Synoptic moisture pathways associated with mean

Climate Dynamics https://doi.org/10.1007/s00382-018-4300-6

Synoptic moisture pathways associated with mean and extreme precipitation over Canada for summer and fall Xuezhi Tan1,2 · Thian Yew Gan2 · Yongqin David Chen3 Received: 3 July 2017 / Accepted: 13 June 2018 © Springer-Verlag GmbH Germany, part of Springer Nature 2018

Abstract Large-scale meteorological patterns (LSMPs), especially vertically integrated water vapor transport (IVT) patterns were identified using the self-organizing map algorithm, and LSMPs were related to seasonal precipitation totals and widespread precipitation extremes for summer and fall seasons over Canada West and East, using the JRA-55 reanalysis and the ANUSPLIN precipitation dataset (1958–2013). Changes in the frequency of occurrences, persistence and maximum duration of each LSMP were detected. Trends in seasonal precipitation totals and extreme precipitation events associated with each LSMP were also detected. Our results show that synoptic settings of precipitation that have occurred over Canada exhibit a variety of spatial clusters of IVT anomalies, pressure highs and lows, troughs and ridges over North America, North Pacific, North Atlantic and Arctic. Extremely high IVT magnitude anomaly (|IVT|′) over central Canada West are associated with the Aleutian low and Gulf cyclone across Canada West and Alaska, which force moisture from North Pacific to Canada West. Widespread positive |IVT|′ over Canada East are related to a strengthened and southwest-shifted trough across Canada East. Annually, only 19.4% of the occurrence characteristics of LSMPs show statistically significant changes. More statistically significant changes in the daily precipitation and occurrences of extreme precipitation related to each LSMP, than changes in the occurrence of LSMPs, have resulted in changes to seasonal precipitation totals and the occurrence of extreme precipitation across Canada. LSMPs associated with a dry climate and less frequent extreme precipitation events over Canada West in summer and fall tend to occur during the negative phase of PNA. LSMPs associated with a wet climate and frequent occurrence of extreme precipitation events over the south (north) of Canada East are more likely to occur during the positive (negative) phase of NAO. Keywords Vertically integrated water vapor transport · Synoptic patterns · Large-scale meteorological patterns · Seasonal precipitation · Extreme precipitation · Self-organizing maps, climate anomalies

1 Introduction

Electronic supplementary material The online version of this article (https://doi.org/10.1007/s00382-018-4300-6) contains supplementary material, which is available to authorized users. * Xuezhi Tan [email protected] 1

Department of Water Resources and Environment, Sun Yat-sen University, Guangzhou 510275, People’s Republic of China

2

Department of Civil and Environmental Engineering, University of Alberta, Edmonton, AB T6G 2W2, Canada

3

Department of Geography and Resource Management, and Institute of Environment, Energy and Sustainability, The Chinese University of Hong Kong, Hong Kong, Hong Kong

The northern high-latitude land area has experienced about twice the rate of mean global warming over past several decades (Cohen et al. 2014; Hartmann et al. 2013; IPCC 2013). This “Arctic amplification”, enhanced warming over the Arctic, primarily resulted from the ice/snow-albedo feedbacks due to interactions between land snow cover, sea ice extent, lapse rate feedback, Planck feedback, and water vapor feedback, which are all positive feedbacks (Wendisch et al. 2017; Pithan and Mauritsen 2014; Screen and Simmonds 2011; Serreze and Francis 2006; Serreze and Barry 2011). Canada, which has a large landmass in the Arctic, has also experienced a rapid warming with a total mean surface temperature increase of about 1.5 °C over 1950–2010 (Vincent et al. 2015). Given that theoretically atmospheric

13

Vol.:(0123456789)

water vapor increases at 7%/°C of warming following the Clausius–Clapeyron relation (Held and Soden 2006), global warming brings about a strengthened moisture transport, a intensified hydrologic cycle, and therefore more precipitation (Huntington 2006; Lenderink and van Meijgaard 2008, 2010; Zahn and Allan 2013). As expected, with land surface warming, precipitation has generally increased in Canada, including an increase in the number of small to moderate precipitation events (Mekis and Vincent 2011; Vincent and Mekis 2006). However, occurrences of extreme precipitation events had decreased in Canadian Prairies over 1950–2010 (Tan and Gan 2017). Climate warming has also significantly reduced the streamflow of major rivers across western Prairie provinces in the warm, summer and fall seasons when water demand for agricultural and industry production, and in-stream flow needs for sustaining the environment are the greatest (Schindler and Donahue 2006; Kerkhoven and Gan 2011; Tanzeeba and Gan 2011). Because the hydrologic response of river basins (runoff production) to precipitation differs widely, which depends on the antecedent soil moisture condition, vegetation, terrain, geology and the seasonal precipitation climatology, particularly the regional and seasonal variability of precipitation under a changing climate across Canada where hydroclimatic and physiographic variability is enormous. The observed changes in annual precipitation totals and extremes over different regions have been examined (e.g., Burn et al. 2011; Tan and Gan 2017; Vincent et al. 2015). However, few studies have analyzed the spatial and temporal variability of seasonal precipitation totals and extremes across Canada associated with synoptic moisture pathways, even though an understanding of the characteristics of seasonal precipitation and their changes is very useful to water resources management and water-related natural hazards such as droughts and floods. In this study, the summer and fall precipitation of Canada divided into two regions, Canada West and Canada East, is investigated. Changes in seasonal precipitation totals and extremes should be partly related to changes in circulation patterns and atmospheric humidity that control the moisture transport and moisture sources for precipitation, respectively. From back-trajectory analyses of moisture transport for 3-day seasonal maximum precipitation over Canada, Tan et al. (2017) shows that moisture sources and pathways associated with precipitation extremes for different seasons and regions of Canada vary widely. This finding motivates us to further investigate large-scale meteorological patterns (LSMPs which include large-scale atmospheric circulations and moisture pathway patterns) that are responsible for seasonal precipitation totals and extremes. Extreme precipitation events had resulted in disastrous flood events over Canada (Buttle et al. 2016), while extreme low seasonal precipitation has led to persistent, extensive droughts (Bonsal et al. 2013;

13

X. Tan et al.

Gobena and Gan 2013). Because droughts and floods have been costliest natural disasters for Canada (White and Etkin 1997), with significant impacts on sectors such as water supply, agriculture, forestry, human health and society, and ecosystems, it will be beneficial to better understand how LSMPs and their variabilities are associated with seasonal precipitation totals and the occurrence of extreme precipitation events in Canada. Moisture pathway patterns are dependent on both circulation patterns and water vapor conditions. Moisture from the Gulf of Mexico can give rise to extreme precipitation in high latitudes of North America when low-level jets from Great Plains to Dakotas occur concurrently with rapid cyclogenesis over the Canadian Prairies (CP) (Brimelow and Reuter 2005). Some droughts over the central and western CP were associated with low moisture transport from the Pacific Ocean (Gulf of Mexico) in the winter (summer) (Liu et al. 2004). As expected, both the Arctic and Pacific Oceans are sources of winter moisture for the Saskatchewan River Basin in CP (Liu and Stewart 2003). For the Mackenzie River basin, the transport of water vapor has high spatiotemporal variabilities, with moisture from the southwest (the subtropical and mid-latitude central Pacific Ocean) transported to the basin during autumn, winter, and spring by extratropical cyclones, while moisture from the northwest (Arctic Ocean) enters the basin during the summer (Smirnov and Moore 1999, 2001). Atmospheric river events in mid-latitude Pacific Ocean characterized by relatively narrow and long bands of concentrated moisture and strong, low-level wind in the atmosphere (Dettinger 2011) are often associated with extreme precipitation in British Columbia (Roberge et al. 2009; Spry et al. 2014; Tan et al. 2017). Past studies on atmospheric moisture transport tend to focus on a particular region even though moisture pathway patterns over a large landmass such as Canada have large spatial variabilities. Therefore, in this study, we have investigated synoptic moisture pathway patterns that are conducive to the occurrence of seasonal precipitation totals and extremes over Canada. Over the past two decades, the self-organizing map (SOM) algorithm (Kohonen 1998), a neural network algorithm, has been widely used to analyze and display characteristic behavior of atmospheric and oceanic circulation patterns as well as for linking regional climate components (i.e., temperature and precipitation) to these patterns (e.g., Cassano et al. 2015; Gibson et al. 2017; Schuenemann and Cassano 2010; Swales et al. 2016). SOM, useful for visualizing low-dimensional views of high-dimensional data, is a discretization of a continuous, spatial field pattern in terms of a two-dimensional array, while a SOM array essentially is a composite of input maps with similar spatial distributions for the field examined and displays features that represent the most temporal variabilities, so that SOM arrays represent archetypal patterns of a field. For meteorological studies,

Synoptic moisture pathways associated with mean and extreme precipitation over Canada for…

SOMs can be used to explore the physical behavior yielding certain climatic events by identifying circulation patterns frequently associated with those events. The SOM method for data analysis with a linear manifold produces comparable patterns to those derived using an EOF analysis (Hannachi et al. 2007) or other clustering methods (Phillippopoulos and Delligiorgi 2012). However, SOM or other neural networkbased cluster analysis performs better when a time series contains nonlinear manifolds and/or large amount of stochastic noise (Liu et al. 2006). In this study, both seasonal precipitation totals and the occurrence of extreme precipitation events over Canada West and East were linked to SOM patterns of vertically integrated water vapor transport (IVT), to identify influence of LSMPs on precipitation of Canada. The two objectives of this study: (1) to identify dominant moisture pathway patterns and their variabilities during warm, summer and fall seasons of Canada West and East, and (2) to relate changes in the occurrence of LSMPs to changes in seasonal precipitation and occurrences of extreme precipitation events in both seasons over Canada. The precipitation data of Canada, atmospheric reanalysis data and research methods used are described in Sect. 2. The discussions of results are presented in Sect. 3, and summary and conclusions are given in Sect. 4.

2 Data and methods 2.1 JRA‑55 dataset In this study, gridded (0.5625° × 0.5616°) daily IVT, surface mean daily temperature, and 500 hPa geopotential height (GPH) fields are taken from the Japanese 55-year Reanalysis (JRA-55) (Kobayashi et al. 2015) data set provided by the Japan Meteorological Agency. JRA-55 is the 2nd global atmospheric reanalysis dataset of Japan suitable for studying multi-decadal variability and climate change impact. The JRA-55 IVT dataset has been shown to well represent the hydrologic cycle and water balance of northern highlatitude regions (Bintanja and Andry 2017; Dufour et al. 2016). Compared to other reanalyses data (e.g., MERRA2, CFSR, and ERA-Interim) that began around 1980s, the JRA55 reanalyses dataset has a relatively long period, starting from 1958 to the present (Bintanja and Andry 2017; Kochtubajda et al. 2017). Therefore, the JRA-55 IVT dataset has been widely used to evaluate the water cycle of northern high-latitude regions including the Arctic (e.g., Bintanja and Andry 2017; Dufour et al. 2016; Hu et al. 2016; Vihma et al. 2015) and Canada (Kochtubajda et al. 2017). For this study, we have analyzed the atmospheric fields for spring (March–May) and summer (June–August) from 1958 to 2013. To better capture the spatial variability of changes in LSMPs and their associated regional precipitation over

a large landmass, we divide Canada to Canada West and East. The study domains extend from 180°W to 100°W for Canada West, and 100°W to 25°W for Canada East, respectively, and 40°N to 80°N for both regions of Canada (Fig. 1).

2.2 ANUSPLIN precipitation dataset High-resolution (~ 10 km) gridded daily Canadian precipitation data (1958–2013) developed by Hutchinson et al. (2009) using the Australian National University Spline (ANUSPLIN) interpolation scheme was used to estimate seasonal precipitation totals and days with extreme precipitation over Canada West and East. This ANUSPLIN dataset for Canada was derived from observed precipitation data of nearly 3000 stations of Environment Canada. Even though there are some minor discrepancies between station extreme precipitation and ANUSPLIN extreme precipitation (Benyahya et al. 2014; Hopkinson et al. 2011), the ANUSPLIN dataset provides representative high-resolution daily precipitation data for Canada. This dataset is the station-based, gridded daily precipitation data set widely used in past studies on Canadian climate (e.g., Benyahya et al. 2014; Cannon et al. 2015; Gizaw and Gan 2016; Newton et al. 2014a, b; Radić et al. 2015; Wong et al. 2017). However, due to an inadequate data network to adequately capture the high spatial complexity of daily precipitation, the ANUSPLIN precipitation data are weak in representing the daily precipitation over highelevation regions such as the Rocky Mountains where the intensity of daily precipitation extremes are underestimated (Hopkinson et al. 2011). However, to ensure ANUSPLIN precipitation data are valid to show the occurrence of percentile-based precipitation extremes for Canada in this study, we have analyzed 164 (131) stations of daily precipitation data over Canada West (East) for 1958–2005 taken from Mekis and Vincent (2011), with the same method adopted in this study. The station-based analysis results are consistent with what we have herein reported for regions with dense stations. Therefore, it is reasonable to use ANUSPLIN precipitation dataset.

2.3 Large‑scale meteorological patterns (LSMPs) For every season and each region, the SOM was trained on the magnitude of the standardized daily IVT anomaly (|IVT|′) data using 3 × 3 SOM nodes. Thus, moisture transport conditions for each region and each season were clustered into 9 different moisture transport patterns. SOM patterns were explored using a variety of SOM node configurations and parameters, such as the neighborhood function, learning rate, and radius. Results show that dominant patterns were not sensitive to the SOM configuration and parameters. The 9-node SOM, which is sufficient to show some infrequent occurrences of |IVT|′,

13

X. Tan et al.

Fig. 1 Provinces and ecoregions of Canada. The red vertical line divides Canada into Canada West and East analyzed in this study. The Provinces or Territories of Canada are: AB Alberta, SK Saskatchewan, MB Manitoba, NL Newfoundland & Labrador, PE Prince

Edward Island, NS Nova Scotia, NT Northwest Territories,, Nunavut, ON Ontario, NB New Brunswick, YT Yukon Territory; British Columbia, QC Quebec

represent the continuum of moisture pathway patterns without excessive intra-node variability. Furthermore, since 9 |IVT|′ patterns are also easier to interpret than those derived from more SOM nodes (e.g. 16 nodes), 9 SOM patterns of the |IVT|′ for precipitation over Canada were derived for further analyses. Once a SOM array for |IVT|′ has been created, the |IVT|′ data were mapped, in which each daily |IVT|′ is compared with nine SOM nodes to identify the node to which a daily |IVT|′ best matches. This mapping procedure generated a list of |IVT|′ associated with each node, which allows us to determine the frequency of occurrence of each SOM |IVT|′ pattern. Therefore, each daily |IVT|′ field corresponds to one of the nine SOM nodes. The composite averages of IVT, GPH and surface temperature on days within each node of the |IVT|′ pattern were also computed by relating each daily GPH field to the node occurring on that day. To clearly identify climatological features of Canada, we spatially expanded the study domain for IVT, GPH and surface temperature composites over the entire country (Canada West and East). In this study, |IVT|′ patterns (nodes) and GPH composites are referred to as LSMPs.

2.4 Mean and extreme precipitation mapping

13

To relate the seasonal precipitation totals and extreme precipitation events to atmospheric SOM patterns (LSMPs) derived, we have identified days of precipitation and extreme precipitation, respectively, in ANUSPLIN precipitation dataset at each grid cell throughout Canada West and East. For each season at every grid cell throughout the domain, we defined seasonal heavy and extreme precipitation events as the top 5 and 1% of days of precipitation larger than 0.2 mm that occurred in that season during the 56-year period, respectively. We first identified trends of seasonal totals by the nonparametric Mann–Kendall (MK) test (Kendall 1975) and trends of seasonal heavy or extreme precipitation events by the Poisson regression method (Tan and Gan 2017). The slope of trends in seasonal precipitation totals and occurrence of heavy or extreme precipitation events were estimated using the Theil–Sen estimator (Sen 1968). For seasonal precipitation totals, the daily precipitation at each grid cell for each day was related to the SOM node occurring on that day, and each day’s circulation is assigned to one of the SOM nodes. A composite of daily precipitation for all days associated with a SOM node was


obtained and thus daily precipitation was mapped to that SOM node. Daily precipitation anomalies for a node were derived by subtracting the mean daily precipitation for that node from the overall mean daily precipitation of each season. For extreme precipitation events, a day with extreme precipitation was related to a SOM node occurred on that day. For each grid cell, the frequency of extreme precipitation events that have occurred on a node was the ratio of the total number of extreme precipitation events for that node to that for all the nine nodes. These analysis produced maps relating each SOM node to seasonal precipitation anomalies and the frequency of extreme precipitation events occurring across Canada. Therefore, we have identified SOM patterns (LSMPs) responsible for the occurrence of seasonal precipitation and extreme precipitation events of Canada.

2.5 Changes in occurrences of LSMPs and precipitation Transition frequencies from each node to other nodes show how nodes related to each other temporally. Because SOM patterns and composites (LSMPs) are results from |IVT|′, and GPH fields of individual days, we further examined all days involved in each node and trends of each LSMP (node) occurred in a year for each season during 1958–2013. Here we calculated four characteristics of pattern occurrences, which are (1) the total number of days on which each SOM pattern had occurred (occurrence, day year− 1) or the occurrence frequency of each SOM pattern (%); (2) the mean length of consecutive occurrences (persistence, day event− 1); and (3) the longest consecutive occurrence (maximum duration, day e vent− 1); (4) the 1-day lead transition frequency for all the nine SOM nodes (transition matrix, %). For the former three characteristics, a trend analysis over 1958–2013 was conducted by the MK test at a 0.05 significance level. For each season, the trends of seasonal precipitation totals and the extreme precipitation events occurred in each node were also estimated using the MK test. We have computed the lag-1 correlation of all the time series and found no statistically significant serial correlations, so the conventional MK test is applicable to our trend tests. The trend magnitude in these variables was estimated using the Theil–Sen estimator (Sen 1968). Changes in seasonal precipitation totals and occurrences of extreme precipitation events (Fig. 2) were also tested by the MK test and the Theil–Sen estimator, which as non-parametric methods, are both independent of the underlying probability distributions of the time series. By relating changes in LSMPs to precipitation changes, we can qualitatively relate precipitation changes to changes in the occurrence of LSMPs and changes in precipitation that occurred within some particular LSMPs (Cassano et al. 2007).

2.6 Teleconnections between LSMPs and large‑scale climate anomalies Some interannual and interdecadal variabilities in dominant atmospheric circulations and hydroclimatic variables over Canada are related to large-scale teleconnections (Coulibaly 2006; Gan et al. 2007; Newton et al. 2014a, b; Tan et al. 2016), such as El Niño-Southern Oscillation (ENSO), Pacific North American (PNA) pattern, Pacific Decadal Oscillation (PDO), North Atlantic Oscillation (NAO), and Arctic Oscillation (AO). We used climate indices that represent the above five climate oscillations to evaluate their influence on occurrences of LSMPs and precipitation for each season over Canada West and East. The multivariate ENSO index (MEI) (Wolter and Timlin 2011) was used to represent ENSO. The cold (warm) ENSO phase, i.e., La Niña (El Niño), is represented by negative (positive) values of the MEI. Daily values of the MEI, PNA, PDO, NAO, and AO indices were downloaded from the Climate Prediction Centre (http://www.cpc.ncep.noaa.gov/). We divided each daily teleconnection modes into three, positive, neutral, and negative phases for the 56 years of study period, respectively. This procedure was widely used to categorize teleconnections (e.g., Bonsal et al. 2001; Newton et al. 2014a, b). We linked MEI, PDO, PNA, NAO and AO conditions to each |IVT|′ pattern as composite teleconnection conditions. To show the possible lag effects of teleconnection on LSMPs, we calculated the 0-, 1-, 2-, and 3-month lag time composites of teleconnection conditions for each |IVT|′ pattern. We used the two-sample nonparametric Kolmogorov–Smirnov (KS) test to evaluate the differences in synoptic type frequency distributions for each positive–negative pair at the 0.05 significance level.

3 Results 3.1 Changes in seasonal precipitation and occurrence of extreme precipitation events Figure 2 shows changes in mean precipitation and occurrences of extreme precipitation events in summer and fall. Summer precipitation has predominantly experienced increasing trends across Canada over 1958–2013, although some areas in the east (eastern Manitoba, northern Ontario, and southern Quebec), southeastern Northwest Territories and northern Alberta had also experienced significant decreasing trends (Fig. 2a). However, fall precipitation mainly decreased in the West but increased in the East of Canada (Fig. 2b). The averaged increase in the summer precipitation over the West and East was significant at 36.6 and 36.3 mm, respectively, while the fall precipitation over the

13

X. Tan et al.

Fig. 2 Changes in the mean seasonal precipitation (a, b), occurrence of heavy precipitation events defined with days when precipitation exceed 95% of daily precipitation in a season (c, d), and occurrence of extreme precipitation events defined with days when precipitation exceed 99% of daily precipitation in a season (e, f). Grid cells with statistically significant trends in heavy and extreme precipitation were denoted with blue contours. All grid cells are with statistically significant trends in mean seasonal precipitation. The left column a, c, e shows trends for the summer season while the right column b, d, f shows trends for the fall season

west (east) decreased (increased) by − 13.5 (73.3) mm, over the 56-year period. There has been more spatial variability in the occurrence of extreme precipitation events. The occurrence frequency of heavy (> 95th percentile) precipitation events is much more significant than that of extreme precipitation events (> 99th percentile) (Fig. 2c–f), although both are regionally consistent regarding the primary location of increasing and decreasing trends. Given results obtained for LSMPs associated with extreme precipitation events based on the 99th

13

percentile are also similar to that based on the 95th percentile, we have only reported results with heavy precipitation which is generally representative of that for the extreme precipitation, unless otherwise specified. Extreme precipitation events in the summer had mainly increased in the north, Saskatchewan and west Manitoba, but decreased in southern British Columbia, Ontario, southern Quebec and Atlantic Canada (Fig. 2c). In the fall, while extreme precipitation events had decreased in southwestern Canada, the far north (> 60°) and Canada East had experienced a considerable


increase in the occurrence of extreme precipitation events (Fig. 2d). Because both Canada West and East have experienced both increasing and decreasing trends in extreme precipitation, the overall increase in areally-weighted average occurrences of extreme precipitation was only 0.089 and 0.001 events over Canada West, and 0.130 and 0.292 events over Canada East, for summer and fall, respectively, over the 56-year period. Note that the average number of extreme seasonal precipitation is only about two events per year.

3.2 LSMPs and their changes and transitions Figures 3 and 4 show results of SOM training with the daily |IVT|′ for summer and fall, respectively, over Canada West (Figs. 3a, 4a) and East (Figs. 3b, 4b), while Figs. 5 and 6 show composites of 500-hPa GPH, surface temperature anomalies, and IVT fields for summer and fall, respectively. These figures show identified climatological features of LSMPs, including a variety of spatial clusters of IVT vectors, surface temperature anomalies, ridges, troughs, pressure highs and lows over North America, North Pacific, North Atlantic and the Arctic. The |IVT|′ patterns across the nine nodes vary from the center of anomalously high and low IVT conditions from top to bottom of SOM grids. The different spatial distribution of |IVT|′ between these nodes shows different pathways for transporting moisture to Canada West and East, which can be identified by combining |IVT|′ patterns to circulation patterns (Figs. 5, 6). To more generally describe the moisture pathway patterns and LSMPs for both seasons and both regions, nine nodes were assigned to one of four pattern clusters based on the central locations and their evolution of extremely large |IVT|′ (dark blue regions in Figs. 3, 4). Table 1 lists the pattern clusters defined for presentation. In addition to a dry pattern cluster, the wet |IVT|′ patterns are north, west and central pattern clusters for Canada West, while north, south and east pattern clusters for Canada East. For LSMPs over Canada West in the summer, the Central and Dry pattern clusters (Nodes S6–S9 in Fig. 5a) are associated with the Aleutian Low and Gulf cyclone across Canada West and Alaska, forcing moisture from North Pacific to the North of Canada West. Nodes S6–S9 are also associated with a strong trough over the Pacific Ocean but a ridge centered over Alaska, stretching south and west across the Rocky Mountains. However, the North pattern cluster (Nodes S1–S2 in Fig. 5a) is associated with a split-flow blocking high which results in the weather pattern between the two jets becoming stagnant over Canada West, resulting in extremely large temperature anomalies. For the West pattern cluster, Node S3 shows a strong ridge over northwestern Canada and northeastern Pacific which is linked with a strong northerly meridional flow, while Nodes S4–S5 show a strong trough over the west coast of Canada (Fig. 5a).

The GPH patterns from Nodes S1 to S9 generally show the transition between highs and lows in pressure in terms of strength and center. All these nine nodes show variations in the wavy nature of atmospheric circulations, which shows to some extent that the pathway for moisture transport to Canada West is associated with the anticyclonic or cyclonic nature of the Rossby wave breaking event that drives the moisture transport (Liu and Barnes 2015). Even though the moisture transport over northern Pacific is generally (> 90%) poleward as shown in other studies on moisture transport to the Arctic (e.g., Liu and Barnes 2015; Dufour et al. 2016), moisture transported to the Canadian landmass are essentially driven by both zonal flow (westerly) from the Pacific and poleward meridional flow from the northern continental United States. Note that we have used 500 hPa geopotential heights data, which at that elevation may not contain much moisture but have been widely used to represent largescale circulations (e.g., Newton et al. 2014b; Swales et al. 2016). From this regard, the 500 hPa GPH pattern cannot fully describe flow directions of IVT derived from the entire atmospheric profile. For LSMPs over Canada West, the wavy nature of atmospheric circulation in the fall is more evident than in the summer. Nodes S1–S5 (Fig. 6a) generally vary in the center locations and are associated with the strength of the Aleutian Low, Gulf cyclone, and North America ridge. The North (Nodes S2–S4) and Central (Nodes S5–S6) pattern clusters are linked with a very strong ridge over North America and extremely large temperature anomalies over Canada West. Nodes S6–S9 show somewhat zonal flow features as indicated by GPH patterns and less wavy circulations than that of Nodes S1–S5. For LSMPs over Canada East in summer and fall, nine nodes (Figs. 5b, 6b) show how the troughs associated with the Arctic Low stretch from northern Canada to southeastern Canada, although there are possible impacts of a high pressure centered over northern Atlantic such as in Nodes S6 and S8 in the summer (Fig. 5b), and Nodes S2, S5, and S6 in the fall (Fig. 6b). This concurrent presence of a low and a high increases the waviness of circulation and facilitates moisture transport from northern Atlantic, northern continental United States, and western Canada to Canada East. Troughs that do not stretch far to Canada East are concurrent with extremely high temperature (e.g., Nodes S1 and S4 in both summer and fall). Troughs stretch far southeast for the North pattern cluster (Nodes S2–4) while far southwest for the Central pattern cluster (Nodes S5–6). As expected, troughs are much stronger in summer than in fall because of larger temperature differences between mid- and highlatitude regions in summer than in fall. The most frequently (> 21.4%; Table 1) occurring pattern in Canada West and East is the predominantly dry pattern (e.g., Nodes S8 and S9 in Figs. 2a, 3a, b). Dry patterns

13

also have the highest persistence and maximum durations of ~ 2–3 and ~ 4–10 days, while wet patterns are of ~ 1–2 and ~ 1–4 days, respectively. The occurrence frequency of three

13

X. Tan et al.

wet pattern clusters is relatively similar (~ 23%) in summer and fall for Canada West, although the North pattern cluster has occurred less frequently than the West and Central

Synoptic moisture pathways associated with mean and extreme precipitation over Canada for… ◂Fig. 3 3 × 3 SOM patterns of standardized integrated vapor transport

anomaly (|IVT|′) in summer over Canada West (a) and Canada East (b). Three numbers in the left top corner of each plot are mean values for three characteristics of each pattern: the occurrence frequency (occurrences; %), the mean length of consecutive occurrence (persistence; day e vent− 1), and maximum duration (duration; day e vent− 1) from top to bottom. The corresponding magnitudes (per year) of trends of the three characteristics are shown in the left bottom corner (middle right) of a and b. Red (blue) numbers shows positive (negative) trends in seasonal time series of these three characteristics. Statistically significant trends in seasonal time series are shown in bold and italic numbers

pattern clusters (Table 1). For Canada East, frequency of occurrence of four pattern clusters differ considerably between summer and fall. The North pattern clusters (34.3%) occurred most frequently in the summer whereas the dry pattern (44.7%) in the fall. This indicates that IVT in northern Canada (across Canada East) tends to be higher (lower) in magnitude in the summer (fall) than the annual mean IVT conditions. Since we have estimated three characteristic features (frequency, persistence and maximum duration) for the occurrence of nine |IVT|′ patterns for two seasons over two regions each year, we have produced 108 (3 × 9 × 2 × 2) seasonal time series of pattern occurrence characteristics. Trend analyses on these 108 time series (Fig. 7) show that only 21 (19.4%) annual characteristics of the occurrence of |IVT|′ patterns show statistically significant changes in 1958–2013 (Fig. 4) for both the two regions and two seasons (summer and fall). More summer |IVT|′ patterns show significant changes in the occurrence characteristics (15 out of 54; 27.8%) than in the fall |IVT|′ patterns (6 out of 54; 11.1%). Marginally more |IVT|′ patterns experienced changes in the occurrence frequency (8 out of 36; 22.2%) than persistence (6 out of 36; 16.7%) and maximum duration (5 out of 36; 13.9%). Canada West (11 out of 54) and East (10 out of 54) show a similar percentage of changes in the occurrence of |IVT|′ patterns. For summer in Canada West, the significant decrease in the occurrence of the dry pattern (Node S9) is concurrent with the significant increase in the occurrence of anomalously high |IVT|′ over the Canadian Arctic (Node S1) and southwestern Canada (Node S7), on which both the occurrence frequency and maximum duration experienced a statistically significant increase (Fig. 3a). For summer in Canada East, more wet patterns (e.g., Nodes S1, S2 and S4) have shifted patterns associated with which widespread and anomalously low IVT (i.e., Nodes S9) (Fig. 3b). The occurrence of anomalously high |IVT|′ over northern Canada in the spring has significantly increased while Nodes S7 and S9, which both have a widespread and anomalously low IVT over Canada West, has occurred less frequently, even though their decrease in the occurrence frequency are not statistically significant (Fig. 4a). Similar to changes to summer patterns in Canada East, decreases in the occurrence

of dry patterns in fall over Canada East (Nodes S7 and S9) coincide with an increase in the occurrence of the North and South pattern clusters (Nodes S1–S5; Fig. 4b). The increased occurrences in patterns where |IVT|′ over Arctic Canada are extremely high is consistent with the increasing water vapor transport to northern high-latitude (e.g., Liu and Barnes 2015; Zhang et al. 2013). Figure 8 shows the one-day lead transition probability between the four |IVT|′ pattern clusters defined in Table 1. The predominantly high probability of a node (Figure S1 in Supplement) or a pattern cluster (Fig. 8) to persist in two days agrees with the persistence of |IVT|′ patterns (~ 2 days). For most possible transitions between different nodes over Canada West, as expected, occur between nodes that are involved in one of four major pattern clusters, since they have similar LSMPs with different centers of extremely high |IVT|′, high-, and low-pressure centers. For example, as summer synoptic patterns evolve over Canada West, an |IVT|′ pattern at Node S1 (S3) is more likely to transit to Node S2 (S6), which shows the movement of a blocking (ridge) over Canada West propagating with the westerly (Fig. 5a). Similarly, in the transitions of patterns involving a low and extratropical cyclone (Nodes S6–S9), the air rich in moisture often moves zonally over the coast away from the surface pressure low because of the cold front of southwesterly which indicates the persistence of moisture transport needed for precipitation to occur over Canada West. In addition to the movement of a ridge from northeastern Pacific to western Canada, the frequent, sequential transition between fall patterns for Canada West also shows the weakening of that ridge (e.g., from Nodes S1 to S9, and from Nodes S2 and S3 to S8; Figure S1 in Supplement and Fig. 6a). For summer in Canada East, a frequent transition from Nodes S2 to S3 indicates a deepening of the Arctic Low, while the frequent, sequential transition from Nodes S5 to S8 the poleward movement of a low-pressure system (Fig. 5b). The troughs for Canada East in the fall tend to be weak as most patterns (Nodes S1–S6) with a trough over Canada East frequently transit to zonal flow patterns (Nodes S8 and S9 in Fig. 6b). For both Canada West and East in both summer and fall, most patterns tend to transit to the dry pattern because of the dominant occurrence of dry patterns in four pattern clusters. However, for Canada East, more transitions of LSMPs in the summer occur between wet patterns (north, south and east pattern clusters) than those in the fall, partly from the relatively low occurrence frequency of dry patterns for Canada East in summer.

3.3 Mapping of seasonal precipitation totals and extremes Figures 9, 10, 11, and 12 show how |IVT|′ patterns are related to spatial distributions of summer and fall

13

X. Tan et al.

Fig. 4 Same as Fig. 3, but for 3 × 3 SOM |IVT|′ patterns of standardized integrated vapor transport anomaly (|IVT|′) in fall over Canada West (a) and Canada East (b)

13


Fig. 5 A 3 × 3 SOM composite of the integrated vapor transport, IVT (shown in vectors), in summer over Canada West (a) and Canada East (b). Overlaid on the shaded colors of surface temperature

anomalies are contours of the 500 hPa geopotential heights (m). For convenience, numbers of the characteristics of the occurrence of each pattern are also shown in this figure as that in Fig. 3

precipitation anomalies (occurrences of extreme precipitation) over Canada, respectively. As expected, regional positive (negative) seasonal precipitation anomalies across

Canada are generally found in regions of mid-tropospheric convergence (divergence) that are located to the right (left) of the ridge axis and left (right) of the trough, in cases

13

X. Tan et al.

Fig. 6 Same as Fig. 5, but for A 3 × 3 SOM composite of the integrated vapor transport, IVT (shown in vectors), in fall over Canada West (a) and Canada East (b)

that the trough points south while the ridge points north. Anomalously high (low) precipitation in a region is associated with pressure lows (highs) centered over that region. Most |IVT|′ patterns associated with regional positive

13

(negative) precipitation anomalies also correspond to high (low) occurrence probabilities of extreme precipitation events in that region in all seasons across Canada.

Synoptic moisture pathways associated with mean and extreme precipitation over Canada for… Table 1 SOM nodes included in each pattern cluster presented in this study and the occurrence frequency of pattern clusters

Canada West

Canada East

Pattern clusters

Summer

Fall

Pattern clusters

Summer

Fall

North West Central Dry

1, 2 (13.5%) 3, 4, 5 (27.3%) 6, 7 (21.6%) 8, 9 (37.4%)

2, 3, 4 (9.5%) 1, 7 (31.3%) 5, 6 (25.4%) 8, 9 (33.8%)

North South East Dry

1, 2, 3, 4 (34.3%) 5, 6 (21.3%) 7, 8 (23.0%) 9 (21.4%)

1, 2, 3 (16.1%) 4, 5 (18.6%) 6, 7 (20.5%) 8, 9 (44.7%)

Fig. 7 Trends in land surface precipitation and moisture transport patterns. Trends are calculated for the Canada West and East in summer and fall seasons. Region domains (see Fig. 1) in which a SOM pattern demonstrate statistically significant, increasing (decreasing) trends in the occurrence (O), persistence (P) and maximum duration (M) of standardized integrated vapor transport anomaly (|IVT|′) is shown in

red (blue). Regional domains with positive (negative) trends in the average seasonal precipitation (A), occurrence of heavy precipitation (H) and extreme precipitation (E) occurred in each SOM pattern of |IVT|′ is shown in red (blue) which are covered with semi-transparent, grey rectangles. White boxes indicate no statistically significant trends

Different LSMPs lead to different spatial variabilities of precipitation across Canada because of different central locations and strength of low- and high-pressure systems, and ridges and troughs for each LSMP. For example, for Canada West in the summer (Fig. 9a), Node S6 shows extremely high and low precipitation over northern and southern parts of Canada West, respectively, while Node S8 shows the opposite spatial precipitation distribution (i.e., low in the north but high in the south). However, Node S7 (S9) shows positive (negative) precipitation anomalies across Canada West, even though some regions in Canadian Prairies show negative (positive) precipitation anomalies (Fig. 9a). However, extremely positive |IVT|′ may or may not necessarily concurrent with positive precipitation anomalies and/or high probabilities of occurrences of extreme precipitation, since intense moisture fluxes may just pass through a region without precipitating over that region. In contrast, some negative |IVT|′ could associate with positive precipitation anomalies and high probabilities of extreme precipitation. Therefore, dry pattern clusters defined by |IVT|′ (Table 1)

do not necessary correspond to widespread low precipitation, and/or low occurrence frequency of extreme precipitation events across Canada West and East. The relationships between LSMPs and Canadian precipitation, and changes in the Canadian precipitation related to the occurrence of each LSMP for summer and fall in Canada West and East. Canada West in summer (Figs. 9a, 11a) Widespread positive precipitation anomalies across Canada West are linked with Nodes S5 and S7 where troughs of low-pressure centered over eastern Pacific and west coast of North America (Fig. 5a), while northern or Arctic Canada shows extremely high precipitation in Nodes S1, S2, and S6 where extremely high |IVT|′ tends to be located. The anomalously low summer precipitation over southern British Columbia and Canadian Prairies generally occur in Nodes S1–S3 and S6. The dry pattern (Node S9) corresponds to widespread negative precipitation anomalies across Canada West. The high occurrence frequency of extreme precipitation also shows a large spatial variability, such as the Canadian Prairies and the Canadian Rockies are linked with Nodes S5 and S9 and

13

X. Tan et al.

Fig. 8 One day lead transitional frequencies between the 4 pattern clusters defined in Table 1

northern Canada with Nodes S6–S8. The unexpected high occurrence probability of extreme precipitation under the dry pattern (Node S9) may be due to the frequent and heavy convective rainfall of the summer, or extremely low atmospheric temperature (Fig. 5a) that facilitates anomalously low moisture to precipitate. Nodes S2 and S3 (S5 and S7) accounting for 13.9% (21.4%) of days correspond to extremely dry (wet) condition over British Columbia and the Canadian Rockies, while Node S6 (S5) corresponds to dry (wet) condition over the Canadian Prairies. Up to 89% of extreme precipitation events throughout southwestern Canada including southern British Columbia, southern Canadian Rockies and the Canadian Prairies are associated with Nodes S5, S7 and S9, while extreme precipitation events over northern Canada are mostly concurrent with Node S6 in which the westerly resulted from a northeastward shift in the Aleutian low, transporting moisture-rich air to northern Canada. Seven out of nine (five out of nine) patterns are related to the increase (decrease) in the extreme precipitation events (seasonal precipitation), but only two (one) are (is) statistically significant for the entire Canada West. However, for Node S9 which occurred most frequently, both the occurrence and the associated extreme precipitation events had decreased over 1958–2013, which was partly why the occurrence of extreme precipitation in summer had decreased over Canadian Prairies (Fig. 2c; Sect. 3.1; Shook and Pomeroy 2012; Tan and Gan 2017). Canada East in summer (Figs. 9b, 11b) Regions showing extremely high precipitation and frequent occurrences of

13

extreme precipitation are generally consistent with regions showing extremely high |IVT|′, except Node S9 where, similar as Node S9 for Canada West in summer, extremely low |IVT|′ are associated with high occurrence probabilities of extreme precipitation. Thus, wet conditions and extreme precipitation events over Canada East are generally associated with North and South pattern clusters (Table 1) where positive |IVT|′ are located in Canada East. Days exhibiting similar LSMPs as the North and South pattern clusters had significantly increased. However, the trough shifted southwest over Canada East (Node S9; Fig. 5b) has also led to positive precipitation anomalies and more frequent occurrences of extreme precipitation events, even though the occurrence of Node S9 and its associated precipitation events (Fig. 9b) had significantly decreased. The summer precipitation extremes which occur in seven out of nine patterns (Fig. 11b) had increased, and four of these increasing trends are statistically significant, implying an increase in occurrences of summer extreme precipitation events over Canada East (Fig. 2c). Canada West in fall (Figs. 10a, 12a) Except the dry pattern Node S9 where widespread low precipitation had occurred, the contrasting regional distribution of low and high precipitation across Canada West in the fall are more significant than that in the summer. For example, Nodes S1, S2, and S8 show anomalous dry condition over northwestern Canada while anomalous wet condition over southwestern Canada West (Fig. 10a). A wet climate over Canadian Prairies (Nodes S2 and S8) is associated with a trough extending from the tropic Pacific and the western United States, while


Fig. 9 Summer daily precipitation anomalies over Canada West (a) and Canada East (b) which was derived from high-resolution ANUSPLIN precipitation dataset mapped onto 3 × 3 SOM |IVT|′ patterns shown in Fig. 3. Daily precipitation anomalies were calculated as the percentage of mean daily precipitation occurred in summer days with a SOM |IVT|′ pattern in each grid cell relative to the 1958–2013 mean daily precipitation at each grid cell in all summer days. The upper and lower numbers in the top left corner of each plot show the area-weighted average daily precipitation (mm d ay− 1) over Canada West (a) and Canada East (b) in each SOM |IVT|′ pattern,

and its magnitude of trend ( 10− 3 mm d ay− 1 year− 1) over 1958–2013, respectively. Red (blue) numbers shows positive (negative) trends in time series of mean daily precipitation. Statistically significant trends are shown in bold and italic text. The mean daily precipitation for all days in summer over Canada West (a) and Canada East are 1.479 and 1.756 mm day− 1, respectively. For convenience, three numbers of the characteristics of the occurrence of each pattern are also shown in Fig. 5 as that in Fig. 3, at the bottom left corner of Canada West (a) and middle right of Canada East (b)

a dry climate over Canadian Prairies resulted from a ridge extending from the central Pacific (Figs. 6a, 10a). The regions with high occurrence probability of extreme precipitation events (northern Canada West in Nodes S3 and S5, southwestern Canada in Node S6, and the east of Canada in Node S8) are consistent with the anomalously high precipitation. As expected, extreme precipitation events over Canada West in fall had occurred more frequently under Nodes S5–S8 (Fig. 12a) where Canada West generally experienced an anomalously high IVT (Fig. 4a), resulting from the zonal flow that drives moisture from the North Pacific to western Canada (Fig. 6a). However, extreme precipitation are very likely to occur under Node S9 where extremely low |IVT|′ and temperature anomalies has occurred across Canada West (Fig. 4a). Fall precipitation had decreased under seven out of nine Nodes (but only the decrease under Node S9 is statistically

significant) but increased under Nodes S3 and S4 where anomalously high precipitation over the North and low precipitation over the South had occurred. The total decreasing magnitude under seven nodes is much higher than the total increasing magnitude under other two nodes, which results in an overall decrease in the fall precipitation over Canada West (Sect. 3.1). Similarly, the occurrence of extreme fall precipitation events had decreased under six out of nine Nodes but had increased under Nodes S2, S3 and S6 where there is a relatively high probability that extreme precipitation events occur over the north and southern British Columbia. As an increase (decrease) in the occurrence of patterns leads to a higher (lower) probability of the occurrence of extreme precipitation events over the northern and southern British Columbia (Canadian Prairies), the occurrence of extreme precipitation events in the fall has increased (decreased) over the

13

X. Tan et al.

Fig. 10 Same as Fig. 9, but for fall daily precipitation anomalies. The mean daily precipitation for all days in summer over Canada West (a) and Canada East (b) is 1.191 and 1.609 mm d ay− 1, respectively

northern and southern British Columbia (Canadian Prairies; Fig. 2d). Canada East in fall (Figs. 10b, 12b) Positive fall precipitation anomalies over Canada East are associated with weak low-pressure systems. The landmass of Canada East with high precipitation in the fall (regions of mid-tropospheric convergence) is located to the left of a trough (Nodes S1, S4 and S5), while low fall precipitation (regions of mid-tropospheric divergence) is located to the right of a trough (Nodes S2, S3, and S6–S9; Fig. 10b). Extreme fall precipitation events have occurred more frequently in regions exposed to extremely high |IVT|′ (Nodes S4 and S5; Fig. 12b) and extremely low |IVT|′ (Nodes S8 and S9; Fig. 12b). There has been an increase in the occurrence of extreme fall precipitation events but a decrease in the fall precipitation over Canada East under most LSMPs identified by SOM |IVT|′ patterns, even though occurrences of LSMPs did not change significantly. However, as the total increase in the magnitude of precipitation under Nodes S1 and S3 (0.026 mm day− 1 year− 1) is much higher than the total decrease in the magnitude of precipitation under other nodes (0.001–0.014 mm day− 1 year− 1), the regional average precipitation in fall over Canada East has increased (Sect. 3.1).

13

3.4 Teleconnections MEI, PDO, PNA, NAO and AO daily indices were teleconnected to each of the nine patterns. Figure 13 shows differences in the synoptic type frequency distribution between positive and negative teleconnection conditions. Although the separation between positive–negative pair frequencies by teleconnection is generally poor, 52% (94 out of 180) positive/negative pairs of synoptic type frequencies were significantly different statistically. The effects of certain teleconnections on the occurrences of LSMPs identified by SOM of |IVT|′ patterns are statistically significant. Further, analyses for 0-, 1-, 2-, and 3-month lag time composites of teleconnection conditions for the occurrence of each |IVT|′ pattern produced similar results. The effects of PNA on the occurrence of patterns associated with precipitation are mostly shown in summer and fall over Canada West. For example, summer Nodes S1–S4 (S6–S8) related to dry (wet) climate and less (more) frequent extreme precipitation events tend to occur during the negative (positive) phase of PNA. Fall Nodes S5–S8 (S9) associated with the high probability of occurrence of extreme precipitation in fall (extremely low IVT) are also more likely


Fig. 11 Probability of heavy precipitation occurred in each SOM |IVT|′ pattern (shown in Fig. 3) in summer over Canada West (a) and Canada East (b) derived from high-resolution ANUSPLIN precipitation dataset. The upper and lower numbers in the top left corner of each plot show the mean occurrence of heavy precipitation events (days) in each SOM |IVT|′ pattern area-weighted averaged over Canada West (a) and Canada East (b), and its magnitude of trend (10− 3 days year− 1) over 1958–2013, respectively. Red (blue)

numbers shows positive (negative) trends in time series of the occurrence. Statistically significant trends are shown in bold and italic text. The mean occurrence of extreme precipitation events in all days of each SOM |IVT|′ pattern for Canada West (a) and Canada East (b) are 0.232 and 0.256 days, respectively, which is the mean seasonal occurrence of extreme precipitation (2.1 and 2.3 days per season based on the threshold of 95th percentile of non-zero precipitations) divided by 9 SOM |IVT|′ patterns

to occur during the positive (negative) phase of PNA. For Canada East, PNA only shows significant influence on the occurrence of particular LSMPs, e.g. Nodes S9 for both summer and fall. There had been strong teleconnections between AO/NAO and the occurrence of various patterns in summer and fall, especially for Canada East. Since AO and NAO are strongly correlated to each other positively, and results are similar to each other, we will only discuss NAO that has exerted more significant influence on occurrences of LSMPs than AO. During summer over Canada East, Nodes S4, S5, S7, and S8, which correspond to a dry climate and less frequent extreme precipitation events over the south of Canada East, mostly occurred during the positive phase of NAO. Conversely, Nodes S1 and S2 which lead to a dry climate and less frequent extreme precipitation events over the north of Canada East are associated with the negative phase of NAO. During fall, Nodes S5 which is linked with a wet climate and more frequent extreme precipitation events occurring over

the south of Canada East mostly occurred during the positive phase of NAO, while Nodes S2 and S3 which lead to a dry climate and less frequent extreme precipitation events widespread over Canada East are associated with the negative phase of NAO. ENSO represented by MEI is significantly linked with the occurrence of some particular patterns, but there were only slight differences in the occurrence frequency of synoptic patterns between positive and negative MEI. The weak correlation between ENSO and synoptic patterns is consistent with results found by other studies which show that the correlation between ENSO and Canadian climate is weak and region-dependent in both summer and fall (e.g., Bonsal and Shabbar 2008; Bonsal et al. 2001; Gan et al. 2007). Positive and negative phases of PDO show similar but slightly stronger influence than ENSO on the occurrence of synoptic patterns over Canada West. In summer, Nodes S8–S9 which are associated with wet climate and frequent occurrence of extreme precipitation events over the south of Canada West

13

X. Tan et al.

Fig. 12 Same as Fig. 11, but for fall over Canada West (a) and Canada East (b). The mean occurrence of heavy precipitation events in each SOM |IVT|′ pattern for Canada West (a) and Canada East (b) are 0.269 and 0.324 days, respectively, which is the mean seasonal occur-

rence of extreme precipitation (2.4 and 2.9 days per season based on the threshold of 95th percentile of non-zero precipitations) divided by 9 SOM |IVT|′ patterns

are more likely to occur during positive phases of ENSO (El Niño) and PDO.

1. Synoptic patterns that had impacted precipitation of Canada West and East in summer and fall were featured by various spatial clusters of |IVT|′, IVT fields, pressure highs and lows, troughs and ridges over North America, North Pacific, North Atlantic and Arctic. The GPH and IVT patterns identified are season dependent, the strength and center of low and high pressure, and the location and direction of ridges and troughs which varied significantly. Extremely high |IVT|′ over the central Canada West tends to be associated with the Aleutian low and Gulf cyclone across Canada West and Alaska, forcing moisture from North Pacific to Canada West. For Canada East, patterns showing widespread positive |IVT|′ tend to relate to a strengthened and southwestshifted trough across Canada East. 2. As expected, the transition between LSMPs generally shows the movement and evolution of ridges and troughs, and low- and high-pressure systems. Regional positive (negative) precipitation anomalies are generally associated with regions of mid-tropospheric convergence (divergence), located to the right (left) of the ridge axis and left (right) of the trough, which is more evident

4 Summary and conclusions This study applied the SOM algorithm on |IVT|′ over Canada West and East to analyze LSMPs associated with seasonal precipitation totals and widespread precipitation extremes in summer and fall seasons. Changes in the occurrence of LSMPs, which consist of large-scale synoptic moisture transport and circulation patterns presented in terms of the frequency, persistence and maximum duration of each LSMP identified by the SOM analysis, were detected. An 1-day lag transition, temporal relationships between LSMPs were estimated. Trends in seasonal precipitation totals and the occurrence frequency of extreme precipitation events associated with each LSMP were also detected. The effects of largescale climate anomalies on the occurrence of LSMPs were also analyzed using composite values of climate indices. The findings of the study are summarized as follows:

13


Fig. 13 Comparison of synoptic type frequency distribution between positive and negative teleconnection conditions. Statistically significant differences in synoptic type frequency distributions detected by

the two-sample Kolmogorov–Smirnov test are denoted with asterisk over the pair-wise bars

in Canada East. Extremely large |IVT|′ are not necessarily associated with positive precipitation anomalies and/or high probabilities of extreme precipitation events, because intensive moisture fluxes sometimes just pass over a region without precipitating. The dry patterns with anomalously low |IVT|′ are sometimes associated with a more frequent occurrence of extreme precipitation, which partly caused by an extremely low ground surface temperature that promotes the moisture flux to precipitate. However, the predominant, underlying cause for the high occurrence probability of extreme precipitation over dry patterns should be further studied. 3. Only 19.4% of the annual occurrence characteristics of LSMPs identified show statistically significant changes in summer and fall. More summer LSMPs show changes than fall LSMPs and more LSMPs throughout the year have shown changes in the frequency of occurrence than persistence and the maximum duration. Predominantly a LSMP shows a persistence over 2 days (~ 2 days). More statistically significant changes in daily precipita-

tion and occurrences of extreme precipitation related to each LSMP identified, than changes in occurrences of LSMPs, have resulted in changes of seasonal precipitation totals and occurrences of extreme precipitation across Canada. 4. LSMPs associated with a dry (wet) climate and less (more) frequent extreme precipitation events over Canada West in summer and fall tend to occur during the negative (positive) phase of PNA. LSMPs associated with a wet climate and more frequent occurrence of extreme precipitation events over the south (north) of Canada East are more likely to occur during the positive (negative) phase of NAO. LSMPs associated with wet climate and frequent occurrence of extreme precipitation events over the south of Canada West in summer are more likely to occur during positive phases of ENSO (El Niño) and PDO.

13

Acknowledgements The first author was partly funded by the China Scholarship Council (CSC) of P. R. China and the University of Alberta. We are grateful to Dan McKenney and Pia Papadopol from the Natural Resources Canada for providing the ANUSPLIN Canadian precipitation data used in this study. All analyses were conducted using the R language and figures were plotted using NCL language. The SOM algorithm was implemented in the kohonen package (Wehrens and Buydens 2007). The JRA-55 reanalysis for geopotential heights and IVT were downloaded from http://rda.ucar.edu/datasets/ds628.0/ Monthly values of the multivariate ENSO index (MEI) (Wolter and Timlin 2011) and the PDO index, and daily values of PNA, NAO and AO indices were obtained from the National Oceanic and Atmospheric Administration (NOAA) Climate Prediction Center.

References Benyahya L, Gachon P, St-Hilaire A, Laprise R (2014) Frequency analysis of seasonal extreme precipitation in southern Quebec (Canada): an evaluation of regional climate model simulation with respect to two gridded datasets. Hydrol Res 45:115–133 Bintanja R, Andry O (2017) Towards a rain-dominated Arctic. Nature Clim Change 7:263–267 Bonsal B, Shabbar A (2008) Impacts of large-scale circulation variability on low streamflows over Canada: a review. Can Water Resour J 33:137–154 Bonsal BR, Shabbar A, Higuchi K (2001) Impacts of low frequency variability modes on Canadian winter temperature. Int J Climatol 21:95–108 Bonsal BR, Aider R, Gachon P, Lapp S (2013) An assessment of Canadian prairie drought: past, present, and future. Clim Dynam 41:501–516 Brimelow JC, Reuter GW (2005) Transport of atmospheric moisture during three extreme rainfall events over the Mackenzie River Basin. J Hydrometeorol 6:423–440 Burn DH, Mansour R, Zhang K, Whitfield PH (2011) Trends and variability in extreme rainfall events in British Columbia. Can Water Resour J 36:67–82 Buttle JM et al (2016) Flood processes in Canada: regional and special aspects. Can Water Resour J 41:7–30 Cannon AJ, Sobie SR, Murdock TQ (2015) Bias correction of GCM precipitation by quantile mapping: how well do methods preserve changes in quantiles and extremes? J Clim 28:6938–6959 Cassano JJ, Uotila P, Lynch AH, Cassano EN, 2007: Predicted changes in synoptic forcing of net precipitation in large Arctic river basins during the 21st century. J Geophys Res 112:G04S49, https://doi. org/10.1029/2006jg000332 Cassano EN, Glisan JM, Cassano JJ, Gutowski WJ, Seefeldt MW (2015) Self-organizing map analysis of widespread temperature extremes in Alaska and Canada. Clim Res 62:199–218 Cohen J et al (2014) Recent Arctic amplification and extreme midlatitude weather. Nat Geosci 7;627–637 Coulibaly P (2006) Spatial and temporal variability of Canadian seasonal precipitation (1900–2000). Adv Water Resour 29:1846–1865 Dettinger M (2011) Climate change, atmospheric rivers, and floods in California—a multimodel analysis of storm frequency and magnitude changes. J Am Water Resour As 47:514–523 Dufour A, Zolina O, Gulev SK (2016) Atmospheric moisture transport to the Arctic: assessment of reanalyses and analysis of transport components. J Clim 29:5061–5081 Gan TY, Gobena AK, Wang Q (2007) Precipitation of southwestern Canada: Wavelet, scaling, multifractal analysis, and teleconnection to climate anomalies. J Geophys Res 112:D10110. https: //doi. org/10.1029/2006jd007157

13

X. Tan et al. Gibson PB, Perkins-Kirkpatrick SE, Uotila P, Pepler AS, Alexander LV (2017) On the use of self-organizing maps for studying climate extremes. J Geophys Res Atmos 122:3891–3903 Gizaw MS, Gan TY (2016) Possible impact of climate change on future extreme precipitation of the Oldman, Bow and Red Deer River Basins of Alberta. Int J Climatol 36:208–224 Gobena AK, Gan TY (2013) Assessment of trends and possible climate change impacts on summer moisture availability in western Canada based on metrics of the palmer drought severity index. J Clim 26(13):4583–4595. https://doi.org/10.1175/ JCLI-D-12-00421.1 Hannachi A, Jolliffe IT, Stephenson DB (2007) Empirical orthogonal functions and related techniques in atmospheric science: a review. Int J Climatol 27:1119–1152 Hartmann DL, et al (2013) Observations: atmosphere and surface. In: Stocker TF et al (eds) Climate change 2013: the physical science basis. Contribution of working group I to the fifth assessment report of the intergovernmental panel on climate change. Cambridge University Press, Cambridge, pp 159–254 Held IM, Soden BJ (2006) Robust responses of the hydrological cycle to global warming. J Clim 19:5686–5699 Hopkinson RF, McKenney DW, Milewska EJ, Hutchinson MF, Papadopol P, Vincent LA (2011) Impact of aligning climatological day on gridding daily maximum–minimum temperature and precipitation over Canada. J Appl Meteorol Clim 50:1654–1665 Hu C et al (2016) Shifting El Nino inhibits summer Arctic warming and Arctic sea-ice melting over the Canada Basin. Nat Commun 7:11721. https://doi.org/10.1038/ncomms11721 Huntington TG (2006) Evidence for intensification of the global water cycle: Review and synthesis. J Hydrol 319:83–95 Hutchinson MF, McKenney DW, Lawrence K, Pedlar JH, Hopkinson RF, Milewska E, Papadopol P (2009) Development and testing of Canada-wide interpolated spatial models of daily minimum– maximum temperature and precipitation for 1961–2003. J Appl Meteorol Clim 48:725–741 IPCC (2013) Summary for policymakers. In: Stocker TF et al (eds) Climate change 2013: the physical science basis. Contribution of working group I to the fifth assessment report of the intergovernmental panel on climate change. Cambridge University Press, Cambridge, pp 1–30 Kendall MG, 1975: Rank correlation methods. Charless Griffin, London Kerkhoven E, Gan TY (2011) Differences and sensitivities in potential hydrologic impact of climate change to regional-scale Athabasca and Fraser river basins of the Leeward and Windward sides of the Canadian rocky mountains respectively. Clim Change 106(4):583– 607. https://doi.org/10.1007/s10584-010-9958-7 Kobayashi S et al (2015) The JRA-55 reanalysis: general specifications and basic characteristics. J Meteorol Soc Jpn Ser II 93:5–48 Kochtubajda B, Mooney C, Stewart R (2017) Characteristics, atmospheric drivers and occurrence patterns of freezing precipitation and ice pellets over the Prairie Provinces and Arctic Territories of Canada: 1964–2005. Atmos Res 191:115–127 Kohonen T (1998) The self-organizing map. Neurocomputing 21:1–6 Lenderink G, van Meijgaard E (2008) Increase in hourly precipitation extremes beyond expectations from temperature changes. Nat Geosci 1:511–514 Lenderink G, van Meijgaard E (2010) Linking increases in hourly precipitation extremes to atmospheric temperature and moisture changes. Environ Res Lett 5:025208. https : //doi. org/10.1088/1748-9326/5/2/025208 Liu C, Barnes EA (2015) Extreme moisture transport into the Arctic linked to Rossby wave breaking. J Geophys Res Atmos 120:3774–3788 Liu J, Stewart RE (2003) Water vapor fluxes over the Saskatchewan River Basin. J Hydrometeorol 4:944–959

Synoptic moisture pathways associated with mean and extreme precipitation over Canada for… Liu J, Stewart RE, Szeto KK (2004) Moisture transport and other hydrometeorological features associated with the severe 2000/01 drought over the western and central Canadian Prairies. J Clim 17:305–319 Liu Y, Weisberg RH, Mooers CNK (2006) Performance evaluation of the self-organizing map for feature extraction. J Geophys Res 111:C05018. https://doi.org/10.1029/2005JC003117 Mekis É, Vincent LA (2011) An overview of the second generation adjusted daily precipitation dataset for trend analysis in Canada. Atmos Ocean 49:163–177 Newton BW, Prowse TD, Bonsal BR (2014a) Evaluating the distribution of water resources in western Canada using synoptic climatology and selected teleconnections. Part 2: summer season. Hydrol Process 28:4235–4249 Newton BW, Prowse TD, Bonsal BR (2014b) Evaluating the distribution of water resources in western Canada using synoptic climatology and selected teleconnections. Part 1: winter season. Hydrol Process 28:4219–4234 Phillippopoulos K, Delligiorgi D (2012) A self-organizing maps multivariate spatio-temporal approach for the classification of atmospheric conditions. In: Huang T, Zeng Z, Li C, Leung CS (eds) Neural information processing. Springer: Berlin, Heidelberg, pp 544–551 Pithan F, Mauritsen T (2014) Arctic amplification dominated by temperature feedbacks in contemporary climate models. Nat Geosci 7:181–184 Radić V, Cannon AJ, Menounos B, Gi N (2015) Future changes in autumn atmospheric river events in British Columbia, Canada, as projected by CMIP5 global climate models. J Geophys Res Atmos 120:9279–9302 Roberge A, Gyakum JR, Atallah EH (2009) Analysis of intense poleward water vapor transports into high latitudes of western North America. Weather Forecast 24:1732–1747 Schindler DW, Donahue WF (2006) An impending water crisis in Canada’s western prairie provinces. Proc Natl Acad Sci USA 103:7210–7216 Schuenemann KC, Cassano JJ (2010) Changes in synoptic weather patterns and Greenland precipitation in the 20th and 21st centuries: 2. Analysis of 21st century atmospheric changes using self-organizing maps. J Geophys Res 115:D05108. https://doi. org/10.1029/2009jd011706 Screen JA, Simmonds I (2011) Declining summer snowfall in the Arctic: causes, impacts and feedbacks. Clim Dynam 38:2243–2256 Sen PK (1968) Estimates of the regression coefficient based on Kendall’s tau. J Am Stat Assoc 63:1379–1389 Serreze MC, Barry RG (2011) Processes and impacts of Arctic amplification: a research synthesis. Global Planet Change 77:85–96 Serreze MC, Francis JA (2006) The Arctic on the fast track of change. Weather 61:65–69 Shook K, Pomeroy J (2012) Changes in the hydrological character of rainfall on the Canadian prairies. Hydrol Process 26:1752–1766 Smirnov VV, Moore GWK (1999) Spatial and temporal structure of atmospheric water vapor transport in the Mackenzie River Basin. J Clim 12:681–696

Smirnov VV, Moore GWK (2001) Short-term and seasonal variability of the atmospheric water vapor transport through the Mackenzie River Basin. J Hydrometeorol 2:441–452 Spry CM, Kohfeld KE, Allen DM, Dunkley D, Lertzman K (2014) Characterizing pineapple express storms in the lower mainland of British Columbia, Canada. Can Water Resour J 39:302–323 Swales D, Alexander M, Hughes M (2016) Examining moisture pathways and extreme precipitation in the U.S. Intermountain West using self organizing maps. Geophys Res Lett 43:1727–1735 Tan X, Gan TY (2017) Non-stationary analysis of the frequency and intensity of heavy precipitation over Canada and their relations to large-scale climate patterns. Clim Dyn 48:2983–3001 Tan X, Gan TY, Shao D (2016) Wavelet analysis of precipitation extremes over Canadian ecoregions and teleconnections to largescale climate anomalies. J Geophys Res Atmos 121:14469–14486 Tan X, Gan TY, Chen YD (2017) Moisture sources and pathways associated with the spatial variability of seasonal extreme precipitation over Canada. Clim Dynam. https://doi.org/10.1007/s0038 2-017-3630-0 Tanzeeba S, Gan TY (2011) Potential impact of climate change on the water availability of South Saskatchewan River Basin. Clim Change 112(2):355–386 Vihma T et al (2015) The atmospheric role in the Arctic water cycle: A review on processes, past and future changes, and their impacts. J Geophys Res Biogeosci 121:586–620 Vincent LA, Mekis É (2006) Changes in daily and extreme temperature and precipitation indices for Canada over the Twentieth Century. Atmos Ocean 44:177–193 Vincent LA et al (2015) Observed trends in Canada’s climate and influence of low-frequency variability modes. J Clim 28:4545–4560 Wehrens R, Buydens L (2007) Self- and super-organizing maps in R: the kohonen package. J Stst Softw 21(5):1–19 Wendisch M et al (2017) Understanding causes and effects of rapid warming in the Arctic. EOS. https: //doi.org/10.1029/2017EO 0648 03 White R, Etkin D (1997) climate change, extreme events and the Canadian insurance industry. Nat Hazards 16:135–163 Wolter K, Timlin MS (2011) El Niño/Southern Oscillation behaviour since 1871 as diagnosed in an extended multivariate ENSO index (MEI.ext). Int J Climatol 31:1074–1087 Wong JS, Razavi S, Bonsal BR, Wheater HS, Asong ZE (2017) Inter-comparison of daily precipitation products for large-scale hydro-climatic applications over Canada. Hydrol Earth Syst Sc 21:2163–2185 Zahn M, Allan RP (2013) Climate warming—related strengthening of the tropical hydrological cycle. J Clim 26:562–574 Zhang X, He J, Zhang J, Polyakov I, Gerdes R, Inoue J, Wu P (2013) Enhanced poleward moisture transport and amplified northern high-latitude wetting trend. Nat Clim Change 3:47–51

13