Motivation Climate is a major determinant of risk Effective water resources management conditioned on: historical data (stream flow, evaporation) current state (reservoir levels, river levels, demands) future predictions (days, seasons, decades, climate change) Predictions allow for advanced warning of extreme conditions (droughts and floods) improved decision-making: prepare, not just react reducing (exploiting) climate-related risks (opportunities)
Motivation Can we leverage applications of climate science to reduce (exploit) negative (positive) impacts of climate variability?
Integrating Management of Climate Risks An operational approach: 1. Identify hazards associated with climate risks (of all time scales) to the water system 2. Characterize the climate risks 3. Propose and Assess portfolio of solutions/adaptations to key climate (plus other) risks
Integrating Management of Climate Risks 1. Identify hazards associated with climate risk to the water system What are the key climate challenges that the system faces now (e.g., frequent drought, flood events, variable flows) What damages occur as functions of these events? Where are the impacts felt? Are there distributional effects? Is the environment considered/protected? Are there opportunity losses due to risk aversion associated with current climate risks?
Integrating Management of Climate Risks 2. Characterize hydroclimatic risk What are the probabilities, recurrence periods, etc. of hazard causing events Is there spatial or temporal structure? Are there probable/predictable changes expected? What are the most plausible future scenarios and the uncertainty associated with them? How do these risks compare to the social, economic, demographic and environmental challenges the water system faces (severity, uncertainty)?
Integrating Management of Climate Risks 3. Propose/Assess portfolio of solutions/adaptations to Climate Risks Incorporate uncertainty of climate futures in the decision process May favor flexibility over structure (soft vs hard approaches) Solutions have spatial and temporal characteristics that modulate appropriateness based on the climate risks Infrastructure Economic instruments: water banks, options, contracts Seasonal forecasts Flexible operating rules Insurance
Goals & Objectives of Water Systems
Goals & Objectives of Water Systems Water Systems include: Natural hydrologic features Infrastructure to divert, store, distribute water Social and institutional frameworks Purposes of large-scale systems: Water supply Flood protection Hydropower generation Fishing and aquaculture Waste assimilation Ecological flows Navigation Recreational activities
Water Use Hydropower Others Urban Irriga8on DEMAND Transporta8on Infrastructure Ecosystem Water Nature Climate SUPPLY
Managing Water Resource Systems Balance Water Supply and Demand Historical rules for resource allocation How much, and when should these rules be modified? How do we assess and communicate potential impacts of action & inaction? Water Energy Climate Health Agriculture Dam 1 Irrigated Farms Electric Grid Dam 2 Irrigated Farms Well Field Dam 3 Human Activity New City Muddy River
Performance Criteria Hydrologic - Economic Benefit Maximization (including uncertainty) Economic Efficiency: Pareto / Kaldor-Hinks Duration Severity Magnitude Reliability Resilience Vulnerability System Cost, Hydrologic Loss (evaporation, spill ) Justice Utilitarian, Equity, Others Objective Performance Economic Environmental Social Political Justice.
Hydrologic Analysis Variability in space and time of precipitation Source: FAO/GIEWS website
Characterizing Streamflow Natural vs. Regulated Short-term events [day(s)] Seasonal patterns Interannual variability Persistent periods Multi-decadal signal
Streamflow Prediction Historical Analogs Simplest approach Take sequences of weather observed during past months as possible scenarios for a predicted season. Advantage: multivariate structure (both spatial and between-variables) is preserved Disadvantage: may not be many sequences in the historical record and every event is different (sampling problems)
Streamflow Prediction Hydrologic Persistence Use of antecedent (current) conditions to forecast SF t+1 follows SF t. SF t+1 follows SF t Relationship between July and August flows on the Chagres River, Panama. Linear regression illustrates one relationship for forecasting.
Streamflow Prediction Beyond analogs and hydrologic persistence Seasonal/interannual predictability comes from factors that exert a continuous influence over a period of time that includes many sequences of weather events. Such factors are: Sea surface temperatures (SST, ENSO, others) Land surface conditions (soil moisture, vegetation) Radiative variations (volcanoes, greenhouse gases) Intraseasonal processes (MJO)
Streamflow Prediction Sea Surface Temperatures Historical Inflow Observations Photo: MWSS Global Climate Model RAINFALL Cross Validated Model WINDS Sta1s1cal Model Forecast Inflow for OND 2002 Source: B. Lyon
Prediction Systems Global Circulation Model Downscaling Regional Climate Model: Dynamic or Stochastic Regional Climate Predictor Stochastic Model Hydrologic Model lump Distributed Reservoir Inflow Reservoir Operation System Demand Model Agriculture/Urban Economic Models
Downscaling Downscaling from GCMs use Caution! Credit: Chong Yu Xu, Uppsala University; WRM 13:369 382, 1999
Prediction Systems Statistical vs Dynamical (Applicable to D/S and Hydro models) Statistical ------- Dynamical (Physical) ADVANTAGES Based on actual, real-world observed data. Knowledge of physical processes not needed. Many climate relationships quasi-linear, quasi-gaussian ------------------------------------ Uses proven laws of physics. Quality observational data not required (but needed for validation). Can handle cases that have never occurred. DISADVANTAGES Depends on quality and length of observed data Does not fully account for climate change, or new climate situations. ------------------------------ Some physical laws must be abbreviated or statistically estimated, leading to errors and biases. Computer intensive.
Prediction Systems Conceptual representation of S/I streamflow forecasting problem SST t Atmos t Precip t Land Hydrology t Flow t SST t+t Atmos t+t Precip t+t Land Hydrology t+t Are the intermediate variables needed in a statistical approach? Flow t+t Is information lost or gained errors accumulate if we step through?
Prediction Systems Bias Correction? Model and Parameter Uncertainty?
Variable Prediction Linear model fit by least squares Nearest neighbor
Variable Prediction Statistical model for the joint distribution Pr(X,Y) Y=f(X)+ε f(x)=e(y X=x) Pr(Y X) depends on X only through f(x)
Application to Brazil Wet Season = January-June
Application to Brazil Multi-model approach
Application to Brazil Dynamical downscaling CPTEC GCM (T42)
Application to Brazil Dynamical downscaling bias correction
Application to Brazil Statistical downscaling Parametric Linear Regression with Principal Components (each month) Step 1: Model PCs based on 1950-1996 ECHAM monthly precipitation ensemble mean (excluding prediction month) Step 2: Identify regression coefficients using PCs and observed values Step 3: Find PCs for prediction month with E found in Step 1 and applying the equation from Step 1. Step 4: Use the PCs from Step 3 and the regression coefficients from Step 2 to determine the monthly precipitation hindcast
Application to Brazil Hydrology models SMAP: conceptual, lumped model contains two reservoirs (subsurface and ground water) four parameters: soil saturation capacity, surface flow, a recharge coefficient, and a base flow recession coefficient. ABCD: nonlinear watershed model represents soil moisture storage, ground water storage, direct runoff, ground water outflow to the stream channel, and actual evapotranspiration.
Application to Brazil Pooling (Barnston et al. 2003; Robertson et al. 2004; Hagedorn et al. 2005) Least Squares Linear Regression (Krishnamurti et al. 1999, 2000; Rajagopalan et al. 2002) Normal Kernel Density Estimator
Ethiopia Courtesy of Dorling Kindersley Minimal HP dev (~5%) 83% no access to Elec Ambitious HP dev strategy Main stem BN development Kiremt season (JJAS) Base Map Courtesy of PLC Map CollecHon, UT
Linked Model System Predictors Precipitation Forecast Rainfall - Runoff Hydropower Benefits, Costs, Electricity, etc JJAS Precip Monthly SF, Evap CRU Climate Data (Block and Rajagopalan 2007) (Block and Strzepek 2010)
Linked Model System Predictors Precipitation Forecast Rainfall - Runoff Hydropower Benefits, Costs, Electricity, etc JJAS Precip Monthly SF, Evap CRU Climate Data Understand and predict interannual variability in P, specifically JJAS Predictors (Mar-May): Pacific: SLP, SST (ENSO) S. Atlantic: GeoP height Local: PDSI
Linked Model System Predictors Precipitation Forecast Rainfall - Runoff Hydropower Benefits, Costs, Electricity, etc JJAS Precip Monthly SF, Evap CRU Climate Data Nonparametric local polynomial approach, one season lead forecast (May)
Linked Model System Predictors Precipitation Forecast Rainfall - Runoff Hydropower Benefits, Costs, Electricity, etc JJAS Precip Monthly SF, Evap CRU Climate Data Dynamical Climate Forecast System (CFS) model from NOAA (NCEP/EMC) (Saha et al. 2006) coupled ocean-atmosphere model hindcast 1981-2004; 15 ensemble members global model strong JJAS monthly bias over Ethiopia (165mm 275mm)
Linked Model System Predictors Precipitation Forecast Rainfall - Runoff Hydropower Benefits, Costs, Electricity, etc JJAS Precip Monthly SF, Evap CRU Climate Data Semi-distributed model Simulates changes in soil moisture and runoff Produces Streamflow and Net PET
Linked Model System Predictors Precipitation Forecast Rainfall - Runoff Hydropower Benefits, Costs, Electricity, etc JJAS Precip Monthly SF, Evap CRU Climate Data IMPEND (Investment Model for Planning Ethiopian Nile Development) Planning/systems model with operational level detail Decision variable is reservoir head level Objective value: maximize net present value (benefits)
Linked Model System Predictors Precipitation Forecast Rainfall - Runoff Hydropower Benefits, Costs, Electricity, etc JJAS Precip Monthly SF, Evap CRU Climate Data Border Project Installed Power (MW) Mabil Mendaia Karadobi Karadobi 1350 Mabil 1200 Mendaia 1620 Border 1400
Linked Model System Predictors Precipitation Forecast Rainfall - Runoff Hydropower Benefits, Costs, Electricity, etc JJAS Precip Monthly SF, Evap CRU Climate Data Compare Modes (1961-2000) Actual Forecast: JJAS precip in May, climatology other months Monitoring: climatological SF all months, current conditions Systems approach: Climate + Hydrology + Water Management
Thank You