM2−ES

6

6

4

4

6

6 CSIRO−Mk3.6.0

MRI−CGCM3

4

4

2 IPSL−CM5A−MR 1900

2

IPSL−CM5A−LR

2000

2100

1900

2000

2100

1900

2000

2100

1900

2000

2100

Fig. 4. Indian summer monsoon seasonal (June–September) mean rainfall for the period 1850–2100 from the 20 comprehensive climate models. Gray shadings represent the yearly values, and red and blue lines represent the nonlinear trend in seasonal mean rain obtained from a singular spectrum analysis with a window width of 21 yr for RCP-8.5 and RCP-2.6 respectively. The non-linear trends are calculated using a routine from Aslak Grinsted, and the method is discussed in Moore et al. (2005). The horizontal black lines represent mean ± standard deviation for each model for the period 1871–2004, which shows the range of baseline variability. 10

10 NorESM1−M

8

8 MIROC−ESM

MIROC−ESM−CHEM

CCSM4

8

8 INM−CM4

JJAS Rainfall (mm/d)

6

6 GFDL−CM3

GFDL−ESM2G

GFDL−ESM2M

8

8 ACCESS1.0

MPI−ESM−LR

FGOALS−s2

CNRM−CM5

6

6

8

8 CanESM2

BCC−CSM1.1

HadGEM2−CC

HadGEM2−ES

6

6

4

4

6

6 CSIRO−Mk3.6.0

MRI−CGCM3

4

4

2 IPSL−CM5A−MR 1900

2000

2

IPSL−CM5A−LR

2100

1900

2000

2100

1900

2000

2100

1900

2000

2100

Fig. 5. Long-term trend in seasonal mean rainfall, same as in Fig. 4, but for RCP-4.5 (green) and RCP-6.0 (orange) scenarios. Horizontal black lines represent the range of baseline variability as in Fig. 4. Please refer Table 1 for the data availability for Figs. 4 and 5.


Earth Syst. Dynam., 4, 287–300, 2013

194

Appendix

294

A. Menon et al.: Consistent increase in Indian monsoon rainfall δ mean (%) −5

0

5

10

0

15

MIROC−ESM

MIROC−ESM

MIROC−ESM−CHEM

MIROC−ESM−CHEM

CCSM4

CCSM4

NorESM1−M

NorESM1−M

GFDL−CM3

GFDL−CM3

GFDL−ESM2G

GFDL−ESM2G

GFDL−ESM2M

GFDL−ESM2M

δ mean (%) /K 4 2

INM−CM4

INM−CM4

ACCESS1.0

ACCESS1.0

MPI−ESM−LR

MPI−ESM−LR

FGOALS−s2

0 2 4

FGOALS−s2

CNRM−CM5

CNRM−CM5

CanESM2

CanESM2

BCC−CSM1.1

BCC−CSM1.1

HadGEM2−CC HadGEM2−CC

HadGEM2−ES

HadGEM2−ES IPSL−CM5A−MR IPSL−CM5A−LR

RCP4.5 RCP6.0 RCP8.5

IPSL−CM5A−MR

RCP8.5 RCP6.0 RCP4.5 RCP2.6

IPSL−CM5A−LR CSIRO−Mk3.6.0

CSIRO−Mk3.6.0

MRI−CGCM3

MRI−CGCM3

0

20

40

Fig. 6. Percentage change in mean rainfall for RCP-8.5 (red), RCP6.0 (orange), RCP-4.5 (green) and RCP-2.6 (blue) for all models under consideration. Relative change is calculated as the change in seasonal (JJAS) mean rainfall for the period 2070–2100with respect to the period 1870–1900, i.e.,

JJASrain2070−2100 · 100 JJASrain1870−1900

5

10

15

20

Fig. 7. Percentage change in mean rainfall per degree change in temperature for RCP-8.5 (red), RCP-6.0 (orange) and RCP-4.5 (green) for all models. As the temperature change in RCP-2.6 is very small, it is not taken into consideration. Transparent bars are the values that are not significant at 95 % confidence level.

− 100. The

gap separates models with rainfall values for 1871–2004 lying within and outside twice the standard deviation of the observed mean as per Fig. 1. Bars on the panels are transparent if the δmean values are not significant at 95 % confidence level.

change in AISMR as a function of global mean temperature change or AISMR change per degree of warming. It is given in Fig. 7. RCP-2.6 scenario is not considered here as the temperature changes are very low under this scenario. Considering only trends that are significant at a 95 % confidence level, all models project an increase in the AISMR with an increase in temperature. The trends are comparatively smaller for the more realistic models. Figure 8 shows the histogram of the trends per degree kelvin. The relative changes in AISMR per degree of warming range from 1 to 19 % K−1 ; 66.5 % of the trends for an ensemble of all models lie in the range of 1– 9 % K−1 with a median increase of 3.2 % K−1 . While considering only the more realistic models in the upper panel of Fig. 7, 66.5 % of the relative changes in AISMR per degree of Earth Syst. Dynam., 4, 287–300, 2013

0

60

warming are in the range 1.3–3 % K−1 . These models show a median increase of 2.3 % K−1 . This value is closer to the projected increase in global mean precipitation per degree of warming (2.2 ± 0.52 % K−1 ) given by Frieler et al. (2011) for CMIP-3. Figure 9 represents the changes in the spatial pattern of rainfall under the RCP-8.5 scenario compared to the historic period. The majority of the models show an increase in rainfall over almost all parts of India by the end of the 21st century compared to the end of the 20th century. GFDL-ESM2M, GFDL-ESM-2G, MPI-ESM-LR and CanESM2 show a reduction in rainfall over central India in the future but capture an increase in rainfall over other parts of India. 2.3

Future evolution of interannual variability

The percentage changes in long-term standard deviation for the period 2050–2100 with respect to 1900–1950 are used to analyze how the interannual variability evolves under global www.earth-syst-dynam.net/4/287/2013/



295 δ std (%)

12

−30 −20 −10

0

10

20

30

40

MIROC−ESM

10

RCP8.5 RCP6.0 RCP4.5 RCP2.6

MIROC−ESM−CHEM CCSM4

8

NorESM1−M GFDL−CM3

6

GFDL−ESM2G GFDL−ESM2M

4

INM−CM4 ACCESS1.0

2 0 0

MPI−ESM−LR FGOALS−s2

2

4

6

8 10 12 δ mean (%)/ K

14

16

18

20

Fig. 8. Histogram of the percentage changes in mean rainfall per degree change in temperature for all significant scenarios of all models as in Fig. 7. Red dots represent the values for the models in the upper panel of Fig. 7. Black dashed line shows the median value for the ensemble of all models, and red dashed line shows the median value for the most realistic models (i.e., the ones in the upper panel of Fig. 7).

CNRM−CM5 CanESM2 BCC−CSM1.1 HadGEM2−CC HadGEM2−ES IPSL−CM5A−MR IPSL−CM5A−LR CSIRO−Mk3.6.0 MRI−CGCM3

−20

0

20

40

60

Fig. 10. The percentage change of standard deviation (STD) during the second half of 21st century to the standard deviation during the first half of the 20th century under RCP-8.5 (red), RCP-6.0 (orange), RCP-4.5 (green)and RCP-2.6 (blue) for all models under consideration. All values greater than 0show an increase invariabilSTD ity. The percentage change is given as STD2050−2100 · 100 − 100. 1900−1950 The upper and lower panels are separated as in Figs. 6 and 7.

Fig. 9. JJAS rainfall composite difference (mm day−1 ) for the period 2070–2100 under RCP-8.5 scenario and for the period 1970– 2000. The majority of the models capture an increase in summer monsoon mean rainfall by about 0–3 mm day−1 in most parts of India.


warming. The standard deviation of seasonal mean rainfall shows a positive trend in most of the models under the RCP8.5 scenario (Fig. 10) indicating an increase in interannual variability in the future. Out of the 20 models under consideration, 17 models show an increase in interannual variability under this scenario. MIROC-ESM-CHEM, HadGEM2CC and IPSL-CM5A-LR show a slight decrease (< 10 %) in standard deviation by the second half of the 21st century compared to the first half of the 20th century. It has to be noted that most models show an increase in interannual variability in the future under various concentration pathways. The largest increase is simulated by FGOALSs2, BCC-CSM1.1 and HadGEM2-ES under the RCP-8.5 scenario. CCSM4 shows a decrease in variability under all scenarios except RCP-8.5. GFDL-ESM2G, GFDL-ESM2M, FGOALS-s2, HadGEM2-ES and MRI-CGCM3 show an increase in interannual variability under all four RCPs. While HadGEM2-ES captures an increase in interannual variability under all four scenarios, HadGEM2-CC captures a decrease in interannual variability in the two available scenarios at the time of the study. But as shown earlier in Fig. 3, these two models did not capture the spatial pattern of monsoon rainfall reasonably well. Out of the few negative trends Earth Syst. Dynam., 4, 287–300, 2013

196

296

Appendix


Fig. 11. June–September 850 hPa wind (m s−1 ) climatology for the period 1970–2000 from NCEP/NCAR reanalysis data (Kalnay et al., 1996).

of interannual variability, most of them are under the RCP2.6 and RCP-4.5 scenarios. The interannual variability has a clear positive trend in most of the models under the higher scenarios RCP-6.0 and RCP-8.5. 2.4

Changes in monsoon circulation in the future

Some studies suggest a weakening of the monsoon circulation in a number of CMIP-3 models under global warming (Tanaka et al., 2005; Ueda et al., 2006). The 850 hPa summer wind climatology from observational data shows the lowlevel monsoon circulation that carries moisture from over ocean to the Indian land region (Fig. 11). Figure 12 depicts the composite difference in the wind anomaly between the end of the 21st century under RCP-8.5 and the end of the 20th century from the CMIP-5 models. The majority of the models show an increase in wind speed (shaded) in the north of India and a decrease in wind speed in southern peninsular India as well as the north equatorial Indian Ocean by the end of the 21st century. Anomalies in wind direction (vectors) are opposite to the direction of the mean wind over the southern peninsular India, and along the direction of the mean wind over central and northern India, in most of the models. This could indicate a northward shift in the monsoon circulation in the future. The ensemble mean over all 19 models under consideration also shows the same pattern (Fig. 13). This pattern resembles that of the wind anomaly from the CMIP-3 models (cf. Fig. 2a in Ueda et al., 2006). The monsoon circulation strengthens over northern India, but it weakens over the south of India. Figure 14 shows the meridional pattern of the zonal wind averaged over the longitudes 50–110◦ E for all 19 models under consideration. The majority of the models show a slight northward shift in monsoon circulation of the order of about 2◦ by the end of the 21st century under RCP-8.5. Kitoh et al. (1997) suggest a similar northward shift in the monsoon circulation under global warming. Such a latitudinal shift of the circulation would be important to consider when Earth Syst. Dynam., 4, 287–300, 2013

Fig. 12. Difference in wind speed (shading, in m s−1 ) and direction (vectors) during June–September for the period 2070–2100 under RCP-8.5 and 1970–2000 for 19 models under consideration. HadGEM2-ES is not shown as wind data for historic period were not available at the time of the study. Wind vector anomalies are in the direction of the mean flow over the northern parts of India and are opposite to the mean flow over the southern parts of India in most of the models.

assessing changes in the total strength of the circulation. Figure 14 shows that the overall magnitude of the zonal monsoon winds decreases in a few models (e.g., MIROC-ESM, MIROC-ESM-CHEM, IPSL-CM5A-MR, IPSL-CM5A-LR, CanESM2), but remains fairly constant in most models, or even increases in some (e.g., INM-CM4).

3

Discussion and conclusions

The future evolution of Indian summer monsoon rainfall and its interannual variability have been analyzed based on global coupled model simulations under the RCP scenarios. This study analyzes whether previous inconsistency between models regarding the long-term trend in the Indian summer monsoon rainfall under transient warming scenarios still exists in the CMIP-5 generation of climate models. By comparison of the models’ performance with the all-India mean monsoon rainfall for the historic period from observations and examination of the spatial patterns of rainfall, we consider some models as more realistic and put more emphasis on them compared to the ones with a very weak monsoon rainfall. For these models a consistent picture arises: Indian www.earth-syst-dynam.net/4/287/2013/



297

1 1

Fig. 13. Differences in June–September 850 hPa winds (m s−1 ) for the period 2070–2000 (under RCP-8.5) compared to the historic period (1970–2000) for the ensemble mean of all 19 models under consideration.

summer monsoon rainfall increases under future warming. All models except MPI-ESM-LR simulate the maximum positive trend in mean monsoon rainfall under the highest concentration pathway RCP-8.5. This result agrees with Fu et al. (1999), who find an increase in the Indian monsoon rainfall during abrupt warming, and suggests a relationship between global temperature increase and the Indian monsoon rainfall. An increase in seasonal mean precipitation can occur due to changes in the inter-tropical convergence zone (Hu et al., 2000). In coupled models under global warming, it has been attributed predominantly to an increase in the water-holding capacity of the atmosphere with an increase in surface temperature (Trenberth, 1998). For example, Meehl et al. (2005) suggest the increase in water vapor content associated with an increase in sea surface temperature in a warmer climate as the reason for enhanced precipitation in the tropics in some IPCC AR4 models. The atmospheric water vapor is projected to increase by 12–16 % over large parts of India (Kripalani et al., 2007) at the time of CO2 doubling. This increased moisture content can lead to enhanced precipitation. In this study, we see that the increase in AISMR per degree change in temperature is about 2.3 % K−1 , which is similar to the projected increase in global mean precipitation per degree change in temperature in CMIP-3 (Frieler et al., 2011). A second trend that emerges consistently across models in CMIP-5 is an increase in interannual variability. The monsoon variability shows a general increasing trend under various RCPs in most of the models. Rainfall variability is particularly important for societal and economic adaptation strategies, defining the required year-to-year flexibility for agricultural management, disaster preparedness, etc. Further studies are needed to understand the physical reasons behind the increase in interannual variability. One of the reasons attributed to the increase is the increase in El Niño Southern Oscillation (ENSO) variability in the future, which is transmitted to www.earth-syst-dynam.net/4/287/2013/

Fig. 14. Meridional pattern of zonal wind (m s−1 ) averaged over the region 50E–110E for the 19 CMIP-5 models during June– September. Black lines represent JJAS mean zonal wind for the period 1970–2000, and red lines represent JJAS mean zonal wind for the period 2070–2100.

South Asian monsoon rainfall through the Walker cell (Hu et al., 2000; Schewe and Levermann, 2012). Another possibility is that the enhanced variability is attributed to the increase in tropical Indian Ocean and Pacific sea surface temperatures, irrespective of the ENSO variability (Meehl and Arblaster, 2003). According to Meehl and Arblaster (2003), the Pacific Ocean SST plays a dominant role, whereas the Indian Ocean plays a secondary role in monsoon interannual variability. Also an observation-based study suggests that the increase in interannual variability of Indian summer monsoon is associated with warmer land and ocean temperatures (Meehl and Washington, 1993). CMIP-5 models show a strengthening of the monsoon circulation in the northern parts of India and a weakening of circulation in the southern parts. The majority of the models show a northward shift in the monsoon circulation under global warming. We do not aim for a consistent physical understanding across all climate models here, but concluded that most of the models that participated in the CMIP-5 show a positive trend in monsoon mean rainfall as well as its interannual variability under future warming. It can be noted that all trends in AISMR that are significant at a 95 % confidence level are positive. The long-term intensification of monsoon rainfall, but even more Earth Syst. Dynam., 4, 287–300, 2013

198

298

Appendix


so the intensification of monsoon variability, requires longterm adaptation strategies to cope with future climate change in India.

Acknowledgements. This work was funded by the BMBF PROGRESS project (support code 03IS2191B). We acknowledge the World Climate Research Programme’s Working Group on Coupled Modelling, which is responsible for CMIP, and we thank the climate modeling groups (listed in Table 1) for producing and making available their model output. For CMIP, the US Department of Energy’s Program for Climate Model Diagnosis and Intercomparison provided coordinating support and led development of software infrastructure in partnership with the Global Organization for Earth System Science Portals. We have used the NCEP Reanalysis Derived data provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, from their website at http://www.esrl.noaa.gov/psd/. Edited by: M. Huber

References Ashfaq, M., Shi, Y., Tung, W., Trapp, R. J., Gao, X., Pal, J. S., and Diffenbaugh, N. S.: Suppression of south Asian summer monsoon precipitation in the 21st century, Geophys. Res. Lett., 36, L01704, doi:10.1029/2008GL036500, 2009. Auffhammer, M., Ramanathan, V., and Vincent, J. R.: Integrated model shows that atmospheric brown clouds and greenhouse gases have reduced rice harvests in India, P. Natl. Acad. Sci., 103, 19668–19672, 2006. Burns, S., Fleitmann, D., Mudelsee, M., Neff, U., Matter, A., and Mangini, A.: A 780-year annually resolved record of Indian Ocean monsoon precipitation from a speleothem from south Oman, J. Geophys. Res., 107, 4434, doi:10.1029/2001JD001281, 2002. Cherchi, A., Alessandri, A., Masina, S., and Navarra, A.: Effects of increased CO2 levels on monsoons, Clim. Dynam., 37, 83–101, doi:10.1007/s00382-010-0801-7, 2011. Cook, E. R., Anchukaitis, K. J., Buckley, B. M., Di Arrigo, R. D., Jacoby, G. C., and Wright, W. E.: Asian Monsoon Failure and Megadrought During the Last Millennium, Science, 328, 486– 439, 2010. Cubasch, U., Meehl, G., Boer, G., Stouffer, R., Dix, M., Noda, A., Senior, C., Raper, S., and Yap, K.: Projections of future climate change, in: Climate Change 2001: The Scientific Basis: Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel, edited by: Houghton, J. T., Ding, Y., Griggs, D. J., Noguer, M., Van der Linden, P. J., Dai, X., Maskell, K., and Johnson, C. A., Cambridge University Press, UK, 526–582, 2001. Duan, K., Yao, T., and Thompson, L.: Low-frequency of southern Asian monsoon variability using a 295-year record from the Dasuopu ice core in the central Himalayas, Geophys. Res. Lett., 31, L16209, doi:10.1029/2004GL020015, 2004. Fan, F., Mann, M. E., Lee, S., and Evans, J. L: Future changes in the south Asian summer monsoon: An analysis of the CMIP3 multi-model projections, J. Climate, 25, 3909–3928, 2012.

Earth Syst. Dynam., 4, 287–300, 2013

Frieler, K., Meinshausen, M., von Deimling, T., Andrews, T., and Forster, P.: Changes in global-mean precipitation in response to warming, greenhouse gas forcing and black carbon, Geophys. Res. Lett., 38, L04702, doi:10.1029/2010GL045953, 2011. Fu, C., Diaz, H., Dong, D., and Fletcher, J.: Changes in atmospheric circulation over Northern Hemisphere oceans associated with the rapid warming of the 1920s, Int. J. Climatol., 19, 581–606, 1999. Goswami, B. N.: Interannual variations of Indian summer monsoon in a GCM: External conditions versus internal feedbacks, J. Climate, 11, 501–522, 1998. Goswami, B. N., Venugopal, V., Sengupta, D., Madhusoodanan, M. S., and Xavier, P. K.: Increasing Trend of Extreme Rain Events Over India in a Warming Environment, Science, 314, 1442–1445, 2006. Guhathakurta, P. and Rajeevan, M.: Trends in the rainfall pattern over India, Int. J. Climatol., 28, 1453–1469, 2008. Hsu, P.-C., Li, T., Murakami, H., and Kitoh, A.: Future change of the global monsoon revealed from 19 CMIP5 models, J. Geophys. Res.-Atmos., 118, 1–14, doi:10.1002/jgrd.50145, 2013. Hu, Z., Latif, M., Roeckner, E., and Bengtsson, L.: Intensified Asian summer monsoon and its variability in a coupled model forced by increasing greenhouse gas concentrations, Geophys. Res. Lett., 27, 2681, doi:10.1029/2000GL011550, 2000. Jagannathan, P. and Parthasarathy, B.: Trends and periodicities of rainfall over India, Mon. Weather Rev., 101, 371–375, 1973. Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., Gandin, L., Iredell, M., Saha, S., White, G., Woollen, J., Zhu, Y., Chelliah, M., Ebisuzaki, W., Higgins, W., Janowiak, J., Mo, K. C., Ropelewski, C., Wang, J., Leetmaa, A., Reynolds, R., Jenne, R., and Joseph, D.: The NCEP/NCAR 40-Year Reanalysis Project, B. Am. Meteorol. Soc., 77, 437–471, 1996. Kitoh, A., Yukimoto, S., Noda, A., and Motoi, T.: Simulated changes in the Asian summer monsoon at times of increased atmospheric CO2 , J. Meteorol. Soc. Jpn., 75, 1019–1031, 1997. Kripalani, R., Kulkarni, A., Sabade, S., and Khandekar, M.: Indian monsoon variability in a global warming scenario, Nat. Hazards, 29, 189–206, 2003. Kripalani, R. H., Oh, J. H., Kulkarni, A., Sabade, S. S., and Chaudhari, H. S.: South Asian summer monsoon precipitation variability: Coupled climate model simulations and projections under IPCC AR4, Theor. Appl. Climatol., 90, 133–159, doi:10.1007/s00704-006-0282-0, 2007. Kucharski, F., Scaife, A. A., Yoo, J. H., Folland, C. K., Kinter, J., Knight, J., Fereday, D., Fischer, A. M., Jin, E. K., Kröger, J., Lau, N.-C., Nakaegawa, T., Nath, M. J., Pegion, P., Rozanov, E., Schubert, S., Sporyshev, P. V., Syktus, J., Voldoire, A., Yoon, J. H., Zeng, N., Zhou, T.: The CLIVAR C20C project: skill of simulating Indian monsoon rainfall on interannual to decadal timescales. Does GHG forcing play a role?, Clim. Dynam., 33, 615–627, 2009. Kumar, K., Pant, G., Parthasarathy, B., and Sontakke, N.: Spatial and subseasonal patterns of the long-term trends of Indian summer monsoon rainfall, Int. J. Climatol., 12, 257–268, 1992. Lal, M., Cubasch, U., and Santer, B.: Effect of global warming on Indian monsoon simulated with a coupled ocean-atmosphere general circulation model, Current Sci., 66, 430–438, 1994. Lal, M., Cubasch, U., Voss, R., and Waszkewitz, J.: Effect of transient increase in greenhouse gases and sulphate aerosols on monsoon climate, Current Sci., 69, 752–763, 1995.



A. Menon et al.: Consistent increase in Indian monsoon rainfall Lal, M., Nozawa, T., Emori, S., Harasawa, H., Takahashi, K., Kimoto, M., Abe-Ouchi, A., Nakajima, T., Takemura, T., and Numaguti, A.: Future climate change: Implications for Indian summer monsoon and its variability, Current Sci., 81, 1196–1207, 2001. Lee, J.-Y. and Wang, B.: Future change of global monsoon in the CMIP5, Clim. Dynam., 39, 1123–1135, 2012. Levermann, A., Schewe, J., Petoukhov, V., and Held, H.: Basic mechanism for abrupt monsoon transitions, P. Natl. Acad. Sci., 106, 20572–20577, 2009. Mahfouf, J., Cariolle, D., Royer, J., Geleyn, J., and Timbal, B.: Response of the Meteo-France climate model to changes in CO2 and sea surface temperature, Clim. Dynam., 9, 345–362, 1994. May, W.: Simulated changes of the Indian summer monsoon under enhanced greenhouse gas conditions in a global time-slice experiment, Geophys. Res. Lett., 29, 1118, doi:10.1029/2001GL013808, 2002. Meehl, G. A. and Arblaster, J. M.: Mechanisms for projected future changes in south Asian monsoon precipitation, Clim. Dynam., 21, 659–675, 2003. Meehl, G. A. and Washington, W. M.: South Asian Summer Monsoon Variability in a Model with Doubled Atmospheric Carbon Dioxide Concentration, Science, 260, 1101–1104, 1993. Meehl, G. A., Arblaster, J. M., and Tebaldi, C.: Understanding future patterns of increased precipitation intensity in climate model simulations, Geophys. Res. Lett., 32, L18719, doi:10.1029/2005GL023680, 2005. Meehl, G. A., Stocker, T. F., Collins, W. D., Friedlingstein, P., Gaye, A. T., Gregory, J. M., Kitoh, A., Knutti, R., Murphy, J. M., Noda, A., Raper, S. C. B., Watterson, I. G., Weaver, A. J., and Zhao, Z.C.: Climate Change 2007: The Physical Science Basis, Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, chap. Global Climate Projections, Cambridge University Press, Cambridge, UK and New York, NY, USA, 2007. Meehl, G. A., Arblaster, J. M., and Collins, W. D.: Effects of black carbon aerosols on the Indian monsoon, J. Climate, 21, 2869– 2882, 2008. Menon, A., Levermann, A., and Schewe, J.: Enhanced future variability during India’s rainy season, Geophys. Res. Lett., 40, 3242–3247, doi:10.1002/grl.50583, 2013. Mooley, D. and Parthasarathy, B.: Fluctuations in All-India summer monsoon rainfall during 1871–1978, Climatic Change, 6, 287– 301, 1984. Moore, J., Grinsted, A., and Jevrejeva, S.: New tools for analyzing time series relationships and trends, EOS Trans. Am. Geophys. Union, 86, 226, 2005. Moss, R. H., Edmonds, J. A., Hibbard, K. A., Manning, M. R., Rose, S. K., van Vuuren, D. P., Carter, T. R., Emori, S., Kainuma, M., Kram, T., Meehl, G. A., Mitchell, J. F. B., Nakicenovic, N., Riahi, K., Smith, S. J., Stouffer, R. J., Thomson, A. M., Weyant, J. P., and Wilbanks, T. J.: The next generation of scenarios for climate change research and assessment, Nature, 463, 747–756, 2010. Parthasarathy, B., Munot, A., and Kothawale, D.: Regression model for estimation of Indian foodgrain production from summer monsoon rainfall, Agr. Forest Meteorol., 42, 167–182, doi:10.1016/0168-1923(88)90075-5, 1988.


299

Parthasarathy, B., Munot, A., and Kothawale, D.: All-India monthly and seasonal rainfall series: 1871–1993, Theor. Appl. Climatol., 49, 217–224, 1994. Rajeevan, M., Bhate, J., Kale, J. A., and Lal, B.: High resolution daily gridded rainfall data for the Indian region: Analysis of break and active monsoon spells, Current Sci., 91, 296–306, 2006. Ramanathan, V., Chung, C., Kim, D., Bettge, T., Kiehl, J. T., Washington, W. M., Fu, Q., Sikka, D. R., and Wild, M.: Atmospheric brown clouds: Impacts on South Asian climate and hydrological cycle, P. Natl. Acad. Sci., 102, 5326–5333, 2005. Sabade, S., Kulkarni, A., and Kripalani, R.: Projected changes in South Asian summer monsoon by multi-model global warming experiments, Theor. Appl. Climatol., 103, 543–565, 2011. Schewe, J. and Levermann, A.: A statistically predictive model for future monsoon failure in India, Environ. Res. Lett., 7, 044023, doi:10.1088/1748-9326/7/4/044023, 2012. Schewe, J., Levermann, A., and Meinshausen, M.: Climate change under a scenario near 1.5 ◦ C of global warming: monsoon intensification, ocean warming and steric sea level rise, Earth Syst. Dynam., 2, 25–35, doi:10.5194/esd-2-25-2011, 2011. Schewe, J., Levermann, A., and Cheng, H.: A critical humidity threshold for monsoon transitions, Clim. Past, 8, 535–544, doi:10.5194/cp-8-535-2012, 2012. Sinha, A., Berkelhammer, M., Stott, L., Mudelsee, M., Cheng, H., and Biswas, J.: The leading mode of Indian Summer Monsoon precipitation variability during the last millennium, Geophys. Res. Lett., 38, L15703, doi:10.1029/2011GL047713, 2011. Sperber, K. and Palmer, T.: Interannual tropical rainfall variability in general circulation model simulations associated with the Atmospheric Model Intercomparison Project, J. Climate, 9, 2727– 2750, 1996. Sperber, K. R., Annamalai, H., Kang, I.-S., Kitoh, A., Moise, A., Turner, A., Wang, B., and Zhou, T.: The Asian summer monsoon: an intercomparison of CMIP5 vs. CMIP3 simulations of the late 20th century, Clim. Dynam., doi:10.1007/s00382-012-1607-6, in press, 2012. Tanaka, H., Ishizaki, N., and Nohara, D.: Intercomparison of the intensities and trends of Hadley, Walker and monsoon circulations in the global warming projections, SOLA, 1, 77–80, 2005. Thompson, L., Yao, T., Mosley-Thompson, E., Davis, M., Henderson, K., and Lin, P.: A high-resolution millennial record of the South Asian monsoon from Himalayan ice cores, Science, 289, 1916–1919, 2000. Timbal, B., Mahfouf, J., Royer, J., and Cariolle, D.: Sensitivity to prescribed changes in sea surface temperature and sea ice in doubled carbon dioxide experiments, Clim. Dynam., 12, 1–20, 1995. Trenberth, K.: Atmospheric moisture residence times and cycling: Implications for rainfall rates and climate change, Climatic Change, 39, 667–694, 1998. Turner, A. and Annamalai, H.: Climate change and the South Asian summer monsoon, Nat. Clim. Change, 2, 587–595, 2012. Ueda, H., Iwai, A., Kuwako, K., and Hori, M.: Impact of anthropogenic forcing on the Asian summer monsoon as simulated by eight GCMs, Geophys. Res. Lett., 33, L06703, doi:10.1029/2005GL025336, 2006. Webster, P.: Response of the tropical atmosphere to local, steady forcing, Mon. Weather Rev., 100, 518–541, 1972.

Earth Syst. Dynam., 4, 287–300, 2013

200

300

Appendix


Xu, H., Hong, Y., and Hong, B.: Decreasing Asian summer monsoon intensity after 1860 AD in the global warming epoch, Clim. Dynam., 39, 2079–2088, doi:10.1007/s00382-012-1378-0, 2012. Zhao, Z. and Kellogg, W.: Sensitivity of soil moisture to doubling of carbon dioxide in climate model experiments, Part II: The Asian monsoon region, J. Climte, 1, 367–378, doi:10.1175/15200442(1988)0012.0.CO;2, 1988.

Earth Syst. Dynam., 4, 287–300, 2013

Zickfeld, K., Knopf, B., Petoukhov, V., and Schellnhuber, H. J.: Is the Indian summer monsoon stable against global change?, Geophys. Res. Lett., 32, L15707, doi:10.1029/2005GL022771, 2005.


A3. The role of the North Atlantic overturning and deep ocean for multi-decadal global-mean-temperature variability Carl Friedrich Schleussner, Jacob Runge, Jascha Lehmann, and Anders Levermann. In this paper, the multi-decadal variability of the Atlantic meridional overturning circulation, the Northern Hemisphere sea-ice extent and the global mean temperature is investigated in an ensemble of CMIP5 models under control conditions. It is shown that the Atlantic meridional overturning circulation contributes to 8 percent of the global mean temperature variability with sea-ice feedbacks playing an important role. Published in Earth System Dynamics, 2014, 5(1), 103-115, doi:10.5194/esd-5-103-2014.

A3. The role of the North Atlantic overturning and deep ocean for multi-decadal

203

global-mean-temperature variability

Earth System Dynamics

Open Access

Earth Syst. Dynam., 5, 103–115, 2014 www.earth-syst-dynam.net/5/103/2014/ doi:10.5194/esd-5-103-2014 © Author(s) 2014. CC Attribution 3.0 License.

The role of the North Atlantic overturning and deep ocean for multi-decadal global-mean-temperature variability C. F. Schleussner1,2 , J. Runge1,3 , J. Lehmann1,2 , and A. Levermann1,2 1 Potsdam

Institute for Climate Impact Research, Telegrafenberg A62, 14473 Potsdam, Germany Institute, Potsdam University, Potsdam, Germany 3 Department of Physics, Humboldt University, Berlin, Germany 2 Physics

Correspondence to: C. F. Schleussner ([email protected]) Received: 7 August 2013 – Published in Earth Syst. Dynam. Discuss.: 6 September 2013 Revised: – Accepted: 11 August 2013 – Published: 20 February 2014

Abstract. Earth’s climate exhibits internal modes of variability on various timescales. Here we investigate multi-decadal variability of the Atlantic meridional overturning circulation (AMOC), Northern Hemisphere sea-ice extent and global mean temperature (GMT) in an ensemble of CMIP5 models under control conditions. We report an inter-annual GMT variability of about ±0.1◦ C originating solely from natural variability in the model ensemble. By decomposing the GMT variance into contributions of the AMOC and Northern Hemisphere sea-ice extent using a graph-theoretical statistical approach, we find the AMOC to contribute 8 % to GMT variability in the ensemble mean. Our results highlight the importance of AMOC sea-ice feedbacks that explain 5 % of the GMT variance, while the contribution solely related to the AMOC is found to be about 3 %. As a consequence of multi-decadal AMOC variability, we report substantial variations in North Atlantic deep-ocean heat content with trends of up to 0.7 × 1022 J decade−1 that are of the order of observed changes over the last decade and consistent with the reduced GMT warming trend over this period. Although these temperature anomalies are largely densitycompensated by salinity changes, we find a robust negative correlation between the AMOC and North Atlantic deepocean density with density lagging the AMOC by 5 to 11 yr in most models. While this would in principle allow for a self-sustained oscillatory behavior of the coupled AMOC– deep-ocean system, our results are inconclusive about the role of this feedback in the model ensemble.

1

Introduction

Multi-decadal variability of the climate system has been studied intensively, with a special focus on the last millennium and climate variability on centennial timescales (e.g., Eby et al., 2013; Fernández-Donado et al., 2013; Ortega et al., 2011; Fidel et al., 2011; Otterå et al., 2010; Hofer et al., 2011). Among others, a dominant mode of global climate variability is the Atlantic multi-decadal oscillation (AMO) (Schlesinger and Ramankutty, 1994) in the North Atlantic that is evident in ocean records over the last 8000 yr (Knudsen et al., 2011). The AMO as a signal of anomalous seasurface temperatures in the North Atlantic has been found to have profound influence on other climate phenomena such as the Atlantic hurricane frequency (Vimont and Kossin, 2007; Zhang and Delworth, 2009), West African monsoon and Sahel rainfall (Mohino et al., 2010). Model studies suggest that the multi-decadal mode in North Atlantic sea-surface temperatures (SSTs) is closely related to variability of the Atlantic meridional overturning circulation (AMOC) (Timmermann and Latif, 1998; Knight et al., 2005; Delworth et al., 2007; Park and Latif, 2011; Ba et al., 2013) with models exhibiting AMOC variability on multi-decadal to multi-centennial timescales (Menary et al., 2011; Wouters et al., 2012). The origin of these multidecadal modes is not yet fully understood. Already simple box models are found to produce multi-decadal variability as a result of delayed advection (Griffies and Tziperman, 1995; Lee and Wang, 2010), and Dijkstra et al. (2006) report a multi-decadal mode in models of different complexity driven by horizontal temperature anomalies (Te Raa and Dijkstra, 2002; Sévellec and Fedorov, 2013). By analyzing

Published by Copernicus Publications on behalf of the European Geosciences Union.

MPI-ESM-MR CanESM2 CCSM4 CESM1-BGC CESM1-CAM5 Mean

0.99 ± 0.01 1.01 ± 0.01 1.01 ± 0.01 0.97 ± 0.01 1.02 ± 0.01

-0.06 ± 0.01 -0.07 ± 0.01 −0.11 ± 0.01 −0.05 ± 0.01 −0.07 ± 0.01

0.57 ± 0.03 0.64 ± 0.03 0.37 ± 0.04 0.3 ± 0.04 0.49 ± 0.05

0.13 ± 0.03 0.13 ± 0.03 0.38 ± 0.04 0.39 ± 0.04 0.24 ± 0.05

0.05 0.03 0.07 0.08 0.03

0.59 0.49 0.56 0.64 0.56

0.07 0.03 0.05 0.05 0.04

8.0 6.0 5.0 11.0 6.0

0.40 0.51 0.45 0.36 0.45

0.95 0.97 0.93 0.92 0.97

1.0

−0.07

0.5

0.24

0.05

0.55

0.05

7.2

0.45

0.95

204

Appendix

104

C. F. Schleussner et al.: Multi-decadal AMOC variability Table 1. List of CMIP5 models analyzed in this study.

22 20 18 16 14 12

SIE [10 6 km 2]

o

GMT [ C]

13.8 13.6 13.4 13.2 13.0

AMOC [Sv]

MPI-ESM-LR MPI-ESM-MR CanESM2 CNRM-CM5

CCSM4 CESM1-BGC CESM1-CAM5

12

Modeling center (or group) Max Planck Institute for Meteorology Canadian Centre for Climate Modelling and Analysis Centre National de Recherches Météorologiques National Center for Atmospheric Research Community Earth System Model Contributors

Model name MPI-ESM-LR MPI-ESM-MR CanESM2 CNRM-CM5 CCSM4 CESM1-BGC CESM1-CAM5

11 10

tal subpolar gyre circulation (Levermann and Born, 2007) in a multi-model study. The existence of such a multi-stability 0.3 was found to greatly enhance subpolar variability close to the 0.0 system’s internal threshold (Mengel et al., 2012). 0.3 Atlantic variability is found to contribute significantly to Northern Hemisphere variations (Zhang et al., 2007; Knight et al., 2005, 2006) and eventually to global mean temper0.2 ature (GMT). Zanchettin et al. (2010) find multi-decadal 0.0 AMOC variations to be a major source of GMT variability − 0.2 over the last millennium, and Feulner et al. (2013) report meridional heat transport in the Atlantic to be the dominant 0 200 400 600 800 process behind the persistent temperature difference between Years the Northern Hemisphere and Southern Hemisphere in an unperturbed climate. Results of a multi-model intercomparFig. 1. Time series for GMT, AMOC, Northern Hemisphere seaFig. 1. Time series GMT,multi-decadal AMOC, Northern Hemisphere Sea-Iceison Extent (SIE),that themulti-decadal Atlantic multi-decadal oscillation (AMO) index and ice extent (SIE), the for Atlantic oscillation (AMO) insuggest variability in the high northNorth Atlantic deep-ocean (NADO) temperature for seven CMIP5 models. A 10 yr butterworth low-pass filter is applied to the time series. dex and North Atlantic deep-ocean (NADO) temperature for seven ern latitudes might be a major source of deviations within CMIP5 models. A 10 yr Butterworth low-pass filter is applied to the the CMIP5 model ensemble and in comparison with obsertime series. vational data over the 20th century (Jones et al., 2013). Besides oceanic contribution, sea-ice retreat has been a dominant contributor to observed northern high-latitude warming over the recent decades (Holland and Bitz, 2003; Screen and observational records, Dima and Lohmann (2010) find two Simmonds, 2010). distinct modes in observed Atlantic SST over the last cenHere, we first disentangle the contributions of AMOC and tury. Consistently, Park and Latif (2008) report multiple sea-ice variability to GMT variations in unperturbed control modes of AMOC variability in the Kiel Climate Model with runs of the CMIP5 model ensemble using graph-theoretical a multi-centennial mode originating in the Southern Ocean statistical models (Runge et al., 2012a; Ebert-Uphoff and and a multi-decadal one of Northern Hemisphere origin. Deng, 2012, denoted graphical models hereafter) and comWhile several model studies find stochastic atmospheric monality analysis (). We relate AMOC variability to North forcing to be the dominant driver of the Northern Hemisphere Atlantic deep-ocean temperature and salinity and investigate variability component (Eden and Willebrand, 2001; Tulloch an internal advective feedback mechanism. The magnitude and Marshall, 2012), others report variations in Labrador Sea of the deep-ocean warming rate that is found in the CMIP5 convection as a possible origin (Bentsen et al., 2004; Medmodels is consistent with the observed changes over the past haug et al., 2011; Persechino et al., 2012). Delworth and decade. The link between GMT reduction and deep-ocean Zeng (2012) and Ba et al. (2013) find salinity advection to warming is quantitatively consistent with the past reduction drive variability in Labrador Sea convection, where salinin the GMT warming trend. ity anomalies of subtropical and Arctic origin are found to play a role (Jungclaus et al., 2005). Born et al. (2012) identified this salinity advection feedback on subpolar convection as a potential driver of multi-stability of the horizono

NADOT [ C]

o

AMO [ C]

9

Earth Syst. Dynam., 5, 103–115, 2014



205

global-mean-temperature variability 12

C. F. S

C. F. Schleussner et al.: Multi-decadal AMOC variability

We analyze the unperturbed control run of seven atmosphere–ocean general circulation models (AOGCMs) from the Coupled Model Intercomparison Project Phase 5 – CMIP5 (Taylor et al., 2012) – which provides all diagnostics required for our analysis and at minimum 300 yr of model data (listed in Table 1). Figure 1 shows the annual time series for GMT, AMOC, AMO and Northern Hemisphere sea-ice extent (SIE) as well as North Atlantic deep ocean (NADO) temperature (45◦ to 65◦ N, z = 1000–2000 m). The AMOC is derived as the vertical maximum of the stream function at 45◦ N and the AMO as the anomaly of Atlantic SSTs between 30◦ and 65◦ N. We find substantial inter-annual variations for all quantities investigated, notably a GMT variability of about ±0.1◦ C (standard deviation) and a AMOC variability of ±1.1 Sv in the ensemble mean. Values for the individual models and the other quantities are given in Table 2. The corresponding power spectra are depicted in Fig. 2. We find two coherent modes in AMOC, GMT and SIE for the MPI-ESM-LR(-MR) model with around 30 and 40 (45) yr period, which are significant at the 95 % level. The CanESM2 spectrum also exhibits significant AMOC modes at around 25 and 40 yr, whereas only the 25 yr mode is present also in the GMT and SIE spectra, though not significant at the 95 % level. A 25 yr mode is also present in the CNRM-CM5 spectrum. The time series of all other models are considerably shorter (500 yr or less), which reduces the signal to noise ratio and thus leads to less or no significant spectral peaks. Despite that, the AMOC spectrum for the CESM1-BGC exhibits significant AMOC modes at around 30 and 60 yr. Cross-correlations between the AMOC and AMO, GMT and SIE reveal a clear relation between AMO and AMOC in all models with some models exhibiting an AMOC lead by several years (Fig. 3). We also find a robust positive correlation at zero or 1 yr lag between AMOC and GMT significant at the 95 % significance level as well as a negative peak between AMOC and SIE. The cross-correlation for the AMOC and GMT in CNRM-CM5 model is significant, but does not exhibit a clear peak as a consequence of a very weak AMOC mean state and weak variability on less than centennial timescales in this model (compare Fig. 1). Significant cross-correlation does not, however, imply physical relevance. To test for the importance of the high northern latitudes for absolute GMT variability, we derived the latitude-dependent contribution of surface air temperature (SAT) anomalies to the GMT signal. Figure 4 depicts the explained variance (R 2 ) of the GMT time series by the zonal and meridional integrated SAT anomalies north of the latitude drawn on the x axis. The expected surface area contribution by latitude, assuming SAT anomalies to be uniform across the globe, is drawn for comparison (in grey). We find about one third of the GMT variance to originate www.earth-syst-dynam.net/5/103/2014/

MPI-ESM-LR

CCSM4

MPI-ESM-MR

CESM1-BGC

CanESM2

CESM1-CAM5

1.5 1.0 0.5 1.5 1.0 0.5 PSD

Temperature–circulation relations in the climate model ensemble CMIP5

1.5 1.0 0.5

CNRM-CM5

1.5

AMOC GMT SIE

1.0 0.5 20

40

60 20 Period [Years]

PSD

40

95% Sign

60

Fig. 2. Power spectral density of AMOC (blue), GMT (red) and SIE

Fig. 2. Power spectral density of AMOC (blue), GMT (red) and SIE (grey). Light (grey). Light lines indicate the 95 % significance levels determined a N=10000 ensemble of a red noise first order auto-regressive process fitted to each by a N = 10 000 ensemble of a red noise first order auto-regressive relative thetocorresponding maximum 95 % process to fitted each of the quantities. In of thisthe plot, all significance amplitudes level. are rescaled relative to the corresponding maximum of the 95 % significance level.

north 0.5 of 45◦ N in the ensemble mean exceeding their share expected purely by the surface area. We applied a forward– 0.0 backward Butterworth low-pass filter with cutoff frequen0.5 10 to 50 yr to our time series and find the northcies−from MPI-ESM-LR CCSM4 ern latitude contribution to increase with increased filtering in all 0.5 models except the CCSM4 and CNRM-CM5. Half of the multi-decadal GMT variance in the MPI-ESM-LR originates0.0 north of 45◦ N, and we hypothesize AMOC and SIE − 0.5 as likely drivers. The CCSM4 actually shows a decrease for MPI-ESM-MR CESM1-BGC multi-decadal variability, but our analysis suggests the high latitudes of the Southern Hemisphere to be dominant con0.5 tributors to GMT variability in this model (compare Fig. 4). Thus,0.0 our results indicate that the high northern latitudes 0.5 an important contributor to GMT variability on are −indeed CanESM2 CESM1-CAM5 annual timescales and that this share increases for multiGMT decadal variability. We performed the same analysis with an 0.5 ensemble of 30 CMIP5 models (not shown) SIE confirming our AMO 0.0 results. Correlation

2

105

− 0.5

CNRM-CM5

NADOT 95 perc Sign

-20 -10 0 10 20 -20 -10 0 10 20 Lag in Years (pos: AMOC leads)

Fig. 3. Cross correlations between AMOC and GMT, SIE, NADO temperature and mark the two-tailed 95 % significance range. Earth Syst. Dynam., 5, 103–115, 2014

12


206

MPI-ESM-LR

CCSM4

MPI-ESM-MR

CESM1-BGC

1.5

Appendix

1.0 0.5

106

1.5


PSD

1.0 2. Inter-annual variability (standard deviation) of GMT, AMOC, Northern Hemisphere sea-ice extent (SIE), AMO and North Atlantic Table deep ocean temperature (NADOT). SAT > 45◦ N denotes the explained GMT variance integrated over the Northern Hemisphere north of 0.5 45◦ N as in Fig. 3. OHC SD: standard deviation of the North Atlantic deep ocean (NADO) (45◦ to 65◦ N, z = 1000–2000 m) heat content (OHC). OHC rate: maximum decadal trend over the full time series. CanESM2

1.5

CESM1-CAM5

GMT [◦ C]

1.0 0.5

AMOC [Sv]

SIE [106 km2 ]

AMO [◦ C]

NADOT [◦ C]

SAT > 45◦ N

OHC SD [1022 J]

OHC RATE [1022 J dec−1 ]

0.24 0.24 95%0.33 Sign 0.40 0.22 0.22 0.21

0.20 0.20 0.16 0.21 0.13 0.13 0.17

0.11 0.10 0.13 0.14 0.06 0.06 0.10

0.27 0.27 0.33 0.33 0.36 0.39 0.39

0.20 0.21 0.23 0.31 0.16 0.20 0.22

1.1 1.0 0.6 0.5 0.5 0.5 0.7

0.17

0.10

0.33

0.22

0.7

MPI-ESM-LR 0.12 1.60 MPI-ESM-MR 0.10 1.35 CNRM-CM5 CanESM2 0.12 0.86 PSD CNRM-CM5 0.11 AMOC 0.99 GMT CCSM4 0.10 0.87 SIE CESM1-BGC 0.10 0.93 CESM1-CAM5 0.11 1.00

1.5 1.0 0.5 20

40

Mean

60 20 40 0.11 1.09 Period [Years]

60

0.27

Fig. 2. Power spectral density of AMOC (blue), GMT (red) and SIE (grey). Light lines indicate the 95 % significance levels determined by 3 Quantifying AMOC contributions to GMT variance a N=10000 ensemble of a red noise first order auto-regressive process fitted to each of the quantities. In this plot, all amplitudes are rescaled relative to the corresponding maximum of the 95 % significance level. Cross correlations need to be interpreted cautiously, since

0.5 0.0 − 0.5

MPI-ESM-LR

CCSM4

MPI-ESM-MR

CESM1-BGC

0.5 0.0 Correlation

− 0.5 0.5

they are subject to several limitations. In the presence of strong auto-correlation, peaks in the cross-correlation can be amplified and shifted to larger time lags, limiting the interpretability of time delays in such systems (Runge et al., 2014). Additionally, in the global climate system with its complex coupled dynamic, also common driver effects and indirect chains can occur. A notional process Z that drives two processes X and Y is called a common driver of X and Y , whereas an indirect chain is present if process X drives Z and Z drives Y , without a direct relation between X and Y . In both cases, a non-zero cross-correlation between X and Y would be detected, even though no direct interaction is present. 3.1

Graphical model analysis

0.0

To overcome these limitations, graphical models can be used to identify causal relations in complex coupled systems CanESM2 CESM1-CAM5 (Ebert-Uphoff and Deng, 2012). Here we apply a graphiGMT cal model approach introduced by Runge et al. (2012a, b). 0.5 SIE While this approach is also applicable to nonlinear interacAMO 0.0 tions, here the linear case as discussed in Runge et al. (2014) NADOT − 0.5 is used to determine linear causal interdependencies. In the 95 perc Sign CNRM-CM5 following, we illustrate the concept with a simple example -20 -10 0 10 20 -20 -10 0 10 20 and refer to the references for further details on the methodLag in Years (pos: AMOC leads) ology. Consider the simple bivariate first-order auto-regressive Fig. 3. Cross-correlations between AMOC and GMT, SIE, NADO stochastic (AR(1)): Fig. 3. Cross correlations between AMOC and GMT, SIE, NADO temperature process and the AMO index (positive: AMOC leads). The grey lines − 0.5

temperature and the AMO index (positive: AMOC leads). The grey

mark two-tailed 95 %95 significance range. lines the mark the two-tailed % significance range.

Xt = a Xt−1 + tX

(1)

Yt = cXt−1 + bYt−1 + tY , where X and Y denote the two variables of the stochastic process with a, b and c as constant coefficients. The coefficients a and b represent the auto-regressive contribution of X and Earth Syst. Dynam., 5, 103–115, 2014


0 10 20

0.75 0.50 0.25 CNRM-CM5 − 45

0

45

− 45 Latitude

0

50 Lat

45

Fig. 4. Meridional dependence of the GMT variance. The horizontal averaged yearly SAT anomalies are integrated from

A3. The role of the North Atlantic overturning and deep ocean multi-decadal The value for a given latitudefor marks the explained variance (R ) of the GMT time series by the integrated SAT variance 207 assuming equally distributed G indicate different low-pass filters applied. The grey line marks the variance contribution explained variances of the unfiltered GMT signal that originates from north of 45 N are listed in Table 2. global-mean-temperature variability 2

◦



R2

1.00 0.75 0.50 0.25 MPI-ESM-LR 1.00 Latitude 0.75 0.50 0.25 MPI-ESM-MR 1.00 Latitude 0.75 0.50 0.25 CanESM2 1.00 Latitude 0.75 0.50 0.25 CNRM-CM5

13

107

CCSM4 Latitude Fig. 5. Representation of the time-series graph of the simple cou-

Fig. 5. Representation of the time series graph of the simple coupled bivariate process (Xt , Yt ) in Eq. (1) with the con pled bivariate process (Xt , Yt )ofintheEq. (1) with the constant coeffib and c. σXY is the non-diagonal element covariance matrix. Process Yt (black node) is driven by Yt−1 and Xt− a, b and is the Due non-diagonal of the covaria cients contemporaneous linkc.to σ XXY to stationarity, element the same causal relation holds for all t. t (hatched).

CESM1-BGC

ance matrix. Process Yt (black node) is driven by Yt−1 and Xt−1 (in grey) and has a contemporaneous link to Xt (hatched). Due to stationarity, the same causal relation holds for all t.

Latitude

CESM1-CAM5

a time-series graph algorithm as introduced in Runge et al. (2012b) can help to identify causalities in such complex coupled systems. As an illustration, consider the estimation of the parents of Y in the time-series graph shown in Fig. 5. The algorithm tests possible links from all processes (including Y ) at all lags up to a maximum delay (which can − 45 0 45 − 45 0 45 be specified accordingly; for the analysis presented below Latitude we use a maximum time lag of 30 yr). Here the hypothetical link “Xt−2 → Yt ” would be tested by first checking Fig. 4. Meridional dependence of the GMT variance. The horiwhether the yearly unconditional cross-correlation ρ(Xfrom Yt )◦ N is southward. t−2 ; 90 ◦ Fig. 4. Meridional dependence of the GMT variance. The horizontal averaged SAT anomalies are integrated zontally averaged yearly SAT anomalies are integrated from 90 N 2 nonzero. Since there exists a path between X and Y , it. Colors t−2 The value for givenforlatitude explained variance (R ) of the GMT time series by the integrated SAT variance northt of southward. Theavalue a givenmarks latitudethe marks the explained vari2 this will be the case. In the next iteration step, the conance (R different ) of the GMT time series the integrated SATline variance indicate low-pass filtersby applied. The grey marks the variance contribution assuming equally distributed GMT variance. The ditional dependence tested. north of it.variances Colors indicate filters that applied. The from explained of the different unfilteredlow-pass GMT signal originates north linear of 45◦ N are listed inisTable 2. As a heuristic criterion for selecting the conditions in each test, we choose the grey line marks the variance contribution assuming equally disconditions sorted by their correlation value in the previous tributed GMT variance. The explained variances of the unfiltered GMT signal that originates from north of 45◦ N are listed in Tastep. Assuming ρ(Yt−1 ; Yt ) > ρ(Xt−1 ; Yt ) > . . . in our exble 2. ample, the partial correlation ρ(Xt−2 ; Yt |Yt−1 ) that excludes the influence of Yt−1 would be tested. This partial correlation would also be non-zero due to the unblocked path Y respectively, and |a|, |b| < 1 is required for the process to “Xt−2 → Xt−1 → Yt ”. Also the test with the next-largest be stationary. c gives the coupling between the two and can condition on Xt−1 yields a non-zero partial correlation. AfX Y in principle attain any finite value. t and t are indepenter some more tests with weaker conditions, the partial cordent, identically distributed Gaussian random variables with relation ρ(Xt−2 ; Yt |Yt−1 , Xt−1 ) would be found to vanish, a given covariance matrix, such that the hypothetical link can be removed. An analogue 2 procedure is applied for contemporaneous links, where all σX σXY 6= (2) 2 directed links are identified and iteratively more and more σXY σY contemporaneous links are conditioned out (Runge et al., 2012b). and zero mean. Following this procedure, we now construct a time-series A time-series graph for this model can be constructed, graph for GMT, AMOC and SIE with a two-sided signifiwhere the nodes are represented by the time-dependent states cance level of 99 % for the partial correlation test. The reXt , Yt , Xt−1 , Yt−1 , (. . .). Directed links (symbolized by →) sulting cross-links between AMOC, GMT and SIE are illusare given by non-zero coefficients a, b, c, and undirected trated in Fig. 6. Across the whole ensemble, we robustly find contemporaneous links are given by non-zero entries in the contemporaneous links with a positive partial correlation beinverse covariance matrix 6. This time-series graph for our Fig. 5. system Representation of thein time graph of time-series the simple coupled bivariate process (Xt , Yand (1) with the constant coefficients a, tween AMOC and GMT a Eq. negative partial correlation 5. Note that model is illustrated Fig.series t ) in bstationarity and c. σXYas isa prerequisite the non-diagonal element of the covariance ProcessSIE Yt (black node)Several is driven by Yt−1 Xt−1 (in grey) and has between and GMT. models alsoand show negative for such an analysis implies thatmatrix. aifcontemporaneous Due the same causal relation t. links at a time lag of 1 yr. “GMT → SIE” andholds “SIEfor → all GMT” “Xt−τ → Yt ”, “Xlink →XYtt 0(hatched). ” is true for anytot 0stationarity, . t 0 −τ to On the one hand sea-ice reduction will lead to atmospheric In real world systems, however, the exact relations bewarming (Holland and Bitz, 2003; Screen and Simmonds, tween the investigated processes are often unknown, and www.earth-syst-dynam.net/5/103/2014/

0 10 20

50 Lat

Earth Syst. Dynam., 5, 103–115, 2014

208

Appendix 14

C. F. Schleussner et al.: Multi-decadal AMOC

108

C. F. Schleussner et al.: Multi-decadal AMOC variability CNRM-CM5

CanESM2

CCSM4

MPI-ESM-MR

Positive Negative MPI-ESM-LR

CESM1-CAM5

CESM1-BGC

Ensemble

robustness 100% 20%

Fig. 6. Resulting time-series graphs for AMOC, GMT and SIE by applying a method introduced in Runge et al. (2012b). Based on the sign of the partial correlation as a result of the graph estimation algorithm, we denote positive (negative) partial correlations in red (blue). Direct Fig. 6. Resulting time graphs AMOC, and SIE contemporaneous by applying a method in Runge et al. (2012b). Based links are indicated by curved lines with theseries associated timefor lags. StraightGMT lines represent links at introduced lag zero.

Density Anomalies [kg/m 3]

of the partial correlation as a result of the graph estimation algorithm, we denote positive (negative) partial correlations in red (b links are indicated by curved lines with the associated time lags. Straight lines represent contemporaneous links at lag zero. 2010), but a warming anomaly could also cause additional contribution at lag zero. Following , the unique contribution sea-ice melt in particular during the summer and early auof AMOC to GMT variability UAMOC derives as tumn season. Thus, influences in both directions are present 2 2 UAMOC = RFULL − RSIE , (3) that could explain the model dependent presence of the lag-1 links. 2 where RFULL is the coefficient of determination (giving The relation between AMOC and SIE differs substanthe explained variance) for the multivariate regression of tially between the models, exhibiting links at zero lag as 2 the GMT on AMOC and SIE at lag zero. RSIE is the exwell as “AMOC → SIE” and “SIE → AMOC” links at time plained variance of the univariate regression of GMT on lag 1. A stronger AMOC leads to increased northward heat SIE at lag zero. The unique SIE contribution USIE can be transport and thus to SIE reduction in the North Atlantic, CCSM4 derived accordingly. 0.04 which could explain the AMOC–SIE links. LevermannMPI-ESM-LR et al. The common contribution C is given by the difference be(2007) identified increased oceanic heat loss as a result of 2 0.00 tween RFULL and the sum of the unique contributions. All reduced SIE in the North Atlantic to be a positive feedback time series have been standardized (time series mean subon AMOC strength −in0.04 a CMIP3 model ensemble, implying tracted and divided by the standard deviation). In the endriving mechanisms that work both ways. Our results indisemble average, we unique AMOC contribution of MPI-ESM-MR CESM1-BGC 250 500 750 100 200find a300 400 cate that the strength of this individual processes differs sub0.04 3 %, SIE 14 % and a common component of 5 %. In total, stantially between the models investigated. Direct coupling the coupled AMOC–SIE system contributes about 21 % to between AMOC and 0.00 SIE is located in the subpolar and poGMT variance (see Table 3 for the coefficients of the individlar Atlantic, while SIE comprises the sea-ice extent over the − 0.04 ual models). We performed the same analysis for low-passfull Northern Hemisphere. Thus, an AMOC imprint might filtered time series 200 and found a decrease in the AMOC unique be present, but not significant in the compound SIE signal. CanESM2 CESM1-CAM5 250 500 750 100 300 400 0.04 contribution to 1 %, while the SIE and the common compoTo account for this, we performed the same analysis for SIE nent increase to 21 and 9 % respectively, in total explaining 0.00 region and found similar results. limited to the North Atlantic 31 % of GMT variability. 0.04 We would like to emphasize that the analysis presented 3.2 Commonality−analysis here is not a comprehensive study of Northern Hemisphere CNRM-CM5 500 100 200 400 We use commonality 0.04 analysis () to 250 disentangle AMOC 750 and climate variability, since 300 important sources of variability Temp-component SIE contributions to GMT variability. Commonality analysis such as multi-decadal variability in the Pacific (Mantua et al., Salt-component [inv] 0.00 allows a distinction of the explained variance by a multivari1997) as well as continental variability are not included. Also Full density ate system into unique components that are explained by the interrelations between modes of atmospheric variability such − 0.04 individual variables alone and common components that give as the North Atlantic and the Arctic oscillation (Hurrell and 250between 500 100 and 200 Deser, 2009) AMOC 300 as well 400 as SIE, which have been the explained variance by the coupling them. 750 Years intensively and shown to have substantial influence As depicted in Fig. 6, the contemporaneous positive studied AMOC–GMT and negative SIE–GMT links are robust on North Atlantic climate variability (e.g., Tulloch and Maracross all models ensemble. we density limit thetime series shall, (black) 2012; Frankcombe et al., 2009; Medhaug al., 2011), (red) related com Fig.in7.the North AtlanticTherefore, Deep Ocean and its salinity (blue, inverted) andettemperature decomposition of the GMT variability to the AMOC and SIE are not resolved. However, we find that SIE and AMOC Earth Syst. Dynam., 5, 103–115, 2014


A3. The role of the North Atlantic overturning and deep ocean for multi-decadal global-mean-temperature variability

209


109

Table 3. Results of the commonality analysis of GMT variability and the AMOC and SIE contribution at lag zero. UAMOC and USIE denote the unique AMOC and SIE contribution and C the common 2 component. RFULL gives the explained GMT variance by the coupled system.

except for the CanESM2), indicating a systematic process behind the phenomenon. As shown in Fig. 9, we find deepocean density changes to be related to the AMOC in our model ensemble. Except for the CNRM-CM5 model that has a very weak AMOC and exhibits no significant multidecadal modes of AMOC variability as discussed above, all models show a positive correlation between the AMOC and the NADO thermal density component at zero lag and a significant anti-correlation with a time lag between 10 and 20 yr. Cross-correlations between AMOC and NADO salinity at lag zero are generally weaker over the model ensemble with only the MPI-ESM-LR and MPI-ESM-MR as well as the CanESM2 model exhibiting a significant negative peak. Still, most models show a positive correlation at multidecadal timescales. Taken together, these results indicate that a strong AMOC leads to a NADO warming and salinification on multi-decadal timescales. In the combined density signal, however, only the MPI-ESM-LR model shows a significant AMOC NADO density peak between lag 10 and 20, while a robust correlation at zero lag is found for most models in the ensemble. Again we performed a time-series graph analysis as in Sect. 3.1 to account for the autocorrelations in both time series. Our analysis yields a positive contemporaneous AMOC NADO density link and directed negative AMOC → NADO density links with 5 to 11 yr time lag that are significant at the 99 % level for all models except the CNRM-CM5 (not shown). The robust zero lag link between AMOC and NADO density suggests a causal relation. A linear relationship between meridional density differences and AMOC strength is assumed in conceptual approaches (Stommel, 1961; Rahmstorf, 1996; Johnson et al., 2007; Marzeion and Drange, 2006; Fürst and Levermann, 2012; Cimatoribus et al., 2012) and confirmed by a variety of studies with models of different levels of complexity (e.g., Manabe and Stouffer, 1988; Thorpe et al., 2001; Griesel and Maqueda, 2006; Dijkstra, 2008; Huisman et al., 2010). The physical mechanism behind this relationship is controversial in a predominantly geostrophic ocean, although a number of explanations for this relation have been proposed (Marotzke, 1997; Gnanadesikan, 1999; Schewe and Levermann, 2009; Sijp et al., 2012). Gregory and Tailleux (2011) present a kinetic energy analysis of the AMOC, finding the pressure gradient force to be a dominant driver of the AMOC in the high northern latitudes by conversion of potential to kinetic energy. They report good correlation between changes in the North Atlantic pressure gradient force and the AMOC under CO2 -forced climate change on decadal timescales (their Fig. 10) and find changes in the pressure gradient force to be dominated by buoyancy changes in the northern North Atlantic, which is confirmed by Saenko (2013) for the CanESM2 model. Consequently, they propose a linear relation 1M ∝ 1ρ in the North Atlantic. Such a scaling is also supported by an analytical study by Sijp et al. (2012). Thus, Gregory and Tailleux (2011) provide a physical mechanism for the zero-lag link between

MPI-ESM-LR MPI-ESM-MR CanESM2 CNRM-CM5 CCSM4 CESM1-BGC CESM1-CAM5 Mean

UAMOC

USIE

C

2 RFULL

0.01 0.02 < 0.01 0.02 0.06 0.06 0.02

0.13 0.09 0.20 0.18 0.08 0.07 0.22

0.03 0.05 0.01 0.06 0.03 0.05 0.08

0.16 0.16 0.21 0.26 0.17 0.19 0.32

0.03

0.14

0.05

0.21

variability already contributes about two thirds to the 33 % explained GMT variance that is found to originate from north of 45◦ N. 4

Deep ocean heat uptake and density

As shown in the previous section, a significant part of multidecadal GMT variability is due to changes in the North Atlantic. But when investigating oceanic contributions to GMT variability, direct surface feedbacks as well as oceanic heat uptake variability have to be considered (Levitus et al., 2000; Meehl et al., 2011; Balmaseda et al., 2013). As shown in Fig. 1, we find substantial variations in North Atlantic deep-ocean (NADO) temperatures (45◦ to 65◦ N, z = 1000– 2000 m). NADO heat content variability ranges from a standard deviation of 0.16×1022 J for the CCSM4 to 0.31×1022 J for the CNRM-CM5 model with an ensemble mean of 0.22× 1022 J. Numbers for the individual models are given in Table 4. Our results compare well with Mauritzen et al. (2012), who find a peak-to-peak variability of about 1×1022 J for the entire extratropical North Atlantic deep ocean (700–2000 m) between 20◦ and 65◦ N. Such anomalies in heat content will also lead to substantial variations in NADO density as depicted in Fig. 7 (red curve). However, density variations caused by temperature change are largely compensated by salinity changes (in blue, inverted) leading to a weaker overall density signal (in black) in line with observational studies (Curry et al., 1998; Yashayaev et al., 2007). Figure 8 depicts the strength of the density compensation, and we find substantial variance between the model ensemble (slope of −1 indicates full compensation), ranging from full or even slight overcompensation for the CNRM-CM5 and the CESM1-CAM5 model to only partial compensation in the CCSM4 and the MPI-ESMLR. Apart from the CNRM-CM5 and the CESM1-CAM5 model, all models show substantial deviation from full compensation (which are significant at one standard deviation www.earth-syst-dynam.net/5/103/2014/

Earth Syst. Dynam., 5, 103–115, 2014

CNRM-CM5

CanESM2

CCSM4

MPI-ESM-MR

Positive Negative

210 MPI-ESM-LR

CESM1-CAM5

CESM1-BGC

Appendix

Ensemble

robustness

100% C. F. Schleussner et al.: Multi-decadal AMOC variability 20%

110

Table 4. Results of the multivariate regression of the conceptual coupled AMOC–NADO density system as in Eq. (4) to the individual models. Fig. 6. Resulting time series graphs for AMOC, GMT and SIE by applying a method introduced in Runge et al. (2012b). Based on the sig of the partial correlation as a result of the graph estimation algorithm, we denote positive (negative) partial correlations in red (blue). Dire 2 2 2 aρ cMρlines with the aMassociatedctime σM τ contemporaneous RM Rρ2 ρM lags. σStraight ρ links are indicated by curved linesσMρ represent links at lag zero. MPI-ESM-LR MPI-ESM-MR CanESM2 CCSM4 CESM1-BGC CESM1-CAM5

0.98 ± 0.01 0.99 ± 0.01 1.01 ± 0.01 1.01 ± 0.01 0.97 ± 0.01 1.02 ± 0.01

−0.08 ± 0.01 −0.06 ± 0.01 −0.07 ± 0.01 −0.11 ± 0.01 −0.05 ± 0.01 −0.07 ± 0.01

0.61 ± 0.03 0.57 ± 0.03 0.64 ± 0.03 0.37 ± 0.04 0.3 ± 0.04 0.49 ± 0.05

0.18 ± 0.03 0.13 ± 0.03 0.13 ± 0.03 0.38 ± 0.04 0.39 ± 0.04 0.24 ± 0.05

0.07 0.05 0.03 0.07 0.08 0.03

0.48 0.59 0.49 0.56 0.64 0.56

0.04 0.07 0.03 0.05 0.05 0.04

7.0 8.0 6.0 5.0 11.0 6.0

0.52 0.40 0.51 0.45 0.36 0.45

0.93 0.95 0.97 0.93 0.92 0.97

1.0

−0.07

0.5

0.24

0.05

0.55

0.05

7.2

0.45

0.95

Mean

MPI-ESM-LR

0.04

CCSM4

0.00 − 0.04

Density Anomalies [kg/m 3]

0.04

MPI-ESM-MR 750

250

500

250

500 CanESM2 750

250

500

250

500

100

200

300CESM1-BGC 400

100

200

CESM1-CAM5 300 400

0.00 − 0.04 0.04 0.00 − 0.04 0.04

CNRM-CM5 750

100

0.00 − 0.04 750

Years

100

200 300 400 Temp-component Salt-component [inv] Full density 200

300

400

Fig. 7. North Atlantic deep ocean density time series (black) and its salinity-related (blue, inverted) and temperature-related (red) compoFig. 7. North Atlantic Deep Ocean density time series (black) and its salinity (blue, inverted) and temperature (red) related components. nents.

NADO and AMOC that is found in all models in the ensemble. Additionally, also a direct feedback between AMOC and NADO temperature and salinity via subpolar convection might play a role. Increased sea surface salinity as a consequence of a strengthened AMOC could lead to enhanced subpolar convection (Mengel et al., 2012) and as a consequence to a deep-ocean freshening and cooling (Yashayaev, 2007). Based on our findings, we study a conceptual bivariate statistical model for the change in NADO density 1ρ and the AMOC change 1M: ρ

1ρt = aρ 1ρt−1 + cMρ 1Mt−τ + t

(4)

1Mt = aM 1Mt−1 + cρM 1ρt + tM , with aρ and aM denoting the auto-regressive coefficients and cρM and cMρ the coupling coefficients for 1ρ → 1M at lag zero and 1M → 1ρ with a time lag τ . Results of the fit to the model ensemble time series are given in Table 4. In the Earth Syst. Dynam., 5, 103–115, 2014

ensemble mean, we find about 95 % of the density and about 45 % of the AMOC variance explained by this simple conceptual model, although these good fitting results are partly due to the very strong auto-correlation in the system. Equation (4) can be written as an auto-regressive system, for which power spectra can be derived analytically (Brockwell and Davis, 2009). Figure 10 depicts analytical spectra for the conceptual model (FULL, green) in comparison with a system without the AMOC–NADO density link (e.g., because of complete density compensation by salinity; cMρ = 0, in blue) and an AR(1) process fitted to the AMOC time series (grey). Note that the cMρ = 0 and AR(1) coefficients may differ from the fully coupled system given in Table 4. We find an enhancement of multi-decadal variability for FULL vs. cMρ = 0 in all models, but only for the MPI-ESM-LR, the CCSM4 and the CESM1-BGC, it also exceeds the spectral density of the fitted AR(1) process for periods between



211

C. F. Schleussner et al.: Multi-decadal AMOC variability global-mean-temperature variability C. F. Schleussner et al.: Multi-decadal AMOC variability

15


0.00 0.04 − 0.04 0.00 − 0.04

MPI-ESM-MR

a = -0.66

MPI-ESM-MR a = -0.77

CanESM2

a = -0.77

a = -0.82

CanESM2

CNRM-CM5

a = -0.82

a = -0.56 a = -0.56

CCSM4

CESM1-BGC

CESM1-BGC a = -0.81 a = -0.81

CESM1-CAM5

CESM1-CAM5 a = -1.07 a = -1.07

CNRM-CM5

a = -1.05 a− =0.04 -1.05 0.00 − 0.04

0.04

− 0.04

0.00

0.04

Temperature [kg/m ] 0.04 0.00 0.04 component − 0.04 0.00 3

Temperature component [kg/m 3]

0.5 1.0 1.5

a = -0.66

MPI-ESM-LR

CCSM4

PSD 0.5 1.0 1.5

3 Salinity component ] Salinity component [kg/m 3[kg/m ]

0.00 0.04 − 0.04 0.00 0.04 − 0.04 0.00 0.04 − 0.04 0.00 0.04 − 0.04 0.00 0.04 − 0.04 0.00 − 0.04 0.04

111

0.5 1.0 1.5

MPI-ESM-LR

0.04

15

MPI-ESM-LR

CCSM4

MPI-ESM-MR

CESM1-BGC

CanESM2

20

40

FULL c Mρ= 0

60

CESM1-CAM5

80 20 Period [Years]

40

AR(1) 2x c Mρ

60

80

Fig. 10. The Analytically derived power spectral densities differFig. of temperaturetemperatureand andsalinity-driven salinity driven NADO density changes. blue linederived marks power the linear fit with slope for a. a = −1 line systems r Fig. 8. 8. Relation Relation of NADO density Fig. 10. Analytically spectral densities forThe different model ent model systems regressed on the individual model time series: is drawn in grey. changes. The blue line marks the linear fit with slope a. The a = −1 coupled AMOC-NADO system, blue: system c = 0, purple Fig. 8. Relation of temperature and salinity driven NADO density changes. The blue line marksdensity the linear fit with slopedecoupled a. The a = −1 line green: fully coupled AMOC–NADO density system, blue: decou-

is drawn in grey. isline drawn in grey.

mρ

coupling (2 × cmρ AMOCAR(1) spectra derived from the model time series rescaled pled system cmρ = ,). 0, purple: process, dashed red: a doubled grey. AMOC–NADO coupling (2 × cmρ ,). AMOC spectra derived from the model time series rescaled relative to their maximum amplitude are shown in light grey.

0.5

0.5 0.0 0.0 − 0.5

MPI-ESM-LR

MPI-ESM-LR

CCSM4

CCSM4

Correlation Correlation

30 and 70 yr. The MPI-ESM-LR and the CCSM4 show the weakest density compensation in the ensemble with slightly higher values for the CESM1-BGC (compare Fig. 8), indi− 0.5 MPI-ESM-MR CESM1-BGC cating that a coupled NADO density–AMOC mode might 0.5 MPI-ESM-MR CESM1-BGC contribute to multi-decadal variability in models with limited 0.5 0.0 density compensation of temperature and salinity anoma0.0 lies. All three models also show pronounced spectral peaks − 0.5 − 0.5 with periods below 30 yr likely as a result of other sources CanESM2 CESM1-CAM5 of AMOC variability. One possible mechanism is variable 0.5 CanESM2 CESM1-CAM5 Labrador Sea convection that has been found to dominate 0.5 0.0 the 20 yr AMOC mode for the CCSM4 model (Danabasoglu 0.0 et al., 2012). For the MPI-ESM-MR, the CanESM2 and the − 0.5 − 0.5 CESM1-CAM5, our conceptual model only shows slightly enhanced variability in comparison with a simple AR(1) proCNRM-CM5 Temp-comp CNRM-CM5 0.5 cess, suggesting minor importance of the process. For ilTemp-comp 0.5 Salt-comp lustration purposes, we also depict spectra with enhanced Salt-comp 0.0 Full Density 0.0 Full Density AMOC–NADO density coupling (2 × cMρ in red, dashed) − 0.5 95 perc Sign − 0.5 95 perc Sign that show enhanced multi-decadal variability for all models. As discussed above, a variety of other processes related − 20 0 20 − 20 0 20 − 20 0 20 − 20 0 20 to horizontal temperature (Te Raa and Dijkstra, 2002), salinLag in Years (pos: AMOC leads) Lag in Years (pos: AMOC leads) ity advection (Jungclaus et al., 2005) and atmospheric variFig. 9. Cross-correlations between AMOC and NADO density and ability (Eden and Willebrand, 2001) are found to be relFig. 9. Cross correlations correlations betweenAMOC AMOCand andNADO NADOdensity density and temperature salinity related components (positive: AMOC leads). Fig. 9. Cross between and itsits temperature andand salinity related components (positive: its temperatureand salinity-related components (positive: AMOC evant for multi-decadal AMOC variability and areAMOC not in-leads). The grey lines mark the two-tailed 95%%significance significance range. The grey lines mark two-tailed 95 range. leads). The grey linesthe mark the two-tailed 95 % significance range. cluded in this simple conceptual system. However, we find substantial changes in the NADO heat content as a result www.earth-syst-dynam.net/5/103/2014/

Earth Syst. Dynam., 5, 103–115, 2014

212

Appendix

112


of multi-decadal AMOC variability, which are only partly density-compensated in most models. Based on these results, we propose a conceptual model relating NADO density changes and the AMOC and find the coupled system to enhance multi-decadal variability in three models in the ensemble.

3 yr in our model ensemble, not shown) and deep-ocean heat content in the northern North Atlantic. Meehl et al. (2011) investigated deep-ocean heat uptake anomalies during decades exhibiting a hiatus in the GMT over the 21st century in global warming model simulations and found a difference of about 1.3 × 1022 J for the global deep-ocean heat uptake rate per decade below 750 m and 0.2 × 1022 J for the Atlantic Basin compared to reference decades. We find anomalous NADO heat uptake with a maximum uptake trend of 0.7×1022 J decade−1 averaged over the model ensemble with a range from 1.1 × 1022 J decade−1 for the MPI-ESM-LR and 0.5×1022 J decade−1 for the CCSM4, CNRM-CM5 and the CESM1-BGC. Our results indicate that North Atlantic deep-ocean heat uptake anomalies connected to a hiatus decade in GMT increase as identified by Meehl et al. (2011) are very well within the range of natural variability of our model ensemble. As a consequence of an AMOC strengthening related to a positives AMO signal in the late 1990s (Parker et al., 2007), our analysis would suggest NADO heat content to rise for more than a decade after a peak in circulation strength, which is in good agreement with observations from the North Atlantic (Mauritzen et al., 2012). Studies by Guemas et al. (2013) and Balmaseda et al. (2013) demonstrate that ocean heat uptake plays a crucial role in understanding the GMT hiatus over the last decade. Although they find the dominant signal in the upper Pacific Ocean likely related to El Niño– Southern Oscillation variability, also the North Atlantic deep ocean contributed substantially to anomalous global ocean heat uptake over the last decade (compare Balmaseda et al., 2013, their Fig. S06). While most of the NADO temperature signal is densitycompensated through changes in salinity, we find a substantial model spread regarding the strength of this compensation. We use time-series graph analysis to extract causal relations out of the highly auto-correlated time series for AMOC and NADO density and found a robust positive link at zero time lag and a negative link between AMOC and NADO density with time lags between 5 and 11 yr (AMOC leads). Based on these results, we propose a stochastic bivariate model that we fit to the time series and find it to explain about 95 % of the density and about 45 % of the AMOC variance in the ensemble average. The AMOC–NADO density feedbacks identified may lead to multi-decadal AMOC variability in models with weak density compensation. Thus, they represent a possible advective mechanism for multi-decadal AMOC variability (Latif, 1998; Te Raa and Dijkstra, 2002; Menary et al., 2011).

5

Discussion and conclusions

By resolving the latitudinal contributions to GMT variance in unperturbed simulations of seven CMIP5 models, we find 33 % of the variance to originate north of 45◦ N in the ensemble mean, which exceeds the share expected by the surface area. Using a time-series graph analysis approach as in Sect. 3.1 (Runge et al., 2012b, 2014), we identify statistically robust couplings between global mean temperature (GMT), Northern Hemisphere sea-ice extent (SIE) and the Atlantic meridional overturning circulation (AMOC) at zero lag and find AMOC and SIE to explain about 21 % of GMT variance in the model ensemble mean. Applying commonality analysis in Sect. 3.2, we disentangle AMOC and SIE contributions and report the contribution that can solely be attributed to the AMOC (SIE) alone to be about 3 % (14 %) in the ensemble mean. Additionally, we find AMOC and SIE coupling to explain 5 % of the GMT variance suggesting that coupled AMOC–SIE feedbacks might play an important role for North Atlantic climate variability. Brönnimann (2009) discusses the influence of natural variability during the early 20th century warming and highlights the importance of Arctic warming for this global phenomenon. While anthropogenic aerosol emissions affecting patterns of atmospheric variability are discussed as a possible driver of this anomaly (Booth et al., 2012), our results suggest that also natural multi-decadal AMOC variability could have played an important role (Zhang et al., 2013). The Arctic temperature anomaly coincides with a positive AMO index (Parker et al., 2007) and a sea-ice retreat in the North Atlantic (Macias Fauria et al., 2009; Bengtsson et al., 2004; Semenov and Latif, 2012) that match well with the AMOC– sea-ice coupling identified. The drastic reduction of Northern Hemisphere SIE as a result of anthropogenic global warming (Stroeve et al., 2011, 2012; Notz and Marotzke, 2012) will likely lead to decreased contribution of the high northern latitudes to GMT variability, since SIE is found to be a dominant contributor to GMT variance in an unperturbed climate. Additionally, AMOC–SIE feedbacks might weaken in warmer future climates, and consequently also the impact of multidecadal AMOC variability on GMT might be reduced. In addition to the AMOC influence on surface temperature variability, we find North Atlantic deep-ocean (NADO) heat content to be highly correlated with the AMOC on decadal timescales. Using observational data, Mauritzen et al. (2012) also report a decadal time lag between upper ocean (which is highly correlated with the AMOC at lags between zero and Earth Syst. Dynam., 5, 103–115, 2014

Acknowledgements. The work was supported by the Federal Ministry for the Environment, Nature Conservation and Nuclear Safety (11 II 093 Global A SIDS and LDCs). Carl-Friedrich Schleussner was funded by the Deutsche Bundesstiftung Umwelt. Jakob Runge appreciates support by the German National Academic Foundation and the DFG grant no. KU34-1. On



213


113

http://tocsy.pik-potsdam.de/tigramite.php we provide a program with a graphical user interface to estimate the time-series graph. We acknowledge the World Climate Research Programme’s Working Group on Coupled Modelling, which is responsible for CMIP, and we thank the climate modeling groups (listed in Table 1 of this paper) for producing and making available their model output. For CMIP the US Department of Energy’s Program for Climate Model Diagnosis and Intercomparison provides coordinating support and led development of software infrastructure in partnership with the Global Organization for Earth System Science Portals.

Delworth, T., Zhang, R., and Mann, M.: Decadal to centennial variability of the Atlantic from observations and models, ocean circulation: mechanisms and impacts, Geophys. Monogr. Ser., 173, 131–148, 2007. Dijkstra, H. A.: Scaling of the Atlantic meridional overturning circulation in a global ocean model, Tellus A, 60, 749–760, 2008. Dijkstra, H. A., Raa, L., Schmeits, M., and Gerrits, J.: On the physics of the Atlantic multidecadal oscillation, Ocean Dynam., 56, 36–50, doi:10.1007/s10236-005-0043-0, 2006. Dima, M. and Lohmann, G.: Evidence for two distinct modes of large-scale ocean circulation changes over the last century, J. Climate, 23, 5–16, doi:10.1175/2009JCLI2867.1, 2010. Ebert-Uphoff, I. and Deng, Y.: Causal discovery for climate research using graphical models, J. Climate, 25, 5648–5665, doi:10.1175/JCLI-D-11-00387.1, 2012. Eby, M., Weaver, A. J., Alexander, K., Zickfeld, K., Abe-Ouchi, A., Cimatoribus, A. A., Crespin, E., Drijfhout, S. S., Edwards, N. R., Eliseev, A. V., Feulner, G., Fichefet, T., Forest, C. E., Goosse, H., Holden, P. B., Joos, F., Kawamiya, M., Kicklighter, D., Kienert, H., Matsumoto, K., Mokhov, I. I., Monier, E., Olsen, S. M., Pedersen, J. O. P., Perrette, M., Philippon-Berthier, G., Ridgwell, A., Schlosser, A., Schneider von Deimling, T., Shaffer, G., Smith, R. S., Spahni, R., Sokolov, A. P., Steinacher, M., Tachiiri, K., Tokos, K., Yoshimori, M., Zeng, N., and Zhao, F.: Historical and idealized climate model experiments: an intercomparison of Earth system models of intermediate complexity, Clim. Past, 9, 1111– 1140, doi:10.5194/cp-9-1111-2013, 2013. Eden, C. and Willebrand, J.: Mechanism of interannual to decadal variability of the North Atlantic circulation, J. Climate, 14, 2266– 2280, 2001. Fernández-Donado, L., González-Rouco, J. F., Raible, C. C., Ammann, C. M., Barriopedro, D., García-Bustamante, E., Jungclaus, J. H., Lorenz, S. J., Luterbacher, J., Phipps, S. J., Servonnat, J., Swingedouw, D., Tett, S. F. B., Wagner, S., Yiou, P., and Zorita, E.: Large-scale temperature response to external forcing in simulations and reconstructions of the last millennium, Clim. Past, 9, 393–421, doi:10.5194/cp-9-393-2013, 2013. Feulner, G., Rahmstorf, S., Levermann, A., and Volkwardt, S.: On the origin of the surface air temperature difference between the hemispheres in earth’s present-day climate, J. Climate, 26, 7136– 7150 doi:10.1175/JCLI-D-12-00636.1, 2013. Fidel, J., González-Rouco, L., Gonzalez-Rouco, J. F., FernándezDonado, L., Raible, C. C., Barriopedro, D., Lutherbacher, J., Jungclaus, J. H., Swingedow, D. Servonnat, J. Zorita, E. Wagner, S., and Ammann, C. M.: Medieval climate anomaly to Little Ice Age transition as simulated by current climate models, Pages News, 19, 3–4, 2011. Frankcombe, L. M., Dijkstra, H. A., and von der Heydt, A.: Noise-induced multidecadal variability in the North Atlantic: excitation of normal modes, J. Phys. Oceanogr., 39, 220–233, doi:10.1175/2008JPO3951.1, 2009. Fürst, J. J., and Levermann, A.: A minimal model for wind- and mixing-driven overturning: threshold behavior for both driving mechanisms, Clim. Dynam., 38, 239–260, doi:10.1007/s00382011-1003-7, 2012. Gnanadesikan, A.: A simple predictive model for the structure of the oceanic pycnocline, Science, 283, 2077–2079, 1999.

Edited by: V. Lucarini

References Ba, J., Keenlyside, N. S., Park, W., Latif, M., Hawkins, E., and Ding, H.: A mechanism for Atlantic multidecadal variability in the Kiel Climate Model, Clim. Dynam., 41, 2133–2144, doi:10.1007/s00382-012-1633-4, 2013. Balmaseda, M. A., Trenberth, K. E., and Källén, E.: Distinctive climate signals in reanalysis of global ocean heat content, Geophys. Res. Lett., 40, 1754–1759, doi:10.1002/grl.50382, 2013. Bengtsson, L., Semenov, V. A., and Johannessen, O. M.: The early twentieth-century warming in the Arctic – a possible mechanism, J. Climate, 17, 4045–4057, 2004. Bentsen, M., Drange, H., Furevik, T., and Zhou, T.: Simulated variability of the Atlantic meridional overturning circulation, Clim. Dynam., 22, 701–720, doi:10.1007/s00382-004-0397-x, 2004. Booth, B. B. B., Dunstone, N. J., Halloran, P. R., Andrews, T., and Bellouin, N.: Aerosols implicated as a prime driver of twentiethcentury North Atlantic climate variability., Nature, 484, 228–32, doi:10.1038/nature10946, 2012. Born, A., Stocker, T. F., Raible, C. C., and Levermann, A.: Is the Atlantic subpolar gyre bistable in comprehensive coupled climate models?, Clim. Dynam., 40, 2993–3007, doi:10.1007/s00382012-1525-7, 2012. Brockwell, P. J. and Davis, R. A.: Time series: theory and methods, Springer, New York, 2009. Brönnimann, S.: Early twentieth-century warming, Nat. Geosci., 2, 735–736, doi:10.1038/ngeo670, 2009. Cimatoribus, A. A., Drijfhout, S. S., and Dijkstra, H. A.: Meridional overturning circulation: stability and ocean feedbacks in a box model, Clim. Dynam., doi:10.1007/s00382-012-1576-9, online first, 2012. Curry, R., McCartney, M., and Joyce, T.: Oceanic transport of subpolar climate signals to mid-depth subtropical waters, Nature, 391, 575–577, 1998. Danabasoglu, G., Yeager, S. G., Kwon, Y.-O., Tribbia, J. J., Phillips, A. S., and Hurrell, J. W.: Variability of the Atlantic meridional overturning circulation in CCSM4, J. Climate, 25, 5153–5172, doi:10.1175/JCLI-D-11-00463.1, 2012. Delworth, T. L. and Zeng, F.: Multicentennial variability of the Atlantic meridional overturning circulation and its climatic influence in a 4000 year simulation of the GFDL CM2.1 climate model, Geophys. Res. Lett., 39, 1–6, doi:10.1029/2012GL052107, 2012.


Earth Syst. Dynam., 5, 103–115, 2014

214

Appendix

114


Gregory, J. M. and Tailleux, R.: Kinetic energy analysis of the response of the Atlantic meridional overturning circulation to CO2 -forced climate change, Clim. Dynam., 37, 893–914, doi:10.1007/s00382-010-0847-6, 2011. Griesel, A. and Maqueda, M. A. M.: The relation of meridional pressure gradients to North Atlantic deep water volume transport in an ocean general circulation model, Clim. Dynam., 26, 781–799, doi:10.1007/s00382-006-0122-z, 2006. Griffies, S. and Tziperman, E.: A linear thermohaline oscillator driven by stochastic atmospheric forcing, J. Climate, 8, 2440– 2453, 1995. Guemas, V., Doblas-Reyes, F. J., Andreu-Burillo, I., and Asif, M.: Retrospective prediction of the global warming slowdown in the past decade, Nat. Clim. Change, 3, 649–653, doi:10.1038/nclimate1863, 2013. Hofer, D., Raible, C. C., and Stocker, T. F.: Variations of the Atlantic meridional overturning circulation in control and transient simulations of the last millennium, Clim. Past, 7, 133–150, doi:10.5194/cp-7-133-2011, 2011. Holland, M. M. and Bitz, C. M.: Polar amplification of climate change in coupled models, Clim. Dynam., 21, 221–232, doi:10.1007/s00382-003-0332-6, 2003. Huisman, S. E., den Toom, M., Dijkstra, H. A., and Drijfhout, S.: An indicator of the multiple equilibria regime of the Atlantic meridional overturning circulation, J. Phys. Oceanogr., 40, 551–567, doi:10.1175/2009JPO4215.1, 2010. Hurrell, J. and Deser, C.: North Atlantic climate variability: the role of the North Atlantic Oscillation, J. Marine Syst., 78, 28–41, doi:10.1016/j.jmarsys.2008.11.026, 2009. Johnson, H. L., Marshall, D. P., and Sproson, D. A. J.: Reconciling theories of a mechanically driven meridional overturning circulation with thermohaline forcing and multiple equilibria, Clim. Dynam., 29, 821–836, doi:10.1007/s00382-007-0262-9, 2007. Jones, G. S., Stott, P. A., and Christidis, N.: Attribution of observed historical near-surface temperature variations to anthropogenic and natural causes using CMIP5 simulations, J. Geophys. Res.Atmos., 118, 4001–4024, doi:10.1002/jgrd.50239, 2013. Jungclaus, J., Haak, H., Latif, M., and Mikolajewicz, U.: Arctic– North Atlantic interactions and multidecadal variability of the meridional overturning circulation, J. Climate, 18, 4013–4031, 2005. Knight, J. R., Allen, R. J., Folland, C. K., Vellinga, M., and Mann, M. E.: A signature of persistent natural thermohaline circulation cycles in observed climate, Geophys. Res. Lett., 32, L20708, doi:10.1029/2005GL024233, 2005. Knight, J. R., Folland, C. K., and Scaife, A. A.: Climate impacts of the Atlantic multidecadal oscillation, Geophys. Res. Lett., 33, L17706, doi:10.1029/2006GL026242, 2006. Knudsen, M. F., Seidenkrantz, M.-S., Jacobsen, B. H., and Kuijpers, A.: Tracking the Atlantic Multidecadal Oscillation through the last 8,000 years, Nat. Comm., 2, 178, doi:10.1038/ncomms1186, 2011. Latif, M.: Dynamics of interdecadal variability in coupled ocean – atmosphere models, J. Climate, 11, 602–624, 1998. Lee, S. and Wang, C.: Delayed advective oscillation of the Atlantic thermohaline circulation, J. Climate, 23, 1254, doi:10.1175/2009JCLI3339.1, 2010.

Levermann, A. and Born, A.: Bistability of the subpolar gyre in a coarse resolution climate model, Geophys. Res. Lett., 34, L24605, doi:10.1029/2007GL031732, 2007. Levermann, A., Mignot, J., Nawrath, S., and Rahmstorf, S.: The role of Northern Sea ice cover for the weakening of the thermohaline circulation under global warming, J. Climate, 20, 4160– 4171, doi:10.1175/JCLI4232.1, 2007. Levitus, S., Antonov, J., Boyer, T., and Stephens, C.: Warming of the world ocean, Science, 287, 2225–2229, doi:10.1126/science.287.5461.2225, 2000. Macias Fauria, M., Grinsted, A., Helama, S., Moore, J., Timonen, M., Martma, T., Isaksson, E., and Eronen, M.: Unprecedented low twentieth century winter sea ice extent in the Western Nordic Seas since A. D. 1200, Clim. Dynam., 34, 781–795, doi:10.1007/s00382-009-0610-z, 2009. Manabe, S. and Stouffer, R. J.: Two stable equilibria of a coupled ocean-atmosphere model, J. Climate, 1, 841–866, 1988. Mantua, N. J., Hare, S. R., Zhang, Y., Wallace, J. M., and Francis, R. C.: A Pacific interdecadal climate oscillation with impacts on salmon production, B. Am. Meteorol. Soc., 78, 1069–1079, 1997. Marotzke, J.: Boundary mixing and the dynamics of threedimensional thermohaline circulations, J. Phys. Oceanogr., 27, 1713–1728, 1997. Marzeion, B. and Drange, H.: Diapycnal mixing in a conceptual model of the Atlantic meridional overturning circulation, DeepSea Res. Pt. II, 53, 226–238, 2006. Mauritzen, C., Melsom, A., and Sutton, R. T.: Importance of density-compensated temperature change for deep North Atlantic Ocean heat uptake, Nat. Geosci.,5, 905–910, doi:10.1038/ngeo1639, 2012. Medhaug, I., Langehaug, H. R., Eldevik, T., Furevik, T., and Bentsen, M.: Mechanisms for decadal scale variability in a simulated Atlantic meridional overturning circulation, Clim. Dynam., 39, 77–93, doi:10.1007/s00382-011-1124-z, 2011. Meehl, G. A., Arblaster, J. M., Fasullo, J. T., Hu, A., and Trenberth, K. E.: Model-based evidence of deep-ocean heat uptake during surface-temperature hiatus periods, Nat. Clim. Change, 1, 360–364, doi:10.1038/nclimate1229, 2011. Menary, M. B., Park, W., Lohmann, K., Vellinga, M., Palmer, M. D., Latif, M., and Jungclaus, J. H.: A multimodel comparison of centennial Atlantic meridional overturning circulation variability, Clim. Dynam., 38, 2377–2388, doi:10.1007/s00382-011-1172-4, 2011. Mengel, M., Levermann, A., Schleussner, C.-F., and Born, A.: Enhanced Atlantic subpolar gyre variability through baroclinic threshold in a coarse resolution model, Earth Syst. Dynam., 3, 189–197, doi:10.5194/esd-3-189-2012, 2012. Mohino, E., Janicot, S., and Bader, J.: Sahel rainfall and decadal to multi-decadal sea surface temperature variability, Clim. Dynam., 37, 419–440, doi:10.1007/s00382-010-0867-2, 2010. Notz, D. and Marotzke, J.: Observations reveal external driver for Arctic sea-ice retreat, Geophys. Res. Lett., 39, L08502, doi:10.1029/2012GL051094, 2012. Ortega, P., Montoya, M., González-Rouco, F., Mignot, J., and Legutke, S.: Variability of the Atlantic meridional overturning circulation in the last millennium and two IPCC scenarios, Clim. Dynam., 38, 1925–1947, doi:10.1007/s00382-011-1081-6, 2011.

Earth Syst. Dynam., 5, 103–115, 2014



215


115

Otterå, O., Bentsen, M., Drange, H., and Suo, L.: External forcing as a metronome for Atlantic multidecadal variability, Nat. Geosci., 3, 688–694, doi:10.1038/ngeo955, 2010. Park, W. and Latif, M.: Multidecadal and multicentennial variability of the meridional overturning circulation, Geophys. Res. Lett., 35, L22703, doi:10.1029/2008GL035779, 2008. Park, W. and Latif, M.: Atlantic Meridional Overturning Circulation response to idealized external forcing, Clim. Dynam., 39, 1709– 1726, doi:10.1007/s00382-011-1212-0, 2011. Parker, D., Folland, C., Scaife, A., Knight, J., Colman, A., Baines, P., and Dong, B.: Decadal to multidecadal variability and the climate change background, J. Geophys. Res., 112, 1– 18, doi:10.1029/2007JD008411, 2007. Persechino, A., Marsh, R., Sinha, B., Megann, A. P., Blaker, A. T., and New, A. L.: Decadal-timescale changes of the Atlantic overturning circulation and climate in a coupled climate model with a hybrid-coordinate ocean component, Clim. Dynam., 39, 1021– 1042, doi:10.1007/s00382-012-1432-y, 2012. Rahmstorf, S.: On the freshwater forcing and transport of the Atlantic thermohaline circulation, Clim. Dynam., 12, 799–811, 1996. Runge, J., Heitzig, J., Marwan, N., and Kurths, J.: Quantifying causal coupling strength: a lag-specific measure for multivariate time series related to transfer entropy, Phys. Rev. E, 86, 061121, doi:10.1103/PhysRevE.86.061121, 2012a. Runge, J., Heitzig, J., Petoukhov, V., and Kurths, J.: Escaping the curse of dimensionality in estimating multivariate transfer entropy, Phys. Rev. Lett., 108, 1–5, doi:10.1103/PhysRevLett.108.258701, 2012b. Runge, J., Petoukhov, V., and Kurths, J.: “Quantifying the strength and delay of climatic interactions: the ambiguities of cross correlation and a novel measure based on graphical models”, J. Climate, 27, 720–739, 2014. Saenko, O.: Energetics of weakening and recovery of the Atlantic overturning in a climate change simulation, Geophys. Res. Lett., 40, 888–892, doi:10.1002/grl.50101, 2013. Schewe, J. and Levermann, A.: The role of meridional density differences for a wind-driven overturning circulation, Clim. Dynam., 34, 547–556, doi:10.1007/s00382-009-0572-1, 2009. Schlesinger, M. and Ramankutty, N.: An oscillation in the global climate system of period 65–70 years, Nature, 367, 723–726, doi:10.1038/367723a0, 1994. Screen, J. A. and Simmonds, I.: The central role of diminishing sea ice in recent Arctic temperature amplification, Nature, 464, 1334–1337, doi:10.1038/nature09051, 2010. Seibold, D. and McPhee, R.: Commonality analysis: a method for decomposing explained variance in multiple regression analyses, Human Comm. Res., 5, 355–365, 1979. Semenov, V. A. and Latif, M.: The early twentieth century warming and winter Arctic sea ice, The Cryosphere, 6, 1231–1237, doi:10.5194/tc-6-1231-2012, 2012. Sévellec, F., and Fedorov, A. V.: The leading, interdecadal eigenmode of the Atlantic meridional overturning circulation in a realistic ocean model, J. Climate, 26, 2160–2183, doi:10.1175/JCLID-11-00023.1, 2013. Sijp, W. P., Gregory, J. M., Tailleux, R., and Spence, P.: The key role of the western boundary in linking the AMOC strength to the north–south pressure gradient, J. Phys. Oceanogr., 42, 628– 643, doi:10.1175/JPO-D-11-0113.1, 2012.

Stommel, H.: Thermohaline convection with two stable regimes of flow, Tellus, 13, 224–230, 1961. Stroeve, J. C., Serreze, M. C., Holland, M. M., Kay, J. E., Malanik, J., and Barrett, A. P.: The Arctic’s rapidly shrinking sea ice cover: a research synthesis, Clim. Change, 110, 1005–1027, doi:10.1007/s10584-011-0101-1, 2011. Stroeve, J. C., Kattsov, V., Barrett, A., Serreze, M., Pavlova, T., Holland, M., and Meier, W. N.: Trends in Arctic sea ice extent from CMIP5, CMIP3 and observations, Geophys. Res. Lett., 39, L16502, doi:10.1029/2012GL052676, 2012. Taylor, K. E., Stouffer, R. J., and Meehl, G. A.: An overview of CMIP5 and the experiment design, B. Am. Meteorol. Soc., 93, 485–498, 2012. Te Raa, L. A. and Dijkstra, H. A.: Instability of the thermohaline ocean circulation on interdecadal timescales, J. Phys. Oceanogr., 32, 138–160, 2002. Thorpe, R., Gregory, J., and Johns, T.: Mechanisms determining the Atlantic thermohaline circulation response to greenhouse gas forcing in a non-flux-adjusted coupled climate model, J. Climate, 14, 3102–3116, 2001. Timmermann, A. and Latif, M.: Northern hemispheric interdecadal variability: a coupled air–sea mode, J. Climate, 11, 1906–1931, 1998. Tulloch, R. and Marshall, J.: Exploring mechanisms of variability and predictability of Atlantic meridional overturning circulation in two coupled climate models, J. Climate, 25, 4067–4080, doi:10.1175/JCLI-D-11-00460.1, 2012. Vimont, D. J. and Kossin, J. P.: The Atlantic Meridional Mode and hurricane activity, Geophys. Res. Lett., 34, L07709, doi:10.1029/2007GL029683, 2007. Wouters, B., Drijfhout, S., and Hazeleger, W.: Interdecadal NorthAtlantic meridional overturning circulation variability in ECEARTH, Clim. Dynam., 39, 2695–2712, doi:10.1007/s00382012-1366-4, 2012. Yashayaev, I.: Hydrographic changes in the Labrador Sea, 1960–2005, Prog. Oceanogr., 73, 242–276, doi:10.1016/j.pocean.2007.04.015, 2007. Yashayaev, I., van Aken, H. M., Holliday, N. P., and Bersch, M.: Transformation of the Labrador sea water in the subpolar North Atlantic, Geophys. Res. Lett., 34, L22605, doi:10.1029/2007GL031812, 2007. Zanchettin, D., Rubino, A., and Jungclaus, J. H.: Intermittent multidecadal-to-centennial fluctuations dominate global temperature evolution over the last millennium, Geophys. Res. Lett., 37, L14702, doi:10.1029/2010GL043717, 2010. Zhang, R. and Delworth, T. L.: A new method for attributing climate variations over the Atlantic Hurricane Basin’s main development region, Geophys. Res. Lett., 36, L06701, doi:10.1029/2009GL037260, 2009. Zhang, R., Delworth, T., and Held, I.: Can the Atlantic Ocean drive the observed multidecadal variability in Northern Hemisphere mean temperature?, Geophys. Res. Lett., 34, L02709, doi:10.1029/2006GL028683, 2007. Zhang, R., Delworth, T. L., Sutton, R., Hodson, D. L., Dixon, K. W., Held, I. M., Kushnir, Y., Marshall, J., Ming, Y., and Msadek, R.: Have aerosols caused the observed Atlantic multidecadal variability?, J. Atmos. Sci., 70, 1135–1144, doi:10.1175/JAS-D-120331.1, 2013.


Earth Syst. Dynam., 5, 103–115, 2014

Bibliography Alexander, L. V. et al. (2006), Global observed changes in daily climate extremes of temperature and precipitation, J. Geophys. Res., 111(D5), D05109, doi:10.1029/2005JD006290. Allan, R. P., B. J. Soden, V. O. John, W. Ingram, and P. Good (2010), Current changes in tropical precipitation, Environ. Res. Lett., 5(2), 025205, doi:10.1088/1748-9326/5/2/025205. Allen, M. R., and W. J. Ingram (2002), Constraints on future changes in climate and the hydrologic cycle., Nature, 419(6903), 224–32, doi:10.1038/nature01092. Barnes, E. a. (2013), Revisiting the evidence linking Arctic amplification to extreme weather in midlatitudes, Geophys. Res. Lett., 40(August), 4734–4739, doi:10.1002/grl.50880. Benestad, R. (2003), How often can we expect a record event?, Clim. Res., 25, 3–13. Bergaoui, K., D. Mitchell, R. Zaaboul, R. McDonnell, F. Otto, and M. Allen (2015), The Contribution of Human-Induced Climate Change To the Drought of 2014 in the Southern Levant Region [in “Explaining Extremes of 2014 from a Climate Perspective”], Bull. Am. Meteorol. Soc., 96(12), S66–S70, doi:10.1175/BAMS-D-15-00129.1. Blackmon, M. (1976), A climatological spectral study of the 500 mb geopotential height of the Northern Hemisphere, J. Atmos. Sci., 33, 1607–1623. Bladé, I., B. Liebmann, D. Fortuny, and G. J. van Oldenborgh (2012), Observed and simulated impacts of the summer NAO in Europe: Implications for projected drying in the Mediterranean region, Clim. Dyn., 39(3), 709–727, doi:10.1007/s00382-011-1195-x. Boer, G. J., and T. G. Shepherd (1983), Large-Scale Two-Dimensional Turbulence in the Atmosphere, J. Atmos. Sci., 40(1), 164–184, doi:10.1175/1520-0469(1983)0402.0.CO;2. Cohen, J. et al. (2014), Recent Arctic amplification and extreme mid-latitude weather, Nat. Geosci., (August), doi:10.1038/ngeo2234. Coumou, D., and S. Rahmstorf (2012), A decade of weather extremes, Nat. Clim. Chang., 2, 491–496, doi:10.1038/nclimate1452. Coumou, D., and A. Robinson (2013), Historic and future increase in the global land area affected by monthly heat extremes, Environ. Res. Lett., 8(3), 034018, doi:10.1088/1748-9326/8/3/034018. Coumou, D., A. Robinson, and S. Rahmstorf (2013), Global increase in record-breaking monthly-mean temperatures, Clim. Change, 118(3-4), 771–782, doi:10.1007/s10584-012-0668-1. Coumou, D., V. Petoukhov, S. Rahmstorf, S. Petri, and H. J. Schellnhuber (2014), Quasi-resonant circulation regimes and hemispheric synchronization of extreme weather in boreal summer, Proc. Natl. Acad. Sci., 1412797111–, doi:10.1073/pnas.1412797111. Coumou, D., J. Lehmann, and J. Beckmann (2015), The weakening summer circulation in the Northern Hemisphere mid-latitudes, Science, 348(6232), 324–327, doi:10.1126/science.1261768. Delworth, T. L., and F. Zeng (2014), Regional rainfall decline in Australia attributed to anthropogenic greenhouse gases and ozone levels, Nat. Geosci., 7(8), 583–587, doi:10.1038/ngeo2201. Diffenbaugh, N. S., D. L. Swain, and D. Touma (2015), Anthropogenic warming has increased drought risk in California, Proc. Natl. Acad. Sci., 112(13), 3931–3936, doi:10.1073/pnas.1422385112. Dong, B., R. T. Sutton, T. Woollings, and K. Hodges (2013), Variability of the North Atlantic summer storm track: mechanisms and impacts on European climate, Environ. Res. Lett., 8(3), 034037, doi:10.1088/1748-9326/8/3/034037. Eady, E. T. (1949), Long Waves and Cyclone Waves, Tellus A, 1, 33–52, doi:10.3402/tellusa.v1i3.8507. Fischer, E. M., and R. Knutti (2015), Anthropogenic contribution to global occurrence of heavy-precipitation and high-temperature extremes, Nat. Clim. Chang., (April), 1–6, doi:10.1038/nclimate2617.

218

Bibliography

Folland, C. K., J. Knight, H. W. Linderholm, D. Fereday, S. Ineson, and J. W. Hurrel (2009), The summer North Atlantic oscillation: Past, present, and future, J. Clim., 22(5), 1082–1103, doi:10.1175/2008JCLI2459.1. Fraedrich, K., and H. Böttger (1978), A Wavenumber-Frequency Analysis of the 500 mb Geopotential at 50°N, J. Atmos. Sci., 35(4), 745–750, doi:10.1175/1520-0469(1978)0352.0.CO;2. Frieler, K., M. Meinshausen, T. Schneider Von Deimling, T. Andrews, and P. Forster (2011), Changes in global-mean precipitation in response to warming, greenhouse gas forcing and black carbon, Geophys. Res. Lett., 38(4), 1–5, doi:10.1029/2010GL045953. GISTEMP (2015), GISS Surface Temperature http://data.giss.nasa.gov/gistemp/

Analysis

(GISTEMP),

Available

from:

Gulev, S., and O. Zolina (2001), Extratropical cyclone variability in the Northern Hemisphere winter from the NCEP/NCAR reanalysis data, Clim. Dyn., 17(10), 795–809. Hansen, J., M. Sato, and R. Ruedy (2012), Perception of climate change, Proc. Natl. Acad. Sci., 109(37), E2415–E2423, doi:10.1073/pnas.1205276109. Harvey, B. J., L. C. Shaffrey, and T. J. Woollings (2013), Equator-to-pole temperature differences and the extra-tropical storm track responses of the CMIP5 climate models, Clim. Dyn., doi:10.1007/s00382013-1883-9. Hawcroft, M. K., L. C. Shaffrey, K. I. Hodges, and H. F. Dacre (2012), How much Northern Hemisphere precipitation is associated with extratropical cyclones?, Geophys. Res. Lett., 39(24), L24809, doi:10.1029/2012GL053866. Held, I. M., and B. J. Soden (2006), Robust responses of the hydrological cycle to global warming, J. Clim., 19(21), 5686–5699, doi:10.1175/JCLI3990.1. Herring, S. C., M. P. Hoerling, J. P. Kossin, T. C. Peterson, and P. A. Stott (2015), Explaining Extreme Events of 2014 from a Climate Perspective, Bull. Am. Meteorol. Soc., 96(12), S1–S172. Holton, J. R. (2004), An Introduction to Dynamic Meteorology, fourth edi., Elsevier Academic Press. Horton, D. E., N. C. Johnson, D. Singh, D. L. Swain, B. Rajaratnam, and N. S. Diffenbaugh (2015), Contribution of changes in atmospheric circulation patterns to extreme temperature trends, Nature, 522(7557), 465–469, doi:10.1038/nature14550. Hoskins, B., and P. Valdes (1990), On the existence of storm-tracks, J. Atmos. Sci., 47(15), 1854–1864, doi:10.1175/ 1520-0469(1990)047,1854:OTEOST.2.0.CO;2. Hurrell, J. & N. C. for A. R. S. (Eds). (2015), The Climate Data Guide: Hurrell North Atlantic Oscillation (NAO) Index (station-based), Available from: https://climatedataguide.ucar.edu/climate-data/hurrell-northatlantic-oscillation-nao-index-station-based#sthash.n4V17FsG.dpuf Hurrell, J. W. (1995), Decadal trends in the north atlantic oscillation: regional temperatures and precipitation., Science, 269(5224), 676–679, doi:10.1126/science.269.5224.676. IPCC (2013), Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, edited by T. F. Stocker, D. Qin, G.-K. Plattner, M. Tignor, S. K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex, and P. M. Midgley, Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. Jacobeit, J., J. Rathmann, A. Philipp, and P. D. Jones (2009), Central European prEcipitation and temperature extremes in relation to large-scale atmospheric circulation types, Meteorol. Zeitschrift, 18(4), 397– 410, doi:10.1127/0941-2948/2009/0390. Joughin, I., B. E. Smith, and B. Medley (2014), Marine Ice Sheet ollapse potentially under way for the Thwaites Glacier Bassin, West Antarctica, Science (80-. )., 344(May), 735–738. Kelley, C. P., S. Mohtadi, M. A. Cane, R. Seager, and Y. Kushnir (2015), Climate change in the Fertile Crescent and implications of the recent Syrian drought, Proc. Natl. Acad. Sci., 112(11), 3241–3246, doi:10.1073/pnas.1421533112.

219

Ki-moon, B. (2012), Secretary General ’ s remarks to UNFCCC COP18 High Level, 2012. Available from: http://www.un.org/sg/statements/index.asp?nid=6482 Kintisch, E. (2014), Into the maelstrom., Science, 344(April), 250–3, doi:10.1126/science.344.6181.250. Knutti, R., and J. Sedláček (2012), Robustness and uncertainties in the new CMIP5 climate model projections, Nat. Clim. Chang., 3(4), 369–373, doi:10.1038/nclimate1716. Kretschmer, M., D. Coumou, J. F. Donges, and J. Runge (2016), Using Causal Effect Networks to analyze different Arctic drivers of mid-latitude winter circulation, J. Clim. (in rev.). Lehmann, J., and D. Coumou (2015), The influence of mid-latitude storm tracks on hot, cold, dry and wet extremes, Sci. Rep., 5, 17491, doi:10.1038/srep17491. Lehmann, J., D. Coumou, K. Frieler, A. V Eliseev, and A. Levermann (2014), Future changes in extratropical storm tracks and baroclinicity under climate change, Environ. Res. Lett., 9(8), 084002, doi:10.1088/1748-9326/9/8/084002. Lehmann, J., D. Coumou, and K. Frieler (2015), Increased record-breaking precipitation events under global warming, Clim. Change, doi:10.1007/s10584-015-1434-y. Lehmann, J., F. Mempel, and D. Coumou (2016), Increased occurrence of record-wet months in high latitudes and record-dry months in subtropics, PNAS. Lewis, S. C., and D. J. Karoly (2013), Anthropogenic contributions to Australia’s record summer temperatures of 2013, Geophys. Res. Lett., 40(14), 3705–3709, doi:10.1002/grl.50673. Lindzen, R., and B. Farrell (1980), A simple approximate result for the maximum growth rate of baroclinic instabilities, J. Atmos. Sci., 37, 1648–1654, doi:0022-4928/80/071648-07$05.75. Liu, C., and R. P. Allan (2013), Observed and simulated precipitation responses in wet and dry regions 1850– 2100, Environ. Res. Lett., 8(3), 034002, doi:10.1088/1748-9326/8/3/034002. Luber, G. et al. (2014), Ch. 9: Human Health. Climate Change Impacts in the United States: The Third National Climate Assessment. Luterbacher, J. et al. (2010), Circulation dynamics and its influence on European and Mediterranean January-April climate over the past half millennium: Results and insights from instrumental data, documentary evidence and coupled climate models, Clim. Change, 101(1), 201–234, doi:10.1007/s10584-009-9782-0. Mahlstein, I., O. Martius, C. Chevalier, and D. Ginsbourger (2012), Changes in the odds of extreme events in the Atlantic basin depending on the position of the extratropical jet, Geophys. Res. Lett., 39(22), L22805, doi:10.1029/2012GL053993. Marshall, J., Y. Kushnir, D. Battisti, P. Chang, A. Czaja, R. Dickson, J. Hurrell, M. McCartney, R. Saravanan, and M. Visbeck (2001), North Atlantic climate variability: phenomena, impacts and mechanisms, Int. J. Climatol., 21(15), 1863–1898, doi:10.1002/joc.693. Min, S.-K., X. Zhang, F. W. Zwiers, and G. C. Hegerl (2011), Human contribution to more-intense precipitation extremes., Nature, 470(7334), 378–81, doi:10.1038/nature09763. Munich Re (2013), Economic Consequences of Natural Catastrophes: Emerging and Developing Economies Particularly Affected - Insurance Cover is Essential, Position Pap., (October), 1–9. Munich Re (2015), NatCatSERVICE Schadenereignisse weltweit 1980 – 2014, NASA (2014), Wild , Wild Winds, Available from: http://svs.gsfc.nasa.gov/vis/a010000/a011400/a011471/ Neu, U. et al. (2013), IMILAST: A Community Effort to Intercompare Extratropical Cyclone Detection and Tracking Algorithms, Bull. Am. Meteorol. Soc., 94(4), 529–547, doi:10.1175/BAMS-D-11-00154.1. Otto, F. E. L., N. Massey, G. J. van Oldenborgh, R. G. Jones, and M. R. Allen (2012), Reconciling two approaches to attribution of the 2010 Russian heat wave, Geophys. Res. Lett., 39(4), n/a–n/a, doi:10.1029/2011GL050422.

220

Bibliography

Pall, P., M. R. Allen, and D. a. Stone (2007), Testing the Clausius–Clapeyron constraint on changes in extreme precipitation under CO2 warming, Clim. Dyn., 28(4), 351–363, doi:10.1007/s00382-006-01802. Peixoto, J. P. and O., and A. H. (1992), Physics of Climate, American Institute of Physics, New York. Petoukhov, V., S. Rahmstorf, S. Petri, and H. J. Schellnhuber (2013), Quasiresonant amplification of planetary waves and recent Northern Hemisphere weather extremes., Proc. Natl. Acad. Sci. U. S. A., 110(14), 5336–41, doi:10.1073/pnas.1222000110. Rahmstorf, S., and D. Coumou (2011), Increase of extreme events in a warming world., Proc. Natl. Acad. Sci. U. S. A., 108(44), 17905–9, doi:10.1073/pnas.1101766108. Rignot, E., J. Mouginot, M. Morlighem, H. Seroussi, and B. Scheuchl (2014), Widespread, rapid grounding line retreat of Pine Island, Thwaites, Smith, and Kohler glaciers, West Antarctica, from 1992 to 2011, Geophys. Res. Lett., 41(10), 3502–3509, doi:10.1002/2014GL060140. Runge, J., V. Petoukhov, J. F. Donges, J. Hlinka, N. Jajcay, M. Vejmelka, D. Hartman, N. Marwan, M. Paluš, and J. Kurths (2015), Identifying causal gateways and mediators in complex spatio-temporal systems, Nat. Commun., 6, 8502, doi:10.1038/ncomms9502. Schleussner, C. F., J. Runge, J. Lehmann, and A. Levermann (2014), The role of the North Atlantic overturning and deep ocean for multi-decadal global-mean-temperature variability, Earth Syst. Dyn., 5(1), 103–115, doi:10.5194/esd-5-103-2014. Screen, J. A., and I. Simmonds (2014), Amplified mid-latitude planetary waves favour particular regional weather extremes, Nat. Clim. Chang., 4(8), 704–709, doi:10.1038/nclimate2271. Seneviratne, S. I. et al. (2012), Changes in climate extremes and their impacts on the natural physical environment, in Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation, edited by C. B. Field et al., pp. 109–230, Cambridge University Press. Shepherd, T. G. (2014), Atmospheric circulation as a source of uncertainty in climate change projections, Nat. Geosci., (September), 6–11, doi:10.1038/ngeo2253. Shepherd, T. G. (2015), Climate science: The dynamics of temperature extremes, Nature, 522(7557), 425– 427, doi:10.1038/522425a. Stephens, E., and H. Cloke (2014), Improving flood forecasts for better flood preparedness in the UK (and beyond), Geogr. J., 180(4), 310–316, doi:10.1111/geoj.12103. Thompson, D. W., and J. M. Wallace (2001), Regional climate impacts of the Northern Hemisphere annular mode., Science, 293(5527), 85–89, doi:10.1126/science.1058958. Tollefson, J. (2014), US cold snap fuels climate debate, Nature, 1–5, doi:10.1038/nature.2014.14485. Trenberth, K. (2011), Changes in precipitation with climate change, Clim. Res., 47(1), 123–138, doi:10.3354/cr00953. Trenberth, K. E., and D. J. Shea (2005), Relationships between precipitation and surface temperature, Geophys. Res. Lett., 32(14), 1–4, doi:10.1029/2005GL022760. Trenberth, K. E., G. W. Branstator, D. Karoly, A. Kumar, N.-C. Lau, and C. Ropelewski (1998), Progress during TOGA in understanding and modeling global teleconnections associated with tropical sea surface temperatures, J. Geophys. Res., 103(C7), 14291, doi:10.1029/97JC01444. Trenberth, K. E., J. T. Fasullo, and T. G. Shepherd (2015), Attribution of climate extreme events, Nat. Clim. Chang., 5(8), 725–730, doi:10.1038/nclimate2657. Ulbrich, U., J. G. Pinto, H. Kupfer, G. C. Leckebusch, T. Spangehl, and M. Reyers (2008), Changing Northern Hemisphere Storm Tracks in an Ensemble of IPCC Climate Change Simulations, J. Clim., 21(8), 1669– 1679, doi:10.1175/2007JCLI1992.1. Ulbrich, U., G. C. Leckebusch, and J. G. Pinto (2009), Extra-tropical cyclones in the present and future climate: a review, Theor. Appl. Climatol., 96(1-2), 117–131, doi:10.1007/s00704-008-0083-8.

221

Wallace, J. M., I. M. Held, D. W. J. Thompson, K. E. Trenberth, and J. E. Walsh (2014), Global Warming and Winter Weather, Science, 343(6172), 729–730, doi:10.1126/science.343.6172.729. Wang, S.-Y., L. Hipps, R. R. Gillies, and J.-H. Yoon (2014), Probable causes of the abnormal ridge accompanying the 2013-2014 California drought: ENSO precursor and anthropogenic warming footprint, Geophys. Res. Lett., 41(9), 3220–3226, doi:10.1002/2014GL059748. Westra, S., L. V. Alexander, and F. W. Zwiers (2013), Global Increasing Trends in Annual Maximum Daily Precipitation, J. Clim., 26(11), 3904–3918, doi:10.1175/JCLI-D-12-00502.1. World Climate Research Programme (2015), WCRP Grand Challenges, Available from: http://www.wcrpclimate.org/grand-challenges World Health Organization (2010), Climate change, extreme weather events and public health – Meeting report. World Meteorological Organization (2015), 2015 is hottest year on record, Press Release. Available from: https://www.wmo.int/media/content/2015-hottest-year-record Yin, J. H. (2005), A consistent poleward shift of the storm tracks in simulations of 21st century climate, Geophys. Res. Lett., 32(18), 2–5, doi:10.1029/2005GL023684. Zappa, G., L. C. Shaffrey, K. I. Hodges, P. G. Sansom, and D. B. Stephenson (2013), A Multimodel Assessment of Future Projections of North Atlantic and European Extratropical Cyclones in the CMIP5 Climate Models*, J. Clim., 26(16), 5846–5862, doi:10.1175/JCLI-D-12-00573.1. Zhang, X., F. W. Zwiers, G. C. Hegerl, F. H. Lambert, N. P. Gillett, S. Solomon, P. a Stott, and T. Nozawa (2007), Detection of human influence on twentieth-century precipitation trends., Nature, 448(7152), 461–5, doi:10.1038/nature06025.

223

Danksagung Diese Arbeit ist nur entstanden, weil ich tolle Unterstützung von vielen Menschen hatte, denen ich herzlich danken möchte. Zunächst einmal möchte ich Matthias Mengel danken. Warum? Weil ich ohne ihn wahrscheinlich gar nicht am PIK gelandet wäre. Eigentlich wollte ich ihn nur auf dem Telegrafenberg besuchen kommen. Und nu sieh, was daraus geworden. Ich freue mich, dass wir nicht nur die Uni zusammen gemeistert haben, sondern auch die Promotion. Und ich hoffe, dass sich das noch weiter fortsetzen wird. Anders Levermann öffnete mir dann quasi die Tür zum PIK und damit in eine Welt voller interessanter Dinge wie Klimamodelle, high-performance Cluster und riesige Datenwüsten. Seine offene und anpackende Art ist höchst ansteckend und von seinen Erfahrungen konnte ich oftmals profitieren. Katja Frieler hat mich gleich zu Beginn der Promotion in die PRIMAP-Gemeinschaft aufgenommen. Ihre Herzlichkeit und Hilfsbereitschaft sind einmalig und der von ihr zur Verfügung gestellte wöchentliche Obstkorb in unserer „WG“-Küche war ein wichtiger Ausgleich zum ungesund hohen Kuchenkonsum. Der Geburtstags-Floß-Ausflug war ein Highlight, an das ich mich immer noch gerne erinnere. Unvergesslich wurde die Zeit im A26-Keller dank meiner tollen Mitbewohner: Lila, Mahé, Jacob, Sabrina, Ines, Kathleen, Robert, Louise, Johannes... Mit einigen von euch war ich auf tollen Konferenzen und habe Zimmer, Schlafwaggon oder sogar das Bett geteilt, wenn z.B. die Wohnung bei Herrn Parisini mal wieder vollkommen überbelegt war. In meinem Büro war ich zum Glück nie alleine. Mit Antonius konnte ich mich prima über Basketball, Beziehungen, R-Codes und neue Ideen für Ausgründungen unterhalten. Affitos‘ Auslaufdrang hat mich dazu angetrieben, zusammen mit Antonius auch die entlegensten Ecken des Telegrafenbergs zu erkundschaften. Und mit Tobias war ich gleich auf einer Wellenlänge und habe mit ihm über alles disktutieren können. Dabei durfte gerne auch mal (natürlich nur in Ausnahmefällen) ein guter Tropfen Whiskey gereicht werden. Unser Presse-Team – bestehend aus Mareike, Jonas und Sarah – hat mir eindrucksvolle Einblicke in die Journalistenwelt gewährt und mich mit viel Geduld und Wiederholung („Wiederholung ist nicht schlimm, Wiederholung ist gut!“) auf Presseanfragen vorbereitet. Ohne euch, hätte ich vor Aufregung wahrscheinlich keinen Ton rausbekommen.

224

Bibliography

Das mir die ganze Papierwirtschaft nicht über den Kopf gewachsen ist, habe ich Gitta und Peggy zu verdanken. Ein großes Lob geht auch an unsere IT-Abteilung. Von meinen drei Telefon- bzw. EMail Jokern (Ciaron für Cluster, Dietmar für Hardware und Roger für eigentlich alles) habe ich reichlich Gebrauch machen dürfen. Ich bin unendlich glücklich darüber, gegen Ende meiner Doktorarbeit Teil der SacreX-Gruppe mit „Pappa“-Dim geworden zu sein. In unserer temporären Exklave in der HMA hatten wir es uns sehr gemütlich gemacht. Gemeinsam haben wir den täglichen Husarenritt zur Kantine überstanden, den Lärm von gefühlt 10.000 Baggern ertragen, uns gegenseitig motiviert und Halt in schweren Zeiten gegeben. Marlene, Kai, Sonja, Giorgia und Peter, ihr seid toll! Es war und ist für mich eine riesengroße Freude mit Dim Coumou zu arbeiten. Seine positive und ruhige Ausstrahlung ist gepaart mit einem tollen Humor und ich habe mich immer sehr gut betreut gefühlt. Ich habe es nie geschafft, ihn aus der Fassung zu bringen. Dabei habe ich es das ein oder andere mal mehr als „versucht“. Dims größte Stärke (neben seinem Fachwissen) ist meiner Meinung nach aber sein Gespür für den Gemütszustand seiner Schützlinge. Ich habe viel von ihm lernen können und es war rundherum einfach eine tolle Zeit. Unsere gemeinsame AGU Reise wird mir immer in Erinnerung bleiben. Ich hätte es nie bis hierher geschafft, ohne meine Familie. Meine Mutter, mein Vater und meine Schwester waren und sind immer für mich da und ich weiß, dass ich mich zu hundert Prozent auf sie verlassen kann. Sie haben mich nie zu etwas gedrängt und mich im Gegenzug bei allem unterstützt und mir gut zugeredet. Es ist schön zu wissen, dass ihr euch so mit mir freut! Danke für alles! Apropos Familie, das Gleiche gilt natürlich auch für meine bald noch größere Familie. Marion, Hasalt, Anni, Jens, Dorian, Steffi, ... eure tatkräftige Unterstützung hat mir sehr geholfen und ihr ward immer super Testpersonen für meine wissenschaftlichen Erklärungsversuche. Mit Mareika habe ich alles gemeinsam erlebt und geteilt. Sie hat mit mir unzählige Vorträge geübt und immer so getan, als fände sie alles total interessant. Sie hat bei Interviews am Radio gehangen und mitgezittert und hat sich über Erfolge fast noch mehr gefreut als ich mich selber. Ich danke dir fürs einfach-da-sein, für deine Unterstützung, für dein Lachen und deine Liebe!

Ich weiß, dass eine Person ganz besonders stolz auf mich sein wird, wenn ich meine Promotion abschließe. Ihr ist diese Arbeit gewidmet.

Erklärung Diese Arbeit ist bisher an keiner anderen Hochschule eingereicht worden. Sie wurde selbständig und ausschließlich mit den angegebenen Mitteln angefertigt. Potsdam, den 29.02.2016

Jascha In-su Lehmann

Diese Arbeit wurde durchgeführt am Potsdam-Institut für Klimafolgenforschung

Changes in extratropical storm track activity and their

Changes in extratropical storm track activity and their

Suggest Documents

Changes in the Extratropical Storm Tracks in Response to ... - UiO

North Atlantic storm track changes during the Last Glacial Maximum

Seasonal Predictability of Extratropical Storm Tracks in ... - AMS Journals

track changes - Geoff-Hart.com

With Track Changes - HSA

Animal track activity sheets

Changes in physical activity and short-term changes in ... - CiteSeerX

Storm track signature in total ozone during ... - Wiley Online Library

Storm track signature in total ozone during northern hemisphere winter

Microsoft Word Track Changes - UTHSC

The Position of the Midlatitude Storm Track and Eddy-Driven ...

Channel morphology changes and their

Developmental changes in spontaneous electrocortical activity and ...

Changes in jump performance and muscle activity

Activity-dependent changes in transporter and ...

Changes in Pigments, Chlorophyllase Activity, and Chloroplast ...

Hospital mortality and changes in activity

Epidemiology Changes in physical activity and modelled

Monsoon-extratropical circulation interactions in

Activity-Dependent Changes in Cholinergic

Asarone and Their Anticancer Activity

Changes In Submersed Macrophytes In Relation To Tidal Storm Surges

Storm time changes in total electron content in ionosphere - Annales ...

Storm time changes in total electron content in ionosphere measured ...