d i s c u s s i o n pa p e r n u m B e r 8 septemBer 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 56801 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 Modeling the Impact of Climate Change on Global Hydrology and Water Availability d i s c u s s i o n pA p E r n u m b E r 8 sEptEmbEr 2010 E c o n o m i c s o f A d A p t A t i o n t o c l i m A t E c h A n g E Modeling the Impact of Climate Change on Global Hydrology and Water Availability Kenneth m. strzepek and Alyssa l. mccluskey University of Colorado and Joint Program on the Science and Policy of Global Change Massachusetts Institute of Technology 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 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. September 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: All images © Shutterstock Images, LLC. All dollars are U.S. dollars unless otherwise indicated. iii Contents Acronyms vi 1. introduction 1 2. frAmEWorK of thE AnAlysis 2 Procedures and Rationale 2 Resolution and scale 2 Model scenarios 4 Reporting 6 Hydrologic variables useful to water planning and investment 7 Indicators based on time series ­ New approach 7 SRES and GCMs 7 Future years 8 Geographic representation 9 3. hydrologic driVErs And dAtA 11 Historical Climate 11 Historical Observed Runoff 11 Historical Modeled Runoff 11 4. sElEctEd climAtE chAngE scEnArios 13 5. runoff 16 Methodology 16 Historical Results 16 Climate Change Results 17 6. bAsin yiEld 18 Methodology 18 Historical Results 20 Climate Change Results 20 iv m od Eling th E impAct of climAtE ch Ang E on g lobA l hydrology And WAt E r AVAilA b i l i t y 7. summAry of rEsults 21 8. conclusions 25 rEfErEncEs 26 AppEndiX: rEfErEncE for modEls And dAtA 28 CLIRUN-II Rainfall Runoff Model 28 Boxes 2.1 Koppen-geiger climate classifications 10 Figures 2.1 description of modeling scale 3 2.2 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 5 2.3 relative change from historical climate for different gcms and average 6 2.4 total global cumulative co2 emissions from 1990 to 2100 and histogram of their distribution by scenario groups 8 2.5 Koppen-geiger climate classifications for World bank regions 10 3.1 mean annual runoff in Afr region gridded at 0.5º latitude/longitude resolution (University of New Hampshire) 12 4.1 climate moisture index spread for each scenario and global land mass and for each region 15 4.2 climate moisture index spread for each scenario and for each region (Willmott and feddema) 15 5.1 historical annual runoff in Afr region (mm/yr) 17 5.2 projected change in annual runoff in Afr region (percent) 17 6.1 impacts of evaporative losses on the storage-yield curve for lake nasser 18 6.3 impacts of the gcm scenarios on the storage-yield curves for nine regions in china 19 6.2 impact of climate change on reservoir yield and adaptations 19 6.4 historical annual basin yield in Afr region (mm/year) 20 6.5 projected changes in annual basin yield in Afr region (percent) 20 c.1 clirun-ii conceptual hydrologic model schematic 29 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 v TaBles 2.1 spatial resolution of ipcc fourth assessment report (Ar4) archived gcms 2 2.2 1º latitude by 1º longitude areas 3 2.3 Available models, scenarios, and variables in ipcc Ar4 9 3.1 sources of data in this study 11 4.1 gcm and associated base cmis used for each scenario and for regions EAp, EcA, and lAc 14 4.2 gcm and associated base cmis used for each scenario and for regions mnA, sAr, and Afr 14 7.1 runoff 21 7.2 10% flood exceedence 22 7.3 90% low flow 23 7.4 baseflow 23 7.5 10% basin yield 24 7.6 Water deficit index 24 vi ACronyms AR4 Fourth Assessment Report of IPCC AFR Africa (World Bank Region) AAA analytic and advisory activities CLIRUN Climate and Runoff Model CMI Climate Moisture Index CRU Climate Research Unit ECHAM4 Fourth-Generation Atmospheric General Circulation Model EAP East Asia and Pacific (World Bank Region) ECA Europe and Central Asia (World Bank Region) ET evapotranspiration GCM global climate model GFDL Geophysical Fluid Dynamics Laboratory GRDC Global Runoff Data Center HadCM2 Hadley Centre Model INM Institute for Numerical Mathematics IPCC Intergovernmental Panel on Climate Change LAC Latin America and the Caribbean (World Bank Region) MNA Middle East and North Africa (World Bank Region) MIROC Model for Interdisciplinary Research on Climate PET potential evapotranspiration SAR South Asia (World Bank Region) SRES Special Report on Emissions Scenarios of IPCC UNH University of New Hampshire USGS United States Geological Survey WMO World Meteorological Organization 1 1. IntroDUCtIon objective of the AAA is to provide analytical, intellec- tual, and strategic support to Bank operations and client countries in order to help them make sound water Climate change can have a profound impact on the investment decisions that account for climate variability water cycle and water availability at the global, regional, and change. The output of this AAA will be a series of basin, and local levels. Indeed, according to the recent reports that will address a number of key questions, Technical Report on Climate Change and Water from such as the impacts of climate variability and change on the Intergovernmental Panel on Climate Change water systems, both natural and engineered; adaptation (IPCC), "Observational records and climate projections strategies to reduce vulnerability of water systems to provide abundant evidence that freshwater resources are these impacts; and how the Bank can assist client coun- vulnerable and have the potential to be strongly tries in making informed decisions in this sector. impacted by climate change, with wide ranging conse- quences on human societies and ecosystems" (Bates et The objective of this technical report is to provide the al. 2008, p. 4). Developing countries are particularly background to the methodology used to model the vulnerable, and estimates show that the negative impact of climate change on runoff for the Global economic impacts of climate change could be significant Track of the EACC project. This report will present (Stern 2006). Appropriate adaptation strategies, findings from computer modeling of the impacts of however, could mitigate some of the adverse impacts of potential climate change on hydrology and water avail- climate change on the water sector in these highly ability (that is, changes in runoff, basin yield, and vulnerable countries. flooding). The World Bank recognizes water as a key affected Chapter 2 provides the framework of analysis. Chapter sector, and potential strategies for adapting to climate 3 provides the hydrological drivers and data. Chapter 4 change have become central to the dialogue on water describes selected climate scenarios. Chapter 5 provides policy reforms and investment programs with client the runoff. Chapter 6 describes the basin yield. Chapter countries. In order to support this process and future 7 summarizes the work and discusses future work. The World Bank initiatives, the Water Anchor is undertak- Appendixes describe the Climate and Runoff ing a two-year analytic and advisory activity (AAA) (CLIRUN) Model. (FY08­09) on Water and Climate Change. The main 2 2. FrAmeWorK oF tHe AnALysIs tAbLe 2.1. sPAtIAL resoLUtIon oF IPCC FoUrtH Assessment rePort (Ar4) Pro Ce DU res An D rAt IonALe ArCHIveD GCms Area at 40º resolution and scale GCM LAT LONG (km2) bccr_bcm2_0 2.81 2.81 75,115 The spatial and temporal scale of hydrologic analyses cccma_cgcm3_1 3.75 3.75 133,538 needed for World Bank investments span a wide range cccma_cgcm3_1_ 2.81 2.81 75,115 from very small (1­10 km2 and daily to weekly) for local t63 village water supply to very large catchments (100,000 cnrm_cm3 2.81 2.81 75,115 km2 and monthly to yearly) for major hydropower reser- csiro_mk3_0 1.88 1.88 33,384 voirs. Climate change will occur at local scales, but pres- csiro_mk3_5 1.88 1.88 33,384 ently models used for projecting climate change due to gfdl_cm2_0 2 2.5 47,480 future greenhouse gas emissions have an average global gfdl_cm2_1 2 2.5 47,480 climate model (GCM) resolution of 2.6° x 3.0°. The highest giss_aom 3 4 113,952 resolution belongs to the Japanese MIROC (Model for giss_model_e_h 3.91 5 185,792 Interdisciplinary Research on Climate) Hires at 1.13° X giss_model_e_r 3.91 5 185,792 1.13° and the lowest resolution belongs to the Russian INM iap_fgoals1_0_g 3 2.81 80,123 (Institute for Numerical Mathematics) at 4.0° X 5.0°. inmcm3_0 4 5 189,920 ipsl_cm4 2.5 3.75 89,025 One potential difficulty with using climate information miroc3_2_hires 1.13 1.13 12,018 in impact assessments, including in the water sector, is miroc3_2_medres 2.81 2.81 75,115 the "mismatch" between the low spatial (and temporal) mpi_echam5 1.88 1.88 33,384 resolution of GCMs and the scale at which assessments mri_cgcm2_3_2a 2.81 2.81 75,115 need typically to be conducted for investment purposes. GCMs provide climate change projections at a low ncar_ccsm3_0 1.41 1.41 18,779 spatial resolution (~2.5° x 2.5° grid; see Table 2.1), while ncar_pcm1 2.81 2.81 75,115 water planning and management analyses often require ukmo_hadcm3 2.47 3.75 87,806 a much finer resolution (~0.5° x 0.5° grid or even finer ukmo_hadgem1 1.24 1.88 22,103 for project level analyses). (Table 2.2 provides a list of average 2.6 3 72,420 areas for 1° grid cells at various latitudes for reference.) Source: IPCC 2007. The goal of the assessment--that is, what is it trying to answer? who is it trying to inform?--should drive the 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 fewer GCMs (see also the following section on Model tAbLe 2.2. 1º LAtItUDe by 1º LonGItUDe Scenarios). AreAs There is no one "best" method; the most appropriate 1 Degree 1 Degree 1 Square Longitude Latitude Degree method for a particular application will strike a careful Latitude km km km2 balance between precision (resolution) and accuracy 0 111 111 12,393 (confidence in projections). Figure 2.1 provides a visual 40 85 111 9,496 representation of the trade-off between precision and 60 56 111 6,181 accuracy. As resolution increases, so too does the uncer- 80 17 111 1,876 tainty associated with the more detailed information. In other words, more "precise" information comes at a cost, Source: IPCC 2007. and the additional uncertainty must be recognized and taken into account in assessing impacts. Given the trade-off, it is critical to establish at the outset of any impact assessment whether the goal is to have finer decisions on the relevant scale and the most appropriate resolution or "better" (that is, more reliable) technique for matching GCM output with that scale. information. There are several methods available for addressing scale The purpose of this assessment is to establish a issues, including statistical downscaling (using empiri- common platform of information on the behavior of cal relationships), dynamical downscaling (using key hydrologic variables across all World Bank regions. regional climate models), and `spatial techniques'1 The catchment level was selected because it is the most (linear interpolation, krigging, spline fitting, and intelli- appropriate scale for water planning and investment. gent interpolation). Downscaling involves methods Figure 2.2 shows the different model scales/outputs used to map the large-scale signals from GCMs to a relative to the catchment level. This figure shows the finer resolution (tens of kilometers versus hundreds of three different scales: 0.5° by 0.5° grid, 2.5° by 2.5° grid, kilometers). and the catchments. Aggregating from the 0.5° by 0.5° Care needs to be taken in selecting a method. Beyond reproducing the underlying uncertainties of GCMs, many introduce additional uncertainty and biases. For FIGUre 2.1. DesCrIPtIon oF moDeLInG example, downscaling techniques increase the detail of sCALe information, but also the uncertainties associated with that information due to fact that the GCM output is manipulated below the scale at which the physics of the +1 Relative error GCM itself are mathematically described. Under some downscaling schemes, mass balances of water and 0 Uncertainty energy over the GCM scale are violated by the down- scaling algorithm. Use of dynamical and statistical ­1 downscaling techniques requires extensive quantification Grid Catchment River basin Region of the sensitivities of the underlying assumptions of Length scale both the GCMs and the downscaling algorithms, Source: Authors. resulting in the need for exhaustive numerical experi- mentation. Time and cost constraints often do not allow use of more than a couple of GCMs in downscaling exercises. Running multiple GCMs at a coarse resolu- 1 "Spatial technique" is often referred to as "spatial downscaling"; tech- nically, however, it does not involve downscaling but rather statistical tion may provide more insight into the range of possible and spatial relationships. The majority of downscaling being done is futures than more detailed information obtained from with this method. 4 m od Eling th E impAct of climAtE ch Ang E on g lobA l hydrology And WAt E r AVAilA b i l i t y grid to the catchment level is appropriate for two ancillary layers derived from the USGS' 30 arc-second reasons: First, the average scale of a catchment is digital elevation model of the world (GTOPO30)." approximately the size of the native spatial grid scale of (http://eros.usgs.gov/products/elevation/gtopo30/hydro/ the GCM (~2.5° x 2.5°), which results in less uncer- index.html). tainty by using a scale that the input data was created for. And second, the indicators are representative of Hydro1K has six levels of catchments. For this analysis, what is occurring at the catchment level (including all level four was selected for all Bank Regions except available runoff and storage in the catchment) and not Africa, which used level three. There are 8,406 catch- at individual 0.5° grid cells. ments covering the World Bank Regions, which means an average of six CRU grids per catchment. Using Given that the goal of this work is a broad-scale analy- geographic information systems, the catchment bound- sis of the exposure of World Bank investments to aries were overlaid with the CRU grids and the cells potential climate change impacts, examining a range of were aggregated by their weighted area in the climate change scenarios at the coarse resolution was catchment. preferred to examining a few selected scenarios in a detailed spatial resolution requiring some form of model scenarios downscaling, as just discussed. For this work the spatial resolution for the use of GCM was at the native grid Another issue with climate change information is how to scale of each GCM (see Table 2.1). The GCM capture the full range of GCM and Special Report on provided relative changes in temperature and precipita- Emissions Scenarios (SRES) model impacts in a manage- tion for the years 2030 and 2050 on a monthly level as able way. Many combinations of GCM and SRES compared with the model baselines of the twentieth scenarios are available. It takes time and money to evalu- century. These relative changes were then applied ate each scenario in an analysis. There are different directly to the historic climate variables from the approaches to choosing which scenarios to use in an analy- Climate Research Unit (CRU) dataset. As seen in sis (for example, multi-model averages, taking extremes, Figure 2.2, there are numerous 0.5° by 0.5° CRU grids probabilistic.) First, the goal of the analysis needs to be within each GCM grid box. For this analysis we apply addressed. In this analysis, with the goal of informing the relative changes from the GCM grid box uniformly water management, the most appropriate approach is to to all CRU grid cells within the GCM grid box. use the model extremes that may contain the riskiest Although this leads to some discontinuities at the aspects of climate change for water resources. The spec- border of grid boxes, it is the most appropriate tech- trum of model projections is being captured here by imple- nique for this work, given the scope and mathematics menting dry, middle, and wet projections. of the GCM models. As mentioned earlier in the context of downscaling, With the uncertainty inherent in the GCM and the relying on results from a single or even just a few GCMs CRU data, it would be unwise to perform this analysis is not advisable. This is because there are model errors in at the lowest level of resolution (0.5° by 0.5°). any one model and natural variability (randomness) in Aggregating to a higher spatial level would reduce the any particular run. A single model, if run multiple times uncertainty in the model indicators and more correctly with differing initial conditions, can provide an estimate reflect the larger-scale climate change projections from of the uncertainty due to natural variability. For any the GCM models. Therefore a catchment scale was given model, however, there are also uncertainties associ- chosen for this analysis. ated with the assumptions made about model physics and parameterizations, as well as with the structural The catchments were obtained from the U.S. Geological aspects of the model itself. Using a group of GCMs Survey (USGS) Hydro1K. "HYDRO1k is a geographic (multi-model ensembles) as opposed to one individual database developed to provide comprehensive and GCM can somewhat correct for biases and errors. The consistent global coverage of topographically derived use of multi-model ensembles raises the question of how data sets, including streams, drainage basins and to capture the full range of results from model runs. 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 2.2. 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 70º00ºE 72º00ºE 75º00ºE 77º30ºN 80º00ºE 30º00ºN 30º00ºN 27º00ºN 27º00ºN 25º00ºN 25º00ºN 22º30ºN 22º30ºN 20º00ºN 20º00ºN 17º30ºN 17º30ºN 15º00ºN 15º00ºN 12º00ºN 12º00ºN 70º00ºE 72º00ºE 75º00ºE 77º30ºN 80º00ºE Source: Authors. 6 modEling thE impAct of climAtE chAngE on globAl hydrology And WAtEr AVAilA b i l i t y In many applications, the mean of multiple models is used, the rationale being that the mean is representative FIGUre 2.3. reLAtIve CHAnGe From of all runs. The problem with relying on the mean is HIstorICAL CLImAte For DIFFerent GCms that it masks extreme values. A model "average" of near AnD AverAGe zero could be the result of models predicting near-zero West Africa change, but also it could also be the result of two oppos- 30 ing changes that differ in sign, as seen in Figure 2.3. In 20 Rainfall change (%) water management, the risk lies in the "tails" of the full 10 0 spectrum of model projections, so failing to capture the ­10 extremes could be dangerous. ­20 ­30 The potential exists for the range of model outcomes to 1950 1975 2000 2025 2050 2075 Year vary so much that it could be construed as "noise." But there is evidence that suggests a degree of consistency in East Africa 30 some of the more salient changes generated by a collec- 20 Rainfall change (%) tion of model outcomes. As an example, the trend in 10 precipitation intervals as simulated by the IPCC AR4 0 models show, statistically and/or probabilistically speak- ­10 ­20 ing, agreement in the projection of the change in ­30 precipitation interval across latitudes as climate warms. 1950 1975 2000 2025 2050 2075 This implies that although no one should rely solely on Year a single model, each run could potentially contain South Africa important information that is more than merely "noise." 30 Indeed, there are regions where a sign change is consis- 20 Rainfall change (%) 10 tent among the climate models--but with a range that 0 is important to consider explicitly for the assessment of ­10 potential impacts. ­20 ­30 1950 1975 2000 2025 2050 2075 A related issue is filtering or screening of GCM and Year SRES scenarios that are implausible or, at the very least, extremely unlikely. This is difficult--if not impossible-- Source: Giannini et. al., 2008 pg. 376, Fig. 6. to unequivocally determine, as there are no definitive criteria for determining whether a given climate change projection can or should be excluded. One approach in not screened. Dry, middle, and wet scenarios were iden- impact assessments is to consider all modeled projec- tified in terms of each World Bank Region (for tions as "equally likely" at the outset of the assessment, instance, the driest scenario for LAC). A wet scenario and then to exclude in a secondary step those scenarios means that the location experienced the smallest impact with minimal or limited impacts (and to focus on those (or change in CMI); a dry scenario, the largest impact; that could cause significant damage/consequence). and a middle scenario, an impact between the two Techniques are being developed for undertaking a full extremes. The advantage of this approach is that it probabilistic analysis of scenarios to determine which provides a representation of the full range of available are most applicable to each region, but these are not yet scenarios in a "manageable" way. (Further details are available for practical use. given in Chapter 4.) In this analysis, the full spread of model projections-- r e P o rt I n G including extremes--is captured by identifying dry, middle, and wet scenarios, as defined by a change in the The potential impacts of climate change on the hydrologic Climate Moisture Index (CMI). Model projections are cycle and, consequently, the water sector are numerous. 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 7 Projected impacts on runoff and basin yield, extreme sample size of data compared with "mean" value indica- events (floods and droughts), and net irrigation demand tors. However, given that mean value indicators do not are assessed here, as these variables are particularly rele- properly reflect the water resource design or investment vant for water planning and investments. Twenty-two and management decisions, it is felt that the appropriate GCMs along with the A1B, A2, and B1 SRES were used indicators with more uncertainty are preferred. to analyze changes in these key hydrologic variables in the years 2030 and 2050. For each World Bank Region, the sres and GCms wettest, driest, and a middle scenario were identif ied based on the climate moisture index. The results are The Special Report on Emissions Scenarios are emis- presented at the catchment level and summarized by sions scenarios that account for a range of possible Koppen-Geiger climate zones. future greenhouse gas emissions. They are based on assumptions about population growth, economic devel- Hydrologic variables useful to water planning and opment, technological advances, policies on interdepen- investment dency, and commitment to environmental protection. There are a total of 40 scenarios, organized into four There are different methods to representing the impacts "scenario families": of climate change on water systems; most involve the use of hydrologic variables to represent impacts. A vari- · A1 assumes a world of rapid economic growth with ety of approaches exist for generating hydrologic vari- the most growth in developing countries (includes ables, such as mean annual and seasonal runoff and three technology scenarios A1F1, A1T, and A1B) groundwater recharge, within the context of climate · A2 assumes very high population growth and modeling. slower economic growth and technological development Waggoner (1990), Faures (1998), Strzepek et al. (2000), · B1 assumes the same population levels as A1, but Kirshen (2005), Smith and Zhang (2007), UN/WWAP with more clean technologies (and the lowest CO2 (2003, 2006, 2009), and Esty (2008) are just a few who emissions) have proposed hydrologic variables or indicators to help · B2 assumes intermediate levels of economic growth, policy makers and planners make decisions related to and less rapid technological development than A1 water resources investment and policies. A number of and B1. themes emerge from this literature suggesting a set of indicators that would provide information on the The SRESs were used as a basis for climate projections performance of water resource development projects in in the IPCC Fourth Assessment Report (2007).2 the near future and under the threat of climate change in the near distant future. The indicators were selected While it may be interesting to use all 40 SRESs in a to provide inputs to those involved with the wide range climate change analysis, this is not feasible on results of water resource development projects at the World that are archived on the IPCC database. The SRES Bank. The indicators chosen provide information on the team identified marker scenarios to represent a given mean and extreme values of runoff, the storage require- scenario family, although they were not considered to ments for reliable basin yield, groundwater recharge, and be any more "likely" than other scenarios. These net irrigation water demand. included A1B1, A2, B1, B2, and two additional scenar- ios for the groups A1F1 and A1T (Nakienovic and Indicators based on time series ­ new approach Swart 2000). The IPCC based its findings on these six scenarios. For the indicators for extreme events, the approach developed in this study is one of the first attempts to develop indicators based on a time series rather than 2 The earlier emissions scenarios that served as a basis for the climate long-term average indicators. There is more uncertainty projections in the IPCC Third Assessment Report are referred to as around extreme event indicators due to the limited the IS92 scenarios. 8 modEling thE impAct of climAtE chAngE on globAl hydrology And WAtEr AVAilA b i l i t y In this analysis, three SRES scenarios are used: A1B, with more clean technologies and less material inten- A2, and B1. These were chosen because they are sity than A1B. CO2 concentrations are the lowest of included in the marker scenarios identified by the IPCC the SRES scenarios: over 500 ppm by 2100. and are in the middle range of SRESs (see Figure 2.4). The three scenarios can be summarized as follows: Similarly to the 40 SRESs, there are 22 GCMs avail- able in the IPCC Fourth Assessment Report to use in · A1B storyline and scenario family (the "B" stand- climate change analyses (see Table 2.3). In this analysis, ing for balanced) assume a world of rapid economic all of the GCMs were evaluated to determine which growth, with the most growth in developing coun- models would represent the dry, medium, and wet tries, the population peaking at 9 billion by mid- scenarios for each of the World Bank regions, as century and then declining to 8 billion by 2100, and discussed in Chapter 4. rapid technological development. It has the highest per capita income of the four storylines. This sce- Future years nario assumes a mix of fossil-intensive and non-fos- sil fuel energy sources. CO2 concentrations would Typical climate change analyses will evaluate impacts be about 700 ppm by 2100. anywhere from the 2030s to the 2100s. It is important · A2 storyline and scenario family assume very high to keep in mind the purpose of the analysis--that is, population growth (about 15 billion people by 2100) near-term planning or long-range potential--to help and slower economic growth and technological guide which future decades are most important to evalu- development than the other storylines. There is also ate. In this study, the years 2030 and 2050 were used to less convergence in the standard of living and tech- evaluate the impacts of climate change on various hydro- nology between developed and developing countries logic variables. These years were chosen for two reasons: than in the other storylines. It results in the lowest this is the relevant timeframe for current infrastructure per capita income of the four storylines. CO2 con- planning, and beyond 2050 uncertainties in projections centrations would be over 800 ppm by 2100. increase dramatically. As shown in Figure 2.4, SRES · B1 storyline and scenario family assume the same scenarios are tightly bunched until 2050, at which time population levels as A1B, but with more of a transi- they start to diverge significantly. By limiting the analy- tion to a service- and information-based economy, sis to 2050, uncertainty beyond 2050 is eliminated. FIGUre 2.4. totAL GLobAL CUmULAtIve Co 2 emIssIons From 1990 to 2100 AnD HIstoGrAm oF tHeIr DIstrIbUtIon by sCenArIo GroUPs 3000 Scenarios grouped by Medium- Medium- Total Cumulative Carbon dioxide emissions (GtC) cumulative emissions Low low high High 18 2500 IS92d IS92b IS92a IS92e IS92c IS92r A1F1 16 14 Number of scenarios 2000 High > 1800 GtC A2 AIT A1F1 12 A1B Medium High 1450­1800 GtC A2 1500 A1B IS92 range 10 B1 Medium Low 1100­1450 GtC B2 B2 AIT 8 1000 Low <1100 GtC B1 6 500 4 2 0 0 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 < 100 200­300 400­500 600­700 800­900 1000­1100 1200­1300 1400­1500 1600­1700 1800­1900 2000­2001 2002­2003 2004­2005 2006­2007 2008­2009 3000­3001 3200­3003 3400­3500 Cumulative emission 1990­2100, GtC Source: IPCC 2007. 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 The years 2030 and 2050 represent decadal averages of tAbLe 2.3. AvAILAbLe moDeLs, sCenArIos, monthly GCM output. In other words, when reporting AnD vArIAbLes In IPCC Ar4 changes in 2030 relative to historical climate, these are actually average changes from 2025 to 2035 relative to MODELS average historical climate. The same is true for 2050 bccr:bcm2[anomalies] (which represents the average from 2045 to 2055). cccmA:cgcm3_1-t47[anomalies] Average monthly changes from the GCMs over the two cccmA:cgcm3_1-t63[anomalies] separate decades are applied to historical monthly hydro-climatology from 1961 to 1990. cnrm:cm3[anomalies] csiro:mK3-5[anomalies] Geographic representation csiro:mK3-0[anomalies] gfdl:cm2[anomalies] The projected impacts on runoff, basin yield, extreme gfdl:cm2_1[anomalies] events, and net irrigation demand for dry, middle, and inm:cm3[anomalies] wet scenarios are presented at the catchment scale. However, for ease of exposition, catchment-level projec- ipsl:cm4[anomalies] tions are discussed for each World Bank Region accord- lAsg:fgoAls-g1_0[anomalies] ing to Koppen's climate classifications (see Figure 2.5). mpim:EchAm5[anomalies] Synthesizing the information in this way allows the mri:cgcm2_3_2[anomalies] identification of broader trends across regions. All nAsA:giss-Aom[anomalies] projections at the catchment scale can be made avail- nAsA:giss-Eh[anomalies] able, upon request. nAsA:giss-Er[anomalies] The Koppen-Geiger climate classification system3 ncAr:ccsm3[anomalies] divides climate into five primary classifications (and ncAr:pcm[anomalies] several types and subtypes), based on regional average niEs:miroc3_2-hi[anomalies] annual temperature and precipitation, as well as the niEs:miroc3_2-mEd[anomalies] seasonality of precipitation. The five primary classifica- uKmo:hAdcm3[anomalies] tions are equatorial, arid, warm temperate, cold, and uKmo:hAdgEm1[anomalies] polar (see Box 2.1). Source: IPCC 2007. 3 The classification system--one of the most well known and widely used--was developed in the early twentieth century by Wladimir Koppen. Revised in 1961 by Rudolph Geiger, Koppen's classification system was again updated in 2006 with data from the Climate Research Unit of University of East Anglia and the Global Precipitation Climatology Centre at the German Weather Service to reflect the 1951­2000 climate. 10 modEling thE impAct of climAtE chAngE on globAl hydrology And WAtEr AVAilA b i l i t y box 2.1 KoPPen-GeIGer CLImAte CLAssIFICAtIons equatorial. the equatorial climate is characterized by relatively hot temperatures; the coldest month in these regions is greater than 18oc. the most water- and heat-demanding crops are typically grown in this climate (fAo 2007). Equatorial climates include northern south America and central Africa. arid. in arid climates, annual evapotranspiration exceeds annual precipitation and there is a distinct dry season. sunshine in these regions is typically high (fAo 2007). Examples of arid climates include the sahara, the sudan, and parts of the middle East. Warm Temperate. the warm temperate climate is characterized by a relatively mild range of temperatures; the average temperature of the coldest month ranges between ­3o and 18oc and the average of the warmest month is greater than 10oc. some regions may be typified by this general temperature range but fall into a different classification due to precipitation characteristics (for instance, north Africa is considered arid due to the very limited rainfall in the region) (fAo 2007). Examples of warm temperate climates include southeastern brazil, southeast- ern south Africa, and southeastern china. Cold. the cold climate experiences colder temperatures than warm temperate regions; the average temperature in the coldest month is less than ­3oc. growing seasons in cold climates are similar to, but shorter than, warm temperate regions and are typically limited by frost (fAo 2007). Examples of cold climates include parts of russia, Kazakhstan, and mongolia. Polar. the average temperature of the warmest month in polar climates is less than 10oc. the biome typical of polar regions is tundra. Examples of polar climates include southwestern china and the Andes (parts of peru, chile, and Argentina). FIGUre 2.5. KoPPen-GeIGer CLImAte CLAssIFICAtIons For WorLD bAnK reGIons Arid Cold Equetorial Polar Warm Temperate Source: Authors. 11 3. HyDroLoGIC DrIvers AnD DAtA tAbLe 3.1. soUrCes oF DAtA In tHIs stUDy Data Description Source Historical hydrologic conditions in this analysis are precipitation cru-ts2 1.0 dataset. A data- http://www. addressed at three different levels: historical climate, and set of mean monthly surface cru.uea. historical observed runoff, and historical modeled runoff. temperature climate over global land ac.uk/cru/ areas, excluding Antarctica. data/hrg/ interpolated from station data cru_ts_2.10 HI stor ICAL CLIm Ate to 0.5° lat./long. monthly from 1901 to 2002. climatic research unit, university of East Anglia The historical climate was taken from a database calibration simulated runoff (monthly http://www. provided by the Climate Research Unit at the runoff totals in mm) 0.5o grid. grdc.sr.unh. University of East Anglia, Norwich, U.K. The CRU 2.1 university of new hampshire edu/ cd ­ unh/grdc composite data set provides a time series of monthly precipitation runoff fields V1.0 data and the climate variables required to compute Source: Author's Data. potential evapotranspiration (PET) from 1901 to 2002. These data, provided on a 0.5° longitude/latitude grid, represent the World Meteorological Organization's preserve the accuracy of the discharge measurements as (WMO) standard reference "baseline" for climate well as the spatial and temporal distribution of simu- change impact studies. The climate change scenarios lated runoff, thereby providing the "best estimate" of (plausible descriptions of how things may change in the terrestrial runoff over large domains. The method future) are expressed as changes from this baseline. applied in the preparation of this data set uses a gridded There are 67,420 grids (0.5° x 0.5°) over the global land river network at 0.5° spatial resolution to represent the area, excluding Antarctica (see Table 3.1). riverine flow pathways and to link the continental land mass to oceans through river channels. This data set HI stor ICAL observeD rU noFF provides 12 monthly mean values and a mean annual value of runoff for over 50,000 grids (0.5° x 0.5°) over Long-term average monthly runoff has been developed the global land area, excluding permanent ice cover such by the University of New Hampshire (UNH) for the as much of Greenland and all of Antarctica (see Figure WMO Global Runoff Data Center (GRDC). This 3.1 and Table 3.1). UNH-GRDC Composite Runoff Fields V1.0 data set is the combination of observed river discharge informa- H I s to rI C A L mo De Le D r Un o F F tion with a climate-driven Water Balance Model to develop composite runoff fields that are consistent with Water resource development and management are observed discharges. Such combined runoff fields primarily a process to reduce the variability in runoff in 12 modEling thE impAct of climAtE chAngE on globAl hydrology And WAtEr AVAilA b i l i t y FIGUre 3.1. meAn AnnUAL rUnoFF In AFr reGIon GrIDDeD At 0.5º LAtItUDe/LonGItUDe resoLUtIon (UNIVERSITY OF NEW HAMPSHIRE) Mean annual runoff (mm) 0­10 11­25 26­50 51­75 76­150 151­200 201­400 401­600 601­10000 No Data source: Authors. order to provide reliable water supplies. In order to used the monthly data from 1961 to 1990 to produce a project the impacts of climate change on water historic or base scenario runoff time series. resources development projects, a time series of global runoff fields is needed to examine the impacts on So using the UNH/GDRC gridded runoff fields and extreme events (droughts and floods) and variability, the CRU gridded climate database, the CLIRUN-II since an observed global runoff time series does not model (Strzepek et al. 2008) was calibrated to match exist. The GRDC has an extensive data base of stream the UNH/GRDC runoff fields using the CRU data flow data, but this is gauged (with all the development from 1961 to 1980. The calibration was good and impacts included) and at a highly varying spatial scale. provided confidence that the model was capturing the The UNH/GRDC global gridded runoff fields are a underlying hydrology at the grid level and even more wonderful resource but are only available for average confidence that the model was reflecting the "catch- monthly runoff. A modeled data set was developed. The ment" level hydrology. University of Colorado has taken the CRU Historical Climate Data base and the UNH-GRDC Composite With a historic runoff time series, statistical and Runoff Fields together with their Global Runoff stochastic process indicators can be estimated and Model­CLIRUN-II (see Appendix C) to produce a compared with modeled climate change runoff scenar- 30-year monthly time series of runoff at the 50,000 plus ios to examine the projected changes in important grids of the UNH-GRDC data set. The time series design and planning indicators. 13 4. seLeCteD CLImAte CHAnGe red line represents the median CMI, and the top of the box represents the 25th percentile while the bottom of sCenArIos the box represents the 75th percentile. The whiskers show the extremes and the cross-hairs show the model The scenario categorization is defined by the Climate outliers. The dashed lines represent the historical CMI Moisture Index, which is an indicator of the aridity of a (averaged from 1960 to 1990). For example, in the LAC region. The CMI depends on average annual precipita- region, there is a 75 percent chance of drying with all tion and average annual potential evapotranspiration.4 three scenarios. The CMI for the SAR region has the If PET is greater than precipitation, the climate is largest spread because of the way the different GCMs considered to be dry, whereas if precipitation is greater model the monsoons. In the MNA region, there is not than PET, the climate is moist. Calculated as CMI = much variation because the area is so dry. (P/PET)­1 {when PET>P} and CMI = 1­(PET/P) {when P>PET}, a CMI of ­1 is very arid and a CMI of It is important to note that the CMI is only calculated +1 is very humid. As a ratio of two depth measure- over land masses and not over the ocean. Many climate ments, CMI is dimensionless. change analyses discuss GCMs with regards to their properties/results over land Identify the dry, middle and As mentioned in the Model Scenarios section in wet as only based on precipitation. Chapter 2, the full spread of model results was captured by selecting the driest, the wettest, and a middle scenario. There was no screening: the historical and all 22 GCM scenarios were analyzed based on their CMI 4 Average annual PET is a parameter that reflects the amount of water for 2050 to identify the dry, middle, and wet scenario lost via evaporation or transpiration (water consumed by vegetation) for each World Bank region (see Tables 4.1 and 4.2). during a typical year for a given area if sufficient water were available at all times. Average annual evapotranspiration (ET) is a measure of The scenarios used for 2050 CMIs were also used for the amount of water lost to the atmosphere from the surface of soils the 2030 analysis. and plants through the combined processes of evaporation and tran- spiration during the year (measured in mm/yr). ET, which is both connected to and limited by the physical environment, is a measure Figure 4.1 shows the range of CMI for all scenarios for that quantifies the available water in a region. Potential evapotranspi- ration is a calculated parameter that represents the maximum rate of the globe and World Bank Regions as a whole and the ET possible for an area completely covered by vegetation with remaining land mass. Figure 4.2 shows the CMI for all adequate moisture available at all times. PET is dependent on several variables, including temperature, humidity, solar radiation, and wind scenarios individually for each World Bank Region. The velocity. If ample water is available, ET should be equal to PET. 14 m od Eling th E impAct of climAtE ch Ang E on g lobA l hydrology And WAt E r AVAilA b i l i t y tAbLe 4.1. GCm AnD AssoCIAteD bAse CmIs UseD For eACH sCenArIo AnD For reGIons eAP, eCA, AnD LAC EAP ­0.069 ECA ­0.205 LAC ­0.075 Base Model CMI Model CMI Model CMI A2-dry csiro mk3 5 (­0.143) ¡psl cm4 (­0.252) gfdl_cm2_0 (­0.228) A2-middle mri_cgcm2_3_2a (­0.082) ukmo hadcm3 (­0.215) ukmo hadcm3 (­0.151) A2-Wet cccma_cgcm3_1_t63 (­0.033) giss_model_e_r (­0.177) cnrm_cm3 (­0.068) A1b-dry csiro_mk3_5 (­0.135) ¡psl_cm4 (­0.251) ukmo_hadgem1 (­0.202) A1b-midde inmcm3_0 (0.097) mpi_echam5 (­0.212) mpi_echam5 (­0.129) A1b-Wet cccma _cgcm3_1 (­0.054) cccma_cgcm3_1 (­0.184) bccr_bcm2_0 (­0.076) b1-dry csiro_mk3_5 (­0.122) ¡psl_cm4 (­0.243) miroc3_2_hires (­0.153) b1-middle mri_cgcm2_3_2a (­0.084) mpi_echam5 (­0.216) cccma_cgcm3_1 (­0.11) b1-Wet cccma_cgcm3_1_t63 (­0.048) gfdl_cm2_1 (­0.177) cnrm_cm3 (­0.074) Source: IPCC 2007. tAbLe 4.2. GCm AnD AssoCIAteD bAse CmIs UseD For eACH sCenArIo AnD For reGIons mnA, sAr, AnD AFr MNA ­0.91 SAR ­0.372 AFR ­0.5 Base Model CMI Model CMI Model CMI A2-dry gldl_cm2_1 ­0.942 ipsl_cm4 ­0.466 inmcm3_0 ­0.552 A2-mlddle ukmo_hadgem1 ­0.920 ukmo_hadcm3 ­0.312 mpi_echam5 ­0.519 A2-Wet ncar_pcm1 ­0.898 mri_cgcm2_3 _2a ­0.055 ncar_ccsm3_0 ­0.488 A1b-dry gfdl_cm2_1 ­0.941 ipsl_cm4 ­0.496 gfdl_cm2_1 ­0.537 A1b-midde ukmo_hadcm3 ­0.916 ukmo_hadgem1 ­0.294 ukmo_hadgem1 ­0.501 A1b-Wet mpi_echam5 ­0.891 mri_cgcm2_3_2a ­0.003 cnrm_cm3 ­0.484 b1-dry gfdl_cm2_1 ­0.930 csiro_mk3_5 ­0.433 ipsl_cm4 ­0.539 b1-middle inmcm3_0 ­0.907 inmcm3_0 ­0.291 miroc3_2_medres ­0.517 b1-Wet mpi_echam5 ­0.882 mri_cgcm2_3_2a ­0.051 cnrm_cm3 ­0.486 Source: IPCC 2007. 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 FIGUre 4.1. CLImAte moIstUre InDex sPreAD For eACH sCenArIo AnD GLobAL LAnD mAss AnD For eACH reGIon CMI Historic and Range of 2050 Climate Change Imapcts ­0.18 ­0.2 ­0.22 Globe ­0.24 WB Regions ­0.26 Non WB ­0.28 ­0.3 ­0.32 ­0.34 A2 A1b B1 A2 A1b B1 A2 A1b B1 Source: Authors. FIGUre 4.2. CLImAte moIstUre InDex sPreAD For eACH sCenArIo AnD For eACH reGIon (WILLmott AnD FeDDemA) Absolute Predicted Change with Original Values 0 ­0.1 ­0.2 EAP ­0.3 LAC ECA ­0.4 ­0.5 SAS ­0.6 SSA ­0.7 MENA ­0.8 ­0.9 A2 A1b B1 A2 A1b B1 A2 A1b B1 A2 A1b B1 A2 A1b B1 A2 A1b B1 Source: Authors. 16 5. rUnoFF runs per model) over 165 river basins with long-term (28­99 years, median 59 years) stream flow measure- ments. While this work was extremely important to As discussed in Procedures and Rationale in Chapter 2, highlight areas of the globe at risk to changes in runoff, there are different methods to representing the impacts it was based on very poor spatial and temporal scale of climate change on water systems. The first goal is to and, more important, on the hydrologic models of the understand the impact of the water resource, namely the GCMs. annual runoff. Analyzing the climate change impact on this key indicator is discussed in this chapter. The analysis presented here addresses the spatial and temporal resolution as well as methodological limita- met HoD o Lo Gy tions of the Milley approach. Outputs of climate change projections on climatic variables are input to a cali- A variety of approaches/models exist for generating brated hydrologic model with a 0.5° by 0.5° resolution, runoff. These include using runoff estimates derived running on a monthly time scale. Simulating monthly directly from GCMs, using GCM output as input into over a 30-year base period, from 1961 to 1990, provides offline macro-scale hydrologic models, and downscaling an excellent estimation of annual runoff but also GCM output and using resulting data in offline hydro- provides time series data to examine other statistical logic models.5 Runoff estimates derived directly from and stochastic variables from the 30-year monthly time GCMs should be interpreted (and used) with a great series. deal of caution; the limitations of downscaling are discussed in Chapter 2. For this analysis, GCM output H I s to rI C A L re sU Lt s was used as input into an offline hydrologic model. Figure 5.1 shows the average annual runoff from the 30 The hydrologic model CLIRUN-II (Strzepek et al. year historic (1961­90) monthly modeling of runoff for 2008) was chosen for this analysis. This model was the AFR Region . The results show the great spatial developed specifically to assess the impact of climate variability of runoff and highlight the arid and semiarid change on runoff and to address extreme events at the conditions at the northern and southern ends of the annual level by modeling low and high flows. A more region and the extreme humid conditions at the region's detailed description of the model can be found in equatorial central region. These conditions are driven by Appendix C. the rainfall associated with the movement of Inter- Tropical Convergence Zone. Milley et al. (2005) spatially integrated annual runoff fields (using runoff from GCMs) from 62 runs of the 5 These various approaches are discussed in depth in the companion 20C3M experiment on 21 different models (one to nine report on the Science of Water and Climate Change. 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 5.1. HIstorICAL AnnUAL rUnoFF In FIGUre 5.2. ProjeCteD CHAnGe In AnnUAL AFr reGIon (mm/yr) rUnoFF In AFr reGIon (PerCent) Source: Authors. CLI m Ate CHA nGe res ULts Figure 5.2 shows the percentages of change of annual runoff at catchment level (2030 and 2050; A1B, A2, and B1--wet, mid, and dry) from the 30-year historic (1961­90) monthly modeling of runoff for the AFR Region. The results show the great spatial variability of changes, but there are a few trends and consistency in the results that are worth noting. There is an increase in runoff in the northeast of the Source: Authors. region for all three scenarios and particular the 2050 dry scenario. West and Southeast Africa experience drying in all scenarios, while West Africa experiences signifi- cant drying even in the wet 2050 scenario. These results again show that "average" conditions for World Bank Regions may be an increase in runoff, but specific countries or regions may be projected to see the opposite conditions. 18 6. bAsIn yIeLD FIGUre 6.1. ImPACts oF evAPorAtIve Losses on tHe storAGe-yIeLD CUrve For Annual runoff is a good measure of the potential water LAKe nAsser resource available in a basin. However, the variability of Aswan storage-yield 90 that runoff within a year and in between years can make 80 the amount available for economic development only a 70 60 Yield (109 m3) small fraction of the total amount. Through the use of 50 dams and reservoirs, water resource engineers have been 40 30 able to increase the percentage of annual runoff that is 20 reliably available for development. An indicator to 10 express the ability and accessibility of runoff for 0 0 50 100 150 200 economic use is the basin yield. Storage (109 m3) met HoD o Lo Gy Source: Authors. The basin yield is a measure of annually reliable water supply from the basin. It is directly related to the amount assumed that one will reliably get the minimum yearly of reservoir storage in a basin. Water resource planners flow or the lowest recorded. The shape of the curve that have developed methodologies to estimate reliable water ends at the average annual runoff is a function of the supply or basin yield as a function of reservoir storage in within-year and year-to-year variability of the stream a basin. The result of these methodologies is a concept flow. A steep curve reflects low variability and a flatter known as the storage-yield curve. This is an estimated curve is reflective of high variability. A highly variable time series of annual or monthly flows in the basin, basin will require more storage for the same basin yield which gives the planner a tool to answer two questions: than a basin with less variability. The storage-yield curve How much storage is needed to provide a certain amount can be presented in absolute terms of volume of storage of annual reliable yield? And for a certain amount of versus annual flow; in some cases, it is preferred to pres- storage, what is the reliable yield from the base? ent it as a ratio to average annual runoff. Climatic change has the potential to affect not only the average annual Figure 6.1 is an example of a storage-yield curve the Nile runoff in a basin but also the annual runoff 's variability in River at Aswan. Two points of the curve are easily esti- the shape of the storage-yield curve. mated. The maximum yield, ignoring evaporation, is the average annual runoff in the basin, while the minimum If the policy is to maximize the head for hydropower yield is the lowest measured or modeled flow in the time while at the same time delivering a reliable yield, then series. Thus without any storage (zero on the x-axis), it is the reliable yield actually declines with increased storage 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 19 at higher storage levels due to the greater surface area and higher evaporation. FIGUre 6.2. ImPACt oF CLImAte CHAnGe on reservoIr yIeLD AnD ADAPtAtIons Figure 6.1 is an illustration of how climate change will shift the storage-yield curve due to changes in annual Yield (MCM/Yr) 2005 Average annual flow runoff. The change in annual runoff translates into the change in maximum basin yield for a fixed storage. Figure 6.2 shows the increase in storage required to Y2005 2050 Average annual flow maintain a constant basin yield. The storage-yield Y2050 curves for nine regions in China are shown in Figure 6.3 as an example. Storage-yield curves were created for each catchment in the analysis and then aggregated to the regions for reporting. The solid line represents the base case, the solid gray line the HadCM2 scenario, the dotted line the K K' Storage (MCM) CGCM1 scenario, and the dashed line the ECHAM4 Source: Authors. scenario. The horizontal axis is storage in billions of cubic meters, and the vertical axis is yield in billions of cubic meters (Wiberg and Strzepek 2006). FIGUre 6.3. ImPACts oF tHe GCm sCenArIos on tHe storAGe-yIeLD CUrves For nIne reGIons In CHInA Source: Authors. 20 modEling thE impAct of climAtE chAngE on globAl hydrology And WAtEr AVAilA b i l i t y HIstorICAL resULts FIGUre 6.5. ProjeCteD CHAnGes In AnnUAL Figure 6.4 shows the basin yield calculated from the bAsIn yIeLD In AFr reGIon (PerCent) 30-year historic (1961­90) monthly modeling of runoff for the AFR Region. The results show similar spatial patterns as the average annual runoff results in Figure 5.1. CLImAte CHAnGe resULts Figure 6.5 presents the climate change impacts on Basin Yield for the AFR Region. Since basin yield is a func- tion of average runoff and variability, it exhibits a non- linear behavior to changes in climate. This can be seen as the spatial pattern of changes are similar but in many case more dramatic for the wet and dry parts of the region. FIGUre 6.4. HIstorICAL AnnUAL bAsIn yIeLD In AFr reGIon (mm/yeAr) Source: Authors. Source: Authors. 21 7. sUmmAry oF resULts It is important to remember that these summaries reflect only the general trend over the entire Region. Subregions and even individual catchments can vary Tables 7.1 through 7.6 summarize the results for the widely from the mean conditions. Another confounding each of the indicators for three IPCC SRES scenar- condition is when two significant Regions exhibit ios (A2, A1B, and B1) and for the decades surround- exactly the opposite climate change impact (for exam- ing 2030 and 2050. The Tables provide results for the ple, significant increase or significant decrease in average of all catchments in the World Bank runoff ), so that the average result is little climate Regions. change impact at all. tAbLe 7.1. rUnoFF SRES scenario Year Projection AFR EAP ECA LAC MENA SAR dry ­5% ­13% 5% ­16% ­17% ­7% 2030 medium 6% ­4% 9% ­3% ­17% 26% Wet 30% 7% 13% 15% 24% 31% A2 dry ­13% ­11% 8% ­31% ­50% ­52% 2050 medium 14% 3% 18% ­9% ­5% 20% Wet 31% 11% 18% 7% 31% 39% dry ­5% 7% 9% ­9% ­25% 18% 2030 medium 3% 5% 13% ­5% 12% 18% Wet ­1% 11% 10% 9% 15% 43% A1b dry ­20% ­12% 14% ­25% ­46% 24% 2050 medium 0% ­2% 21% ­16% ­16% 24% Wet 6% 21% 19% 7% 20% 37% dry ­1% ­8% 6% ­13% ­15% ­20% 2030 medium ­9% 2% 11% ­11% 12% ­5% Wet 10% 14% 16% 11% 18% 36% b1 dry ­8% ­8% 5% ­9% ­30% ­14% 2050 medium ­12% ­1% 14% ­9% 19% 38% Wet 23% 15% 19% 9% 49% 33% Source: Authors. 22 m od Eling th E impAct of climAtE ch Ang E on g lobA l hydrology And WAt E r AVAilA b i l i t y The results show the indicators tend to be well corre- precipitation and lower response of potential evapo- lated and generally show the same sign and magnitude transpiration to temperature increases at the colder for each indicator for each Region. One exception is temperatures of the Region. that in most cases the water deficit index increases for all scenarios, especially in 2050, due to the non-linear The other Regions exhibit reduction in runoff for dry increase of potential evapotranspiration. scenarios and increase in runoff for wet scenarios. The results are mixed for the mid scenarios. All Regions tend to show decreases in runoff for dry and mid scenarios. However, the Europe and Central Detailed results by catchment area are available for Asia Region shows an increase in runoff for all scenar- anyone interested by contacting the World Bank Water ios wet to dry. This is due to the significant increase in Anchor. tAbLe 7.2. 10% FLooD exCeeDenCe SRES scenario Year Projection AFR EAP ECA LAC MENA SAR dry ­4% ­13% 5% ­16% ­15% ­8% 2030 medium 5% ­3% 8% ­2% ­16% 23% Wet 27% 7% 13% 14% 25% 32% A2 dry ­11% ­10% 7% ­28% ­48% ­50% 2050 medium 13% 3% 17% ­9% ­4% 21% Wet 27% 11% 17% 7% 36% 39% dry ­4% 7% 8% ­8% ­24% 18% 2030 medium 3% 5% 13% ­4% 10% 18% Wet ­1% 10% 10% 9% 13% 43% A1b dry ­18% ­12% 12% ­21% ­43% 24% 2050 medium 1% ­2% 20% ­14% ­13% 24% Wet 6% 20% 19% 7% 27% 35% dry ­1% ­8% 6% ­11% ­13% ­18% 2030 medium ­7% 2% 10% ­9% 12% ­5% Wet 9% 13% 17% 11% 22% 36% b1 dry ­7% ­8% 4% ­7% ­26% ­12% 2050 medium ­10% ­1% 13% ­7% 20% 34% Wet 20% 15% 19% 8% 52% 32% Source: Authors. 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 23 tAbLe 7.3. 90% LoW FLoW SRES scenario Year Projection AFR EAP ECA LAC MENA SAR dry ­4% ­14% 7% ­18% ­19% ­5% 2030 medium 6% ­4% 10% ­2% ­18% 31% Wet 37% 8% 14% 19% 18% 31% A2 dry ­15% ­11% 10% ­35% ­52% ­55% 2050 medium 19% 3% 21% ­11% ­8% 21% Wet 40% 13% 19% 9% 17% 41% dry ­5% 7% 11% ­11% ­29% 20% 2030 medium 3% 5% 15% ­5% 9% 20% Wet ­1% 12% 10% 9% 13% 47% A1b dry ­23% ­13% 18% ­32% ­50% 25% 2050 medium 0% ­2% 23% ­18% ­24% 25% Wet 7% 21% 19% 8% 3% 40% dry 0% ­9% 9% ­17% ­19% ­22% 2030 medium ­12% 2% 13% ­15% 9% ­3% Wet 13% 14% 17% 10% 12% 37% b1 dry ­8% ­9% 7% ­13% ­34% ­17% 2050 medium ­14% ­1% 17% ­13% 8% 48% Wet 30% 14% 20% 9% 37% 37% Source: Authors. tAbLe 7.4. bAseFLoW SRES scenario Year Projection AFR EAP ECA LAC MENA SAR dry ­14% ­32% 21% ­24% ­7% ­14 2030 medium ­7% ­13% 12% ­7% ­19% 33% Wet 27% 6% 19% 17% 4% 32% A2 dry ­12% ­30% 26% ­37% ­51% ­44% 2050 medium ­8% ­6% 33% ­18% ­9% 25% Wet 12% 10% 33% 21% 58% 35% dry ­9% 2% 20% ­11% ­30% 27% 2030 medium 5% 2% 24% ­10% 4% 27% Wet 0% 9% 23% 16% 5% 47% A1b dry ­17% ­34% 33% ­33% ­57% 27% 2050 medium ­3% ­12% 40% ­19% ­17% 27% Wet 13% 35% 29% 8% 17% 37% dry ­4% ­19% 19% ­14% ­18% ­10% 2030 medium ­10% ­2% 6% ­6% ­13% ­16% Wet 8% 23% 20% 21% 15% 28% b1 dry ­14% ­21% 18% ­14% ­39% ­5% 2050 medium ­13% ­13% 33% ­2% 12% 24% Wet 10% 21% 23% 2% 87% 29% Source: Authors. 24 m od Eling th E impAct of climAtE ch Ang E on g lobA l hydrology And WAt E r AVAilA b i l i t y tAbLe 7.5. 10% bAsIn yIeLD SRES Scenario Year Projection AFR EAP ECA LAC MENA SAR dry 21% 42% ­2% 44% 61% 40% 2030 medium 9% 14% ­9% 17% 60% ­14% Wet ­22% ­7% ­13% ­10% ­4% ­22% A2 dry 38% 42% ­3% 74% 136% 53% 2050 medium ­2% 0% ­14% 39% 24% ­5% Wet ­19% ­13% ­18% 5% 4% ­29% dry 26% ­4% ­9% 29% 85% ­19% 2030 medium 3% ­2% ­16% 26% ­17% ­19% Wet 7% ­12% ­7% ­1% ­11% ­29% A1b dry 53% 49% ­12% 57% 141% ­18% 2050 medium 12% 10% ­24% 53% 55% ­18% Wet ­4% ­25% ­21% 6% ­19% ­32% dry 8% 35% ­2% 38% 46% 45% 2030 medium 27% 1% ­15% 35% 13% 27% Wet ­5% ­17% ­15% ­6% ­7% ­26% b1 dry 30% 26% 4% 27% 103% 30% 2050 medium 37% 7% ­16% 30% 26% ­31% Wet ­14% ­20% ­21% ­5% ­29% ­28% Source: Authors. tAbLe 7.6. WAter DeFICIt InDex SRES Scenario Year Projection AFR EAP ECA LAC MENA SAR dry 0% 122% 10% 40% ­5% 9% 2030 medium ­1% 91% ­15% 30% ­1% ­6% Wet ­10% 70% ­18% 12% ­4% ­10% A2 dry ­2% 127% 17% 69% 6% 26% 2050 medium 4% 87% ­16% 37% 0% ­3% Wet 5% 74% ­16% 21% ­12% ­5% dry 13% 106% 30% 59% 14% 33% 2030 medium 13% 97% 19% 59% 7% 4% Wet 10% 90% 19% 25% 7% 3% A1b dry 29% 141% 44% 87% 23% 33% 2050 medium 18% 114% 28% 64% 15% 4% Wet 5% 65% 19% 39% 11% 11% dry 15% 125% 35% 41% 10% 23% 2030 medium 16% 104% 9% 23% 10% 22% Wet 9% 71% 6% 13% 8% 5% b1 dry 20% 132% 44% 48% 14% 17% 2050 medium 18% 117% 25% 22% 8% 12% Wet 16% 67% 5% 44% ­5% 10% Source: Authors. 25 8. ConCLUsIons GCM data are raw output at an average spatial scale of 2.5° by 3°. This provided significant uncertainty for climate change at the subgrid scale level. It is difficult to come up with one simple result from this analysis. But there are a series of universal messages Future work that could build on this analysis should about potential climate change impacts over the World focus on three areas of improvement: climate scenarios, Bank Regions of operation: hydro climatic data, and hydrologic modeling: 1. There is a wide range of Climate Change impacts 1. Probabilistic Climate Scenarios. Since it is suggested within each Region. to cast climate change in a risk planning frame- 2. The IPCC SRES scenario and the specific GCMs work, this calls for explicit probabilistic character- analyzed greatly influence the results of climate ization of GCM results. impact modeling. 2. Statistical and Dynamic Downscaling. Current 2.5° x 3° 3. For a risk-based approach to planning, it is impor- GCM grids are not sufficient to capture many oro- tant to use a wide range of SRES and GCM sce- graphic climate processes. An effort to statistically or narios rather than focusing on a few to address the dynamically (regional climate models) a wide range full range of uncertainty regarding future climate of IPCC CGM is very costly. The result is that just that the water resource planner is facing. for one or two models are done whilte a wide range 4. A wide range of climate change indicators are of models, say a minimum of 10 models is needed. needed to assist the assessment of operational risk Finer Temporal Resolution of Climate Model Result. to the varied Water System investment projects Hydrology and flood flows occur at the hourly and undertaken by the World Bank. daily scale. Working with climate modelers to 5. Some results suggest that World Bank projects archive data at the daily and lower time step for lon- may be facing a much different climatic and thus ger time periods and for more variables needed for hydrologic regime threatening their economic per- the potential evapotranspiration calculations (such as formance as early as 2030. tmin, tmax, vapor pressure) is an important step 6. Climate change is an additional uncertainty facing needed to solve this issuesFiner-scale Hydrologic water resource planners and should be included as Models. This involves development of daily hydrologic a regular part in any hydrologic assessment. model at the 500 to 1000 km 2 watershed scale. 3. Improvement in Global Scale Stream-flow Time Series The data and models used are monthly data, and the Data for calibration of hydrologic model at the stream flow data used for calibration are monthly aver- monthly and daily levels is needed with detailed ages not time series. This produces increased uncer- meta data on time periods and human influences tainty in extreme values or tails of the distributions. The on the stream records. 26 reFerenCes ------. 1985. "Relationship between Data and the Precision of Parameter Estimates of Hydrologic Models." Journal of Hydrology 81 (1/2): 57­77. Block, P., and B. Rajagopalan. 2007. "Interannual Variability and Ensemble Forecast of Upper Blue Nile Huber-Lee, A., D. Yates, D. Purkey, W. Yu, C. Young, Basin Kiremt Season Precipitation." Journal of and B. Runkie. 2005. "How Can We Sustain Hydrometeorology 8 (3): 327­43. Agriculture and Ecosystems? The Sacramento Basin (California, USA)." In Climate Change in Contrasting Esty, Daniel C., and Andrew S. Winston. 2009. Green to River Basins: Adaptation Strategies for Water, Food and Gold: How Smart Companies Use Environmental Strategy Environment, ed. J. C. J. H. Aerts and P. Droogers. to Innovate, Create Value, and Build Competitive CABI Publishing, Wallingford, U.K. Advantage. Hoboken, NJ: Wiley. IPCC (Intergovernmental Panel on Climate Change). FAO (Food and Agriculture Organization). 1996. Crop 2007. Climate Change 2007: Synthesis Report. Geneva. Evapotranspiration (Guidelines for Computing Crop Water Requirements). Irrigation and Drainage Paper, No. 56. Kaczmarek, Z. 1993. "Water Balance Model for Climate Rome. Impact Analysis." Acta Geophysical Polonica 41 (4): 423­37. Faurès, J. M., D. C. Goodrich, D. A. Woolhiser, and S. ------. 1998. Human Impact on Yellow River Water Sorooshian. 1995. "Impact of small-scale rainfall vari- Management, Interim Report IR-98-016. International ability on runoff modeling." Journal of Hydrology, 173 Institute for Applied Systems Analysis, Laxenburg, (1995) Austria. Giannini A., M. Biasutti, I . M . Held, and A. H. Sobel. Kirshen, P., M. McCluskey, R. Vogel, and K. Strzepek. 2008. A global perspective on African Climate. Climatic 2005. "Global Analysis of Changes in River Basin Change 90 (4): 359­383. Water Supply Yields and Costs under Climate Change: A Case Study of China." Climatic Change, 68:303-330. Gupta, V. K., and S. Sorooshian. 1983. "Uniqueness and Observability of Conceptual Rainfall­Runoff Model Milley, P., K. Dunne, and A. Vecchia. 2005. "Global Parameters: The Percolation Process Examined." Water Pattern of Trends in Streamflow and Water Availability Resources Research 19 (1): 269­76. in a Changing Climate" (letter). Nature. 438: 347­50. 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 27 McCabe, G., and D. Wolock. 1999. "General- Strzepek, K., and D. Yates. 2000. Responses and Circulation-Model Simulations of Future Snowpack in Thresholds of the Egyptian Economy to Climate the Western United States." Journal of the American Change Impacts on the Water Resources of the Nile Water Resources Association 35 (6): 1473­84. River, Climatic Change 46(3): 339­56. Nakienovic, N., and R. Swart, eds. 2000. Special Report UN/WWAP (United Nations/World Water Assessment on Emissions Scenarios. Cambridge University Press, Programme). 2003. 1st UN World Water Development Cambridge, U.K. Report: Water for People, Water for Life. Paris, New York and Oxford. UNESCO (United Nations Smith, Ethan T., and Harry F. Zhang. 2007. "Evolution Educational, Scientific and Cultural Organization) and of Sustainable Water Resources Indicators." Proceedings Berghahn Books. of the Water Environment Federation 31 (2007): 2624- 2649. Water Environment Federation. ------. 2006. The 2nd UN World Water Development Report: "Water, A Shared Responsibility." Strzepek, K., C. Rosenzweig, D. Major, A. Iglesias, D. Yates, A. Holt, and D. Hillel. 1999. "New Methods of ------. 2009. The 3rd UN World Water Development Modeling Water Availability for Agriculture under Report: "Water in a Changing World." Climate Change." Journal of the American Water Resources Association 35 (6): 1639­55. Waggoner, Paul E. 1990. Climate Change and U.S. Water Resources. New York: Wiley. Strzepek, K. , A. McCluskey, J. Hoogeveen, and J. van Dam. 2005. "Food Demand and Production: A Global Willmott, C. J., and J. J. Feddema. 1992. "A More and Regional Perspective." In Climate Change in Rational Climatic Moisture Index." The Professional Contrasting River Basins: Adaptation Strategies for Water, Geographer 44.1: 84­88. Food and Environment, ed. J. C. J. H. Aerts and P. Droogers. CABI Publishing, Wallingford, U.K. Yates, D. 1996. "WatBal: An Integrated Water Balance Model for Climate Impact Assessment of River Basin Strzepek, K., R. Balaji, H. Rajaram, and J. Strzepek. Runoff." International Journal of Water Resources 2008. "A Water Balance Model for Climate Impact Development 12(2): 121­39. Analysis of Runoff with Emphasis on Extreme Events." In preparation. 28 APPenDIx: reFerenCe For (Gupta and Sorooshian 1983, 1985), and a unique moDeLs AnD DAtA conditional parameter estimation procedure was used. This Appendix presents a brief description of the components of the model. CLI r Un-II r AInFALL rU noFF moDeL Spatial and Temporal Scale. CLIRUN II is models runoff CLIRUN-II is the latest model in a family of hydro- as a lumped watershed with climate inputs and soil logic models developed specifically for the analysis of characteristics averaged over the watershed simulating the impact of climate change on runoff. Kaczmarek runoff at a gauged location at the mouth of the catch- (1993) presents the theoretical development for a ment. CLIRUN can run on a daily or monthly time single-layer lumped watershed rainfall runoff model- step. For this study, climate and runoff data were avail- CLIRUN. Kaczmarek (1998) presents the application able only on a monthly basis so monthly was used. of CLIRUN to the Yellow River in China. Snow-Balance Model. The snow accumulation and melt Yates (1996) expanded on the basic CLIRUN by adding model used in this study is based on concepts frequently a snow-balance model and providing a suite of possible used in monthly water balance models (McCabe and PET models and packaged it in a tool WATBAL. Wolock 1999). Inputs to the model are monthly WATBAL has been used on a wide variety of spatial temperature (T) and precipitation (P). The occurrence scales from small to large watersheds and globally in of snow is computed as a function of average watershed 0.5° by 0.5° grid (Strzepek et al. 1999; Huber-Lee et al. temperature and two parameters, Temp_snow and 2005 ; Strzepek et al. 2005). Temp_rain. These two parameters are calibrated for each watershed. Snowmelt is added to any monthly CLIRUN-II (Strzepek et al. 2008) is the latest in the precipitation to form effective precipitation available for "Kaczmarek School" of hydrologic models. It incorpo- infiltration or direct runoff. rates most of the features of WATBAL and CLIRUN but was developed specifically to address extreme events The figure shows the mass balance of water in the at the annual level modeling low and high flows. CLIRUNII system. Water enters via precipitation and CLIRUN and WATBAL did very well in modeling leaves via evapotranspiration and runoff. The difference mean monthly and annual runoff, important for water between inflow and outflow is reflected as change in supply studies, but they did not model well the tails of storage in the soil or groundwater. runoff distribution. Evapotranspiration. A suite of potential evapotranspira- CLIRUN-II has adopted a two-layer approach follow- tion models is available for use in CLIRUNII. For this ing the framework of the SIXPAR hydrologic model study the Blaney-Criddle (temperature based) method 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 29 and spring) and non-frozen (summer and fall). The FIGUre C.1. CLIrUn-II ConCePtUAL infiltration then enters the soil layer. A non-linear set of HyDroLoGIC moDeL sCHemAtIC equations determines how much water leaves the soil as Precipitation runoff, how much is percolated to the groundwater, and how much goes into soil storage. The runoff is a linear Temperature relation of soil water storage, and percolation is a non- linear relationship of both soil and groundwater Evapotranspiration storages. Snow Slow Runoff. The groundwater receives percolation from Melt Discharge the soil layer, and runoff is generated as a linear func- Surface storage quick flow Soil storage tion of groundwater storage. Groundwater storage slow flow The soil water processes have six parameters similar to the SIXPAR model (Gupta and Sorooshian 1983) that are determined via calibration of each watershed. Modeling Dry and Wet Years. When CLIRUNII is cali- brated in a classical rainfall-runoff framework, the results are very good for the 25th to 75th percentile of (FAO 1996) was used to be consistent with State of the observed stream flows, producing R2 of 0.3 to 0.7 Colorado practices. Actual evapotranspiration is a func- For most water resource systems, however, the tails of tion of potential and soil moisture state following the the stream flow distribution are important. for design FAO method. and operation planning. To address this issue, a concept developed by Block and Rajagopalan (2007) Soil Water Modeling. Soil water is modeled as a two- for hydrologic modeling of the Nile River, know as layer system: a soil layer and a groundwater layer. These localized polynomial, was extended to calibration of two components correspond to a quick and a slow rainfall runoff modeling in CLIRUNII (Strzepek et al. runoff response to effective precipitation. 2008). Quick Runoff. The soil layer generates runoff in two Briefly, when calibrating, each observed year is catego- ways. First there is a direct runoff component, which is rized as to whether it falls into a dry year 0­25 percent the portion of the effective precipitation (precipitation of the distribution, a normal year 25­75 percent, or wet plus snowmelt) that directly enters the stream systems. year greater than 75 percent. A separate set of model The remaining effective precipitation is infiltration to parameters was estimated for the three different class of the soil layer. The direct runoff is a function of the soil annual stream flow. This increased the R2 from 0.7 to surface, and models differently for frozen soil (winter 0.92. 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