E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
ECLAT-2 A Concerted Action Towards the Improved Understanding and Application of Results from Climate Model Experiments in European Climate Change Impacts Research
Climate scenarios for agricultural, forest and ecosystem impacts
ECLAT-2 Workshop Report No. 2 Potsdam, Germany, 13-15 October 1999
Edited by: Wolfgang Cramer, Ruth Doherty, Mike Hulme and David Viner
Published by the Climatic Research Unit, UEA, Norwich, UK. July 2000 (Further copies of this report can be obtained from
[email protected])
1
W O R K S H O P
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Contents page numbers Introduction to ECLAT-2 Mike Hulme and David Viner
4
CLIMATE SCENARIOS FOR AGRICULTURAL, FOREST AND ECOSYSTEM IMPACTS WORKSHOP INTRODUCTION Wolfgang Cramer
8
Key Findings from the first ECLAT-2 Workshop - “Representing Uncertainties in Climate Change Scenarios and Impact Studies” (Helsinki, Finland) Timothy R. Carter and Mike Hulme
10
Creating Climate Scenarios for the European ACACIA assessment Mike Hulme and Timothy R. Carter (presented by Wolfgang Cramer)
15
KEYNOTE PAPERS 1. Agriculture and Climate Change Scenarios Mark D.A. Rounsevell
21
2. Assessing the Impact of Climate Change on Forests and Forestry using Climate Scenarios: Current Practice and Future Directions David T. Price and Mike D. Flannigan
32
3. Development of Climate Change Scenarios for Agricultural Applications Mikhail A. Semenov and Elaine M. Barrow
50
4. Incorporating Dynamic Vegetation Cover within Global Climate Models 1 Jonathan A. Foley, Samuel Levis, Marcos Heil Costa, Wolfgang Cramer and David Pollard
59
WORKING GROUP REPORTS Introduction to Working Groups Wolfgang Cramer
77
1. Climate Scenarios for Biospheric Impact Assessments Stephen Sitch and Navin Ramankutty
78
2. Climate Scenarios for Forests Impact Assessments Harald Bugmann
80
1
During the workshop, Navin Ramankutty (Univ. of Wisconsin) gave an oral presentation with related material under the title
"Integrated modelling of the Earth system: bi-directional atmosphere-biosphere interactions" - we here reproduce, with permission 1
by the Ecological Society of America, the text of a review paper summarising the topic due to appear in Ecological Applications during 2000.
2
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
page numbers
3. Climate Scenarios for Agricultural Impact Assessments Ana Iglesias and David Favis Mortlock
84
4. Global and Continental Scale Impact Assessments Benjamin Smith
87
5. Regional Scale Impact Assessments Paula Harrison
90
6. Stand-level Scale Impact Assessments Frank Wechsung
95
POSTER PRESENTATIONS 1. Modelling the Risk of Windthrow in Forestry under Changing Climatic conditions Kristina Blennow, Markku Rummukainen and Ola Sallnäs 2. Impact of Crops on Biomass and Soil Carbon: Steady State Simulations Bernard Nemry, Louis François, Jean-Claude Gérard, Dominique Otto, Daniel Rasse and Pierre Warnant
CLIMATE SCENARIOS FOR AGRICULTURAL, FOREST AND ECOSYSTEM IMPACTS - WORKSHOP SUMMARY Wolfgang Cramer
98
101
107
APPENDICES 1. Workshop Agenda
110
2. List of Participants
113
3. Useful Web Sites
118
3
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Introduction to ECLAT-2 Mike Hulme and David Viner Climatic Research Unit, School of Environmental Sciences, University of East Anglia, Norwich, Norfolk NR4 7TJ. (
[email protected]) ECLAT-2 is a Concerted Action Initiative funded through the Climate and Environment Programme of DGXII of the European Commission. The Initiative started in July 1998 and will run until June 2001. ECLAT-2 is co-ordinated by the Climatic Research Unit at the University of East Anglia, Norwich, UK, and has a Steering Committee comprising representatives from 12 other organisations in Europe. ECLAT-2 Objectives Some of the world’s leading climate modelling centres are based in Europe and with a large number of high quality interdisciplinary research teams examining the effects of climate change on a wide range of natural, managed and social systems, Europe can also claim to be at the forefront of research into the impacts of climate change. The European Union continues to fund both of these areas of research through the Climate and Environment Programme of DGXII. Further enhancement and exploitation of these research successes requires improved co-ordination and efficiency of flows of information between the climate modelling, observed climate data and impacts research communities. More importantly, users of climate model results need to be kept abreast of developments in climate modelling and, through working together with scientists engaged in climate modelling, need to improve the ways in which climate model results are applied and interpreted in impacts research. ECLAT-2 therefore has two specific objectives: Primary Objective: to improve the understanding and application of results from climate model experiments in EU climate change impacts research projects. Secondary Objective: to keep EU researchers into the impacts of climate change abreast of developments in climate modelling and informed about the availability of results from new climate change experiments performed in Europe and worldwide. These two objectives are being achieved primarily through a series of four workshops initiated, designed and run by the ECLAT-2 Steering Committee, in association with climate impacts researchers throughout Europe. Each of these workshops involves between 20 and 40 participants drawn from the main European climate modelling centres and from the European climate change impacts research community. A larger ECLAT-2 Forum meeting will be organised towards the end of the Concerted Action to review the advancement of knowledge and methodology represented through the four ECLAT-2 workshops. An ECLAT-2 web site has been established (http://www.cru.uea.ac.uk/eclat) and the four workshop reports will be published and widely distributed within Europe and beyond. ECLAT-2 therefore provides a focal activity for improving the understanding and application of climatological data within on-going and proposed future climate change impacts projects of the EU Climate and Environment Programme. The ECLAT-2 workshops are also complementary to, and supportive of, other major international climate scenario activities, such as those occurring under the Intergovernmental Panel on Climate Change (IPCC), the International Geosphere Biosphere Programme (IGBP) and the Human Dimensions of Global Environmental Change Programme (HDP).
4
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
The Management of ECLAT-2 The ECLAT-2 Partners and Steering Committee consist of representatives from each of the four major climate modelling centres in Europe (the Hadley Centre, MPI/DKRZ, Meteo-France, LMD/IPSL) and representatives from the climate scenario construction and climate change impacts communities. All of the ECLAT-2 Partners (see list below) are automatically members of the Steering Committee, but additional people may be co-opted onto the Steering Committee during the lifetime of ECLAT-2. We also list four organisations who have a special consultative role in ECLAT-2, but which for various reasons are not full Partners to the Initiative. The Steering Committee liaises primarily through electronic means and through occasional meetings that coincide with other Workshops or Conferences. One full meeting of the Steering Committee is organised each year. The Steering Committee formulate and guide the ECLAT-2 Workshop series and the Forum meeting, although these Workshops are hosted locally by members of the ECLAT-2 Network. The ECLAT-2 Project is co-ordinated by the Climatic Research Unit (CRU) from where the ECLAT-2 email list and website will be run and from where the Workshop Reports will be published. Readers of this Report can be added to the ECLAT-2 email list by contacting Dr David Viner at
[email protected]. The ECLAT-2 Contracted Partners • • • • • • • • •
CRU, Climatic Research Unit, Dr Mike Hulme and Dr David Viner Department of Geography University of Southampton, Dr Nigel Arnell KNMI, Royal Netherlands Meteorological Institute, Dr Jules Beersma MTT, Agricultural Research Centre of Finland, Dr Timothy Carter ICAT, University of Lisbon, Professor Joao Corte-Real PIK, Potsdam Institüt für Klimafolgenforschung, Professor Wolfgang Cramer DMI, Danish Meteorological Institute, Dr Eigal Kaas LMD, Laboratoire de Meteorologie Dynamique, Professor Katia Laval Meteo-France, Dr Serge Planton
ECLAT-2 Non-Contracted Partners • • • •
Hadley Centre, Dr Geoff Jenkins MEDIAS-France, Michel Hoepffner EUMETET/ECSN, José Diaz DKRZ, Deutsches Klimarechenzentrum, Dr Ulrich Cubasch
The ECLAT-2 Workshops These meetings form the main activity of the ECLAT-2 Concerted Action. A series of four Technical Workshops are planned and these will be held at approximately six monthly intervals. A larger ECLAT-2 Forum meeting will be held towards the end of the three year Concerted Action. Each ECLAT-2 Workshop will possess the following characteristics and will last between two and three days: • Workshop themes and broad design agreed by the ECLAT-2 Steering Committee. • Discussion papers (either single or multi-authored) on the relevant Workshop theme(s) commissioned by the organisers. • Workshop organisation and hosting shall be delegated to members/institutes in the ECLAT-2 Network. These hosts need not be Partners to the proposal. 5
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
• Representatives of climate modelling centres and climate change impacts research projects to be present at each Workshop. • Participation to be limited to between 20 and 40 people, some places being reserved for invitation only. • Expenses of participants to be met by ECLAT-2 up to an agreed maximum limit. Workshop hosts will be responsible for distributing travel and subsistence costs, with a set budget for each Workshop. • A consolidated Workshop report to be prepared and published after each Workshop. The four Workshop themes, venues and approximate dates are as follows: EW-1 (RED): Helsinki (FMI), Finland, April 14th-16th 1999 “Representing Uncertainty in Climate Change Scenarios and Impact Studies” This workshop will (1) review the various sources of uncertainty in climate change and related scenarios, their relative importance and their representation in climate change impact studies, and (2) work towards a series of recommendations for a more systematic treatment of uncertainty when designing and applying climatic scenarios for impact studies. Specific topics considered will include: uncertainties in social and economic projections, uncertainties in representing observed climate, climate modelling uncertainties, uncertainties in climate impact modelling, and an assessment of techniques for estimating uncertainty in climate change studies. The two-day workshop will comprise a series of short invited papers and commentaries, with a substantial amount of time dedicated to group discussion and the preparation of the workshop report. EW-2 (GREEN): Potsdam (PIK), Germany, October 13th-15th 1999 “Climate Scenarios for Agricultural, Ecosystem and Biological Impacts” This workshop will review the construction and application of climate change scenarios in areas related to agriculture, ecosystems and other biological indicators. Specific topics to be covered will include the incorporation of different time-scales of climate variability in climate change scenarios, enhancing consistency between climate and impact models, and representing the direct effects of CO2 and other atmospheric constituents alongside the impacts of climate change. EW-3 (BLUE): de Bilt (KNMI), Netherlands, May 10th-12th 2000 “Climate Scenarios for Water-related and Coastal Impacts” This workshop will review the construction and application of climate change scenarios in areas related to water and coastal indicators. Specific topics to be covered will include the specification and incorporation of daily and sub-daily weather extremes in climate change scenarios, the use of weather generators and statistical downscaling techniques, and the interaction between sea-level change, storms and storm surges. EW-4 (BROWN): Toulouse, France, October 25th-27th 2000 “Applying Climate Scenarios for Regional Studies: with particular reference to the Mediterranean” This workshop will review the construction and application of climate change scenarios in regional-scale impacts assessments. Particular reference will be made to the Mediterranean region, but examples will be drawn from other regional studies in Europe. Specific topics to be covered will include the use of Regional Climate Models in regional scenario construction, the achievement of consistency between point, local and regional-scale scenarios, and the integration of climate impacts results across regions. Towards the end of ECLAT-2, a larger Forum meeting will be held. This meeting will be designed to draw together the conclusions from the Workshop series and to present these to a wider audience, including representation from DGXII and maybe other relevant EU Directorates, and representatives from all of the 6
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
climate impacts projects funded by the EU, especially new projects funded under the 5th Framework Programme. It will also be an opportunity to identify priorities for future research in this area, again drawing upon the conclusions of the four ECLAT-2 Workshops. This Forum meeting would be held over two days with participation of around 50 people. The Benefits of ECLAT-2 The primary objective of the ECLAT-2 Concerted Action is to facilitate the better understanding and application of climate model results within on-going and future EU Climate and Environment Programme’s climate change impacts research projects. The secondary objective is to enable improved communication between climate modellers and users of climate model results and to allow climate change impacts researchers to be kept abreast of developments in climate modelling in Europe and about the availability of results from climate model experiments. Improved construction, and more consistent application, of climate change scenarios within EU climate impacts research projects allows for more effective integration and continent-wide interpretation of the results from these research projects. It also enables the strength of European climate change modelling activities to be better exploited by EU climate impacts researchers. ECLAT-2 strengthens the European contribution to international climate change activities such as the IPCC and the policy development under the UN Framework Convention on Climate Change. Within the IPCC, for example, a requirement for comparability in regional impacts assessments for the Third Assessment Report (TAR) due in the year 2001 has already been recognised and one or more chapters dealing with issues of climate scenario construction and application are planned for the TAR. ECLAT-2 activities coalesce European expertise in these areas and enable European scientists to make a more effective contribution to IPCC. The ECLAT-2 Network allows for better communication within the impacts projects of the EU Environment and Climate Programme and between scientists involved in climate modelling, climate scenario construction and climate scenario application.
7
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Climate Scenarios for Agricultural, Forest and Ecosystem Impacts - Workshop Introduction Wolfgang Cramer Department of Global Change and Natural Systems, Potsdam Institute for Climate Impact Research, PO BOX 60 12 03, D-14412 Potsdam, Germany (
[email protected]) Moving from occasional studies of ‘some’ possible change in climate and ‘some’ interesting sector of impacts, climate change impact assessment studies now are entering a new level of technical and scientific maturity. The reports of the IPCC Working Group II as well as many national and international assessments, have confirmed that the sensitivity of ecosystems, and of the economic sectors depending on them, is generally high for the currently projected range of climate change. It is also highly variable across regions, and for different conceivable emission pathways and climate scenarios. At this point, expert assessments, which are in great demand by policy-makers and others, face a serious difficulty, since they are requested to provide quantitative information in such a way that adaptation options can be tested. Underlying this difficulty is, however, not ‘just’ the problem of predicting climate and atmospheric CO2 concentrations some decades into the future, but also the need for consistency with respect to the techniques to process such information. If one scenario is used for checking the sensitivity of one sector, and a different scenario for another sector (as has frequently occurred in IPCC reports), then little generic information can be provided to policymakers. The ECLAT-2 “Green” Workshop was initiated to address this situation, to discuss available scenario construction techniques and to provide a blueprint of ‘best practice’ for the use of climate change scenarios in agricultural, forest and ecosystem impacts analyses. It is hardly possible to overemphasise the importance of the study of managed and natural terrestrial ecosystems in this context. They provide the basis of human food supply, as well as other goods (e.g., timber) and non-commercial values (e.g., water resources). Ecosystems are well adapted to regional climatic characteristics, which vary widely, e.g., in Europe from semi-arid, fire-prone scrub through rich temperate mesic ecosystems to the high Arctic cold desert. A shift of climatic regimes across the continent, as is projected by several climate models, would not only cause change in productive potential at an unprecedented rate, but might also result in serious shortages of products in regions that depend on them for a significant part of their economy. These changes can presently not be predicted in a strict sense, for several reasons: • despite more than a decade of climate impact research, the sensitivity of ecosystems to climate change is incompletely known, • the interactions between multiple climatic trends in particular are complex and not well understood (e.g., how does an increase in atmospheric CO2 affect a trend towards water shortages in certain regions?), • after a certain point, regional changes in vegetation cover are known to affect climate processes as well, i.e., feedbacks between the terrestrial and atmospheric systems add further complexity to the issue. Most important, however, is that ecosystems are regionally patchy due both to climatic gradients and also to the high regional heterogeneity of topography, soils and land use. Mostly due to computational limitations, climate models, and hence the scenarios derived from them, can only reflect the broadest features of this pattern – and these features tend to be more uncertain than the overall trends shown by climate models. Explicit ‘predictions’ of the kind that are requested by many stakeholders are therefore presently impossible. Nevertheless, to provide a measure of sensitivity to likely changes, it has become
8
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
common place to circumvent the problem of scale by one of two methods: either by limiting the scale of the assessment to the coarse pattern provided by the climate model or by applying some means of systematic re-scaling between the coarse scaled scenario and the finer scaled pattern of possible impacts. Since the former, more rigorous, method only provides results of very limited value, a growing body of literature has focused attention towards the development of suitable re-scaling techniques. Many different methods of scaling exist and a range of them is reviewed in this report. Common to them are the following issues which need to be resolved in every assessment: • the temporal structure of climate information is related to the spatial structure of the information, e.g., if the impact assessment requires the precise resolution of individual rainstorms, then this places strong demands not only on estimates of spatial extent of these storms, but also of their duration and frequency – scaling model output with respect to its spatial structure only, will likely give an inappropriate picture. • requirements for spatial and temporal structure of the climate scenario differ widely between different impact sectors, e.g., forestry-related assessments often require long series of approximate climatic information without the need for high temporal or spatial accuracy, while flood-risk assessments usually need the opposite. • despite centuries of measurements, the present-day climate remains incompletely known with respect to spatial and temporal patterns – there is therefore no rigorously and adequately defined baseline with which model-derived climate scenarios can be compared at all scales. The ECLAT-2 Concerted Action represents an effort to consolidate these issues. The “Green” workshop presented in this report and held at the Potsdam Institute for Climate Impact Research in Potsdam, Germany, on October 13-15, 1999, pursued the goal of looking at ecosystem-related impact sectors with respect to their requirements of climate change scenarios. The workshop brought together keynote speakers who reviewed scaling techniques for different sectors, and it provided a platform for Working Groups exchanging experiences and ideas for ‘best practice’. As an introduction, participants first reconsidered the findings of the first ECLAT-2 workshop the “Red” workshop, held in Helsinki, Finland, on the crucial issue of uncertainty in climate scenarios and impact studies based on them. Then, a very brief presentation outlined the climate scenario construction practice adopted by the European ACACIA impact assessment. Two sectoral keynote papers were then given, on agriculture (Mark Rounsevell) and on forestry (David Price). A third keynote paper on downscaling techniques in general was then presented (Mikhail Semenov). After the Working Group discussions (introduced later), a final outlook-oriented keynote paper was given by Navin Ramankutty on global feedbacks between climate change and the terrestrial biosphere.
9
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Key Findings from the First ECLAT-2 workshop “Representing Uncertainties in Climate Change Scenarios and Impact studies” 1
2
Timothy R. Carter and Mike Hulme
1. Impacts Research Division, Finnish Enviroment Institute, Box140, Fin-00251, Helsinki, Finland. (
[email protected]) 2. Climatic Research Unit, School of Environmental Sciences, University of East Anglia, Norwich NR4 7TJ, UK (
[email protected])
“If you don’t know what you want, all roads will lead you there” General observations on uncertainty A number of general conclusions emerged from discussions at the first ECLAT-2 Workshop concerning the nature and magnitude of uncertainties surrounding estimates of future climate change and its impacts. These can be categorised according to different links in the causal chain of effects from socio-economic driving factors, through consequent climate changes and sea-level rise, to the impacts of, and adaptations to, climate change. • Driving factors. Future developments of socio-economic driving factors are largely unknowable, and cannot be assigned objective probabilities. For this reason, it is suggested that a range of scenarios (e.g., the SRES emissions scenarios) be applied in impact assessments rather than a single best guess or average case. • Climate change. Uncertainties in estimates of future climate change due to large-scale climate processes are probably more important for assessing the likely range of most impacts than uncertainties in resolving sub-grid scale details about future climate change. While regional climate models may provide more credible information on changes in climate than GCMs in regions of heterogeneous terrain (e.g., land-sea boundaries, mountain areas), these models nevertheless exhibit uncertainties in estimates of large-scale climate change that are comparable to those found in GCMs. • Sea-level rise. Uncertainties in estimated regional sea-level rise due to thermal expansion for a given global-mean rise are comparable to uncertainties in estimates of mean global sea-level rise. • Impacts. Uncertainties in estimates of future impacts are highly dependent on the sector, exposure unit and region of interest. For instance, climate change and direct CO2 effects on crop productivity are often stronger (and the uncertainty lower) in climatically marginal regions than in core producing regions. However, these uncertainties can be dwarfed by uncertainties in future productivity associated with future agricultural policy, management and technology. Furthermore, the relationship between key scenario/impact uncertainties may change and switch priority depending on the time horizon considered (for example, global-mean temperature changes estimated for the four SRES “marker” emissions scenarios describe only a narrow range up to 2050, but a much wider range by 2100). Nevertheless, it is not necessarily the case that uncertainties always increase at subsequent stages in the scenarios/impacts analytical cascade; they might narrow due to differing levels of adaptive capacity. 10
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
How effectively do scientists treat uncertainty? A consensus view from the first ECLAT-2 Workshop was that scientists have failed to address uncertainties adequately in the great majority of climate impact assessments conducted to date. A general point concerns the identification of different types of uncertainty. Various typologies were proposed at this first Workshop, including: 1.
2. 3.
A threefold classification of primary sources of uncertainty: (i) measurement error, (ii) variability (including inherent randomness and variability over temporal and spatial scales), and (iii) model structure (i.e, model simplifications, omissions and mis-specifications). A fourfold classification, according to the general sources of uncertainty: (i) theory, (ii) data, (iii) assumptions, and (iv) scenarios. A twofold classification distinguishing (i) ‘probable error’, which is uncertainty attributable to factors that are inherently unpredictable (also referred to as unknowable knowledge), and (ii) ‘probable bias’, which is uncertainty due to imperfect knowledge and which is liable to change over time with improved understanding.
These alternatives are not exhaustive, but it is evident that many previous impact studies have either ignored these types of distinctions or have confused them. More specific points concerning the treatment of uncertainty can be illustrated for the same categories as outlined in the previous section: • Driving factors. There has been a temptation to adopt ‘best guess’ scenarios of future driving factors of global change, based either on averaging of alternative scenarios or on selection of the middle or other preferred case. The former approach can lead to inconsistencies between scenario components that are not present in the individual scenarios being averaged. The latter approach can lend the appearance of higher likelihood to the central case, as is well illustrated by the widespread adoption of the IS92a scenario of greenhouse gas emissions as a reference, although it was only one of six alternative emissions scenarios produced by the IPCC in 1992. It is now acknowledged that uncertainties can only be explored adequately if a set of alternative scenarios are adopted, each offering a self-consistent view or storyline of the future. • Climate change. Errors and homogeneities of observed climatological data are fairly well recognised and many of them have been quantified. These data are commonly applied as a reference in impact assessments, but their uncertainties are rarely accounted for even though it would be a fairly straightforward exercise to evaluate them. Scenarios of future climate from models are commonly applied in impact assessments, and there are emerging techniques to account for the range of uncertainty in emissions, climate sensitivity, regional climate change and natural climatic variability. However, these are seldom applied systematically in impact assessments. The application of expert judgement-based methods to obtain probabilistic assessments of future climate is a controversial approach that is gaining in acceptance in some quarters. However, it exhibits many pitfalls, and in the opinion of some participants at the Workshop these methods rely on ‘belief’ as much as on scientific understanding. Assumptions about prior probabilities in such exercises need to be made explicit. • Sea-level. Risk assessment methods, employing the use of simple models and expert judgement, have been applied to global sea-level rise to obtain quasi-probability distributions of future outcomes. However, this approach is not yet feasible for regional sea-level rise, due to the requirement to apply more complex global models, and the limited number and differing assumptions of model simulations conducted to date.
11
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
• Impacts. Some impact assessments include systematic uncertainty analyses and sensitivity analyses of impact models. These exercises can be very helpful in defining the level of confidence in model estimates, but are not commonly undertaken, either because of excessive model complexity (for example, in some integrated assessment models), or because of time and resource constraints. Many of the largest uncertainties in model-based estimates of impacts relate to factors not included in models, but which may assume increasing importance over the time scales of relevance for a changing climate (e.g., effects of pests, diseases and weeds on crop production; effects of nitrogen deposition on forests). What information on uncertainty is required by stakeholders? One of the questions posed at the first Workshop concerned the value of information on uncertainties for stakeholders. It was recognised in the different Breakout Groups that there are many different categories of stakeholder, each having different requirements for information on climate change and its impacts (see Section 4). However, although generalisations are difficult, a few themes surfaced in discussion: • Robustness of results. It was thought important to stress the robust aspects of assessments (including levels of confidence) as well as the uncertainties. An excessive emphasis on uncertainties might detract from important messages about likely consequences of climate change. • Operational application. Several participants argued that uncertainty information is likely to be most useful to stakeholders if posed in an operational or decision-making context. If one goal of a stakeholder is to minimise risks, then useful information might be obtained, for example, by distinguishing between the relative sizes of the ‘probable error’ and ‘probable bias’ (see the twofold classification of uncertainty described above). • Focus for scientists. The question was posed, but not resolved, as to whether scientists should channel their investigations of uncertainty towards narrowing the range of uncertainty, or rather towards providing information on uncertainties in a more effective way. Given that uncertainty will remain inherent in all descriptions of future climate, efforts in the latter area will always be important. What research is required to improve the treatment of uncertainty? Uncertainties in assessing vulnerability Research into climate change impacts and adaptation is undergoing a shift away from scenario-driven, “dose-response” type impact assessments towards more integrated studies that address system vulnerability and adaptability to climatic perturbations. This shift, which is likely to accelerate, has been largely policy driven, recognising the need to evaluate levels of climate change that threaten key social, economic and ecological systems. It requires a better understanding of the uncertainties involved in assessing vulnerability (i.e., the performance of impact models and the quality of input and validation data) to lay alongside the uncertainties in projecting alternative storylines of the driving factors of global change and uncertainties in future atmospheric composition, climate and sea-level. Some specific research recommendations from the Workshop included:
12
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
• The testing and intercomparison of impact models should be improved and their suitability for application under conditions of changed climate evaluated more critically than hitherto. If necessary, models should be excluded from consideration in impact assessments if their performance against past observations is inadequate or their range of operation is limited to present-day conditions. Model ‘invalidation’ may thus be a more relevant goal of model testing than model ‘validation’, and could serve to narrow the uncertainty range of estimates by excluding inappropriate models. • The development of simple empirical relationships or models that demonstrate similar behaviour to more complex process-based models, may provide an opportunity for undertaking detailed uncertainty analysis that is prohibited in complex models by insufficient computer power and resources. • More studies are required to examine how the statistical properties of observed and modelled responses change as values are scaled up or down. • Continued improvements to GCMs should be encouraged. As well as providing more confidence in estimates of future climate, improvements to land parametrisation schemes and to ocean dynamics offer the promise for obtaining new information from GCMs that has hitherto been derived in impact models. Examples include GCM estimates of the surface water balance, including evapotranspiration, and regional estimates of sea-level rise due to thermal expansion. Uncertainties in assessing scenarios of future change In some sectors for which the possible impacts of future climate change have been evaluated (e.g., water resources), the uncertainties in estimates of systemic response to climate change may be narrower than the uncertainties attributable to alternative views of the future (e.g., estimates of future streamflow under different climatic scenarios). In these cases, the choice of scenarios has a critical bearing on the predicted outcomes. A number of key research tasks were identified for enhancing the quality and usefulness of scenarios: • The driving factors underlying new impact assessments conducted during the first few years of the 21st century should be based on the SRES emissions scenarios or related exercises. These should be used to force climate model runs to obtain scenarios of future regional climate, and downscaled to regions, as appropriate, for characterising future climates consistent with these non-climatic assumptions of the future. • There should be continued intensive research into abrupt and/or non-linear events (e.g., a re-orientation of the thermohaline circulation of the ocean or rapid deglaciation of the West Antarctic) which need to be accounted for in any uncertainty analysis. • The downscaling of climate information from GCMs is a key requirement for some impact applications, especially those requiring information on weather-related events at high temporal resolution (e.g., high intensity, short duration rainfall events; high winds, storm surges). It should not be assumed, however, that downscaling large-scale climate change information is always a necessary or cost-effective strategy for scenario construction. In addition to intensified research to refine downscaling methods, more impact studies are needed that assess the relative value of different downscaling approaches. • More research is needed into new methods, including expert judgement and statistically-based approaches, for obtaining probability density functions of future outcomes where these are unknown. Improved information on the probability distribution of future climate will require model simulations for a wider range of emissions scenarios and for a greater number of ensembles. One option for obtaining the latter at relatively low cost, is to use time-slice experiments with global climate models. 13
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
• Decadal-scale natural climatic variability needs to be better-quantified, using palaeoclimatic reconstructions, observations and models, so that it can be accounted for in climate scenarios and impacts assessments. What guidance is required for scientists in addressing uncertainty? One of the clear messages to emerge from this first ECLAT-2 Workshop was that a proper treatment of uncertainty is merited at all stages of a climate change impact assessment. This applies to the development of climate change and non-climatic scenarios, the evaluation of impact models or alternative experimental methods, the quality control and application of measured data, the analysis of model-based or experimental results, and the presentation and communication of results to diverse audiences. On the other hand, however, there was also substantial agreement that the current practice of impacts assessment falls disappointingly short of this desirable goal. Numerous examples were presented throughout the first Workshop to illustrate important uncertainties needing to be accounted for at different stages of an impacts assessment. A number of standard techniques were also described for addressing these uncertainties. The fact that climate change impact studies seldom apply such procedures suggests that proper guidelines are urgently required by the research community. It also draws a shroud over much of the published research on climate change impacts which may, at best, be understating the uncertainties, and at worst, providing blatantly misleading information. This is an uncomfortable situation for informing the decision process. Some of the features of assessments that were identified as requiring guidance included: • Goals of an assessment. In defining the aims of an assessment, stakeholder involvement was regarded as critical. Dialogue is important to ensure that any evaluation of uncertainties undertaken by scientists is both relevant and expressed in appropriate terms for the stakeholder. • Selection of scenarios. It would be useful to have improved guidance (e.g., a ‘performance index’) on the trustworthiness of different climate model estimates of future climate for use in impacts assessment. There should be better descriptive and interpretative information on scenarios and related inputs to impact studies. • Selection of impact models. Guidance is required on how to enhance the documentation and transparency of impact models. There is also a need for improved access to such documentation, and to results of model intercomparison activities and information on key model parameters. A clearing house for different model types would be useful to assist researchers in selecting appropriate models for their impacts work. • Evaluation of uncertainties. One of the major gaps in the methodology of climate impact assessment, is a comprehensive guide to tools and procedures for estimating uncertainties at each stage of an impact assessment. The Discussion Paper by Katz in the first ECLAT-2 “Red” report offers a taxonomy of approaches to uncertainty estimation. A next step would be to identify appropriate tools for different types of climate impact assessment. • Presentation of uncertainties. There are many alternative methods of representing uncertainty, but no agreed guidelines on “best practice”. There was a general feeling in the Workshop that reports of impacts assessments should make clear which uncertainties are considered and which are suppressed. Studies should also pay more careful attention to presentation - both for stakeholders and for other scientists. Moreover, impact analysts and policy makers should be prepared to work with scenarios that comprise probability distributions. This is an emerging research area, and guidance will be required in both the application of risk assessment techniques and the interpretation of such information. 14
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Creating Climate Scenarios for the European ACACIA Assessment 1
Mike Hulme and Timothy R. Carter
2-2
1. Climatic Research Unit, School of Environmental Sciences, University of East Anglia, Norwich NR4 7TJ, UK (
[email protected]) 2. Impacts Research Division, Finnish Environment Institute, Box 140, FIN-00251 Helsinki, Finland Background Climate change scenarios present coherent, systematic and internally-consistent descriptions of changing climates. These scenarios will typically be used as an input into climate change vulnerability, impact or adaptation assessments. Scenarios are used in many different ways by many different individuals or organisations. Some require only a semi-quantitative description of future climate, perhaps as part of a scoping study. Others may need explicit quantification of a range of future climates, perhaps with explicit probabilities attached, as part of a risk assessment exercise. Others still may require information for very specific geographical areas. There is also a range of time horizons that may be considered depending on the type of decision to be made. Water companies may be concerned with operating conditions over the near-term (10-20 years), while coastal engineers or forestry investment decisions may need to consider longer-term (50-80 years) horizons. No single set of scenarios can satisfy all of these needs. As a context for further discussions at this ECLAT-2 “Green” workshop on climate scenarios for agricultural, forest and ecosystem impacts, this note briefly describes the climate scenario construction approach developed for 3 the European ACACIA Concerted Action (Parry, 1999). The ACACIA project - A Concerted Action Towards a Comprehensive Climate Impacts and Adaptation Assessment for the European Union - assesses the potential impacts of climate change across Europe, and identifies a range of adaptive response to those impacts. The climate scenarios developed for ACACIA (Hulme and Carter, 1999) represent a synthesis of current knowledge about future climate change in Europe, as well as providing a common scenario framework for the ACACIA assessment.
The European ACACIA Scenarios Climate change scenarios are most commonly constructed using results from global climate model (GCM) experiments. The main modelling uncertainties in future climate change predictions stem from the contrasting behaviour of different climate models in their simulation of global and regional climate change. These differences are largely a function of the relatively coarse resolutions of the models and the different schemes employed to represent important processes in the atmosphere, biosphere and oceans. Another important uncertainty in describing future climate, however, is unrelated to the difficulties of climate modelling and stems from the unknown world future. How will global greenhouse gas emissions change in the future? Will we continue to be dominated by a carbon-intensive energy system? What environmental regulation may be introduced to control such emissions? Different answers to these questions lead to a wide range of emissions scenarios being created. Since all future climate change model experiments need to choose an emissions scenario, different choices can lead to quite different climate outcomes. For both of these reasons - unknown future emissions and differing model behaviour - it is preferable to talk about future climate change scenarios rather than future climate predictions. 2 3
At the workshop this presentation was given by Wolfgang Cramer. The European ACACIA project has no relation to the ACACIA - A Consortium for the Application of Climate Impact Assessment - run from NCAR, Boulder, USA.
15
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Much research in Europe has been undertaken developing methods for the construction of climate change scenarios and in constructing the scenarios themselves. A substantial proportion of this work has been funded through successive European Union DG-Research projects. Much of this research into climate scenario construction has concentrated on either: • creating one or more Europe-wide scenarios for a particular impact sector (e.g., agriculture in the CLAIRE, Harrison et al., 1995, and CLIVARA projects, Downing et al., 2000); • creating one or more scenarios for a specific nation, region or catchment (e.g., Johannesson et al. 1995); • exploring new methodologies for downscaling from global climate model outputs to regional and local scale (e.g., von Storch and Reichardt, 1997; Semenov and Barrow, 1997; Kilsby et al., 1998). Much less work has been done, however, on creating European-wide overviews of the range of possible climate change scenarios and how any of these scenarios relate to the natural multi-decadal variability of climate within Europe. Two rather simple examples of quite generalised European scenario overviews are provided by the work of the European Climate Support Network in their 1995 assessment (ECSN, 1995) and the European Environment Agency (1998) in its Dobris+3 Report. It is our belief that for an adequate assessment of the impacts of climate change, and consequently an adequate assessment of the policy significance of these impacts, we need to move toward a fully quantified probability assessment of future climate outcomes. Such an objective was discussed in the first ECLAT-2 workshop the “Red” workshop held in Helsinki in April 1999 (Carter et al., 1999), where it was concluded that further research is needed to achieve this goal. The ACACIA climate change scenarios represent a step in this direction. The approach followed is rather similar to that used at national level in the Finnish Research Programme on Climate Change (Carter et al. 1996) and more fully expressed and refined in the recent UKCIP98 climate change scenarios for the UK (Hulme and Jenkins, 1998). In developing the ACACIA scenarios, we were not content to define a single ('best guess') scenario of climate change for Europe, nor even to provide a number of arbitrarily selected alternative scenarios. Instead, we presented a range of future climates that attempt to account for some of the major uncertainties in climate prediction. In a further advance over previous scenario development exercises, we also placed these climate scenarios in the context of natural variations in European climate. A natural further development of this approach is to develop fully probabilistic regional climate change scenarios for seasonal-mean temperature and precipitation, as illustrated by New and Hulme (2000). We selected the period 1961-90 as the baseline period from which our climate changes are calculated. This period has been adopted by the climate scenarios Data Distribution Centre (DDC) of the Intergovernmental Panel on Climate Change (IPCC). This Centre is making consistent sets of scenarios of climate and related environmental and socio-economic factors, along with global observational climate data, available for use in climate change impact assessments (DDC, 1999). It is useful to adopt one or more future time periods for which to design the scenarios. The 2020s, 2050s and 2080s have been adopted by the IPCC DDC and are being increasingly used in related impact and adaptation studies. These dates refer to the 30-year periods centred on 2025 (2010-2039), 2055 (2040-2069) and 2085 (2070-2099) and were adopted in the ACACIA scenarios (see Box 1 for the summary highlights).
16
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
BOX 1: KEY FEATURES OF THE EUROPEAN ACACIA SCENARIOS The ACACIA climate change scenarios have the following characteristics: • Four climate change scenarios were created, ranging from slow (0.1ºC per decade) to rapid (0.4ºC per decade) rates of future global warming. This range results from combining uncertainties about future greenhouse gas emissions with uncertainties about the sensitivity of the climate system. • The climate scenarios used the four draft marker emissions scenarios prepared for the Special Report on Emissions Scenarios (SRES) and global climate model results from the IPCC Data Distribution Centre (DDC). • Europe-wide maps showed the median modelled mean seasonal changes in temperature and precipitation, alongside maps which measured the inter-model spread of the responses. The magnitude of these changes was related to model-based estimates of natural multi-decadal climate variability. These scenario uncertainties were also depicted for sub-regions of Europe. • The climate scenarios considered only the effects of elevated greenhouse gas concentrations and not the additional effects that may be induced by changes in sulphate aerosol concentrations. The main features of future climate change in Europe, as described by these scenarios, were: • Annual temperatures over Europe warm at a rate of between 0.1ºC/decade and 0.4ºC/decade. This warming of future annual climate is greatest over southern Europe (Spain, Italy, Greece) and north-east Europe (Finland, western Russia), and least along the Atlantic coastline of the continent. In the winter season, the continental interior of eastern Europe and western Russia warms more rapidly than elsewhere. • Winters currently classified as cold (occurring 1 year in 10 during 1961-1990) become much rarer by the 2020s and disappear almost entirely by the 2080s. In contrast, hot summers become much more frequent. Under the 2080s scenario, nearly every summer is hotter than the 1-in-10 hot summer as defined under the present climate. • The general pattern of future change in annual precipitation over Europe is for widespread increases in northern Europe, rather smaller decreases across southern Europe, and small or ambiguous changes in central Europe. Most of Europe gets wetter in the winter season (between +1 and +4 per cent/decade). In summer, there is a strong gradient of change between northern Europe (wetting of up to +2 per cent/decade) and southern Europe (drying of up to -5 per cent/decade). The inter-model range of seasonal precipitation changes implies that sign differences frequently exist between the precipitation changes simulated by different climate models. • Global-mean sea-level rises by the 2050s by between 13 and 68cm. These estimates make no allowance for natural vertical land movements. Owing to tectonic adjustments following the last glaciation, there are regional differences across Europe in the natural rates of relative sea-level change. • The ACACIA scenarios did not explicitly quantify changes in daily weather extremes. However, it is very likely that frequencies and intensities of summer heatwaves will increase throughout Europe, likely that intense precipitation events will increase in frequency, especially in winter, and that summer drought risk will increase in central and southern Europe, and possible that gale frequencies will increase.
17
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Further Research Needed While we have a good appreciation of the historical changes in mean climate across Europe, and in some large-scale circulation features such as the North Atlantic Oscillation (e.g., the European Climate Support Network assessments), there is less comprehensive information available for Europe concerning long-term changes in extreme event frequencies such as heatwaves, intense precipitation, gales, etc. A number of studies have been conducted, but mostly at a national level. It is therefore not yet possible to build up a continental-scale picture of such changes in extreme events because these national studies use different event definitions and different methodologies for studying them. Fundamental to the improvement of future climate prediction is the continuing development and improvement of global climate models. This objective is being met through a large number of national and international programme initiatives. Of particular importance is to improve the treatment of biospheric feedbacks in climate models, for example those relating to land cover change and changes in net primary productivity (NPP). Providing detailed climate change scenarios with high spatial resolution for use in climate impacts research has long been a fundamental rationale for statistical and dynamical downscaling. Outputs of climate change experiments using GCMs are often viewed as an inadequate basis for assessing land-surface impacts at regional scales because their spatial resolution is too coarse to resolve important sub-grid scale processes, and because GCM output is unreliable at individual gridbox scales. While there is also a growing appreciation of the limitations inherent to statistical downscaling, relatively little is known about the 'valueadded' aspect of using downscaled as opposed to raw GCM scenarios for impact studies. Studies are needed that: • provide critiques and/or intercomparisons of different statistical downscaling methods (including comparisons with nested regional modelling experiments); • validate typical downscaling predictor-predictand relationships within GCM output; • compare climate change impacts which result from downscaled climate change scenarios with impacts resulting from climate change scenarios taken directly from GCM output. There is a need to analyse and interpret more systematically a wider range of outputs from climate models - both GCMs and RCMs - that relate to extreme climatic events. Not only do differences between global and regional climate model outputs on these shorter time-scales need evaluating and interpreting, but investigations need to be conducted into how 'extreme events' in climate models - operating at a resolution of one to many thousands of km2 - relate to observed extreme climatic events. More research should be conducted into the development of risk assessment frameworks for the development of climate change scenarios (Carter et al., 1999; Jones, 2000; New and Hulme, 2000). The objective here would be to use Bayesian logic and Monte Carlo simulation approaches to enable a more consistent expression of climate change scenario uncertainties in quantitative terms. Such approaches should encompass climate variability on daily to decadal time-scales.
Acknowledgements The European ACACIA Concerted Action is funded by DGXII of the European Commission (Contract No. ENV4-CT97-0531).
18
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
References Carter, T.R., Posch, M. and Tuomenvirta, H. (1996) The SILMU scenarios: specifying Finland’s future climate for use in impact assessment, Geophysica 32, 235-260. Carter, T., Hulme, M. and Viner, D. (eds.) (1999) Representing uncertainty in climate change scenarios and impact studies, ECLAT-2 Report No.1, Helsinki Workshop Proceedings, Climatic Research Unit, Norwich, UK, 128pp. Downing, T.E., Harrison, P.A., Butterfield, R.E. and Lonsdale, K.G. (eds.). (2000) Climate Change, Climatic Variability and Agriculture in Europe: An Integrated Assessment, Research Report No. 21, Environmental Change Unit, University of Oxford, Oxford, UK (in press). ECSN, (1995) Climate of Europe: Recent Variation, Present State and Future Prospects, European Climate Support Network, KNMI, de Bilt, Netherlands, 72pp. European Environment Agency, (1998) Europe's environment: the second assessment European Environment Agency/Elsevier, Copenhagen, Denmark, 293 pp. Harrison, P.A., Butterfield, R.E. and Downing, T.E. (eds.) (1995) Climate Change and Agriculture in Europe: Assessments of Impacts and Adaptations. Research Report No. 9, Environmental Change Unit, University of Oxford, UK, 414 pp. Hulme, M. and Jenkins, G.J. (1998) Climate change scenarios for the United Kingdom. UKCIP Technical Note No.1, United Kingdom Climate Impacts Programme, Climatic Research Unit, Norwich, UK, 80 pp. Hulme, M. and Carter, T.R. (1999) The changing climate of Europe. In, An Assessment of the potential effects of climate change in Europe Parry, M.L. (ed.). The draft report of the ACACIA Concerted Action, October, 1999, UEA, Norwich, pp 51-90. Johannesson, T., Jonsson, T., Källen, E. and Kaas, E. (1995) Climate change scenarios for the Nordic countries, Climate Research 5, 181-195. Jones, R.N. (2000) Analysing the risk of climate change using an irrigation demand model, Climate Research 14, 89-100. Kilsby, C.G., Cowpertwait, P.S.P., O'Connell, P.E. and Jones, P.D. (1998) Predicting rainfall statistics in England and Wales using atmospheric circulation variables, International Journal of Climatology 18, 523-540. New, M.G. and Hulme, M. (2000) Representing uncertainties in climate change scenarios: a Monte Carlo approach, Integrated Assessment (in press) Parry, M.L. (ed.) (1999) Assessment of the potential effects of climate change in Europe, The draft report of the ACACIA Concerted Action, October, 1999, UEA, Norwich , 350pp. Semenov, M.A. and Barrow, E.M. (1997) Use of a stochastic weather generator in the development of climate change scenarios, Climatic Change 35, 397-414. von Storch, H. and Reichardt, H. (1997) A scenario of storm surge statistics for the German Bight at the expected time of doubled atmospheric carbon dioxide concentration, Journal of Climate 10, 2653-2662.
19
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Keynote Papers page numbers 1. Agriculture and Climate Change Scenarios Mark D.A. Rounsevell
21
2. Assessing the Impact of Climate Change on Forests and Forestry using Climate Scenarios- Current Practice and Future Directions David T. Price and Mike D. Flannigan
32
3. Development of Climate Change Scenarios for Agricultural Applications Mikhail A. Semenov and Elaine M. Barrow 4. Incorporating Dynamic Vegetation Cover within Global Climate Models Jonathan A. Foley, Samuel Levis, Marcos Heil Costa, Wolfgang Cramer and David Pollard
4
4
50
59
During the workshop, Navin Ramankutty (Univ. of Wisconsin) gave an oral presentation with related material under the title "Integrated modelling of the Earth system: bi-directional atmosphere-biosphere interactions" - we here reproduce, with permission by the Ecological Society of America, the text of a review paper summarising the topic due to appear in Ecological Applications during 2000.
20
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Keynote Paper 1 Agriculture and Climate Change Scenarios Mark D.A. Rounsevell Department of Geography, Université Catholique de Louvain, Place Louis Pasteur 3, Louvain-La-Neuve, B-1348 Belgium (
[email protected])
INTRODUCTION Agriculture is one of the sectors that has been most widely researched by the climate change impact community worldwide. This reflects perhaps the central role that agriculture as a socio-economic activity plays within many regions, as well as the direct relationship between agriculture and the weather, that has given agricultural scientists a strong interest in climate. Many of these agricultural impact studies have been based on modelling approaches that have used climate change scenarios. The specific climate data requirements of these studies, however, have often been very different, and have not always matched what has been available from global climate models (GCMs) in terms of variables and their spatial and temporal resolution. This, inevitably, has led to the use of a wide range of different approaches to the interpretation of climate change data for agricultural impact models with the result that large inconsistencies have appeared in the application of climate change scenarios in agricultural assessments. The aim of the ECLAT2 workshop, therefore, is timely in giving us an opportunity to review what has been achieved so far in the agriculture sector, in order to benefit from this knowledge in developing improved methods of interpreting and representing future climate in impact studies. This paper draws on the experience of previous research on agriculture and climate change, as an introduction to a discussion on where we need to take future research. The intention is not to prescribe solutions to specific technical problems, but to introduce and highlight a range of issues for agriculture and climate change, as a way of initiating discussion during the ECLAT-2 workshop on the use of climate scenarios. Further presentations during the workshop will discuss specific techniques, for example, on scaling and ‘spatialisation’ issues (e.g., see the keynote paper by Semenov and Barrow in this volume). Because the ECLAT-2 workshop aims to examine the use of climate change scenarios amongst the impacts community, the main focus of this paper will be on the use of such data within agricultural assessments. Furthermore, because ECLAT-2 is funded by the European Commission, this paper will also focus primarily on Europe as a geographic region, although many of the points raised will be of direct relevance to other regions of the world. The paper also draws heavily on the work recently undertaken within the EC-funded IMPEL project that aimed to assess climate change impacts on agricultural land use within contrasting European regions (Rounsevell et al., 1998).
Why are we interested in climate change impacts on agriculture? In addition to its fundamental role of sustaining human populations through the provision of food and fibre, potential climate-induced changes in agriculture are important for a number of reasons: • Agriculture is a large user of the land surface itself (i.e., Food and Agriculture Organization- FAO, statistics suggest that, with the exception of the Nordic countries, between 50 and 70 % of the total land area of most European countries is used for agriculture).
21
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
• Agricultural land use has important implications for other sectors, e.g., water resources/quality, biodiversity and species distribution, regional economies. • Agriculture impinges on many parts of society from the farming community itself, to the consumer, and the policy-maker concerned with both agricultural and social policy (i.e., changes in rural communities). • Agricultural communities are often perceived by society as the stewards of the countryside. • Agriculture is a productive rather than consumptive land use, which implies the requirement to maintain production into the future, i.e., sustainable development. • Agricultural land use influences the climate system (through sources and sinks of greenhouse gases and surface albedo effects), so that land management changes may possibly contribute to the mitigation of climate change.
What are the specific issues for agriculture in relation to climate change impacts? Whilst agricultural impact studies share many of the methodological problems faced by other sectors, some issues are specific to agriculture. In particular: • Agriculture is a highly managed ecosystem in which one of the principal management activities is dealing with problems (and opportunities) arising from the weather. Therefore, farmers are skilled managers of risk and uncertainty (issues that are important for climate change). • Furthermore, farmers are currently and constantly required to adapt to a changing economic environment, and these responses can give us useful insights into potential adaptation strategies to climate change in the future. • Adaptation of agriculture to climate change will involve the autonomous responses of individual farmers to their changing environment. Thus, agricultural assessments need to be set within a framework that recognises the ‘choice’ for individual farmers in the use and management of agricultural resources, which implies an integrated socio-economic and biophysical approach to understanding change. • We are not only interested in estimating changes in the use (or potential) of natural resources for agriculture, but also the intensity of that use. • Agricultural assessments are undertaken over a wide range of spatial scales depending on the specific questions asked of research from leaf/crop/animal/field scales through to the farm system and regional scales and national/international scales (e.g., reforms to the Common Agricultural Policy- CAP, international trade agreements).
22
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
ISSUES FOR AGRICULTURAL ASSESSMENTS The drivers of change Climate change will influence agricultural production systems at specific geographic locations, as well as modifying the spatial distribution of agricultural land use. In trying to understand the impact of climate change on agriculture we need to recognise the fundamental role of the farmer as the central decisionmaker who will respond to and manage the consequences of climate change. Whilst this seems an obvious statement to make, in practice very little research on agriculture and climate change has been ‘farmer/decision-maker orientated’ probably because of the difficulties in developing multidisciplinary research. Decision-making in agriculture, driven by considerations of profitability, is undertaken within the context of the opportunities and constraints offered by a specific geographic location. Thus, climate change may affect either directly or indirectly: • Biophysical production constraints and opportunities (the productivity and quality of crops and animals); • Management constraints and opportunities; • Socio-economic constraints and opportunities.
The biophysical production constraints and opportunities Climate change is expected to modify crop yields and quality as well as growing season lengths through the effects of temperature, CO2 and precipitation changes. Crop yields may be affected both directly by climate change, and indirectly through changes in a) water stress (drought) or water excess, b) the incidence of pest and diseases and c) levels of land degradation (e.g., soil erosion, salinisation, peat wastage). Livestock farming will also be influenced by indirect climate-induced changes in forage productivity and quality, as well as direct changes in climate including extreme events on animal welfare. Much previous research on agriculture and climate change has focused on the modelling of crop growth and yields at different spatial scales (e.g., Mearns et al., 1999; Downing et al., 1999; Wassenaar et al., 1999 to name but a few). However, the biophysical components of the agricultural production system are still difficult to model in a satisfactory way. This is because models do not predict well the observed regional differences in agricultural production (e.g., crop yields) in response to a wide range of environmental factors, and also data that is required for model calibration and application are limited. Uncertainties associated with future climate scenarios as well as the coarse-scale resolution of such data, are amongst the types of data limitations that exist for biophysical agricultural assessments (e.g., Mearns et al., 1997).
The management constraints and opportunities Changes in the soil water balance (arising from temperature, CO2 and precipitation effects) will affect the opportunities for soil tillage at more northerly latitudes (Rounsevell and Brignall, 1994). This will have potentially important consequences for the scheduling of farm operations (e.g., tillage, spraying, etc.), which are an important influence on the areal distribution of arable crops. Changes in the distribution and incidence of pests and diseases will also determine how agricultural production is managed, as will factors such as fertilization, irrigation, erosion control, harvest conditions and snow cover effects on sowing at high latitudes. Management effects are also important in the context of local socio-economic factors including, for example, the availability of machinery and labour, and farm sizes.
23
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
The socio-economic constraints and opportunities Whilst the productivity of the physical environment, and the management constraints that confront farmers play an important role in agricultural production systems and the distribution of land use, the prevailing economic environment (prices and costs) and the socio-cultural context will ultimately determine what farmers will cultivate where. Whilst socio-economic constraints depend primarily on policy and macro-economics (i.e., international trade), climate change may have an indirect effect through changes to the supply and demand balance of agricultural commodities. The interaction between climate and socio-economic change is likely to be complex, with one perhaps either exacerbating or ameliorating the other. It is likely, however, that the comparative advantage between different land uses and/or policies of regional support or subsidy will be important in relation to shifting zones of agricultural potential under a changing climate.
Multidisciplinary scenarios The points above highlight the importance of undertaking agricultural impacts within a multidisciplinary setting. This implies that a discussion of the use of climate change scenarios should also recognise and address the need for scenarios of socio-economic change. More recent developments in climate impact assessments have begun to adopt this approach. For example, the EC-funded ACACIA project (Parry, 1999), the UK Climate Impacts Programme (DETR, 1999), the US Country Studies Programme (e.g., Government of Pakistan, 1998) and the IPCC have all begun to tackle the problem of developing integrated climate and socio-economic-change scenarios. The basic challenge is to develop plausible scenarios that cover a range of possible changes in the climate as well as the socio-economic baseline, in an internally-consistent way. At the simplest level, this may be achieved by the construction of scenarios that are linked through the assumed political and societal attitudes towards both greenhouse-gas emission and mitigation as well as the management of economies and landscapes. The principal benefit of this approach is that it allows impact scientists to test their models with data inputs describing a limited range of plausible futures. However, the quantification of such scenarios can be difficult in practice, as can the issue of defining differences in attitudes that could occur between local or global scales. Furthermore, scenarios are limited by their inability to incorporate feedbacks and/or autonomous adaptation to change, although some integrated assessment models can address this problem at coarse resolutions (e.g., Alcamo et al., 1996; Morita et al., 1994). An example of the feedback problem in agricultural impact assessments is the relationship between land use (change) and prices of agricultural products. In a free market situation, an increase in the cultivated area of a particular crop, for example arising from favourable changes in the climate or increases in prices, is likely to result in over-supply of that crop with a subsequent reduction in its price. This is likely to reduce the cultivated area. Similar arguments could be put forward for the feedback between changes in agricultural land use and management, and the climate system itself. Thus, scenarios may be limited in their application where feedback processes are important.
The spatial aspects of change The points listed above also demonstrate that in undertaking agricultural assessments, we are normally confronted with the issue of spatial scale. Biophysical impacts will occur at the scale of individual plants, animals or fields; management impacts occur at the scale of fields and farms; social influences occur at farm and regional scales; whereas economic influences occur over regions and the globe. This variability in the spatial scale of impacts needs to be considered in relation to the resolution of available climate change data.
24
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Many agricultural studies have ignored the influence of spatial variability on agricultural production, focusing instead on ‘representative’ sites or farms. This has partly reflected the level of scientific development in terms of suitable models as well as the availability of spatially-variable data (especially climate data) at the required resolution. For example, calibrating crop models for a wide range of environmental conditions is often limited by the availability of data. Increasingly, however, there is a recognition that sites are not normally ‘representative’ and so, there is a need to develop spatially-explicit methodologies. This has lead to the fundamental problem of whether to develop more finely resolved climate datasets as input to existing agricultural models (at the risk of implying a greater understanding of local climate variability than actually exists in practice), or to modify the models to accept spatially-coarser climate data whilst attempting to maintain the validity of the model in representing landscape processes. In practice, therefore, the problem of developing spatially-explicit models (that represent landscapes and regions rather than points) has led to the development of a range of alternative methodologies. These methods have included, for example, techniques such as weather generators (e.g., Wilks, 1992 ; Semenov and Barrow, 1997 and also keynote paper 3 this volume; Mearns et al., 1997; Carter et al., 1996; Downing et all., 1999), ‘regionalised’ high resolution climate models (e.g., Mearns et al., 1999), statistical methods of interpolation/extrapolation (e.g., Wassenaar, et al., 1999), or the use of reduced-form models (e.g., Brignall and Rounsevell, 1995; Harrison and Butterfield, 1996; Mayr et al., 1996). An example, of the statistical approach is presented in Figures 1 and 2 for a wheat crop in southern France (Wassenaar, et al., 1999). This work was based on the establishment of statistical relationships between a) a spatially-variable and easily mapped soil parameter- the available water capacity, and b) changes in crop yields for a number of experimental sites, where these changes are based on the modelled difference in yields between the baseline climate and a climate scenario. The statistical function derived from application of the crop model at sites within three different regions (Figure 1), was used to create a spatially-variable map of yields based on the available water capacity (Figure 2). In Figure 1, the points represent individual soils with available water capacities that are assumed not to change with climate change. Thus, the map in Figure 2 can be created from the derived statistical function, and the known current distribution of soil types and their available water capacities. Some discontinuities can be observed in Figure 2, which result from the different statistical functions derived for each region. The wide range of yields demonstrated by the map emphasise the need to consider spatial variability in agricultural assessments. A statistical approach was possible in this example because in the region of study, southern France, crop productivity is strongly limited by soil water availability. In other regions this might not be the case, and finding alternative ‘indicators’ of change might be difficult. Furthermore, application of the derived statistical function is limited to the study area, and the two stages of a) running the model and then b) deriving the statistical relationship needs to be repeated for each climate scenario considered.
Figure 1. Mean change of wheat yield (UKHI, IS92e emissions scenario, high climate sensitivity) for each soil profile as a function of the available water capacity of the soil in three regions of southern France (Pèzenas, Béziers and Aniane) (after Wassenaar et al., 1999).
25
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Further discussion of the technical aspects of scaling techniques will be introduced in keynote paper 3 by Semenov and Barrow, this volume.
wheat yield -0.7 -1.0 -1.3 -1.6 -
c hange -0.4 t/h -0.7 t/h -1.0 t/h -1.3 t/h
N
4
0
4
8 Ki lomètres
Figure 2.Wheat yield change yield (UKHI, IS92e emissions scenario, high climate sensitivity), estimated on the basis of the available water capacity of the dominant soil type within each map unit for three climatic regions: Béziers, Pézenas, Aniane (after Wassenaar et al., 1999).
The temporal aspects of agricultural assessments and climate change scenarios In comparison with questions related to the spatial scale of agricultural impact assessments, temporal scale has received relatively little attention. However, we know that the different components of the agricultural production system respond over different time scales, for example : • biophysical responses occur over the short to medium term (days to weeks/months), including responses to extreme events such as hail and wind that can occur on very short time scales; • farm operations and management strategies are implemented over the medium term (weeks/months to years); and, • policy-making and changes in farmer perceptions occur over the longer term (years to decades). In the shorter term, the effect of weather extremes can have significant, if difficult to anticipate, effects on crop and animal production. Furthermore, modelling changes in mean crop yields, for example, may be of less interest if the interannual climatic variability is great, as is the case in many Mediterranean environments. Arguably, however, the most important influence of climatic variability on agriculture are the effects on farmer decision-making. There are two points for consideration: a) the influence of risk and risk perception, and b) the influence of rates of change of decision-making. Climate variability, risk and risk perception In attempting to maximise profits from agricultural land use farmers are strongly aware of the influence of fluctuations in prices and the weather. The potential for ‘good’ profits in one year may be mitigated by the impact of a ‘bad’ year (e.g., because of drought, waterlogging or disease). This often leads farmers to selecting a land use that is less than optimal in terms of profit, but which is also less ‘risky’, i.e., it produces
26
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
a consistent return between years. The consequence is that this perception of risk can strongly influence the distribution of agricultural land use. De Baets et al. (1999), for example, showed this effect on agricultural land use in Bedfordshire, eastern England. Figure 3 shows that in comparison with the observed statistics (for 1991 and 1995) the areal extent of the cultivation of crops such as peas and beans is overestimated by a model of agricultural land use at the expense of cereals (wheat and barley), which are underestimated. Conventional wisdom within this region is that peas and beans carry a high risk because in wet years pest and disease problems can create substantial losses. Cereals are perceived as more consistent ‘performers’.
Bedfordshire 1991 and 1995 70
60
arable land cover (%)
50
40
30
20
10
0
wheat
barley
oats
potato census 91
sugar b. census 95
dr. peas model 91
rape
beans
linseed
model 95
Figure 3. Change in arable land cover in Bedfordshire per crop (ha) from 1991 to 1995 in comparison with outputs from the IMPEL model (after De Baets et al., 1999).
In climate change terms, risk perception may become increasingly important in understanding future land use change because a changing climate (and variability in the weather) may create different risk perceptions in land use managers. The challenge for agricultural assessments that use climate change scenarios is how best to represent risk perception in models, and to define for the climate modelling community, the data required to apply these models. It seems reasonable to suppose, however, that the representation of climatic variability within climate change scenarios will be an important future need of the climate impacts community. Rates of change In addition to the effects of risk perception on agricultural decision-making, we need to recognise that most farm-based decisions do not occur instantaneously, and are part of a longer-term planning strategy. Such strategies are usually developed in light of past experience. For example, the introduction of a new, profitable crop to a region is not normally met with widespread and immediate ‘take-up’ by the farming community. There is normally a time-lag before the crop is cultivated, the duration of which is also a function of local socio-cultural conditions. For example, a land use modelling study undertaken in Greece (Koutsidou et al., 1999) based on price data until 1992 suggested that cotton should be a profitable crop in the region of Thrace and, therefore, extensively cultivated. In practice, however, very little cotton was cultivated in Thrace at this time. Examination of published agricultural statistics (Figure 4), however, showed that cotton production increased substantially after the period of the model experiment. It appears,
27
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
therefore, that the change in land use had been anticipated by the model before the effect was realised in practice. Such time-lags have been observed in other regions of Europe and potentially can occur over periods of several years to decades Cotton Thrace 400000 350000
Cultivated area Ha
300000 250000 200000 150000 100000 50000 0
1978
1979
1980
1981
1982 1983 1984 1985 1986 1987 1988 1989 Subsidies to farmers: 2/3 of the total price /kgr ----- 1092 ECU/Ha
1990
1991
1992
1993
1994
Figure 4.The progress of cotton cultivation in the Thrace area of Greece (data provided by E. Koutsidou, University of the Aegean, Mytilene)
In the context of climate change impact studies, the rate of ‘take-up’ of crops or agricultural production systems that are new to a region could be extremely important. Models that assess what could be grown in a region based on current climate change scenarios (the potential land use), may not be recognising that farmers are unlikely to anticipate climate change and will only respond after a change in the climate is perceptible and this becomes part of their ‘experience’. Thus, understanding the rate at which the climate will change will be very important in estimating the rates of change in agricultural land use. Once again, the challenge for agricultural assessments that use climate change scenarios is how best to represent such processes in impacts models, and to define for the climate modelling community, the data requirements for such models.
DISCUSSION What should we be trying to do with agricultural assessments in the future? Clearly in improving understanding of climate change impacts on agriculture (as for other sectors) responsibilities lie with both the impacts and climate science communities. An important part of this development is a continuous dialogue between the two communities to encourage understanding of the key issues. Initiatives such as ECLAT-2, therefore, have an important role to play in encouraging this dialogue.
28
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Possible practical developments could include : • The construction of a common set of universally applied, integrated scenarios (e.g., IPCC, ACACIA). This would facilitate comparison across regions, and sectors (e.g., the ecosystem modellers could access data describing plausible agricultural land use changes that are referenced to a particular climate/socioeconomic storyline). This implies, however, not only the development of appropriate (internallyconsistent) storylines of plausible futures, but also the development of spatially-explicit datasets that represent these future. Such databases need to be put together within the framework of their use within impacts studies and models, so that the requirements of these studies are addressed directly. • Initiatives that seek to develop, and make available, spatially and temporally scaled climate/weather data (both baseline and climate scenario) for a wide-range of impact studies. Such work should consider the needs of the impact community, but within the context of what is realistic in terms of the representation of the climate. At what level should we generalise and simplify? By deriving increasingly finer resolution data (in time and space) or by modifying the impacts models to accept coarser resolution data? There is a need for the agricultural research community to focus on what is important in assessing the response of agriculture to climate change (e.g., weather-related management constraints within a farmer decision-making framework) in order to be able to assess autonomous adaptation. The development of models of these processes needs also to recognise the limitations of climate data resolution, but at the same time impact scientists should work with the climate community in improving the climate data for their purposes. The climate science community could assist the impacts community through the development of methods to bridge the gap between the results of GCM experiments and the requirement for climate data in impact studies. The onus is also on the impacts community to be open about the capability of impacts methodologies. Whilst impact studies must generally face the problem of the quality and availability of climate data, it is often the ‘quality’ of the impacts models and methodologies themselves, which limit our ability to understand the potential impact of climate change. One of the great values of the climate change debate, in terms of scientific development, has been the realisation that we are not fully equipped to answer questions about the interaction between human activities, such as agriculture, and climate. We need to highlight the great uncertainty that exists at all levels of impact assessment, from the use of climate scenarios through to the representation of socio-economic sectors in models.
CONCLUSIONS • Temporal scale. It is likely that the use of time-series of climate change data would lead to substantial progress in understanding climate change impacts on agriculture. This suggests that the climate impacts community needs to develop their methodologies to better take account of temporal effects, and to articulate their needs for climate data to the climate modelling community. • Integrated scenarios. Whilst recognising the potential limitations of a scenario approach, the use of integrated climate and socio-economic scenarios would provide both internally-consistent views of possible futures, and a reference point for the comparison of different impact assessments. • Spatial scale. There is a need to bridge the data gap between impact model requirements and climate data availability based either on the modification of impact methods and models or the generation of spatially more finely resolved climate data. However, as well as encouraging improvements in the quality and availability of climate change data, the climate impacts community needs to be more open about the limitations of their own models and methodologies for use in climate impact assessments. 29
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
ACKNOWLEDGEMENTS Some of the examples used in this paper were based on the IMPEL project (Integrated Model to Predict European Land use), which was funded by DGXII of the European Commission under the Environment and Climate Programme of Framework IV (contract nos. ENV4-CT95-0114 and IC20-CT96-0013). In particular, I would like to thank the IMPEL participants Eric Audsley, Anton de Baets, Eugenia Koutsidou, Tom Wassenaar and Philippe Lagacherie for the examples I have used in this paper. I would also like to thank Jørgen Olesen and Marco Bindi for contributions through their involvement in the European ACACIA project.
REFERENCES Alcamo, J., Kreileman, G.J.J., Bollen, J.C., van den Born, G.J., Gerlagh, R., Krol, M.S., Toet, A.M.C. and de Vries, H.J.M. (1996) Baseline scenarios of global environmental change, Global Environmental Change 6, 261-303. Brignall, A.P. and Rounsevell, M.D.A. (1995) Land evaluation modelling to assess the effects of climate change on winter wheat potential in England and Wales, Journal of Agricultural Science 124, 159-172. Carter, T.R., Saarikko, R.A. and Niemi, K.J. (1996) Assessing the risks and uncertainties of regional crop potential under a changing climate, Agricultural and Food Science in Finland 5, 329-350. De Baets, A., Audsley, E., Mayr, T.R. and Rounsevell, M.D.A. (1999) The effects of climate and socio-economic change on agricultural land use: a case study in eastern England. In Spatial Modelling of the Response and Adaptation of Soils and Land Use Systems to Climate Change – An Integrated Model to Predict European Land Use (IMPEL), Rounsevell, M.D.A. (ed.) Final report to the European Commission, contract nos. ENV4-CT95-0114 and IC20-CT96-0013, pp. 103-119 (unpublished). DETR (1999) Socio-Economic Scenarios for Climate Impact Assessment. Report produced for the UK DETR. Department of Environment, Transport and the Regions, London, UK. Downing, T.E., Harrison, P.A., Butterfield R.E. and Lonsdale, K.G. (eds.) (1999) Climate Change, Climatic Variability and Agriculture in Europe: An Integrated Assessment. Research Report No. 21, Environmental Change Unit, University of Oxford, Oxford, UK (in press). Government of Pakistan (1998) Study on Climate Impact and Adaptation Strategies for Pakistan. Ministry of Environment, Local Government and Rural Development, Islamabad, Pakistan. Harrison, P.A. and Butterfield, R.E. (1996) Effects of climate change on Europe-wide winter wheat and sunflower productivity, Climate Research 7, 225-241 Koutsidou, E., Margaris, N.S. and Loumou, A. (1999) The effects of climate and socio-economic change on agricultural land use: a case study in Thrace, Greece. In Spatial Modelling of the Response and Adaptation of Soils and Land Use Systems to Climate Change – An Integrated Model to Predict European Land Use (IMPEL), Rounsevell, M.D.A. (ed.) Final report to the European Commission, contract nos. ENV4-CT95-0114 and IC20-CT96-0013, pp. 185-194 (unpublished). Mayr, T.R., Rounsevell, M.D.A., Loveland, P.J. and Simota, C. (1996) Agroclimatic Change and European Soil Suitability: regional modelling at monthly time-steps, International Agrophysics 10, 155-170. Mearns, L.O., Rosenzweig, C. and Goldberg, R. (1997) Mean and variance change in climate scenarios: methods, agricultural applications, and measures of uncertainty, Climatic Change 35, 367-396. Mearns, L.O., Mavromatis, T., Tsvetsinskaya, E., Hays, C. and Easterling, W. (1999) Comparative responses of EPIC and CERES crop models to high and low resolution climate change scenarios, Journal of Geophysical Research 104(D6), 6623-6646. Morita, Y., Matsuoka, Y. Kainuma, M. Harasawa, H. and Kai, H. (1994) AIM - Asian Pacific integrated model for evaluating policy options to reduce GHG emissions and global warming impacts. In Global Warming Issues in Asia Bhattacharya, S. et al. (eds.), AIT, Bangkok, pp. 254-273.
30
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Parry, M.L. (ed.) (1999) ACACIA: A Concerted Action Towards A Comprehensive Climate Impacts and Adaptations Assessment for the European Union. Draft Report to DG XII of the European Commission, Jackson Environment Institute, Norwich, UK. Rounsevell, M.D.A. and Brignall, A.P. (1994) The potential effects of climate change on autumn soil tillage opportunities in England and Wales, Soil and Tillage Research 32, 275-289. Rounsevell, M.D.A., Brignall, A.P. and Siddons, P.A. (1996) Potential climate change effects on the distribution of agricultural grassland in England and Wales, Soil Use and Management 12, 44-51. Rounsevell, M.D.A., Evans, S.P. and Bullock, P. (1999) Climate change and agricultural soils: impacts and adaptation, Climatic Change 43, 683-709. Rounsevell, M.D.A., Evans, S.P., Mayr, T.R. and Audsley, E. (1998) Integrating biophysical and socio-economic models for land use studies. In Proceedings of the ITC – ISSS Conference on Geo-informationfor Sustainable Land Management, Enschede, The Netherlands, 17-21 August 1997, pp. 368. Semenov, M.A. and Barrow, E.M. (1997) Use of a stochastic weather generator in the development of climate change scenarios, Climatic Change 35, 397-414 Wassenaar, T. Lagacherie, P. Legros, J-P. and Rounsevell, M.D.A. (1999) Modelling wheat yield responses to soil and climate variability at the regional scale, Climate Research 11, 209-220. Wilks, D.S. (1992) Adapting stochastic weather generation algorithms for climate change studies, Climatic Change 22, 67-84.
31
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Keynote Paper 2 Modelling Impacts of Climate Change on Forests and Forestry using Climate Scenarios David T Price and Mike D Flannigan Canadian Forest Service, 5320-122 Street, Edmonton, Alberta, T6H 3S5, Canada (
[email protected])
ABSTRACT This paper focuses on the use of climate scenarios for simulating possible impacts of change on forest ecosystems in Europe. Forests and other extensively managed ecosystems are relatively complex, with many dynamic and interacting relationships among natural and human influences. Recognising that there are many difficulties in capturing all these complexities, ecosystem researchers have developed a wide range of forest process models to assess the effects of environmental change. The main classes of these models are reviewed, highlighting their strengths, limitations and data needs. Experienced impacts modellers generally recognize that there are fundamental limitations in climate data and global climate model (GCM) simulations, although undoubtedly as a group we could be better informed. On the other hand, there are many serious problems in the ecological models that need to be addressed. Hence we argue that imperfect climate data, poor agreement between GCMs and observations, and inadequate downscaling of output, will not prevent us from making progress in understanding responses to plausible changes in climate. Some of the possible approaches to building climate scenarios are reviewed. Spatial data sets are in demand though not essential for all modelling applications. In general, simple methods of downscaling are preferred, if only because no method is perfect and none seem to be clearly superior. Climatic variability is an important concern, both spatially and in the expectation of uncertain changes in the future. Two key areas are identified: frequency and intensity of Atlantic storms, and the occurrence of hot dry spells causing severe droughts and increased risk of forest fires. Limitations in downscaling approaches notwithstanding, higher resolution data sets provide a number of advantages, not least in mountainous terrain where orography has important effects on the climate of natural ecosystems. Small-scale variations in soil, hydrology and elevation are factors that interact strongly with climatic variables, greatly increasing the range of conditions occurring within a GCM grid cell. Hence even a poor representation of the true spatial variability in climate is still likely to increase the sampling of environmental variability within realistic ranges.
INTRODUCTION The earliest global climate models (GCMs) date back to the early 1970s. Thirty years later, few would doubt that the representations of the physical processes that drive these models and, hence, our understanding of the global climate system, have improved enormously. Similar models are used routinely for weather forecasting, and their short-term predictive ability can be tested and refined using reanalysis of past weather data. Predictive skill has increased significantly in the last few years, both for short-term (2-15 day) and for seasonal forecasts (e.g., see Kerr, 1998). But long-term climate statistics generated from extended 32
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
runs of GCMs remain problematic. Even the most sophisticated fully coupled transient-mode GCMs exhibit significant regional and seasonal biases when compared to reality (e.g., see Doherty and Mearns, 1999). In spite of these limitations, however, there is great demand from the global community of researchers investigating climate change impacts to use GCM output to create scenarios of near-future changes in global climate. Amongst impacts-researchers, ecological modellers attempt to simulate the possible effects of environmental change on natural ecosystems around the world. Most recognise that the climate scenario data they are using are inherently ‘inaccurate’ and must affect the credibility of their own simulations. (The report of the first ECLAT-2 workshop on uncertainty (Carter et al.,1999) provides an excellent introduction to these issues.) Moreover, the impacts models are simplifications of reality, and generally developed and tested using climatic and ecological data obtained for relatively short periods at a few carefully selected study sites. It is highly unlikely, therefore, that these models will capture the full range of ecological responses to a changing climate. Hence, the usefulness of ecosystem simulations for impact assessment is limited both by the uncertainty of the information captured in the models and by the availability of accurate climate data (for the past as well as the future). The purpose of this paper is to provide a general review of how climate scenarios are used for investigating possible effects of climate change on natural ecosystems (primarily forests), and how the results may be interpreted. Most of the discussion will focus on GCM simulations as the source of climate scenarios. Further, it will be assumed that all impact studies are based on computer simulations driven by climate scenario data. Aside from simple interpretations of the effects of average changes in temperature and precipitation derived from GCM simulations, it is hard to conceive of a meaningful impacts study that does not make some use of computer models. In general, such studies are carried out with one or more of these key objectives: • to assess absolute and relative ecosystem sensitivities to environmental change (including changes in the frequency and intensity of extreme weather events) • for those ecosystems which appear vulnerable in some way, to assess the magnitude, rate of change and nature of the impacts (positive or negative)—possibly under a range of plausible scenarios, and • for those ecosystems where the impacts are considered serious, to develop and test the effects of alternative management and policy responses.
FORESTS AND CLIMATE CHANGE Across Europe, forests may be more or less sensitive to a changing climate than agricultural systems (see also Rounsevell, 2000, this volume). On the one hand, established forests are relatively complex ecosystems (compared to agriculture), which probably gives them greater resilience to small changes in their environment. Hence forests are generally able to withstand some levels of climatic change that could seriously affect the productivity (and profitability) of seasonal crops. For example, the cumulative effects of more frequent and more severe droughts might result in farmers changing from their traditional crops simply to avoid bankruptcy. Over the same period the only observable effects on a forest might be small changes in growth and perhaps slightly increased tree mortality. On the other hand, most ecosystems can only be perturbed so far: if climate-related forest mortality were to become a serious problem, adapting present-day forest management practices to deal with it would likely be a slow and expensive process. Virtually all of the world’s temperate forests destined to reach maturity within the next 50 years are already established, and they will be exposed to whatever changed environmental conditions occur during that time. It is not feasible to change forest species over a single growing season, and in most cases we simply do not know the best new species or management methods to select for an uncertain future climate. 33
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
In some regions, unmanaged (or ‘extensively managed’) forests are subject to large-scale fires and other natural disturbances, although this is not typically the case for most of central Europe where centuries of human activity have resulted in a highly fragmented, intensively managed forest landscape. Fragmentation creates effective fire-breaks and reduces the tendency for insect pests to migrate to new stands before foresters have an opportunity to control them! Less fragmented forests are more typical for northern Scandinavia and Finland where fires can pose serious problems and would be expected to worsen with warmer, drier conditions. In the Mediterranean region, forest fires are more frequent and can cause serious damage, in some cases over extensive areas. Extreme events are also an important concern for those parts of western Europe exposed to the full force of Atlantic storms. These often cause extensive damage to forests, particularly in the more northerly regions where wet soils make plantations susceptible to ’windthrow‘. Climate also affects the vulnerability of the ecosystem to disturbance, e.g., warm winds in early spring dry out damp dead grass to create a fire hazard, and warmer summers are likely to accelerate phenological development of insect herbivores, and hence increase the risk of major tree defoliation. Following disturbance (including harvesting and salvage of timber after a storm or insect attack), the young forest (established either naturally or artificially) is likely to be more sensitive to the new climate and climatic extremes than the mature stand it replaced. If followed by a persistent change in climate, disturbance is the key mechanism causing changes in species assemblages and hence in the vertical structure and spatial distribution of forests. The interacting effects of climate on forest ecosystems, and the economic infrastructure they support, are illustrated in Figure 1. The main vegetation feedbacks to climate operate through changes in leaf area and canopy structure, which affect transfer resistances and optical characteristics—hence changing surface radiation and energy budgets, and contributing to regional weather. Climate-related changes in the activity of soil organisms and vegetation can affect rates of release and uptake of CO2 and other greenhouse gases. Disturbances can trigger immediate releases of gases and particulates, as well as longer-term changes in fluxes. It is important to note also that anthropogenic increases in concentrations of atmospheric CO2 and other air-borne pollutants could have significant effects on forest ecosystems (of which some may even be considered ‘beneficial’), but we will not address these further. The connections among these system components are far from completely understood, but some are clearly important. Hence one of the key issues in the context of climate change is how these dynamic interactions may shift in response to a systematic change in climate forcing. Added to these are a number of human influences that result from our desire to manipulate forests for specific objectives (timber, wildlife habitat, recreation). By accident or by design, forest management and the policy decisions that drive it, also contribute to ecosystem feedbacks that may affect the global climate. Studies are needed both to assess the interacting effects of present-day management practices and environmental change on forests, and as a precursor to planning changes in management to mitigate these effects.
34
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Climate Plant Processes Photosynthesis Respiration Water balance Growth Competition
Soil Factors Temperature
Disturbances Type Type
Water content
Frequency Frequency
Aeration
Timing Timing
Intensity Intensity
Extent Extent
Feedbacks Stand Structure
Rainfall Temperature Temperature
Spp. composition
Radiation Balance Radiation balance
Vertical structure
CO CO
Forest management
Spatial distribution
2
exchange exchange
2
Silviculture Harvesting Protection Biodiversity
Socio-Economics Dependant communities Dependent communities
forest Industries Forest industries
Other forest values... Other forest values....
Policy implications Policy implications
Figure 1. Diagram showing effects of climate on forest ecosystems, and possible feedbacks in response to change.These effects could be mediated or exacerbated by human management of forests, which in turn is driven by social and economic needs that may also be affected by changes in the ecosystem.
CLASSES OF FOREST MODELS USED IN CLIMATIC IMPACT ASSESSMENT It is generally believed that for ecosystem models to be able to predict correct responses to changes in climate, they must be process-based. Process models compare to empirical models, such as those used to construct forest yield tables, where measurements of stand structure (tree height, diameter, stem density, etc.) are related to site quality, e.g., characterized on the basis of the ground flora and soils. When properly calibrated, traditional yield models can be quite accurate predictors of growth over small regions, and they have the benefit of being relatively simple to construct. On the other hand, because climate affects several important site attributes (including temperature and moisture availability), a shift in climate could easily invalidate the established relationships and therefore make these models unreliable. In comparison, process models are relatively complex, often requiring numerous parameter values and large amounts of input data. These models attempt to explain plant and ecosystem behaviour in terms of quantitative responses to environmental variables, inferred from experiments in the laboratory and in the field. In theory, process models have the advantage that once developed and tested at a few sites, they can be used to estimate responses at others using only observed environmental data. In practice, such models are never completely successful, but, when extrapolated outside their region of calibration, good process models will generally perform better than the best empirical models. For the same reason, we assume that those ecosystem models that consider the important climate-sensitive processes, will be better predictors of responses to simulated changes in climate. The application of climate scenarios to assess impacts on forests and other ecosystems is largely predicated on this assumption.
35
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Gap dynamics models Perhaps the earliest attempt at simulating entire forest ecosystems, gap dynamics models originated with JABOWA (Botkin et al., 1972), and thirty years later are still the subject of much interest and research. Gap (or patch) models simulate successional dynamics of forests over relatively long periods of time, and though based on some questionable assumptions, can be reasonably accurate predictors. Traditional gap models use simple representations of tree growth, driven by monthly (in some cases daily) climate data. Tree regeneration and survivorship are simulated as stochastic rather than deterministic processes, which requires that simulations be repeated over a large number of small (~0.1 ha) patches within the simulated forest. Each tree species is represented by a set of parameters, often derived from local observations, which allow them to ‘compete’ for site occupancy, based on the limitations imposed by local climate, soils and shading from surrounding trees. Hence vegetation composition is defined by the simulated presence of particular tree species. These models are prone to fail when applied outside the region in which they were developed, although with time and effort they have been applied successfully over subcontinental regions. Sykes et al. (1996) used the FORSKA2 model to simulate distributions of natural forests across western Europe, and then applied GCM-derived scenarios to investigate possible effects of climate warming on these simulated forests. Lindner and co-workers have also adapted this model to examine interactions of forest management and climate change on forests in Germany (Lindner, 2000). The SIMA model of Kellomäki and co-workers tracks soil organic matter accumulation and decomposition in simulations of Finnish forest responses to climate change and various management scenarios (e.g., see Karjalainen, 1996). Many gap dynamic models also include simple representations of disturbances as a factor influencing succession, although attempts to relate the incidence of disturbance to climate are relatively rare.
Ecophysiological process models A large number of ecosystem models do not simulate successional dynamics, but focus more on the physiological responses of one or two species in a stand, as they interact with the physical (and in some cases, the chemical) environment in soils and the atmosphere. Typically driven by daily, or even hourly, weather data, such models can be highly detailed and often require considerable effort to be applied to new species, or even to stands of the same species at different stages of development, e.g., Grant’s ecosys model (Grant and Nalder, 2000). On the other hand, they can often be transplanted over long distances and work very successfully with only a few changes to stand, soil and climate parameters. The major strength of these models is that they can be expected to respond realistically to simulated changes in climate (within certain limits). The major weaknesses are that they take considerable time and effort to develop, usually involve both extensive model development and calibration, and are limited to one or two species at particular developmental stages, while being computationally expensive. There are several distinct subgroupings of these models, which we will attempt to categorize briefly. “Green sponge” models are physiologically less detailed but take more account of changes in tree size and allocation of biomass among roots, stems and foliage, as the stand develops over periods of decades to centuries (e.g., FOREST-BGC, Running et al., 1988). With these models, successional dynamics are not considered, but in some, changes in stand structure as a function of age and site conditions can be simulated relatively successfully. A second closely related group are the large scale biogeochemistry (BGC) models such as TEM (McGuire et al., 1993) which are used to simulate regional and global responses to climate, but include detailed simulations of soil and plant interactions including nutrient cycling. A third group are the SWAT (“soil-water-atmosphere-transfer”) models used in a) hydrological modelling and b) at coarser resolution as the land surface parameterisation schemes used to simulate terrestrial surface exchanges of radiation and sensible and latent heat in GCMs (e.g., SiB of Sellers et al., 1986).
36
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Equilibrium projection models These models, such as BIOME3 (Prentice et al., 1992), share some of the characteristics of gap dynamics models, but are more generalised and applied at much larger scales (continental to global on coarse resolution grids). In these models vegetation is defined by assemblages of a very few distinct plant functional types (PFTs), rather than by individual species. Empirical or phenomenological relationships are established between the present-day geographic distributions of vegetation and interpolated climate (primarily monthly mean temperature and precipitation). It is thus implicitly assumed that at the large scale, existing natural vegetation is fully adapted to present-day average climate. The relationships are then used to map the corresponding ‘equilibrium’ distribution of vegetation for different scenarios of stable future climate derived from GCM simulations. The major weakness in these models is that even if a fully stabilised climate (e.g., for a 2xCO2 scenario) were to occur in reality, it would take a much longer time centuries to millennia - for the vegetation to fully adapt. Hence these models are useful indicators of possible ‘end-points’ for future forest vegetation distribution, but tell us little about the short-term transient responses and feedbacks. For most researchers and others concerned about climate change impacts, it is the latter effects, likely to occur within the next 50-100 years, which are of greatest interest. Dynamic global vegetation models (DGVM) Among the most recent, and possibly the most comprehensive, vegetation models, Dynamic Global Vegetation Models (DGVM) are related to the equilibrium projection models in their representation of vegetation as PFTs. They also combine much of the detailed physics and physiology described in SWAT models as well as representing stand development for forest vegetation in a manner similar to that of FOREST-BGC. Although intended primarily for coarse resolution simulations of the transient responses of global vegetation to environmental change, they are being used for continental-scale modeling as well. They are seen as the successors to static land surface parameterization schemes in fully coupled GCMs, where changes in vegetation due to changes in radiative forcing could result in feedbacks to the atmosphere (e.g., through changes in albedo and the net CO2 flux). An example is the Integrated BIosphere Simulator (IBIS) of Foley et al. (1996; 2000 this volume). IBIS attempts to simulate all the important climate-sensitive ecosystem processes, but each part of the model is driven by climate data of an appropriate temporal resolution. For example, competition processes are driven on a monthly timestep, but canopy exchange processes require daily data created by an internal weather generator. In addition to the relatively standard drivers of temperature, precipitation and solar radiation, IBIS uses a record of rainy days per month to simulate rainfall distributions as well as data on mean windspeed (to estimate aerodynamic transfer coefficients) and atmospheric humidity (for calculating evapotranspiration). Such data needs indicate what variables should be included in a comprehensive climate scenario. Some DGVMs also incorporate results from fire research where fire occurrence, severity and size, are simulated as direct functions of fuel and weather conditions (Lenihan et al., 1998). The main point of the foregoing was not to assess the relative merits of these different types of models, but rather to show that there are several distinct types of ecosystem model in use. With the exception of the traditional forest yield models, all have the potential to generate useful information about the effects of climatic change on forests (e.g., on wood production, geographic distribution or species diversity). The choice of model or models, and therefore, of the data to drive them, will depend on the questions being posed. Any discussion of climate scenarios for use in assessing impacts of climate change on forest ecosystems should consider all of these model types as potential applications for the data.
37
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
CLIMATE SCENARIOS IN ECOSYSTEM IMPACT ASSESSMENTS Many ecological researchers investigating the application of GCM data to climate change impacts studies have concluded either from their own experience, or from the wisdom of climatologists and GCM modellers, that it is unwise or even completely misleading to use GCM data as direct representations of reality. Even so, it is worth reiterating some remarks of Hulme et al. (1995): “. . . it is often assumed by scientists whose specialty is not in the handling and use of climate data either that (comprehensive characterizations of the current climate) are already available or that GCM experiments, as well as providing the climate change scenarios, can also provide adequate information about current climate. Neither of these two assumptions is valid.” Some climatologists even argue that ecosystem modellers should not be attempting to use climate scenarios in impact assessment, because the inherent inaccuracies in the baseline climatology (quite aside from limitations in the GCM output) will cause misleading results (see New, 1999, for a discussion of these issues). Obviously it is crucial to continue refining the data (not only to calibrate the GCMs, but also to determine how much the climate is changing, and which regions are most affected). But we believe (perhaps provoking some argument?) that a highly accurate climatology is not essential to obtain useful information from the impacts models. There are so many uncertainties and inherent errors in the ecosystem models that random errors in temperature data (say < 0.5ºC) or precipitation data (say < 10%) are surely inconsequential. Systematic biases in climate data are of greater concern, naturally, but they are more likely to be known and therefore easier to correct approximately (e.g., undercatch of snow by rain gauges). If there are, as yet undetected, serious systematic errors in climate records, then they are likely to have been ‘corrected’ in the normal tuning and tweaking to which most models are subjected as they are extrapolated to larger regions. To be clear, we are not suggesting that these problems should be ignored, but simply that it could take years of effort before there is sufficient confidence in the ecosystem models to prove that their predictions are of limited value because of uncertainty in the climate data. To pursue this a little further, ecosystem modellers are generally obliged to assume that competent field scientists made the measurements they use to develop and test their models. Thus it is assumed that the field researchers know how to use their instruments properly, calibrate them regularly and check them occasionally against independent reference instruments. Any unknown biases in these measurements are likely also to be present in similar measurements made elsewhere, say at climate stations. It is possible that some biases creep in at larger scales, e.g., because the model is developed using continuous observations at the research site, but is applied over a wider region driven by climate data derived from observations taken at 6 or 12 hourly intervals. Such systematic differences will then manifest as a ‘scaling error’, but in general problems with climate data will be small compared to many of the other concerns due to scaling up to the landscape or region. For example, even if all the important plant and soil processes affecting CO2 exchange in a ‘homogeneous’ forest stand could be captured accurately by an ecosystem model, a typical landscape is composed of many stands of different types and different ages, growing on a range of sites with differing topography, hydrology and nutrition. The model would require parameters of comparable accuracy for every stand and every site before a spatially aggregated estimate of net CO2 flux was limited by poor climate data.
What do ecosystem modellers need? Before we can use any GCM simulation to create a scenario of future climate, we generally need a properly constructed ‘baseline’ climatology corresponding to observed reality for the region of interest. Several global and many regional data sets of past climate have been constructed from station records, of which many are freely available (although the climatologists may still be wrestling with imperfections in the 38
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
source data!). The common approach is then to calculate differentials between GCM control runs and the future projections (usually some averaging is required, particularly with transient simulations). Calculated differentials are then downscaled to the locations of the observed data and merged with them—by simple addition for most variables, but generally using ratios in the case of precipitation and windspeed. Several researchers attending the workshop (Bugmann, Lindner, Lasch, Sykes, Price) have used this kind of approach to create scenarios for point-based simulations with gap dynamics models—particularly along transects represented by individual climate stations. This was particularly attractive when computing power seemed expensive: a simple change in the driving variables (e.g., as simulated for stable 1xCO2 and 2xCO2 climates) could be applied to observed climate data to create a scenario that retained all the essential characteristics (variability, autocorrelation, covariance). (This approach still makes an implicit assumption, however, that these relationships do not alter, or perhaps are linearly related to changes in the means.) Simulations based on known site locations (particularly where long-term climate data are available) will have some distinct advantages when compared to spatial simulations that require data interpolation. With GIS databases (both raster and polygon) and more powerful computers, spatially distributed simulations are now feasible and becoming increasingly common. Most spatial simulations are driven by gridded data which are themselves the end-product of some form of downscaling. In general the underlying principles or statistics used in the downscaling algorithm can be tested exhaustively (for present-day climate) before the final data sets are used to drive an ecosystem simulator. Recalling that GCMs are still not able to simulate present-day climate adequately, gridded scenarios generally suffer from the problem that two data sets need to be downscaled. First, observed climate data must be interpolated somehow from climate station coordinates to the target locations (grid nodes or polygon centres); and second, the climate model output must be downscaled to the same target locations. Although temperature and precipitation regime are obviously the most crucial indicators of climate change for many impacts studies, for some applications GCM forecasts of changes in other variables may also be important. For example, occurrence of forest fires is strongly correlated to fuel dryness, which is commonly estimated using daily or even hourly atmospheric humidity data. Wotton et al. (1998) report that several GCMs tend to overestimate mean daily specific humidity, which tends to cause fire weather index (FWI) models to underestimate the flammability of fuels, and hence the likelihood of serious fires. Additionally, a GCM can provide output on the three dimensional structure of the atmosphere which can be important in statistical models of fire activity. Hence the state of the 50 kPa pressure level has been found a useful predictor for models driven primarily by surface variables (Skinner et al., 1999; Flannigan and Harrington, 1988).
Importance of climate variability It is customary when using forest gap dynamics models to perform multiple runs on identical patches, often 100 or more, to allow a proper assessment of the stochastic components. Monte Carlo methods are typically used in these models to represent seedling establishment, tree mortality and disturbance events, rather than to sample spatial variability in climate. Some of the causes of apparently random events (e.g., death of individual trees) can be explained in reality by microclimatic and soil variations, so it may be possible to reduce the number of stochastic simulations where the representation of environmental variability can be improved. For spatially distributed simulations at fine resolutions (grid or polygon), however, it becomes essential to have a model capable of resolving spatial variability in the key environmental factors. For example, Lindner et al. (1997) found that the gap model they used in their study of forests in central Germany was not generally sensitive to the fine-scale variations in site types in that region.
39
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
More sophisticated approaches include stochastic sampling of the climate and soil variables assuming they are normally distributed within the region of the ‘site’ (e.g., King, 1993). Where a gap dynamics model is being applied on a spatial grid, it may be possible to reduce the number of runs per grid point because the grid itself contributes to the sampling of the environment space. It is tempting to argue that one should therefore invest increased computing power in working at higher resolution grids rather than increase the replicates per grid point. In practice, however, given the inherent stochasticity in these models, it will be necessary to perform a minimum number of replicate runs to obtain a statistically reliable estimate for each pixel. Simulating ecosystem dynamics on regular grids, or for small homogeneous polygons, still has the important advantage that the spatial variability is based on reality, and therefore preferable to random sampling from ‘fake’ normal distributions. Temporal variability in climate is likely to be just as important as spatial variability in site conditions, with the added complication that frequency and magnitude of extreme events may change as the climate changes. Storm events are an obvious example of great importance to forest management in western Europe. To simulate such events and to estimate the damage they cause will require information on the frequency and intensity of extreme winds both for the past and the predicted future. Hence we need either daily data or if using longer (monthly) periods, the statistics which enable daily data characteristics and extremes to be estimated, i.e., by using weather generators (Richardson, 1981; Wilks, 1992, 1999; Semenov et al., 1998). Semenov and Barrow (2000, this volume) discuss the strengths and limitations of weather generators, for assessing climate change impacts on agricultural production. In general, the requirements for modelling temporal climate variability in agricultural systems are more exacting than those for natural ecosystems, so weather generators applied successfully to agricultural applications should be useful for simulating longterm daily climate (variability and extremes) in natural ecosystems. Spatially correlated (i.e., synchronous) weather data will be needed for problems involving spatial propagation of short-term phenomena such as fires and insect dispersal, but the models, such as FIRE-BGC (Keane et al., 1996) are typically applied over 2 much smaller domains—rarely exceeding 1000 km . The traditional approach to modelling effects of climate change on potential fire activity has been to use monthly data (i.e., GCM-future minus GCM-control) as anomalies superimposed on the observed daily climatology in the estimation of future FWIs, and longer-term (seasonal) indices such as the Seasonal Severity Rating (SSR) (e.g., Stocks et al., 1998). This approach is problematic, however, because fire activity is highly sensitive to changes in extremes of temperature and precipitation. For present-day operational purposes the typical range for forecasting and planning is 7-10 days. Hence, daily 90-percentile and extreme minimum and maximum data are typically used in preference to daily means. Shifts in frequency and magnitudes of these daily extremes predicted by the GCM cannot be captured if only the monthly average anomalies are used. Hence in fire research, there has been a recent trend toward use of raw daily (6- or 12-hourly) output from GCMs - typically ensemble averages of three or more separate runs driven by similar forcings. While such data are not generally considered trustworthy (i.e., statistics for control runs do not agree well with observation, and there are often considerable differences among the different runs within each ensemble), they are useful for modelling effects on fire weather. Using such data Flannigan et al. (1998) have created FWI and SSR maps for North America, Europe and Siberia, under both current and future climate simulations (Figures 2 and 3). Systematic biases between observed data and the GCM output can be removed at the final stage by expressing the final results as ratios of future index/control index (Figure 3).
40
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Figure 2: Effects of doubled CO2 climate on estimated mean Fire Weather Index (FWI) for Europe derived from output of the Canadian Climate Centre GCM1 model. Map data are ratios of mean FWIs for the 2xCO2 climate to the ratio for the 1xCO2 climate. Data from Flannigan et al. (1998). [Reproduced by permission of Journal of Vegetation Science.]
Figure 3:As Figure 2, but showing the ratio of maximum FWIs for the 2xCO2 climate to the ratio for the 1xCO2 climate. [Reproduced by permission of Journal of Vegetation Science.]
41
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Spatial resolution of climate scenarios How much spatial detail do we actually need in forecasting ecosystem responses to a changing climate? Although the answer to this question will depend on the objective of the modelling exercise and a number of technical factors such as the area of the domain, quality of source data and whether the terrain is flat or mountainous, it seems that there are some guidelines that can be offered in building data sets: • It is easier to aggregate from a fine resolution than it is to interpolate from a coarser one. • Higher resolution allows more of the climate variable space (temperature x precipitation x radiation x windspeed x humidity etc.) to be sampled, which in turn leads to more checks on model stability and the robustness of output. Furthermore, other variables, such as soil texture and water content, often vary on much finer scales, so higher resolution grids enable better sampling of the multivariate space of climate and other factors. • Intercomparison of studies performed at different resolutions is easier if the climate data are derived from the same fine-scale data set. • There may be a need to think on a scale that has some meaning to the ‘end-user’. For example, impact assessments generally need to be reported at a resolution useful to forest managers and/or regional government policy makers. It is generally agreed that accounting for topographic effects in spatial climate interpolation is important (McKenney et al., 1996; Hutchinson, 1995; Daly et al., 1994). That said, a critical problem in interpolating GCM data to higher resolutions is how to correct for the effects of the relatively poor representation of topography inherent in most present day GCM outputs (unavoidable when model elevation is typically 2 averaged over grid cell areas of more than 100,000 km ). This can create completely unrealistic resultsparticularly in the vicinity of large mountain ranges. Some form of vertical scaling must presumably be included in the downscaling algorithm.
METHODS OF DOWNSCALING GCM SIMULATIONS There appear to be at least three distinct approaches to downscaling raw GCM simulation output to construct climate scenarios of higher spatial resolution than the GCM grid. These are: • addition of interpolated GCM data to observed climatology—the latter based on spatial interpolation using either purely statistical methods, or statistics plus expert knowledge; • statistical downscaling of simulated meteorology to present-day and projected future climate by various analytical or statistical methods; and • dynamical downscaling using regional climate (or limited area) models (RCMs) nested inside one or more of the GCM grid cells (see also Giorgi and Mearns, 1999).
Interpolation For large regions, interpolation approaches range from purely statistical interpolators such as ANUSPLIN (Hutchinson, 1995, 1997; McKenney et al., 1996) and MTCLIM (Thornton et al., 1997) to relatively sophisticated methods such as PRISM (Daly, 1994), which require additional expert knowledge to enhance 42
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
the interpolation, and are therefore more labour-intensive and costly. PRISM works around data limitations, e.g., to account for rain-shadow effects of mountain ranges. There is a fair amount of debate as to whether the additional expertise employed in PRISM results in a significantly better product. We have recently completed a comparison of data interpolated using ANUSPLIN with observations at the locations of stations excluded from the input data and found generally good agreement, with the exception of highelevation regions where station coverage is sparse (Price et al., 2000). From the Canadian perspective, although more complicated methods of interpolation may give noticeably better results, this is a rather different issue from getting useful scenario data for the future. There are so many errors inherent in any use of data derived from GCMs that will overwhelm the gains obtained from a better interpolator that it could be more practical to use a simpler (i.e., cheaper) method. Conversely in smaller regions where climate records have been maintained for long periods across a dense network of stations, such as in the Alps, there are a lot more data with which to work. This allows gridded climatologies of high accuracy and high resolution to be built relatively easily, which can be used for a range of applications in addition to development of future climate scenarios. In such cases, it may well be appropriate to invest the resources needed to do the best possible interpolation and then use the product as the basis for creating scenarios of future climate. As an example the U.S. Vegetation/Ecosystem Modeling and Analysis Project (VEMAP Members, 1995; Kittel et al., 1997; http://www.cgd.ucar.edu/vemap/) has invested significant resources in spatial mapping of climate data, as a prerequisite for carrying out comprehensive inter-comparisons of several large-scale ecological models applied to the continental USA. Both historical climate observations and transient climate scenarios (derived from Hadley Centre and Canadian Climate Centre coupled GCM simulations) have been interpolated to a common 0.5º x 0.5º grid. The Phase 2 historical gridded record (1895-1993) provides spatially consistent data for monthly minimum and maximum temperatures and precipitation for the continental USA, based on data from up to 8000 climate stations (VEMAP members, 1995). Missing data were estimated statistically to create continuous records at all locations. These records were then spatially interpolated using PRISM (Daly, 1994; Kittel et al., 1995). Differentials (or ratios) between GCM-simulated data and GCM-simulated 1961-90 means were interpolated to a 0.5º x 0.5º grid without topographic or other corrections, and then added to (or multiplied by) the 1961-90 means of the observed monthly values. An important feature of the VEMAP data sets is that the variability simulated in the GCM time series data was retained in the gridded scenarios (although model biases in mean values were removed). Hence, all biases in the modelled variances (i.e., compared to observed variances) were also retained in the scenarios (except in the case of precipitation ratios, which were modulated by the corrections for bias in the means). There does not appear to be an established method for correcting GCM-derived estimates of changes in variability in an interpolated scenario. In the VEMAP2 data sets, for example, Kittel et al (in prep.) decided to accept the variances generated by the GCMs, even though there were known biases in the frequency distributions of these variances compared to the observed record. Semenov and Barrow (2000, this volume) also note the lack of a robust procedure, but propose one approach where GCM daily data were used to calculate changes in parameters used to drive the LARS-WG weather generator.
Statistical downscaling Much climate variability is caused by sub-GCM grid scale meteorological phenomena, such as convective storm precipitation. Statistical downscaling attempts to compensate for this problem, generally by developing relationships between past regional weather patterns and observed climate statistics, and then using these to predict the effect of GCM simulations of future synoptic patterns on the regional climate. There are a number of approaches reported in the literature. Some are purely statistical (e.g., Gyalistras et al., 1994; Huth, 2000), some use stochastic methods (e.g., Hughes et al., 1993; Zorita et al., 1995) and some employ neural networks to approximate nonlinear relationships (Weichart and Bürger, 1998; Snell et al., 43
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
2000). Such methods can produce significantly better results than interpolation for smaller and topographically complex regions such as the Alps, particularly when spatial and temporal coherence among a set of hourly weather variables is important. A major concern, frequently expressed in the literature, is that relationships established from past observations of weather systems and climate may not be valid for a changed climate system in the future. Moreover, many statistical downscaling methods do not reproduce all the characteristics of observed climate which are ecologically significant, e.g., intra- and interannual extremes, and covariances among variables such as rainfall, temperature and radiation. This apparent limitation is a focus of ongoing work and some of these problems are likely to be resolved (e.g., see Wilby et al. 1998, for a comparison of these techniques). Nevertheless these approaches clearly hold a lot of promise, particularly when downscaling GCM data to regions with complex terrain. Weichart and Bürger (1998) suggest that the best method may be a combination of the linear statistical and stochastic approaches.
Dynamical downscaling: Regional Climate Models Poor spatial resolution, inherent in the operation of most present-day GCMs, means that it is misleading and impractical to apply simulation output to precise locations, such as urban centres, farms and forest stands. A potential solution to this problem is the relatively recent development of RCMs, nested inside GCM grid-cells, which promise higher resolution coverage for particular regions of the globe (see Giorgi and Mearns, 1999 for a review of the state-of-the-art in RCMs). The main benefit of using RCMs to downscale GCM simulations, as compared to statistical downscaling, is that they are based on the same systems of physical equations used to formulate the GCMs, and which are assumed to best capture the collective knowledge of atmospheric processes. For example, the Canadian RCM (CRCM), currently covers much of Canada at 45 km resolution—an order of magnitude higher resolution than most GCMs (Laprise et al., 1998), but still inadequate for many impact studies. Some possibilities exist for running CRCM at higher resolutions, but these are currently limited to very small areas or very short simulations because of the computational expense. As with GCMs, most RCMs are currently unable to reproduce the full range of spatial and temporal variability of observed climate. This could be important for many studies of climate change impacts on ecosystems that must differentiate the effects of natural variability from the systematic changes attributed to increased atmospheric greenhouse gases. At the risk of being controversial, we suggest that RCMs are already showing their potential as a means of developing more realistic regional climate scenarios than GCMs (although there is still a long way to go). This view is reflected in Giorgi and Mearns (1999), who compare simulated precipitation fields in the western USA with observed data. Compared to the ‘host’ GCM running alone, the NCAR RegCM does a markedly better job in capturing both the magnitudes and distributions of precipitation east of the Rocky Mountains. Wotton et al. (1998) observed similar differences for the Canadian Prairie regions where CRCM generates more plausible estimates of precipitation and hence of regional FWIs.
THE FUTURE:WHAT SHOULD WE BE DOING DIFFERENTLY OR BETTER? Undoubtedly forest ecosystem models will continue to evolve, both within the broad classes described earlier, and as new hybrids between them. The general trend will be towards better representations of processes, even though modellers will continue to look for more efficient ways of implementing them. Process models will emerge that provide complete and relatively detailed linkages among physical and biological processes including climate-driven disturbances and atmospheric feed-back effects. In particular, dynamic vegetation models will be routinely coupled to the GCMs and/or nested RCMs to create complete ‘earth system models’ (see also Giorgi and Mearns’ (1999) reference to the concept of nested regional earth system models).
44
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Not all these developments will be of direct interest to ecosystem modellers investigating climate change impacts on forests and natural ecosystems. There will continue to be a strong need to analyse effects of climate scenarios treated as exogenous factors influencing the natural vegetation for a country or region. Hence, we would like to see: • higher resolution, topographically corrected, climate scenarios (RCM grids are tending towards 10 km and even finer resolutions). • some agreement on acceptable methods for performing downscaling of means and variances (but probably dependent on application)—leading to the production of some standardized procedures for getting state-of-the-art climate scenarios to impacts modellers in a convenient form. On this second point, while it is possible for individual researchers to perform the downscaling themselves, it seems sensible to construct and distribute high-quality data sets of general use to researchers in many fields. First, such an approach avoids duplication of effort and lack of consistency in the way information is extracted from data. Second, it allows results of different impact studies to be compared directly without confounding due to methodological differences in manipulating the climate model output. The preparation of such data sets requires a strong collaboration (and perhaps some compromise) between the climatologists who would like to see the limitations in climate data and models treated very cautiously, and the impacts modellers who need the climate scenarios and are willing to take some risks with them. A good way for impacts researchers to appreciate the limitations in the available climate data sets is to perform studies with as wide a range of scenarios as possible. Apart from helping to quantify effects of uncertainty, this allows the relative sensitivities of different ecosystems to a range of possible future climates to be explored. It may be difficult to prevent planners and resource managers from treating individual results as predictions, but a wider range of scenarios will help convey the level of confidence that can be attached to any one outcome. The resulting management prescriptions and longer term policy planning may be less precise, but presumably they can be made more flexible where the uncertainty is greatest. On a more practical level, perhaps any study of climate change impacts should use at least two scenarios before it can be considered acceptable for peer-review?
CONCLUDING REMARKS Although current GCMs may appear disappointingly incapable of simulating present-day climate, the data sets they generate are still the best-available encapsulation of current knowledge for estimating future climate trends. GCM scenarios are useful to impacts researchers for testing the sensitivity and plausibility of ecosystem models in ways that cannot be investigated using observed data alone. We recognise that there may be systematic differences between GCM-simulated mean values and observed data, but most can be removed relatively easily (although significant biases in simulated variances remain an unresolved problem at present). Moreover, the limitations inherent in most present-day impacts models probably outweigh many of the systematic biases likely to occur in both observations and normalized GCM data. Hence data sets of poor accuracy can still be very useful for examining the behaviour of ecosystem models and for making comparative assessments of simulated ecosystem responses to alternative scenarios of future climate. (Obviously individual researchers should still examine scenario data sets very carefully for consistency and plausibility before using them to carry out complex impact studies.) It is conceivable that aberrant behaviour of the impact model can indicate problems in these data sets that will lead to improvements in the GCMs, particularly when vegetation feedbacks to the atmosphere are being simulated by fully coupled dynamic models.
45
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Various techniques exist for downscaling GCM simulations to a scale appropriate for most ecological impacts studies. We look to the climatologists for guidance on the strengths and weaknesses of each approach. Interpolation is attractive compared to statistical downscaling because it seems relatively simple and makes few assumptions about the relationships between future weather phenomena and surface climate. Dynamical downscaling, through the use of regional climate models, appears to offer the best prospects for creating coherent regional climate scenarios at high spatial resolution, but a lot of work is still required. In turn, impacts modellers have an important responsibility to policy makers and managers concerned about the effects of climatic change on ecosystems. These researchers must convey the fact that both the ecosystem models and the climate scenarios used to investigate effects of climate change, currently provide a very poor predictive capability. At this stage we can offer only a range of possibilities, with expert opinions on which are more likely. At the same time it is important to show that the situation is not hopeless, and that continued effort invested in improving and testing both the GCMs and the impacts models will enable us to narrow the range of uncertainty. Acknowledgements The authors greatly appreciate the thoughtful and constructive review of the original draft of this paper by Harald Bugmann. DTP gratefully acknowledges funding to prepare and present this paper from the Potsdam Institute for Climate Impact Research and the ECLAT-2 program. References Botkin, D.B., Janak, J.F. and Wallis, J.R. (1972) Some ecological consequences of a computer model of forest growth, Journal of Ecology 60, 849–872. Bugmann, H. (1997) Sensitivity of forests in the European Alps to future climatic change, Climate Research 8, 35–44. Carter, T.R., Hulme, M. and Viner, D. (eds.) (1999) Representing Uncertainty in Climate Change Scenarios and Impact Studies. ECLAT-2 Workshop Report No. 1. Helsinki, Finland, 14-16 April 1999. Climatic Research Unit, University of East Anglia, UK, 127 pp. Daly, C., Neilson, R.P. and Phillips, D.L. (1994) A statistical-topographic model for mapping climatological precipitation over mountainous terrain, Journal of Applied Meteorology 33, 140–158. Doherty, R. and Mearns, L.O. (1999) A Comparison of Simulations of Current Climate from Two Coupled Atmosphere-Ocean Global Climate Models Against Observations and Evaluation of their Future Climates. Report to the National Institute for Global Environmental Change (NIGEC) in support of the US National Assessment. Draft available at http://www.dir.ucar.edu/esig/doherty/text.html. Flannigan, M.D., Bergeron, Y. Engelmark, O. and Wotton, B.M. 1998. Future wildfire in circumboreal forests in relation to global warming, Journal of Vegetation Science 9, 469–476. Flannigan, M.D. and Harrington, J.B. (1998) A study of the relation of meteorological variables to monthly provincial area burned by wildfire in Canada 1953-80, Journal of Applied Meteorology 27, 441–452. Foley, J.A., Prentice, I.C., Ramankutty, N., Levis, S., Pollard, D., Sitch, S. and Haxeltine, A. (1996) An integrated biosphere model of land surface processes, terrestrial carbon balance, and vegetation dynamics, Global Biogeochemical Cycles 10, 603–628. Giorgi, F. and Mearns, L.O. (1999) Introduction to special section: Regional climate modeling revisited, Journal of Geophysical Research 104(D6), 6335–6352. Grant, R.F. and Nalder, I.A. (2000). Climate change effects on net carbon exchange of a boreal aspen-hazelnut forest: estimates from the ecosystem model ecosys., Global Change Biology 6, 183–200. Gyalistras, D., Von Storch, H., Fischlin, A. and Beniston, M. (1994) Linking GCM generated climate scenarios to ecosystems: case studies of statistical downscaling in the Alps, Climatic Research 4, 167–189. 46
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Hughes, J.P., Lettenmaier, D.P. and Guttorp, P. (1993) A stochastic approach for assessing the effect of changes in synoptic circulation patterns on gauge precipitation, Water Resources Research 29(10), 3303–3315. Hulme, M., Conway, D., Jones, P.D., Jiang, T., Barrow, E.M. and Turney, C. (1995) Construction of a 1961–1990 European climatology for climate change modelling and impact applications, International Journal of Climatology 15, 1333–1363. Hutchinson, M.F. (1995) Interpolation of mean rainfall using thin-plate splines, International Journal of GIS 9, 385–403. Hutchinson, M.F. (1997) ANUSPLIN Version 3.2. http://cres.anu.edu.au/software/anusplin.html Huth, R. (1999) Statistical downscaling in central Europe: evaluation of methods and potential predictors, Climate Research 13, 91–101. Karjalainen, T. (1996) Dynamics and potentials of carbon sequestration in managed stands and wood products in Finland under changing climatic conditions, Forest Ecology and Management 80, 113–132. Keane, R.E., Ryan, K. and Running, S.W. (1996) Simulating effects of fire on northern Rocky Mountain landscapes using the ecological process model FIRE-BGC, Tree Physiology 16, 319–331. Kerr, R.A. (1998) Models win big in forecasting El Niño, Science 280, 522–523. King, A. (1993) Quantifying regional changes in terrestrial carbon storage by extrapolation from local ecosystem models. In Carbon Cycling in Boreal Forests and Subarctic Ecosystems: Biospheric Responses and Feedbacks to Global Climate Change, Kolchugina, T.P. and Vinson, T.E. (eds.). U.S. EPA Report EPA/600R-93/084, Office of Research and Development, Washington DC, pp. 221–234. Kittel, T.G.F., Rosenbloom, N.A., Painter, T.H., Schimel, D.S. and VEMAP Modeling Participants. (1995) The VEMAP integrated database for modeling United States ecosystem/vegetation sensitivity to climate change, Journal of Biogeography 22, 857–862. Kittel, T.G.F., Royle, J.A., Daly, C., Rosenbloom, N.A., Gibson, W.P., Fisher, H.H., Schimel, D.S., Berliner, L.M. and VEMAP2 Participants. (1997) A gridded historical (1895–1993) bioclimate data set for the conterminous United States. In Proceedings of the 10th Conference on Applied Climatology, Reno, NV, 20–24 October 1997. American Meteorological Society, Boston. MA, pp. 219–222. Kittel, T.G.F., Rosenbloom, N.A., Fisher, H.H., Schimel, D.S. and VEMAP2 Participants. The VEMAP Phase 2 dataset: II. Gridded transient bioclimate change scenarios for the conterminous United States and Alaska. In preparation. Laprise, R., Caya, D., Giguère, M., Bergeron, G., Côté, H., Blanchet, J.-P., Boer, G.J. and McFarlane, N. (1998) Climate and climate change in western Canada as simulated by the Canadian Regional Climate Model, Atmosphere-Ocean 36, 119–167. Lenihan, J. M., Daly, C., Bachelet, D. and Neilson, R. P. (1998) Simulating broad-scale fire severity in a dynamic global vegetation model, Northwest Science 72, 91–103. Lindner, M. (2000) Developing adaptive forest management strategies to cope with climatic change, Tree Physiology 20, 299-307. Lindner, M., Bugmann, H., Lasch, P., Flechsig, M. and Cramer, W. (1997) Regional impacts of climatic change on forests in the state of Brandenburg, Germany, Agricultural and Forest Meteorology 84, 123–135. McGuire, A.D., Joyce, L.A., Kicklighter, D.W., Melillo, J.M., Esser, G. and Vörösmarty, C.J. (1993) Productivity response of climax temperate forests to elevated temperature and carbon dioxide: A North American comparison between two global models, Climatic Change 24, 287–310. McKenney, D.W., Mackey, B.G. and Sims, R.A. (1996) Primary databases for forest ecosystem managementexamples from Ontario and possibilities for Canada: NatGRID, Environmental Monitoring and Assessment 39, 399–415. New, M. (1999) Uncertainty in representing observed climate. In Representing Uncertainty in Climate Change Scenarios and Impact Studies, Carter, T., Hulme, M. and Viner, D. (eds.), ECLAT-2 Workshop Report No. 1, Helsinki, Finland, 14-16 April 1999. Climatic Research Unit, University of East Anglia, UK, pp. 59–66. Price, D.T., McKenney, D.W., Nalder, I.A., Hutchinson, M.F. and Kestevan, J.L. (2000) A comparison of statistical and thin-plate spline methods for spatial interpolation of Canadian monthly climate data, Agricultural and Forest Meteorology 101, 81–94. 47
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Prentice, I.C., Cramer, W., Harrison, S.P., Leemans, R., Monserud, R.A. and Solomon, A.M. (1992) A global biome model based on plant physiology and dominance, soil properties and climate, Journal of Biogeography 19, 117–134. Richardson, C.W. (1981) Stochastic simulation of daily precipitation, temperature and solar radiation, Water Resources Research 17, 182–190. Rounsevell, M.D.A. (2000) Agriculture and Climate Change Scenarios. In Climate Scenarios for Agricultural, Forest and Ecosystem Impacts, Cramer, W., Doherty, R., Hulme, M. and Viner, D. (eds.), ECLAT-2 Workshop Report No. 2, Potsdam, Germany, 13–15 October 1999. Climatic Research Unit, University of East Anglia, Norwich, UK, pp21-31. Running, S.W. and Coughlan, J.C. (1988) A general model of forest ecosystem processes for regional applications. I. Hydrologic balance, canopy gas exchange and primary production processes, Ecological Modelling 42, 125–154. Semenov, M.A., and Barrow, E.M. (2000) Development of Climate Change Scenarios for Agricultural Applications. In Climate Scenarios for Agricultural, Forest and Ecosystem Impacts, Cramer, W., Doherty, R., Hulme, M. and Viner, D. (eds.), ECLAT-2 Workshop Report No. 2, Potsdam, Germany, 13–15 October 1999. Climatic Research Unit, University of East Anglia, Norwich, UK, pp50-58. Semenov, M.A., Brooks, R.J., Barrow, E.M. and Richardson, C.W. (1998) Comparison of the WGEN and LARSWG stochastic weather generators for diverse climates, Climate Research 10, 95–107. Sellers, P.J., Mintz, Y., Sud, Y.C. and Dalcher, A. (1986) A Simple Biosphere model (SiB) for use within General Circulation Models, Journal of Atmospheric Science 43, 505–531. Sellers, P.J., Hall, F.G., Collatz, G.J., Kelly, R.D., Black, T.A., Baldocchi, D., Berry, J., Ryan, M., Ranson, K.J., Crill, P.M., Lettenmaier, D.P., Margolis, H., Cihlar, J., Newcomer, J., Fitzjarrald, D., Jarvis, P.G., Gower, S.T., Halliwell, D., Williams, D., Goodison, B., Wickland, D.E. and Guertin, F.E. (1997) BOREAS in 1997: Experiment overview, scientific results, and future directions, Journal of Geophysical Research 102 (D24), 28731–28769. Skinner, W.R., Stocks, B.J., Martell, D.L., Bonsal, B. and Shabbar, A. (1999) The association between circulationanomalies in the mid-troposphere and area burned by wildland fire in Canada, Theoretical Applied Climatology 63, 89–105. Snell, S.E., Gopal, S. and Kaufmann, R.K. (2000) Spatial interpolation of surface air temperatures using artificial neural networks: evaluating their use for downscaling GCMs, Journal of Climate 13(5), 886-895. Stocks, B.J., Fosberg, M.A., Lynham, T.J., Mearns, L., Wotton, B.M., Yang, Q., Jin, J.-Z., Lawrence, K., Hartley, G.R., Mason, J.A. and McKenney, D.W. (1998) Climate change and forest fire potential in Russian and Canadian boreal forests, Climatic Change 38, 1–13. Sykes, M.T., Prentice, I.C. and Cramer, W. (1996) A bioclimatic model for the potential distribution of North European tree species under present and future climates, Journal of Biogeography 23, 203–233. Thornton, P.E., Running, S.W. and White, M.A. (1997) Generating surfaces of daily meteorological variables over large regions of complex terrain, Journal of Hydrology 190, 214–251. VEMAP Members. (1995) Vegetation/Ecosystem Modeling and Analysis Project (VEMAP): Comparing biogeography and biogeochemistry models in a continental-scale study of terrestrial ecosystem responses to climate change and CO2 doubling, Global Biogeochemical Cycles 9, 407–437. Weichart, A. and Bürger, G. (1998) Linear versus nonlinear techniques in downscaling, Climate Research 10, 83–93. Wilby, R.L., Wigley, T.M.L., Conway, D., Jones, P.D., Hewitson, B.C., Main, J. and Wilks, D.S. (1998) Statistical downscaling of general circulation model output: a comparison of methods, Water Resources Research 34(11), 2995–3008. Wilks, D.S. (1992) Adapting stochastic weather generation algorithms for climate change studies, Climatic Change 22, 67–84. Wilks, D.S. (1999) Interannual variability and extreme value characteristics of several stochastic daily precipitation models, Agricultural and Forest Meteorology 93, 153–169.
48
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Wotton, B.M., Stocks, B.J., Flannigan, M.D., Laprise, R. and Blanchet, J.-P. (1998) Estimating current and future fire climates in the boreal forest of Canada using a Regional Climate Model. In Proceedings of the Third International Conference on Forest Fire Research and Fourteenth Conference on Fire and Forest Meteorology, November 16–20, 1998, Luso, Portugal, pp. 1207–1221 Zorita, E., Hughes, J.P., Lettenmaier, D.P. and von Storch, H. (1995) Stochastic characterization of regional circulation patterns for climate model diagnosis and estimation of local precipitation, Journal of Climate 8, 1023–1042.
49
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Keynote Paper 3 Development of Climate Change Scenarios for Agricultural Applications 1
2
Mikhail A. Semenov and Elaine M. Barrow
1. IACR-Long Ashton Research Station, Department of Agricultural Sciences, University of Bristol, Bristol, BS41 9AF, UK (
[email protected]) 2. Canadian Climate Impacts Scenarios Group, c/o Environment Canada- Prairie and Northern Region, Atmosphere and Hydrologic Science Division, 2365 Albert Street, Room 300, Regina, Saskatchewan S4P 4KI, Canada.
INTRODUCTION In order to develop scenarios of climate change which are of greatest use in impacts assessment the scenarios should be tailored to their area of application. The first stage in this process is a sensitivity analysis of the impact model in question to changes in the relevant climate variables. Changes in those variables which may result in noticeable changes in the output of the impact model, should be incorporated in the climate scenarios in order to produce realistic climate change scenarios. Very often in modelling studies investigating the impact of climate change on crop production changes in only the means of the climate variables have been considered (Hulme et al., 1999). These changes, derived from global climate models (GCMs), were usually applied to historical weather data to construct scenarios of climate change relevant to agricultural applications. Analyses of the sensitivity of crop simulation models to changes in climate variables has clearly shown that changes in climate variability can have a significant effect on crop growth and associated agricultural risk (Semenov and Porter, 1995; Mearns et al., 1996). Crop simulation models incorporate non-linear responses of the crop to its environment, and it is equally important for impact assessments to include changes in climate variability as well as changes in mean climate. The tools, which are most widely used to construct scenarios of climate change for impacts assessment, are GCMs (Giorgi and Mearns, 1991; Viner and Hulme, 1994). These complex computer models describe the climatological conditions of the Earth at a finite number of grid points. The limiting factor for running GCMs is computational power; a compromise must be reached between the spatial resolution of the model and the computer time required to perform a single experiment. Hence, most GCMs tend to have a coarse spatial resolution that leads to approximations in the model representation of meteorological variables at the regional or local scale. These so-called ‘sub-grid scale’ processes have to be parameterised in the model rather than solved realistically as a function of the fundamental equations. However, despite these limitations, GCMs still provide an opportunity to examine the evolution of climate under a variety of conditions. There are a number of factors, which limit the direct use of GCMs output in scenario development. These include: • The ability of the control experiment to adequately simulate the larger-scale features of the present-day climate. This is one of the reasons that the difference between the control and perturbed integrations is used in climate scenario development, rather than the raw data from the integrations themselves.
50
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
• The coarse spatial grid - output is on the scale of hundreds of kilometres rather than the tens of kilometres needed for impact assessment. This coarse resolution also means that sub-grid scale processes, such as precipitation, are not adequately represented and important regional topographic features are also omitted. Hence, although GCMs may be able to simulate large-scale features of climate well, their simulation of regional climate is considerably poorer. In this study the output from a GCM experiment was combined with a stochastic weather generator, LARS-WG, in order to produce climate change scenarios which were suitable for use in agricultural impact assessment. The requirements of such scenarios may be summarised as follows: • scenarios should be site-specific with daily temporal resolution; • they should include the full set of climate variables required by the impacts model; • they should include changes in means and climate variability; • they should contain an adequate number of years to permit risk analysis; and • the set of scenarios used should encompass a range of future plausible climates rather than a single scenario In this study climate change scenarios were constructed in two ways. First, climate change scenarios with high spatial resolution were constructed using regression downscaling to obtain site-specific climate data from the coarse grid-scale GCM data. Second, changes in climate variability were incorporated into the scenarios. In this latter case, the GCM data were utilised without any downscaling in the absence of a robust method to downscale coarse resolution variability to the site-specific scale. The basic method of producing the climate change scenarios is the same regardless of whether downscaling or climate variability are included. Climate change information, derived from GCMs, was used to perturb the parameters of the stochastic weather generator, LARS-WG, which had previously been calibrated for each site using observed daily climate data. Daily scenario data were then generated from these perturbed parameters. Results are reported for two sites, namely Rothamsted, UK and Seville, Spain.
THE LARS-WG STOCHASTIC WEATHER GENERATOR Models for the simulation of time-series of a suite of climate variables with certain statistical properties have a long history. The first examples are found in the early 1960s (e.g., Gabriel and Neumann, 1962). Initially models were developed to simulate a single variable, most often daily precipitation for use in hydrological applications. From the beginning of the 1980s models, which could generate a whole suite of climate variables, stochastic weather generators, became available (Richardson, 1981; Racsko et al., 1991). Stochastic weather generators may be site-specific, i.e., they generate weather time-series for a single site, or spatial, i.e., they generate weather for a number of locations simultaneously, reflecting the spatial correlation of the different climate variables (Bardossy and Plate, 1991; Hutchinson, 1995; Wilks, 1999). Originally there were two main reasons for the development of stochastic weather generators. The first was the provision of a means of simulating synthetic weather time-series with certain statistical properties which were long enough to be used in an assessment of risk in hydrological or agricultural applications. The observed weather series normally required as input into mathematical models of hydrological processes or simulation models of crop growth are often insufficiently long to allow the estimation of the probability functions of rare events. The second purpose was to provide the means of extending the simulation of weather time-series to unobserved locations. It is worth noting that a stochastic weather generator is not a 51
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
predictive tool, which can be used in weather forecasting, but is a means of generating time-series of synthetic weather statistically ‘identical’ to the observations. It must be borne in mind that statistical ‘identity’ depends on the number of statistics used for the comparison. New interest in local stochastic weather simulation has arisen as a result of climate change studies. Output from GCMs cannot be used directly as climate change scenarios for the reasons mentioned earlier. The weather generator, however, can serve as a computationally inexpensive tool to produce multiple-year climate change scenarios at the daily timescale, which incorporate changes in means and climate variability (Semenov and Barrow, 1997; Wilby et al, 1998; Wilks, 1999). 5
In this study the LARS-WG stochastic weather generator has been used (Semenov and Barrow, 1997; Semenov et al, 1998; Semenov and Brooks, 1999). It generates a suite of climate variables, namely precipitation, maximum and minimum temperature and solar radiation. Precipitation is considered as the primary variable and the other three variables on a given day are conditioned on whether the day is wet or dry. The simulation of precipitation occurrence is based on distributions of the length of continuous sequences, or series, of wet and dry days. This is different from the approach suggested by Bailey (1964) and re-used by Richardson (1981), which applies a first-order Markov chain to describe the occurrence of wet and dry days. The main limitation of the ‘Markovian’ approach is that the Markov chain has a ‘limited memory’ of rare events and, for example, could fail to simulate accurately long dry series at certain locations (Racsko et al., 1991). This problem was resolved by using the series approach, where the distribution of wet and dry series is derived by accumulating information from the observations. Consideration of long dry series is important in agricultural studies since long droughts significantly affect crop growth and can dramatically decrease yields. Semi-empirical distributions were used to model the dry and wet series so that LARS-WG would be applicable over a wide range of climates (Semenov et al., 1998). The amount of rain on wet days was also simulated using a semi-empirical distribution. Daily minimum and maximum temperatures are considered as stochastic processes with daily means and daily standard deviations conditioned on the wet or dry status of the day. The technique used to simulate the process is very similar to that presented in Yevjevich (1972) and Richardson (1981). The seasonal cycles of means and standard deviations are modelled by finite Fourier series of order 3 and the residuals are approximated by a normal distribution. The Fourier series for the mean is fitted to the observed mean values for each month. Before fitting the standard deviation Fourier series, the observed standard deviations for each month are adjusted to give an estimated average daily standard deviation by removing the estimated effect of the changes in the mean within the month. The adjustment is calculated using the fitted Fourier series already obtained for the mean. The observed residuals, obtained by removing the fitted mean value from the observed data, are used to analyse a time auto-correlation for minimum temperature and a time auto-correlation for maximum temperature. For simplicity both of these are assumed to be constant through the whole year for both dry and wet days with the average value from the observed data being used. The analysis of daily solar radiation over many locations showed that the normal distribution for daily solar radiation, commonly used in other generators, is unsuitable for certain climates. The same results have been found in Australia for the distribution of cloud cover (Chia and Hutchinson, 1991). The distribution of solar radiation also varies significantly on wet and dry days. Therefore, separate semi-empirical distributions were used to describe solar radiation on wet and dry days. An auto-correlation coefficient was calculated for solar radiation and assumed to be constant throughout the year.
5
The LARS-WG is in a public domain and a version for Windows 9x/NT is available from www/lars.bbsrc.ac.uk/model/larswg.html.
52
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
CONSTRUCTION OF THE CLIMATE CHANGE SCENARIOS Data from the UK Met. Office transient GCM (UKTR; Murphy, 1995; Murphy and Mitchell, 1995) experiments were used in the construction of the climate change scenarios. Construction of climate change scenarios from the transient experiment was not straightforward. One of the problems of UKTR is climate drift in its control integration - there is a noticeable deviation (approximately 1°C) from the initial ten-year average over the 75-year period of the simulation. How this drift is handled affects the way in which the scenarios are constructed and thus there are a number of different ways of calculating the change fields, each of which makes assumptions about the climate variability and control integration drift. For our purposes the change fields from UKTR were constructed by calculating the difference between a period in the climate change integration and the corresponding years of the control integration. This definition is appropriate if it is assumed that both the control and climate change integrations exhibit similar drift and long-term variability. Data from UKTR were available only as decadal time-slices and the last decade, model years 66-75, was selected for use. The global-mean temperature change corresponding to this decade is 1.76ºC. Depending on assumptions concerning future greenhouse gas emissions and climate sensitivity a range of dates as to when this temperature change may occur can be calculated, but a best estimate is towards the middle of the next century. Scenarios using regression downscaling In order to produce scenarios of climate change at the scale required by crop-growth simulation models, it was necessary to ‘downscale’ the coarse resolution GCM data to specific sites. This procedure involved the development of relationships between the coarse- and local-scale data for the climate variables concerned. There are currently a number of downscaling methodologies in use, including circulation patterns (e.g., Bardossy and Plate, 1991; Matyasovszky et al., 1993) and regression techniques (e.g., Kim et al., 1984; Wigley et al., 1990; Karl et al., 1990; von Storch et al., 1993). Both methods use existing instrumental databases to determine the relationships between large-scale and local climate. Regression techniques develop statistical relationships between local station data and grid-box scale, area-average values of say, temperature and precipitation and other meteorological variables. The circulation pattern approach classifies atmospheric circulation according to type and then determines links between the circulation type, e.g., westerly, and climate variable, e.g., precipitation. There are a number of reservations, however, which need to be considered when using circulation patterns as part of climate change studies including the problems that some GCMs have in simulating the correct frequencies of weather type and also the observed relationships between particular circulation patterns and temperature and precipitation. Also, the relationships between circulation patterns and, for example, temperature and precipitation, in one area of Europe may not be applicable in another location, so for these reasons it was decided to use the regression approach to downscaling. At Rothamsted and Seville, sites selected for downscaling, regression relationships were calculated between local station data (mean temperature and precipitation; i.e., the predictands) and grid-box scale, monthly anomalies of mean sea level pressure (MSLP), the north-south and east-west pressure gradients, temperature and precipitation (i.e., the predictors). The regression relationships were based on anomalies from the long-term mean in order to facilitate the use of the GCM-derived changes in the equations. Observed area-averages corresponding to the grid-box area of UKTR were calculated for Rothamsted and Seville for mean temperature and precipitation. Anomalies from the 1961-90 mean were then calculated for each month for each of the five predictor variables. The data set was split into two time periods, one of which was used to calibrate the regression equations whilst the other was used to verify their performance. 53
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Regression relationships were then calculated between the local (i.e., site) and regional (i.e., grid-box) climate. The next step in the procedure was the calculation of the changes in the predictor variables from UKTR. At both Rothamsted and Seville some of these changes, particularly mean temperature, were outside of the anomaly ranges originally used to calibrate the regression models. Despite this, it was decided to continue the downscaling process, but to add a caveat regarding the confidence placed in the downscaled results because of the combination of poor performance of some of the regression models and of the grid-box changes being outside of the calibration range in some instances. The downscaled changes in mean temperature and precipitation were then used to perturb the parameters of LARS-WG (all other parameters were kept unchanged). No changes in variability were included in these scenarios.
Scenarios incorporating changes in inter-daily climate variability The climate change scenarios incorporating changes in inter-daily climate variability were constructed without any downscaling of the GCM information for the two sites, Rothamsted and Seville. This was because a robust procedure for downscaling the variability parameters was not available. Daily data for the appropriate grid boxes from the control and perturbed integrations of the UKTR experiment were used to calculate changes in precipitation intensity, duration of wet and dry spells and temperature means and variances. These changes were then applied to the LARS-WG parameters previously calculated from the observed daily data at each site. For comparison, a corresponding scenario without variability change was also constructed by applying changes in monthly mean precipitation and monthly mean temperature to the LARS-WG parameters. The implications and importance of including changes in inter-daily climate variability in scenarios of climate change was then demonstrated by comparing the effect of scenarios with and without variability on simulated grain yield by using SIRIUS Wheat, a crop-growth simulation model for wheat (Jamieson et al., 1996). Incorporation of inter-daily variability into climate change scenarios should not make any difference to monthly statistics such as, for example, monthly total precipitation or monthly mean temperature. In Table 1 these means are compared for the UKTR scenarios with and without variability for Seville. There is no significant difference between monthly mean temperatures for the scenarios with and without changes in inter-daily climate variability for all months. Results from a t-test indicate that precipitation totals were significantly different for four months out of seven during the vegetation period for winter wheat (January - July). The differences in the totals are most probably due to the way in which the scenarios were constructed, rather than to the weather generator. For three of these months (May, June and July) precipitation for both scenarios was so low that it did not make a big difference to total precipitation over the vegetation period, 184 mm and 210 mm with and without inter-daily variability changes, respectively, compared to 496 mm for the base climate.
54
E C L A T - 2
R E P O R T
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
N O . 2
P O T S D A M
Without variability Precip. Temp. 55.5 12.0 64.5 15.5 34.6 15.6 40.6 17.0 7.8 21.7 7.5 28.6 0.0 31.6 2.3 32.5 3.5 31.2 48.8 24.0 23.6 17.2 43.1 12.2
W O R K S H O P
With variability Precip. Temp. 34.4* 11.9 60.6 15.6 23.7 15.7 38.4 17.3 15.0* 21.9 3.0* 28.0 9.2* 31.5 0.3* 32.5 0.7* 31.6 30.3 24.4 3.5* 16.9 20.4* 12.0
Table 1. Mean monthly total precipitation (mm) and maximum temperature (ºC) at Seville, Spain, for the UKTR scenario without downscaling. Values marked with an asterisk indicate where the hypothesis of equal monthly means was rejected with 95% confidence level.
For the base climate the grain yield simulated by SIRIUS Wheat was 5.6 t/ha and its coefficient of variation (CV) was 0.24 (Table 2). According to the UKTR scenario without variability changes, the grain yield does not change much (5.2 t/ha) and the CV remains about the same (0.23). If changes in climate variability are considered the results are very different. The grain yield drops to 3.9 t/ha and the CV almost doubles to 0.48. The reason for this is not the total amount of precipitation, but the change in precipitation distribution over the vegetation period and the prolonged dry spells. Base Grain yield, t/ha CV of yield Total precipitation January-July, mm Cum. Temperature January-July, ºC
5.6 0.24 296 3630
UKTR
UKTR with variability 5.2 3.9 0.23 0.48 210 184 4293 4323
Table 2. The effect of inter-daily climate variability on crop yield and its coefficient of variation (CV), as simulated by SIRIUS Wheat, for UKTR scenario at Seville, Spain.Total precipitation and cumulative mean temperature were calculated for the winter wheat vegetation period from January to July.
The probability of producing yields less than 3.5 t/ha is almost 50% for the UKTR scenario with variability changes and only about 10% for the UKTR scenario without variability changes or for the baseline climate (Figure 1). The high probability of obtaining low grain yields may make wheat an economically unsuitable crop in Spain under this climate change scenario. A detailed comparison of five wheat models (AFRCWHEAT2, CERES, NWHEAT, SIRIUS and SOILN), including model sensitivity to changes in means and variances of weather variables and model performances for a set of climate change scenarios, are presented in Wolf et al. (1996) and Semenov et al. (1996).
55
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
1.0
cumulative probability
0.8 0.6 0.4 0.2 0.0
0
1
2
3
4
5
6
7
8
yield, t/ha base
UKTR no var
UKTR
Figure 1. Cumulative probability functions of grain yield as simulated by SIRIUS Wheat for the base climate and for the UKTR scenarios with and without changes in climatic variability.
CONCLUSIONS A stochastic weather generator has been used in this climate change study as a computationally inexpensive tool to construct site-specific climate change scenarios which incorporate changes in climate means and climate variability, as indicated by UK Met. Office GCM experiments UKTR and which are suitable for agricultural impact assessment. Site-specific scenarios were produced using regression downscaling techniques, whilst scenarios incorporating changes in variability used only the GCM grid-box changes. The daily time-series for both types of scenario were produced by the LARS-WG stochastic weather generator. The GCM-derived changes were then applied to the parameters of the weather generator for each site and 30 years of daily data generated. This study has demonstrated that the different methods of scenario construction produce significantly different climate change scenarios which, in the case of Seville, imply quite different conclusions concerning the suitability of wheat cultivation in this area of Spain as a result of climate change. The disadvantage of using regression downscaling is that it is rather data intensive; observed data from several sites are required in order to calculate observed areal means and anomalies. Construction of sitespecific scenarios of climate change may be aided by the current development of Regional Climate and High Resolution Limited Area Models (RegCMs and HRLAMs, respectively). This methodology has been developed for climate change studies (Giorgi, 1990; Mearns et al., 1999a,b). The basic idea of the approach is to run a RegCM with a high grid resolution (approximately 50km) but only over a limited area of interest. The RegCM is a physically-based model nested into the GCM and is able to reproduce regional climate in more detail than the GCM itself. However, recent work on the validation of a RegCM has shown that there may be still large differences between model output and observed weather statistics, especially in the case of climate variability (Mearns et al., 1995a,b). This means that the construction of local climate change scenarios from these models may be as problematic as from GCMs. Hence, the need for local stochastic weather generators in climate change studies will remain in the near future.
ACKNOWLEDGEMENTS The UK Met. Office GCM data were provided by the Climate Impacts LINK Project on behalf of the UK Met. Office. This work was partly funded by the Commission of the European Communities’ Environment Programme (Contract EV5V-CT93-0294). IACR receives grant-aided support from the Biotechnology and Biological Sciences Research Council in the UK. 56
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
REFERENCES Bardossy, A. and Plate, E.J. (1991) Modelling daily rainfall using a semi-Markov representation of circulation pattern occurrence, Journal of Hydrology 122, 33-47. Bailey, N.T.J. (1964) The Elements of Stochastic Processes, Wiley: New York. Barrow, E.M., Hulme, M. and Semenov, M.A. (1996) The effect of using different methods in the construction of climate change scenarios: examples from Europe, Climate Research 7,195-211 Gabriel, R. and Neumann, J. (1962) A Markov chain model for daily rainfall occurrence in Tel Aviv, Israel, Quarterly Journal of the Royal Meteorological Society 88, 90-95. Giorgi, F. (1990) Simulation of regional climate using a limited area model nested in a GCM, Journal of Climate 3, 941-963. Giorgi, F. and Mearns, L.O. (1991) Approaches to the simulation of regional climate change: a review, Reviews of Geophysics 29, 191-216. Houghton, J.T., Meira Filho, L.G., Callander, B.A., Kattenberg, A. and Maskell, K. (eds): (1996) Climate change 1995: the science of climate change, Cambridge University Press, Cambridge. Hulme, M., Briffa, K.R., Jones, P.D. and Senior, C.A. (1993) Validation of GCM control simulations using indices of daily airflow types over the British Isles, Climate Dynamics 9, 95-105. Hulme M., Barrow E.M., Arnell N.W., Harrison P.A., Johns T.C., Downing T.E. (1999) Relative impacts of human-induced climate change and natural climate variability, Nature, 397, No.6721, pp.688-691. Hutchinson, M. (1995) Stochastic space-time weather models from ground-based data, Agricultural and Forest Meteorology 73, 237-265. Jamieson, P.D., Semenov, M.A., Brooking, I.R. and Francis, G.S. (1996) Sirius: a mechanistic model of wheat response to environmental variation, European Journal of Agronomy 8, 161-179 Karl, T.R., Wang, W.C., Schlesinger, M.E., Knight, R.W. and Portman, D. (1990) A method of relating general circulation model simulated climate to the observed local climate. Part I: Seasonal statistics, Journal of Climate 3, 1063-1079. Kim, J.W., Chang, J.T., Baker, N.L., Wilks, D.S. and Gates, W.L. (1984) The statistical problem of climate inversion: Determination of the relationship between local and large-scale climate, Monthly Weather Review 112, 2069-2077. Matyasovszky, I., Bogardi, I., Bardossy, A. and Duckstein, L. (1993) Space-time precipitation reflecting climate change, Hydrological Sciences 38, 539-558. Mearns, L.O., Giorgi, F., Mcdaniel, L. and Shields, C. (1995a) Analysis of variability and diurnal range of daily temperature in a nested regional climate model - comparison with observations and doubled CO2 results, Climate Dynamics 11, 193-209. Mearns, L.O., Giorgi, F., Mcdaniel, L. and Shields, C. (1995b) Analysis of daily variability of precipitation in a nested regional climate model - comparison with observations and doubled CO2 results, Global and Planetary Change 10, 55-78. Mearns, L.O., Rosenzweig, C. and Goldberg, R. (1996) The effect of changes in daily and interannual climatic variability on CERES-wheat yields. A sensitivity study, Climatic Change 32, 257-292. Mearns, L.O., Bogardi, I., Giorgi, F., Matyasovszky, I. and Palecki, M. (1999a) Comparison of climate change scenarios generated from regional climate model experiments and statistical downscaling, Journal of Geophysics Research 104, 6603-6621. Mearns, L.O., Mavromatis, T., Tsvetsinskaya, E., Hays, C. and Easterling, W. (1999b) Comparative responses of EPIC and CERES crop models to high and low spatial resolution climate change scenarios, Journal of Geophysics Research 104, 6623-6646. Murphy, J.M. (1995) Transient response of the Hadley Centre coupled ocean-atmosphere model to increasing carbon dioxide. Part I: Control climate and flux adjustment, Journal of Climate 8, 36-56. Murphy, J.M. and Mitchell, J.F.B. (1995) Transient response of the Hadley Centre coupled ocean-atmosphere model to increasing carbon dioxide. Part II: Spatial and temporal structure of the response, Journal of Climate 8, 57-80.
57
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Porter J.R. and Semenov M.A. (1999) Climate variability and crop yields in Europe, Nature 400, No.6746, p.724 Racsko, P., Szeidl, L. and Semenov, M. (1991) A serial approach to local stochastic weather models, Ecological Modelling 57, 27-41. Richardson, C.W. (1981) Stochastic simulation of daily precipitation, temperature and solar radiation, Water Resources Research 17, 182-190. Semenov, M.A. and Porter, J.R. (1995) Climatic variability and the modelling of crop yields, Agricultural and Forest Meteorology 73, 265-283 Semenov, M.A., Wolf, J, Evans, L.G, Eckersten, H. and Iglesias, A. (1996) Comparison of wheat simulation models under climate change. II. Application of climate change scenarios, Climate Research 7, 271-281 Semenov M.A. and Barrow. E.M. (1997) Use of a stochastic weather generator in the development of climate change scenarios, Climatic Change 35, 397-414 Semenov M.A., Brooks R.J., Barrow E.M. and Richardson C.W (1998) Comparison of the WGEN and LARSWG stochastic weather generators in diverse climates, Climate Research 10:95-107. Semenov M.A. and Brooks R.J. (1999) Spatial interpolation of the LARS-WG stochastic weather generator in Great Britain, Climate Research 11,137-148 Viner, D. and Hulme, M. (1994) The Climate Impacts LINK Project: providing climate change scenarios for impacts assessment in the UK, DoE/CRU Report, Norwich. von Storch, H., Zorita, E. and Cubasch, U. (1993) Downscaling of global climate change estimates to regional scales: an application to Iberian rainfall in wintertime, Journal of Climate 6, 1161-1171. Wilby, R.L., Wigley, T.M.L., Conway D., Jones, P.D. Hewitson B.C., Main, J. and Wilks D.S. (1998) Statistical downscaling of general circulation model output: A comparison of methods, Water Resources Research 34, 2995-3008 Wilks, D. (1999) Multisite downscaling of daily precipitation with a stochastic weather generator, Climate Research 11,125-136 Wigley, T.M.L., Jones, P.D., Briffa, K.R. and Smith, G. (1990), Obtaining sub-grid-scale information from coarse-resolution general circulation model output, Journal of Geophysical Research 95, 1943-1953. Wolf, J., Evans, L.G, Semenov, M.A., Eckersten, H. and Iglesias, A. (1996) Comparison of wheat simulation models under climate change. II. Model calibration and sensitivity analysis, Climate Research 7, 253-270 Yevjevich, V. (1972) Structural analysis of hydrological time series, Hydrology Paper 56, Color. State Univ., Fort Collins, 59 pp.
58
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Keynote Paper 4 Incorporating Dynamic Vegetation Cover within Global Climate Models 1, 2,
2, 3,
2, 4,
Jonathan A. Foley Samuel Levis Marcos Heil Costa 5 6 Wolfgang Cramer and David Pollard
This paper was presented as a background paper to the keynote presentation of Navin Ramankutty (
[email protected]). It is currently in press with the journal Ecological Applications. The text is reprinted here with kind permission of the Ecological Society of America. 1 corresponding author 2 Climate, People, and Environment Program (CPEP), Institute for Environmental Studies, University of Wisconsin, 1225 West Dayton Street, Madison, WI 53706, USA 3 now at the National Center for Atmospheric Research (NCAR), P.O. Box 3000, Boulder, CO 80307-3000, USA 4 now at the Departamento de Engenharia Agrícola, Universidade Federal de Viçosa, Av. P. H. Rolfs, s/n, Viçosa - MG - 36.571-000, Brazil 5 Potsdam Institut für Klimafolgenforschung e.V. (PIK), Telegrafenberg, P.O. Box 60 12 03, D -144 12 Potsdam, Germany 6 Earth System Science Center, Penn State University, University Park, PA 16802, USA ABSTRACT Numerical models of the Earth’s climate system must consider the atmosphere and terrestrial biosphere as a coupled system, with biogeophysical and biogeochemical processes occurring across a range of timescales. On short timescales (i.e., seconds to hours), the coupled system is dominated by the rapid biophysical and biogeochemical processes that exchange energy, water, carbon dioxide, and momentum between the atmosphere and the land surface. Intermediate timescale (i.e., days to months) processes include changes in the store of soil moisture, changes in carbon allocation, and vegetation phenology (e.g., budburst, leaf-out, senescence, dormancy). On longer timescales (i.e., seasons, years, and decades), there can be fundamental changes in the vegetation structure itself (disturbance, land use, stand growth). In order to consider the full range of coupled atmosphere-biosphere processes, we must extend climate models to include intermediate and long-term ecological phenomena. This paper reviews early attempts at linking climate and equilibrium vegetation models through iterative coupling techniques, and some important insights gained through this procedure. We then summarize recent developments in coupling global vegetation and climate models, and some of the applications of these tools to modeling climate change. Furthermore, we discuss more recent developments in vegetation models (including a new class of models called Dynamic Global Vegetation Models), and how these models are incorporated with atmospheric general circulation models. Fully coupled climate-vegetation models are still in the very early stages of development. Nevertheless, these prototype models have already indicated the importance of considering vegetation cover as an interactive part of the climate system. Key phrases: atmosphere-biosphere interactions, dynamic global vegetation models Key words: climate, vegetation, coupled models 59
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
1. INTRODUCTION During the last decade, the international scientific community has become increasingly concerned with the interconnections between terrestrial ecosystems and the atmosphere. For example, there is significant interest in how climatic variability and climate change affect the structure and functioning of ecosystems on regional and global scales. Furthermore, there is increasing interest in how changes in terrestrial ecosystems can, in turn, affect the atmosphere. One of the most obvious manifestations of atmosphere-ecosystem interactions is the relationship between global patterns of vegetation cover and climate. The location of deserts, tropical rainforests, and tundra ecosystems, for example, are obviously dictated by climate. In fact, the use of a few basic climate parameters (e.g., growing degree-days, minimum wintertime-temperatures, soil moisture availability) allows the successful prediction of the geographic distribution of many plant functional types on continental and global scales (Box 1981, Woodward 1987). Following this logic, several models of global vegetation patterns, based on the relationships between climate and vegetation, have been developed (e.g., BIOME of Prentice et al. 1992, BIOME-3 of Haxeltine and Prentice 1996, MAPSS of Neilson 1995, DOLY of Woodward et al. 1995). Changes in climate affect the geographic distribution of global vegetation communities in fundamental ways. For example, climatic changes during the late Quaternary and the Holocene drove large-scale changes in global biome distributions (e.g., Prentice et al. 1993, Foley 1994, TEMPO 1996, Claussen and Gayler 1997, Texier et al. 1997, Kutzbach et al. 1998). Moreover, projected changes in future climate associated with an increase in greenhouse gases are expected to be large enough to cause fundamental changes in global vegetation distribution (Solomon and Cramer 1993). Changes in vegetation structure may also significantly influence the climate (Pielke and Avissar 1990). The physical characteristics of vegetation and soils have a strong influence on the exchange of energy, water, and momentum between the land surface and the atmosphere. Changes in vegetation therefore imply changes of the physical properties of the land surface, including surface albedo, surface roughness, leaf area index, rooting depth, and the availability of soil moisture. Consequently, the atmosphere and terrestrial biosphere must be considered as a coupled system, with exchange processes acting on a range of timescales. On short timescales (i.e., seconds to hours), the coupled system is dominated by the rapid biophysical and physiological processes that exchange energy, water, carbon dioxide, and momentum between the atmosphere and the land surface. Intermediate timescale (i.e., days to months) processes include changes in the store of soil moisture, changes in carbon allocation, and vegetation phenology phenomena (e.g., budburst, leaf-out, senescence, dormancy). On longer timescales (i.e., seasons, years, and decades), there can be fundamental changes in the vegetation cover itself (disturbance, land use, stand growth). Most climate models already describe the rapid biophysical processes occurring at the land-atmosphere interface, but they do not consider longer-term ecological phenomena.
2 INFLUENCE OF VEGETATION COVER ON CLIMATE Global climate models, including atmospheric general circulation models (GCMs), require some specification of the fluxes of energy (including radiative and turbulent fluxes), water vapor, and momentum between the lower atmosphere and the underlying surface. Over land, these fluxes are determined generally by biophysical and physiological processes occurring between the soil, plant canopies, and the atmospheric boundary layer. Nearly all GCMs now include land surface parameterization schemes to simulate rapid landatmosphere exchange processes. A large number of land surface models now exist; most of them are based at least in part on the early BATS (Dickinson et al. 1986) and SiB (Sellers et al. 1986, 1996) land surface parameterization schemes. 60
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
GCM land surface models typically operate by prescribing the geographic distribution of vegetation and soil characteristics across the globe. Vegetation and soil properties are taken as prescribed ‘boundary conditions’, which are not allowed to change with the climate, and thus neglect long-term changes in vegetation cover and resultant feedbacks. In studies of short-term climate variability, where changes in vegetation and soils are unlikely to be of major importance in most areas, this is a reasonable approach. However, in long-term, transient climate simulations, including scenarios of future global warming, where changes in climate may drive fundamental changes in vegetation distribution, this approach is no longer inadequate. To demonstrate the importance of vegetation structure, a number of studies have considered the sensitivity of atmospheric processes to changing global vegetation distribution. Below, we present three examples of how changes in vegetation structure in different biomes are likely to influence broad-scale climate.
2.1 Tropical Deforestation The clearing of tropical rainforests for pasture and croplands is one of the most prominent environmental issues of our time. Many authors have discussed the possible effects of tropical deforestation on global climate processes. For example, exploratory GCM modeling studies have considered the sensitivity of the climate system to a complete conversion of Amazonian rainforests to pastures (e.g., Dickinson and Henderson-Sellers 1988, Shukla et al. 1990, Nobre et al. 1991, and Henderson-Sellers et al. 1993). Table 1 compares several recent GCM simulations of Amazonian deforestation. Dickinson & Kennedy (1992) GCM CCM1 Resolution 4.5° x 7.5° Surface model BATS (Dickinson et al., 1986) Ocean mixed layer Simulation length 3 yrs Roughness length 2.00/0.05 Albedo 0.12/0.19
Reference
Henderson- Lean & Sellers et al. Rowntree (1993) (1993) CCM1-OZ UKMO 4.5° x 7.5° 2.5° x 3.75° BATS Warrilow (Dickinson (1986) et al., 1986) mixed prescribed layer SST 6 yrs 3 yrs 2.00/0.20 0.80/0.04 0.12/0.19 0.14/0.19
Polcher & Laval (1994) LMD 2.0° x 5.6° SECHIBA (Ducroudé et al., 1993) prescribed SST 1.1 yr no change 0.098/0.177
Sud Manzi & Lean & et al. Planton Rowntree (1996) (1996) (1997) GLA EMERAUDE UKMO 4.0° x 5.0° 2.8° x 2.8° 2.5° X 3.75° SSiB ISBA (Noilhan Warrilow (Xue et al., & (1986) 1991) Planton 1989) modified prescribed prescribed prescribed SST SST SST 3 yrs 3 yrs 10 yrs 2.65/0.077 2.00/0.026 2.10/0.026 0.092/0.142 0.12/0.163 0.13/0.18
Costa & Foley in press) GENESIS 4.5° x 7.5° IBIS (Foley et al., 1996) mixed layer 15 yrs interactive interactive
Table 1 – Summary of recent climate model simulations of Amazonian deforestation (from Costa and Foley, in press).
Replacing a tropical rainforest with a pasture typically increases the surface albedo, lowers the surface roughness, and reduces the leaf area index (and associated canopy interception) and available soil moisture (mainly because pasture plants often have shallower roots than rainforest trees). As a consequence, tropical deforestation is expected to lower the ability of the land surface to maintain a high rate of evapotranspiration throughout the year, leading to changes in the latent heating of the atmospheric boundary layer and the strength of tropical convection. Generally speaking, this change in the surface energy and water balance (and the associated changes to the atmospheric boundary layer) leads to a significant reduction in rainfall and an increase in surface temperature. Costa and Foley (2000) recently conducted a series of GCM sensitivity studies to compare the potential effects of large-scale deforestation and CO2-induced global warming on the climate of the Amazon basin. In
61
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
these simulations, deforestation caused a significant decrease in regional precipitation, in association with a general decrease in latent heating and vertical motion over the deforested area. On the other hand, the overall effects of increased CO2 concentrations (including both the CO2 radiative forcing and the physiological effects of increased CO2 on stomatal conductance) caused an increase in precipitation. Together, the combined effects of deforestation and doubled CO2, including the interactions among the -1 processes, caused a precipitation decrease of approximately 0.4 mm day (Figure 1). While the effects of deforestation and increasing CO2 concentrations on precipitation tended to counteract each other, both processes worked to warm the Amazon basin. Deforestation caused an increase in surface temperature, largely because of decreases in evapotranspiration. In addition, the physiological effects of CO2 also tended to increase the surface temperature, associated with the decrease in canopy conductance and transpiration. Finally, the radiative effect of CO2 also increased the surface temperature. The combined effect of deforestation and doubling CO2 concentrations, including the interactions among the processes, was a temperature increase of roughly 3.5 °C. 0 -1 0
9°N -2
0° 1
9°S
1
9°S
0
1 2
0 1
9°S 75°
0°
0°
January/February/March
-4
60° -3
75°
45° -2
9°N
-1 0
0°
July/August/September
-4 -1
9°S
(b) (a) Response of Amazonian Precipitation to Deforestation
January/February/March
0
(a)
1
0
9°N
9°N
60°
45° 3
3
2
1 0
July/August/September
0
(b) Response of Amazonian Precipitation to Global Warming
Figure 1. Potential Climatic Impact of Global Warming and Amazonian Deforestation. Costa and Foley(2000) used the GENESIS Global Climate Model to compare the impact of tropical deforestation and global warming on the climate of Amazonia. (a) Changes in precipitation resulting from the complete deforestation of Amazonia alone. (b) Changes in precipitation resulting from a doubling of atmospheric CO2 concentrations alone. -1 Dashed contour lines indicate a decrease of precipitation. Units are in mm day . (Adapted from Costa and Foley, 2000).
Climate model sensitivity studies have clearly established the importance of tropical forests in influencing the Earth’s climate. According to the current generation of climate models, a complete deforestation of tropical land-masses could have global climatic significance. However, future studies should concentrate on more realistic scenarios of land use and deforestation, rather than conducting only total-deforestation simulations (Chu et al. 1994). Furthermore, other modes of land use (including mid-latitude deforestation and agricultural expansion) should be further evaluated for their climatic significance (e.g., Pielke et al. 1991, Chase et al. 1996, Bonan 1997).
2.2 Boreal Forests and Tundra The boundary between boreal forests and Arctic tundra represents a dramatic change in land surface properties, with a known history of climate-driven fluctuations (Huntley and Cramer 1997). For example, the albedo (or reflectivity) of boreal forests and tundra is markedly different, especially in spring when the tundra is covered by snow (Robinson and Kukla 1985, Laine and Heikinheimo 1996, Sharratt 1998). Furthermore, boreal forest and tundra have significantly different surface roughness and surface emissivity (Bonan et al. 1995, Kurvonen et al. 1998, Levis et al. 1999). The differences in the physical properties of boreal forests and tundra ecosystems may have great significance to the climate system. For example, an increase in forest area (at the expense of tundra) would 62
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
result in a decreased albedo, thereby warming the land surface with additional absorbed solar radiation. The reverse, caused by an expansion of tundra replacing boreal forest, would therefore act to cool the surface through albedo increases. Across the high northern latitudes, the relative amount of boreal forest and tundra could therefore have a significant impact on the overall energy balance, temperature, and general circulation of the atmosphere. On smaller spatial scales, Pielke and Vidale (1995) noted that the forest/tundra boundary could have a significant impact on local atmospheric circulations, which may lead to the enhancement of frontal boundaries along the forest/tundra ecotone. Using the GENESIS GCM (Thompson and Pollard 1995, 1997), Bonan et al. (1992) examined the potential impact of complete boreal deforestation on the climate system, finding that temperatures were dramatically cooler year-round. While this study uses an extreme and unlikely deforestation scenario, it does point to the significance of vegetation cover in affecting high latitude climates. Foley et al. (1994) therefore considered how changes in the boreal forest – tundra boundary during the early and middle Holocene (roughly 5,000 to 10,000 years before present) might have affected climate. During this time, northern high latitude regions were substantially warmer than present (TEMPO 1996), presumably in response to increases in incoming solar radiation during summer that resulted from changes in Earth’s orbital geometry (Kutzbach et al. 1998). Paleobotanical evidence indicates that the boundary between boreal forest and tundra was at least in some high-latitude locations, farther north than present (TEMPO 1996). Foley et al. (1994) showed that the poleward movement of the boreal forest – tundra ecotone might have greatly amplified the orbitally-induced warming of the high northern latitudes during the middle Holocene. They found that the high-latitude warming caused by orbital forcing was nearly doubled by the poleward expansion of boreal forests (Figure 2). This vegetation feedback mechanism acted to amplify the initial warming (induced by changes in Earth’s orbit) through changes in the surface albedo associated with boreal forest and tundra.
(b) additional warming due to vegetation feedback
(a) initial warming due to orbital forcing
Figure 2.Vegetation Feedbacks on High Latitude Warming During the Middle Holocene. Foley et al. (1994) used the GENESIS global climate model to examine how changes in the border of boreal forests and tundra may have affected the mid-Holocene climate. (a) Annually-averaged simulated mid-Holocene temperature changes (compared to modern) associated with changes in Earth’s orbital geometry. (b) The additional simulated warming associated with the northward movement of boreal forests, replacing tundra, during the middle Holocene. (Adapted from Foley et al., 1994).
These and other studies show that changes in the boundary between boreal forests and tundra may lead to important climatic feedbacks. Considering that future warming may be strongest in the high latitudes, we need to further evaluate the possibility of high-latitude vegetation feedbacks and not neglect them in the assessment of global warming scenarios.
63
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
2.3 Sahara / Sahel Vegetation Transition The regional differences in albedo, surface roughness, leaf area index, and soil moisture holding capacity, between the Sahara and the vegetation of the Sahel zone have a significant impact on surface energy balance, and the consequent rate of atmospheric heating. Furthermore, the shrublands and grasslands to the south can maintain a higher rate of evapotranspiration than deserts, and therefore have a greater ability to re-circulate precipitation (and, therefore, adding latent heat) back into the atmosphere. Changes in the vegetation cover of northern Africa, whether due to climate change or human activity, are therefore likely to have profound impacts on the atmosphere. For example, Charney (1975) suggested that human-induced desertification in the Sahel could reduce the latent heating and moisture recirculation to the atmosphere, and partially explain increasing drought conditions occurring in the region during the 1970s. Recent studies by Zheng and Eltahir (1997, 1998) indicated that deforestation along the coastal tropical forests of West Africa might also contribute to reduced rainfall over the Sahel. However, the importance of land surface feedbacks in changing Sahelian rainfall, relative to the possible influence of changes in ocean conditions (Eltahir and Gong 1996), is still being debated. For the recent geologic record, several authors have discussed how changes in the vegetation cover over the Sahara might have affected climate during the early to middle Holocene period (Kutzbach et al. 1996, Broströmet al. 1997, Claussen and Gayler 1997, Broström et al. 1998, Pollard et al. 1998, Claussen et al. 1999). Changes in the Earth’s orbit during this time increased the seasonal variation of incoming solar radiation in the northern hemisphere (compared to present day), leading to an increase in the heating (in summer) and cooling (in winter) of continental land masses. The oceans, because of their large heat capacity, show relatively little change in temperature during the seasonal cycle. As a result, increases in the seasonal cycle of land temperatures (compared to the ocean) enhanced the strength of subtropical monsoon circulations in Africa and south Asia, relative to the present-day (Kutzbach et al. 1998). Paleobotanical data from northern Africa demonstrates that much of the modern desert was covered with grasslands, savannas, and lakes until approximately 5000-6000 years before present (e.g., Ritchie and Haynes 1987, COHMAP 1988; Lézine, 1989; Hoelzmann et al. 1998; Jolly et al. 1998a, 1998b). These large changes in vegetation structure must have dramatically altered patterns of atmospheric heating, and also led to increased re-circulation of rainfall back to the atmosphere (through increases in evapotranspiration), which may have been a significant feedback on the African monsoon. Kutzbach et al. (1996) and Broström et al. (1997) conducted simple GCM sensitivity studies, showing that changing vegetation about 6000 years before present (increasing grassland and shrubland area, at the expense of desert) led to a significant enhancement to the monsoon circulation and associated rainfall.
3 LINKING CLIMATE AND EQUILIBRIUM VEGETATION MODELS While the influence of vegetation cover on climate can be documented on a case-by-case basis with existing GCMs, there is a need for a more generalized framework for exploring climate-vegetation interactions across the globe. In particular, climate models need to consider vegetation cover as an interactive surface boundary, which can change in response to changes in climate. Interactions take place regarding fluxes of energy and momentum (physical feedbacks) and of matter, mostly carbon and water (biogeochemical feedbacks).
64
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
3.1 Physical feedbacks The first attempt to link climate and vegetation models to address physical feedbacks was made by Henderson-Sellers (1993), using the CCM1-Oz climate model together with the Holdridge (1947) bioclimatic scheme. The Holdridge scheme uses simple delineations of annual temperature and rainfall to predict the geographic distribution of vegetation types in relation to climate. In this study, the Holdridge scheme is applied at the end of each year to update the geographic pattern of vegetation types within the GCM. Hence, the coupling procedure iterates between the climate and vegetation models (Figure 3), where a single year of the climate simulation is used to predict changes in equilibrium vegetation cover. The changes in vegetation cover are, in turn, used by the GCM to simulate the next year of climate. This asynchronous coupling cycle is repeated until the results of the two models converge to equilibrium.
ATMOSPHERE (GENESIS R15 GCM)
Coupled GENESIS-IBIS Model
canopy physics
VEGETATION DYNAMICS MODULE
energy water aerobalance balance dynamics
soil physics
energy water balance balance
plant physiology
photosynthesis stomatal & leaf respiration conductance
temperature, photosynthesis
LAND SURFACE MODULE (BASED ON LSX LAND SURFACE PACKAGE)
gross primary total net primary production respiration production
allocation growth of leaves, mortality & and turnover stems & roots disturbance
canopy nitrogen allocation
daily LAI
VEGETATION PHENOLOGY MODULE budburst senescence
timescale: minutes to hours
gross photosynthesis, foliage respiration, & optimal C:N ratios for each plant functional type
dormancy
timescale: days to weeks
vegetation structure and biomass
timescale: years
Figure 3. Coupling of the IBIS Dynamic Global Vegetation Model to the GENESIS Global Climate Model. (Adapted from Foley et al., 1998).
Since asynchronous coupling may introduce more interannual variability into the simulated vegetation structure than actual buffering of the biosphere permits, a somewhat more elaborate approach was put forth by Claussen (1994), wherein he iteratively coupled the ECHAM GCM with the BIOME equilibrium vegetation model of Prentice et al. (1992). Instead of individual years, Claussen used multi-year averages of the climate model simulation to drive changes in vegetation. This coupled model was used to study the sensitivity of climate and vegetation to the choice of initial conditions. Specifically, he examined how changes in the initial distribution of deserts and tropical forests could lead to different equilibrium states. In follow-on simulations, Claussen (1997, 1998) and Claussen and Gayler (1997) noted how the Western Sahara region could, under certain conditions, maintain two alternative stable equilibria: one as a desert, another as a grassland. Several other investigators have employed the iterative, asynchronous coupling of GCMs and equilibrium vegetation models. For example, de Noblet et al. (1996) linked the LMD GCM to the BIOME equilibrium vegetation model to study how changes in boreal forest and tundra may have helped initiate glacial conditions during the Quaternary. Texier et al. (1997) showed that climate-induced changes in the vegetation of northern Africa influenced the African monsoon during the middle Holocene epoch. De Conto et al. (in press) used the GENESIS GCM, linked to the EVE equilibrium vegetation model, to examine high-latitude climate-vegetation interactions of the Cretaceous (~ 80 million years BP). 65
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
3.2 Biogeochemical feedbacks All of the above-cited studies exclusively addressed physical feedback processes caused by changing vegetation structure. It has only recently become possible to also study the exchange of matter between atmosphere and biosphere, particularly the flux of carbon as plant nutrient and as powerful greenhouse gas. This development has become possible through the development of biogeochemical models of the land biosphere—these models simulate the flux of carbon and water through plants and soil using physiological process models (the breadth of these models was recently reviewed by Cramer et al. 1999). Off-line calculations have illustrated the sensitivity of biospheric carbon fluxes to the state of the atmosphere for some time already, the most direct factor being the reduction of stomatal conductance which is assumed to take place in many plants at higher CO2 levels. Cao and Woodward (1998) showed that terrestrial ecosystems, when influenced by both changing climate and increased atmospheric CO2, might take up additional carbon for some time—an important element in the calculation of future radiative forcing scenarios, since clearly these have to be adjusted by the (nonlinear) temporal evolution of terrestrial carbon uptake. Using an equilibrium vegetation model that contains physiological as well as structural responses, Betts et al. (1997) evaluated the potential for vegetation feedbacks on CO2-induced future climate change. They found that the feedbacks due to structural changes (i.e., the physical feedbacks) might be significantly affected by changes in physiology (e.g., changes in stomatal conductance), but the overall vegetation feedbacks remain considerable.
4 LINKING CLIMATE AND DYNAMIC GLOBAL VEGETATION MODELS The exploratory modeling studies discussed above all demonstrate the need to incorporate representations of changing vegetation within global climate models. Nevertheless, the technique of iteratively coupling dynamic climate models to equilibrium vegetation models poses two fundamental problems. First, there are differences in how climate and vegetation models consider the surface energy and water balance. GCM land surface models are typically based upon more detailed biophysical parameterizations than the water balance treatments of equilibrium vegetation models. As a result, the behavior of these linked climatevegetation model systems is not necessarily physically consistent, and does not guarantee that water and energy are properly conserved. Secondly, since structural and physiological vegetation dynamics (disturbance, land use, stand growth) are not considered, the use of equilibrium vegetation models prohibits realistic considerations of long-term climate variability or transient climate change. To overcome this limitation, a new generation of ecological models, termed Dynamic Global Vegetation Models (DGVMs), is under development (e.g., Steffen et al. 1992, Walker 1994, Foley et al. 1996, Beerling et al. 1997, Friend et al. 1997). Conceptually similar to earlier generations of forest dynamics models (Shugart 1984), these models are designed to simulate transient changes in vegetation cover (and the associated ecosystem processes) using considerations of both physiological and plant population processes. Plant growth, resulting from the assimilation of carbon through photosynthesis minus the respiration of the plants, contributes to the build-up of a canopy, which is moderated by competition-related mortality. Carbon pools in live and dead biomass, and in the soil, are continuously updated and provide the ‘memory’ of the state of the system over a series of years to decades. In several DGVMs, the canopy may be affected by disturbance such as fire or wind-throw. Implicitly, a DGVM can account for land use as an additional disturbance, although this process needs to be driven by additional forcing (such as a map of contemporary agricultural activity). For the study of interactions between the atmosphere and biosphere, DGVMs may therefore be expected to reliably simulate transient changes in both climate and vegetation, as well as their interactions, rather than instantaneous jumps in global vegetation patterns. So far, only exploratory modeling studies have linked 66
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
DGVMs directly with GCMs. For example, Foley et al. (1998) incorporated the IBIS DGVM (Foley et al. 1996) within the GENESIS atmospheric GCM of Thompson and Pollard (1997), using a R15 spectral grid (spatial resolution of approximately 4.5° latitude by 7.5° longitude). The two models exchanged information through a common land surface model with a 45 minute interval (‘synchronous coupling’, Figure 3). For this study, the coupled model was exercised for current climate conditions, and it captured the broad, globalscale patterns of climate and vegetation fairly well. However, because of biases in the GCM climate simulation, there were a number of regional-scale errors in the simulation. For example, the coupled model correctly simulated the general placement of most biomes, but there was too little boreal forest in North America (associated with a dry bias in the climate model) and too much grassland in northern Africa (associated with a wet bias in the climate model). Clearly, the development of fully coupled climate-vegetation models is still in the early stages. Moreover, model testing and evaluation becomes a problem with new dimensions, since spatially-explicit dynamics of the land biosphere have not been observed over long time-scales. Below, we review two preliminary studies that have been conducted with this new generation of coupled climate-vegetation models.
4.1 High Latitude Vegetation Feedbacks on Global Warming Many simulations of global warming driven by fossil fuel emissions show the greatest warming in northern high-latitude regions, where positive albedo feedbacks (resulting from large decreases in high-latitude snow and sea-ice cover) are acting to amplify the initial CO2-induced warming (Kattenberg et al. 1996). Levis et al. (1999) used the coupled GENESIS/IBIS model to examine the potential for climatic feedbacks resulting from changes in vegetation cover, due to both physical and biogeochemical feedbacks. By comparing ‘interactive vegetation’ and ‘fixed vegetation’ global warming simulations, they found that there are several climatic feedbacks introduced by changes in vegetation cover in response to CO2-induced warming and CO2–induced changes in plant physiology. The most pronounced of these vegetation feedbacks resulted from the northward expansion of boreal forests into tundra. Replacing tundra with boreal forest greatly lowered the surface albedo, especially in spring when tundra becomes snow covered. This vegetation feedback mechanism therefore acted to amplify the initial CO2-induced warming, primarily in spring and summer. In particular, the northern high-latitude land surfaces are warmed between 3.5 and 5.0°C in spring and summer by a doubling of CO2 alone; the vegetation feedback produced an additional warming of approximately 1.0 to 2.5°C (Figure 4a). The additional vegetation-induced warming is concentrated over the high-latitude land masses, particularly Eurasia (Figure 4b). Ice free land points from 45 to 90°N 7
2xCO2 R Control 2xCO2 RPV Control
6
Temperature Difference (°C)
5
4
3
2
1
0
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Ann
Figure 4a. Potential Vegetation Feedbacks on Global Warming. Levis et al. (1999) simulated changes in temperature resulting from CO2-induced global warming (R-Control), and the changes in temperature resulting from global warming and vegetation feedbacks (RPV-Control). Temperatures are reported over ice-free land in the northern high latitudes (between 45 and 90 °N) from the lowest GCM level (~50 m above the surface). Statistically significant changes of RPV-R (at the 95% confidence level, using a student’s t-test) are shown with open circles.
67
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
March
April
W O R K S H O P
May
15 N 30 N 45 N 60 N
75 N
90 N
Temperature: RPV R (°C) 5
4
3
2
1
0
1
2
3
4
5
Figure 4b. Potential Vegetation Feedbacks on Global Warming in Springtime. Levis et al. (1999) simulated the additional warming (during the months of March, April, and May) that results from vegetation feedbacks on CO2 - induced global warming in the high latitudes. Statistically significant changes of RPV-R (at the 95% confidence level, using a student’s t-test) are surrounded by dashed lines.
These experiments corroborate the hypothesis suggested by Bonan et al. (1992) and Foley et al. (1994), where changes in the location of the boreal forest - tundra boundary were shown to have significant climatic impacts, using a model with a considerably higher level of physical and biological realism.
4.2 Vegetation Feedbacks on the Climate of the Last Glacial Maximum There is ample evidence showing that the Earth’s vegetation distribution was substantially different during the Last Glacial Maximum (LGM), nearly 21,000 years before present. Vegetation patterns were significantly altered not only by the presence of large continental ice sheets, but also due to changes in climate and sea level; however, the details are still debated. In particular, several modelling studies (e.g., Friedlingstein et al. 1992, Prentice et al. 1993, Crowley 1995, Jolly and Haxeltine 1997, Kubatzki and Claussen 1998) have suggested that tropical rainforests were greatly diminished during the LGM. Many field studies have also suggested that the modern tropical rainforests may have been largely replaced by drought resistant trees and grasses (van der Hammen and Absy 1994, van de Kaars and Dam 1997, Flenley, 1998). Nevertheless, other field studies have confirmed that a core of tropical rainforest remained at the LGM (Salo 1987, Colinvaux 1989, Colinvaux et al. 1996). Crowley and Baum (1997) conducted a series of GCM sensitivity studies to determine if the climate of the LGM was sensitive to changes in tropical forest cover. They found that a reduction in tropical rainforest area had a significant impact on climate. Specifically, tropical land temperatures increased significantly (by as much as 2-4°C) and rainfall greatly decreased (by 10-35%). Crowley and Baum pointed out that these results were somewhat analogous to simulations of modern-day tropical deforestation. Levis et al. (in press) used the GENESIS/IBIS model to examine the potential for vegetation feedbacks on the LGM climate, including those arising from the fact that the LGM had significantly lower CO2 concentrations (~200 ppm) than present. It is likely that lowered CO2 concentrations would have had a significant impact on stomatal conductance, and the relative competitive balance between C3 (e.g., tropical evergreen trees) and C4 (e.g., tropical grasses) plants. Levis et al. (in press) found a dramatic reduction in rainforest area in the tropical latitudes. Sensitivity studies with the model indicated that the lowering of CO2 has a more dramatic effect on tropical rainforest than the change in climate alone, a result that was suggested previously by Jolly and 68
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Haxeltine (1997). There was a 60% reduction in forest LAI, with an associated increase in tropical grasslands and savannas. As suggested by Crowley and Baum (1997), Levis et al. found that these changes in vegetation cover have a strong effect on climate, such as lowering precipitation in the tropical forests, particularly in -1 Amazonia and Africa, by nearly an additional 3% (0.1 mm day ) (Figures 5a and b). Land points from 22.5°S to 22.5°N 0.2
21 ka R Control 21 ka RPV Control
0
Precipitation Difference (mm day 1)
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Ann
Figure 5a. Potential Vegetation Feedbacks on Last Glacial Maximum (LGM) Tropical Precipitation. Levis et al. (in press) simulated changes in precipitation during the LGM (R-Control), and the changes in precipitation resulting considering vegetation feedbacks (RPV-Control). Precipitation is reported over land between 22.5 °S and 22.5 °N. Statistically significant changes of RPV-R (at the 95% confidence level, using a student’s t-test) are shown with open circles.
Annual Average 90 N 75 N 60 N 45 N 30 N 15 N 0 15 S 30 S 45 S 60 S 75 S 90 S
Precipitation (21 ka BP): RPV - R (mm day - 1) 1.5
1
0.5
0
0.5
1
1.5
Figure 5b. Potential Vegetation Feedbacks on Tropical Precipitation during the Last Glacial Maximum (LGM). Levis et al. (in press) simulated the additional changes in global precipitation patterns (averaged over the entire year) resulting from vegetation feedbacks on LGM climate. Statistically significant changes of RPV-R (at the 95% confidence level, using a student’s t-test) are surrounded by dashed lines.
These results indicate that there may be a significant vegetation feedback on LGM climates, but that it is mainly due to the physiological effect of lowering of atmospheric CO2 concentrations. However, this must still be considered preliminary as long as the fossil evidence cannot unequivocally determine the condition of tropical forests at the LGM.
69
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
5 CONCLUSIONS Our understanding of the interactions between the atmosphere and terrestrial biosphere has advanced during recent years. Modelling studies are increasingly enhanced by the availability of observations from a range of sources, and historical reconstructions with unprecedented accuracy. For example, high-resolution global land cover maps from satellites are now available, which greatly improves our ability to characterize the nature of land surfaces across the globe. In addition, international field campaigns and observing networks of terrestrial ecosystem processes have been (e.g., ABRACOS, FIFE, BOREAS) or are being (e.g., LBA, AmeriFlux, EuroFlux, FLUXNET) formed. Large-scale ecosystem manipulation experiments, testing the response of whole plant communities to increasing CO2 concentrations and warmer temperatures, are underway across the globe. Finally, the wealth of paleoecological information is now being synthesized and organized in a manner which allows the use of such data in comprehensive broad-scale studies (e.g., BIOME 6000, Prentice and Webb 1998). Together, these developments have come in parallel with, and supported, the considerable advances in computer simulation models in ecology, particularly in examining globalscale ecological phenomena and their coupling to the atmosphere. Here we have reported how a prototype coupled GCM-DGVM has given us insight into the interactions between climate and vegetation across a wide range of climatic regimes, ranging from the Last Glacial Maximum to a global warming scenario. To address feedbacks between structural change in vegetation and climate, coupled climate-vegetation models must be run for multiple decades and centuries, sometimes at high resolution, which poses limitations due to computational costs. More efficient Earth System models (EMICs – Earth System Models of Intermediate Complexity) are now being developed that allow coupling of atmospheric, vegetation, ocean and ice models with coarser resolution. For example, Ganopolski et al. (1998a, b) showed that CLIMBER-2, a synchronously coupled ocean-atmosphere-vegetation model of intermediate complexity, could successfully simulate modern, mid-Holocene and glacial climates, as well as many of the associated features of the biosphere. Specifically, they noted that the time evolution of climate and vegetation changes from the LGM to the present (which cannot presently be simulated with any comprehensive coupled GCMDGVM) could be simulated with many critical aspects of its nonlinear behavior. The advent of such computationally efficient climate models, as alternatives to traditional GCMs, will allow for a more evenhanded treatment of the atmosphere, ocean, and terrestrial biosphere, and their respective roles in the Earth system. Future developments in atmosphere-biosphere modeling will have to consider the dynamics of the global carbon cycle more comprehensively than earlier. Global patterns of primary productivity and terrestrial carbon storage have been shown to be tightly linked to climate (Melillo et al. 1993, Foley 1994a). Changes in climate could therefore have profound effects on the ability of the terrestrial biosphere to store carbon, and to maintain the current biospheric sink of anthropogenic carbon. Recent modeling studies (e.g., Cao and Woodward 1998) suggest that global warming is likely to alter the carbon balance of the terrestrial biosphere. In order to better evaluate the potential for greenhouse warming, we must develop an Earth system model that simultaneously considers the coupled dynamics of the physical climate system and the global carbon cycle.
70
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
REFERENCES Beerling, D.J., Woodward, F.I., Lomas, M. and Jenkins, A.J. (1997) Testing the responses of a dynamic global vegetation model to environmental change: a comparison of observations and predictions, Global Ecology and Biogeography Letters 6, 439-450. Betts, R.A., Cox, P.M., Lee, S.E. and Woodward, F.I. (1997) Contrasting physiological and structural vegetation feedbacks in climate change simulations, Nature 387, 796-799. Bonan, G.B., Pollard, D. and Thompson, S.L. (1992) Effects of boreal forest vegetation on global climate, Nature 359, 716-718. Bonan, G.B., Chapin, F.S. and Thompson, S.L.(1995) Boreal forest and tundra ecosystems as components of the climate system, Climatic Change 29,145-167. Bonan, G.B. (1997) Effects of land use on the climate of the United States, Climatic Change 37, 449-486. Box, E. O. (1981) Macroclimate and Plant Forms: An Introduction to Predictive Modeling in Phytogeography. The Hague, Dr. W. Junk Publishers. Broström A., Coe M., Harrison S.P., Gallimore R., Kutzbach J.E., Foley J.A., Prentice I.C. and Behling, P. (1997) Land surface feedbacks and paleomonsoons in northern Africa, Geophysical Research Letters 25, 3615-3618. Cao, M. and Woodward, F.I. (1998) Dynamic responses of terrestrial ecosystem carbon cycling to global climate change, Nature 393, 249-252. Charney, J.G. (1975) Dynamics of deserts and drought in the Sahel, Quarterly Journal of the Royal Meteorological Society 101, 193-202. Chase, T.N., Pielke, R.A., Kittel, T.G.F., Nemani, R. and Running, S.W. (1996), Sensitivity of a general circulation model to global changes in leaf area index, Journal of Geophysical Research 101, 7393-7408. Chu, P.-S., Yu, Z.-P. and Hastenrath, S. (1994) Detecting climate change concurrent with deforestation in the Amazon Basin: Which way has it gone? Bulletin of the American Meteorological Society 75, 579-583. Claussen, M. (1994) Coupling global biome models with climate models, Climate Research 4, 203-221. Claussen, M. (1997) Modeling bio-geophysical feedback in the African and Indian monsoon region, Climate Dynamics 13, 247-257. Claussen, M. and V. Gayler. (1997) The greening of the Sahara during the mid-Holocene: results of an interactive atmosphere-biome model, Global Ecology and Biogeography Letters 6, 369-377. Claussen, M. (1998) On multiple solutions of the atmosphere-vegetation system in present-day climate, Global Change Biology 4, 549-559. Claussen M., Kubatzki C., Brovkin V., Ganopolski A., Hoelzmann P. and Pachur, H.-J. (1999) Simulation of an abrupt change in Sahara vegetation in the mid-Holocene, Geophysical Research Letters 26, 2037-2040. COHMAP Authors. (1988) Climatic changes of the last 18,000 years: Observations and model simulations, Science 241, 1043-1052 Colinvaux, P.A. (1989) Ice-age Amazon revisited, Nature 340, 188-189. Colinvaux, P.A., De Oliveira, P.E., Moreno, J.E., Miller, M.C. and Bush, M.B (1996) A long pollen record from lowland Amazonia: Forest and cooling in glacial times, Science 274, 85-88. Costa, M.H. and Foley, J.A. (2000) Combined effects of deforestation and doubled atmospheric CO2 concentrations on the climate of Amazonia, Journal of Climate 13(1), 18-34. Cramer W., Kicklighter D.W., Bondeau A., Moore B., III, Churkina G., Nemry B., Ruimy A., Schloss A.L. and participants of the "Potsdam NPP Model Intercomparison" (1999) Comparing global models of terrestrial net primary productivity (NPP): Overview and key results, Global Change Biology 5 (Suppl. 1), 1-15 Crowley, T.J. (1995) Ice age terrestrial carbon changes revisited, Global Biogeochemical Cycles 9, 377-389. Crowley, T.J. and Baum, S.K. (1997) Effect of vegetation on an ice-age climate model simulation, Journal of Geophysical Research 102(D14), 16463-16480. De Conto, R.M., Thompson S.L. and Pollard, D. (in press) Recent advances in paleoclimate modeling: toward better simulations of warm paleoclimates. In Warm Climates in Earth History, Huber B.T., Wing, W. and Macleod, K. (eds.), Cambridge University Press. 71
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
de Noblet, N.I., Prentice, I.C., Joussaume, S., Texier, D., Botta, A. and Haxeltine, A. (1996) Possible role of atmosphere-biosphere interactions in triggering the last glaciation, Geophysical Research Letters 23, 3191-3194. Dickinson, R.E., Henderson-Sellers, A., Kennedy, P.J. and Wilson, M.F. (1986) Biosphere-Atmosphere Transfer Scheme (BATS) for the NCAR CCM, NCAR/TN-275-STR, National Center for Atmospheric Research, Boulder CO. Dickinson, R.E. and Henderson-Sellers, A. (1988) Modeling tropical deforestation: A study of GCM landsurface parameterizations, Quarterly Journal of the Royal Meteorological Society 114, 439-462. Dickinson, R. E. and Kennedy, P.J. (1992) Impacts on regional climate of amazon deforestation, Geophysical Research Letters 19, 1947-1950. Eltahir, E.A.B. and Gong, C. (1996) Dynamics of wet and dry years in West Africa, Journal of Climate 9, 1030-1042. Flenley, J.R. (1998) Tropical forests under the climate of the last 30000 years, Climatic Change 39, 177-197. Foley, J.A., Kutzbach, J.E., Coe, M.T. and Levis, S. (1994) Feedbacks between climate and boreal forests during the Holocene epoch, Nature 371, 52-54. Foley, J.A., Prentice, I.C., Ramankutty, N., Levis S., Pollard, D., Sitch, S. and Haxeltine, A. (1996) An integrated biosphere model of land surface processes, terrestrial carbon balance, and vegetation dynamics, Global Biogeochemical Cycles 10, 603-628. Foley, J.A., Levis, S., Prentice, I.C., Pollard, D. and Thompson, S.L. (1998) Coupling dynamic models of climate and vegetation, Global Change Biology 4,561-579. Friedlingstein, P., Delire, C., Müller, J.F. and Gérard, J.-C. (1992) The climate induced variation of the continental biosphere: a model simulation of the last glacial maximum, Geophysical Research Letters 19, 897-900. Friend, A.D., Stevens, A.K., Knox, R.G. and Cannell, M.G.R. (1997) A process-based, terrestrial biosphere model of ecosystem dynamics (Hybrid v3.0), Ecological Modelling 95, 249-287. Ganopolski, A., Rahmstorf, S., Petoukhov, V. and Claussen, M. (1998a) Simulation of modern and glacial climates with a coupled global model of intermediate complexity, Nature 391, 351-356. Ganopolski, A., Kubatzki, C., Claussen, M., Brovkin, V. and Petoukhov, V. (1998b) The influence of vegetationatmosphere-ocean interaction on climate during the mid-Holocene, Science 280, 1916-1919. Haxeltine, A. and Prentice, I.C. (1996) BIOME3: An equilibrium terrestrial biosphere model based on ecophysiological constraints, resource availability and competition among plant functional types, Global Biogeochemical Cycles 10, 693-709. Henderson-Sellers, A. (1993) Continental vegetation as a dynamic component of a global climate model: a preliminary assessment, Climatic Change 23, 337-377. Henderson-Sellers, A., Dickinson, R.E., Durbridge, T.B., Kennedy, P.J., McGuffie, K. and Pitman, A.J. (1993) Tropical deforestation: Modeling local to regional-scale climate change, Journal of Geophysical Research 98(D4), 7289-7315. Hoelzmann, P., Jolly, D., Harrison, S.P., Laarif, F., Bonnefille, R. and Pachur, H.-J. (1998) Mid-holocene landsurface conditions in northern Africa and the Arabian peninsula: a data set for AGCM simulations, Global Biogeochemical Cycles 12, 35-52. Holdridge, L.R. (1947) Determination of world plant formations from simple climatic data, Science 105, 367-368. Huntley, B. and Cramer, W. (1997) Arctic ecosystems and environmental change: Perceptions from the past and predictions for the future. In Disturbance and Recovery in Arctic Lands: An Ecological Perspective Crawford, R.M.M. (ed), Dordrecht, The Netherlands, Kluwer Academic Publishers, pp 1-24. Jolly, D., and Haxeltine, A. (1997) Effect of low glacial atmospheric CO2 on tropical African montane vegetation, Science 276, 786-788. Jolly, D., Harrison, S.P., Damnati, B. and Bonnefille, R. (1998a) Simulated climate and biomes of Africa during the Late Quaternary: comparison with pollen and lake status data, Quaternary Science Reviews 17, 629-658.
72
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Jolly, D., Prentice, I.C., Bonnefille, R., Ballouche, A., Bengo, M., Brenac, P., Buchet, G., Burney, D., Cazet, J.P., Cheddadi, R., Edorh, T., Elenga, H., Elmoutaki, S., Guiot, J., Laarif, F., Lamb, H., Lézine, A.-M., Maley, J., Mbenza, M., Petron, O., Reille, M., Reynaud-Farrera, I., Riollet, G., Ritchie, J.C., Roche, E., Scott, L., Ssemmanda, I., Straka, H., Umer, M., Van Campo, E., Vilimumbalo, S., Vincens, A. and Waller, M. (1998) Biome reconstruction from pollen and plant macrofossil data for Africa and the Arabian peninsula at 0 and 6000 years, Journal of Biogeography 25, 1007-1028. Kattenberg, A., Giorgi, F., Grassl, H., Meehl, G.A., Mitchell, J.F.B., Stouffer, R.J., Kokioka, T., Weaver, A.J. and Wigley, T.M.L. (1996) Climate models - projections of future climate. In Climate change 1995 - the science of climate change Houghton, J.T., Meira Filho, L.G., Callander, B.A., Harris, N., Kattenberg, A. and Maskell, K. (eds.). Cambridge University Press, pp 285-357. Kubatzki, C. and Claussen, M. (1998) Simulation of the global bio-geophysical interactions during the Last Glacial Maximum, Climate Dynamics 14, 461-471. Kurvonen, L., Pulliainen, J. and Hallikainen, M. (1998) Monitoring of boreal forests with multitemporal special microwave imager data, Radio Science 33, 731-744. Kutzbach, J.E., Bonan, G., Foley, J.A. and Harrison, S.P. (1996) Vegetation and soil feedbacks on the response of the African monsoon to orbital forcing in the early to middle Holocene, Nature 384, 623-626. Kutzbach, J.E., Gallimore, R., Harrison, S.P., Behling, P., Selin, R. and Laarif, F. (1998) Climate and biome simulations for the past 21,000 years, Quaternary Science Reviews 17, 473-506. Laine, V. and Heikinheimo, M. (1996) Estimation of surface albedo from NOAA AVHRR data in high latitudes, Tellus A48, 424-441. Lean, J. and Rowntree, P.R. (1993) A GCM simulation of the impact of Amazonian deforestation on climate using an improved canopy representation, Quarterly Journal of the Royal Meteorological Society 119, 509-530. Lean, J. and Rowntree, P.R. (1997) Understanding the sensitivity of a GCM simulation of Amazonian deforestation to the specification of vegetation and soil characteristics, Journal of Climate 10, 1216-1235. Levis, S., Foley, J.A. and Pollard, D. (1999) Potential high-latitude vegetation feedbacks on CO2 -induced climate change, Journal of Geophysical Research 26(6), 747-750. Levis, S., Foley, J.A. and Pollard, D. (in press) Climate Vegetation Feedbacks at the Last Glacial Maximum. Journal of Geophysical Research (Atmospheres). Levis, S., Foley, J.A. and Pollard, D. (in press) Large scale vegetation feedbacks on a doubled CO2 climate. Journal of Climate. Lézine, A.-M. (1989) Late Quaternary vegetation and climate of the Sahel, Quaternary Research 32, 317-334. Manzi, O. and Planton, S. (1996) Calibration of a GCM using ABRACOS and ARME data and simulation of Amazonian deforestation. In Amazonian deforestation and climate, Gash, J.H.C., Nobre, C.A., Roberts, J.M. and Victoria, R.L. (eds.) John Wiley and Sons, 1st ed, pp 505-530. Melillo J.M., McGuire A.D., Kicklighter D.W., Moore B., III, Vörösmarty, C.J. and Schloss, A.L. (1993) Global climate change and terrestrial net primary production, Nature 363, 234-240. Neilson, R.P. (1995) A model for predicting continental scale vegetation distribution and water balance, Ecological Applications 5, 362-385. Nobre, C.A., Sellers, P.J. and Shukla, J. (1991) Amazonian deforestation and regional climate change, Journal of Climate 4, 957-988. Pielke, R. A. and Avissar, R. (1990) Influence of landscape structure on local and regional climate, Landscape Ecology 4, 133-155. Pielke, R. A., Dalu, G.A., Snook, J.S., Lee. T.J. and Kittel, T.G.F. (1991) Nonlinear influence of mesoscale land use on weather and climate, Bulletin of the American Meteorological Society 1-18. Pielke, R.A. and Vidale, P.L. (1995) The boreal forest and the polar front. Journal of Geophysical Research 100(D12), 25755-25758. Polcher, J. and Laval, K. (1994) The impact of African and Amazonian deforestation on tropical climate Journal of Hydrology 155, 389-405.
73
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Pollard, D., Bergengren, J.C., Stillwell-Soller, L.M., Felzer, B. and Thompson, S.L. (1998) Climate simulations for 10000 and 6000 years BP using the GENESIS global climate model, Paleoclimates 2, 183-218. Prentice, I.C., Cramer, W., Harrison, S.P., Leemans, R., Monserud, R.A. and Solomon, A.M. (1992) A global biome model based on plant physiology and dominance, soil properties and climate, Journal of Biogeography 19, 117-134. Prentice, I.C., Sykes, M.T., Lautenschlager, M., Harrison, S.P., Denissenko, O. and Bartlein, P.J. (1993) Modelling global vegetation patterns and terrestrial carbon storage at the last glacial maximum, Global Ecology and Biogeography Letters 3, 67-76. Prentice, I.C. and Webb, T., III. (1998) BIOME 6000: reconstructing global mid-Holocene vegetation patterns from palaeoecological records, Journal of Biogeography 25, 997-1005 Ritchie, J.C. and Haynes, C.V. (1987) Holocene vegetation zonation in the Eastern Sahara, Nature 330, 645-647. Robinson, D.A. and Kukla, G. (1985) Maximum surface albedo of seasonally snow-covered lands in the Northern Hemisphere, Journal of Climatology and Applied Meteorology 24, 402-411. Salo, J. (1987) Pleistocene forest refuges in the Amazon: evaluation of the biostratigraphical, lithostratigraphical and geomorphological data, Annales Zoologicae Fennici 24, 203-211. Sellers, P.J., Mintz, Y., Sud, Y.C. and Dalcher, A. (1986) A simple biosphere model (SiB) for use within general circulation models, Journal of Atmospheric Science 43, 505-531. Sellers, P. J., Los, S.O., Tucker, C.J., Justice, C.O., Dazlich, D.A., Collatz, G.J. and Randall, D.A. (1996) A revised land surface parameterization (SiB2) for atmospheric GCMs. Part II: the generation of global fields of terrestrial biophysical parameters from satellite data, Journal of Climate 9, 706-737. Sharratt, B.S. (1998) Radiative exchange, near-surface temperature and soil water of forest and cropland in interior Alaska, Agricultural and Forest Meteorology 89, 269-280. Shugart, H.H. (1984) A Theory of Forest Dynamics: The Ecological Implications of Forest Succession Models. New York: Springer-Verlag. 278 pp Shukla, J.L., Nobre, C. and Sellers, P.J. (1990) Amazon deforestation and climate change, Science 247, 1322-1325. Solomon, A.M. and Cramer, W. (1993) Biospheric implications of global environmental change. In Solomon, A.M. and Shugart, H.H. (eds.) Vegetation Dynamics and Global Change. New York, Chapman and Hall, pp 25-52. Steffen, W.L., Walker, B.H., Ingram, J.S. and Koch, G.W. (eds.) (1992) Global changes and terrestrial ecosystems: The operational plan: Global Change Report 21. International Geosphere-Biosphere Programme, Stockholm, Sweden, 95pp. Sud, Y. C., Walker, G.K., Kim, J.H., Liston, G.E., Sellers, P.J. and Lau, W.K.M. (1996) Biogeophysical consequences of a tropical deforestation scenario: a GCM simulation study, Journal of Climate 9, 3225-3247. TEMPO Authors (1996) Potential role of vegetation feedback in the climate sensitivity of high-latitude regions: A case study at 6000 years B.P, Global Biogeochemical Cycles 10, 727-736. Texier, D., de Noblet, N., Harrison, S.P., Haxeltine, A., Jolly, D., Joussaume, S., Laarif, F., Prentice, I.C. and Tarasov, P. (1997) Quantifying the role of biosphere-atmosphere feedbacks in climate change: a coupled model simulation for 6000 years BP and comparison with palaeodata for northern Eurasia and northern Africa, Climate Dynamics 13, 865-882. Thompson, S. L. and Pollard, D. (1995) A global climate model (GENESIS) with a land-surface transfer scheme (LSX). Part I: present climate simulation, Journal of Climate 8, 732-761. Thompson, S.L. and Pollard, D. (1997) Greenland and Antarctic mass balances for present and doubled atmospheric CO2 from the GENESIS version 2 global climate model, Journal of Climate 10, 871-900. van der Hammen, T. and Absy, M.L. (1994) Amazonia during the last glacial, Palaeography, Palaeoclimatology, Palaeoecology 109, 247-261. Van der Kaars, W.A. and Dam, M.A.C. (1997) Vegetation and climate change in West Java, Indonesia during the last 135000 years, Quaternary International 37, 67-71.
74
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Walker, B. (1994) Landscape to regional-scale responses of terrestrial ecosystems to global change, Ambio 23, 67-73. Woodward, F.I. (1987) Climate and Plant Distribution, New York: Cambridge University Press, 174pp. Woodward, F.I., Smith, T.M. and Emanuel, W.R. (1995) A global primary productivity and phytogeography model, Global Biogeochemical Cycles 9, 471-490. Zheng, X.Y. and Eltahir, E.A.B. (1997) The response to deforestation and desertification in a model of West African monsoons, Geophysical Research Letters 24, 155-158. Zheng, X.Y. and Eltahir, E.A.B. (1998) The role of vegetation in the dynamics of West African monsoons, Journal of Climate 11, 2078-2096.
75
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Working Groups Page numbers Introduction to Working Groups Wolfgang Cramer
77
Working Group 1: Climate Scenarios for Biospheric Impact Assessments Stephen Sitch and Navin Ramankutty
78
Working Group 2: Climate Scenarios for Forest Impact Assessments Harald Bugmann
80
Working Group 3: Climate Scenarios for Agricultural Impact Assessments Ana Iglesias and David Favis-Mortlock
84
Working Group 4: Global and Continental Scale Impact Assessments Benjamin Smith
87
Working Group 5: Regional Scale Impact Assessments Paula Harrison
90
Working Group 6: Stand-level Scale Impact Assessments Frank Wechsung
95
76
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Introduction to Working Groups Wolfgang Cramer Department of Global Change and Natural Systems, Potsdam Institute for Climate Impact Research, PO BOX 60 12 03, D-14412 Potsdam, Germany (
[email protected])
All Working Groups were asked to consider the following questions: 1. What does ‘good practice’ mean with respect to using climate change scenarios for impact assessments in our sector / at our scale? 2. What are current ‘state of the art’ practices? 3. What are achievable objectives of climate change impact assessment (and how are they limited by presently available climate scenario information)? 4. What advances are urgently needed to achieve a higher quality of impact assessments? In a first session, three Working Groups considered these questions with respect to the main sectors biospheric, forests/forestry and agricultural management. In a second session, with a different mix of the same participants, cross-sectoral issues were discussed at three different scales: global, regional (landscape) and stand-level.
77
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Working Group 1 Climate Scenarios for Biospheric Impact Assessments Stephen Sitch (chair), Navin Ramankutty (rapporteur) Participants: Ruth Doherty, Hervé Douville, Wolfgang Lucht, Bernard Nemry, Patricia de Rosnay, Benjamin Smith This Working Group comprised scientists from the Global Climate Model (GCM), biosphere impacts (terrestrial vegetation modelling, biodiversity), and the climate data communities. The group discussed the following in relation to the biospheric impacts community: • Current good practice in biospheric impacts assessments • Achievable objectives of the community with respect to using climate scenarios • Advances urgently needed to improve the interface and feedbacks between the biospheric impacts and climate scenarios communities
Current and Good Practice Until recently, many biosphere impact studies have used one so called ‘best guess’ GCM projection (itself forced by one future emission scenario, e.g., IPCC IS92a) which was often the only scenario available at that time. In fact the impact community has no real ‘best guess’ methodology until now. With the different spatial results between GCMs for some key variables and considering the number of possible future emission scenarios this is clearly an unsatisfactory situation. This has led to a progression towards a multiGCM, multi-emission scenarios suite of GCM scenarios (including coupled models with atmosphere, ocean and possibly biosphere components) and a need for ‘data services’ facilities to bridge the gap between the GCM and impact communities, in providing a large number of data sets with a common format. Concern was raised over the scaling issue. For terrestrial biosphere studies, GCM resolution is in general too coarse, with perhaps the exception of applications with global terrestrial biosphere models. Therefore, in order to study regional terrestrial biosphere impacts, appropriate scaling techniques are required. It is good practice for the impacts community to be self-informed about both a) relevant scaling techniques and b) with regard to the best set of GCM scenarios to use for the individual study. Both require a good knowledge of the climate scenario literature.
Achievable Objectives The need for adequate sensitivity analyses by the impact community would help to identify the key driving variables and appropriate spatial and temporal resolutions, which need to be supplied by the GCM community. In addition, the impact community should assess the uncertainties that they can afford in their driving variables. These sensitivities and other results should feedback to the climate modellers. In turn the climate community should provide a sufficient number of projections so that it is possible to access the actual uncertainties in the projections and thereby evaluate the feasibility of the impact study. They should also provide reliability ‘indicators’ for these individual climate scenarios, e.g., regional strengths and weaknesses, sharing their expertise to facilitate the choice of a ‘best set’ of GCM projections for individual impact studies. 78
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Advances Urgently needed Individual impact modellers must evaluate scenarios based on literature sources (informing themselves on both the GCMs and emission scenarios), and their own analyses of the GCM scenarios and control climates with a view to reducing the number of simulations. The impact community must not rely on one single GCM scenario. To find an intermediate solution between one and a large number of projections at least three projections, if not more, covering the range of GCM scenarios (lower, medium and upper responses) are required. In GCM studies, ‘ensemble simulations’ consider a number of realisations with the same forcing starting from different points along a ‘reference’ simulation i.e., starting from different initial conditions. Figure 1 compares the flow of information and data between the present situation and the final goal. Firstly, GCM groups have supplied output data to the data services, the Climate Research Unit (CRU). Here the data have been put into a common format and made easily accessible to the impacts community. The impacts community uses a set of GCM scenarios (as described above), supplying feedback either directly to the GCM group(s), thus sponsoring collaboration activities, or through the data services. The final goal (right hand side) is a harmonised system with easy and rapid flow of information and dialogue between the communities.
GCM groups
GCM Groups
Data Service (CRU)
Data Service (CRU)
Impact Community
Impact Community
Figure 1. Flow of information and data between the communities. Current and future interactions are represented by green and red arrows, respectively. (a) present situation (b) final goal.
79
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Working Group 2 Climate Scenarios for Forest Impact Assessments Harald Bugmann (Chair) Participants: Franz Badeck, Sten Bergström, Kristina Blennow, Jean-Luc Borel, Markus Erhard, Corine Hoff, Petra Lasch, Marcus Lindner, Merylyn McKenzie-Hedger, Christophe Neff, David Price, Adele Rowden, Roger Street, Martin Sykes In its discussion, this Working Group tried to answer four questions, which were phrased as follows: • What is the state-of-the-art in methodology of applying climate scenarios, and what advances are urgently needed? • What are objectives that are currently difficult to achieve in impact assessments that would be of high priority? • What does the term ‘good practice’ mean with respect to impact assessments in the forest/forestry sector? • How do we deal with differentiating climatic and non-climatic forcings that act upon the systems under consideration? Throughout its discussion, the group emphasized the fact that there is a considerable ‘internal diversity’ of the forest sector, implying that not only ecological aspects (e.g., forest structure, composition, growth, diversity) have to be dealt with, but that management aspects (e.g., traditional local practices, constraints at the level of the forest enterprise) as well as policy aspects (e.g., regional to national economies, global markets) are equally important. The variables considered and the methods used in impact assessments vary strongly depending on the aspect that is emphasized in any study. Hence, it is nearly impossible to make sweeping general statements about the use of climate scenarios in sectoral assessments that deal with forests. Rather, the answers to the above four questions often depend on the target variable(s) and the end users of an assessment. The group also agreed that few if any forest impact assessments at present provide predictions of the future state of these resources; while we may be working towards achieving true predictive capabilities, our methodologies (including, but not restricted to the climate scenarios) have not reached this stage – at least not yet. State-of-the-art in methodologies of applying climate scenarios In the forest sector (FS), the selection of climate scenarios is typically done on a pragmatic basis, dictated primarily by access to the scenario data, and sometimes also by political considerations (e.g., an assessment carried out in a given country typically will [need to] work with Global Climate Model (GCM) data from groups in the same country, if available). In the past, most impact assessments were carried out based on single climate scenarios. More recently, however, the trend is towards using several scenarios in comparative mode, which is an important step towards better exploring the range of ‘possible futures’. To date, the most widespread approach is to use anomalies between a simulated future climate and the simulated current climate from a GCM, and to impose those on the measured time series of the climatic variables of interest. There is less agreement, however, on the downscaling procedure that must be 80
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
employed for doing so. Some studies have used the anomalies from a single GCM grid-box, others have used average anomalies from several grid cells, whereas others still have used anomalies from Regional Climate Models (RCMs). Other methods that have been used include: • statistical downscaling, which – depending on one's perspective – comes at great cost (relative to the simple anomaly approach) or very cheaply (relative to RCMs); it provides a wealth of detailed data, but the major drawback may be the assumption that current relationship between the large-scale phenomena and the local time series will hold into the future • incremental changes to the observed records (e.g., an increase of 0.1 K per decade), which are easy to achieve, but lack physical consistency among variables • analogue climates (e.g., using the Bordeaux region as a possible future for the UK, or Florida as a possible future for eastern Canada); while this preserves the physical consistency between variables, it may convey a false sense of a ‘desirable future’ with the end users of a study • pure, systematic sensitivity analyses, which are possible for only a small subset of the climate variables under consideration (e.g., changes of long-term temperature and precipitation averages). Historically, most of the impacts research have dealt with assessing the effects of changes in the long-term averages of climatic parameters, and much less on changes of the higher orders (e.g., standard deviations), although we know that forests and the forestry sector are at least as sensitive and vulnerable to changes in the frequency and magnitude of extreme events. The above considerations are valid mainly for studies that focused on one single aspect of the FS; studies that involve a coupled system ranging from ecological aspects across management issues to forest policy analyses are faced with many scenarios other than climatic ones (e.g., technology development, land use, markets), which makes it often technically impossible to consider more than one or two climate scenarios. Also, such studies tend to focus on the development within the coming decades, whereas many ecological studies consider longer time frames up to several centuries in a hypothetical set-up, by assuming that climate after the year 2100 (or so) will remain constant. This is done to explore time lags and the long-term response of the system to (relatively) short-term changes. What advances are urgently needed? • increased communication regarding appropriate downscaling methods between GCM modellers (climate researchers) and those who perform impact assessments • more focus on changes in climatic parameters other than long-term averages • consider the relative capabilities of impact models and climate models – impact models need to be able to respond to the variables provided by climate scenarios, so along with improvements that are required in climate scenarios, impact models also need to become more sophisticated.
81
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Objectives that are currently difficult to achieve in impact assessments The members of this Working Group identified the following objectives that are difficult to achieve based on current methodologies: • pests and diseases and their impacts on ecosystem dynamics; this is valid for most except a few wellstudied cases (e.g., spruce budworm) • the combination of biological invasions, their pests, and new climates • CO2 impacts, especially at large spatial and temporal scales • CO2 /drought interactions • communication of uncertainties and assumptions inherent in impact assessments to the ‘end users’ of the studies, including the general public.
What does the term ‘good practice’ mean in our sector? From our discussions, the following key features of a good practice of using climate scenarios in forestrelated impact assessments arose: • sensitivity analysis: focus not only on behavior of the complete impacts model, but also on the sensitivity of sub-models with respect to climate and climatic change, so as to increase our understanding of the functioning of the complete model, including its limitations • ensemble simulations: perform ensemble simulations based on a range of different climate scenarios, in the sense of using GCM data for a guided, informed sensitivity analysis of forest models • uncertainty: try to derive probabilities and/or error bars as the outcome of impact assessments instead of single values that may convey a false sense of certainty in the results • communication of uncertainty: convey the potential, but also the limitations of the climatic input (scenarios) and of the forest model results to the end user • housekeeping: evaluate our own practice of using climate scenarios through meetings of impact assessment community to exchange ideas, discuss problems, etc; such activities would be very valuable, but are likely to be fairly difficult to fund. How do we deal with differentiating climatic and non-climatic forcings? Many past and ongoing changes in forests are not directly related to climatic forcings. For example, currently there is a shift in the dominance of many old-world Mediterranean systems from evergreen (Quercus ilex) to summergreen species (Quercus pubescens), which is induced by a combination of changes in forestry practices, land-use, and disturbance regimes (fires). These different effects need to be disentangled and understood in addition to understanding the impacts of a changing climate per se on forest and the forest sector.
82
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Towards this end, research is needed into the sensitivities of the forest sector to non-climatic forcings; that is, we need to evaluate what the key forcings are, and how sensitive the forest sector is with respect to these forcings. Also, work is needed into developing forest management scenarios and socio-economic scenarios for those non-climatic forcings that are considered significant to the forest sector (e.g., land-use - both consumptive and passive - , demand for forest resources, etc.). On a more technical level, the following points can be made: • Management practices and changes thereof can be incorporated into sectoral assessments in the form of additional scenarios (but see limitations discussed under state-of- the-art methodology section). At least with stand-alone forest models it is feasible to run sensitivity analysis, as outlined in the section above on good practice, with regard to CO2, Nitrogen fertilization, and management practices. It is then possible to compare the climate sensitivity of the forest ecosystem with other important influencing factors • Impacts of changing CO2 and Nitrogen fertilization pose no fundamental problem in process-based ecological models, as long as scenarios for the future trajectories of these drivers are available (there is no problem with including CO2 effects but perhaps it is less straightforward to include Nitrogen effects). • For practical purposes, we need to separate land cover changes (e.g., abandonment of agricultural land, deforestation) from the consideration of ecologically-caused changes that go on within currently forested areas. The former needs to be considered from a land use/land cover perspective that is largely driven by population size, economic status, etc., and much less by direct or indirect climatic causes.
83
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Working Group 3 Climate Scenarios for Agricultural Impact Assessments Ana Iglesias (chair), David Favis-Mortlock (rapporteur), Participants: Vesselin Alexandrov, Jules Beersma, Andy Bootsma, Martin Dubrovsky, Josef Eitzinger, Yehia Yehia Hafez, Paula Harrison, Márta Hunkár, Geoff Jenkins, Thomas Kartschall, Svetlana Nikulina, Seppo Rekolainen, Mark Rounsevell, Mikhail Semenov, John Taylor, David Viner Introduction The participants in this Working Group represented a wide diversity in their approaches to analysing the impacts of climate change on agriculture, both in relation to the scale of the studies – from local to national – and in relation to the objectives – from soil quality to socio-economic implications. The research interests of the group members together represented a broad set of methodologies for evaluating crop and land interactions with climate, but these research interests did not include animal production or crop-pest interactions. The fundamental agreement is that agricultural impact studies need to continue to be diverse and analyse both the crop ecosystem response and the socio-economic response to global change. The choice of scale of the analysis depends on the processes that need to be considered. Site scale analysis need to be integrated with spatial analysis to provide observations for model validation. The aim of the discussion was to analyse current practices and future expectations for using climate scenarios in agricultural impact assessments. Since the final goal of an impact assessment is to determine the effects that environmental change may have in society, the discussion also considered the use of integrated global change scenarios. The group discussed four questions: • What is ‘good practice’ in agricultural impact assessments? • What are the achievable objectives of climate change impact assessments in agriculture? • What methodological advances are desirable? • How should the difference between climatic and non-climatic forcings be approached?
Current use of scenarios in agricultural impact assessment The group identified a number of issues for applying climate change scenarios in agricultural impact studies. A fundamental premise for any evaluation was testing the methodological tools in the current climate conditions.
84
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
• Global Climate Model (GCM) scenarios. The studies should incorporate a minimum set of scenarios derived from GCMs (at least two), to adequately provide a range of meteorological variables that represent the current understanding of the anthropogenic climate forcing. GCM scenarios provide the framework for comparison of studies across regions and sectors. The use of composite scenarios, derived from more than one GCM, was discouraged since the resulting climate variables are not physically consistent, and therefore the results of the impact studies cannot be compared across regions. • Sensitivity analysis. It was strongly recommended to provide an understanding of the response of the system to individually perturbed meteorological variables. • Variability scenarios. The studies should incorporate and contrast the response of the system to climate change and climatic variability since there is increasing evidence that the effects of climate variability are a key factor for assessing the risk and vulnerability of many agricultural systems. Current scientific knowledge of climate variability and its prediction (e.g., the North Atlantic Oscillation and the El Niño/Southern Oscillation) should be incorporated into the impact evaluations. The use and validation of weather generators is an essential component for analyzing the effect of climate variability. • Downscaling. The studies should incorporate scenarios derived from direct perturbations of current climate variables with GCM output. The group did not consider (essential to include) more complex downscaled scenarios, due to the uncertainty associated to current downscaling techniques. Providing uncertainty estimates, derived from the use of different downscaled techniques, is essential for analysing the results of the impact studies. • Evaluating uncertainty. The studies should evaluate the uncertainty derived from the key assumptions and coupling of different models, and contrast this uncertainty with that derived from the use of different climate scenarios (provided by the atmospheric scientists). • Integrating scenarios. It is essential that the scenarios integrate internal consistency between climate and non-climate assumptions. Due to the wide range of objectives in agricultural impact studies it is not possible to recommend a consistent set of non-climatic scenarios. Nevertheless, water availability and water regulations in the countries of southern Europe will determine the future sustainability of their current agricultural systems and will influence land use. Policy should be considered explicitly since it is a main driver for change in current agricultural systems in the European Union.
Achievable objectives During the last ten years agricultural scientists and resource economists have advanced in their understanding of the processes that drive the interactions of agriculture with environmental factors and land use. The current methodologies define a set of short term (1 to 3 years) achievable objectives. Medium and long term (3 years or more) objectives require investment in further basic research. The group agreed to the following achievable objectives: • Understanding the effect of climate variability (short term). • Understanding the direct effect of CO2 on crop production (short term). • Developing and testing adequate downscaling methodologies (short term). • Analysing the signal and noise components of the impact evaluations (e.g., with respect to changes in both CO2 and climate) (short term).
85
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
• Linking closely impact and climate models (long term). • Incorporating feedbacks (e.g., change in CO2 resulting form changes in vegetation and land use) into the emissions scenarios used to drive the GCMs (long term).
Desirable methodological advances This discussion focussed on the desirable methodological advances that the impact community demands from the atmospheric science community. The group agreed that these two scientific communities need to define common objectives on the following issues: • Development of scenarios more relevant for regional analysis of impacts including those derived from statistical downscaling techniques and Regional Climate Models (RCMs). • Comparison of the output obtained from different downscaling methods is urgently needed for interpreting the impact results. This would permit the definition of guidelines for using downscaled scenarios in different impact assessments. • Research on the internal consistency of simulated climate variables and the GCM output. In some impact studies, climate data is needed that is either a) not available from the GCMs, or b) is available but not readily usable in regional-scale impact studies e.g., changes in rainfall intensity. In such cases, the only alternative is to use sensitivity-based perturbations of the observed parameters. The impacts community needs to understand if their simulated variables are consistent with GCM output.
Climatic and non-climatic forcings Current impact studies generally do not explicitly address or even ignore the issue of consistency of climatic and non-climatic forcings. In the past this has not been a major concern, since most studies have evaluated first-order impacts (e.g., on crop production). As the research community increasingly focuses on higherorder impacts (e.g., on agricultural pests, runoff available, or soil erosion) this lack of internal consistency will be more difficult to justify. The group agreed that development of integrated global change scenarios are essential for advancing the evaluation of impacts, vulnerability, and adaptation.
86
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Working Group 4 Continental and Global Scale Impacts Assessments Benjamin Smith (Chair) Participants: Hervé Douville, Hans-Martin Füssel, Bernard Nemry, Navin Ramankutty, Patricia de Rosnaye, Stephen Sitch, Roger Street, Christoph Zöckler. This Working Group considered the relevance of scale-related issues, particularly approaches to downscaling of climate scenario data, to impact assessment studies at continental to global scales. By ‘global scale’ we envisage both the global aggregate scale, at which a single global value is sought for each output variable at a given point in time, and what is here termed the ‘global grid’ scale, at which output variables may be mapped with sufficient spatial detail for global patterns to be discernible.
Impact models applied at continental to global scales Impact models applied at these scales are of necessity more general in both the internal representation of processes and mechanisms and in the types of output they produce. This makes them, in general, less demanding of driver variable precision and resolution compared with stand- to regional-scale models. The main categories include: •
Biosphere models, including equilibrium vegetation models, dynamic vegetation models and biogeochemical models which assume fixed vegetation. These models have been used particularly to characterise the global carbon cycle. Recent developments include interactive coupling to GCMs to incorporate vegetation feedbacks on climate.
• Hydrological models • Models of glacier/ice-sheet advance/retreat • Models of agriculture and food supply • Models of vector-borne diseases • Integrated Assessment Models (IAMs) linking representations of the physical, biological and socio-economic impacts on climate systems.
What target resolution is desirable at different scales? The most desirable resolution in a given study will be defined by trade-offs among: the ‘added value’ of downscaling the ‘raw’ GCM data, the intrinsic scale of processes and mechanisms represented in the impacts model (which may include upscaling of much finer scale processes, e.g., from the leaf to ecosystem for photosynthesis in vegetation models), the spatial and temporal precision of output variables in the impacts model, and the availability and resolution of non-climate driver data sets required by the impact model. 87
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
• Global aggregate scale: GCM resolution (e.g., the HadCM2 grid resolution of 2.5°x3.75°) may well be adequate to achieve summary statistics for the globe within the precision level of the impact model (e.g., for modelling of the global carbon cycle). • Global grid scale: A 0.5°x0.5° grid has been used in many studies. Because it is so widely used, there is an abundance of both driver and calibration data sets available at this resolution. It is also an appropriate scale for the inclusion of topographic information as required by nested Regional Climate Models (RCMs, see below). There have also been applications (e.g., of carbon cycle models) using the ‘raw’ GCM resolution. • Continental grid scale: 0.5°x0.5° grids have commonly been applied (e.g., the VEMAP project examining potential climate change impacts on US vegetation). However, this might be too coarse for Europe (some countries encompass less than 20 0.5° grid cells). Some data sets are available at considerably finer resolutions (e.g., 5’ soils for the USA) and a multi-resolution approach to take advantage of the extra detail might be desirable in some cases. It should be noted that not all models are grid-based, but may use e.g., political regions as the spatial units (e.g., many socio-economic models). In general, non-grid-based applications will have a courser equivalent resolution, and so be less demanding in terms of driver precision and resolution.
What downscaling techniques are appropriate? A common technique has been to apply changes in climate relative to ‘present-day climate’, derived from GCM output, to baseline (observed) data at the target resolution. This ‘anomaly approach’ has the advantage that it is simple and does not introduce any new model assumptions (as, for example, with certain more sophisticated statistical downscaling techniques). A disadvantage is that the physical consistency between variables is lost. The approach accommodates changes in the means of variables, but not their variances (at least when anomalies are applied arithmetically). Comparatively little is known about the sensitivity of impact models to changes in variability of the climatic drivers, but effects could be important, e.g., in dynamic vegetation models. Nevertheless, the anomaly approach may be adequate and appropriate for many studies downscaling by less than order of magnitude in spatial dimension. More sophisticated techniques include the use of statistical weather generators trained on regional climate in conjunction with GCM output; and nesting of RCMs within GCMs. Nested RCMs are justified especially for continental-scale assessments for which ‘raw’ GCM data may fail to accomodate sufficient detail as to meso-scale climate heterogeneity associated with topography and large water bodies (e.g., coarse or nonrepresentation of Rocky Mountains and Great Lakes in GCM surfaces for North America). Both weather generators and RCMs have limitations, however; they should not be regarded as a panacea for downscaling.
88
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Recommendations • For many applications 0.5°x0.5° is the finest resolution that is necessary, possible and/or sensible (i.e., no extra ‘added value’ is gained by going to finer scales). This is also an appropriate scale for nesting of RCMs to obtain finer resolution climate driver data. • At the global aggregate and in some cases global gridded scale, raw GCM resolutions may be adequate and appropriate, with regard to the intrinsic precision of the impact model used. • For many purposes the traditional ‘anomaly’ downscaling approach using change fields from GCM output may be adequate. However, this approach may be inappropriate especially for driving models that are sensitive to temporal variability in the boundary conditions. For these purposes other downscaling approaches, such as RCMs or statistical weather generators, may be preferable. • Studies are needed to understand more about the sensitivities of various families of impact models to resolution (temporal and spatial) of the climate drivers.
89
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Working Group 5 Regional Scale Impacts Assessments Paula Harrison (chair) Participants: Sten Bergström, Kristina Blennow, Andy Bootsma, Jean-Luc Borel, Gerd Bürger, Richenda Connell (rapporteur), Ruth Doherty, Markus Erhard, Yehia Yehia Hafez, Corine Hoff, Ana Iglesias, Marcus Lindner, Svetlana Nikulina, David Price, Seppo Rekolainen, Mark Rounsevell, Martin Sykes, John Taylor. Introduction The composition of this Working Group represented a wide diversity of disciplines and interests, which included regional climate modelling, downscaling and impacts assessments in the agriculture, forestry, vegetation and water sectors. All participants were involved in regional assessments at national or subnational scales, but the spatial scale of studies varied enormously from 100 to 10 million square kilometres. Four questions were discussed by the group: • What spatial resolution is feasible in regional impact assessments? • What are the critical temporal scales in regional impact assessments? • How should climate change scenarios be applied in regional impact assessments? • Is a common set of integrated scenarios desirable?
What spatial resolution is feasible in regional impact assessments? A fundamental problem in undertaking regional impact assessments is whether to develop more finely resolved climate data sets as input to existing impact models, or to modify the models to accept spatiallycoarser data (see discussion by Rounsevell, this volume). Both these approaches are being developed and several methods have been reported for scaling-up site models to regional assessments (van Gardingen et al, 1997; Downing et al., 1999). At the simplest level, results for representative sites are aggregated to a regional value (e.g., Easterling et al. 1993; Wolf, 1993; Rosenzweig and Parry, 1994). More spatially explicit approaches involve applying a site model or simplified model to input data sets which are spatially interpolated to regular grids and/or coherent polygons (e.g., Brklacich et al. 1996; Rounsevell et al. 1996; Harrison and Butterfield, 1996; Easterling et al. 1998; Carter et al. 1999). More quantitative methodologies rely on relating the site characteristics to its spatial domain using remotely sensed and other environmental data sets (e.g., Bindi et al. 1999; Delécolle, 1999). Discussions within this Working Group focussed on spatially explicit methodologies, assuming that climate change scenario requirements for the representative site approach would be covered in the stand-level assessments Working Group.
90
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
The group identified five issues which determine the spatial resolution of a regional assessment: • It is related to the purpose and end-use of the study, i.e., a higher spatial resolution may be required to provide information on changes in species/habitat composition to local interest groups, than to examine changes in national resources for political stakeholders. • It is driven by the process to be modelled, i.e., what resolution is needed to explain observed distributions or statistics? • It depends on the heterogeneity of the region, i.e., more topographically complex regions may need to be studied at higher spatial resolutions. • It should be consistent with the available resolution of the key environmental data sets, e.g., climate, soils, management, calibration/validation data sets. • It can be limited by the resolution of available data sets. For climate data, better communication is needed between the impacts community and climatologists in order to provide spatial data for the observed climate and climate change scenarios at appropriate resolutions for regional assessments.
What are the critical temporal scales in regional impact assessments? The group recognised that the spatial and temporal scale of a study are typically inter-linked and often related to the specification of the impact model. For example, at small spatial scales it is frequently important to capture rapid organism-level responses, such as sub-hourly variations in photosynthesis. Conversely, at larger scales the important phenomena are typically slower and can be adequately represented with data on daily or longer time steps (e.g., phenological development). At these different scales different complexities of modelling are appropriate. At the small scale, the response of enzymes and chemical reactions within a plant cell are significant, whilst at larger scales, where the objective may be to predict the broad-scale suitability of crops or plant functional types, such details can be ignored or parameterised. Despite a large variation in spatial scales within regional assessments, the majority of impact models operate at either a daily or monthly temporal resolution. Extreme events and short-term (hourly to daily) variations in climate are also often important determinants of biological responses, particularly in marginal environments where they may be the key drivers of change. Hence, daily extremes and other statistics of climatic variability are important fields which should be included in regional climate scenarios. Most impact assessments in the forest and natural vegetation sectors are now conducted in a transient mode because of the long time scales involved. However, the majority of agricultural assessments are still undertaken using an equilibrium modelling approach. In such studies, an impact model is first run for current climatic conditions and then run for future climatic conditions represented by the period-mean change field from several time slices of a transient climate model. The difference between the two model runs is the equilibrium response of the exposure unit to climate change. This approach has proved useful for studying agricultural impacts that operate on annual time scales. However, additional impacts which occur over longer time scales may be highlighted by exploring the transient approach for agriculture, such as impacts from cumulative soil moisture estimates, and the effects of rates of change on farmer’s risk perception, rotations and the take-up of new crops.
91
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
How should climate change scenarios be applied in regional impact assessments? The simplest method that is commonly used to construct climate change scenarios for regional impact assessments is to calculate Global Climate Model (GCM) change fields as the difference (absolute or relative) between the climate change experiment and its control run. The difference fields are determined at each GCM’s original spatial resolution and then may be simply applied to the finer spatial resolution of the observed climatology. Alternatively, the change fields may be first interpolated to this finer resolution before they are applied to the observed climate data. This process (sometimes termed ‘unintelligent’ downscaling; Hulme and Jenkins, 1998) adds no real geographic detail, but merely smoothes the GCM output. Such procedures may be appropriate for impact studies based on monthly climatic data. However, they should not be applied to daily or extreme climatic data sets, because the meteorological processes operating at sub-GCM grid scales are not represented. More sophisticated downscaling techniques based on statistical methods (regression/circulation types) were not generally recommended as the highly specific nature of these methods means that output can not be made generally available to the impacts community. Further, such techniques are data intensive, may add greater uncertainties, and the impacts community does not have the correct expertise to apply such techniques themselves. The development of a spatial version of the LARS-WG stochastic weather generator (or other similar weather generators, see Semenov this volume) may be useful for generating daily data sets and manipulating high frequency climatic variability. Nevertheless, a limited number of situations were identified where more detailed climate change information is necessary: • In areas of complex topography. • In regions where precipitation (or any variable derived from a sub-GCM grid scale process) is the key driving factor. • In regions classified as ocean grid boxes in the GCM, e.g., southern Italy in the HadCM2 GCM. In these situations, the use of results from Regional Climate Models (RCMs) was recommended. RCMs can produce output at spatial resolutions down to ~10 km with current technology, albeit over relatively small domains (say, 2500 grid points). However, the results are still limited by being driven by the boundary conditions of a GCM, although earlier problems are being resolved and work is now being undertaken using multiple GCMs to drive multiple RCMs. A major drawback of RCMs for impacts assessment is that only a limited number of experiments and time slices are available. Hence, a full exploration of uncertainties is not possible. It may however be possible to scale RCM output, in a similar manner to GCM output, to cover further sources of uncertainty. Nevertheless, this Working Group recommended a dual approach to the application of climate change scenarios in regional impact assessments where RCMs were useful for providing extra spatial detail (where this is necessary) and GCMs were useful for capturing a fuller range of uncertainty.
92
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Is a core set of integrated scenarios desirable? The group recommended that multiple scenarios should be applied in regional impact assessments. There was some discussion over whether a core set of integrated scenarios that would be applied in all assessments was desirable. Some thought this was very useful enabling more consistency and easier comparisons between studies, but care would need to be taken to ensure the impacts community were not endorsing a particular climate model. The conclusion of this discussion was that this item should be further reviewed at subsequent ECLAT-2 workshops. The provision of climate model output in easily accessible formats was encouraged by the impacts community. Insufficient time was available to define an appropriate format or a mechanism for making such data widely available.
References Bindi, M., Fibbi, L., Maselli, F. and Miglietta, F. (1999) Modelling climate change impacts on grapevine in Tuscany. In Climate Change, Climatic Variability and Agriculture in Europe: An Integrated Assessment, Downing, T.E., Harrison, P.A., Butterfield. R.E. and Lonsdale, K.G. (eds.), Research Report No. 21, Environmental Change Unit, Oxford, pp. 191-216. Brklacich, M., Curran, P. and Brunt, D. (1996) The application of agricultural land rating and crop models to CO2 and climate change issues in northern regions: the Mackenzie Basin case study, Agricultural and Food Science in Finland 5, 351-365. Carter, T.R., Saarikko, R.A. and Joukainen, S.K.H. (1999) Modelling climate change impacts on wheat and potato in Finland. In Climate Change, Climatic Variability and Agriculture in Europe: An Integrated Assessment, Downing, T.E., Harrison, P.A., Butterfield. R.E. and Lonsdale, K.G. (eds.) Research Report No. 21, Environmental Change Unit, Oxford, UK, pp. 287-310. Delécolle, R. (1999) Modelling climate change impacts on winter wheat in the Paris Basin. In Climate Change, Climatic Variability and Agriculture in Europe: An Integrated Assessment, Downing. T.E., Harrison, P.A., Butterfield. R.E. and Lonsdale, K.G. (eds.). Research Report No. 21, Environmental Change Unit, Oxford, UK, pp. 179-189. Downing, T.E., Harrison, P.A., Butterfield, R.E. and Lonsdale, K.G. (1999) (eds.), Climate Change, Climatic Variability and Agriculture in Europe: An Integrated Assessment, Research Report No. 21, Environmental Change Unit, Oxford. Easterling, W.E., Crosson, P.R., Rosenberg, N.J., McKenney, M.S., Katz, L.A. and Lemon, K.M. (1993) Agricultural impacts of and responses to climate change in the Missouri-Iowa-Nebraska-Kansas (MINK) region, Climatic Change 24, 23-61. Easterling, W.E., Weiss, A., Hays, C.J. and Mearns, L.O. (1998) Spatial scales of climate information for simulating wheat and maize productivity: the case of the US Great Plains, Agricultural and Forest Meteorology 90, 51-63. Harrison, P.A. and Butterfield, R.E. (1996) Effects of climate change on Europe-wide winter wheat and sunflower productivity, Climate Research 7(3), 225-241. Hulme, M. and Jenkins, G. (1998) Climate Change Scenarios for the UK: Scientific Report. UKCIP Technical Report No. 1, Climatic Research Unit, Norwich, 80pp. Rosenzweig, C. and Parry, M.L. (1994) Potential impact of climate change on world food supply, Nature 367, 133-138. Rounsevell, M.D.A., Loveland, P.J., Mayr, T.R., Armstrong, A.C., de la Rosa, D., Legros, J-P., Simota, C. and Sobczuk, H. (1996) ACCESS: a spatially-distributed, soil water and crop development model for climate change research, Aspects of Applied Biology 45, 85-91. van Gardingen, P.R., Foody, G.M. and Curran, P.J. (1997) (eds.) Scaling-up: From Cell to Landscape, Cambridge University Press, Cambridge. 93
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Wolf, J. (1993) Effects of climate change on wheat and maize production potential in the EC. In The Effect of Climate Change on Agricultural and Horticultural Potential in Europe Kenny, G.J., Harrison, P.A. and Parry, M.L. (eds.), Research Report No. 2, Environmental Change Unit, Oxford, pp. 93-120.
94
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Working Group 6 Stand-level Scale Impacts Assessments Frank Wechsung (chair) Participants: Vesselin Alexandrov, Jules Beersma, Harald Bugmann, Martin Dubrovsky, Josef Eitzinger, David Favis-Mortlock, Marta Hunkar, Thomas Kartschall, Christophe Neff, Adele Rowden Climate change scenarios for a stand- level impact assessment regarding agricultural and forest vegetation were discussed from the impact modellers perspective with regard to status, tasks and requests. Stand-level assessments of possible climate change impacts are conventionally focused on a virtual point, representing an area considered uniform with respect to the climatic conditions. This area can be a plot, a patch, a landscape, a river basin, etc. In many cases, an ensemble of stand-level assessments is used to characterise the impact on a larger area by aggregation of these single point assessments. Stand-level assessments are well suited for the analysis of processes that are largely independent of horizontal exchange processes to other sites, (e.g., crop growth and development). Also, the initiation of horizontal processes such as erosion may be considered at this scale. Stand-level assessments can be integrated within ensemble studies. However, limits for this integration appear if we have to consider horizontal exchange processes between single test sites with a routing pattern. This largely depends on the spatial characteristics of climate change. Examples are run-off, erosion and shifts in the spatial distribution of spores and insects. From an early stage (end of 1980’s, early 1990’s) stand-level studies proved to be valuable for estimating the impact of climate change scenarios on crop yield potential, crop yield risk and food security. When these first climate impact studies were carried out, the resolution of climate models was quite coarse and methods for further regionalisation were not available. This has changed over the last decade. Global Climate Models (GCMs) now perform at a higher resolution and downscaling methods allow further regionalisation of GCM outputs. Beside the original pragmatic motivation for stand-level studies there are still advantages that make them valuable in their own right. First of all, stand-level simulations may very often be more realistic and can be evaluated more thoroughly than regional-scale simulations. In addition, when a range of possible climate scenarios are to be considered stand-level studies present an opportunity, to limit the effort to a number of sites as opposed to a large area. For example, this could be the case if an integrated ‘best guess’ estimate about the general direction of change of an impact variable was required, when considering a number of future climate scenarios. Topics that could be studied in the near future at the stand-level scale rather than at other larger scales could be: • the effect of changed interannual, seasonal and daily climate variability on soil and vegetation processes • high temporal resolution impact studies (e.g., hourly) to explore climate change impacts on intensitydependent processes like erosion • detailed studies of local feedbacks between vegetation and climate.
95
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Considering these possible directions two challenges for the climate research community, that are specific to the needs of the stand-level research community were identified by this Working Group: • delivery of a high temporal resolution weather generator, which produces climate change scenarios for weather processes at time-scales of an hour and less • a vertical transfer weather generator that is able to translate a surface climate change signal into a vertical weather profile that interacts with the vegetation below. Beside these challenges, the need to test the performance of current state-of-the-art climate models and their regionalisation presents sufficient reason for carrying out stand-level studies.
96
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Poster Presentations Page numbers 1. Modelling the Risk of Windthrow in Forestry under Changing Climatic Conditions Kristina Blennow, Markku Rummukainen and Ola Sallnäs
98
2. Impact of Crops on Biomass and Soil Carbon: Steady State Simulations Bernard Nemry, Louis François, Jean-Claude Gérard, Dominique Otto, Daniel Rasse and Pierre Warnant
101
97
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Poster Presentation 1 Modelling the Risk of Windthrow in Forestry under Changing Climatic Conditions 1
2
1
Kristina Blennow , Markku Rummukainen and Ola Sallnäs
1. Southern Swedish Forest Research Centre, Swedish University of Agricultural Sciences, P.O. Box 49, SE-230 53 Alnarp, Sweden. (
[email protected]) 2. Swedish Meteorological and Hydrological Institute, Rossby Centre, SE-601 76 Norrköping, Sweden
Windthrow and Swedish forestry Wind and snow damage causes average production losses corresponding to 150 million ECU to Swedish forestry each year (Valinger and Fridman, 1997). The amount of damage is however, highly variable between years and in space. The landscape level distribution of wind damage depends on both the wind field over the terrain and the stability of the trees (Figure 1).
Figure 1.The distribution of wind damage caused by the autumn storms of 1969 in Tönnersjöheden experimental forest, Halland, Sweden.The distribution depends on both the wind field over the terrain and the stability of the trees.
Quantifying the risk of windthrow Within the MISTRA-funded research programme ‘Sustainable Forestry in Southern Sweden’ (SUFOR), a system of models is constructed for quantification of the risk of windthrow and for evaluation of different management programs over time in relation to this risk (Figure 2, Blennow et al., 1999). The system includes a model for describing the up-wind aerodynamical conditions of each stand, models for simulating the wind field over the terrain (Mortensen et al., 1998; Peltola et al., 1999), a model for calculating the critical wind speed for up-rooting or breakage (Peltola et al., 1999) and a model for dynamic simulation of the development of forest stands in a landscape (Dahlin et al., 1997). 98
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Soil conditions Wind climate Forest
Terrain
management
Exposure
Stability
to wind
tree/stand
(WASP + HWIND)
(HWIND)
PROBABILITY of windthrow
Figure 2. Our system of models for assessing the risk of windthrow in forestry. The system is constructed to enable quantification of the risk of windthrow under both present and changed climate conditions. This is achieved by feeding the models with probability distributions for climatic elements over time based on climate station data and based on down-scaled scenarios of future climate, respectively.
Windthrow and climate change Regional climate simulations for the Nordic countries indicate increased future windiness. The system of models will be used for simulating effects of a changed wind climate on the risk of windthrow, a response which is expected to be dependent on the wind exposure in a non-linear way. The effects of a changed wind climate will be simulated by feeding the models with downscaled climate data from regional climate change scenarios produced by the MISTRA-funded programme ‘Swedish regional climate modelling programme’ (SWECLIM). The regional climate change scenarios are done with dynamical downscaling of global model results (SWECLIM, 1998). Figure 3 illustrates some recent results, in this case a 22 km regional model interpretation of two 10-year time slices from the UKMO HadCM2 coupled model transient GHG-simulation (Johns et al., 1997; Mitchell and Johns, 1997).
Figure 3. A 22 km regional model interpretation for southern Sweden of two 10-year time slices from the UKMO HadCM2 coupled model transient GHG-simulation. Differences between two time slices, separated by a ~150% increase in equivalent CO2, are shown.The top panel shows 2 m air temperature changes (°C) depicted for winter, spring, summer and fall.The bottom panel shows the relative change (%) in the 10 m wind for the same seasons.The colouring indicates windiness changes >3%, roughly corresponding to significant changes in the mean regional wind conditions between the time slices.
99
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
References Blennow, K., Carlsson, M. and Sallnäs, O. (1999) Kan riskhantering hjälpa oss att undvika skador? (In Swedish.) Sustainable Forestry in Southern Sweden. Annual report 1998. Dahlin, B., Andersson, M., Ask, P., Carlsson, M., Eiderbrant, D. and Sallnäs, O. (1997) Konsekvenser av olika naturvårdsstrategier i skogsbruket: en studie av åtta typområden. (In Swedish) Rapport no. 4754. Naturvårdsverket, Stockholm. 97 pp. Johns, T. C., Carnell, R. E., Crossley, J. F., Gregory, J. M., Mitchell, J. F. B., Senior, C. A., Tett, S. F. B. and Wood, R. A. (1997) The second Hadley Centre coupled ocean-atmosphere GCM: model description, spinup and validation, Climate Dynamics 13, 103-134. Mitchell, J. F. B. and Johns, T. C (1997) On modification of global warming by sulphate aerosols, Journal of Climate 10, 245-267. Mortensen, N. G., Landberg, L., Troen, I. and Petersen, E. L. (1998) Wind Atlas Analysis and Application Program (WASP). Risø National Laboratory, Roskilde, Denmark. Peltola, H., Kellomäki, S., Väisänen, H. and Ikonen, V. -P. (1999) A mechanistic model for assessing the risk of wind and snow damage to single trees and stands of Scots pine, Norway spruce, and birch, Canadian Journal of Forest Research 29, 647–661. SWECLIM (1998) Regional climate simulations for the Nordic region – first results from SWECLIM. Swedish Meteorological and Hydrological Institute. Valinger, E. and Fridman, J. (1997) Modelling probability of snow and wind damage in Scots pine stands using tree characteristics, Forest Ecology and Management 97, 215–222.
100
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Poster Presentation 2 Impact of Crops on Biomass and Soil Carbon: Steady State Simulations Bernard Nemry, Louis François, Jean-Claude Gérard, Dominique Otto, Daniel Rasse and Pierre Warnant Laboratory for Planetary and Atmospheric Physics Institut d'Astrophysique et de Géophysique, Université de Liège B-4000 Liege, Belgium (
[email protected])
Abstract Crop impacts on carbon stored in living biomass, litter and soil were estimated by comparing three simulated steady states of the continental biosphere. These steady states were characterised by the crop distribution and the atmospheric carbon dioxide concentration. Increase in the carbon contents of the biospheric pools by CO2 fertilisation largely dominated the carbon lost when forested areas were converted to agriculture. As far as regional averages are concerned, the results are less influenced by spatial resolution than temporal resolution. In particular, biospheric simulations from stochastic or linear distributions of daily temperature and precipitation exhibit large differences in carbon fluxes and contents.
Introduction While gross primary productivity (GPP) is the input carbon flux to living biomass by photosynthetic production, autotrophic respiration (RA) and mortality remove carbon from this pool. Dead matter accumulates in litter and deep soil layers and is decomposed by bacterial colonies which release carbon to the atmosphere through the heterotrophic respiration (RH). All these carbon fluxes depend on the seasonal and interannual changes in climate variables. Photosynthesis is also enhanced by increased levels of atmospheric CO2 concentration, provided that water and nutrients are available (IPCC, 1995). Due to anthropogenic changes in land use, atmospheric CO2 concentration and climate, the present amounts of carbon in the continental biospheric pools are transient. In particular, increases in net primary productivity (NPP = GPP - RA) by CO2 fertilisation induces a rise in the amount of carbon contained in living biomass. The subsequent increase in litter and soil carbon contents intensifies RH. During such a transient episode, the change in NPP leads the change in RH and this flux difference results in a net annual exchange of carbon between the biosphere and the atmosphere. As a consequence, pools of continental carbon contained in biomass, litter and soil are changing from year to year. A new steady state will be reached when climate, land use and atmospheric CO2 concentration stabilise. A steady state simulation consists in determining the amounts of carbon contained in the pools of biomass, litter and soil in such a way that annual values of NPP and RH balance each other (Nemry et al., 1996). In order to estimate the influence of crops on the biospheric carbon budget, we analysed the differences between simulated steady states of the present-day continental vegetation and continental vegetation without crops. Impacts of CO2 fertilisation were also estimated.
101
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Materials and methods We used a standard version of the terrestrial biospheric model CARAIB (Warnant et al. 1994), where the continental cover is divided into grid boxes of fixed longitudinal and latitudinal extensions. Fractional areas of functional plant types (FPTs) are ascribed to each grid box: C3 and C4 grasslands, C3 and C4 crops, needleleaf forests (evergreen and deciduous) and broadleaf forests (evergreen and deciduous). GPP is calculated from the models of Farquhar et al. (1980) and Collatz et al. (1992) for C3 and C4 plants, respectively. Both RA and RH depend exponentially on temperature. Carbon budget is calculated in biomass, litter and soil for each FPT. Biomass is composed of green and non green components. Green biomass includes foliage and grassy matter while non green biomass refers to woody matter and stems. Green and non-green litter pools are fed by the mortality in the corresponding biomass pools. Soil organic carbon pool is fed by both the litter pools. Heterotrophic respiration results from the bacterial oxidation of litter and soil organic carbon.
90
90
60
60
30
30 Latitude
Latitude
A ‘no crop world’ scenario was built from the present vegetation distribution, by replacing crops with neighbouring forests. The grassland areas were not modified. Present contributions of C3 and C4 crops to the 6 2 6 2 6 2 continental vegetation cover (127.41x10 km ) are 11.60x10 km and 2.06x10 km , respectively (Figure 1). 6 2 Together, they have reduced the area of needleleaf evergreen forests by 2.38x10 km between 40°N and 6 2 60°N, and the area of broadleaf deciduous forests by 9.44x10 km , between 0° and 60°N. Other forested areas have been reduced to a lesser extent: broadleaf evergreen forests and needleleaf deciduous forests 6 2 6 2 have lost 1.49x10 km and 0.35x10 km , respectively.
0 -30 -60
C3 grassland 6 2
-60
43.58 10 km
90
60
60
30
30 Latitude
Latitude
-90 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 106 km2 per degree of latitude
90
0 -30 -60
C4 grassland 6 2
-60
23.70 10 km
2.23 10 km -> 2.26 10 km
90
90
60
60
30
30
0 -30
0 -30
C3 crops6
-60
11.60 10 km2 -> 0.00 106 km2
-90 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 106 km2 per degree of latitude 90
60
60
30
30 Latitude
90
0 -30
Broadleaf evergreen forests 6 2 6 2 10.63 10 km -> 11.96 10 km
-90 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 106 km2 per degree of latitude
-60
Needleleaf deciduous forests 6 2 6 2
-90 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 106 km2 per degree of latitude
Latitude
Latitude
0 -30
-90 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 106 km2 per degree of latitude
-60
Needleleaf evergreen forests 6 2 6 2 8.38 10 km -> 11.43 10 km
-90 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 106 km2 per degree of latitude
Latitude
0 -30
0 -30
C4 crops 6
-60
2.06 10 km2 -> 0.00 106 km2
-90 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 106 km2 per degree of latitude
Broadleaf deciduous forests 6 2 6 2 25.23 10 km -> 34.48 10 km
-90 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 106 km2 per degree of latitude
Figure 1. Comparison of vegetation areas between the present distribution (solid lines) and a 'no crop world' (dashed lines).The area distributions are presented for the eight functional plant types (FPTs) considered in each CARAIB grid cell.The change in global area is indicated from the ‘no crop world’ to now. Grassland distributions are not modified.
102
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Fertilization of the biosphere by increasing atmospheric CO2 concentration was estimated by comparing biospheric steady states simulated at two levels of CO2 concentration. Increments of biomass and soil carbon contents from the 'no crop world' at 280 ppmv to the present vegetation at 360 ppmv were calculated at both spatial resolutions of 3°x3° and 5°x5° in longitude and latitude. François et al. (1996) and Gérard et al. (1999) used CARAIB to analyse the influence of interannual climate changes on the biosphere. In the assessment of the present results, we focussed on the direct impacts of crops and CO2 fertilization and deliberately neglected any interannual climatic change, including greenhouse warming. All the simulations were performed on the basis of the average seasonal climatology from Cramer and Leemans (personal communication). Nevertheless, two daily weather inputs were considered. A standard distribution of daily temperature and precipitation was produced by the stochastic weather generator included in CARAIB (Hubert et al., 1998). In an alternative weather distribution, temperature was changed linearly from day to day between two successive months, while precipitation did not change during any month.
Results and discussion -1
A global annual value of 67 Gt C yr was calculated for the sink of carbon associated with the NPP flux under an atmospheric CO2 concentration of 360 ppmv. This result is slightly higher than the upper values found by most terrestrial biospheric models with CO2 concentration between 340 and 350 ppmv (Cramer et al., -1 1999). Under 280 ppmv, the estimated global annual NPP sink was reduced by 21%; a value of 53 Gt C yr was calculated whether the considered biosphere was the ‘no crop world’ or the present vegetation (Figure 2). When atmospheric CO2 concentration was doubled from 330 ppmv to 660 ppmv, the simulated NPP increased by 67% (Warnant, 1999). This biospheric response to a doubling in atmospheric CO2 exceeds the 20 to 40% range measured in small-scale experiments. This range can be reduced due to unavailability in water or nutrient (IPCC, 1995).
Figure 2. Latitudinal distribution of the annual carbon sink associated with the NPP flux simulated by CARAIB.Three steady states are compared. Vegetation is considered to include crops (‘now’) or not (‘no crop world’).Two atmospheric CO2 concentrations have been used.
103
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
As far as latitudinal distribution was considered- in bands 30° wide, our results were rather insensitive to the spatial resolution. Sensitivity to temporal resolution was higher. When the stochastic sequence of daily weather was replaced with the linear sequence, the global annual amount of carbon absorbed by NPP flux was reduced by 30 %. As a consequence, stochastic daily weather can be a cause of overestimation in NPP values. It allowed the simulated atmospheric CO2 concentration to match the observation, as long as seasonal amplitudes and phases are considered in the Northern Hemisphere (Nemry et al., 1996). Nevertheless, the current “Ecosystem Model-Data Intercomparison” exercise of the International Geosphere-Biosphere Programme indicates that simulated annual NPP values can differ significantly from site measurements. Further analysis should improve the NPP modelling. The transition from the ‘no crop world’ at 280 ppmv to the present vegetation at 360 ppmv can be decomposed into two transitions: (i) an ‘atmospheric carbon change’ (ACC) transition of the ‘no crop world’ from 280 ppmv to 360 ppmv ; (ii) a ‘land use change’ (LUC) transition from the 'no crop world' to the present vegetation at the constant CO2 concentration of 360 ppmv. In the ACC transition, the carbon content increased by 175 Gt C in the biomass, 60 Gt C in the litter pools and 492 Gt C in the soil. Fertilisation of the 'no crop world' induced a total increase in biospheric carbon of 727 Gt C. These results were produced with stochastic daily weather (Figure 3). Non green biomass 90
75
75
60
60
45
45
30
30 Latitude
Latitude
Green biomass 90
15
15
0
0
-15
-15
-30
-30
-45 -60 -1.0
-45
-0.5
0.0
0.5 1.0 1.5 2.0 2.5 Gt C per latitude band 15o wide
3.0
-60 -50
3.5
-40
-30
90
90
75
75
60
60
45
45
30
30
15
40
50
40
50
175
200
15
0
0
-15
-15
-30
-30
-45 -60 -1.0
-20 -10 0 10 20 30 Gt C per latitude band 15o wide
Non green litter
Latitude
Latitude
Green litter
-45
-0.5
0.0
0.5 1.0 1.5 2.0 2.5 Gt C per latitude band 15o wide
3.0
-60 -50
3.5
-40
-30
-20 -10 0 10 20 30 Gt C per latitude band 15o wide
Soil organic carbon 90 75 60
ACC + LUC transition: Now (360 ppmv) - NCW (280 ppmv)
45
Latitude
30
LUC transition: Now (360 ppmv) - NCW (360 ppmv)
15 0
ACC transition: NCW (360 ppmv) - NCW (280 ppmv)
-15 -30 -45 -60 -25
0
25
50 75 100 125 150 Gt C per latitude band 15o wide
Figure 3. Latitudinal distribution of the carbon increments simulated by CARAIB in five pools : green and non green biomass, green and non green litter, and soil. Three transitions are considered : (i) Atmospheric CO2 Change (ACC) transition : ‘no crop world’ (NCW) from 280 ppmv to 360 ppmv; (ii) Land Use Change (LUC) transition : from NCW to now, at 360 ppmv ; (iii) ACC + LUC transition : from NCW at 280 ppmv to now at 360 ppmv.
104
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
In the LUC transition, the loss of woody biomass associated with the substitution of crops for forested areas amounted to 160 Gt C. All latitudes were affected, especially in the northern temperate regions. Green biomass was lost to a lesser extent (around 1 Gt C). It happened mainly between 45°N and 60°N, where the largest areas of crops have been planted at the expense of needleleaf evergreen and broadleaf deciduous forests. Non green litter carbon decreased by 3 Gt C while green litter carbon increased by an equivalent quantity. Soil organic carbon increased by 146 Gt C. The global budget of the biospheric carbon pools was a loss of about 15 Gt C. According to ACC and LUC transitions, the biospheric carbon increase induced by CO2 fertilisation prevailed on the loss of carbon due to the substitution of crops for forests. The net result of these transitions was an increase of 712 Gt in the global amount of biospheric carbon. The increase in carbon contents from the ‘no crop world’ at 280 ppmv to the present vegetation at 360 ppmv occurred in all pools at all latitudes, except for the northern temperate latitudes where the loss of woody biomass dominated the fertilisation effect by about 30 Gt C.
Conclusion We have compared simulated steady states of the continental biosphere based on two vegetation distributions. Present vegetation distribution was considered with present atmospheric CO2 concentration. In the second scenario, forests replaced present crops and the atmospheric CO2 concentration was set to the pre-industrial level. This comparison relies on some uncertain hypotheses. First, the steady states were simulated with the same seasonal cycle and annual average of the climatic variables. Second, crop harvest and subsequent biomass removal were not simulated. Finally, the disagreement of annual NPP values between current model results and recent data has to be explained and reduced. Regional results are less influenced by spatial resolution than temporal resolution. In particular, simulations largely differ when daily temperature and precipitation are derived from stochastic or linear distributions. Subject to these limitations, the increase in biospheric carbon content due to CO2 fertilization dominates by several orders of magnitude the loss of carbon associated with the substitution of crops for forested areas.
References Collatz G.J., Ribas-Carbo, M. and Berry, J.A. (1992) Coupled photosynthesis-stomatal conductance model for leaves of C4 plants, Australian Journal of Plant Physiology 19, 519-538. Cramer W., Kicklighter, D., Bondeau, A., Moore III, B., Churkina, G., Ruimy, A., Schloss, A. and participants of “POTSDAM 95” (1999) Comparing global models of terrestrial net primary productivity (NPP): Overview and key results, Global Change Biology 5 (Suppl. 1), 1-15. Farquhar G.D., von Caemmerer, S. and Berry, J.A. (1980) A biochemical model of photosynthesis CO2 assimilation in leaves of C3 species, Planta 149, 78-90. François L., Nemry, B., Warnant, P. and Gérard, J.-C. (1996) Seasonal and interannual influences of the terrestrial ecosystems on atmospheric CO2 : A model study, Physics and Chemistry in the Earth 21, 537-544. Gérard J.-C., Nemry, B., François, L. and Warnant, P. (1999) The interannual change of atmospheric CO2: contribution of subtropical ecosystems? Journal of Geophysical Research 26, 243-246. Houghton J.T., Meira Filho, L.G., Bruce, J., Lee, H., Callander, B.A., Haites, E., Harris, N. and Maskell, K. (eds.) Climate Change 1994: Radiative Forcing of Climate Change and An Evaluation of the IPCC IS92 Emission Scenarios, Cambridge University Press, Cambridge. Hubert B., François, L., Warnant, P. and Strivay, D. (1998) Stochastic generation of meteorological variables effects on global models of water and carbon cycles in vegetation and soils, Journal of Hydrology 212-213, 318-334. 105
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Nemry B., François, L., Warnant, P., Robinet, F. and Gérard, J.-C (1996) The seasonality of the CO2 exchange between the atmosphere and the land biosphere : A study with a global mechanistic vegetation model, Journal of Geophysical Research 101, 7111-7125. Warnant P., François, L., Strivay, D. and Gérard J.-C. (1994) CARAIB: A global model of terrestrial biological productivity, Global Biogeochemical Cycles 8, 255-270. Warnant P., François, L., Nemry, B., Hubert, B., Molitor, N., Colinet, G. and Gérard, J.-C. (1995) A new distribution of vegetation types and its inclusion on a global biosphere model. Proceedings of the International Colloquium Photosynthesis and Remote Sensing, 28-30 August 1995, Montpellier. G. Guyot, EARSeL, Paris. Warnant P (1999) Modélisation du cycle du carbone dans la biosphère continentale à l’échelle globale. Ph. D. Université de Liège, Belgique. 97, 215–222.
106
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Climate Scenarios for Agricultural, Forest and Ecosystem Impacts - Workshop Summary Wolfgang Cramer Department of Global Change and Natural Systems, Potsdam Institute for Climate Impact Research, PO BOX 60 12 03, D-14412 Potsdam, Germany (
[email protected]) The topic of this workshop: climate scenarios for agricultural, forest and ecosystem impacts, may have given the impression to many people of a rather technical issue, which requires specialists exchanging knowledge about the practical application of models and techniques. Not surprisingly, the discussion took a quite different approach and encompassed the following issues: What is climate impact assessment? Why are we doing it? Who are we doing it for? At what scale? What is the uncertainty of our results? In addressing these rather more fundamental issues, participants clearly assisted each other in finding out each individual’s position in a rapidly growing field of fundamental and applied science – and this was indeed the main underlying goal of the meeting. Clearly, none of these broad questions has a simple answer. It is precisely the continuous revisiting of the logical foundations and objectives of climate change impacts assessments that eventually may lead to insights which are both, scientifically sound and politically relevant. As far as the more narrow problem- the definition and application of suitable climate scenarios- was concerned, participants were specifically looking for methodologically stable solutions that could be applied, if possible, across sectors and across scales. Trying to identify a ‘good practice’ posed difficult, due to a variety of constraints – these are covered in the Working Group reports and shall not be repeated here. Some additional problems, which arose repeatedly in formal discussions, as well as during coffee breaks, are given here: • Access to climate data is still a significant problem for many groups working on climate impact studies. This limitation may concern both observations of historical and present-day conditions, and output from scenario generation activities elsewhere. It is surprising that this remains an issue, given the amount of resources that have been used for networking and database creation in recent years. One effort that has been successful in this regard is the IPCC Data Distribution Centre for scenario information, which provides observed climate data, socio-economic scenario information and results from GCM experiments; see http://ipcc-ddc.cru.uea.ac.uk. The workshop had the impression that many such limitations were in fact more due to either the commercialisation of weather services in some countries, or else the reluctance of some model development teams to see ‘their results’ being used in an inappropriate way. • Climate scenario data are often generated from model simulations that had other primary goals than climate change impacts assessment studies. This is no surprise since climate modellers have the primary and well-justified needs to better understand dynamical processes in the atmosphere and its associated systems. Climate change scenarios for impact assessments are usually an offspring rather than an intended result from such studies. It is important that climate impact modellers understand this situation: Global Climate Model (GCM) experiments are necessary to investigate the kinds of changes that could occur in the future. Being explorative rather than predictive they can therefore only serve as indications of ‘kinds of futures’.
107
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
• The intimate coupling between greenhouse gas emissions, atmospheric processes and modifications of the land surface, either through land use or through impacts of climate change, in fact prohibits the possibility of ‘accurate’ predictions to ever exist. Increased rainfall, for example, ultimately leads to increased water availability which will modify ecosystem processes. These feedback to the atmosphere by altered evapotranspiration and other fluxes. Off-line scenarios, as they are usually encountered in impact assessments, will always suffer from this inconsistency which should be recognised. • Few recent climate impacts assessments have included a consistent consideration of climate variability. This is unfortunate, since variability is likely to confound attribution of climate change related impacts for quite some time (even in the presence of global change, see Hulme et al. 1999). This is a severe limitation, since many impact sectors are sensitive to changes in variability, and the potential of climate models to simulate the various dimensions of climate variability is still limited. In conclusion, there is no single ‘best practice’ in the development and use of climate scenarios, but there are a number of pitfalls that need to be avoided in the future. Most prominent among these are the wrongful presentations of some impact scenario calculations as ‘predictions’ of change. All we can really do is exploit the scenario technique to explore a range of different conditions, within which we are likely to find an estimate of the sensitivity of the sector being studied. Determining this sensitivity however yields, if it is done following scientific standards, a powerful message. Policy concerning abatement of future emissions of greenhouse gases will have to consider whether the climate system potentially could develop conditions that are detrimental for services provided to human society by its environment. ‘Good practice’ would then be equivalent with policy that minimises and/or avoids this risk being taken. Reference Hulme M, Barrow, E. M., Arnell, N.W., Harrison, P. A., Johns, T.C. and Downing, T.E. (1999) Relative impacts of human-induced climate change and natural climate variability, Nature 397, 688-691.
108
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Appendices Page numbers Appendix 1: Workshop Agenda
110
Appendix 2: List of Participants
113
Appendix 3: Relevant Web sites
118
109
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Appendix 1:Workshop Agenda Wednesday 13 October 1999 Morning-
Arrival in Berlin/Potsdam and registration
12:30
Lunch
14:00
Workshop Opening, Welcome, Setting the stage Wolfgang Cramer (Germany)
14:15
Goals of ECLAT-2 and short summary from the red ECLAT-2 Workshop Representing in climate scenarios and impact studies in Helsinki David Viner (UK)
14:30
The IPCC and ACACIA perspective on scenario generation for climate change impact assessments Wolfgang Cramer (Germany)
14:50
Using climate model simulations for the generation of scenarios for impact models Gerd Bürger (Germany)
15:10
Keynote paper: Agriculture and climate change scenarios Mark Rounsevell (Belgium)
15:50
Coffee break and poster session
16:20
Keynote paper: Assessing the impact of climate change on forests and forestry using climate scenarios: current practice and future directions David Price (Canada)
17:00
Open discussion
17:30
1st Working Group Assignments: Consistent application of climate and CO2 scenarios in different sectors (agriculture, forestry, global biosphere) (Wolfgang Cramer)
19:30
Dinner (restaurant Minsk)
110
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Thursday 14 October 1999 9:00
Short plenary (reminder of Working Group assignments, logistics)
9:10
1st set of (three) Working Groups meet, guided by the rapporteurs (chairpersons appointed by groups if necessary) Coffee available all morning
12:30
Plenary: Building climate change scenarios on the basis of meteorological observations and climate model output Friedrich-Wilhelm Gerstengarbe and Peter C. Werner (Germany)
13:00
Lunch
14:00
Plenary – Working Group reports:
14:00
Climate scenarios for biospheric impact assessments Stephen Sitch (Germany)
14:20
Climate scenarios for forest impact assessments Harald Bugmann (Switzerland)
14:40
Climate scenarios for agricultural impact assessments Ana Iglesias (Spain)
15:00
Coffee break and poster session
15:30
Keynote paper: Development of climate change scenarios for agricultural applications Mikhail Semenov (UK)
16:15
2nd Working Group Assignments: Stand level, regional, global assessments Jules Beersma (The Netherlands)
16:30
2nd set of Working Groups meet for rest of day, guided by rapporteurs Coffee available all afternoon
19:30
Dinner (restaurant Fliegender Holländer)
111
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Friday 15 October 1999 9:00
Plenary – Working Group reports:
9:00
Continental and global scale impact assessments Ben Smith (Sweden)
9:20
Regional scale impact assessments Paula Harrison (UK)
9:40
Stand-level scale impact assessments Frank Wechsung (Germany)
10:00
Concluding discussion of Working Group findings
10:30
Coffee break and poster session
11:00
Keynote paper: Integrated modelling of the Earth system: bi-directional atmosphere-biosphere interactions Navin Ramankutty (USA)
11:45
Final concluding discussion
13:00
Lunch
13:30
Closing the workshop, outlook Wolfgang Cramer (Germany)
112
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Appendix 2: List of Participants Alexandrov VESSELIN University of Agriculture (BOKU) Institute of Meteorology and Physics Türkenschanzstr. 18 1180 Wien AUSTRIA Tel: +43-1-470-5820 Fax: +43-1-470-582060 E-mail:
[email protected]
Jean-Luc BOREL Laboratoire "Ecosystèmes Alpins" Centre de Biologie Alpine Université Joseph Fourier B.P. 53, F-38041 Grenoble Cedex 9 FRANCE Tel: +33-7663-5624 Fax: +33-7651-4463 E-mail:
[email protected]
Jules BEERSMA Royal Netherlands Meteorological Institute (KNMI) P.O. Box 201 3730 AE De Bilt THE NETHERLANDS Tel.: +31-30-2206475 Fax: +31-30-2210407 E-mail:
[email protected]
Franz BADECK Potsdam Institute for Climate Impact Research P.O. Box 60 12 03 14412 Potsdam GERMANY Tel: +49-331-288-2575 Fax: +49-331-288-2600 E-mail:
[email protected] Harald BUGMANN Mountain Forest Ecology Department of Forest Sciences Swiss Fed. Inst. of Technology ETH-Zentrum HG G21.3 CH-8092 Zürich SWITZERLAND Tel:+41-1-632-3239 Fax: +41-1-632-1146 E-mail:
[email protected]
Sten BERGSTRÖM Swedish Meteorological and Hydrological Institute 601 76 Norrköping SWEDEN Tel: +46-11-495-8000 Fax: +46-11-495-8001 E-mail:
[email protected] Kristina BLENNOW Southern Swedish Forest Research Centre Swedish University of Agricultural Sciences Box 49, 23053 Alnarp SWEDEN Tel: +46-40-415230 Fax: +46-40-462325 E-mail:
[email protected]
Gerd BÜRGER Potsdam Institute for Climate Impact Research P.O. Box 60 12 03 14412 Potsdam GERMANY Tel: +49-331-288-2574 Fax: +49-331-288-2600 E-mail:
[email protected]
Andrew BOOTSMA Agriculture and Agri-Food Canada Eastern Cereal and Oilseed Research Centre K.W. Neatby Bldg Rm 4129B, 960 Carling Ave Ottawa, K1A 0C6 CANADA Tel: +1-613-759-1526 Fax: +1-613-759-1924 E-mail:
[email protected]
Wolfgang CRAMER Potsdam Institute for Climate Impact Research P.O. Box 60 12 03 14412 Potsdam GERMANY Tel: +49-331-288-2521 Fax: +49-331-288-2600 E-mail:
[email protected]
113
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
David FAVIS-MORTLOCK Environmental Change Unit Oxford University 5 South Parks Road Oxford OX1 3UB UK Tel:+44-(0)1865-281180 Fax: +44-(0)1865-281202 E-mail:
[email protected]
Richenda CONNELL UK Climate Impacts Programme Union House 12-16 St. Michael's Street Oxford, OX1 2DU UK Tel: +44-1865-432073 Fax: +44-1865-432077 E-mail:
[email protected] Ruth DOHERTY Climatic Research Unit, University of East Anglia Norwich UK Tel.: +44-1603-593818 Fax: +44-1603-507784 E-mail:
[email protected]
Hans-Martin FÜSSEL Potsdam Institute for Climate Impact Research P.O. Box 60 12 03 14412 Potsdam GERMANY Tel: +49-331-288-2537 Fax: +49-331-288-2600 E-mail:
[email protected]
Hervé DOUVILLE CNRM/GMGEC/UDC 42 Avenue Coriolis F-31057 Toulouse Cedex 1 FRANCE Tel: +33-561-079625 Fax: +33-561-079610 E-mail:
[email protected]
Friedrich-Wilhelm GERSTENGARBE Potsdam Institute for Climate Impact Research P.O. Box 60 12 03 14412 Potsdam GERMANY Tel: +49-331-288-2586 Fax: +49-331-288-2600 E-mail:
[email protected]
Martin DUBROVSKY Institute of Atmospheric Physics Husova 456 50008 Hradec Kralove CZECHIA Tel: +420-49-526-3759 Fax: +420-49-526-3759 E-mail:
[email protected]
Uwe HABERLANDT Potsdam Institute for Climate Impact Research P.O. Box 60 12 03 14412 Potsdam GERMANY Tel: +49-331-288-2577 Fax: +49-331-288-2600 E-mail:
[email protected]
Josef EITZINGER University of Agriculture Institute of Meteorology and Physics 1180 Wien AUSTRIA Tel.: +43-1-470-5820-34 Fax: +43-1-470-5820-60 E-mail :
[email protected]
Yehia Yehia HAFEZ Astronomy & Meteorology Department Faculty of Science - Cairo University Giza EGYPT Tel: +20-221-531-74 Fax: +20-257-275-58 E-mail:
[email protected]
Markus ERHARD Potsdam Institute for Climate Impact Research P.O. Box 60 12 03 14412 Potsdam GERMANY Tel: +49-331-288-2678 Fax: +49-331-288-2600 E-mail:
[email protected]
114
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Valentina KRYSANOVA Potsdam Institute for Climate Impact Research P.O. Box 60 12 03 14412 Potsdam GERMANY Tel: +49-331-288-2515 Fax: +49-331-288-2600 E-mail:
[email protected]
Paula HARRISON Environmental Change Unit Oxford University c/o 23 Meadowcroft Honley Huddersfield West Yorkshire HD7 2GJ UK Tel: +44-1484-66 03 78 Fax: +44-1484-66 03 91 E-mail:
[email protected]
Petra LASCH Potsdam Institute for Climate Impact Research P.O. Box 60 12 03 14412 Potsdam GERMANY Tel: +49-331-288-2662 Fax: +49-331-288-2600 E-mail:
[email protected]
Corine HOFF Centre d'Ecologie Fonctionnelle et Evolutive / CNRS Route de Mende F-34033 Montpellier Cedex FRANCE Tel: +33-467-61-3271 E-mail:
[email protected]
Marcus LINDNER Potsdam Institute for Climate Impact Research P.O. Box 60 12 03 14412 Potsdam GERMANY Tel: +49-331-288-2677 Fax: +49-331-288-2600 E-mail:
[email protected]
Márta HUNKÁR Hungarian Meteorological Service Budapest Kitaibel u.1 HUNGARY Tel:36-1-3464619 Fax: +36-1-3464669 E-mail:
[email protected]
Wolfgang LUCHT Potsdam Institute for Climate Impact Research P.O. Box 60 12 03 14412 Potsdam GERMANY Tel: +49-331-288-2533 Fax: +49-331-288-2600 E-mail:
[email protected]
Ana IGLESIAS Departamento de Proyectos y Planificacion Rural E.T.S. Ingenieros Agronomos Universidad Politecnica de Madrid 28040 Madrid SPAIN Tel: +34-91-3365832 Fax: +34-91-3365835 E-mail:
[email protected]
Merylyn MCKENZIE HEDGER UK Climate Impacts Programme Union House 12-16 St. Michael's Street Oxford, OX1 2DU UK Tel: +44-1865-432072 Fax: +44-1865-432077 E-mail:
[email protected]
Geoff JENKINS Hadley Centre The Meteorological Office Bracknell RG12 2SZ UK Tel: +44-1344-856-653 Fax: +44-1344-854-898 E-mail:
[email protected]
Christophe NEFF Geographisches Institut LS f. Physische Geographie u. Länderkunde Universität Mannheim L 9, 1-2, 68131 Mannheim GERMANY Tel: +49-621-181-1968 E-mail:
[email protected]
Thomas KARTSCHALL Potsdam Institute for Climate Impact Research P.O. Box 60 12 03 14412 Potsdam GERMANY Tel: +49-331-288-2680 Fax: +49-331-288-2600 E-mail:
[email protected] 115
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Patricia DE ROSNAY Laboratoire de Meteorologie Dynamique, Université P. et M. Curie Tour 25-15, 5ieme etage BP 99 4 place Jussieu 75252 Paris cedex 05 FRANCE Tel: +33-1-4427-2315 Fax: +33-1-4427-6272 E-mail:
[email protected]
Bernard NEMRY Institut d'Astrophysique et de Geophysique University de Liege 5 avenue de Cointe 4000 Liege BELGIUM Tel: +32-4-254-7577 Fax: +32-4-254-7575 E-mail:
[email protected] Svetlana NIKULINA Chief of International Cooperation Department Central Asian Research Hydromoteorological Institute (SANIGMI) K. Makhsumov st 72, Tashkent UZBEKISTAN Tel.: +998-71-235-83-29 Fax: +998-71-133-11-50/133-20-25 E-mail:
[email protected],
[email protected]
Mark ROUNSEVELL Department of Geography Place Louis Pasteur 3, Université Catholique de Louvain B-1348 Louvain-la-Neuve BELGIUM Tel. +32-(0)1047-2872 Fax. +32-(0)1047-2877 E-mail:
[email protected] Adele ROWDEN Edinburgh Centre for Carbon Management Ltd 64 The Causeway Duddingston Village Edinburgh, EH15 3PZ UK Tel: +44-131-661-6222 or +44-131-650-6480 E-mail:
[email protected]
David PRICE Natural Resources Canada Canadian Forest Service 5320 -122 Street Edmonton Alberta, T6H 3S5 CANADA Tel.: +1-780-435-7249 Fax: +1-780-435-7359 E-mail:
[email protected]
Mikhail SEMENOV IACR Long Ashton Research Station University of Bristol Long Ashton Bristol, BS41 9AF UK Tel: +44-1275-549421 Fax: +44-1275-394007 E-mail:
[email protected]
Navin RAMANKUTTY Climate, People, and Environment Program IES - Center for Climatic Research 1225 W. Dayton Street Madison WI - 53706 USA Tel: +1-608-265-8720 Fax: +1-608-263-4190 E-mail:
[email protected]
Stephen SITCH Potsdam Institute for Climate Impact Research P.O. Box 60 12 03 14412 Potsdam GERMANY Tel: +49-331-288-2536 Fax: +49-331-288-2600 E-mail:
[email protected]
Seppo REKOLAINEN Finnish Environment Institute P.O.Box 140 00251 Helsinki FINLAND Tel: +358-9-40300246 Fax: +358-9-40300291 E-mail:
[email protected]
116
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Frank WECHSUNG Potsdam Institute for Climate Impact Research P.O. Box 60 12 03 14412 Potsdam GERMANY Tel: +49-331-288-2663 Fax: +49-331-288-2600 E-mail:
[email protected]
Ben SMITH Climate Impacts Group Department of Plant Ecology Lund University Sölvegatan 37 22362 Lund SWEDEN Tel: +46-46222-4354 Fax: +46-46222-3742 E-mail:
[email protected]
Kirsten ZICKFELD Potsdam Institute for Climate Impact Research P.O. Box 60 12 03 14412 Potsdam GERMANY Tel: +49-331-288-25318 Fax: +49-331-288-2600 E-mail:
[email protected]
Roger STREET Director, Adaptation and Impacts Research Group Atmospheric Environment Service 4905 Dufferin Street Downsview Ontario M3H 5T4 CANADA Tel: +001-416 739-4271 Fax: +001-416 739-4297 E-mail:
[email protected]
Christoph ZÖCKLER World Conservation Monitoring Centre 219 Huntingdon Rd Cambridge CB3 0DL UK Tel: +44-1223-277314 Fax: +44-1223-277136 E-mail:
[email protected]
Martin SYKES Climate Impacts Group Department of Plant Ecology Lund University Sölvegatan 37 22362 Lund SWEDEN Tel: +46- 46222-9298 Fax: +46- 46222-3742 E-mail:
[email protected] John TAYLOR Mathematics and Computer Science and Environmental Research Divisions Argonne National Laboratory 9700 South Cass Avenue Bld. 221, Argonne IL 60439-4844 USA Tel: 001-630-252-7162 Fax: 001-630- 252-6104 E-mail:
[email protected] David VINER Climate Impacts LINK Project Climatic Research Unit University of East Anglia Norwich NR4 7TJ UK Tel: +44-1603-592089 Fax: +44-1603-507784 E-mail:
[email protected]
117
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
Appendix 3: Relevant web sites Climate Impacts Group, Plant Ecology, Lund University, Sweden Modelling climate impacts on terrestrial ecosystems http://www.planteco.lu.se/CIG/CIGram.html Climate Impacts LINK Project Provision of the results from the Hadley Centre’s climate change experiments (e.g. HadCM2 and HadCM3) and accompanying observational datasets. http://www.cru.uea.ac.uk/link/ Climate People Environment Program, University of Wisconsin, USA Global environmental processes, land-use, ecosystems and natural resources http://cpep.aos.wisc.edu/index.html Climatic Research Unit (CRU) Research into natural and anthropogenic climate change. http://www.cru.uea.ac.uk/ Deutsches Klimarechenzentrum (DKRZ) Service centre for climate researchers in Germany, responsible for the installation and operation of a highperformance computer centre for basic and applied research into climatology and related areas. http://www.dkrz.de ECLAT-2 Improving the understanding of the application of results from climate model simulations to studies of climate change impacts. /www.cru.uea.ac.uk/eclat Environmental Change Institute (Environmental Change Unit), University of Oxford, UK Centre for teaching and interdisciplinary research on the environment and sustainability http://www.eci.ox.ac.uk/ Environmental and Societal Impacts Group, National Centre for Atmospheric Research, Colorado, USA Environmental change and responses to such change http://www.esig.ucar.edu/esig.html European Commission Directorate General XII (DGXII) Development and implementation of European Union Policy on Research and Technological Development, promotion of public knowledge of science and technology. http://europa.eu.int/comm/dg12/ Hadley Centre Provision of an authoritative assessment of natural and anthropogenic climate change for the United Kingdom government. http://www.meto.govt.uk/sec5/sec5pg1.html Inter-governmental Panel on Climate Change - Data Distribution Centre (IPCC-DDC) Distribution of consistent and up-to-date scenarios of changes in climate and related environmental and socio-economic factors for use in climate impacts assessments.
118
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
W O R K S H O P
http://ipcc-ddc.cru.uea.ac.uk/ IPCC-Geneva Established by WMO and UNEP to assess the scientific, technical and socio-economic information relevant for the understanding of the risk of human-induced climate change. http://www.ipcc.ch/ IPCC Special Report on Emissions Scenarios (SRES) Development of emissions scenarios and scenario database. http://www.sres.ciesin.org Koninklijk Nederlands Meteorologisch Instituut (KNMI) Netherlands national research and information centre for climate, climate change and seismology; operational centre for weather and climate observation, weather forecasting and monitoring of seismic activity. http://www.knmi.nl Laboratoire de Météorologie Dynamique (LMD) Mechanisms, evolution and prediction of meteorological and climatic phenomena http://www.lmd.jussieu.fr/ Long Aston Research Station, UK The LARS-WG stochastic weather generator http://www.lars.bbsrc.ac.uk/model/larswg.html METEO-France Research and services related to weather and climate. http://www.meteo.fr/ Mountain Forest Ecology, Department of Forest Sciences, Swiss Federal Institute of Technology, Switzerland Forest ecosystems in mountainous areas http://www.fowi.ethz.ch/pgw/research.html NASA Goddard Institute for Space Studies Climate impacts on Mediterranean agriculture http://www.giss.nasa.gov/research/intro/rosenzweig.01/ Natural Resources Canada, Canadian forest Service Climate change and impacts of climate change on forestry in Canada http://climatechange.nrcan.gc.ca/english/html/index.html, http://www.nofc.forestry.ca/climate/ Potsdam Institute for Climate Impact Research (PIK) Research into the dynamics of global change, bringing together natural and socio-economic processes. http://www.pik-potsdam.de/ The VEMAP PROJECT, National Center for Atmospheric Research, Colorado, USA The Vegetation/Ecosystem Modelling and Analysis Project http://www.cgd.ucar.edu/vemap/index.html
119
E C L A T - 2
R E P O R T
N O . 2
P O T S D A M
120
W O R K S H O P
