SAFETY OF DAMS AND DOWNSTREAM COMMUNITIES Enhancing the Safety and Resilience of Dams in the Context of Climate Change and Extreme Hydrological Events Detailed Methodologies and Case Studies About the Global Department for Water The World Bank Group’s Global Department for Water brings together financing, knowledge, and implementation in one platform. By combining the Bank’s global knowledge with country investments, this model generates more firepower for transformational solutions to help countries grow sustainably. Please visit us at www.worldbank.org/water or follow us on : @WorldBankWater. About GWSP This publication received the support of the Global Water Security & Sanitation Partnership (GWSP). GWSP is a multidonor trust fund administered by the World Bank’s Global Department for Water and supported by Australia’s Department of Foreign Affairs and Trade; Austria’s Federal Ministry of Finance; Denmark’s Ministry of Foreign Affairs; the Netherlands’ Ministry of Foreign Affairs, the Gates Foundation; Spain’s Ministry of Economy, Trade and Business; the Swedish International Development Cooperation Agency, Switzerland’s State Secretariat for Economic Affairs; the Swiss Agency for Development and Cooperation; and the United Kingdom Foreign, Commonwealth and Development Office. Please visit us at www.worldbank.org/gwsp or follow us on : @TheGwsp About the Safety of Dams and Downstream ­Communities series This publication is part of a series produced by the Global Dams and Reservoirs Team of the World Bank and has been developed under a cohesive program focused on the introduction of innovative tools and techniques, generation of new knowledge, and the development and dissemination of good international industry practices. This is achieved through a combination of strategic analyses, advisory activities, and the provision of technical assistance. This work aims to promote collaboration and knowledge exchange among World Bank task teams and with other stakeholders working on resilient water security, disaster risk management, and quality infrastructure improvements. Enhancing the Safety and Resilience of Dams in the Context of Climate Change and Extreme Hydrological Events Detailed Methodologies and Case Studies October 2025 © 2025 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy, completeness, or currency of the data included in this work and does not assume responsibility for any errors, omissions, or discrepancies in the information, or liability with respect to the use of or failure to use the information, methods, processes, or conclusions set forth. The boundaries, colors, denominations, links/ footnotes and other information shown in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. The citation of works authored by others does not mean the World Bank endorses the views expressed by those authors or the content of their works. Nothing herein shall constitute or be construed or considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Please cite the work as follows: World Bank. 2025. “Enhancing the Safety and Resilience of Dams in the Context of Climate Change and Extreme Hydrological Events: Detailed Methodologies and Case Studies.” World Bank, Washington, DC. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. Cover photo: © Marcus Wishart / World Bank. Further permission required for reuse. Cover design: Bill Pragluski, Critical Stages, LLC. Contents Acknowledgments..........................................................................................................................................vii Abbreviations...................................................................................................................................................ix Introduction ....................................................................................................................................................xi PART I: DETAILED METHODOLOGIES OF CLIMATE CHANGE IMPACT ASSESSMENT Chapter 1 Estimation of the Probable Maximum Precipitation............................................................2 Methods for PMP Estimation.........................................................................................................................................2 Statistical PMP Estimation........................................................................................................................................2 Physical PMP Estimation...........................................................................................................................................3 Methodologies for Large Watersheds..................................................................................................................3 Alternative Probabilistic and Stochastic and NWP-based PMP............................................................... 4 PMP Calculations.............................................................................................................................................................. 4 Hershfield Statistical PMP Estimation................................................................................................................. 4 Aspects of Physical PMP Estimation....................................................................................................................5 Generalized PMP Estimation....................................................................................................................................6 Chapter 2 Climate Change Impacts on Design Rainfall and PMP.........................................................7 Notes......................................................................................................................................................................................8 Chapter 3 The Importance of Large GCM Ensemble Experiments......................................................9 Note....................................................................................................................................................................................... 11 Chapter 4 Downscaling of GCM Output................................................................................................. 12 Need for Downscaling of GCM Output................................................................................................................... 12 Statistical Downscaling................................................................................................................................................. 13 Dynamical Downscaling................................................................................................................................................14 Comparison of Dynamical and Statistical Downscaling Methodologies.................................................... 15 Stochastic Weather Generators ................................................................................................................................ 16 CORDEX Dynamical Downscaling Experiments.................................................................................................. 19 Notes.................................................................................................................................................................................... 21 Chapter 5 Methods of Bias Correction...................................................................................................22 Notes................................................................................................................................................................................... 25 Chapter 6 Design Flow Estimation......................................................................................................... 26 Overview............................................................................................................................................................................ 26 Hydrological Simulation Models............................................................................................................................... 26 Hydrological Simulations for Special Conditions............................................................................................... 28 Statistical Flood Frequency Analysis......................................................................................................................30 Envelope Curve Methods for Extreme Flood Estimation................................................................................ 32 Notes...................................................................................................................................................................................34 iii iv Enhancing the Safety and Resilience of Dams in the Context of Climate Change PART II: CASE STUDIES ON IMPACT ASSESSMENT AND ADAPTATIVE RESERVOIR OPERATION Chapter 7 Case Studies on the Assessment of Climate Change Impacts on Floods...................... 36 Australia—Impact of Climate Change on Operational PMP Estimates...................................................... 36 India—The PMP in a Changing Climate................................................................................................................... 39 Viet Nam—Climate Impact Assessment on Large Floods and Dam Safety............................................ 40 Climate Change Projections.................................................................................................................................. 42 Climate Change Impacts on Hydrological Safety of the Dam ................................................................43 Japan—Nationwide Climate Change Impact Assessment on Large Floods............................................44 Sri Lanka—Climate Change Impact Assessment and Future Inundation Analysis................................ 47 Extreme Flood Event Analysis.............................................................................................................................48 Inundation Analysis..................................................................................................................................................50 Notes................................................................................................................................................................................... 53 Chapter 8 Case Studies on the Application of Advanced Rainfall and Reservoir Inflow Forecasting and Optimal Reservoir Operation Systems.................................................. 54 Italy—Enhanced Water Supply and Flood Control for Lake Como.............................................................54 United States—Enhanced Water Supply and Flood Control for Lake Mendocino and Coyote Valley Dam, California .................................................................................................................................................. 55 Operational Testing for Lake Mendocino......................................................................................................... 56 Ensemble Streamflow Predictions and Ensemble Forecast Operations............................................. 57 Continued Enhanced Technological Development and Collaboration toward FIRO2.0............... 58 Japan—Enhanced Flood Control and Hydropower Generation of Hatanagi-I Dam on Oi River...... 59 Japan—Pre-Flood Reservoir Drawdown Operation of the Kusaki Multipurpose Dam......................... 63 Australia—Wivenhoe Dam Flood Operation Incident, Brisbane...................................................................66 Notes.................................................................................................................................................................................... 71 References......................................................................................................................................................................... 73 Box 4.1 Stochastic Weather Generators.......................................................................................................................... 16 Figures 1.1 Example of Depth-Area-Duration Curves.........................................................................................................6 6.1 Overview of the Hydrological Modeling Process for the Assessment of Design Floods............. 27 6.2 Extreme Flood Estimation Using Different Flood Frequency Assessment Methods..................... 31 6.3 Example of a Francou-Rodier Graph for River Basins in Central Africa............................................. 33 7.1 Density Plot of Projected Increases in PMP Estimates, Australia......................................................... 39 7.2 Case Study Dam Reservoir Peak Water Level as a Function of Flood Frequency (AEP) for RCP4.5 and RCP8.5 Climate Change Scenarios...................................................................................44 7.3 Assessment Framework........................................................................................................................................48 7.4 First 20 Peak-Flow Events of Daily Average Discharge Values of the Past and Future 20 Years of the Selected GCMs...........................................................................................................49 7.5 Past vs. Future All-Time Seasonal Maximum Inundation during the NE Monsoon of the Selected GCMs.......................................................................................................................................................... 52 8.1 Lake Mendocino Storage Rule Curve and FIRO Operations in WY 2020 ........................................ 56 8.2 Atmospheric River................................................................................................................................................... 58 8.3 Details of Flooding with a Gate Discharge Not Exceeding 600m3/s.................................................. 62 8.4 Kusaki Dam’s Rainfall Forecast for Flood Control Operation toward Typhoon Hagibis (October 2019).........................................................................................................................................................64 8.5 Reduced Flood Water Level at a Downstream River Section Resulting from Kusaki Dam Drawdown Operation............................................................................................................................................. 65 8.6 Flood Forecast at 3 p.m. October 11, 2019 Prompting Emergency Reservoir Drawdown........... 65 8.7 Impact of Pre-Flood Reservoir Drawdown Operation..............................................................................66 8.8 Inflow and Release of Water during the January 2011 Flood Event—Wivenhoe Dam..................69 8.9 Modeled Wivenhoe Dam Lake Levels at 8 p.m. January 9, 2011...........................................................70 8.10 Brisbane River Water Level at the City Gauge between January 8–16, 2011..................................... 71 Contents v Maps 7.1 Future Increases in the Maximum Dew Point Temperature and Moisture Adjustment Factor in Australia................................................................................................................................................... 37 7.2 Location of Large Dams across Australia by ANCOLD............................................................................ 38 7.3 Regional Classification of Japan........................................................................................................................46 7.4 The Largest and Perennial River Basin in Sri Lanka................................................................................... 47 7.5 Past Inundation Depth and Future Inundation Depth Changes Projection by Different Climate Models in Sri Lanka................................................................................................................................50 8.1 Location Map............................................................................................................................................................. 59 8.2 Hatanagi-I Dam on the Oi River, Japan...........................................................................................................60 8.3 Brisbane River Basin...............................................................................................................................................68 Photos 8.1 Olginate Dam on the Adda River at the Outlet of Lake Como, Seen from Both Sides................54 8.2 Lake Mendocino....................................................................................................................................................... 57 8.3 Hatanagi I Dam.......................................................................................................................................................... 61 8.4 Kusaki Dam................................................................................................................................................................ 63 8.5 Brisbane 2011 Flood Disaster.............................................................................................................................. 67 8.6 Wivenhoe Dam Flood Discharge on January 11, 2011................................................................................69 Tables 2.1 Options for Estimating Climate Change Impacts on the PMP..................................................................7 3.1 List of Large Ensemble Experiments................................................................................................................10 4.1 Different Aspects of Dynamical and Statistical Downscaling Approaches....................................... 15 7.1 Details of Future Simulated Precipitation Data Used in This Study................................................... 40 7.2 Percentage Increase in Mean and Standard Deviation of Annual Maximum Daily Precipitation, and PMP in Post-1970, Compared to Pre-1970................................................................ 40 7.3 Percentage Area Showing Increase in Mean and Standard Deviation of Annual Maximum Daily Precipitation, and PMP in Post-1970 as Compared to Pre-1970................................................ 40 7.4 Average Percentage Increase in PMP Estimates, Mean, and Standard Deviation of AMDP Series and Corresponding Percentage of Area Showing Increase with Respect to the Post-1970 Period for Three Future Time Periods, Following Three Models and Two RCP Scenarios...................................................................................................................................................................... 41 7.5 Change in Climate Change Parameters Compared with Baseline Period (1980–2000 in Current Study).......................................................................................................................................................... 42 7.6 Projected Percentage Change to 1 in 100 and 1 in 1,000 AEP Peak Flood Discharge..................43 7.7 Estimated Increase Rate In Rainfall with 100-Year Return Period In Japan......................................46 7.8 Estimated Average Increase Rates in Flood Discharge and Frequency of Floods........................46 Acknowledgments This report is one in a series of publications and knowledge products prepared under the Programmatic ASA (Advisory Services and Analytics) on Enhancing the Resilience and Safety of Dams and Downstream Communities (P171966). The World Bank team was led by Satoru Ueda (Lead Dam Specialist) and included Johan Grijsen (Hydrologist, Consultant),  Felipe Lazaro (Senior Dam Specialist), Gregory Calner Felter (Water Resources Consultant), Kimberly Lyon (Senior Water Resources Management Specialist), Ximing Zhang (Senior Water Specialist), Marcus Wishart (Lead Water Resources Specialist), Ayelen Becker (Consultant), Yong Nyam (Consultant), and Georgine Adjibola Badou (Program Assistant). The report is based on several technical background reports prepared by the International Centre for Water Hazard and Risk Management (ICHARM) under the auspices of the United Nations Educational, Scientific, and Cultural Organization (UNESCO), Tsukuba, Japan, and by the Japan Water Agency (JWA). ICHARM’s team was led by Professor Toshio Koike, executive director of ICHARM, and included Gusyev Maksym, Hirosato Yoshino, Katsunori Tamakawa, Rasmy Mohamed, Acierto Ralf, Tomoki Ushiyama, Kentaro Aida, Katsuhiro Onuma, and Keijiro Kubota. JWA’s team was led by Nobuyuki Ichihara and included Kazumi Sasaki, Kentaro Kido, and Kenichiro Yamaguchi. The team is grateful for the guidance provided by peer reviewers, including Amal Talbi (Global Lead for Food and Lead Water Specialist), Rikard Liden (Lead Energy Specialist—Hydropower), Chabungbam Rajagopal Singh (Senior Water Resource Specialist), and Homero Alejandro Paltan Lopez (Climate Specialist, Consultant). Eileen Burke (Global Lead for Water Resources and Lead Water Resources Management Specialist) also provided valuable feedback. The team also acknowledges the strategic direction provided by Saroj Kumar Jha (Global Director, Global Water Practice), Yogita Mumssen (Practice Manager for the Global Water Unit) and Soma Ghosh Moulik (Practice Manager for the Heart of Africa Water Unit and former Practice Manager for the Global Water Unit). The team appreciates the editing provided by GW, Inc. and the support from the Global Department for Water Knowledge and Learning Team, including Erin Barrett (Knowledge Management Analyst), David Gray (Senior Knowledge and Learning Officer), and Ayse Boybeyi (Senior Knowledge and Learning Officer). vii Abbreviations AEP annual exceedance probability CDF cumulative distribution function CDF-t cumulative distribution function transform CESM Community Earth System Model CORDEX Coordinated Regional Climate Downscaling Experiment CRCM Canadian Regional Climate Model d4PDF Database for Policy Decision Making for Future Climate Change DAD depth-area-duration EDCDF equidistant cumulative distribution function matching EFO Ensemble Forecast Operations ESF ensemble streamflow predictions FIRO Forecast Informed Reservoir Operations GCM general circulation model IPCC Intergovernmental Panel on Climate Change LE large ensemble MAF moisture adjustment factors MPI-GE Max Planck Institute Grand Ensemble NWP Numerical Weather Prediction PDF probability distribution function PMF probable maximum flood PMP probable maximum precipitation QDM quantile delta mapping QM quantile mapping RCM regional climate model RCP Representative Concentration Pathway RTFM real-time flood model SEFM stochastic event flood model SMHI-LENS Swedish Meteorological and Hydrological Institute Large Ensemble SSP shared socioeconomic pathway SWG stochastic weather generator WY water year ix Introduction This volume is a companion to the Technical Note Enhancing the Safety and Resilience of Dams in the Context of Climate Change and Extreme Hydrological Events (World Bank 2025). Part I provides the detailed methodologies of climate change impact assessment and Part II provides case studies from different regions on impact assessment and adaptative reservoir operation. The six sections of Part I of this document provide a description of methodologies and tools for the assessment of climate change impacts on extreme precipitation and extreme flows. First, the document presents the estimation of the probable maximum precipitation (PMP) and methodologies for the assessment of climate change impacts on the design precipitation, PMP, and design flows. Subsequently, the issues highlighted by ICOLD (2022) during its 27th Congress are presented, particularly (a) the importance—especially for high-risk dam projects— of the statistical and dynamic downscaling for adjusting output obtained from coarse resolution. General Circulation Models (GCMs) to observed watershed-scale spatial and time distributions of hydrometeorological variables and (b) the importance of considering an ensemble of projected climate scenarios for a given time horizon by not relying on the analysis of a single scenario, which may not reveal the variety of potential hydroclimatic responses to greenhouse gas emissions. Finally, hydrological methods for translating the design precipitation and PMP into, respectively, the design flood or safety check flood and probable maximum flood (PMF) will be summarized. These methodologies play a crucial role in assessing the potential impacts of climate change on precipitation patterns, flood frequency, and extreme events. The two sections of Part II of this document assemble selected case studies relevant to managing floods and reservoirs in the face of changing climate conditions. Those presented in this collection highlight real-world examples from different regions, including Japan, India, Viet Nam, the United States, Australia, and other countries. They demonstrate the importance of effective reservoir operations, pre-flood drawdown operations, and the use of advanced forecasting and decision support systems in mitigating flood risks and ensuring the safety of downstream communities. Overall, these sections aim to provide practical guidance for water resources managers, practitioners, and engineers involved in flood control, reservoir operations, and dam safety issues. By understanding the challenges and solutions presented in these case studies, stakeholders can make informed decisions and develop effective strategies to adapt to the changing climate and ensure the resilience of their water management systems. xi PART I DETAILED METHODOLOGIES OF CLIMATE CHANGE IMPACT ASSESSMENT CHAPTER 1 Estimation of the Probable Maximum Precipitation The concept of PMP is essential in relation to dam safety and flood management. It is defined as “the greatest depth of precipitation for a given duration meteorologically possible for a design watershed or a given storm area, at a particular location and at a particular time of the year, with no allowance made for long-term climatic trends” (WMO 2009). PMP estimates are used in designing infrastructure such as dams, levees, and flood control systems to ensure they can withstand the most severe weather events and catastrophic flooding likely to occur in a given region. Procedures for estimating the PMP cannot be standardized; they vary with the amount and quality of data available, the basin size and location, the regional topography and orographic effects, the storm types producing extreme precipitation, and the climate. This section provides a brief summary of the PMP estimation methodologies described in great detail in WMO 2009. METHODS FOR PMP ESTIMATION The methods described in this appendix, except the statistical method, are based on a hydrome- teorological approach, which consists essentially of moisture maximization and the transposition of observed storms or combinations of storms. WMO (2009) distinguishes six PMP estimation methods for general use and two additional methods for extremely large watersheds: • Statistical method (statistical estimation) • Local method (local storm maximization) • Transposition method (storm transposition model) • Combination method (temporal and spatial maximization of storms or storm combination model) • Inferential method (theoretical or ratiocination model) • Generalized method (generalized estimation) • Major temporal and spatial combination method • Storm simulation method based on historical floods Statistical PMP Estimation • Statistical method: The PMP is estimated by a method proposed by Hershfield (1965), with data from numerous gauging stations in a meteorologically homogeneous zone and using the hydrological frequency analysis method together with the regional generalized method. The critical parameter represents the statistical maximum of the observed storm values. It is applicable for small watersheds less than 1,000 km2. 2 Estimation of the Probable Maximum Precipitation 3 Physical PMP Estimation • Local method: The PMP is estimated using available observed maximum storm data for the study area and is applicable if there are several years of observed data. This is also referred to as local storm maximization. • Transposition method: The PMP is estimated by transposing extreme maximum storms observed in adjacent areas to the study area. The critical consideration for this method is to determine the storm transposition probability, while making a variety of adjustments for the transposed storm, based on the differences in geographic and topographic conditions between the area where the original storm occurred and the design area. This method is also referred to as the storm transposition model. • Combination method: The PMP is estimated from a reasonable combination of two or more storms in the study area, based on principles of synoptic meteorology and experi- ence of synoptic forecasting, to form a sequence of artificial storms with a long duration. A key aspect of the method is determining a reasonable storm combination, which requires strong meteorological knowledge. This method is typically used for large watersheds with storms of long duration. This is also referred to as temporal and spatial maximization of a storm model. • Inferential method: The PMP is estimated using a simplified storm equation based on  three-dimensional storm weather systems in the study area, which requires strong  ­meteorological knowledge and upper-air meteorological observation data. This method is applicable for watersheds with an area of hundreds to thousands of square kilometers. • Generalized method: The PMP is estimated by generalizing depth-area-duration (DAD) curves of transposed storms. The spatial distribution is based on generalized isohyets, and the temporal distribution is a single-peak generalized hyetograph. This method requires a significant amount of observed data and is applicable for watersheds less than 13,000 km2 in orographic regions and less than 52,000 km2 in nonorographic regions. Methodologies for Large Watersheds In addition, two other methods can be used for deriving PMP/PMF in extremely large watersheds. • Major temporal and spatial combination method: This is similar to the combination and generalized methods. It is a combination of detailed and rough estimation methods divided into two areas depending on its effects on the PMF. Parts of high impact on the PMF are estimated using detailed physical methods as described earlier, whereas the remaining parts with lower impact on the PMF are estimated using common correlation and flood distribution methods. This methodology is applicable for large watersheds with large differences between upstream and downstream weather conditions. • Storm simulation method based on historical floods: Through reverse engineering and hydrological modeling, this method recreates a storm that could potentially have created a selected historical flood. It is inherently based on the incomplete temporal and spatial distri- bution information of a known extraordinary flood. The PMP is subsequently derived through moisture maximization. This method requires data availability on the flood hydrograph at the design section and knowledge of the rainfall, hydrological, and flooding conditions in parts of the basin. 4 Enhancing the Safety and Resilience of Dams in the Context of Climate Change Alternative Probabilistic and Stochastic and NWP-based PMP There are a number of complex alternative methodologies for estimating the PMP, including the use of Numerical Weather Prediction (NWP) models, as follows: • Stochastic storm transposition method (Kunkel, Novak, and Steurer 1999): The PMP esti- mation is similar to the typical storm transposition method, but it integrates the probability of occurrence of the storm, which is assumed to have the same probability of occurrence any- where in a large meteorological homogeneous area. • Multifractal scaling method (Lovejoy and Schertzer 1992): The PMP is estimated based on defined scaling laws of observed maximum precipitation and uses multifractal scaling in estimat- ing extreme precipitation. Analytical expressions for intensity-duration-frequency estimation using multifractal scale invariance theory were developed similarly to probabilistic frequency estimation methods for estimating extreme precipitation. • Frequency-based estimation method (WMO 2009): The PMP is estimated using frequency analysis for given return periods or exceedance probabilities. The main approaches for fre- quency analysis include block-annual-maximum using three limiting distributions (Gumbel, Fréchet, and reversed Weibull) and the peak-over-threshold method using limiting distributions (generalized extreme value, generalized Pareto) and assuming flexible distributions (general- ized gamma, Burr XII, Burr III) with consistent descriptions of all nonzero and extreme values. The key issue is estimating the parameters used in the probability distributions. The estimation of the tails is the most crucial part of this method in estimating maximum precipitation. • Storm simulation method using atmospheric model (Schär et al. 1996): The PMP is esti- mated using an NWP model. This is a re-creation of the precipitation pattern as a result of storm simulations based on three-dimensional meteorological variables, with reanalysis data and the output of global climate models (GCMs) as input. It is noted that general circulation models (GCMs) are typically not directly used for estimating the PMP, which is a localized extreme precipitation event that can occur in specific areas and is not fully captured by global or regional atmospheric models. However, GCMs can provide information on large-scale climate patterns and atmospheric conditions that can influence extreme precipita- tion events, such as the El Niño Southern Oscillation or other teleconnection patterns. This infor- mation can be used to support PMP estimation by providing context and identifying the potential for extreme precipitation events in a given area. Whereas these models can be used to simulate future climate scenarios and the potential changes in precipitation patterns and extreme events, useful for assessing future PMP conditions, such simulations should be evaluated with caution and calibrated against observed precipitation data to ensure their accuracy and reliability. PMP CALCULATIONS Hershfield Statistical PMP Estimation For the statistical PMP calculation, Hershfield (1965) proposed a statistical equation for the esti- mation of the PMP, based on the general frequency equation, as follows: (1.1) Estimation of the Probable Maximum Precipitation 5 where PMPS is the PMP estimate for the station S and and Sn are, respectively, the mean and standard deviation of a series of annual maximum rainfall data for n year. Km is the frequency factor, which is a statistical representation of the maximum value in the observed storm series, given by equation (1.2): (1.2) where Xm is the maximum value in the annual maximum series of rainfall data at the station, and are the mean and standard deviation of the annual maximum series of rainfall data, respectively, for (N-1) years after removing the extraordinarily large value Xm. The essence of the Hershfield method is storm transposition, but the abstracted statistic Km is transposed instead of the specific rainfall amount of one storm. Standard literature practices suggest a maximum value of 12 as the threshold of Km. Aspects of Physical PMP Estimation Equation (1.3) shows the generalized equation for physical PMP estimation: PMPp = Pmax * MF * AF (1.3) Pmax is the maximum observed precipitation; MF is the maximization factor—that is, the ratio of the maximum precipitable water to the actual precipitable water of major observed storms; and AF is an adjustment factor for topography and location. Topographic adjustments are necessary when orographic effects influence precipitation patterns and storm characteristics. Moisture maximization plays a critical role in preparing PMP estimates because it involves maxi- mizing the moisture content in the atmosphere to determine the maximum amount of precipita- tion that could occur under extreme conditions as the worst-case scenario. The crux of moisture maximization lies in understanding and incorporating all possible factors that could contribute to moisture in the atmosphere, such as temperature, humidity, wind patterns, and geographic features. By maximizing these factors within reasonable bounds, a theoretical upper limit on precipitation can be estimated for a given area and time frame. Sequential and spatial maximization involves the development of hypothetical flood-producing storms by combining observed individual storms or rainfall bursts within individual or separate storms that may also be transposed or geographically adjusted, referred to as spatial maximiza- tion. This step is mainly used in combination methods and application to large watersheds. Enveloping is the process of selecting the maximum values in DAD relation plots for maxi- mum precipitation depths for a given duration for various areas (figure 1.1). This process is also applied in the selection and determination of the parameter Km used for estimating the final value of the statistical PMP. DAD analysis is a method used to estimate the precipitation depth over a particular area for a specific duration of time. It is useful for understanding and predicting the magnitude and 6 Enhancing the Safety and Resilience of Dams in the Context of Climate Change distribution of rainfall in a given area. The maximum depth-area curves for a given duration of a storm are prepared from isohyet maps for selected storm durations. FIGURE 1.1 Example of Depth-Area-Duration Curves Maximum rainfall (mm) 1000 900 800 700 600 500 400 300 200 100 0 50,000 100,000 150,000 200,000 250,000 Area (km ) 2 6hrs 12hrs 18hrs 24hrs 36hrs 48hrs 72hrs Source: Original figure for this publication. Generalized PMP Estimation The generalized PMP estimation method typically follows these steps: • Determine the possible transposition range of storms based on the differences in geographic and topographic conditions between the original storm occurrence area and the design area by using various maps (weather, topography, and preliminary total storm isohyets). • Acquire precipitation data for different storms within the limits of the transposition region. • Prepare the DAD analysis according to the WMO Manual (WMO 1969). • Determine the persisting 12-hour dew point and 24-hour mean wind speed for each storm. • Determine the highest maximum 12-hour dew point and the highest maximum 24-hour mean wind speed on location and on the transposition site, determining the moisture-inflow maximization. • Calculate the combined transposition and maximization ratio of precipitable water and the moisture index corresponding to the 12-hour persisting dew point. • Multiply values for appropriate areas using DAD data with the maximum moisture index and the appropriate precipitable water amount. • Plot graphs of transposed maximized DAD values. • Apply orographic adjustments as necessary. • Consider seasonal, areal, and temporal variations of PMP estimates when the maximum flood is likely to result from a combination of snowmelt and rainfall. CHAPTER 2 Climate Change Impacts on Design Rainfall and PMP It is known with high confidence that global surface-level temperatures will continue to increase with increasing atmospheric greenhouse gas concentrations. The Intergovernmental Panel on Climate Change (IPCC), (2023)1 and that the increasing temperature caused by climate change causes an intensification of precipitation and streamflow extremes (Wasti and Ray 2021). For example, the Clausius-Clapeyron equation predicts the rate at which the saturation vapor pressure increases per unit increase in temperature (Clapeyron 1834; Clausius 1850; Koutsoyiannis 2012). Its relevance to meteorology and climatology is the increase of the atmospheric moisture holding capacity by about 7 percent for every 1°C rise in temperature. It is used to predict the intensification of extreme precipitation under climate change globally (Allen and Ingram 2002; O’Gorman and Muller 2010). However, the upper extreme precipitation (99.9th percentile events) is observed to scale significantly faster (between 6 percent and 15 percent per °C) depending on model parameters and other details (Bao et al. 2017). Moreover, the increase in storm runoff extremes is higher than the increase pro- jected using the Clausius-Clapeyron scaling for the intensification of precipitation extremes (Yin et al. 2018) because of a precipitation elasticity2 > 1 (Grijsen 2014, 2022) of runoff and peak flows. Table 2.1 shows four options for estimating climate change impacts on design rainfall and the PMP. Methods include the application of change ratios between the present and future precip- itation, incorporating GCM outputs, the use of stochastic or physical PMP estimation methods, and the use of storm simulations with the input from precipitation reanalysis modeling and GCMs. The first option is to use statistical methods, in which the input data consist of rainfall data derived from GCM-based outputs. The impact of climate change on the design rainfall and PMP is inherently included in the estimation of Pn, Km, and Sn and the maximization of the pre- cipitation is included in the estimation of the factor Km. Change ratios between the future and the present can be applied to this factor. The advantage of this method is that it is simple in its application, as provided by Sarkar and Maity (2020). TABLE 2.1 Options for Estimating Climate Change Impacts on the PMP Climate change Design precipitation and Options Data used Variables used uncertainty PMP estimation 1 GCM-based outputs Rainfall GCM ensemble Statistical method 2 GCM-based outputs Rainfall, dew point, wind GCM ensemble Physical method 3 GCM-based outputs Rainfall and other variables GCM ensemble Probabilistic and stochastic method 4 GCM-based outputs Meteorological variables GCM ensemble NWP-based method Note: GCM-based outputs can include raw GCM data outputs, statistically downscaled and bias-corrected GCM outputs, or dynamically downscaled and statistically bias-corrected RCM outputs. GCM = General Circulation Model; NWP = numerical weather prediction; PMP = probable maximum precipitation; RCM = regional climate model. 7 8 Enhancing the Safety and Resilience of Dams in the Context of Climate Change The second option is to use a physical estimation method and use several variables, such as rainfall, dew point, and wind data from GCM-based outputs. The impact of climate change can be estimated through the change ratios for these variables, which can be estimated and easily applied to equation (C.3). The main advantage of this method is that it allows an understanding of the future changes from climate change caused by several variables. The disadvantage is that it will require more data and computational resources compared with the first option. Examples are provided by several studies (Kunkel et al. 2013; Micovic, Schaefer, and Taylor 2015; Rouhani et al. 2017; Thanh Thuy, Kawagoe, and Sarukkalige 2019). The third option is to employ probabilistic and stochastic design rainfall and PMP estimators using data from GCM-based outputs. The selected frequency distributions can be used for the rainfall variable and for other variables relevant for estimating storm characteristics. The fre- quency distributions can incorporate nonstationarity assumptions that are not considered in statistical and physical design rainfall estimators. In addition, return periods or the exceedance probability of the estimates can be calculated and compared with existing intensity-duration-­ frequency curves. Examples of this option were also provided by several studies (Alaya, Zwiers, and Zhang 2018; Papalexiou and Koutsoyiannis 2006). The fourth option is to use storm simulations using NWP models with the input of meteoro- logical variables from GCMs. This option enables the investigation of more changes in storm characteristics in the study area. The main advantage is its ability to thoroughly investigate the mechanism of the maximized storm or PMP because of climate change. However, the main disadvantage is the high requirements of data, experience with climate (change) modeling, and computational resources as compared with the previous assessment methods. Examples of this last option are provided in several studies (Chen et al. 2016; Ishida, Kavvas et al. 2015; Ishida, Ohara et al. 2018; Ohara et al. 2011, 2012). An in-depth comprehensive review of technical issues related to incorporating climate change in the PMP estimation can be found in several studies (Chen and Hossain 2019; Fluixá-Sanmartín et al. 2018; Hultstrand and Kappel 2020; Johnson and Smithers 2020; Salas, Anderson et al. 2020). For uncertainty analysis on PMP estimation, several examples of studies (Alaya, Zwiers, and Zhang 2018; Kim et al. 2016; Micovic, Schaefer, and Taylor 2015; Salas, Gavilán et al. 2014) have incorporated uncertainty in PMP estimation in novel ways as compared with the typical uncertainty analysis through the usage of multiple GCMs. NOTES 1. https://www.ipcc.ch/assessment-report/ar6. 2. A precipitation elasticity of, for example, 1.5 for extreme flows indicates that a 10 percent increase in the maximum precipitation causes 15 percent increase in the maximum runoff or peak flow. CHAPTER 3 The Importance of Large GCM Ensemble Experiments The importance of large-ensemble climate projection experiments has emerged because such experiments can (a) separate uncertainty between natural variability and noises resulting from our incomplete understanding of the climate system and its imperfect representation in models, (b) reduce uncertainty in the estimated occurrence frequency of an extreme event, and (c) perform factor analyses of climate variations, such as detection and event attribution. This section discusses several large-ensemble climate projection experiments conducted in recent years for dealing with climate model and emission scenario uncertainties (table 3.1). Jenkins et al. (2009) addressed important data limitations that users need to consider. First, global and regional projections cannot cover all potential future climate outcomes—for example, some potential influences on the future climate are not yet understood well enough to be included in climate models. And the number of future greenhouse gas emission scenarios is limited. Second, GCMs use different spatial resolutions and methods and display biases. Third, GCMs do not explicitly represent smaller-scale processes, such as atmospheric convection. GCMs adopt a set of methods, including a modeling approach, treatment of structural uncertainty, and use of a simple energy balance climate model. Each stage of the methodologies has many inherent assumptions and relies on expert judgment. UKCP18 provides projections for HadGEM3-GC3.05, with a 60-km grid for fifteen members plus thirteen members of CMIP5 results (Met Office 2018).1 The GC3.05 simulations were produced by generating a perturbed parameter ensemble from 1900 to 2100. UKCP18 also includes down- scaling by regional climate models (RCM) to 12-km grids for the United Kingdom and Europe. For the UK area, further downscaling to 2.2 km is provided. The warming trend, rainfall increase, and sea level rise were carefully studied. For example, the UK area will experience an increase in annual average rainfall; the recent decade (2008–17) has been 11 percent wetter than 1961–90 and 4 percent wetter than 1981–2010, with overlapping natural variability. The climate would shift toward warmer, wetter winters and hotter, drier summers in the twenty-first century. Another attempt at large-ensemble projection is the Community Earth System Model (CESM) Large Ensemble Project (Kay et al. 2015). It consists of thirty ensemble simulations for 1920–2100 with minor atmospheric initialization differences while using the same model (CESM1) at approximately 1˚ horizontal resolution. The CESM1 model combines components of atmo- sphere, ocean, land, and sea ice. In addition to land carbon cycle calculations, the CESM Large Ensemble (CESM-LE) simulations also include diagnostic biogeochemistry calculations for the ocean ecosystem and the atmospheric carbon dioxide cycle. The CESM-LE provided many more ensemble members than UKCP18 and produced more robust results, enabling the separation of forced variability from natural climate variability. Kay et al. (2015) compared the warming trend found from their experiments with Coupled Model Intercomparison Project Phase 5 (CMIP5) results and concluded that the spread was generated by 9 10 Enhancing the Safety and Resilience of Dams in the Context of Climate Change TABLE 3.1 List of Large Ensemble Experiments Project Country Model Scenario Resolution Year UKCP18 United Coupled RCP2.6, RCP8.5 60 km 1900–2100 x 15 (HadGEM3-GC3.05) Kingdom CESM-LE United Coupled RCP8.5, 1.0 deg 1920–2100 x 40 CESM-ME States RCP4.5 1920–2100 x 15 CanESM2 Canada Coupled RCP8.5 200 km 1950–2100 x 50 MPI-GE Germany Coupled RCP2.6, RCP4.5, A:1.8 deg, 1850–2005, 2006–99 x 100 RCP8.5 O:1.5 deg d4PDF Japan Atmos +4°C, +2°C, +1.5°C 60 km 1951–2010 x 100, 60 x 90 (+4K), 60 x 54 (+2K, +1.5K) SMHI-LENS Sweden Coupled SSP1-1.9, SSP3-3.4, A:80 km, 1970–2100 x 50 (EC-Earth3) SSP5-3.4-OS, SSP5-8.5 O:1 deg Note: CESM = Community Earth System Model; MPI-GE = Max Planck Institute Grand Ensemble; RCP = Representative Concentration Pathway; SSP = Shared Socioeconomic Pathway. internal climate variability alone. The internal variability of GCMs refers to the range of natural fluctuations in climate that are simulated within a model in the absence of any external forc- ing. These fluctuations arise from the complex interactions among the various components of the climate system, such as the atmosphere, oceans, land surface, and cryosphere. The internal vari- ability of GCMs can manifest in a range of temporal and spatial scales, ranging from short-term fluctuations in temperature and precipitation to longer-term variations in the frequency and inten- sity of extreme events, such as heat waves and droughts. It is important to distinguish the internal climate model variability from the forced climate variability, driven by greenhouse gas emissions. Similarly, the Canadian Earth System Model 2, Canadian Earth System Model version 2 - Large Ensemble (CanESM2-LE) was produced by the Canadian Centre for Climate Modeling and Analysis and consists of fifty members for the period 1950–2100 under the Representative Concentration Pathway (RCP) 8.5 scenario (Fyfe et al. 2017). They generated two fifty-member ensemble sets: one with natural and anthropogenic forcing and another with only natural forcing. After studying the observed trend of annual maximum snow-water equivalent in North America, the authors found that only the LE simulation with the natural and anthropogenic forcing could explain the observed trend. The Canadian Regional Climate Model 5 (CRCM5) Large Ensemble (CRCM5-LE) provides a dynamical downscaling of CanESM2 results in a 12-km grid resolution over Europe and Northeastern North America (Leduc et al. 2019). The downscaled version (CRCM5-LE) allows a more realistic representation of local extremes than CanESM2. Innocenti et al. (2019) studied the annual maximum precipitation using the fifty-member CRCM5-LE. They showed large increases in annual maximum precipitation, especially for the shortest duration and for more extreme events of the longest return periods over Northeastern North America. The Max Planck Institute Grand Ensemble (MPI-GE) simulated large ensemble projections using one hundred members for 1850–2005 and three future scenarios (RCP2.6, 4.5, 8.5, 2006–99) to provide large ensembles (Maher et al. 2019). They intensively studied separating the model responses to internal climate variability from the responses to forced climate variability. The dependence of the pattern of temperature increases on the RCP scenarios was rather small— that is, the warming pattern was generally consistent among the historical, RCP2.6, RCP4.5, and RCP8.5 scenarios. The pattern of precipitation changes was less consistent among the scenarios than the pattern of temperature increases. Several areas (parts of Africa, Southeastern Asia, and The Importance of Large GCM Ensemble Experiments 11 Eastern South America) showed a different precipitation tendency across the RCP scenarios, which were larger than the intermember variability. This indicated that the scenario differences matter in these regions. A Japanese team provided the largest ensemble projection with the highest resolution, named Database for Policy Decision Making for Future Climate Change (d4PDF; Ishii and Mori 2020; Mizuta et al. 2017). Because complex mountain ranges in the Japanese Islands play a key role in precipitation distribution, the team gave the highest priority to spatial resolution. The d4PDF has a different approach from previous large-ensemble projections. It is based on the atmospheric MRI- AGCM3.2 GCM (60-km horizontal resolution), and the ocean lower-boundary conditions are given by observations (historical) and six CMIP5 models (future) plus perturbations. The historical and nonwarming simulation provided one hundred members for the period 1951–2010, whereas the future +4°C simulation has ninety members for the climate state where the global mean tem- perature is 4°C warmer than the preindustrial climate. In the future simulation, the amplitude of the warming is kept constant throughout the sixty-year integration. Later on, +2°C and +1.5°C simulations with fifty-four members were provided (Fujita et al. 2019; Nosaka et al. 2020). The d4PDF reproduced the South Asian and Southeastern Asian monsoon precipitation patterns and tropical cyclone trajectory distributions well but underestimated the tropical cyclone intensity. The d4PDF subsets provide dynamical downscaling to 20-km and 5-km resolution for the area of Japan. This data set is useful for the detection of extreme precipitation for long return periods, such as 200 years, which is required to design hard infrastructure counter measures for natu- ral hazards. Tachikawa et al. (2017) projected future changes of extreme river discharges for three main river basins in Japan using the d4PDF rainfall data set downscaled to a 20-km grid. The authors showed that the probability of occurrence of 1 in 200 years events for the annual maximum twenty-four-hour rainfall of the +4°C simulations can be 1.3 to 1.4 times larger than past events. They also found that the annual maximum discharge can be 1.5 to 1.7 times larger in the future than in the past (suggesting a precipitation elasticity of extreme flood events of about 1.2). Their research showed also that the 200-year annual maximum twenty-four-hour rainfall event in the +4°C future climate is equivalent to the 900-year event in the past climate. Large-ensemble projection data thus enable the quantitative assessment of frequency estimation and statistics of natural hazards. Finally, the Swedish Meteorological and Hydrological Institute performed large-ensemble simu- lations (so called SMHI-LENS) with fifty members, covering the period 1970–2100 (Wyser et al. 2021). Their model forcing comprises the shared socioeconomic pathways (SSP) defined for CMIP6 (O’Neill et al. 2016), which takes into account societal development. Specifically, simu- lations were conducted using SSP1-1.9, SSP3-3.4, SSP5-3.4-OS, and SSP5-8.5 scenarios, all of which except SSP5-8.5 were from Tier 2 (second priority scenarios) of CMIP6. The institute used the EC-Earth3 GCM, with 80-km horizontal resolution for the atmosphere and 1˚ resolution for the ocean. None of the scenarios is able to keep global warming below 1.5°C, and only SSP1-1.9 can keep 2°C warming with a 70 percent probability. NOTE 1. For more information on the UKCP18, visit the Met Office website at https://www.metoffice.gov.uk​ /research/approach/collaboration/ukcp/. CHAPTER 4 Downscaling of GCM Output NEED FOR DOWNSCALING OF GCM OUTPUT There is a large gap between the grid resolution of GCMs and the resolution of catchment-scale hydrological models. For example, the aforementioned d4PDF global projection data set pro- vides projections with a return period of 5,400 years for all 4°C temperature increase scenarios, which is the maximum length of data available. However, its 60-km horizontal resolution limits the use of the data for small catchments upstream of dams, whereas most GCMs have resolu- tions in the order of 100 km (10x10). To assess extreme flows for catchments of 1,000 to 10,000 km2, a horizontal resolution of 20 km is needed (Tachikawa et al. 2017). Moreover, to simulate heavy precipitation, an RCM having a grid interval of 5 km or less is required because cumulo- nimbus clouds have dimensions of about 20km. The model can generally resolve meteorological phenomena with its scale up to five times of the model’s grid interval. For example, if the grid size of the model is 4-5 km, it can resolve meteorological phenomena with its scale of around 20-25 km. GCMs used for climate change projections cannot provide ones with high spatial resolutions because of computational limitations. Although each component of the physical, chemical, and biological processes is calculated with the high temporal resolutions required to simulate each process, not all calculated outcomes are stored as output because they usually require extremely large storage volumes. Only selected output is stored with rather low temporal and spatial res- olutions. For these reasons, GCM output generally consists of limited components with low spatial and temporal resolutions. On the other hand, the water cycle phenomenon over the land area depends on a complex spatial distribution characterized by topography, geology, soil, vegetation, land use, and other characteristics, whereas extreme events, such as floods, have a short time scale. For these reasons, high spatial and temporal resolutions are required to fully understand the details of the hydrological cycle. Thus, to enable the use of GCM outputs as input for hydrological models, it is necessary to con- vert output with low spatial and temporal resolutions to output with high resolutions and derive the input needed for hydrological modeling from the limited data stored as GCM output. As such, downscaling techniques are used to bridge the gap between the coarse resolution of GCMs and the finer resolution needed for regional or local climate impact assessments. There are two main types of downscaling of GCM output data—that is, dynamical and statistical downscaling. The former refers to the nesting of a fine-scale resolution—through an RCM—in the large-scale resolution of GCMs while preserving spatial correlation. However, this method 12 Downscaling of GCM Output 13 is computationally demanding and impractical for multidecade simulations of different GCMs. The latter is instead based on the relationship between the large-scale circulation and local-scale phenomena and can be implemented at reasonable computational costs. If there is a gap (bias) between GCM outputs and observed data regarding the current climate conditions, bias correction needs to be performed as well, which is described separately in the “Methods of Bias Correction” section. This is sometimes done as part of the statistical downscal- ing process. STATISTICAL DOWNSCALING The statistical downscaling of temporal and spatial resolutions requires synoptic-scale climate data that cover the climate characteristics of the target region and long-term, quality-controlled observed in situ meteorological data that can be used to calculate the statistical properties of the climate characteristics of the target domain. Statistical downscaling methods (Iizumi et al. 2010; Wilby and Charles 2004) include the following: • Empirical statistical downscaling for the downscaling of spatial resolutions based on the statistical relationships between the large-scale atmospheric variables simulated by GCMs and the local climate variables at a specific location. Statistical techniques, such as simple or multiple regression equations, neural networks, or principal component analysis are used to establish the relationship between historic climate observations and GCM output for the refer- ence period and then apply those to GCM output for future periods. Multiple linear regression involves developing a linear relationship between large-scale atmospheric variables (predictors) from GCM outputs and local/regional climate variables (predictands) observed at a finer scale. Generalized linear models are extensions of linear regression models that can accommodate nonnormal error distributions and nonlinear relationships between predictors and predictands. Artificial neural networks can learn complex relationships between GCM predictors and pre- dictands without needing explicit assumptions about the underlying statistical distributions. Empirical quantile mapping (QM; discussed in “Methods of Bias Correction”) involves matching the cumulative distribution functions (CDFs) of GCM outputs with observed data to correct biases and match variability in the distribution of climate variables. Because of its computational advantages, empirical statistical downscaling is the most frequently used downscaling method in combination with bias correction. • Weather typing groups large-scale atmospheric patterns into a limited number of discrete weather types, which are associated with local climate conditions. It correlates the distri- bution of barometric pressure and potential temperature with local temperature and pre- cipitation. This approach can be combined with statistical methods to provide downscaling information. • Stochastic weather generators (SWGs) simulate daily weather variables, such as tempera- ture, precipitation, and humidity, at fine spatial and temporal resolutions based on statistical properties derived from GCM outputs. The model is calibrated using historical weather obser- vations and large-scale atmospheric variables simulated by GCMs. Although no atmospheric dynamical relationship exists between temporal changes and the elements that are generated by the weather generator, this method ensures statistical validity for the results. Weather generators are discussed in a later section. 14 Enhancing the Safety and Resilience of Dams in the Context of Climate Change DYNAMICAL DOWNSCALING Methodology: Dynamical downscaling uses RCMs, which cover limited areas but have higher spatial and temporal resolutions than GCMs. An RCM is essentially equivalent to a regional model for numerical weather prediction and provides variables that are physically consistent with each other, which is not guaranteed in statistical downscaling. Higher-resolution RCMs can account for orographic precipitation enhancement because of complex terrain, land use, and water distribution. They can simulate a more realistic atmospheric convection, which has a stronger updraft within a smaller horizontal extent. Nonhydrostatic models are used for cases with grid intervals of less than 10 km to allow for severe convection with strong updrafts. GCMs provide the boundary conditions. The RCM simulations can provide more detailed and accurate information about the future climate conditions at a regional scale, which is important for many climate-related applications, such as water resources management, agriculture, and infrastruc- ture planning. However, the projected variables may still display a bias, which requires a bias correction method to be applied based on observed data. Requirements: Dynamical downscaling requires considerable computational resources. For example, when the horizontal grid intervals are cut in half, the computational time increases at least eight times, which requires high-performance computing. Lateral and lower boundary conditions are provided by GCMs, including principal variables, such as wind speed, tempera- ture, pressure, and specific humidity, for the three-dimensional space up to and including the troposphere1 and for the two-dimensional near-surface space. Time intervals of six hours are required to reproduce diurnal weather variations. Higher-resolution RCMs are required in case of complex terrains or complex coastlines. Historical observation data are required to evaluate projected outputs, such as precipitation and temperature. For bias correction, long-term obser- vation data are required. Downscaling steps: GCMs provide boundary conditions for RCMs. It is generally recom- mended to bias-correct GCM output before using it as input for an RCM to remove systematic errors or biases that may affect the accuracy of the simulations. GCM biases can be magnified when output data are used to drive RCMs at higher spatial resolutions. The bias-corrected data provide the boundary conditions for the RCM, which then simulates the regional-scale climate conditions based on the GCM output. Simulation results obtained from the RCMs, such as precipitation and temperature, should rep- resent the statistical features of observed variables for the reference period. Statistical evaluation should be applied to compare RCM output to observed variables. A popular evaluation method is the frequency of appearance, which plots scatter diagrams with daily rainfall intensity on the X-axis and its relative frequency on the Y-axis (De Troch et al. 2013). This plot represents the characteristics of daily precipitation, and the correspondence of observations and simulation results guarantees good model performance. Spatial rainfall distribution or seasonal variation can be evaluated as well. If the evaluations do not show good agreement between observations and simulations, the RCM needs to be further fine-tuned until the simulated results are satisfactory. However, even after adequately calibrating RCMs, obtained output variables may still display bias because of incomplete representations by the parent GCM and/or the RCM. In this case, further bias correction techniques may be applied to reduce the remaining bias. Such techniques (for example, Teutschbein and Seibert 2012) are discussed in “Methods of Bias Correction.” Downscaling of GCM Output 15 COMPARISON OF DYNAMICAL AND STATISTICAL DOWNSCALING METHODOLOGIES Outputs, requirements, advantages, disadvantages and applications of dynamical and statistical downscaling are summarized in table 4.1. The merit of dynamical downscaling is that it provides climate variables that are internally con- sistent with one another and satisfy atmospheric governing equations. The output variables consider complex terrains, land use, and water distributions. However, a major disadvantage of dynamical downscaling by RCMs is that it is computationally expensive and resource inten- sive, demanding much expertise and access to high-performance computing resources, as well as extensive input data, such as land use and topography data. In addition, RCM simulations are sensitive to the choice of model configuration, including the choice of physical parame- terizations, spatial resolution, and domain size. These choices can affect the simulation results and introduce uncertainties into the downscaling process. Another disadvantage of dynamical downscaling is that it can be subject to biases and errors that are inherent in the RCMs them- selves. For example, RCMs may have biases in the simulation of certain climate variables or phe- nomena, such as precipitation or atmospheric circulation patterns. These biases can be magnified when RCMs are used to simulate climate conditions at higher spatial resolutions. TABLE 4.1 Different Aspects of Dynamical and Statistical Downscaling Approaches Dynamical downscaling Statistical downscaling Provides • 3- to 25-km grid cell information • Any scale, down to station-level information • Information at sites with no observational data • Daily time series (only some methods) • Daily time series • Monthly time series • Monthly time series • Scenarios for extreme events (some methods) • Scenarios for extreme events • Scenarios for consistently observed variables Requires • High-performance computing • Medium/low computational resources • Large volume of data inputs • Medium/low volume of data inputs • Reliable GCM simulations • Sufficient and good quality observational data • High expertise in climate modeling • Reliable GCM simulations Advantages • Based on consistent physical equations • Computationally efficient, allowing for multiple • Resolves atmospheric and surface processes emissions scenarios and GCM pairings occurring at sub-GCM grid scale following complex • Methods range from simple to elaborate and are terrains, land use, and water distributions flexible enough to tailor for specific purposes • Not constrained by historical records so that novel • The same method can be applied across scenarios can be simulated regions or the entire globe, which facilitates • Experiments involving an ensemble of RCMs are comparisons across different case studies becoming available for uncertainty analysis • Methods rely on observed climate statistics as a basis for driving future projections • Provides point-scale climatic variables Disadvantages • Computationally intensive • High-quality observed data may be unavailable • RCMs are typically driven by only one or two GMC/ for many areas or variables emission scenario simulations • Assumes that relationships between large and • Limited number of RCMs available, and model results local-scale processes will remain the same in are not available for many parts of the globe the future (stationarity assumptions) • May require further downscaling and bias correction • The simplest methods may provide only of RCM outputs projections at a monthly resolution • Results depend on RCM assumptions • Affected by bias of driving GCM Applications • Country- or regional-level assessments with • Weather generators in widespread use for significant government support and resources crop-yield, water, and other natural resource • Future planning by government agencies across modeling and management multiple sectors • Future planning considering uncertainty by • Impact studies that involve various geographic areas government agencies across multiple sectors Note: GCM = general circulation model; RCM = regional climate model. 16 Enhancing the Safety and Resilience of Dams in the Context of Climate Change STOCHASTIC WEATHER GENERATORS SWGs produce synthetic time series of weather data of unlimited length for a location based on the statistical characteristics of observed weather at that location (box 4.1). They are used in var- ious applications, such as climate change research, hydrological modeling, agricultural planning, risk assessment for natural disasters, and infrastructure design. They provide valuable insights into the potential impacts of climate variability and change on different sectors of society. SWGs are important for climate stress tests and temporal downscaling when the temporal scale of GCM output needs to be adapted to the needs of watershed modeling (Doblas-Reyes et al. 2021). Weather generators are often used in temporal downscaling—for example, when GCM out- puts are available on a monthly time step yet watershed models usually require a daily time step. Weather generators are models that simulate the variability and randomness inherent in weather patterns based on stochastic processes and mimic the behavior of weather vaiables, such as temperature, precipitation, humidity, wind speed, and so on. Model parameters are adjusted to match the statistical properties of observed weather data to ensure that the simulated weather closely resembles real-world conditions. The future climate is simulated by perturbing the stochastic model parameters. The simulation process involves generating random sequences of weather variables based on the stochastic models and their parameters. Weather generators are thus parametric models relying on mathematical formulae with explicit ran- dom elements for mimicking actual weather data, usually at a daily time scale and for individual weather stations. Temporal downscaling using weather generators requires adjusting the climate parameters based on the changes between current observed and future projected GCM climate sta- tistics. The most commonly used stochastic precipitation generator is based on a first-order Markov chain process in which daily binary (dry or wet) precipitation occurrences are generated and the amount of precipitation on wet days is generated independently based on a regional or local proba- bility distribution. As with most statistical downscaling methods, a major benefit of weather gener- ators is their ability to rapidly develop long time series of climate scenarios for studying the impact of rare climate events. These weather generators also ensure that combinations of small-scale condi- tions (local grid) remain consistent with large-scale models (for example, GCM grid). BOX 4.1 Stochastic Weather Generators Weather generators are computer algorithms designed to generate—based on statistical models derived from historical weather observations—long series of synthetic weather data, such as temperature, precipitation, humidity, wind speed, and solar radiation, for a specific location or region and time frame. The parameters of the model are conditioned on existing meteorological records to ensure that the characteristics of historical weather emerge in the daily stochastic processes. Weather generators are a common tool for extending meteorological records, supplementing weather data in a region of data scarcity, disaggregating seasonal hydroclimatic forecasts, and downscaling coarse, long-term climate (box continues next page ) Downscaling of GCM Output 17 BOX 4.1 Stochastic Weather Generators (Continued) projections to fine-resolution daily weather for impact studies. The generators allow users to simulate and analyze weather patterns over time for a specific location or region. They find application in climate research, agriculture, hydrology, urban planning, and various industries requiring weather data for planning and decision making and for risk assessment based on simulated weather scenarios. The generation of long-time series permits the design of adaptation strategies that are robust to a wide range of potential future climatic scenarios. The World Bank has used weather generators to characterize climate risk in conditions in which data scarcity and uncertainty prevail. Various open source software is freely available, even though basic skills in terms of programming and climate sciences may be required for use: • The WGEN-Weather-Simulatora is a stochastic weather generator (SWG), originally developed at the U.S. Department of Agriculture, using monthly and annual statistics to generate daily time series of precipitation, minimum temperature, maximum temperature, and solar radiation. • WeaGETSb (Weather Generator for Time Series) is a MATLAB-based versatile SWG for producing daily precipitation and maximum and minimum temperatures series of unlimited length, permitting impact studies of rare occurrences of meteorological variables. It implements various models such as the Richardson (1981), the modified Richardson, and the Bartlett- Lewis rectangular pulse. It can also be used in climate change studies as a downscaling tool by perturbing parameters to account for expected changes in precipitation and temperature. • WeatherGenRc is an SWG originally developed by Steinschneider and Brown (2013) and maintained by Deltares. It is a package for deploying a rapid weather generator for semiparametric gridded multivariate, multisite, and multivariable weather data for use in climate risk assessments and stress tests. It contains functions for generating synthetic daily time series of precipitation and air temperature under future climate change scenarios. • LARS-WGd version 8.0 is a computationally inexpensive SWG and downscaling tool to generate local-scale climate scenarios based on global climate models for impact assessments of climate change. It has been well validated in diverse climates around the world and incorporates CMIP6 climate projections (LARS- WGe Manual). • CLIGENf is a weather generator developed by the Agricultural Research Service of the U.S. Department of Agriculture and widely used in hydrological modeling and watershed management studies. It generates daily estimates (box continues next page ) 18 Enhancing the Safety and Resilience of Dams in the Context of Climate Change BOX 4.1 Stochastic Weather Generators (Continued) of precipitation, temperature, dew point, wind, and solar radiation for a single geographic point using monthly parameters (means, standard deviations, skewness, and so on) derived from the historic measurements. a. https://support.goldsim.com/hc/en-us/articles/115012797188-WGEN-Weather-Simulator. b. https://www.mathworks.com/matlabcentral/fileexchange/29136-stochastic-weather-generator​ -weagets. c. https://github.com/Deltares-research/weathergenr. d. https://sites.google.com/view/lars-wg/. e. Spatial resolution: 50 km; time resolution: hourly; forecasting update: every six hours, for eighty- four hours ahead. f. https://www.ars.usda.gov/midwest-area/west-lafayette-in/national-soil-erosion-research/docs​ /­wepp/cligen/#:~:text=Cligen%20is%20a%20stochastic%20weather,derived%20from%20the%20 historic%20measurements. Models for generating stochastic weather data are conventionally developed in two steps. The first is to model daily precipitation, and the second is to model the remaining variables of inter- est, such as daily maximum and minimum temperature, solar radiation, humidity, and wind speed, conditional on precipitation occurrence. Different model parameters are usually required for each month to reflect seasonal variations both in the values of the variables themselves and in their cross-correlations. The Intergovernmental Panel on Climate Change (IPCC) provides a useful summary of key features and characteristics of SWGs on its webpage,2 noting as a precaution that “because Weather Generator time series are usually site-independent and ignore the observed spatial correlation of climate, this can limit the value of some spatial impact assessments … . Although the Stochastic Weather Generator (SWG) may provide an accurate statistical representation of the observed situation at each individual site (i.e. the risk of drought and its local impact), taken together, the droughts are not simultaneous and the aggregate impact (e.g. on water resources or agriculture) is likely to be less severe than in the real situation, where widespread drought affects a large area.” Use of an SWG for impact assessment and stress tests: Weather generators are particularly useful when a climate change impact assessment requires (a) long time series of daily weather, which are not available from observations; (b) daily weather data in a data-sparse region; (c) gridded daily weather data for spatial analysis of climate risks; and (d) the ability to investigate changes in both the mean climate and its interdaily variability. As pointed out in IHA (2019), “when weather generators are used in climate stress tests to systematically test the climate sensi- tivity of impact models, systematic shifts are applied to produce new sequences of weather vari- ables (for instance precipitation) that exhibit a wide range of change in their characteristics (such as average amount, frequency, intensity, and duration). In the context of a climate stress test, a stochastic weather generator can be built for a region of interest and used to generate several Downscaling of GCM Output 19 ­ limate within which a water resources system can be tested. The flexibility of scenarios of daily c stochastic weather generators enables many climate permutations to be generated, each of which can exhibit a different type of climate alteration that the analyst may be interested in. The permu- tations created by the weather generator are not dependent on any climate projections, thereby allowing a wide range of possible future climates to be generated while avoiding biases propagated from the projections. However, the particular permutations generated can be informed by avail- able projections to ensure that they more than encompass the range of GCM projections.” Generators are most commonly applied for specific sites, ignoring the observed spatial correla- tion of climate, which can limit the value of large-scale spatial impact assessments. However, methods have been developed to interpolate site-specific parameters of SWGs over space, facil- itating spatial analysis. It may thus be necessary to test a number of SWG models to assess their suitability for the subject impact assessment. Several analysis steps are required to parameterize and test an SWG: • Data collection: Observed daily climatological data for the variables and site(s) of interest should be collected and quality controlled to parameterize the SWG for the baseline period. In case it is important to model low-frequency, high-magnitude events, it will be desirable to obtain the longest possible observed time series. For spatial applications, consistency of the observational time period between sites would also be important. • Parameterization: The parameters of the model are estimated using methods documented for the selected SWG. If spatial analysis is also being undertaken, this will require parameter esti- mation at many sites and subsequent interpolation of the parameters to a grid or other spatial field. Some SWG programs include automatic procedures for parameter estimation. • Model testing: Time series of weather are generated and their statistics analyzed and com- pared with the observed data on which they were parameterized. The significance of any dis- crepancies between the SWG-derived and observed series can be assessed by running both series through an impact model. Automatic model testing procedures are built into some public domain SWG programs. • Climate scenarios: If the SWG is to be used to create weather time series representing a changed climate, procedures will be required for applying climate change information from GCMs as adjustments to the parameters of the SWG. Some SWG software also handles climate scenarios. CORDEX DYNAMICAL DOWNSCALING EXPERIMENTS The section on “The Importance of Large GCM Ensemble Experiments” introduced several large-ensemble projection experiments using GCMs, some of which were accompanied by dynamical downscaling by RCMs. This section introduces the well-known Coordinated Regional Climate Downscaling Experiment (CORDEX; Giorgi and Gutowski 2015; Giorgi, Jones, and Asrar 2009; Gutowski et al. 2016; Pichelli et al. 2021). Because GCMs typically have horizontal resolutions of about 100 to 200 km, they cannot very well capture the effects of local forc- ings, such as complex topography and land surface characteristics. The coarse resolutions also preclude GCMs from providing an accurate description of extreme events. Aiming to develop a framework to evaluate and improve regional climate downscaling techniques and foster an 20 Enhancing the Safety and Resilience of Dams in the Context of Climate Change international coordinated effort for multimodel high resolution downscaling studies to contrib- ute to IPCC AR5, the CORDEX program was formed in 2009 under the World Climate Research Programme (WCRP3). It covers several subcontinental scale domains of the world and provides a framework for scientists to cooperate within each domain to provide an ensemble of downscaled simulations to assess the uncertainty of climate variability. CORDEX generally features a horizontal resolution of 25 km. CORDEX-South East Asia (CORDEX-SEA), for example, provides a comprehensive climate study for that region (Tangang et al. 2020), with the downscaling of eleven GCMs with seven RCMs and 25-km horizontal resolution for the past (1976–2005) and for the early (2011–40), mid- (2041–70), and late (2071–2100) twenty-first century for RCP4.5 and 8.5. The European CORDEX (EURO-CORDEX) initiative (Jacob et al. 2020) is a large voluntary effort that seeks to advance regional climate and Earth system science in Europe. As part of CORDEX, it shares the broader goals of providing a model evaluation and climate projection framework and improving communication with both the general circulation model (GCM) and climate data user communities. EURO-CORDEX oversees the design and coordination of ongoing ensembles of regional climate projections of unprecedented size and resolution (0.11° or 12 km and 0.44° or 50 km), covering all of Europe. Aalbers et al. (2018) studied a large single-model RCM-GCM ensemble over Western Europe with sixteen members for 1950–2100 at 0.11° (12 km) resolution (KNMI-RACMO2) and driven by EC-EARTH 2.3, a GCM based on the operational seasonal forecast system of the European Centre for Medium-Range Weather Forecasts. They found that winter extreme precipitation increases linearly with global warming and summer precipitation changes nonlinearly. However, the number of ensembles was not enough to provide robust results, and further simulations are needed to confirm. Sieck et al. (2021) provided a larger ensemble (125) of ten-year simulations driven by two GCMs for moderate global warming conditions in the EURO-CORDEX domain with 0.44° (50 km) horizontal resolution. They concluded that their large-ensemble projections can provide robust estimates of changes in extreme temperature and precipitation indices. Ehmele et al. (2020) provided a long-term, large-ensemble, regional-scale data set for central Europe. To study heavy precipitation, they conducted simulations for 12,500 years, from the past to the near future, using the COSMO-CLM RCM with a horizontal resolution of 0.22° (25 km) driven by MPI-ESM. They concluded that they can capture improved and more robust statistics of precipitation and a better estimation of extreme values. Ban et al. (2021) concluded that simulations at a 3-km scale are superior to the coarse resolution of RCM simulations in representing precipitation in the present-day climate and thus offer a promising way forward for investigations of climate change impacts at local to regional scales. Finally, Sørland (2021) reviewed the contributions of the Climate Limited-area Modeling Community (CLM-Community) to the CORDEX experiment through an extensive set of regional climate simulations. Using several versions of the regional COSMO-CLM-Community model, ERA-Interim reanalysis data sets and eight GCMs from CMIP5 were dynamically downscaled with a horizontal grid spacing of 0.44° (~50 km), 0.22° (~ 25 km), and 0.11° (~ 12 km) over the CORDEX domains of Europe, South Asia, East Asia, Australasia, and Africa. This major effort Downscaling of GCM Output 21 resulted in eighty regional climate simulations publicly available through the Earth System Grid Federation web portals for use in impact studies and climate scenario assessments. It was found that increasing horizontal model resolution (from 50- to 25- or 12-km grid spacing) alone does not always improve the performance of the simulations. Moreover, the COSMO-CLM perfor- mance depended on the driving data, which is generally more important than the dependence on horizontal resolution, model version, and configuration. NOTES 1. The troposphere is the lowest layer of the Earth’s atmosphere, extending up to an average altitude of about 12 km at the equator. It is the space where most of the Earth’s weather occurs because it contains the majority of the Earth’s water vapor and is subject to the greatest temperature variations. 2. https://www.ipcc-data.org/guidelines/pages/weather_generators.html. 3. https://www.wcrp-climate.org/. CHAPTER 5 Methods of Bias Correction Systemic biases: Climate models exhibit systemic biases (errors) because of limited spatial ­ resolution, simplifications of physics and thermodynamic processes, numerical schemes, or incomplete knowledge of climate system processes. The biases in GCM simulations relative to historical observations can be large, and bias correction is thus particularly essential when future climate projections are used to drive hydrological models in the context of climate risk assess- ments. Bias correction is a statistical post-processing technique used to reduce the mismatch between the statistics of climate model output and observations. The technique estimates the bias or relative error between a chosen simulated statistical property (for example, specific quan- tiles of the climatological distribution) and that observed over the calibration period. The simu- lated statistic is adjusted, taking into account the simulated deviation for the calibration period (Doblas-Reyes et al. 2021). Bias adjustment methods are often applied on a spatial scale similar to that of the simulation being adjusted, but they are also used as a simple downscaling method by calibrating them between coarse resolution model output and finer resolution observations. Typical implementations of bias adjustment are (a) additive adjustments, where the model data is adjusted by adding a constant, (b) rescaling, where the model data is adjusted by a factor, and (c) more flexible QM approaches that adjust different ranges of a distribution individually. The effectiveness of bias correction is assessed by comparing adjusted GCM output for the refer- ence period to observed data. Common evaluation metrics include the mean bias, the root mean square error, the correlation coefficient, and various statistical metrics for the fit of statistical distributions. Delta change method: A simple bias correction method is the delta change approach, in which selected observations are modified according to corresponding changes derived from dynam- ical model simulations. It is a commonly used statistical approach in climate change analysis to estimate how climate variables, such as temperature and precipitation, may change in the future. The method involves calculating the difference (delta) between the future projections of a climate variable and the corresponding simulations (hindcasting) for the historical refer- ence period and then applying this delta to historical observations for estimating future climate conditions. This assumes that the same systemic modeling errors affect the simulations for the reference period and the simulations for the considered future period, particularly when climate changes are expressed as relative (percentage) changes in variables, for example, in precipita- tion. The delta change method is widely used in impact assessments and adaptation planning because it provides a simple and transparent way to estimate the magnitude of future climate change impacts on different sectors. It must be noted that the delta change method does not account for climate change–induced nonlinear changes in the variability or in extreme events of the climate variable being analyzed. 22 Methods of Bias Correction 23 It assumes that the relationship between the climate variable and the future climate projec- tions is linear, which may not be the case in all situations. Therefore, it is important to consider the limitations and uncertainties of the delta change method when using it for climate change analysis. Bias correction methods using variable values: Bias correction considers three types of sta- tistical populations: past GCM output (reference period), future GCM output, and observation (reference) data. A conversion function is based on the statistical relationship between two sta- tistical populations (that is, past GCM output and observations) and applied to bias correct the future GCM population, as follows: • Adjusting mean values: The simplest bias correction method, the difference in the means of the past GCM and observation populations is applied to the future GCM population, either as an absolute difference or as a relative difference (ratio). Absolute differences are generally used to bias correct temperatures, whereas the relative differences (ratios) are used to bias correct precipitation. • Adjusting probability distributions: This method calculates the means and variance of two populations of model outputs to find a correction rate for application to a probability distribu- tion suitable for the target data. • Quantile mapping:1 Bias correction through QM is a widely used method for bias correction, involving the comparison of the statistical distributions of a variable in the past GCM output and in the observation data. The method then adjusts the GCM output such that its distribution closely matches the observed distribution. The adjustment is typically done by mapping the CDF of the observation data onto the CDF of the past GCM output and then using it to adjust the future GCM data at each grid point. The mapping function relates quantiles from the past GCM simulations to quantiles from the observation data and can be established using various techniques, such as lin- ear regression, spline interpolation, or other statistical methods. The adjustments ensure that the future GCM output retains its spatial and temporal coherence while also reducing any systemic biases that may be present. QM adjusts the entire distribution of model outputs, not just the mean or variance, preserves the spatial and temporal patterns of the GCM simulations, and is computa- tionally efficient. However, the assumption that the relationship between observed and simulated data remains constant over time may not hold true under changing climate conditions. Overall, QM is a powerful tool for improving the reliability of climate change projections by aligning model outputs more closely with observed climatic conditions. QM can be applied to different time scales, from daily to monthly or seasonal averages, and can be used to adjust multiple variables simultaneously. The method is widely used in climate research to improve the accuracy of GCM simulations and is often applied as a preprocessing step before using GCM output for impact assessments as input for hydrological models or for other applications. QM involves the following steps: • Establishing empirical distributions: Empirical CDFs are constructed for both the observed data and the model-simulated data. This involves calculating the quantiles (percentiles) for each data set. • Mapping quantiles: For each quantile in the model-simulated data, the corresponding quan- tile in the observed data is identified. This creates a mapping between the model’s simulated values and the observed values based on their respective quantiles. 24 Enhancing the Safety and Resilience of Dams in the Context of Climate Change • Bias correction: Using the established mapping, the values in the model projections are adjusted. This can be done by direct mapping of simulated and observed quantiles or by creating a transfer function based on the relationship between the quantiles of the observed and simulated data sets. • Adjusting future projections: The bias-corrected projections are then generated by applying the direct mapping or transfer functions to the raw model outputs for future periods. Bias correction methods using occurrence rates: The following methods correct biases based on the occurrence rate of a variable. The corrections corresponding to the occurrence rate are determined by assessing the statistical characteristics of the past GCM, future GCM, and observations populations through one of the following methods: • Cumulative distribution function transform (CDF-t): The CDF-t is a statistical method involving the transformation of the GCM output data to match the observed data in terms of their cumulative distribution functions. It represents the probability of a certain value or range of values occurring, and the transformation involves shifting or scaling the past GCM data to match the observed CDF. Linear or nonlinear regression models are used to relate the quantiles of GCM data to the corresponding quantiles of observed data, which are subsequently applied to the future GCM data to obtain a bias-corrected data set that matches the observed CDF. The main difference is in how the two compare the distributions of the past GCM and observation data, with the CDF-t method involving the cumulative distribution functions of the past GCM and observation data and the QM method using the quantiles. • Equidistant CDF matching (EDCDFm) is a variant of the CDF-t method and involves divid- ing the observed data and the GCM output into equally spaced quantiles, such as deciles or per- centiles, and then adjusting the past GCM output to match the cumulative distribution function of the observation data. This is typically achieved by fitting a transfer function—such as a linear or nonlinear regression—in each quantile interval and then using it to adjust the future GCM data within each interval. The main advantage of this method is that it is relatively simple and easy to implement, and it preserves the shape and variability of the observed data. However, it has some limitations, such as the assumption of linearity within each quantile interval, which may not always hold for some climate variables, particularly for extreme events. • Quantile delta mapping (QDM) is a variant of the QM method that is specifically designed to correct biases in the tails of the distribution of climate variables, such as extreme events. QDM extends the concept of QM by not directly adjusting the quantiles of the model output but rather by focusing on the differences (deltas) between the quantiles of the past GCM data and the corresponding observation data for the same reference time period. These delta values are then approximated with transfer functions, typically a linear or nonlinear regression model. The transfer function is applied separately to different quantiles of the data, such as the tenth, fiftieth, and ninetieth percentiles, to correct for any biases in the tails of the distribution. QDM provides insights into how the model’s quantiles are changing over time and allows for more nuanced adjustments that account for shifts in the data distribution. This can be particularly useful when dealing with nonstationary climate conditions or when assessing the impacts of climate change on extreme events. • Probability distribution parameter correction: The probability distribution parameter correction method involves adjusting the probability distribution function (PDF) of the past GCM output to match that of the observed data by correcting the parameters of the distribu- tion. The method assumes that the PDF of the GCM output can be modeled by a parametric distribution, such as the Gaussian or gamma distribution. The parameters of the distribution, Methods of Bias Correction 25 such  as  the mean and standard deviation, are estimated from the past GCM output for the reference period and adjusted to match the corresponding parameters of the observation data using a transfer function, such as a linear or nonlinear regression. The adjusted parameters are then used to obtain the corrected future GCM output. Advantages of this method are its ability to preserve the shape of the GCM output and its flexibility to accommodate different types of distributions. It also has limitations, such as the assumption of a parametric distribution, which may not always accurately represent the underlying distribution of the data. Open software tools: Various open source software packages for bias correction are freely available, even though basic skills in terms of programming and climate sciences may be required for their use: • The Robust Multivariate Bias Correction (RoMBC)2 is a package designed to deal with the distri- bution and dependence biases in multivariate time series. It starts with a simple, single timescale univariate bias correction and—depending upon the requirement of the data at hand—grows into a comprehensive multivariate multi-timescale bias correction. RoMBC is a comprehensive package that offers a wide variety of in-built options by including variants of standard quantile matching and other routinely used bias correction approaches in a time- and cross-dependence nesting (Mehrotra and Sharma 2021). • The Climate Data Operators (CDO)3 software is a collection of many operators for the standard processing of climate and forecast model data; it provides functionalities for bias correction among other data processing tasks. • The Climate Data Toolbox for MATLAB (CDT)4 is a toolbox that offers various functions for climate data analysis, including bias correction methods. It contains a standard set of MATLAB functions for analyzing and displaying climate data. The functions are computationally efficient and easy to use, and they come with many tutorials that not only describe how to use CDT functions but also offer guidance on how to interpret the results in the context of Earth science processes. • Climate Data Analysis Tools (CDAT)5 is a Python package that provides functionalities for work- ing with climate data, including bias correction techniques. NOTES 1. NASA’s Earth Exchange Global Daily Downscaled Projections data set provides a readily available set of global, high-resolution (0.25 degrees x 0.25 degrees), bias-corrected climate change projections for the period 1950–2100 that can be used to evaluate climate change impacts on processes that are sensitive to finer-scale climate gradients and the effects of local topography on climate conditions. The bias correction was based on a Princeton reanalyzed precipitation data set. The NEX-GDDP-CMIP6 data set is available at www.nccs.nasa.gov/services/data-collections/land-based-products/nex-gddp-cmip6. 2. https://www.unsw.edu.au/research/hydrology-group/our-resources/robust-multivariate ​ - bias​ -correction-rombc. 3. https://code.mpimet.mpg.de/projects/cdo/embedded/index.html. 4. https://www.mathworks.com/matlabcentral/fileexchange/70338-climate-data-toolbox-for-matlab. 5. https://cdat.llnl.gov/pdf/UV-CDAT_February_2013_Presentation_CDAT.pdf. CHAPTER 6 Design Flow Estimation OVERVIEW Determining adequate design flows—commonly referred to as design floods and safety check floods—for water storage and hydroenergy-generating infrastructure is critically important for the safety and proper performance of the structures. A design flow is commonly based on a storm or flood event that will have a recurrence frequency or return period selected depend- ing on the type and value of the structure, downstream conditions, and local regulations. The magnitude of the selected design flood can vary significantly, from a 100- or 200-year flood for low-consequence dams to a 1,000-year flood and typically a 10,000-year flood or the PMF for high-consequence dams. These are extremely large floods, and their estimates come with high uncertainty, whether statistical or because of climate change. It may be noted that the 1,000-year return flood still has probability of occurrence of 9.5 percent within the lifetime of a 100-year old person, which reduces to 1 percent for the 10,000-year flood. Figure 6.1 provides an overview of the hydrological modeling process commonly used for esti- mating the design flood or PMF. There are two principal methods (that is, a hydrological method and a statistical method) practiced in the field of hydrology for assessing design floods and PMF estimates for a particular catchment (Okoli 2019). This section discusses hydrological simulation models and statistical flood frequency analysis. HYDROLOGICAL SIMULATION MODELS There are two modes of hydrological simulations: • Event-based model simulation requires the estimation of the PMP or the selection of a design rainfall from an intensity-duration-frequency curve of rainfall with a given duration and assumed profile and uses it as input for the rainfall-runoff model to derive the flood hydrograph (Rogger et al. 2012). Event-based approaches are limited to design flow estimations. These approaches do not account for the role of antecedent moisture content or water storages in runoff generation. • Continuous model simulation utilizes the long-term rainfall records from observations and/or climate models as input for rainfall-runoff models or couples a stochastic weather model with a rainfall-runoff model. This method can address the influence of threshold effects (for example, catchment storage, such as soil water content and interception) on flood frequency curves, and it can be implemented using only physically based hydrological models (Rogger et al. 2012). Hydrological models can also be designed to simulate the impacts of overland flooding on flood wave attenuation, which can be essential, particularly for extreme floods and PMFs. 26 Design Flow Estimation 27 FIGURE 6.1 Overview of the Hydrological Modeling Process for the Assessment of Design Floods Hydrological simulation Statistical method Rainfall records from Hydrological method Observed observed or climate discharged models or probable records maximum precipitation Rainfall-runoff models Snow hydrology Continuous model Event-based model simulation: simulation: Water and energy Rational method, budget-based conceptual models, Inundation processes models water budget models Flood frequency analysis Simulated flow at dam sites Design flood or probable Max flood Safety assessment Source: Original figure for this publication. Classification of hydrological models: Hydrological models are classified based on model input, model parameters, and the extent of physical principles represented or formulated within the model. Event-based simulations generally employ the simplest forms of ­ models— hydrological ­ for example, the rational method, a unit hydrograph method, or a conceptual ­ rainfall-runoff model (tank model)—to estimate peak discharges. The rational method provides only the peak discharge and cannot produce a hydrograph. However, dam safety (for example, against over- topping) depends on both the peak flow and the total volume of water arriving at the dam during the flood. Accordingly, an exact estimate of the design flood and extreme flood hydro- graph is required for the proper design of dam safety measures, such as spillway, flood control capacity, and dam and reservoir operation rules. To estimate the peak flow volume over a given period, a representative flood hydrograph derived from a unit hydrograph or other conceptual or physically based methods have to be used. The applicability of the unit hydrograph approach and conceptual model approaches has several limitations, particularly with respect to the system response to climate change, which affects the accuracy of the peak flow estimations. Physically based rainfall-runoff models are preferred over other types of models (that is, empirical and conceptual models) because they include the principles of physical processes that are domi- nant in the flow generation across a watershed and can generate both peak flow discharges and flood hydrographs, underpinning hydrologic hazard curves for dam safety assessments. Physical models can overcome many shortcomings of the empirical and conceptual models because of the use of parameters with a physical interpretation. They can be applied to a wide range of situations, including the extrapolation for extreme precipitation and flow conditions. Moreover, the continuous model simulation can incorporate the water cycle variability and climate change 28 Enhancing the Safety and Resilience of Dams in the Context of Climate Change information over time. The proper selection of a physically based model representing the physical processes and hydrological characteristics of a watershed (for example, water and energy-related processes) and the selection of the associated model input parameters are crucial in the design of reservoirs, flood estimation, and dam operation under conditions of future changing climate. Lumped and distributed hydrologic models: Ray and Brown (2015) present an overview of generally trusted and commonly used lumped and distributed physically based hydrologic mod- els, inter alia the Precipitation Runoff Modeling System (PRMS1), the Sacramento Soil Moisture Accounting (SAC-SMA2) model, the Community Land Model (CLM3), the Variable Infiltration Capacity Model (VIC4), the Soil and Water Assessment Tool (SWAT5), the Water Evaluation and Planning system (WEAP6) with its hydrologic modeling component WATBAL), the Hydrologic Modeling System (HEC-HMS7), Topography-based Hydrological Model (TOPMODEL8), SOBEK Rainfall-Runoff (SOBEK RR9) (library of rainfall-runoff modules), WFLOW10, and WEB-DHM11 (Water and Energy Budget-based Distributed hydrological model). Some of these are physically based and distributed, approximately representing actual natural processes, whereas others are conceptual and keep track of the water balance without reproducing complicated relationships with energy and soil. Hydrological models do not always produce reliable estimates of the impacts of climate change on runoff (for example, Grijsen 2014; Sankarasubramanian et al. 2001; Vano, Das, and Lettenmaier 2012). Hagemann et al. (2013) and Schewe et al. (2014) showed that the choice of the hydro- logic model can result in a larger spread in simulated future streamflow than the spread origi- nating from GCMs, and van Vliet et al. (2016) also demonstrated that the variability between adopted hydrologic models explained a considerable part of the uncertainties in projected future runoff and water temperatures. Counterintuitively, it appears that hydrologic models may not always be necessarily well suited to determine the impacts of climate change on basin runoff, particularly concerning the impacts of temperature changes. Hence, great care must be taken in the selection of a suitable hydrological model. HYDROLOGICAL SIMULATIONS FOR SPECIAL CONDITIONS Model-related uncertainties: For certain regions, hydrological simulations carry an addi- tional layer of complexity. For example, in cold, high-elevation regions, the surface water and energy balance are controlled by seasonal snow cover, air temperature, surface moisture con- tent, incoming shortwave and longwave radiation, and vegetation, whereas in arid and semiarid regions, evapotranspiration and soil moisture dynamics control the partitioning of surface water into surface runoff, evapotranspiration, and infiltration processes. Moreover, model-related uncertainty may arise because of the negligence of several fundamen- tal and critical water and energy-related parameters and physical processes (for example, snow- melt, soil moisture, canopy interception, evapotranspiration, and soil-vegetation-atmosphere interactions), which cannot be overlooked, particularly in arid and semiarid regions as well as for high elevation and cold regions (Rasmy, Sayama, and Koike 2019; Shrestha et al. 2014). The negligence of essential physical formulations for water and energy budget components and land-vegetation-atmosphere interactions within rainfall-runoff models will result in sev- eral critical issues in applying these types of models for design flow estimation and operational Design Flow Estimation 29 flood-related applications under changing climate. For example, the snowmelt and amount of moisture within the soil layers control the surface runoff and infiltration ratio, whereas the accurate estimation of the evaporation component of the water budget is crucial for simulating accurate soil-vegetation water storages, river discharges, flood peaks, and flood depths. Accordingly, the uncertainties in snowmelt as well as the initial soil and vegetation water stor- ages will impact the reliability of simulated flood-related parameters and associated flood-related risk assessments. In operational flood monitoring and forecasting activities, unreliable initial storage estimates will introduce larger uncertainties in the timing of the simulated flood onset, peak discharges and inundation depths; therefore, its output needs to be handled with great caution. Moreover, the severity of flood and drought conditions are expected to increase in the future, and the extreme climate sequences of dry and wet periods are also expected to be mod- ified (Madakumbura et al. 2019) because of global warming and an increasing rate of evapo- transpiration. As a result, the application of these models to the projected future climate will introduce additional uncertainties in flood-related simulation and risk assessment. Therefore, it is crucial for the selected rainfall-runoff model to properly represent the dominant processes of the water and energy budgets, such as snowfall, snowmelt, evapotranspiration, and soil mois- ture dynamics, to correctly estimate the inflow to reservoirs for assessing the future flood dis- charge and reservoir inflow patterns based on projected rainfall information. Snowmelt processes in hydrological modeling: Precipitation contributes the greatest uncer- tainty among all forcings of distributed hydrological models, and this is even greater for snowfall than for rainfall. Multiple factors play a role, such as wind, topography, blowing and drifting, wetting, and evaporation losses at point scale, as well as the wind distribution and orographic dependencies at basin scale. Several correction methods have been developed to overcome sys- tematic errors in snowfall measurements at point scale. Some studies were designed to avoid both systematic and nonsystematic (site-specific) biases associated with snow gauges, using observed snow depth and a physics-based land surface model to solve an inverse problem for snowfall. It reconstructs snowfall by calculating what must have occurred to produce the observed snow depth, given the physics of the model. However, all these methods deal with snowfall measure- ment errors at the point scale. The areal distribution of snowfall is still a major problem when extending modeling methods based on a point scale to distributed snowmelt modeling because of an insufficient gauge density across watersheds and the applied methods of interpolating point data. Shrestha et al. (2014) developed a new approach to the correction of snowfall at the basin scale. This approach uses a comprehensive multilayer, energy balance-based snowmelt model, with both water and energy balance closure in a snow-soil-vegetation-atmosphere trans- fer-based distributed biosphere hydrologic modeling framework (WEB-DHM-S). Inundation and flooding processes in hydrological modeling: The inclusion of fluvial and pluvial flood conditions is critical in estimating the inflow to dams or reservoirs because large-scale flooding will attenuate peak flows and delay the arrival of flood waves at dam sites. Therefore, models must be selected that incorporate inundation processes in the catchment, can be used for the mapping of flood hazards, and are able to provide crucial flood-related information for flood risk management and flood damage assessment. This type of model is capable of providing additional information (for example, inundation extents, flooding depths, and, most importantly, the direction of flow paths as the flood water travels) compared with typical (that is, kinematic wave) distributed hydrological models in which water flows along with the defined flow directions 30 Enhancing the Safety and Resilience of Dams in the Context of Climate Change (Bates, Horritt, and Fewtrell 2010; Hunter et al. 2007; Sayama et al. 2012; Wilson et al. 2007). Recently, Rasmy, Sayama, and Koike (2019) developed the Water and Energy Budget-based Rainfall-Runoff-Inundation (WEB-RRI) model and validated it using discharge data, satellite-based terrestrial evapotranspiration, and ground data, as well as satellite-based inundation maps. These types of models have great potential to be applied for both basinwide integrated water resources management and water-related disaster risk reduction, including flood- and drought-related risk assessments, by employing the model to operational (for example, flood forecasting and seasonal flow prediction) and long-term applications (for example, the catchment responses to past and future climatology, water cycle variability, hydrological extremes, and land-use change). STATISTICAL FLOOD FREQUENCY ANALYSIS Flood frequency analysis provides information about the magnitude and frequency of flood flows based on the records of annual maximum peak flows collected at gauging stations close to or at a project site, or simulated with rainfall-runoff models for long observed or reanalyzed precipitation data series or for future hydroclimatic conditions as projected by GCMs. For the baseline conditions, the direct analysis of observed flood peaks is preferred, provided enough data—preferably at least peak flows for thirty years—are available for analysis. Observed peak flows reflect the output of the entire hydrological cycle, whereas rainfall data series constitute the input to the hydrological cycle, which then needs to be modeled, introducing uncertainties and errors in the simulation results. In the absence of river flow data near a dam location, flow data for gauges in nearby catchments or upstream/downstream of dams can be used to infer flood peak data for the dam sites based on correlations or regional flood analyses. Flood frequency analysis consists of fitting selected probability distributions (such as the Gumbel or log-normal distributions) to observed or simulated flood frequencies and using these distributions to estimate extreme flood events with high return periods. Statistical tests and visual inspection are used to select the distribution of choice. Although the statistical methods of design flood esti- mation are straightforward and computationally inexpensive, they carry several issues because of (often) limited data availability and the inadequacy of the hydrological modeling of flood extremes (Klemeš 2000). The main challenges of this method can be summarized as follows: • Historic records of reliable observations are relatively short (often at best thirty to fifty years) and data on extremes flows are usually limited, resulting in large uncertainties when estimating the extremes extrapolated beyond the range of observations. In case the historical hydrome- teorological data sets are insufficient, one can use the CHIRPS, ERA5, or CRU TS4.0612 rean- alyzed precipitation data sets, with long records, to extend the available maximum flow data series through correlation analysis to enhance the flood frequency analysis. Moreover, statistical uncertainty can be expressed through the 95 percent confidence interval, which should be taken into account when selecting design discharges. • The assumption of stationary data is no longer valid because climate and catchment condi- tions are rapidly changing from natural and human activities, causing more frequent and larger extremes accelerated by climate change, urbanization, and deforestation. This results in an increase of both the mean and interannual variability of annual maximum flows, which can be incorporated in the parameters of the statistical distributions used for flood frequency analysis Design Flow Estimation 31 (approach 1 in figure 3.1 of the main Technical Note). Similarly, the uncertainties of climate change–induced changes in the mean and interannual variability of annual maximum flows can be integrated in the estimation of the 95 percent confidence interval. Such extreme flow estimates for a return period of 100,000 years and the 97.5 percent confidence level may be considered as an alternative for the PMF estimate and can possibly be used as an estimate of the safety check flood. The changes in the mean and interannual variability of annual maximum flows, and the inherent uncertainties in their estimates, can also be estimated from extreme precipitation projections from a large ensemble of GCMs. • Floods can be driven by both rainfall and snowfall (Tarasova et al. 2019), and peak discharge estimations are affected by various hydrological processes (for example, land-atmosphere water-energy interaction, soil moisture dynamics, evapotranspiration, vegetation, and inunda- tion). Plots of extreme flood data on a probability scale often show a convex shape, as shown, for example, in figure 6.2, caused for example by enhanced flood attenuation because of over- land flooding during extreme floods. FIGURE 6.2 Extreme Flood Estimation Using Different Flood Frequency Assessment Methods Inflow into upper campbell lake (m /s) 5000 PMF (24-Hour peak inflow) 4500 4000 3500 3000 2500 2000 1500 1000 500 0 0 10 100 1,000 100,00 100,000 Return period (years) Observed 24-hour peak inflows (1963-2010) GEV (3 parameter) frequency distribution Gumbel frequency distribution Log-Pearson type 3 frequency distribution Gamma frequency distribution Weibull frequency distribution Compound Poisson/Exponential frequency distribution SCHADEX (24-hour peak inflow) Log-Normal (3 parameter) frequency distribution SEFM (24-hour peak inflow) Source: Adapted from ICOLD 2018. Note: The bold dashed red lines (representing the upper and lower limits of the 95 percent confidence interval) are superimposed on the original graph. The methods shown in the graph include flood frequency analysis, SEFM, the SCHADEX hybrid modeling procedure and the PMF. SEFM = stochastic event flood model; PMF = probable maximum flood. 32 Enhancing the Safety and Resilience of Dams in the Context of Climate Change Figure 6.2 also illustrates the differences in results arising from different classes of methods applied to a catchment in Canada (ICOLD 2018). Flood estimates based on the flood frequency analysis of historic maxima yielded results that are well below the estimates obtained using two rainfall-based techniques—that is, the stochastic event flood model (SEFM; MGS Engineering Consultants 2009) and the SCHADEX probabilistic method for extreme flood estimation (Paquet et al. 2013). The estimated PMF is also shown, exceeding the results of the flood frequency anal- yses by about 50 percent. The estimates were constrained by the short length of observed flood maxima (forty-seven years) and the need to left-censor (discard) the lowest 50 percent of the observed data, whereas the rainfall-based methods were able to take advantage of rainfall data obtained over a larger region and time frame. It is seen that flood frequency analyses may, as such, underestimate exceedingly rare flood events, but this can be compensated to a great extent by considering the statistical uncertainty of flood estimates (95 percent confidence interval), particularly when also incorporating the uncertainty and changes induced by climate change impacts on the means and variability of extreme flows. Therefore, flood frequency analysis remains a powerful and low-cost tool, which may always be applied for the preliminary design of each dam project and can be followed up with more sophisticated analyses—as shown in figure 6.2—as the need arises. ENVELOPE CURVE METHODS FOR EXTREME FLOOD ESTIMATION Because of the uncertainties involved in the estimation of extreme floods, such as the PMF, it is useful to check the design and safety check floods (m3/s) by using an envelope curve approach. A  frequently used method is the Francou-Rodier approach (Francou and Rodier 1967). Its essence is that the maximum observed floods can provide some guidance as to what reasonable values of design floods might be. The method entails plotting estimated extreme flows (m3/s) derived from flood frequency analyses for high return periods against the catchment area (km2) on a log-log scale graph, as shown in figure 6.3. Francou-Rodier equation (6.1) describes the relationship between peak flow and catchment area (ICOLD 2019): (6.1) where: Q = peak flow (m3/s); Q0 = 106 m3/s A = catchment area (km2); A0 = 108 km2 K = Francou-Rodier coefficient The Francou-Rodier curve is thus a graphical representation used to explore the relationship between flood magnitudes and the size of river basins. Typically, the curve reveals a power-law relationship, where the flood magnitude increases with the size of the basin. It provides insights into the scaling behavior of flood events across different basin sizes and helps in understanding Design Flow Estimation 33 how changes in the size of a drainage basin influence the potential severity of floods. The curve facilitates comparisons among different river basins and provides valuable data for the develop- ment of flood frequency analyses and hydrological models. When this analysis was conducted by Francou and Rodier for all known maximum flood peak flows in the world, it was found that the curves for all regions with homogeneous flood ­ conditions tended to align through the upper-right corner of the graph, characterized by Q0 = 106 m3/s and A0 = 108 km2. The six estimated maximum flow values plotted in figure 6.3 for a region with homogeneous flood flow conditions in Central Africa, for basin areas vary- ing between 2,000 and 125,000 km2, match a K-value of 3.65. The data points represent the Gumbel flood estimates for a return period of 100,000 years, whereby climate change–induced projected changes by 2050 in the means and variabilities of annual extreme flows are incorpo- rated in the analysis (as elaborated in the section on “Statistical Flood Frequency Analysis”), and the uncertainties embedded in the climate change projections are included in the 95 per- cent confidence intervals. The plotted data points represent the upper boundary of the 95 per- cent confidence interval. For one of the subject river basins, it was found that the accordingly determined extreme flow matched the earlier determined PMF. The Francou-Rodier curve helps ensure that consistent extreme flows are obtained for regions with homogeneous hydro- logical conditions. FIGURE 6.3 Example of a Francou-Rodier Graph for River Basins in Central Africa Maximum flood discharge (m3/s) 1,000,000 100,000 10,000 1,000 5 K=6 3.6 K= 100 K=5 10 K=4 K=3 K=2 K=1 K=0 1 1 10 100 00 00 0 0 0 0 00 0 00 0 1,0 ,0 ,0 ,0 0, 0, 10 00 00 10 00 1,0 ,0 0, 10 10 River basin area (km2) Source: Original for this publication. Note: T = 100,000 years + climate change impacts by 2050 at 97.5 percent. 34 Enhancing the Safety and Resilience of Dams in the Context of Climate Change NOTES 1. https://www.usgs.gov/software/precipitation-runoff-modeling-system-prms. 2. http://www.appsolutelydigital.com/ModelPrimer/chapter5_section10.html. 3. https://www.cesm.ucar.edu/models/clm. 4. https://github.com/UW-Hydro/VIC. 5. https://swat.tamu.edu/. 6. https://www.weap21.org/. 7. https://www.hec.usace.army.mil/software/hec-hms/downloads.aspx. 8. https://csdms.colorado.edu/wiki/Model:TOPMODEL. 9. https://www.deltares.nl/en/software-and-data/products/sobek-suite/modules/rr-rainfall-runoff-open​ -water. 10. https://www.deltares.nl/en/software-and-data/products/wflow-catchment-hydrology. 11. https://ui.adsabs.harvard.edu/abs/2008AGUFM.H32D..08W/abstract. 12. CHIRPS = Climate Hazards Group InfraRed Precipitation with Station data; ERA5 = fifth generation atmospheric reanalysis of the global climate (1940 to present), produced by the Copernicus Climate Change Service at the European Centre for Medium-Range Weather Forecasts; CRU TS4.06 = Climate Research Unit Time Series, University of East Anglia, United Kingdom. When using these reanalyzed data sets, it is necessary to check their skills for the area it is intended to be used for because one data set may perform better than another for a specific region. PART II CASE STUDIES ON IMPACT ASSESSMENT AND ADAPTATIVE RESERVOIR OPERATION CHAPTER 7 Case Studies on the Assessment of Climate Change Impacts on Floods AUSTRALIA—IMPACT OF CLIMATE CHANGE ON OPERATIONAL PMP ESTIMATES Visser et al. (2022) assessed the impact of climate change on PMP estimates for Australia. A key shortcoming of traditional PMP estimates is the assumption of a stationary climate. Emerging evidence indicates that key meteorological conditions relevant to the magnitude of extreme storms, such as atmospheric moisture, are changing in a warming climate. Because of the prag- matic nature of PMP methods derived for design purposes, inferring potential changes in PMP estimates based solely on trends or projections of atmospheric variables can ignore complexities and constraints. The study team explored how different traditional PMP methods will respond to a potential increase in atmospheric moisture and found that increases in the persisting dew point temperature will lead to increases in PMP estimates, the magnitude of which depends on whether the moisture maximization step is based on local or on transposed regional information. A historical trend analysis reveals annual maximum persisting dew point temperatures have increased continuously over Australia over the past sixty years, with further increases predicted over the coming decades for all SSPs. Based on the optimistic SSP1-2.6 scenario, PMP estimates across Australia are predicted to increase by an average of 13 percent by 2100, compared with 33 percent for the pessimistic SSP5-8.5 scenario. The team concluded that PMP methods will require regular updating to account for changing persisting dewpoints and likely progressive increases in PMP and the ensuing flood estimates. The study showed significant increases in the maximum persisting 24-hour 1,000-hPa dew point (map 7.1, panels a and b) and the moisture adjustment factor (MAF; map 7.1, panels c and d) over the historical simulation period (1950–2014) and for the future scenarios SSP1-2.6 and SSP5-8.5 up to 2100, as simulated by the ACCESS-CM2 model. Map 7.2 shows the location of large dams across Australia based on the Australian National Committee on Large Dams (ANCOLD) register of large dams. Figure 7.1 shows the density plot of predicted increases in PMP estimates for Australia (no outline) and for large dam locations (black outline), based on the ACCESS-CM2 model simulations for the SSP1-2.6 (green shades) and SSP5-8.5 (purple shades) scenarios by 2100. Dashed lines represent the average values for all of Australia (white dashed line) and for the large dam locations only (black dashed line). Density lines are smoothed using a multiplicate bandwidth adjustment factor of 2. 36 Case Studies on the Assessment of Climate Change Impacts on Floods 37 MAP 7.1 Future Increases in the Maximum Dew Point Temperature and Moisture Adjustment Factor in Australia a. Increases by 2100 in the maximum b. Increases by 2100 in the maximum persisting 24-hr 1,000-hPa dew point persisting 24-hr 1,000-hPa dew point temperature over the historical temperature over the historical simulation period (1950–2014) simulation period (1950–2014) for future scenarios SSP1-2.6 for future scenarios SSP5-8.5 [°C] 5.0 4.5 Increase in 4.0 3.5 maximum 3.0 persisting 2.5 24-h 2.0 1000-hPa 1.5 dewpoint 1.0 0.5 >0 c. MAF over the historical d. MAF over the historical simulation period (1950–2014) simulation period (1950–2014) for future scenarios SSP1-2.6 for future scenarios SSP5-8.5 [%] 55 50 45 Percentage 40 increase in 35 30 MAF and 25 PMP 20 15 10 5 >0 Source: Visser et al. 2022. Note: MAF = moisture adjustment factor; SSP = shared socioeconomic pathway. PMP estimates focused on areas where large dams are located and PMPs are projected to increase by an average value of 16 percent by 2100 based on the optimistic SSP1-2.6 scenario, compared with 38 percent for the pessimistic SSP5-8.5 scenario, and compared with the afore- mentioned 13 percent and 33 percent, respectively, across the entirety of Australia. These pro- jections indicate serious implications for the hydrological safety of dams under climate change impacts in Australia, and the same range of impacts can be expected for many other regions in the world. 38 Enhancing the Safety and Resilience of Dams in the Context of Climate Change It must be noted that this analysis used only climate change projections for the ACCESS-CM2 model, and it is not known how well this GCM reflects the future climate conditions over Australia. Further research is required based on climate change projections of multiple GCMs (ensemble modeling). However, the results shown in map 7.2 and figure 7.1 clearly indicate that one can expect a significant increase in PMP values because of future climate change, even under the optimistic SSP-2.6 scenario. MAP 7.2 Location of Large Dams across Australia by ANCOLD 15°S 30°S Dam location Administrative boundary 135°E 45°S 120°E 150°E Source: Visser et al. 2022. Note: ANCOLD = Australian National Committee on Large Dams. Case Studies on the Assessment of Climate Change Impacts on Floods 39 FIGURE 7.1 Density Plot of Projected Increases in PMP Estimates, Australia Density 0.125 0.100 0.075 0.050 0.025 0 0 20 40 60 Increase PMP by 2100 (%) SSP1-2.6 SSP5-8.5 All Australia Dam locations Source: Visser et al. 2022. Note: Australia is shown with no outline, whereas large dam locations are shown with blue outline. SSP1-2.6 are represented by blue shades, and SSP5-8.5 are in orange shades, both up to 2100. Dashed lines represent average values for all of Australia, and dotted lines represent average values for dam locations. INDIA—THE PMP IN A CHANGING CLIMATE Sarkar and Maity (2020) assessed the increase in the PMP in a changing climate over India. To study its temporal change, they developed one-day PMP maps for India for five consecutive time periods—that is, for two historical periods (1901–70 and 1971–2010) and for three future periods (2010–39, 2040–69, and 2070–2100). Observed gridded daily precipitation data from the India Meteorological Department and results of precipitation projections from three differ- ent GCM-RCM combinations for two RCP scenarios were used to develop historical and future PMP maps, as shown in table 7.1. The modified Hershfield method was used for the estimation of PMPs with a modified enveloping technique. The results show a clear increasing trend in PMP across India as shown in tables 7.2 and 7.3. Specifically, 84 percent of the area of India’s mainland exhibits an increasing trend in PMP with an average of about 35 percent increase in the post-1970 (1971–2010) period compared with the pre-1970 (1901–70) period. This indicates a significant impact of the shift in the global climate regime in the 1970s on the PMP in India. Table 7.4 provides the average percentage increase in PMP estimates and the corresponding percentage of the area showing an increase with respect to the post-1970 period for three future time periods, according to three GCM-RCM model combinations and two RCP scenarios. It is also seen that the mean and particularly the variability of the annual maximum daily precipita- tion (AMDP) data series increase gradually over time, which is eventually causing a substantial increase in PMP throughout India. 40 Enhancing the Safety and Resilience of Dams in the Context of Climate Change TABLE 7.1 Details of Future Simulated Precipitation Data Used in This Study Model Used Data Abbreviation Driving GCM Source Institute scenarios RCM resolution M1 MPI-ESM-LR Max Planck Institute for RCP4.5 MPI-CSC- 0.5° x 0.5° Meteorology RCP8.5 REM02009 M2 CCCma- Swedish Meteorological RCP4.5 SMHI-RCA4 0.44° x 0.44° CanESM2 and Hydrological Institute RCP8.5 M3 MOHC- Swedish Meteorological RCP4.5 SMHI-RCA4 0.44° x 0.44° HadGEM2 and Hydrological Institute RCP8.5 Source: Sarkar and Maity 2020. Used with permission. Further permission required for reuse. TABLE 7.2 Percentage Increase in Mean and Standard Deviation of Annual Maximum Daily Precipitation, and PMP in Post-1970, Compared to Pre-1970 Description Average Maximum Minimum % increase in mean 8.78 359.37 -56.93 % increase in standard deviation 24.78 565.2 -58.15 % increase in PMP 35.40 318.11 -47.38 TABLE 7.3 Percentage Area Showing Increase in Mean and Standard Deviation of Annual Maximum Daily Precipitation, and PMP in Post-1970 as Compared to Pre-1970 Description Quantity % of area showing increase in mean 61.85 % of area showing increase in standard deviation 67.59 % of area showing increase in PMP 83.92 Similar to the historical period, the increasing trend persists also in the future as shown in table  7.4. In the far future period (2071–2100), about 70 percent to 80 percent of the area is showing an increase with an approximate average increase of 30 percent to 35 percent in PMP across different models with respect to the recent past (1971–2010) following the RCP8.5 scenario. These results clearly indicate the significant increases in the PMP because of climate change, demonstrating that it must be factored into the revised planning and design of water resources and hydropower engineering projects. VIET NAM—CLIMATE IMPACT ASSESSMENT ON LARGE FLOODS AND DAM SAFETY This case study is based on an assessment carried out as part of World Bank support to the Dam Rehabilitation and Safety Improvement Project in Viet Nam (Veale et al. 2020), which evaluated the potential climate change impacts on the hydrological safety of an approximately 50-m-high earthfill dam with its storage capacity of about 800 million m3 in a region that experiences fre- quent major flooding. The reservoir catchment area is approximately 800 km2. TABLE 7.4 Average Percentage Increase in PMP Estimates, Mean, and Standard Deviation of AMDP Series and Corresponding Percentage of Area Showing Increase with Respect to the Post-1970 Period for Three Future Time Periods, Following Three Models and Two RCP Scenarios Average percentage increase in PMP Percentage of area showing increase in PMP Near future Future Far future Near future Future Far future Scenarios (2011–40) (2041–70) (2071–2100) (2011–40) (2041–70) (2071–2100) M-1 M-2 M-3 M-1 M-2 M-3 M-1 M-2 M-3 M-1 M-2 M-3 M-1 M-2 M-3 M-1 M-2 M-3 RCP4.5 7.7 9.2 11.9 11.7 15.5 16.6 19.3 23.5 28.1 58.7 69.7 64.7 67.0 72.0 70.1 72.6 76.2 74.3 RCP8.5 8.4 11.7 15.9 23.7 23.6 22.6 29.1 34.4 33.9 62.9 63.8 69.7 76.3 78.9 72.8 81.0 86.0 81.0 Average percentage increase in mean Percentage of area showing increase in mean RCP4.5 7.6 5.7 6.3 10.0 10.6 14.6 17.7 15.9 26.3 54.7 57.6 63.2 61.9 67.6 77.9 77.1 74.9 90.2 RCP8.5 8.4 7.6 9.3 16.5 18.4 24.3 23.7 29.8 33.7 54.1 57.1 64.2 69.6 73.0 87.7 80.1 89.9 95.7 Average percentage increase in standard deviation Percentage of area showing increase in standard deviation RCP4.5 8.9 8.0 9.6 14.0 18.4 15.0 20.6 25.5 28.3 51.9 59.6 54.9 61.9 68.2 63.1 68.9 71.4 69.3 RCP8.5 16.7 11.1 12.7 25.1 24.4 26.3 32.4 32.7 36.7 60.4 58.9 60.5 66.2 73.1 70.3 73.5 77.5 73.9 Source: Sarkar and Maity 2020. Used with permission. Further permission required for reuse. Note: AMDP = annual maximum daily precipitation; RCP = Representative Concentration Pathway; PMP = probable maximum precipitation. Case Studies on the Assessment of Climate Change Impacts on Floods 41 42 Enhancing the Safety and Resilience of Dams in the Context of Climate Change Climate Change Projections The projection of future changes in temperature and rainfall for the catchment upstream of the dam involved the downscaling of outputs from multiple global circulation models (GCMs) by Thuyloi University in Hanoi. Climate change projections of daily rainfall and temperature for the concerned catchment were derived from a selection of eleven GCMs from the Coupled Model Intercomparison Project Phase 5. These were selected for their relative capability in simulating temperature and precipitation for Southeast Asia. Projections were used for middle and upper- bound greenhouse gas emission pathways (that is, RCP4.51 and RCP8.5), covering periods span- ning the middle part (2040–69) and the end of the twenty-first century (2070–99). The resolution of GCM model outputs is relatively coarse, with model grid cells spanning a latitude/ longitude in the range of 1° to 3° (or about 110 to 330 km). This spatial resolution is too coarse for the modeling of hydrological processes at a local scale. Therefore, a statistical downscaling technique known as quantile mapping was used to scale the large-scale GCM outputs down to twenty-four rainfall stations and six temperature stations in the region of the dam. The average of the future projections for the period 2040–99 of the annual minimum tempera- ture (that is, winter, during the flood season) and the maximum annual five-day rainfall in the dam’s catchment were assessed from the downscaled GCM models, as shown in table 7.5. The downscaled, projected time series of daily rainfall and temperatures were introduced into a rainfall-runoff model of the case study river basin. The calibrated model was used to convert the future changes in rainfall and temperatures to peak daily river discharge at the gauging station closest to the dam. These estimated changes in the peak discharge for future flood events, relative to the base- line period, are summarized in table 7.6 for the climate scenarios (that is, RCP4.5 and RCP8.5) and future periods (that is, 2040–65 and 2080–99) considered. The results indicate a signifi- cant increase in the peak discharge for the future 1 in 100 and 1 in 1,000 annual exceedance probability (AEP) events relative to the baseline period. There is also considerable scatter in the predicted change in the peak discharge, which reflects the uncertainty in the climate change projections of daily temperature and rainfall. TABLE 7.5 Change in Climate Change Parameters Compared with Baseline Period (1980–2000 in Current Study) RCP4.5 scenario RCP8.5 scenario Projection 2040–65 2080–99 2040–65 2080–99 Change in annual minimum winter (during flood 1.5 2.0 2.1 3.3 season) temperature (degree in Celsius) Change in maximum annual five-day rainfall 21 31 27 45 (percent increase) Source: Adapted from Veale et al. 2020. Note: Although the projected temperatures and rainfall parameters for both the RCP4.5 and RCP8.5 are calculated in terms of the twenty-fifth (lower), fiftieth (median), and seventy-fifth (upper) percentiles evaluated from the eleven GCM model outputs, this table shows only the fiftieth (median) percentile result based on the analysis of the closest rainfall or river gauging station to the case study dam. GCM = global circulation model; RCP = Representative Concentration Pathway. Case Studies on the Assessment of Climate Change Impacts on Floods 43 TABLE 7.6 Projected Percentage Change to 1 in 100 and 1 in 1,000 AEP Peak Flood Discharge Percentage change in peak discharge (ΔQp) compared to baselinea,b Climate scenario 1980–2000 2040–69 2080–99 1 in 100 AEP RCP4.5 0% (Baseline period) +24% (−14% to +67%) +35% (+13 to +57) RCP8.5 0% (Baseline period) +19% (+3% to +39%) +78% (+45% to +110%) 1 in 1,000 AEP RCP4.5 0% (Baseline period) +24% (−3% to +41%) +41% (+20% to +68%) RCP8.5 0% (Baseline period) +42% (−17% to +85%) +96% (+57% to +143%) Source: Adapted from Veale et al. 2020. Note: (ΔQp) at closest river gauging station to the case study dam, as a function of climate change scenarios (RCP4.5 and RCP8.5) for 2040–69 and 2080–99 periods. AEP = annual exceedance probability; GCM = global circulation model; RCP = Representative Concentration Pathway. a. Fiftieth percentile (median) of GCM multimodel ensemble values. Values in parentheses provide range between twenty-fifth (lower) and seventy-fifth (upper) percentiles. b. Percentage change from baseline values ΔQp = (Qp_future – Qp_baseline)/(Qp_baseline)*100. Climate Change Impacts on Hydrological Safety of the Dam The projected changes in the peak discharge (ΔQp) for future flood events relative to the baseline period (refer to table 7.6) were used as scaling factors to adjust the peak discharges of reservoir design flood hydrograph for the case study dam. The design flood hydrographs were routed through the reservoir to determine the peak spillway outflow and reservoir water level for cur- rent and future climate change scenarios. Figure 7.2 summarizes the results from this assessment in the form of a hydrological hazard curve and plots the peak reservoir water level as a function of flood frequency for both the baseline climate (1980–2000) and for future climate scenarios (RCP4.5 and RCP8.5) at the end of the century (2080–99). The solid lines on figure 7.2 represent the fiftieth percentile (median) from multimodel GCM ensemble values. The twenty-fifth and seventy-fifth percentile values are also plotted with dashed lines and shading on the figure. The twenty-fifth, fiftieth, and seventy-fifth percentile bands provide an indication of the variability in ensemble model climate change projections. Figure 7.2 indicates that the hydrological safety of the dam is projected to reduce signifi- cantly toward the end of the century. By the end of the century, the current 1 in 1,000 AEP reservoir water level reduces to approximately a 1 in 200 AEP flood under the RCP4.5 sce- nario. Under the RCP8.5 emission scenario, the current 1 in 1,000 AEP flood could reduce by the end of the century to less than a 1 in 100 AEP flood. These results clearly indicate the increasing hydrological risks of the dam as a result of projected future climatic changes toward the end of the century. 44 Enhancing the Safety and Resilience of Dams in the Context of Climate Change FIGURE 7.2 Case Study Dam Reservoir Peak Water Level as a Function of Flood Frequency (AEP) for RCP4.5 and RCP8.5 Climate Change Scenarios Reservoir water level (m) 61 60 Main dam - parapet wall crest 59 58 57 56 55 54 53 Maximum operation level 52 0 0 10 100 1,000 10,000 Annual exceedance probability (1 in Y) Baseline (1980–2000) RCP4.5 end century (2070–99) RCP8.5 end century (2070–99) Source: Adapted from Veale et al. 2020. Note: Fiftieth percentile (mean) of GCM multimodel ensemble values shown with thick lines, twenty-fifth and seventy-fifth percentile values shown with dashed lines and shading. JAPAN—NATIONWIDE CLIMATE CHANGE IMPACT ASSESSMENT ON LARGE FLOODS In Japan, many metropolitan areas have been developed in highly flood-prone areas at the downstream ends of river basins. Therefore, the country has established a river basin–based comprehensive and integrated flood management system, upon which river basin administra- tors provide guidance regarding dam operations to dam owners and operators during floods. Koike (2021) introduced Japan’s new flood management policy and system to deal with climate change impacts on floods, based on new GCM outputs and referred to as “the database for policy decision making for future climate change (d4PDF)” (MEXT 2019). The database contains the output of a set of highly accurate model experiments using high-resolution atmospheric models with meshes of 60 km and 20 km covering, respectively, the world and the surrounding area of Japan. Ensemble projections for 6,000 years (3,000 years for Japan and its surrounding area) represent the historical climate. Outputs for the future climate have been calculated for 3,240 and 5,400 years based, respectively, on RCP2.6 (equivalent to a 2°C rise in temperature) and RCP8.5 (equivalent to a 4°C rise). Case Studies on the Assessment of Climate Change Impacts on Floods 45 The Technical Working Group for Flood Management Planning, established in April 2018 under the River Sub-Committee of the National Land Development Council of Japan, identified the optimistic RCP2.6 as the primary scenario for future flood control planning, taking into account that the Paris Agreement in 2015 sets a goal of limiting global warming to 1.5°C above pre-in- dustrial levels and considering that the current levels of flood control safety of many rivers are lower than the target levels. The outputs for 1951–2010 are used to define the present climate conditions because the current flood control plans are mainly based on post-World War II data. The future climate conditions are, on the other hand, defined using the 2040 outputs because optimistic RCP2.6-based projections indicate that temperature increases after 2040 will stay more or less at the same level. It should be noted that whereas most RCP scenarios yield similar results till the 2050s, projections for various RCP scenarios diverge significantly thereafter, and it would have been wise to also consider the RCP4.5 and RCP8.5 scenarios to project possible future flood conditions across the country. A rainfall data set created by dynamic downscaling produced 5-km gridded model outputs for Japan and its surrounding area, which made it possible to compare the probabilities of extreme events under the present and future climate conditions. The whole country is divided into fifteen climate zones. In each, the probabilities of heavy rain events have been compared for the present and future climate conditions. The rates of change used for multiplying the design rainfall, and statistically derived from the observed and simulated data up to 2100, have been calculated to be 1.15 for Hokkaido and 1.1 for the other regions of the country (map 7.3 and table 7.7) for a 2°C increase in temperature. For a 4°C increase in temperature, the ratios are 1.5 for Hokkaido and northwestern Kyusyu and 1.3 for the other regions of the country (for three to nine hours’ short-range rainfall). Table 7.8 shows that severe floods will increase accordingly and that the frequency of severe floods will increase significantly as well. GCM data sets are huge and multidimensional and have different temporal and spatial reso- lutions, though their output in terms of meteorological parameters is not unified. These char- acteristics make the data difficult to process for climate change impact assessments. The Data Integration and Analysis System (DIAS), developed at the Japan Agency for Marine-Earth Science and Technology2; Kawasaki et al. 2017), makes GCM data more accessible by providing analysis tools with the functions of data selection and visualization for a given period, domain, scenario, climate variable, and ensemble DIAS can also compare model outputs with observed data (reference data), download daily and monthly data, and bias correct projected GCM rainfall data. DIAS archives all the data from the IPCC Fourth to Sixth Reports (IPCC 2007, 2014, 2023) and is equipped with functions for selecting models that reflect the climate characteristics of different regions, supporting the registration, quality control, and archiving of observation data; supplementing missing observed data with global data; and performing statistical downscaling and bias correction. 46 Enhancing the Safety and Resilience of Dams in the Context of Climate Change MAP 7.3 Regional Classification of Japan Source: Adapted from MLIT 2021. TABLE 7.7 Estimated Increase Rate In Rainfall with 100-Year Return Period In Japan +2°C +4°C +4°C(SR) Nationwide average 1.1 1.3 1.4 Hokkaido 1.15 1.4 1.5 Northwestern Kyushu 1.1 1.4 1.5 Other areas 1.1 1.2 1.3 Source: Adapted from MLIT 2021. Notes: Short-range (SR) means rainfall duration of more than three hours but less than 12 hours. Applicable to flood control plans for floods less than or equal to 200-year return period in the catchment area greater than or equal to 100 km2. Colors of areas correspond to map 7.3. TABLE 7.8 Estimated Average Increase Rates in Flood Discharge and Frequency of Floods Scenario Rainfall Flood discharge Frequency of floods +2°C 1.1 ≈1.2 ≈2 +4°C 1.3 ≈1.4 ≈4 Source: Adapted from MLIT 2021. Case Studies on the Assessment of Climate Change Impacts on Floods 47 SRI LANKA—CLIMATE CHANGE IMPACT ASSESSMENT AND FUTURE INUNDATION ANALYSIS Selvarajah et al. (2021) assessed climate change impacts on the hydrometeorological charac- teristics of the Mahaweli River Basin in Sri Lanka (map 7.4 and figure 7.3) using downscaled and bias-corrected GCM output of four selected GCMs for the RCP8.5 scenario. The DIAS of Japan and the Water Energy Budget-based Rainfall-Runoff-Inundation model (WEB-RRI) were utilized to develop an integrated approach, which was then applied to the basin to investigate climate change impacts on its hydrometeorological characteristics. Mahaweli River Basin is the largest and perennial river basin in Sri Lanka, which drains 10,300 km2, as in map 7.4. Rainfall in the basin varies greatly over space and time. Because of the seasonal variability of rainfall, the island’s weather seasons are classified into four catego- ries: northeast (NE) monsoon, southwest (SW) monsoon, intermediate monsoon-1 (IM-1), and intermediate monsoon-2 (IM-2). The NE monsoon (December to February) brings more precip- itation to the downstream region of Mahaweli River Basin (500 to 1,200 mm). MAP 7.4 The Largest and Perennial River Basin in Sri Lanka a. Topographical map of b. Topography of c. Land use map with discharge Sri Lanka with the demarcation of the Mahaweli River basin measurement locations the Mahaweli River Basin Elevation (m) 1–100 100–1000 1000–1500 River path 1500–2500 River basin Land use Broadleaf evergreen trees Broadleaf and needleleaf trees Broadleaf shrubs with bare soil Agriculture/C3 grassland Water, wetlands Source: Selvarajah et al. 2021. 48 Enhancing the Safety and Resilience of Dams in the Context of Climate Change Figure 7.3 shows the overall research framework and methodology adopted for this study, which included five main components: (a) GCM selection, statistical downscaling, and bias correc- tion of rainfall projections; (b) assessment of climate change impact on future precipitation and temperature; (c) development of a hydrological model aimed at simulating the hydrological responses of the basin; (d) assessment of climate change impacts on basin runoff and extreme flows; and (e) facilitation of decision-making procedures. The results for the RCP8.5 scenario from four selected GCMs showed that, with an average tem- perature increase of 1.1°C over the twenty years in future (2026–45), the basin will experience more extreme rainfall (increase ranging 204 to 476 mm/year) and intense flood disasters. The results of more detailed assessment of the climate change impact on flood events are summa- rized in the next section. Extreme Flood Event Analysis The daily average discharge simulated for the past and future climates were rank ordered, and the first twenty were analyzed for extreme floods. FIGURE 7.3 Assessment Framework 1. Model selection, downscaling, GCM and bias correction Global observed/ Model reanalysis selection products Rainfall (downscaled Future temperature and bias corrected) change Static data, dynamic 3. Hydrologic model development vegetation data, dynamic forcing data, and response In situ evaluation data rainfall data 2. Meteorological assessment impact Past Future Calibration and validation Hydrologic model 4. Hyrological assessment 5. impact Decision Past Future making Source: Selvarajah et al. 2021. Note: GCM = General Circulation Model. Case Studies on the Assessment of Climate Change Impacts on Floods 49 Figure 7.4, panels a through d, shows the first twenty peak discharges estimated by a physi- cally based distributed hydrological model for the inputs from CanESM2, CNRM-CM5, GFDL- ESM2G, and MPI-ESM-LR, respectively. As shown in the figures, the projected future peak discharges are significantly higher in CanESM2 and MPI-ESM-LR and marginally higher in the other two models. Therefore, peak flood discharges are highly likely to increase in the future in this basin. Similar findings were reported in the Assessment Report 5 of the IPCC, explaining that GCMs have projected the impacts associated with increased flooding with high confidence in the case of small islands (Bertoli, Balzarolo, and Todini 2022). FIGURE 7.4 First 20 Peak-Flow Events of Daily Average Discharge Values of the Past and Future 20 Years of the Selected GCMs a. CanESM2 b. CNRM-CM5 First 20 peak flow (m3/S) First 20 peak flow (m3/S) 1200 1200 1000 1000 800 800 600 600 400 400 200 200 0 0 5 10 15 20 5 10 15 20 Rank Rank c. GFDL-ESM2G d. MPI-ESM-LR First 20 peak flow (m3/S) First 20 peak flow (m3/S) 1200 1200 1000 1000 800 800 600 600 400 400 200 200 0 0 5 10 15 20 5 10 15 20 Rank Rank Past Future Source: Selvarajah et al. 2021. 50 Enhancing the Safety and Resilience of Dams in the Context of Climate Change Inundation Analysis The study assessed the difference in flood inundation depth between past and future climate con- ditions, simulated with the hydrological model for each GCM individually. The difference between past and future climates in all-time maximum inundation depth at each grid cell was estimated by the model-simulated inundation gridded data for each GCM individually and is plotted in map 7.5, panels e through h, and the past all-time inundation depth gridded data are given in map 7.5, panels a through d, for reference. As shown in the figure, except for the CNRM-CM5 model, the other three models project an increase in the inundation extent for all-time flood depth. Overall, it is likely that the flood inundation area and risk will increase in the basin under the future climate. MAP 7.5 Past Inundation Depth and Future Inundation Depth Changes Projection by Different Climate Models in Sri Lanka a. Past all-time b. Past all-time c. Past all-time inundation depth for inundation depth for inundation depth for CanESM2 CNRM-CM5 GFDL-ESM2G In undation Past Past Past depth in m 12.0 10.0 8.4 8.4 8.4 8.0 7.0 8.2 8.2 8.2 6.0 5.5 8.0 8.0 8.0 5.0 4.5 7.8 7.8 7.8 4.0 3.5 3.0 7.6 7.6 7.6 2.5 2.0 7.4 7.4 7.4 1.5 1.0 7.2 7.2 7.2 0.5 0.0 –0.5 7.0 7.0 7.0 –1.0 –1.5 6.8 6.8 6.8 –2.0 80.6 80.8 81.0 81.2 80.6 80.8 81.0 81.2 80.6 80.8 81.0 81.2 d. Past all-time e. Inundation depth f. Inundation depth inundation depth for deference between future deference between future MPIESM-LR and past of CanESM2 and past of CNRM-CM5 In undation Past Future-past Future-past depth in m 12.0 8.4 8.4 8.4 10.0 8.0 7.0 8.2 8.2 8.2 6.0 5.5 8.0 8.0 8.0 5.0 4.5 7.8 7.8 7.8 4.0 3.5 3.0 7.6 7.6 7.6 2.5 2.0 7.4 7.4 7.4 1.5 1.0 7.2 7.2 7.2 0.5 0.0 7.0 –0.5 7.0 7.0 –1.0 –1.5 6.8 6.8 6.8 –2.0 80.6 80.8 81.0 81.2 80.6 80.8 81.0 81.2 80.6 80.8 81.0 81.2 (figure continues next page ) Case Studies on the Assessment of Climate Change Impacts on Floods 51 MAP 7.5 Past Inundation Depth and Future Inundation Depth Changes Projection by Different Climate Models in Sri Lanka (Continued) g. Inundation depth h. Inundation depth deference between future deference between future and past of GFDL-ESM2G and past of MPIESM-LR In undation Future-past Future-past depth in m 12.0 8.4 8.4 10.0 8.0 7.0 8.2 8.2 6.0 5.5 8.0 8.0 5.0 4.5 7.8 7.8 4.0 3.5 3.0 7.6 7.6 2.5 2.0 7.4 7.4 1.5 1.0 7.2 7.2 0.5 0.0 –0.5 7.0 7.0 –1.0 –1.5 6.8 6.8 –2.0 80.6 80.8 81.0 81.2 80.6 80.8 81.0 81.2 Source: Selvarajah et al. 2021. To verify the simulation results further, the all-time seasonal (that is, NE monsoon) maximum inundation depths simulated by the hydrological model were estimated. Figure 7.5 shows the scatter plot of past versus future all-time seasonal maximum inundation depths during the NE monsoon for all the model grids. As shown in the figure, CNRM-CM5 projects a decrease in all-time seasonal inundation depth in the future climate compared with those in the past climate during the NE monsoon, whereas the other three models project clear increases in the future inundation depth. The contrasting behavior of a decreased inundation depth from CNRM-CM5 is considered to be a result of uncertainties in rainfall projections during the NE monsoon season. 52 Enhancing the Safety and Resilience of Dams in the Context of Climate Change FIGURE 7.5 Past vs. Future All-Time Seasonal Maximum Inundation during the NE Monsoon of the Selected GCMs a. CanESM2 b. CNRM-CM5 Future all-time maximum inundation (m) Future all-time maximum inundation (m) 12 12 10 10 8 8 6 6 4 4 2 2 0 0 0 2 4 6 8 10 12 0 2 4 6 8 10 12 Past all-time maximum inundation (m) Past all-time maximum inundation (m) c. GFDL-ESM2G d. MPI-ESM-LR Future all-time maximum inundation (m) Future all-time maximum inundation (m) 12 12 10 10 8 8 6 6 4 4 2 2 0 0 0 2 4 6 8 10 12 0 2 4 6 8 10 12 Past all-time maximum inundation (m) Past all-time maximum inundation (m) No trend Source: Selvarajah et al. 2021. Note: GCM = General Circulation Model; NE = northeast. Case Studies on the Assessment of Climate Change Impacts on Floods 53 NOTES   1. It is noted that since the case study assessment described here, IPCC published its Sixth Assessment Report in 2021, for which it used SSPs instead of RCPs. Although not available at the time of the case study assessment, any subsequent climate change impact assessment for dams should utilize climate change projections for multiple SSPs rather than for RCPs.   2. http://www.diasjp.net/en/. CHAPTER 8 Case Studies on the Application of Advanced Rainfall and Reservoir Inflow Forecasting and Optimal Reservoir Operation Systems ITALY—ENHANCED WATER SUPPLY AND FLOOD CONTROL FOR LAKE COMO Lake Como is a natural reservoir with a weir with multiple gates controlling its outlet (photo 8.1). Although the lake water is mainly used for irrigation and hydropower generation, protection against floods has become an increasingly important additional purpose. The live storage of the reservoir is 246 mm³ and is only 5 percent of the lake’s mean annual inflow volume. Combined with the small discharge capacity of the control gates, this renders reservoir management challenging. An adaptive reservoir management approach was implemented in 1997, incorporating short- term predictive probability density forecasting information into decision making (Todini 1999). Operating rules for meeting long-term objectives for hydropower generation and irrigation water supply were established on a ten-day basis using a stochastic dynamical programming algorithm. The short-term operating rule, based on the reservoir level, was derived conditionally to daily inflow forecasts with the objective of minimizing total expected losses. PHOTO 8.1 Olginate Dam on the Adda River at the Outlet of Lake Como, Seen from Both Sides a. With fully open sluices b. With partly closed sluices Source: Bertoli, Balzarolo and Todini 2022. 54 Case Studies on the Application of Advanced Rainfall and Reservoir 55 Despite conflicting management objectives of the multipurpose reservoir, the decision support system demonstrated significant operational improvements: (a) the frequency of the flooding of the city of Lake Como reduced by more than 30 percent; (b) the water deficit reduced by 110 million m³/year—that is, on average, by 12 percent; and (c) the hydropower generation increased, on average, by 3 percent. The evaluation of Lake Como’s management provided a satisfactory outcome by fully matching the decision support system–anticipated results during the drought years of 2000–06 (ICOLD 2016). The decision support system aims at optimizing conflicting operational objectives with an advanced flood forecasting system. Whereas the system’s operating rules focus on maximizing long-term benefits, it also guides operators on the amount of water to be released from the res- ervoir during floods in order to reduce dam failure and downstream flooding risks. Under flood conditions, dam managers will focus mostly on short-term objectives instead of on long-term expected benefits. Long-term operating rules need to be temporarily relaxed by minimizing the expected damages based on short-term flood forecasts. It is then necessary to decide how much water must be preventively released in a trade-off between the expected benefits of avoiding downstream flooding, estimated using predictive probability density functions, and the expected losses from the reduced availability of water because of the preventive releases. The decision support system makes extensive use of probabilistic forecasts, stochastic optimi- zation, and Bayesian decision techniques implemented in a user-friendly environment. The experience also illustrates the importance of collaboration among reservoir managers, s­ cientists, and other stakeholders for developing an advanced rainfall and flood forecasting system integrated with a system for optimizing reservoir operation (Bertoli, Balzarolo, and Todini 2022; ­ Todini 2018). UNITED STATES—ENHANCED WATER SUPPLY AND FLOOD CONTROL FOR LAKE MENDOCINO AND COYOTE VALLEY DAM, CALIFORNIA In a significant policy shift, the United States Army Corps of Engineers and others developed the Forecast Informed Reservoir Operations (FIRO) program to utilize forecasted weather con- ditions for better addressing dam operations (NOAA 2017). The intent is to provide a reservoir operation system that better informs decision making on retaining or releasing water by inte- grating additional flexibility in operation policies and rules based on enhanced monitoring and improved weather and flow forecasts, with the aim to maximize various development objectives, such as water supply, hydropower generation, and flood attenuation. One of the first applications of the FIRO program was for the Coyote Valley Dam on Lake Mendocino in California, with a reservoir capacity of 137 million m3. The Russian River Basin, which feeds Lake Mendocino, has one of the most variable climates in the United States, and water storage in the reservoir fluctuates greatly by season. Over time, it became clear that Lake Mendocino was meeting neither water resources nor flood control demands efficiently. Through a systems approach, it was determined that Lake Mendocino could be managed more efficiently by integrating reservoir inflow forecasts into decisions on the release schedule by modifying the Water Control Manual. 56 Enhancing the Safety and Resilience of Dams in the Context of Climate Change Operational Testing for Lake Mendocino The United States Army Corps of Engineers was to pilot test the system, allowing for major deviations by applying the hybrid Ensemble Forecast Operations (EFO) model during water years (WYs) 2019 and 2020 (note that, for example, WY 2020 runs from October 1, 2019, to September 30, 2020). The hybrid EFO adds a variable buffer pool to the guide curve and uses a fifteen-day streamflow ensemble forecast to recommend flood releases. The test period coincided with an extreme drought in the region. Whereas WY 2019 was a rel- atively wet year, WY 2020 was the third driest year on record for 127 years. FIRO increased the water supply benefits and managed flood risks in both years. Figure 8.1 shows the outcome for WY 2020, when FIRO enabled a 19 percent (more than 13.6 million m³ or 11,175 acre-feet) increase in water storage by the end of March. This allowed the service provider to maintain water supplies throughout the subsequent very dry period in the region. The incremental water supply was enough to provide approximately 22,000 homes with water for the entire year. FIGURE 8.1 Lake Mendocino Storage Rule Curve and FIRO Operations in WY 2020 Acre-feet 100,000 100,000 95,000 95,000 90,000 90,000 85,000 85,000 80,000 80,000 FIRO and major deviation 75,000 75,000 variable buffer pool 70,000 ~11,175 70,000 acre-feet 65,000 (~19% 65,000 Water supply pool increase) 60,000 60,000 55,000 55,000 50,000 50,000 Oct. 2019 Nov. 2019 Dec. 2019 Jan. 2020 Feb. 2020 Mar. 2020 Apr. 2020 Date Major deviation encroachment curve Storage curve Modeled 2020 WY storage without FIRO + Major deviation Actual 2020 WY storage Source: Jasperse et al. 2020. Note: As of March 30, 2020, precipitation since October 2019 is about 38.4 percent of average and water supply storage is about 19 percent greater than modeled storage without FIRO and major deviation. FIRO = Forecast Informed Reservoir Operations; WY = water year. Case Studies on the Application of Advanced Rainfall and Reservoir 57 Ensemble Streamflow Predictions and Ensemble Forecast Operations The FIRO system uses EFO, which is a risk-based approach of reservoir flood control operations that incorporates ensemble streamflow predictions made by the California-Nevada River Forecast Center. Reservoir operations for each member are individually modeled to forecast system condi- tions and calculate risk of reaching critical operational thresholds (Delaney et al. 2020). Reservoir release decisions are simulated to better manage forecasted risks with respect to established risk tolerance levels. The EFO system was developed for Lake Mendocino (photo 8.2, figure 8.2, and map 8.1) to provide foundational inputs for the FIRO implementation and to evaluate its viability for improving the reliability of reservoir storage without increasing downstream flood risks. PHOTO 8.2 Lake Mendocino Source: Sonoma Water. Lake Mendocino is a dual-use reservoir, owned and operated for flood control by the United States Army Corps of Engineers and operated for water supply by Sonoma Water. EFO was cali- brated using a twenty-six-year (1985–2010) ensemble streamflow prediction hindcast generated by the California-Nevada River Forecast Center, which provides sixty-one-member ensembles of fifteen-day flow forecasts. EFO simulations generally yield higher storage levels during the flood management season while maintaining the needed flood storage capacity by strategically prere- leasing water in advance of forecasted storms. Model results demonstrate a 33 percent increase in median storage at the end of the flood management season (May 10) over existing operations, without marked changes in flood frequency for locations downstream from Lake Mendocino. EFO has been demonstrated to be a viable alternative for managing flood control operations at Lake Mendocino, which provides multiple benefits for water supply, flood mitigation, and eco- systems and provides a management framework that could be adapted and applied to other flood control reservoirs. 58 Enhancing the Safety and Resilience of Dams in the Context of Climate Change FIGURE 8.2 Atmospheric River Source: Brian Kawzenick, Center for Western Weather and Water Extremes. Used with permission. Further permission required for reuse. Continued Enhanced Technological Development and Collaboration toward FIRO2.0 Work continues for advancing the FIRO system to raise the water supply level to a higher ele- vation to increase the water supply with the aid of further advanced rainfall forecasting along with a flood forecasting system under FIRO 2.0. Given the potential risks for the safety of the dam and for downstream communities, FIRO 2.0 will require the application of state-of-the-art technologies for data collection and monitoring of the watershed and the atmosphere, includ- ing additional soil moisture, precipitation and stream gauges, vertically profiling radars, radio- sondes, and so on. It will also require an observational database for evaluating the performance of the numerical weather prediction and flood forecasting models, which are still constrained by uncertainties in the model’s initial state and in its numerical approximations, physical process parameterizations, and subgrid-scale unresolved processes. The case of Lake Mendocino illustrates the importance of the introduction of advanced tech- nologies, building the capacity of responsible agencies, and ensuring close collaboration among key stakeholders, including technical and research agencies for climate and hydrometeorology. Case Studies on the Application of Advanced Rainfall and Reservoir 59 MAP 8.1 Location Map Eel River Lake Potter Valley Project Powerhouse Pillsbury $ 1 PG&E Potter Valley Project £ ¤ 101 Coyote Valley Dam $ 1 Lake Mendocino Ukiah Russ U V253 ia n Riv $ Hopland1 V U29 ¯ er U V Hopland 175 Legend Russian River $ 1 Discharge Gage $ Cloverdale 1 V U 53 Cloverdale Lake Sonoma U V 175 U V Dr y 1 C r ee k Healdsburg $ 1 Healdsburg R i ver Windsor sian Rus U V116 £ ¤101 V U 128 Santa Rosa µ Sebastopol V U 12 Rohnert Park 0 2.5 5 10 Miles Cotati U V 12 Source: Sonoma Water. Used with permission. Further permission required for reuse. £ ¤ 121 JAPAN—ENHANCED FLOOD CONTROL AND HYDROPOWER GENERATION OF HATANAGI-I DAM ON OI RIVER The Hatanagi-1 Dam (map 8.2 and photo 8.3) is a single-purpose hydropower dam located in the upper Oi River in Japan. The dam is a hollow-core concrete gravity dam with a height of 125 m, 107 million m³ storage capacity, and 86 MW installed capacity, owned by the Chubu Electric Power Co., Inc. (CEP). The catchment area of the dam is 318 km², formed by rigged mountain- ous terrain. The project was commissioned in 1962. The government of Japan has issued a guideline requesting nongovernment owners of dams that do not have flood control functions to contribute to flood management at the basin scale, with the aid of an advanced rainfall and flood forecasting system. In advance of the develop- ment, CEP has recently implemented a proof of concept (PoC) for an advanced rainfall, flood forecasting, and reservoir operation system aimed at reducing downstream flood disasters under 60 Enhancing the Safety and Resilience of Dams in the Context of Climate Change MAP 8.2 Hatanagi-I Dam on the Oi River, Japan a. Location of Hatanagi I Dam in Japan b. Location and Basin Area of Hatanagi I Dam N 35°40'N Sea of Japan 35°35'N Study area Pacific Japan ocean 35°30'N 0 150 300 600 900 Km 35°25'N Legend Elevation (meter) Dams 265–600 Mountain (>3100m) 600–1,000 35°20'N Oi river 1,000–1,400 Temperature gauges 1,400–1,800 Rain gauges 1,800–2,200 Study area 2,200–2,600 35°15'N 2,600–3,000 3,000–3,185 138°5'E 138°10'E 138°15'E 138°20'E Source: Naseer at al. 2019. climate change and increasing hydropower generation. In collaboration with the International Centre for Water Hazard and Risk Management (ICHARM), CEP utilized the Data Integration Analysis System (DIAS) developed by the University of Tokyo for the PoC (ICHARM 2022; Koike et al. 2021; Namakura et al. 2022). The system applies a high-resolution regional ensemble rainfall forecasting system, developed using the Weather Research and Forecasting (WRF) Model and the local ensemble transform Kalman filter (LETKF) with thirty-two ensemble members being updated every six hours for thirty-nine-hour duration forecasts, with 3-km spatial and 1-hour temporal resolution. The inflow forecasting system consists of two steps: Step 1 calculates the surface soil moisture, the groundwater conditions, and discharge up to the present time, and step 2 conducts the ensemble forecasting of future discharges up to thirty-nine hours ahead. The system uses the WRF model for numerical weather prediction. The initial and boundary conditions are derived from the Japan Meteorological Agency’s global forecast, which is based on the Global Spectral Model1. The LETKF is used to generate ensemble perturbation and data assimilation. Data assimilation uses the observation data (PREPBUFR)2 that is employed in the operational numerical weather fore- casts produced by the United States and released to the world in near real time. Furthermore, a distributed hydrological model for the runoff and reservoir inflow forecasting is used, with 1-km spatial and 1-hour temporal resolution, and calibrated and validated using observed rainfall, Automated Meteorological Data Acquisition System rainfall data, and dam inflow data provided by CEP. The model system proved to perform well with a Nash-Sutcliffe coefficient of 0.71. Case Studies on the Application of Advanced Rainfall and Reservoir 61 PHOTO 8.3 Hatanagi I Dam Source: Adobe Stock. Making use of the bandwidth of the ensemble of thirty-two inflow forecasts, the dam operation framework tests the general dam operation procedures: • Use the upper 25 percent subensemble of reservoir inflow time series if flood control is the pri- ority, or use the lower 25 percent subensemble of time series if the priority is securing water for more hydroenergy generation. • Check whether the reservoir inflow exceeds the amount that can be safely discharged through the dam during the forecasting period under the existing operating rules. • Check whether the accumulated inflow within the forecasting period exceeds the storage capac- ity at the forecasting start time. • If the inflow exceeds the designated amount in steps 2 and 3, the excess amount will be released evenly while prioritizing securing water for power generation within the forecasting period. If the inflow does not exceed the designated amount in steps 2 and 3, the average value of the predicted inflow up to thirty-nine hours ahead will be applied as the average hourly amount of water to be used for power generation. 62 Enhancing the Safety and Resilience of Dams in the Context of Climate Change The performance evaluation of this upgraded operational framework demonstrated that the res- ervoir operation system can effectively control flood discharges to downstream areas within the specified discharge control targets and, at the same time, increase average hydropower genera- tion. The evaluation indicator for flood control was set to keep the reservoir level below the full water storage level when the flood events exceed 600 m³/s. According to the power generation volume index, average energy generation from July to October (rainy season) showed increases of 13 percent in 2018 and 4 percent in 2019. As an example, figure 8.3 shows the results of the pilot application of the operation system in which the proposed dam operating framework was applied continuously during the warm seasons (July to October) of 2018 and 2019. The flood events exceeding 600 m³/s occurred once during each of the two warm seasons: September 30 – October 1, 2018, and October 12–13, 2019. The 2018 event was a typical case in which the dam water level was high at the start of the forecast- ing period. In this event, the inflow was forecast while the dam water level was almost the same as the prerelease water level, and prereleases were performed on the morning of September 29 (pink range). As a result, the simulated dam water level (orange dotted line) decreased, and the simulated peak flow (pink range) was successfully controlled to remain below the 600 m³/s level when the inflow (solid blue line) peaked. If the prereleases had not been performed, the flood discharge would have exceeded 600m³/s, as shown in the orange range, causing flood events in downstream areas. FIGURE 8.3 Details of Flooding with a Gate Discharge Not Exceeding 600m3/s a. September 30 – October 1, 2018 950 3,000 0 945 940 2,500 Precipitation (mm/h) 935 50 Discharge (m3/s) Water level (m) 930 2,000 925 920 1,500 100 915 910 1,000 905 150 900 500 895 890 0 200 9/29/2018 9/30/2018 10/1/2018 10/2/2018 Date b. October 12–13, 2019 950 3,000 0 945 940 2,500 Precipitation (mm/h) 935 50 Discharge (m3/s) Water level (m) 930 2,000 925 920 1,500 100 915 910 1,000 905 150 900 500 895 890 0 200 10/11/2019 10/12/2019 10/13/2019 10/14/2019 Date Dam water level Dam water level (sim.) Top water level Preliminary discharge water level Dam inflow Predicted dam inflow Total discharge Discharge for power generation Total discharge (Sim.) Discharge for power generation (Sim.) Precipitation Source: Koike et al. 2021. Case Studies on the Application of Advanced Rainfall and Reservoir 63 The 2019 event was a typical case in which the dam water level was low at the forecasting start time. With the water storage capacity still available, the inflow was stored effectively up to the full water storage level (red dotted line) in advance, and the flood discharge was successfully controlled to remain below 600 m³/s. In addition to the advanced operation system based on ensemble forecasting with a lead time of thirty-nine hours, the applicability of seasonal ensemble forecasting is now under review. By downscaling the three-month ensemble weather forecasts provided by the Japan Meteorological Agency, Nakamura et al. (2022) applied thirteen sets of ensemble forecast with a 15-km spatial resolution and a ninety-day lead time and checked the system’s performance. The system with a thirty-day lead time showed the possibility of a 6 percent increase in annual hydropower generation. Further research is in progress to determine the optimal combination of lead time selection and dam operation rules. JAPAN—PRE-FLOOD RESERVOIR DRAWDOWN OPERATION OF THE KUSAKI MULTIPURPOSE DAM The Kusaki Dam is a concrete gravity dam with a height of 140 m and a reservoir capacity of 60.5 mm³ (photo 8.4 and figure 8.4). The project was commissioned in 1976 and is managed by the Japan Water Agency. The project serves several purposes, including 20-MW hydropower generation, water supply to the Tokyo metropolitan area, and irrigation and flood control of the downstream Tone and Watarase Rivers. PHOTO 8.4 Kusaki Dam The Kusaki Dam Kusaki Bridge Kusaki Lake 2 yN o. 12 Highwa ional Nat Source: Watarase River Office, MLIT 2019. 64 Enhancing the Safety and Resilience of Dams in the Context of Climate Change FIGURE 8.4 Kusaki Dam’s Rainfall Forecast for Flood Control Operation toward Typhoon Hagibis (October 2019) Rainfall forecast scenario Total rainfall for next 10 days (mm) 500 450 400 350 300 250 200 150 100 50 0 10/7 10/8 10/9 10/10 10/11 10/12 Forecasting time point Maximum case Minimum case Maximum Median Minimum Source: JWA 2023. Used with permission. Further permission required for reuse. During Typhoon Hagibis from October 11–13, 2019, the Kusaki Dam conducted an emergency pre-flood reservoir drawdown based on weather and flood forecasting information (JWA 2023). A pre-flood release of about 15 mm³ of stored water, corresponding to 30 percent of the total effective reservoir capacity, was carried out by lowering the reservoir water level by 14 m. Because the designated storage capacity of the dam for flood control is 20 mm³, the additional storage capacity of 15 mm³ amounted to a significant 75 percent increase. Without this pre-flood drawdown operation, the dam would have had to move into emergency operation, requiring the dam operator to discharge the full amount of inflow to prevent the reservoir level from exceed- ing the maximum surcharge flood water level and to ensure dam safety. Thus, the operator was able to limit the maximum outflow to about 600 m³/s compared with about 1,700 m³/s without such drawdown. Accordingly, the maximum flood reservoir level was reduced by about 2.6 m (figure 8.5). Without such operation, critical downstream areas protected by embankment dikes could have been flooded as a result of breaches. As shown in figure 8.6, an extreme amount of rainfall (as much as 437 mm) was predicted, and the dam operator was required to conduct an additional amount of predischarge in coordination with the river manager. The preliminary discharge,3 per the standard operation rules, is part of the original dam operation procedure for flood control. On the other hand, the ad hoc predischarge4 is an additional safety measure for drawing down the reservoir level and sacrificing stored water for flood storage based on an advanced rainfall and flood forecasting system. As shown in figure 8.7, the dam operator conducted the prelimi- nary discharge and predischarge in the amount of about 15.5 million m³, in addition to the allo- cated flood control capacity of 20 million m³ in coordination with the river manager. Case Studies on the Application of Advanced Rainfall and Reservoir 65 FIGURE 8.5 Reduced Flood Water Level at a Downstream River Section Resulting from Kusaki Dam Drawdown Operation River section at Takatsudo point Kusaki Dam Water level (m) Without any Without pre-flood Takatsudo point 8 measures at drawdown Kusaki Dam (pre-discharge) 7 Tokyo 6 Water level drop by Kusaki Dam: 2.6 m W.L. indicating The Japan Society 5 danger of inundation of Civil Engineers 4 Actual water level: 4.29 m Technology Award W.L. for judging Group I) for the 3 evacuation Kusaki Dam’s 2 achievement during W.L. indicating Typhoon Hagibis attention to inundation 1 W.L. for flood prevention 0 team standby –1 Without the Kusaki Dam, there would have been more flood damage downstream. The water level would have reached the river water level indicating danger of inundation, if pre-flood drawdown (pre-discharge) operation had not conducted. Source: JWA 2023. Used with permission. Further permission required for reuse. FIGURE 8.6 Flood Forecast at 3 p.m. October 11, 2019 Prompting Emergency Reservoir Drawdown Hourly rainfall (mm/h) Total rainfall (mm) 0 0 10 100 20 200 30 300 Result Forecast 40 437.3 mm 437 mm 400 50 500 4: 0 6: 0 8: 0 10 00 12 00 14 00 16 00 18 00 20:00 22:00 0 4: 0 6: 0 8: 0 10 00 12 00 14 00 16 00 18 00 20:00 22:00 0 4: 0 6: 0 8: 0 10 00 12 00 14 0 16 00 18 0 20:00 0 0 0 0 :0 0 0 0 :0 0 0 0 :0 :0 :0 2: : : : : 2: : : : : 2: : : 10/11 10/12 10/13 Flow rate (m3/s) Water level (EL. m) 2,000 460 Highest water level during a flood: 454.0 m Max. inflow 1,793 m3/s 1,800 455 1,600 Max. discharge 450 Water level 1,400 1,687 m3/s 445 1,200 440 Flood season basic water level 1,000 435 800 430 600 425 400 Shift to 420 Preliminary discharge emergency 200 spillway gate 415 operation 0 410 4: 0 6: 0 8: 0 10 00 12 00 14 00 16 00 18 00 20:00 22:00 0 4: 0 6: 0 8: 0 10 00 12 00 14 00 16 00 18 00 20:00 22:00 0 4: 0 6: 0 8: 0 10 00 12 00 14 0 16 00 18 00 20 00 0 0 0 0 :0 0 0 0 :0 0 0 0 :0 :0 2: : : : : 2: : : : : 2: : : : : 10/11 10/12 10/13 Source: JWA 2023. Used with permission. Further permission required for reuse. 66 Enhancing the Safety and Resilience of Dams in the Context of Climate Change FIGURE 8.7 Impact of Pre-Flood Reservoir Drawdown Operation Hourly rainfall (mm/h) Total rainfall (mm) 0 0 10 100 20 200 30 300 40 356.1 mm 400 50 500 4: 0 6: 0 8: 0 10 00 12 00 14 0 16 00 18 0 20:00 22:00 0 4: 0 00 8: 0 10 00 12 00 14 0 16 00 18 0 20:00 22:00 0 4: 0 6: 0 8: 0 10 00 12 00 14 0 16 00 18 0 20:00 0 0 0 0 :0 :0 :0 0 0 :0 :0 :0 0 0 0 :0 :0 :0 2: : : 2: 6: : : 2: : : 10/11 10/12 10/13 Flow rate (m3/s) Water level (EL. m) 2,000 460 Highest water level during a flood: 454.0 m Inflow Water level 1,800 455 1,600 Preliminary discharge start Flood control 450 EL.443.94 m EL.443.94 m volume 1,400 20,000,000 m3 445 Flood season basic 3.34 m 1,200 water level: 440.6 m 440 21,840,000 m3 1,000 m3/s 1,000 Preliminary discharge storage Flood reduction 435 4,520,000 m3 14.17 m control 800 430 600 Pre-flood drawdown EL.426.43 m 425 (predischarge) 400 15,460,000 m3 Pre-flood Discharge 420 200 Preliminary discharge drawdown 415 (predischarge) 0 410 4: 0 6: 0 8: 0 10 00 12 00 14 00 16 00 18 00 20:00 22:00 0 4: 0 6: 0 00 10 00 12 00 14 00 16 00 18 00 20:00 22:00 0 4: 0 6: 0 8: 0 10 00 12 00 14 0 16 00 18 00 20:00 0 0 0 0 :0 0 0 :0 0 0 0 :0 :0 2: : : : : 2: 8: : : : : 2: : : : 10/11 10/12 10/13 Source: JWA 2023. Used with permission. Further permission required for reuse. AUSTRALIA—WIVENHOE DAM FLOOD OPERATION INCIDENT, BRISBANE A suboptimal flood control operation of the Wivenhoe Dam in Australia caused a severe flooding incident because it gave more priority to storing water for water supply than to releasing water for flood control. The 2011 flood affected southeast Queensland in that (a) thirty-three people lost their lives, (b) flood damages amounted to $3.2 billion across ninety towns, and (c) more than 20,000 homes were flooded in Brisbane and 3,000 homes in nearby Ipswich (see photo 8.5). This case highlights the importance of establishing a proper decision support system and oper- ation manual for multipurpose dam operations, which inherently include conflicting demands for water supply and flood control. The introduction of advanced rainfall and inflow forecasting systems is equally important for supporting the decision support systems for reservoir operation. Because Australia is one of the most advanced countries regarding dam safety in the world, with a low number of cases of dam failures and incidents, this incident caught the attention from the public and was relatively well studied for learning purposes (Thompson 2011; Van den Honert and McAneney 2011). Case Studies on the Application of Advanced Rainfall and Reservoir 67 PHOTO 8.5 Brisbane 2011 Flood Disaster a. Flooded businesses, South Brisbane b. Flooding among houses, Ipswich Source: State Library of Queensland. Photographers Leif Ekstrom (a) and Dean Saffron (b). Creative Commons Attribution 4.0 International (CC-BY 4.0) license. The Wivenhoe Dam is situated on the Brisbane River in Queensland, Australia, 80 km by road from the center of Brisbane, with a total catchment area of 7,020 km² (map 8.3). Somerset Dam is located upstream of the Wivenhoe Dam with a catchment area of 1,347 km². The pri- mary purpose of the Wivenhoe Dam is to supply potable water for the Brisbane and Ipswich regions. It also provides for flood mitigation control, hydropower generation, and recreation. The dams are operated as water supply and flood control reservoirs by Seqwater, the Queensland Government Bulk Water Supply Authority.5 Wivenhoe Dam was planned in the early 1970s, and the 1974 Brisbane flood highlighted the need of flood protection for the southeast Queensland area. Construction was completed in 1984. The Wivenhoe Dam is a rock and earthfill embankment dam with a concrete spillway 59 m in height and 2,300 m in crest length. The dam has a normal storage capacity (full supply storage) of 1.1 billion m³ and a flood storage capacity of 2.0 billion m³—thus a total storage capacity of about 3.1 billion m³. The gated spillway, with five steel crest gates, has a discharge capacity of 12,000 m³/s. On January 13, 2011, major flooding occurred throughout most of the Brisbane River catch- ment (photo 8.6). Figure 8.8 indicates the reservoir operation of the dam, with the sharp increase in flood discharge from the dam as the water level rose toward 75.0 masl. It seems that had the operator anticipated the second peak flood relying on the rainfall forecast, he could have released a larger amount of water and lowered the reservoir water level so that the reservoir could have attenuated the peak flood, reducing the peak flood discharge to down- stream areas. Indeed, questions were raised about the flood control operation of the dam. Some analyses indicated that the operator did not free up storage space for the anticipated increasing inflow as they thought that the rainfall forecast information was not reliable. The main focus of the operator seemed to be to fill the reservoir to its full supply level for the purpose of maintaining water supply after years of drought in Queensland. This ended up being a suboptimal decision for retaining the water in the reservoir without preparing for the second peak flood inflow by assuming a no-rainfall scenario and neglecting forecasts of increasing rainfall by the Bureau of Meteorology. 68 Enhancing the Safety and Resilience of Dams in the Context of Climate Change According to Seqwater’s “Manual of Operational Procedures for Flood Mitigation” (2009), it operates a real-time flood monitoring and forecasting system that employs radio telemetry to collect, transmit, and receive rainfall and stream flow information in real time. Once received in the Flood Operations Center, the data are processed using a real-time flood model (RTFM) to estimate likely dam inflows and evaluate a range of possible inflow scenarios based on forecast and potential rainfall in the catchment. Seqwater is responsible for providing and maintaining the RTFM and for ensuring that sufficient data are available to allow proper operation of the RTFM during a flood event. The analyses indicated that a poorly drafted operations manual was blamed for causing confusion among the flood operators during the event. This incident points to the challenges of dam reservoir operation during floods as well as to the importance of peri- odically reassessing O&M manuals and the need for updates and revisions. MAP 8.3 Brisbane River Basin Source: Queensland Department of Energy and Water Supply 2014. Attribution 3.0 Australia (CC BY) license. Case Studies on the Application of Advanced Rainfall and Reservoir 69 PHOTO 8.6 Wivenhoe Dam Flood Discharge on January 11, 2011 Source: Dean Saffron, State Library of Queensland. Creative Commons Attribution 4.0 International (CC-BY 4.0) license. FIGURE 8.8 Inflow and Release of Water during the January 2011 Flood Event—Wivenhoe Dam Flow (m3/s) Elevation (m AHD) 13,000 79.0 12,000 78.0 11.000 77.0 10,000 76.0 9,000 75.0 74.0 8,000 73.0 7,000 72.0 6,000 71.0 5,000 70.0 4,000 69.0 3,000 68.0 2,000 67.0 1,000 66.0 0 0 11 1 1 11 1 1 1 1 11 01 01 01 01 01 01 20 20 20 /2 /2 /2 /2 /2 /2 1/ 1/ /1 1 1 /1 /1 /1 /1 2/ 4/ 6/ 8/ /1 10 14 16 12 18 Time Dam flow Dam outflow Headwater elevation Source: Van Den Honert and McAneney 2011. 70 Enhancing the Safety and Resilience of Dams in the Context of Climate Change Van Den Honert and McAneney (2011) collected information from the Bureau of Meteorology and the Queensland Floods Commission of Inquiry and summarized commentary on the oper- ational management of the Wivenhoe Dam, insurance, and land use planning issues. The anal- yses showed that the RTFM model generated two forecasts—that is, a with-forecast projection that incorporates the Bureau of Meteorology rainfall forecasts and a without-forecast projec- tion that assumes zero rainfall. The former is tracked in blue, and the latter is shown in red in figure 8.9. The with-forecast prediction suggests that the level of the lake would exceed 74.0 m amsl, the tipping point for the purposes of moving to a water release strategy where the primary consideration is the structural safety of the dam. This strategy, according to the manual, has no upper limit on the maximum release rate of water from the dam. However, the red line—that is, the scenario without-forecast projection, remains well below the 74.0 m threshold. In short, by not using all the information available to them, the dam operators seemed to make a subop- timal decision. It is noted that the Seqwater manual (2009) does not specify unequivocally that the Bureau of Meteorology rainfall forecasts need to be utilized, even though a range of possible inflow scenarios need to be evaluated based on forecasts and potential rainfall. It seems apparent that the second major flood event occurred on January 13, 2011, when dam releases from Wivenhoe Dam flooded the city of Brisbane and its surrounding areas. Figure 8.10 shows a hydrograph of the Brisbane River at the Brisbane City gauge between January 8–16. This suggests that the release of water from the Wivenhoe Dam was the principal cause of flood- ing along the mainstream and tributaries of the Brisbane River downstream of the dam over the period January 11–12, 2011. Again, although it was reported that the dam operators were acting in accordance with the operations manual for the dam, their modeling did not take account of forecast rainfall in determining the predicted dam water level, and this resulted in a suboptimal water release strategy. FIGURE 8.9 Modeled Wivenhoe Dam Lake Levels at 8 p.m. January 9, 2011 Elevation (m AHD) 76 75 74 73 72 71 70 69 68 67 66 1 1 1 1 1 1 01 01 01 01 01 01 /2 /2 /2 /2 /2 /2 1 1 /1 /1 /1 /1 6/ 8/ 10 14 16 12 Date and time Without forecast rain With forecast rain Time of run Source: Van Den Honert and McAneney 2011. Case Studies on the Application of Advanced Rainfall and Reservoir 71 FIGURE 8.10 Brisbane River Water Level at the City Gauge between January 8–16, 2011 4.5 Brisbane river at city gauge (mAHD) 4.0 Station number: 540198 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0 –0.5 00 00 00 00 00 00 00 00 00 00 0: 0: 0: 0: 0: 0: 0: 0: 0: 0: n n n n n n n n n n Ja Ja Ja Ja Ja Ja Ja Ja Ja Ja 8 09 10 11 12 3 4 5 16 17 s1 t1 t0 r1 e on ed n on Tu n Sa Fi u Su Sa Su Th W M M Source: Van Den Honert and McAneney 2011. A group of nearly 7,000 local flood victims brought a historic class action6 to court over the management of the Wivenhoe Dam, and in November 2019, the group won their case against Seqwater, Sunwater, and the Queensland government, with a judge finding that they were victims of negligence. The New South Wales (NSW) Supreme Court then ruled that the dam’s flood engineers relied too closely on “rain on the ground” estimates in 2011 while not appro- priately using forecast rainfall for the Wivenhoe Dam’s catchment area, as required by its man- ual. However, in September 2021, the flood victims lost their claims against the dam operator in the same NSW Supreme Court with the finding that “Applying the ... standard, the four flood engineers acted by way of consensus, ultimately following the strategy determined by the Senior Flood Operations Engineer. Failure by Seqwater’s flood engineers to depart from that strategy was not proven to be in breach of the ... standard. Even if their conduct departed from the Manual, that did not of itself entail a breach of that standard.7” Subsequently, in 2022, Australia’s High Court refused to hear the case against Seqwater, which closed the door for further action. Nonetheless, the flood victims still shared a AU$ 440 million settlement from the Queensland government and Sunwater—because Seqwater had not participated in the pre- vious settlement—which is half of what the victims could have received had the dam operator Seqwater also been found liable. NOTES 1. https://www.wis-jma.go.jp/cms/gsm/ 2. PREPBUFR stands for Prepared Binary Universal Form for the Representation of meteorological data. It is a specialized data format developed by the National Centers for Environmental Prediction, USA to prepare observational data for assimilation into weather models. 3. Preliminary discharge: Per the originally agreed protocol among the dam operators in charge of flood control and the private water users, this dam operation rule prescribes the specific method of predischarges such that the water level is lowered to a certain level in a prefixed manner prior to floods. Although the reservoir storage is usually filled for water use, the reservoir space is also allowed to be 72 Enhancing the Safety and Resilience of Dams in the Context of Climate Change emptied and used for flood control, and the joint rights are reflected in the original cost allocation and project agreement. Water users are thus not compensated for a reduced amount of water, even if the reservoir is not filled up after anticipated storms. 4. Predischarge: The river management authority and/or multipurpose dam operator decides the timing and the amount of water to be discharged based on flood forecasting information in the upstream and downstream areas prior to severe storm and flood events. The river management authority can request private dam owners and water users to release a certain amount of stored water in case of flood emergency, as per Article 52 of the River Act, seeking their understanding and cooperation. The Ministry of Land, Infrastructure, Transportation, and Tourism in April, 2020 issued a guideline to provide compensation in case of financial damage for water users by this drawdown operation. 5. “While both dams are jointly operated for flood control and water supply by Seqwater, this study is focused on the Wivenhoe Dam as its operation during the 2011 flood was highlighted as a primary cause for downstream inundation.” 6. https://www.brisbanetimes.com.au/national/queensland/queensland-2011-flood-victims-win-class​ -action​-20191129-p53fd9.html. 7. https://www.brisbanetimes.com.au/national/queensland/2011-qld-flood-victims-lose-final-appeal​ -against-dam​-operator-20220412-p5acw9.html. REFERENCES Aalbers, E. E., G. Lenderink, E. van Meijgaard, and B. J. J. M. van den Hurk. 2018. “Local-scale Changes in Mean and Heavy Precipitation in Western Europe, Climate Change or Internal Variability?” Climate Dynamics 50: 4745–66. doi:10.1007/s00382-017-3901-9. Alaya, M. A. B., F. Zwiers, and X. Zhang. 2018. “Probable Maximum Precipitation: Its Estimation and Uncertainty Quantification Using Bivariate Extreme Value Analysis.” Journal of Hydrometeorology 19 (4): 679–94. Alln, M. R., and W. J. Ingram. 2002. “Constraints on Future Changes in Climate and the Hydrologic Cycle.” Nature 419 (6903):  224–232. https://doi.org/10.1038/nature01092. Ban, N., C. Caillaud, E. Coppola, E. Pichelli, S. Sobolowski, M. Adinolfi, B. Ahrens, A. Alias, I. Anders, S. Bastin, D. Belušić, S. Berthou, E. Brisson, R. M. Cardoso, S. C. Chan, O. B. Christensen, J. Fernández, L. Fita, T. Frisius, G. Gašparac, F. Giorgi, K. Goergen, J. E. Haugen, Ø. Hodnebrog, S. Kartsios, E. Katragkou, E. J. Kendon, K. Keuler, A. Lavin-Gullon, G. Lenderink, D. Leutwyler, T. Lorenz, D. Maraun, P. Mercogliano, J. Milovac, H.-J. Panitz, M. Raffa, A. R. Remedio, C. Schär, P. M. M Soares, L. Srnec, B. M. Steensen, P. Stocchi, M. H. Tölle, H. Truhetz, J. Vergara-Temprado, H. de Vries, K. Warrach-Sagi, V. Wulfmeyer, and M. J. Zander. 2021. “The First Multi Model Ensemble of Regional Climate Simulations at Kilometer Scale Resolution, Part I: Evaluation of Precipitation.” Climate Dynamics 57: 275–302. doi:10.1007/s00382-021-05708-w. Bao, J., S. C. Sherwood, L. V. Alexander, and J. P. Evans. 2017. “Future Increases in Extreme Precipitation Exceed Observed Scaling Rates.” Nature Climate Change 7 (2): 128-132. https://doi.org/10.1038/NCLIMATE3201. Bates, P. D., M. S. Horritt, and T. J. Fewtrell. 2010. “A Simple Inertial Formulation of the Shallow Water Equations for Efficient Two-dimensional Flood Inundation Modelling.” Journal of Hydrology 387: 33–45. doi:10.1016/j​ .jhydrol.2010.03.027. Bertoli, L., D. Balzarolo, and E. Todini. 2022. “On the Benefits of Collaboration between Decision Makers and Scientists: The Case of Lake Como.” Hydrology 9 (11): 187. https://doi.org/10.3390/hydrology9110187. Chen, J., Kavvas, M.L., Ishida, K., Trinh, T., Ohara, N., Anderson, M.L. and Chen, Z.R. 2016. Role of snowmelt in determining whether the maximum precipitation always results in the maximum flood.  Journal of Hydrologic Engineering, 21(10), p.04016032. Chen, X., and F. Hossain. 2019. “Understanding Future Safety of Dams in a Changing Climate.” Bulletin of American Meteorological Society 100 (8): 1395–404. Clapeyron, M. C. 1834.  “Mémoire sur la Puissance Motrice de la Chaleur” [in French].  Journal de l’École Polytechnique 23: 153–90. Clausius, R. 1850. “On the Motive Power of Heat and the Laws Which Can Be Deduced There from Regarding the Theory of Heat” [in German]. Annalen der Physik 155 (4): 500–24. Delaney, C. J., R. K. Hartman, J. Mendoza, M. Dettinger, L. Delle Monache, J. Jasperse, F. M. Ralph, C. Talbot, J. Brown, D. Reynolds, and S. Evett. 2020. “Forecast Informed Reservoir Operations Using Ensemble Streamflow Predictions for a Multipurpose Reservoir in Northern California.” Water Resources Research 56 (9), e2019WR026604. doi:10.1029/2019WR026604. De Troch, R., R. Hamdi, H. V. De Vyver, J.-F. Geleyn, and P. Termonia. 2013. “Multiscale Performance of the ALARO-0 Model for Simulating Extreme Summer Precipitation Climatology in Belgium.” Journal of Climate 15: 8895–15. doi:10.1175/JCLI-D-12-00844.1. Doblas-Reyes, F. J., A. A. Sorensson, M. Almazroui, A. Dosio, W. J. Gutowski, R. Haarsma, R. Hamdi, B. Hewitson, W.-T. Kwon, B. L. Lamptey, D. Maraun, T. S. Stephenson, I. Takayabu, L. Terray, A. Turner, Image and Z. Zuo 2021.”  Linking global to regional climate change.”  In:  Masson-Delmotte, V.,  Zhai, P.,  Pirani, A.,  Connors, S. L.,  Pean, C.,  Berger, S.,  Caud, N.,  Chen, Y.,  Goldfarb, L.,  Gomis, M. I.,  Huang, M.,  Leitzell, K.,  Lonnoy, E., Matthews, J. B. R., Maycock, T. K., Waterfield, T., Yelekci, O., Yu, R. and Zhou, B. (eds.) Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press. Ehmele, F., L.-A. Kautz, H. Feldmann, and J. G. Pinto. 2020. “Long-term Variance of Heavy Precipitation Across Central Europe Using a Large Ensemble of Regional Climate Model Simulations.” Earth System Dynamics 11: 469–90. doi:10.5194/esd-11-469-2020. Fluixá-Sanmartín, J., L. Altarejos-García, A. Morales-Torres, and I. Escuder-Bueno. 2018. “Review Article: Climate Change Impacts on Dam Safety.” Natural Hazards and Earth System Sciences 18 (9): 2471–88. Copernicus GmbH. Francou, J. and J. A. Rodier. 1967. Essai de classification des Crues Maximales Observées dans le Monde. In: Cahier ORSTOM, séries Hydrologie, 4(3), 9-46” 73 74 Enhancing the Safety and Resilience of Dams in the Context of Climate Change Fujita, M., R. Mizuta, M. Ishii, H. Endo, T. Sato, Y. Okada, S. Kawazoe, S. Sugimoto, K. Ishihara, and S. Watanabe. 2019. “Precipitation Changes in a Climate with 2-K Surface Warming from Large Ensemble Simulations Using 60-km Global and 20-km Regional Atmospheric Models.” Geophysical Research Letters 46 (1): 435–42. doi:10.1029/2018GL079885. Fyfe, J. C., C. Derksen, L. Mudryk, G. M. Flato, B. D. Santer, N. C. Swart, N. P. Molotch, X. Zhang, H. Wan, V. K. Arora, J. Scinocca, and Y. Jiao. 2017. “Large Near-term Projected Snowpack Loss over the Western United States.” Nature Communications 8: 14996. doi:10.1038/ncomms14996. Giorgi, F., and W. J. Gutowski. 2015. “Regional Dynamical Downscaling and the CORDEX Initiative.” Annual Review of Environment and Resources 40: 467–490. doi:10.1146/annurev-environ-102014-021217. Giorgi, F., C. Jones, and G. R. Asrar. 2009. “Addressing Climate Information Needs at the Regional Level: The CORDEX Framework.” Bulletin 58, World Health Organization (WHO), Trieste, Italy. Grijsen, J. G. 2014. “Understanding the Impact of Climate Change on Hydropower: The Case of Cameroon— Climate Risk Assessment for Hydropower Generation in Cameroon.” World Bank, Washington, DC. Grijsen, J. G. 2022. “Rapid Climate Risk and Vulnerability Assessment Methodology for Hydropower Project.” In The Handbook of Behavioral Economics and Climate Change, edited by S. N. Seo. Edward Elgar Publishing Ltd. London, UK Gutowski, J. W., F. Giorgi, B. Timbal, A. Frigon, D. Jacob, H.-S. Kang, K. Raghavan, B. Lee, C. Lennard, G. Nikulin, E. O’Rourke, M. Rixen, S. Solman, T. Stephenson, and F. Tangang. 2016. “WCRP Coordinated Regional Downscaling Experiment (CORDEX): A Diagnostic MIP for CMIP6.” Geoscientific Model Development 9  (11): 4087–95. doi:10.5194/gmd-9-4087-2016. Hagemann, S., C. Chen, D. B. Clark, S. Folwell, S. N. Gosling, I. Haddeland, N. Hanasaki, J. Heinke, F. Ludwig, F. Voß, and A. J. Wiltshire. 2012. Climate change impact on available water resources obtained using multiple global climate and hydrology models. Earth System Dynamics Discussions, 3(2). Hershfield, D. M. 1965. “Method for Estimating Probable Maximum Rainfall.” Journal of the American Water Works Association 57 (8): 965–72. Hultstrand, D. M., and B. D. Kappel. 2020. “A Methodology to Evaluate Probable Maximum Precipitation (PMP) under Changing Climate Conditions.” In Association of State Dam Safety Officials Annual Conference, Dam Safety 2020, 2020. 18-22 September 2022, Baltimore, Maryland, USA. Hunter, N. M., P. D. Bates, M. S. Horritt, and M. Wilson. 2007. “Simple Spatially-distributed Models for Predicting Flood Inundation: A Review.” Geomorphology 90: 208–25. doi:10.1016/j.geomorph.2006.10.021. ICHARM (International Centre for Water Hazard and Risk Management). 2022. Climate Change Impact Assessment on Extreme Hydrological Events and Resilience Enhancement for Dam Safety. Tsukuba, Japan: ICHARM. ICOLD (International Commission on Large Dams). 2016. “Multipurpose Water Storage—Essential Elements and Emerging Trends.” Bulletin 171, ICOLD, Paris. ICOLD. 2018. “Flood Evaluation and Dam Safety.” Bulletin 170, ICOLD Chaou, France. ICOLD. 2019. “Flood Evaluation, Hazard Determination and Risk Management.” Bulletin 187, ICOLD. ICOLD. 2022. “Climate Change, Dams and Reservoirs. General Report, Q107.” 27th Congress on Large Dams, Chaou, France, 27 May – 03 June, 2022. IHA (International Hydropower Association). 2019. “Hydropower Sector Climate Resilience Guide.” IHA, London. Iizumi, T., M. Nishimori, Y. Ishigooka, and M. Yokozawa. 2010. “Introduction to Climate Change Scenario Derived by Statistical Downscaling.” Journal of Agricultural Meteorology 66: 131–43. Innocenti, S., A. Mailhot, M. Leduc, A. J. Cannon, and A. Frigon. 2019. “Projected Changes in the Probability Distributions, Seasonality, and Spatiotemporal Scaling of Daily and Sub-daily Extreme Precipitation Simulated by a 50-member Ensemble over Northeastern North America.” Journal of Geophysical Research: Atmospheres 124 (10): 427–49. IPCC, 2018: Global Warming of 1.5°C.An IPCC Special Report on the impacts of global warming of 1.5°C above pre-industrial levels and related global greenhouse gas emission pathways, in the context of strengthening the global response to the threat of climate change, sustainable development, and efforts to eradicate poverty [Masson-Delmotte, V., P. Zhai, H.-O. Pörtner, D. Roberts, J. Skea, P.R. Shukla, A. Pirani, W. Moufouma-Okia, C. Péan, R. Pidcock, S. Connors, J.B.R. Matthews, Y. Chen, X. Zhou, M.I. Gomis, E. Lonnoy, T. Maycock, M. Tignor, and T. Waterfield (eds.)]. Cambridge University Press, Cambridge, UK and New York, NY, USA, 616 pp. https://doi.org/ 10.1017/9781009157940. IPCC (Intergovernmental Panel on Climate Change). 2023. Sixth Assessment Report. Cambridge, UK: Cambridge University Press. Ishida, K., M. L. Kavvas, S. Jang, Z. Q. Chen, N. Ohara, and M. L. Anderson. 2015. “Physically Based Estimation of Maximum Precipitation over Three Watersheds in Northern California: Relative Humidity Maximization Method.” Journal of Hydrologic Engineering 20(10), 04015014. References 75 Ishida, K., N. Ohara, M. L. Kavvas, Z. Q. Chen, and M. L. Anderson. 2018. “Impact of Air Temperature on Physically-based Maximum Precipitation Estimation through Change in Moisture Holding Capacity of Air.” Journal of Hydrology 556: 1050–63. Ishii, M., and N. Mori. 2020. “d4PDF: Large-ensemble and High-resolution Climate Model Simulations for Global Warming Risk Assessment.” Progress in Earth and Planetary Science (2020) 69(8): 413-429. Jacob, D., Teichmann, C., Sobolowski, S., Katragkou, E., Anders, I., Belda, M., ... & Wulfmeyer, V. 2020. Regional climate downscaling over Europe: perspectives from the EURO-CORDEX community. Regional environmental change, 20, 1-20. https://link.springer.com/article/10.1007/s10113-020-01606-9. Jasperse, J., F. M. Ralph, M. Anderson, L. Brekke, N. Malasavage, M. D. Dettinger, J. Forbis, J. Fuller, C. Talbot, R. Webb, and A. Haynes. 2020. Lake Mendocino Forecast Informed Reservoir Operations Final Viability Assessment. UC San Diego. Retrieved from https://escholarship.org/uc/item/3b63q04n Jenkins, G. J., Murphy, J. M., Sexton, D. M. H., Lowe, J. A., Jones, P. and Kilsby, C. G. 2009. UK Climate Projections: Briefing report. Met Office Hadley Centre, Exeter, UK. Johnson, K. A., and J. C. Smithers. 2020. “Updating the Estimation of 1-day Probable Maximum Precipitation in South Africa.” Journal of Hydrology Regional Studies 32, 100736. JWA (Japan Water Agency). 2023. “Technical Support for Optimized Reservoir Operations with Advanced Rainfall and Inflow Prediction Based on the JWA’s Operation System and Experiences” (unpublished). JWA, Chuo-ku, Saitama City, Japan. Kay, J. E., C. Deser, A. Phillips, A. Mai, C. Hannay, G. Strand, J. M. Arblaster, S. C. Bates, G. Danabasoglu, J.  Edwards, M. Holland, P. Kushner, J.-F. Lamarque, D. Lawrence, K. Lindsay, A. Middleton, E. Munoz, R.  Neale, K. Oleson, L. Polvani, and M. Vertenstein. 2015. “The Community Earth System Model (CESM) Large Ensemble Project: A Community Resource for Studying Climate Change in the Presence of Internal Climate Variability: Review and Evaluation of Different Methods.” Bulletin of the American Meteorological Society 96: 1333–49. doi:10.1175/BAMS-D-13-00255.1. Kim, Y., W. Yu, S. Oh, and K. Jung. 2016. “Development of Basin-scale PMP Estimation Method Using Grid-based Rain Search Method.” Procedia Engineering 154: 1243–50. Klemeš, V. 2000. “Tall Tales about Tails of Hydrological Distributions.” Journal of Hydrologic Engineering 1: 227–31. https://doi.org/10.1061/(ASCE)1084-0699(2000)5:3(232). Koike, T. 2021. “Evolution of Japan’s Flood Control Planning and Policy in Response to Climate Change Risks and Social Changes.” Water Policy 23 (S1): 77–84. doi:10.2166/wp.2021.287. Koike, T., S. Nakamura, C. T. Nyunt, T. Ushiyama, M. Rasmy, K. Takakawa, H. Ito, K. Ikeuchi, E. Ikoma, and M. Kitsuregawa. 2021. “An Operation Support System for Hydroelectric Dams for Improving Flood Control and Power Generation” [in Japanese]. Journal of Japan Society of Civil Engineers, Series B1 (Hydraulic Engineering), Vol. 77, Issue 2, pp. I_79- I_84, https://doi.org/10.2208/jscejhe.77.2_I_79. Koutsoyiannis, D. 2012. “Clausius-Clapeyron Equation and Saturation Vapour Pressure: Simple Theory Reconciled with Practice.” European Journal of Physics 33 (2): 295–305. https://doi.org/10.1088/0143-0807/33/2/295. Kunkel, K. E., Karl, T. R., Easterling, D. R., Redmond, K., Young, J., Yin, X. and Hennon, P., 2013. Probable maximum precipitation and climate change. Geophysical Research Letters, 40(7): 1402-1408. Kunkel, K. E., R. J. Novak, and P. M. Steurer. 1999. “Diagnosis of a Long-term, Non-stationary, Stochastic Storm Transposition Model.” Journal of Hydrology 219 (3–4): 169–87. Leduc, M., A. Mailhot, A. Frigon, J.-L. Martel, R. Ludwig, G. B. Brietzke, M. Giguère, F. Brissette, R. Turcotte, M. Braun, and J. Scinocca. 2019. “The ClimEx Project: A 50-member Ensemble of Climate Change Projections at 12-km Resolution over Europe and Northeastern North America with the Canadian Regional Climate Model (CRCM5).” Journal of Applied Meteorology and Climatology 58 (4): 663–93. doi:10.1175/JAMC-D-18-0021.1. Lovejoy, S. and D. Schertzer. 1995. Multifractals and Rain, New Uncertainty Concepts in Hydrology and Water Resources, Z. W. Kunzewicz, University Press, Cambridge, 61-103. Madakumbura, G. D., H. Kim, N. Utsumi, H. Shiogama, E. Fischer, Ø. Seland, J. F. Scinocca, D. M. Mitchell, Y. Hirabayashi, and T. Oki. 2019. “Event-to-event Intensification of the Hydrologic Cycle from 1.5°C to a 2°C Warmer World.” Scientific Reports 9 (3483): 2045–322. Maher, N., Milinski, S., Suarez‐Gutierrez, L., Botzet, M., Dobrynin, M., Kornblueh, L., ... & Marotzke, J.. 2019. “The Max Plank Institute Grand Ensemble: Enabling the Exploration of Climate System Variability.” Journal of Advances in Modeling Earth Systems 11(7), 2050-2069. Mehrotra, R. and Sharma. 2021. “A Robust Alternative for Correcting Systematic Biases in Multi-variable Climate Model Simulations.” Environmental Modeling Software RoMBC 139-105019. https://doi.org/10.1016/j​ .envsoft.2021.105019. MEXT (Ministry of Education, Culture, Sports, Science and Technology of Japan). 2019. Program for Risk Information on Climate Change (SOUSEI). Database for Policy Decision Making for Future Climate Change (d4PDF). Available at: http://www.miroc-gcm.jp/-pub/d4PDF/. MGS Engineering Consultants. 2009. “General Stochastic Event Flood Model (SEFM), Technical Support Model.” U.S. Department of Interior, Bureau of Reclamation, Hydrology Group, Washington, DC. 76 Enhancing the Safety and Resilience of Dams in the Context of Climate Change Micovic, Z., M. G. Schaefer, and G. H. Taylor. 2015. “Uncertainty Analysis for Probable Maximum Precipitation Estimates.” Journal of Hydrology 521: 360–73. Mizuta, R., A. Murata, M. Ishii, H. Shiogama, K. Hibino, N. Mori, O. Arakawa, Y. Imada, K. Yoshida, and T. Aoyagi. 2017. “Over 5,000 Years of Ensemble Future Climate Simulations by 60-km Global and 20-km Regional Atmospheric Models.” Bulletin of American Meteorological Society 98 (7): 1383–98. MLIT. 2021. “Revision of the Suggestion for “Appropriate Flood Control Planning in light of Climate Change” [in Japanese]. Press Release, April 30, 2021. Available at: https://www.mlit.go.jp/report/press /mizukokudo03 _hh_001060.html. Nakamura, S., T. Koike, C. T. Nyunt, T. Ushiyama, M. Rasmy, K. Takakawa, H. Ito, K. Ikeuchi, E. Ikoma, and M. Kitsuregawa. 2022. “An Operation Support System for Hydroelectric Dams—Long and Short Term Ensemble Predictions and Their Applications” [in Japanese]. Journal of Japan Society of Civil Engineers B1 78(2), I_1195-I_1200. Naseer, A., T. Koike, M. Rasmy, T. Ushiyama, and M. Shrestha. 2019. “Distributed Hydrological Modeling Framework for Quantitative and Spatial Bias Correction for Rainfall, Snowfall, and Mixed-Phase Precipitation Using Vertical Profile of Temperature.” Journal of Geophysical Research: Atmospheres 124(9), 4985-5009. NOAA (National Oceanic and Atmospheric Administration). 2017. “Improving Reliability for Droughts and Floods: Forecast-Informed Reservoir Operations (FIRO).” National Oceanic and Atmospheric Administration, Washington, DC, United States of America. Nosaka, M., M. Ishii, H. Shiogama, R. Mizuta, A. Murata, H. Kawase, and H. Sasaki. 2020. “Scalability of Future Climate Changes for +1.5K, +2K, and +4K Global Warming in NHRCM Large Ensemble Simulations.” Progress in Earth and Planetary Science 7: 27. doi:10.1186/s40645-020-00341-3. O’Gorman, P. A., and C. J. Muller. 2010. “How Closely Do Changes in Surface and Column Water Vapor Follow Clausius-Clapeyron Scaling in Climate Change Simulations?” Environmental Research Letters 5 (2): 025207. https://doi.org/10.1088/1748-9326/5/2/025207. Ohara, N., M. L. Kavvas, S. Kure, Z. Q. Chen, S. Jang, and E. Tan. 2011. “Physically Based Estimation of Maximum Precipitation over American River Watershed, California.” Journal of Hydrologic Engineering 16(4), 351-361. Ohara, N., Kavvas, M.L., Kure, S., Chen, Z.Q., Jang, S. and Tan, E., 2012, September. Maximization of precipitation for the 1997 storm event over american river watershed, California, using a regional atmospheric model. In World Environmental and Water Resources Congress 2012: Crossing Boundaries (pp. 1909-1916). Okoli, K. 2019. “Design Flood Estimation under Uncertainty.” Dissertation, Uppsala University. http://www.diva​ -portal.org/smash/get/diva2:1341494/FULLTEXT01.pdf. O’Neill, B. C., C. Tebaldi, D. P. van Vuuren, V. Eyring, P. Friedlingstein, G. Hurtt, R. Knutti, E. Kriegler, J.-F. Lamarque, J. Lowe, G. A. Meehl, R. Moss, K. Riahi, and B. M. Sanderson. 2016. “The Scenario Model Intercomparison Project (ScenarioMIP) for CMIP6.” Geoscientific Model Development 9: 3461–82. doi:10.5194​ /­gmd-9-3461-2016. Papalexiou, S.M. and Koutsoyiannis, D., 2006. A probabilistic approach to the concept of probable maximum precipitation. Advances in Geosciences, 7, 51-54. Paquet, E., F. Garavaglia, J. Gailhard, and R. Garcon. 2013. “The SCHADEX Method: A Semi-continuous Rainfall- runoff Simulation for Extreme Flood Estimation.” Journal of Hydrology 495: 23–37. Pichelli, E., E. Coppola, S. Sobolowski, N. Ban, F. Giorgi, P. Stocchi, A. Alias, D. Belušić, S. Berthou, C. Caillaud, R. M. Cardoso, S. Chan, O. B. Christensen, A. Dobler, H. de Vries, K. Goergen, E. J. Kendon, K. Keuler, G. Lenderink, T. Lorenz, A. N. Mishra, H.-J. Panitz, C. Schär, P. M. M. Soares, H. Truhetz, and J. Vergara- Temprado. 2021. “The First Multi-model Ensemble of Regional Climate Simulations at Kilometer-Scale Resolution Part 2: Historical and Future Simulations of Precipitation.” Climate Dynamics 56: 3581–602. doi:10.1007/s00382-021-05657-4. Queensland Department of Energy and Water Supply, 2014. “Wivenhoe and Somerset Dams Optimisation Study” Report. State of Queensland, 2014. https://damfailures.org/sites/default/files/wp-pdf/OR_Wivenhoe-and​ Somerset-Dams-Optimisation-Study.pdf Rasmy, M., T. Sayama, and T. Koike. 2019. “Development of Water and Energy Budget-based Rainfall-Runoff- Inundation Model (WEB-RRI) and Its Verification in the Kalu and Mundeni River Basins, Sri Lanka.” Journal of Hydrology 579, 124163. Ray, P. A., and C. M. Brown. 2015. “Confronting Climate Uncertainty in Water Resources Planning and Project Design: The Decision Tree Framework.” World Bank, Washington, DC. Richardson, C. W. 1981. “Stochastic Simulation of Daily Precipitation, Temperature and Solar Radiation.” Water Resources Research 17 (1): 182–90.  Rogger, M., H. Pirkl, A. Viglione, J. Komma, B. Kohl, R. Kirnbauer, R. Merz, and G. Blöschl. 2012. “Step Changes in the Flood Frequency Curve: Process Controls.” Water Resources Research 48 (5): 1–15. doi:10.1029/2011WR011187. Rouhani, H., & Leconte, R. 2016. A novel method to estimate the maximization ratio of the Probable Maximum Precipitation (PMP) using regional climate model output. Water Resources Research, 52(9), 7347-7365. References 77 Salas, J. D., M. L. Anderson, S. M. Papalexiou, and F. Frances. 2020. “PMP and Climate Variability and Change: A Review.” Journal of Hydrologic Engineering 25(12), 03120002). Salas, J. D., G. Gavilán, F. R. Salas, P. Y. Julien, and J. Abdullah. 2014. “Uncertainty of the PMP and PMF.” In  Handbook of Engineering Hydrology: Modeling, Climate Change, and Variability Eslamian, S. (Ed) CRC Press. 2, 575-603. Salas, J. D., J. Obeysekera, and R. M. Vogel. 2018. “Techniques for Assessing Water Infrastructure for Non- stationary Extreme Events: A Review.” Hydrological Sciences Journal 63 (3)  325-352. https://doi.org/10.1080​ /02626667.2018.1426858. Sankarasubramanian, A., R. M. Vogel, and J. F. Limbrunner. 2001. “Climate Elasticity of Streamflow in the United States.” Water Resources Research 37 (6): 1771–81. Sarkar, S., and R. Maity. 2020. “Increase in Probable Maximum Precipitation in a Changing Climate over India.” Journal of Hydrology 585: 124806. doi:10.1016/j.jhydrol.2020.124806. https://www.sciencedirect.com/science​ /article/abs/pii/S0022169420302663?via%3Dihub. Sayama, T., G. Ozawa, T. Kawakami, S. Nabesaka, and K. Fukami. 2012. “Rainfall-runoff-inundation Analysis of Pakistan Flood 2010 at the Kabul River Basin.” Hydrological Sciences Journal 57 (2): 298–312. Schär, C., C. Frei, D. Lüthi, and H. C. Davies. 1996. “Surrogate Climate-change Scenarios for Regional Climate Models.” Geophysical Research Letters 23 (6): 669–72. Schewe, J., Heinke, J., Gerten, D., Haddeland, I., Arnell, N. W., Clark, D. B., ... & Kabat, P. 2014. Multimodel assessment of water scarcity under climate change. Proceedings of the National Academy of Sciences, 111(9), 3245-3250. https://www.pnas.org/content/pnas/111/9/3245.full.pdf Selvarajah, H., T. Koike, M. Rasmy, K. Tamakawa, A. Yamamoto, M. Kitsuregawa, and L. Zhou. 2021. “Development of an Integrated Approach for the Assessment of Climate Change Impacts on the Hydro- meteorological Characteristics of the Mahaweli River Basin, Sri Lanka.” Water 13: 1218. Seqwater. 2009. “Manual of Operational Procedures for Flood Mitigation.” Seqwater, Queensland, Australia. https://damfailures.org/wp-content/uploads/2022/03/OR_Rev7-Manual-of-Operational-Procedures-for​ -Flood-Mitigation.pdf. Shrestha, M., L. Wang, T. Koike, H. Tsutsui, Y. Xue, and Y. Hirabayashi. 2014. “Correcting Basin-scale Snowfall in a Mountainous Basin Using a Distributed Snowmelt Model and Remote-sensing Data.” Hydrology and Earth System Sciences 18: 747–61. https://doi.org/10.5194/hess-18-747-2014. Sieck, K., C. Nam, L. M. Bouwer, D. Rechid, and D. Jacob. 2021. “Weather Extremes over Europe under 1.5 and  2.0°C Global Warming from HAPPI Regional Climate Ensemble Simulations.” Earth System Dynamics 12: 457–468. doi:10.5194/esd-12-457-2021. Sørland, S., R. Brogli, P. K. Pothapakula, E. Russo, J. Van de Walle, B. Ahrens, I. Anders, E. Bucchignani, E. L. Davin, M. E. Demory, A. Dosio, H. Feldmann, B. Früh, B. Geyer, K. Keuler, D. Lee, D. Li, N. P. M. van Lipzig, S. K. Min, H. J. Panitz, B. Rockel, C. Schär, C. Steger, and W. Thiery. 2021. “COSMO-CLM Regional Climate Simulations in the CORDEX Framework: A Review.” Geoscientific Model Development 14: 5125–54. Steinschneider, S., and C. Brown. 2013. “A Semi-parametric Multivariate, Multisite Weather Generator with Low-frequency Variability for Use in Climate Risk Assessments.” Water Resources Research 49: 7205–220. https:// agupubs.onlinelibrary.wiley.com/doi/epdf/10.1002/wrcr.20528. Tachikawa, Y., K. Miyawaki, T. Tanaka, K. Yorozu, M. Kato, Y. Ichikawa, and S. Kim. 2017. “Future Change Analysis of Extreme Floods Using Large Ensemble Climate Simulation Data” [in Japanese]. Journal of Japan Society of Civil Engineers, Series B1 (Hydraulic Engineering) 73 (3): 77–90. doi:10.2208/jscejhe.73.77. Tangang, F., J. X. Chung, L. Juneng, Supari, E. Salimun, S. T. Ngai, A. F. Jamaluddin, M. S. F. Mohd, F.  Cruz, G.  Narisma, J. Santisirisomboon, T. Ngo-Duc, P. Van Tan, P. Singhruck, D. Gunawan, E. Aldrian, A. Sopaheluwakan, N. Grigory, A. R. C. Remedio, D. V. Sein, D. Hein-Griggs, J. L. McGregor, H. Yang, H. Sasaki, and P. Kumar. 2020. “Projected Future Changes in Rainfall in Southeast Asia Based on CORDEX-SEA Multi Model Simulations.” Climate Dynamics 55: 1247–67. doi:10.1007/s00382-020-05322-2. Teutschbein, C., and J. Seibert. 2012. “Bias Correction of Regional Model Simulations for Hydrological Climate- change Impact Studies: Review and Evaluation of Different Methods.” Journal of Hydrology 456–457, 12-29. doi:10.1016/j.jhydrol.2012.05.052. Thanh Thuy, L. T., S. Kawagoe S., and R. Sarukkalige R. 2019. “Estimation of Probable Maximum Precipitation at Three Provinces in Northeast Vietnam Using Historical Data and Future Climate Change Scenarios.” Journal of Hydrology Regional Studies, Vol. 23. Thuyloi. 2022. Viet Nam—New Zealand Dam Safety Project. Retrieved from http://en.tlu.edu.vn​ /cooperationinternational-projects/viet-nam--new-zealand-dam-safety-project-10906#. Todini, E. 1999. “Using Phase-space Modelling for Inferring Forecasting Uncertainty in Non-linear Stochastic Decision Schemes.” Journal of Hydroinformatics 1 (2): 75–82. Todini, E. 2018. “Paradigmatic Changes Required in Water Resources Management to Benefit from Probabilistic Forecasts.” Elsevier Water Security 3: 9–17. doi:10.1016/j.wasec.2018.08.001. 78 Enhancing the Safety and Resilience of Dams in the Context of Climate Change Van den Honert, R., and J. McAneney. 2011. “The 2011 Brisbane Floods: Causes, Impacts and Implications.” Water 3 (4): 1149–1173. mdpi.com/2073-4441/3/4/1149. Vano, J., T. Das, and D. Lettenmaier. 2012. “Hydrologic Sensitivities of Colorado River Runoff to Changes in Precipitation and Temperature.” Journal of Hydrometeorology, 13 (3): 932–49. Van Vliet, M. T. H., L. P. H. Van Beek, S. Eisner, M. Flörke, Y. Wada, and M. F. P. Bierkens. 2016. “Multi-model Assessment of Global Hydropower and Cooling Water Discharge Potential under Climate Change.” Global Environmental Change 40: 156–70. Veale, B., P. Amos, N. C. Thai, and M. Wishart. 2022. “Assessment of Climate Change Impacts to a Large Dam in Vietnam.” ANCOLD and NZSOLD Conference Sydney, Australia, 26-28 October 2022. Retrieved from https:// ancold.org.au/product/2022-assessment-of-climate-change-impacts-to-a-large-dam-in-vietnam/. Visser, J. B., S. Kim, C. Wasko, R. Nathan, and A. Sharma. 2022. “The Impact of Climate Change on Operational Probable Maximum Precipitation Estimates.” Water Resources Research 581-18. https://doi​ .org/10.1029/2022WR032247. Wikipedia. 2024. Hatanagi-I Hydroelectric Dam. Last modified March 14, 2024. https://en.wikipedia.org/wiki​ /Hatanagi-I_Dam. Wilby, R.L., Charles, S.P., Zorita, E., Timbal, B., Whetton, P. and Mearns, L.O., 2004. Guidelines for use of climate scenarios developed from statistical downscaling methods. Supporting material of the Intergovernmental Panel on Climate Change, available from the DDC of IPCC TGCIA, 27. Wilson, M., P. Bates, D. Alsdorf, B. Forsberg, M. Horritt, J. Melack, F. Frappart, and J. Famiglietti. 2007. “Modeling Large-scale Inundation of Amazonian Seasonally Flooded Wetlands.” Geophysical Research Letters 34 (15): 1-6 (L15404). doi:10.1029/2007GL030156. WMO (World Meteorological Organization). 1969. “Estimation of Maximum Floods.” WMO-No. 233, TP 126, Technical Note No. 98, WMO, Geneva. WMO. 2009. Manual on Estimation of Probable Maximum Precipitation (PMP). WMO-No. 1045, WMO, Geneva. World Bank. 2025. “Enhancing the Safety and Resilience of Dams in the Context of Climate Change and Extreme Hydrological Events.” World Bank, Washington, DC. Wyser, K., T. Koenigk, U. Fladrich, R. Fuentes-Franco, M. P. Karami, and T. Kruschke. 2021. “The SMHI Large Ensemble (SMHI-LENS) with EC-Earth3.3.1.” Geoscientific Model Development 1: 4781–96. Yin, J., P. Gentine, S. Zhou, S. C. Sullivan, R. Wang, Y. Zhang, and S. Guo. 2018. “Large Increase in Global Storm Runoff Extremes Driven by Climate and Anthropogenic Changes.” Nature Communications 9 (1): 4389. SKU W24026