d i s c u s s i o n pa p e r n u m B e r 9 octoBer 2010 d e v e l o p m e n t a n d c l i m at e c h a n g e d i s c u s s i o n pa p e r s 1 57558 d e v e l o p m e n t a n d c l i m a t e c h a n g e Climate Change Scenarios and Climate Data d i s c u s s i o n pa p e r n u m B e r 9 octoBer 2010 d e v e l o p m e n t a n d c l i m a t e c h a n g e Climate Change Scenarios and Climate Data Kenneth strzepek and c. adam schlosser massachusetts institute of technology Joint program for the science and policy of global change Papers in this series are not formal publications of the World Bank. They are circulated to encourage thought and discussion. The use and citation of this paper should take this into account. The views expressed are those of the authors and should not be attributed to the World Bank. Copies are available from the Environment Department of the World Bank by calling 202-473-3641. © 2010 The International Bank for Reconstruction and Development / THE WORLD BANK 1818 H Street, NW Washington, DC 20433, U.S.A. Telephone: 202-473-1000 Internet: www.worldbank.org/climatechange E-mail: feedback@worldbank.org All rights reserved. October 2010 This volume is a product of the staff of the International Bank for Reconstruction and Development / The World Bank. The findings, interpretations, and conclusions expressed in this volume do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgement on the part of the World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. RIGHTS AND PERMISSIONS The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applica- ble law. The International Bank for Reconstruction and Development / The World Bank encourages dissemination of its work and will normally grant permission to reproduce portions of the work promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center Inc., 222 Rosewood Drive, Danvers, MA 01923, USA; telephone 978-750-8400; fax 978-750-4470; Internet: www.copyright.com. Cover graphic: IPCC 4th Assessment Report, 2007. All dollars are U.S. dollars unless otherwise indicated. taBle oF contents section 1. Background 1 section 2. climate data needs for the eacc 1 section 3. climate data for the eacc global track 2 section 4. selection of climate change scenarios 4 Regional Patterns and Extremes in Climate Models: Implications for Climate Change Projections 4 section 5. use of climate change scenarios 16 section 6. summary 17 references 17 Figures 1. eacc study structure global track 2 2. cru average annual precipitation 3 3. cru average annual temperature 3 4. annual mean precipitation (cm), observed (a) and simulated (b), based on the multimodel mean 4 5. annual global precipition of ar4 gcm and model mean 5 6. examples from two climate models of projected changes in annual precipitation, evaporation, runoff, and soil moisture 7 7a. percentage change in precipitation interval (i.e. inverse of precipitation frequency), as represented by the participating climate models in the ipcc 4th assessment report 8 7b. percentage change in precipitation, as represented by the participating climate models in the ipcc 4th assessment report 9 8. average cmi 1961­90 11 9. climate m oisture index spread for each scenario and global land mass and each region 12 10. climate moisture index spread for each scenario and each World Bank region 13 11. precipitation change for 2050 for a2 sres csiro gcm 14 iv c limate c h an ge sc e n a r ios an d c limate c h an ge d ata 12. precipitation change for 2050 for a2 sres ncar gcm 14 13. maximum temperature change for 2050 for a2 sres csiro gcm 15 14. maximum temperature change for 2050 for a2 sres ncar gcm 15 15. SectionoftheSARRegionShowing0.5°by0.5°(historichydrology)GridScale and2.5°by2.5°(GCM)GridScaleRelativetoCatchmentBoundaries 17 Tables 1. sector analysis climate data needs 2 2. LatitudebyLongitudeAreas(0.5°x0.5°) 2 d e v e l o p m e n t a n d c l i m at e c h a n g e d i s c u s s i o n pa p e r s 1 1. Background 2. climate data needs for the eacc The overall objective of the EACC study is to better understand what adaptation to climate change really is Climate change impacts that will require adaptation and how--without such adaptation--development include (a) increased incidence of extreme weather progress will be threatened and may even be reversed. events, which will draw resources away from develop- The study has two broad objectives. The first objective ment initiatives; (b) increased incidence of infectious is to develop a "global" estimate of adaptation costs. and diarrheal diseases, which will reverse development This will aid the international community's efforts to gains in health standards; (c) changes in temperature help those developing countries most vulnerable to and precipitation, which will affect agricultural produc- climate change meet adaptation costs. The second tivity, making investments in this sector less productive: objective is to help decision makers in developing coun- and (d) sea-level rise, which will lead to loss of lives and tries better understand and assess the risks posed by assets. Under the global track, adaptation costs for all climate change and create better design strategies to developing countries are estimated by major economic adapt to climate change. sectors using country-level data sets that have global coverage. The agriculture, forestry, fisheries, infrastruc- These two objectives require any quantitative scientific ture, water resources, coastal zones, health, and ecosys- or economic analysis to (a) be spatially comprehensive, tem services sectors are covered. The analysis also with globally comprehensive and consistent data sets; considers the cost implications of changes in the and (b) include plausibly extreme climate projections frequency of extreme weather events, including the that span the plausible futures. implications for social protection programs (Figure 1). Due to the very large uncertainties of the scientific The climate data needed for these analyses are summa- projections of climate change, it is important that any rized in Table 1. analysis provide information on the "tails" or extremes of any probability distribution of future climates. This The majority of sectors use a consistent set of future allows decision makers in developing countries to better climate and water runoff projections to establish the understand and assess the risks posed by climate change. nature of climate change. A consistent set of GDP and population projections also are used to establish a base- This document reports on the rationale for selection of line of development in the absence of climate change. climate change scenarios and climate data for the The analysis subsequently estimates the economic and EACC global analysis. social impacts and the corresponding costs of adaptation. 2 c limate c h an ge sc e n a r ios an d c limate c h an ge d ata For historical climate data, we used a database provided figure 1. eacc study structure gloBal by the Climate Research Unit (CRU) at the University track of East Anglia in the United Kingdom. The CRU 2.1 data set provides a time series of monthly climate vari- Projections ables from 1901 to 2002 (http://www.cru.uea.ac.uk/cru/ Climate Projections Sectors Water Run-off Global data sets data/hrg/cru_ts_2.10). We extracted a subset of monthly Agriculture Baseline GDP/ climate variables required to compute potential evapo- Forestry Population Fisheries transpiration (average daily T minimum, average daily T Infrastructure maximum, vapor pressure, cloudiness) and precipitation Water Resources Economic and . This data, provided on a 0.5° longitude/latitude grid, is Coastal Zones social impacts the most frequently used standard reference "baseline" Health Ecosystem Services for global, regional, and local climate change impact studies. Excluding Antarctica, there are 67,420 grids Identification of Cross-Sectors adaptation measures (0.5° x 0.5°) over the global land area. The 0.5° degree Extreme weather events longitude/latitude grids are not constant in area due to Social Decision rule the spherical shape of the earth. Table 2 provides infor- Cost of adaptation mation on the size of the grid cells from the equator to the poles. Source: 2009. eacc-the cost of developing countries of adapting to climate change. taBle 2. latitude By longitude areas (0.5° x 0.5°) taBle 1. sector analysis climate data needs 0.5 degree 0.5 degree 0.25 square longitude latitude degrees Sector Climate Variables Time Step latitude km km km^2 agriculture tmin, tmax, precipitation daily, monthly 0 55.5 55.5 3080 Water resources tavg, precipitation monthly +/-40 42.5 55.5 2359 health tmax, precipitation daily, monthly +/-60 28 55.5 1554 infrastructure tmax, precipitation max monthly +/-80 8.5 55.5 472 Forestry tmin, tmax, precipitation monthly coastal sea level Yearly A visual presentation of the annual averages (1901 to 2002, 102 years) of precipitation and temperature is displayed in Figures 2 and 3. This provided a database used by all sectors. For some sectors, additional sources of daily data were needed. 3. climate data for the eacc These data sources and characteristics are described in gloBal track the sectoral reports. The analysis required a consistent global data set that covered data needs across all sectors and provided data for the entire global landmass with a time series of at least 30 years of monthly data. After examining a number of options, the team selected the CRU TS2.1 data set. d e v e l o p m e n t a n d c l i m at e c h a n g e d i s c u s s i o n pa p e r s 3 figure 2. cru average annual PreciPitation Base Annual Precipitation (mm) 3000 2500 2000 1500 1000 500 0 Source: authors. figure 3. cru average annual temPerature Base Annual Average Temperature (Deg. C) 50 40 30 20 10 0 -10 -20 -30 Source: authors. 4 c limate c h an ge sc e n a r ios an d c limate c h an ge d ata 4. selection of climate change ocean physics made by the model building team. Embedded within all these processes is also the ability scenarios of any given climate model to faithfully represent the diurnal cycle of these mechanisms. The IPCC 4th Assessment (AR4) has archived climate change scenarios for three SRES scenarios and twenty- two global circulation models, resulting in fifty-six climate change scenarios. This was far too many scenar- figure 4. annual mean PreciPitation ios to perform a global multisectoral adaptation analysis. (cm), Observed (a) and simulated (b), This section aims to answer the following question: Is Based on the multimodel mean there information from the climate science community that may assist in filtering 56 GCMS to a more manageable number? regional patterns and extrem es in climate model s : im plicat ions For climate chan ge p roJe ctions Evaluation of Models of the Present Climate An extensive volume of published research has evalu- ated climate models that simulated past climate varia- tions and trends and predicted climate anomalies at the seasonal to inter-annual scales. We clearly have no shortage of evidence to indicate that climate models have the capability to simulate continental-scale coher- ent patterns of temperature--and to a lesser degree precipitation--and anomalies associated with large-scale climate phenomenon such as the El Niño/Southern Oscillation. However, key uncertainties--that is, disagreements with observations among climate models--still exist regarding the precise location, timing, and/or magnitude of temporal climate varia- tions. One only needs to look at global precipitation Source: ipcc 4th assessment report, Figure 8.5. patterns to recognize these quantifiable deficiencies in Notes: the climate prediction center merged analysis of pre- climate models. Among the most prominent shortcom- cipitation (xie and arkin 1997) observation-based climatology for ings of most climate models (Figure 4 shows the mean 1980 to 1999 is shown. the model results are for the same peri- of the IPCC AR4 climate models) is the notable split od in the 20th-century simulations in the mmd at pcmdi. in (a), in maximum precipitation in the eastern tropical Pacific, observations were not available for the grey regions. and to a lesser extent, in the western tropical Atlantic basin. In addition, the location and magnitude of the Indo-Asian Monsoon, as well as the South American rain-forest maximum precipitation as given by the model mean, disagree with observations. The model- mean results shown here also average out considerably larger differences in regional patterns (especially in the tropical regions) from individual models. These larger differences among the individual models (Figure 5) stem largely from assumption about atmospheric and d e v e l o p m e n t a n d c l i m at e c h a n g e d i s c u s s i o n pa p e r s 5 figure 5. annual gloBal PreciPition of ar4 gcm and model mean Annual Precipitation Source: ipcc 4th assessment report, Figure s8.9. Notes: note that the observed (top left map) and model-mean (top center map) are also shown. This is of critical importance for convective precipita- Evaluation Implications for Climate Change Projections tion processes. At present, all climate models show When considering the ability of climate models to considerable deficiencies in representing the observed accurately reproduce observed climate trends and anom- night-time maximum precipitation (most models release alies, and how these results may then be confidently too much precipitation far too early, with a maximum applied to projected (and potential) changes in climate occurring in the late afternoon). Nevertheless, we can be over the next century, we can draw some insight and confident that the fundamental physics explicitly repre- guidance from the following passage in chapter 8 of the sented in climate models is sound, and the information IPCC 4th Assessment Report: they give us--as a whole or as a continuum of plausible outcomes--is useful. We must consider all their projec- tions in the context of assessing potential impacts on "What does the accuracy of a climate model's simula- and responses from our natural and managed systems. tion of past or contemporary climate say about the 6 c limate c h an ge sc e n a r ios an d c limate c h an ge d ata accuracy of its projections of climate change? This ques- from this practice, simply because a model "average" of tion is just beginning to be addressed... studies show near zero is not necessarily the result of a combination promise that quantitative metrics for the likelihood of of models predicting a near-zero change, but could also model projections may be developed, but... the develop- be a result of two opposing changes that differ in sign ment of robust metrics is still at an early stage..." (as seen in Figure 6). This could result in a substantial risk and/or consequence, and the abatement of the As these remarks suggest, we cannot make a direct risk--no matter how slight the chance of climate inference that because climate models can reproduce change might be--may be worthwhile to consider for large-scale observed climatologies and accurately predict policy and/or mitigation strategies. (or reproduce) climate anomalies , we should then have complete confidence in any one climate model or even a We must therefore consider all possible outcomes that collection of climate models. This even holds for models are being produced by climate models, and feed these that have been run at an exceptionally high level of changes individually through impact assessment models, spatial detail (such as the MIROC-Hires). Experience which may also provide a critical coupled feedback to in climate evaluation, particularly in the area of hydro- the climate models via land-use change or water fluxes. logic variables, indicates that there is no "best model" While the potential exists for the range of simulated and we must consider every model simulation individu- climate-change projections to vary so much among one ally. A simple example of this with respect to simulated another that they could be construed as "noise," there is climate change can be seen in the projected changes in evidence to suggest that even when considering the precipitation, evaporation, runoff, and soil moisture collection of model outcomes, we see a consistency in shown in Figure 6. In the top set of panels, we see that some of the more salient changes. As an example, the the MIROC-Hires model would indicate a dramatic trend in precipitation interval as simulated by the IPCC decrease in precipitation over much of the southern AR4 models show--statistically and/or probabilistically United States (extending north in the middle section of speaking--agreement in their projection of the change U.S.), which would pose a significant concern for water in precipitation interval across latitudes as the climate resource impacts. However, looking at the same result warms (Figures 7a, 7b). from the CCSM3 model, there is a complete reversal in the sign of the precipitation response over much of the What this implies is that although we cannot tie southern U.S. (as well as the middle U.S.). The only ourselves to one particular model outcome, we can have area of consistency in sign to the MIROC-Hires result some confidence that we are not merely looking at is in the southwestern U.S., but the CCSM precipita- "noise," and there are regions where a sign change is tion decrease is considerably reduced. A similar situa- consistent among the climate models--but with a range tion can be seen for parts of Africa (MIROC showing that is important to explicitly consider for assessment of large decreases, CCMS showing increases over much of potential impacts and adaptation. Africa). Similarly, the changes in runoff (which can be regarded as the "available surface water") show similar While an approach that explicitly considers the range of marked discrepancies. possible climate-change outcomes as projected by global climate models is desirable, it is important to carefully This simple example in the mean annual fields of the consider whether there are some projections that should major hydrologic variables raises an important issue be filtered out as nearly implausible or at the very least, regarding impact and response assessments for water extremely unlikely. This is a difficult task, as there is no resources. If annual mean fields of changes show clear metric with appropriate or sufficient data that can marked differences in sign and magnitude among identify whether a given climate change projection can climate models, the model average of these changes, not or should be excluded, particularly when considering a only at annual scales but also down to subdiurnal scales, specific region of interest. Further, if there is any chance is not an adequate input for impact, response, and adap- that any given climate change simulation can occur, it tation assessments. To date, many studies have used the should be considered within the spectrum of potential model mean as inputs, but it is important to move away impact and/or adaptation--particularly if the risk of CCSM3 MIROC3.2 (HIRES) d e v e l o p m e n t a n d c l i m at e c h a n g e d i s c u s s i o n pa p e r s precipitatiOn, evapOratiOn, runOFF, and sOil mOisture Source: Figures obtained from ch. 10, ipcc ar4 supplementary materials (accessible at: ). Note: changes are annual means for the sres a1B scenario for the period 2080 to 2099 relative to 1980 to 1999. 7 8 c limate c h an ge sc e n a r ios an d c limate c h an ge d ata figure 7a. Percentage change in PreciPitation interval (i.e. inverse of PreciPita- tiOn Frequency), as represented by the participating climate mOdels in the ipcc 4th assessment rePort Change in Precipitation Event Interval (%/°C) Source: schlosser et al. 2008. Note: shown are the zonally averaged percent changes in the period between "wet" days per degree of global warming for each cli- mate model. results are based on differences between pre-industrial conditions and the time of co2 doubling. d e v e l o p m e n t a n d c l i m at e c h a n g e d i s c u s s i o n pa p e r s 9 Figure 7b. percentage change in precipitatiOn, as represented by the ParticiPating climate models in the iPcc 4th assessment rePort Source: ipcc 2007. Note: results based on differences between pre-industrial conditions and the time of co2 doubling. 10 c limate c h an ge sc e n a r ios an d c limate c h an ge d ata catastrophic damages and/or consequences is high. is greater than PET, the climate is moist. Calculated as Therefore, it may be a worthwhile effort at the outset of CMI = (P/PET)-1 {when PET>P} and CMI = any risk/impact assessment to consider all climate 1-(PET/P) {when P>PET}, a CMI of ­1 is very arid model projections (at very little computational cost), and a CMI of +1 is very humid. As a ratio of two depth and that the exclusion of certain climate change scenar- measurements, CMI is dimensionless. The CMI can be ios might be considered if the costs and/or damage of evaluated at each of the 67420 grid cells of the CRU its associated impacts were very low. It may also high- database (Figure 8). light outcomes that are beneficial, and these also are important in providing a comprehensive assessment of It is a straightforward calculation to determine the climate impacts and risks. Overall, the method of using CMI for each of the 56 GCMs. For the EACC, it was model-mean as inputs to impact and risk assessments decided to use the average changes in monthly precipi- should be replaced with the usage of the full spectrum tation and temperature for the years 2046 to 2055 for of results, as the "tails" of the full spectrum of model each GCM. These average monthly changes were projections may contain the riskiest aspects of climate applied to the CRU 1961­90 average monthly tempera- change. ture and precipitation to determine a CMI for each CGM. The CMI can be averaged over any spatial Criteria for Selection of Climate Change Projections region. It was decided to select the two GCMs that Given this context and the EACC study objective to represented the wet and dry GCM in 2050 over the help decision makers in developing countries to better land mass of the globe. Wet is defined as the GCM that understand and assess the risks posed by climate had the largest increase in average CMI over the 1961­ change, the team agreed to select two GCM scenarios 90 CMI over the globe, while dry was defined as the that represented the extremes of climate change impacts CGM that had the largest decrease in average CMI on the globe. This section describes the criteria that over the 1961­90 CMI over the globe. should be used to evaluate climate change impacts. The reason the globe was selected rather than just over The key economic sectors being evaluated in the EACC developing countries is that for the agricultural sector include agriculture, water resources, infrastructure, analysis, global food trade is a key element. Climate ecosystems, and health. All sectors are heavily influ- change impacts on agriculture in the developed world is enced by soil moisture. Therefore the Climate Moisture a very important part of any analysis of climate change Index (CMI), which is an indicator of the aridity of a impacts on food, agriculture, and hunger in the develop- region, was selected as an indicator to measure climate ing world. change impacts that would best represent impacts on the key sectors. The CMI depends on average annual Figure 9 shows the range of CMI for all scenarios for precipitation and average annual potential evapotranspi- the globe and World Bank Regions as a whole and the ration (PET).1 If PET is greater than precipitation, the remaining land mass. Figure 10 show the CMI for all climate is considered to be dry, whereas if precipitation scenarios individually for each World Bank region. The red line represents the median CMI. The top of the box represents the 25th percentile, while the bottom of the box represents the 75th percentile. The whiskers show 1. Average annual PET is a parameter that reflects the amount of water lost via evaporation or transpiration (water consumed by vegetation) the extremes and the cross-hairs show the model outli- during a typical year for a given area if sufficient water were available ers. The dashed lines represent the historical CMI at all times. Average annual evapotranspiration (ET) is a measure of the amount of water lost to the atmosphere from the surface of soils and (averaged from 1960­90). For example, in the LCR plants through the combined processes of evaporation and transpira- region, there is a 75 percent chance of drying with all tion during the year (measured in mm/yr). ET, which is both connected to and limited by the physical environment, is a measure that quantifies three scenarios. The CMI for the SAR region has the the available water in a region. Potential evapotranspiration is a calcu- largest spread because of the way the different GCMs lated parameter that represents the maximum rate of ET possible for an area completely covered by vegetation with adequate moisture avail- model the monsoons. In the MNA region, there is not able at all times. PET is dependent on several variables, including tem- perature, humidity, solar radiation, and wind velocity. If ample water is much variation because the area is so dry. available, ET should be equal to PET. d e v e l o p m e n t a n d c l i m at e c h a n g e d i s c u s s i o n pa p e r s 11 figure 8. average cmi 1961­90 Base Climate Moisture Index -- Annual Average Source: authors 2010. It is important to note that the CMI is only calculated Selection of Climate Change Projections over land masses and not over the ocean. Many climate Figures 9 and 10 show that the 56 GCMs provide for a change analyses discuss GCMs with regard to their large range of CMI for the globe, developing countries, properties/results over land and sea, but for hydrology and each individual World Bank region. However, for and water resources impacts all that matters is what is reasons related to socioeconomic baseline consistency, occurring over the land. One may notice that all the the EACC team decided to limit GCM scenarios to CMI in the box and whiskers plots are negative. These those run for the A2 SRES scenario. This reduced the results show that warming leading to increases in possible number of GCMs to 17. potential evapotranspiration dominates over increased precipitation over the globe's land mass, leading to an A further constraint was placed on the suite of candi- increase in aridity or a drying for more than 75 percent date GCMs. The agricultural sector required that the of all GCMs. GCM selected must provide monthly Tmin and Tmax to the AR4 archive. This limited the number of SRES A2 to six that met all constraints. 12 c limate c h an ge sc e n a r ios an d c limate c h an ge d ata Figure 9. climate mOisture index spread FOr each scenariO and glObal land mass and each region Source: c. J. Willmotta and J. J. Feddema. 1992. "a more rational climatic moisture index." The Professional Geographer 44.1: 84­88. d e v e l o p m e n t a n d c l i m at e c h a n g e d i s c u s s i o n pa p e r s 13 Figure 10. climate mOisture index spread FOr each scenariO and each wOrld bank region Source: c. J. Willmotta and J. J. Feddema. 1992. "a more rational climatic moisture index." The Professional Geographer 44.1: 84­88. 14 c limate c h an ge sc e n a r ios an d c limate c h an ge d ata Based on the global CMI analysis, the CSIRO climate scenario. Figures 11 and 12 are a plot of the MK_3_0 was selected as the EACC dry climate annual precipitation changes in 2050 for the dry and scenario and NCAR_CCSM was selected as the wet wet scenarios, respectively. figure 11. PreciPitation change for 2050 for a2 sres csiro gcm figure 12. PreciPitation change for 2050 for a2 sres ncar gcm Source: Figure 11, 12, 13, and 14 maps are based on data developed at the massachusetts institute of technology Joint program for the science and policy of global change using the Wcrp's cmip3 multimodel dataset. maps were produced by the international Food policy research institute. d e v e l o p m e n t a n d c l i m at e c h a n g e d i s c u s s i o n pa p e r s 15 Figures 13 and 14 are plots of the annual maximum temperature, calculated modified Penman-Monteith temperature changes in 2050 for the dry and wet potential evapotranspiration (PET), and calculated scenarios, respectively. CMI for 2030 and 2050. Annual results are in Appendix A, seasonal results are in Appendix B, and A series of appendices are provided with this document monthly results are in Appendix C. that provide detail plots for global precipitation, Figure 13. maximum temperature change FOr 2050 FOr a2 sres csirO gcm Figure 14. maximum temperature change FOr 2050 FOr a2 sres ncar gcm Figure 13 & 14: source: maps are based on data developed at the massachusetts institute of technology Joint program for the science and policy of global change using the Wcrp's cmip3 multimodel dataset. maps were produced by the international Food policy research institute. 16 c limate c h an ge sc e n a r ios an d c limate c h an ge d ata 5. use of climate change downscaling algorithms, resulting in the need for exhaustive numerical experimentation. Time and cost scenarios constraints often do not allow use of more than a couple of GCMs in downscaling exercises. Running The spatial and temporal scale of impact and adaptation multiple GCMs at a coarse resolution may provide analyses needed for the EACC sectoral analyses span a more insight into the range of possible futures than wide range from very small (1­10 km2 and daily to more detailed information obtained from fewer GCMs weekly) for local village water supply to very large catchments (100,000 km2 and monthly to yearly) for major reservoirs. Climate change will occur at local There is no one "best" method; the most appropriate scales, but models currently used for projecting climate method for a particular application will strike a careful change due to future GHG emissions have an average balance between precision (resolution) and accuracy resolution of 2.6° x 3.0°. One potential difficulty with (confidence in projections). Figure 15 provides a visual using climate information in impact assessments is the representation of the trade-off between precision and mis-match between the low spatial (and temporal) reso- accuracy. As resolution increases, so too does the uncer- lution of GCMs, on the one hand, and the scale at tainty associated with the more detailed information. In which assessments need typically to be conducted, on other words, more "precise" information comes at a cost, the other. GCMs provide climate change projections at and the additional uncertainty must be recognized and a low spatial resolution (~2.5° x 2.5° grid), while the taken into account in assessing impacts. Given the EACC is using a much finer resolution (~0.5° x 0.5° trade-off, it is critical to establish at the outset of any grid ) impact assessment whether the goal is to have finer resolution or "better"--that is, more There are several methods available for addressing scale reliable--information. issues, including statistical downscaling (using empirical relationships), dynamical downscaling (using regional Sub-GCM Grid Scale Climate Change Projections climate models), and spatial techniques2 (linear interpo- For this work the spatial resolution for the use of GCM lation, krigging, spline fitting, and intelligent interpola- was at the native grid scale of each GCM. The GCM tion). Downscaling involves methods used to map the provided relative changes in temperature and precipita- large-scale signals from GCMs to a finer resolution tion for the years 2030 (average over 2026 to 2035) and (tens of kilometers versus hundreds of kilometers). 2050 (average over 2046 to 2055) on a monthly level as compared to the model baseline (1961­90) of the 20th Care needs to be taken in selecting a method. Beyond century. These relative changes were then applied reproducing the underlying uncertainties of GCMs, directly to the historic climate variables from the CRU many introduce additional uncertainty and biases. For data set. As seen in Figure 15, there are numerous half example, downscaling techniques increase the detail of degree by half degree CRU grids within each GCM information, but also the uncertainties associated with grid box. For this analysis, we apply the relative changes that information since the GCM output is manipulated from the GCM grid box uniformly to all CRU grid below the scale at which the physics of the GCM itself cells within the GCM grid box. While this leads to are mathematically described. Under some downscaling some discontinuities at the border of grid boxes, it is the schemes, mass balances of water and energy over the most appropriate technique for this work given the GCM scale are violated by the downscaling algorithm. scope and mathematics of the GCM models. Use of dynamic and statistical downscaling techniques requires extensive quantification of the sensitivities of With the uncertainty inherent in the GCM and the the underlying assumptions of both the GCMs and the CRU data, it would be unwise to perform this analysis at the lowest level of resolution; half degree by half degree. Aggregating to a higher spatial level would reduce the uncertainty in the model indicators and 2. "Spatial technique" is often referred to as "spatial downscaling." Technically, it does not involve downscaling, but rather statistical and more correctly reflect the larger scale climate change spatial relationships. The majority of "downscaling" being done is with this method. projections from the GCM models. For the agricultual d e v e l o p m e n t a n d c l i m at e c h a n g e d i s c u s s i o n pa p e r s 17 Figure 15. sectiOn OF the sar regiOn shOwing 0.5° by 0.5° (histOric hydrOlOgy) grid scale and 2.5° by 2.5° (gcm) grid scale relative tO catchment bOundaries sector analysis, a large-scale food producing unit (FPU) scenarios support the objective of the EACC study--"to is the unit of analysis; for the water resources analysis, a better understand what adaptation to climate change catchment scale (refer to Figure 15) is used; and for really is and how without such adaptation, development infrastructure and health, a national-scale analysis is progress will be threatened and may even be reversed." used. These scenarios are described in detail in the sectoral analysis reports. 6. summary references Adler, R.f., G.J. Huffman, A. Chang, et al. 2003. "The This report has documented the process used by the Version-2 Global Precipitation Climatology Project EACC team to select climate change scenarios for use (GPCP) monthly precipitation analysis (1979-present)." in the sectoral analyses that were "spatially comprehen- J. Hydrometeor. 4 (6): 1147­1167. sive" and "plausible extreme climate projections." The 18 c limate c h an ge sc e n a r ios an d c limate c h an ge d ata Forest, C. E., P.H. Stone, and A.P. Sokolov. 2008. change assessments. Report #147. Cambridge, MA: "Constraining Climate Model Parameters from MIT Joint Program for the Science and Policy of Observed 20th Century Changes." Tellus 60A: Global Change. Accessible at: globalchange.mit.edu 911­920. Schlosser, C. A., M.D. Webster, and C. E. Forest. Hsu, K., X. Gao, S. Sorooshian, and H.V. Gupta. 1997. (forthcoming). A Probabilistic Description of Precipitation "Precipitation estimation from remotely sensed infor- Frequency Change in Climate Models. Cambridge, MA: mation using artificial neural networks." Journal of MIT Joint Program for the Science and Policy of Applied Meteorology 36: 1176­1190. Global Change. Accessible at: globalchange.mit.edu Huffman, G. J., R. F. Adler, D. T. Bolvin, G. Gu, E. Sokolov, A. P., P.H. Stone, C.E. Forest, R. Prinn, M.C. Nelkin, K. P. Bowman, Y. Hong, E. F. Stocker, and D. B. Sarofim, M. Webster, S. Paltsev, C.A. Schlosser, D. Wolff. 2007. "The TRMM Multisatellite Precipitation Kicklighter, S. Dutkiewicz, J. Reilly, C. Wang, B Felzer, Analysis (TMPA): Quasi-global, multiyear, combined- and H.D. Jacoby. 2008. "Probabilistic forecast for 21st sensor precipitation estimates at fine scales." J. century climate based on uncertainties in emissions Hydrometeor. 8: 38­55. (without policy) and climate parameters." J. Climate (submitted). Iman, R. L., and J.C. Helton. 1988. "An Investigation of Uncertainty and Sensitivity Analysis Techniques for Stern, David I. 2006. "Reversal of the trend in global Computer Models." Risk Analysis 8(1): 71­90. anthropogenic sulfur emissions." Global Environmental Change 16: 207­220. IPCC. (Intergovernmental Panel on Climate Change). 2007. "Climate Change 2007: The Physical Science Stern, David I. 2005. "Beyond the Environmental Basis," Contribution of Working Group I to the Fourth Kuznets Curve: Diffusion of Sulfur-Emissions-Abating Assessment Report of the Intergovernmental Panel on Technology." Journal of Environment and Development Climate Change (Cambridge University Press, 14 (1): 101­124. Cambridge, UK). Webster, M., S. Paltsev, J. Parsons, J. Reilly, and H. Joyce, R. J., J. E. Janowiak, P. A. Arkin, and P. Xie. 2004. Jacoby. 2008. Uncertainty in Greenhouse Emissions and "CMORPH: A method that produces global precipita- Costs of Atmospheric Stabilization. Report #165. tion estimates from passive microwave and infrared data Cambridge, MA: MIT Joint Program for the Science at high spatial and temporal resolution." J. Hydrometeor. and Policy of Global Change. Accessible at: Paltsev, S., J.M. Reilly, H.D. Jacoby, R.S. Eckaus, J. McFarland, M. Sarofim, M. Asadoorian, and M. Xie, Pingping, and Phillip A. Arkin. 1997: Global Babiker. 2005. The MIT Emissions Prediction and Precipitation: A 17-Year Monthly Analysis Based on Policy Analysis (EPPA) Model: Version 4. Report #125. Gauge Observations, Satellite Estimates, and Numerical Cambridge, MA: MIT Joint Program for the Science Model Outputs. Bull. Amer. Meteor. Soc., 78, 2539- and Policy of Global Change. Accessible at: global- 2558. change.mit.edu Schlosser, C. A., D. Kicklighter, and A. Sokolov. 2007. A global land system framework for integrated climate- The World Bank Group 1818 H Street, NW Washington, D.C. 20433 USA Tel: 202-473-1000 Fax: 202-477-6391 Internet: www.worldbank.org/climatechange