MITIGATING THE EFFECTS OF CLIMATE CHANGE ON THE WATER RESOURCES OF THE COLUMBIA RIVER BASIN

Transcription

1 MITIGATING THE EFFECTS OF CLIMATE CHANGE ON THE WATER RESOURCES OF THE COLUMBIA RIVER BASIN JEFFREY T. PAYNE, ANDREW W. WOOD, ALAN F. HAMLET, RICHARD N. PALMER and DENNIS P. LETTENMAIER Department of Civil Engineering, 164 Wilcox Hall, P.O. Box , University of Washington, Seattle, WA , U.S.A. Abstract. The potential effects of climate change on the hydrology and water resources of the Columbia River Basin (CRB) were evaluated using simulations from the U.S. Department of Energy and National Center for Atmospheric Research Parallel Climate Model (DOE/NCAR PCM). This study focuses on three climate projections for the 21st century based on a business as usual (BAU) global emissions scenario, evaluated with respect to a control climate scenario based on static 1995 emissions. Time-varying monthly PCM temperature and precipitation changes were statistically downscaled and temporally disaggregated to produce daily forcings that drove a macroscale hydrologic simulation model of the Columbia River basin at 1 /4-degree spatial resolution. For comparison with the direct statistical downscaling approach, a dynamical downscaling approach using a regional climate model (RCM) was also used to derive hydrologic model forcings for 20- year subsets from the PCM control climate ( ) scenario and from the three BAU climate ( ) projections. The statistically downscaled PCM scenario results were assessed for three analysis periods (denoted Periods 1 3: , , ) in which changes in annual average temperature were +0.5, +1.3 and +2.1 C, respectively, while critical winter season precipitation changes were 3, +5 and +1 percent. For RCM, the predicted temperature change for the period was +1.2 C and the average winter precipitation change was 3 percent, relative to the RCM control climate. Due to the modest changes in winter precipitation, temperature changes dominated the simulated hydrologic effects by reducing winter snow accumulation, thus shifting summer streamflow to the winter. The hydrologic changes caused increased competition for reservoir storage between firm hydropower and instream flow targets developed pursuant to the Endangered Species Act listing of Columbia River salmonids. We examined several alternative reservoir operating policies designed to mitigate reservoir system performance losses. In general, the combination of earlier reservoir refill with greater storage allocations for instream flow targets mitigated some of the negative impacts to flow, but only with significant losses in firm hydropower production (ranging from 9 percent in Period 1 to 35 percent for RCM). Simulated hydropower revenue changes were less than 5 percent for all scenarios, however, primarily due to small changes in annual runoff. 1. Introduction The Columbia River basin (CRB) covers portions of seven western states and the Canadian province of British Columbia, and has a total drainage area about Corresponding author. Climatic Change 62: , Kluwer Academic Publishers. Printed in the Netherlands.

2 234 JEFFREY T. PAYNE ET AL. Figure 1. Map of study area with streamflow routing network used by the VIC hydrology model and a schematic of the ColSim water resource model. the size of the state of Texas (Figure 1). The water resources of the CRB have been extensively developed over the past 60 years for flood control, hydropower production, irrigation and navigation. There are 36,400 MW (average annual generation is about 16,500 MW) of installed hydropower generation capacity at 214 dams (federal, municipal, and independent) within the CRB, and some 1.4 million hectares (3.5 million acres) are irrigated by the Columbia River and its tributaries. The Pacific Northwest (PNW) has by far the largest fraction of hydropower use in the U.S. (around 70%), most of which is produced at thirty federally owned dams. Flood control is likewise an important consideration. The CRB reservoir system operating policies help reduce flood damages throughout the basin, particularly along the lower river near Portland, OR. Ecological considerations such as preservation and enhancement of salmon under the U.S. Endangered Species Act, and use of the river and reservoirs for recreation, are also important in the management of the river s water resources (Bonneville Power Administration: BPA, 2001). The hydrology of the CRB is dominated by the temperature-sensitive cycle of snow accumulation and melt (Leung and Ghan, 1998), hence substantial shifts in runoff patterns can result from a relatively modest warming (Gleick and Chaleki, 1999; McCabe and Wolock, 1999; Hamlet and Lettenmaier, 1999; Lettenmaier et al., 1992; Lettenmaier and Gan, 1990). A seasonality shift in snowmelt runoff is associated with reduced winter snow accumulation, earlier peak snowmelt, higher winter runoff and higher evapotranspiration, and leads to lower streamflows during the low precipitation months of summer and autumn, with deleterious consequences for managed water resources (e.g., Lettenmaier and Sheer, 1991). The CRB s climate sensitivity is likely to add stress to the already complicated and conflicting objectives faced by CRB water resources management (Hamlet and

3 THE EFFECTS OF CLIMATE CHANGE ON THE COLUMBIA RIVER BASIN 235 Lettenmaier, 1999; Miles et al., 2000). Hamlet and Lettenmaier (1999) found that shifts in the seasonality of inflows to the CRB s reservoir system reduced the reliability of spring and summer non-firm energy production, irrigation, summer instream flow targets, and summer recreation in Lake Roosevelt (formed by Grand Coulee Dam, and the largest U.S. reservoir in the CRB). Firm power, the system s contracted minimum hydropower production, was most affected by the decreasing ability of the system to maintain storages through the summer for late autumn drawdown. This paper uses methods similar to those described by Hamlet and Lettenmaier (1999) to establish the sensitivity of the CRB s water resources to climate projections for the 21st century, and explores the potential of reservoir system operations to mitigate negative performance impacts. The basic study framework is a sequence of modeling steps of climate, hydrology and water resources that has been widely employed in climate change impact analyses. The climate change projections were produced by a state-of-the-art coupled land-atmosphereocean general circulation model (GCM), the DOE/NCAR Parallel Climate Model (PCM) (Washington et al., 2000; Dai et al., 2004; Pierce et al., 2004). Unlike previous studies, our results are based on three ensemble representations of transient future climate for a single global emissions scenario that can be averaged to reduce the compounding effects of interannual variability. Like companion studies in the Sacramento-San Joaquin basin (VanRheenen et al., 2004) and the Colorado River system (Christensen et al., 2004), this study utilizes statistical downscaling approaches for translating climate model outputs into hydrologic model inputs. In addition, we include an analysis of a dynamic downscaling experiment using a nested, meso-scale climate model driven by boundary conditions from the PCM simulations. We do not, in this paper, address potential non-climate related changes within the Columbia River basin that would affect the performance of the water resources system, such as energy and water demand changes due to demographic shifts, evolving power alternatives, and the re-prioritization of current operating objectives. Nonetheless, the climate change-related effects on the region s water resources form a foundation upon which a broader study including these aspects might be built. 2. Approach 2.1. PCM SCENARIOS PCM is a coupled land-atmosphere-ocean GCM that simulates the evolution of climate and its dependence on greenhouse gas (GHG) concentrations (Washington et al., 2000). The conclusions of this study are based on an ensemble of three PCM business-as-usual (BAU) future climate scenarios and one climate control

4 236 JEFFREY T. PAYNE ET AL. Table I PCM simulations used in this study Run Description Period B06.22, B06.28 Historical (CO 2 + aerosols evolving from pre-industrial levels) B06.45 Climate Control (CO 2 + aerosols at 1995 levels) B06.44, B06.46, Climate Change (BAU6, future scenario forcing) B06.47 scenario, representing 1995 climate, as described in Dai et al. (2004) and Pierce et al. (2004). In addition to the control and future climate runs, two 130-year historical PCM simulations, in which GHG emissions evolved from pre-industrial to contemporary levels, were used to derive statistics needed for downscaling the PCM and RCM control and climate change runs. A 20-year segment of B06.22 ( ) was used for the RCM downscaling and a 50-year segment of B06.28 ( ) was used for PCM downscaling. The PCM runs are summarized for convenience in Table I. The results for the statistical downscaled PCM scenarios were averaged across the BAU ensembles for three 30-year periods (denoted Periods 1 3): , , and Our assessment approach, however, is based on a transient (or continuous, i.e., using an evolving time series of monthly climate variables) assessment of these warming scenarios, rather than the more common quasi-transient approach exemplified by Lettenmaier et al. (1999), in which the average changes (relative to a control run mean) from decade-long segments of the warming scenarios were used to adjust historic climate observations. The method used for translating the transient monthly climate model output into continuous daily forcings for the hydrologic analysis are outlined in Wood et al. (2002, 2004), to which the reader is referred for details. In brief, however, it is based on a probability mapping approach for bias-correcting ensemble climate forecasts via utilization of probability distributions of both climate model and observed climate variables. The bias-correction is followed by spatial disaggregation (downscaling) of the monthly climate model output to the hydrology model scale ( 1 / 4-degree) and temporally disaggregating the resulting forcing variable time series to the daily time step used by the hydrologic model. The bias-correction and downscaling approach was developed for this project (Wood et al., 2004) and is used in the two companion papers addressing climate change sensitivities of water resources in the Colorado River basin (Christensen et al., 2004) and the Sacramento San Joaquin River basin (VanRheenen et al., 2004). It is, however, closely related to the downscaling method described by Wood et al.

5 THE EFFECTS OF CLIMATE CHANGE ON THE COLUMBIA RIVER BASIN 237 (2002) for seasonal to interannual streamflow forecasting. The significant difference required to apply the approach to future climate scenarios, compared to the previous retrospective and seasonal forecast applications, is the preservation of the climate model warming trend (relative to the control run) by removing it before, and replacing it after, the bias-correction step RCM SCENARIOS The Regional Climate Model (RCM) of Leung et al. (2003) and Leung and Ghan (1998) is a mesoscale model run at 1 / 2-degree spatial resolution for the western U.S. and southern British Columbia, Canada. RCM was used to dynamically downscale 20-year segments (July 2040 June 2060) of each PCM BAU ensemble member, of the PCM control run ( ), and of the PCM historical run from the PCM scale to 1 / 2-degree spatial resolution (see Leung et al., 2004 for details). The BAU and control climate scenario segments were subsequently bias-corrected and further downscaled to the 1 / 4-degree input resolution of the hydrologic model (Section 2.3) using the same methods used for the PCM scenarios (Section 2.1). The approach for bias-correcting and downscaling the PCM and RCM differed slightly in that the runs generating the necessary statistics were 20 years for RCM, compared to 50 years for PCM, and in the handling of the temperature trend removal and replacement (also on account of the different run lengths) CRB VIC MODEL APPLICATION The hydrologic model used in this study is the Variable Infiltration Capacity model (VIC) (Liang et al., 1994). VIC is a grid cell based model that represents fluxes of water and energy at the land surface, and has been widely used for simulation of large river basins (Maurer et al., 1999; Nijssen et al., 2001) including the Columbia (Nijssen et al., 1997). VIC was applied at 1 / 4-degree spatial resolution (1,668 grid cells totaling approximately 1.2 million km 2 ), in a mode requiring daily precipitation and minimum and maximum temperature as input. Most model parameters were taken from previous simulations of the continental U.S. reported in Maurer et al. (2002), which were run at 1 / 8-degree spatial resolution. For calibration (a discussion of the procedure is given in Nijssen et al., 1997), monthly streamflows at the set of locations (corresponding to inflow nodes for the water resources model) shown in Figure 1 were compared with naturalized streamflows (water management effects removed). Discrepancies between observed and predicted (monthly) streamflows were resolved by adjusting VIC parameters governing infiltration rate and base flow recession to improve the agreement of simulated and observed flows CRB WATER RESOURCE MANAGEMENT MODEL APPLICATION This study used the Columbia River Simulation Model (ColSim) to simulate (at a monthly time step) the major physical characteristics of the Columbia River

6 238 JEFFREY T. PAYNE ET AL. water resource system (reservoirs, run-of-river dams, diversions and return flows) and its reservoir operating policies (see Hamlet and Lettenmaier, 1999, for model description and validation). ColSim has been used to assess various aspects of the Columbia River s water resource system, including the effects of climate variability and operating system design (Miles et al., 2000) and the economic value of long-lead streamflow forecasting for hydropower (Hamlet et al., 2002). ColSim explicitly simulates reservoir operations by using naturalized inflows (observed or simulated with the VIC model) at each river node (Figure 1). ColSim assumes perfect foresight (forecasts) of unregulated runoff (i.e., ColSim looks forward in its input streamflow time series to anticipate perfectly future inflows) for the period from April 1 to August 31 to generate a set of reservoir rule curves affecting flood control and hydropower (BPA, 2001). ColSim is able to evaluate the sensitivity of the system to both changes in climate and operating policies for the following objectives: Flood Control. The U.S. Army Corps of Engineers (COE) evacuates major reservoirs in the autumn and early winter according to variable flood rule curves. The volume of draft depends on spring runoff forecasts (COE NPD, 1991), which ColSim represents by a perfect forecast of runoff volumes for April August or May August at specific river checkpoints. Similar policies apply to Canadian storage via international agreements. ColSim operations follow monthly reservoir evacuation schedules specified by the current operating policies, which are a function of expected runoff in spring and summer. Monthly regulated flows are used to estimate peak flood events by using a factor to translate average monthly outflows into peak daily flows (based on historic relationships between monthly flows and daily peak flows). When indexed daily flows exceed the COE threshold for minor damage, flood damages are calculated using COE estimates of historic damage avoided (COE NPD, 1991). Hydropower. Participating Canadian and U.S. projects are operated as a system to meet regional energy requirements, with flood control evacuation as a primary constraint. ColSim simulates support of a fixed system-wide energy target that varies monthly according to a typical load shape. Federal hydropower is marketed through the BPA, traditionally as firm or non-firm power. Firm power (a contractual designation referring to an assured quantity in the driest hydrologic conditions: BPA, 2001) is a reasonable surrogate for the worst case (at least in a fixed historic period) performance of the hydropower system. Recently, market responses to energy shortages and deregulation of wholesale markets have eroded the distinction between firm and non-firm power. Nonetheless, firm power is a concept that continues to be used to measure system performance. For this study, the implications of deregulation are acknowledged by using the same seasonal prices when estimating firm and non-firm proceeds. Instream Flow Targets for Fish. ColSim simulates the support of flow targets for fisheries habitat protection and enhancement at Columbia Falls, Priest Rapids Dam, Lower Granite Dam, McNary Dam, and Bonneville Dam control points

7 THE EFFECTS OF CLIMATE CHANGE ON THE COLUMBIA RIVER BASIN 239 (Figure 1), as well as minimum outflows at each reservoir. The McNary Dam (lower Columbia) and Lower Granite Dam (lower Snake) targets corresponds to the instream goals specified by the National Marine Fisheries Service (NMFS) Biological Opinion Papers (BiOP) (NMFS, 1995, 2000) written pursuant to Endangered Species Act listing of certain Columbia River salmonids. These system-wide flow targets are currently supported by limited storage in a few reservoirs. Agricultural Withdrawals. Withdrawals from the main stem of the Columbia River are relatively small in comparison with annual flow (about 6% in 1990) and are simulated by subtracting them from system inflows before reservoir simulation takes place. By contrast, Snake River irrigation withdrawals are large relative to annual flow (28%), and 1989 diversions are explicitly simulated by ColSim as a fixed diversion schedule constrained by natural inflows and reservoir storage. Recreation. Colsim simulates recreation-related operations only at Grand Coulee Dam (Lake Roosevelt), which restricts the Energy Content Curve (a primary rule curve limiting discretionary energy production) to 1,280 ft of elevation from June 30 to Labor Day to help preserve lake elevations appropriate for recreation. For reporting purposes, water resource results were produced for the same three periods as were used for reporting of the hydrology model results (Periods 1 3; , , and ). The VIC streamflow simulations, however, were run as continuous 105-year simulations through ColSim. The Col- Sim performance metrics for each period are averages over the three ensemble segments CRB WATER RESOURCE POLICY ALTERNATIVES Hamlet and Lettenmaier (1999) demonstrated that climate change-induced system performance decreases in the CRB arose mainly from inflow seasonality shifts. Therefore, we investigated four alternative reservoir operating policies for mitigating performance reductions: (1) lower flood evacuation quantities and earlier reservoir refill schedules; (2) the reallocation of firm hydropower production from winter demands to summer months; (3) increases in reservoir storage allocations for environmental targets; and (4) a combination of the previous three alternatives using heuristic combinatorial techniques. These alternatives are described below. Changes in Flood Evacuation and Refill. Altered flood evacuation and reservoir refill schedules (Martin, 2001) were tested as adaptive strategies to address the increased winter runoff and the lower spring inflows, both of which diminish the ability of the system to refill and sustain late summer operations. Five flood evacuation policies were combined with three system refill dates, yielding fifteen combinations of operating policies within this category. The COE-defined flood evacuations at each dam were multiplied by the sequence [0.8, 0.9, 1.0, 1.1, 1.2] (e.g., the 80% alternative requires 20% less draft than the present system does; and 100% represents current operations). The refill timing was adjusted by re-

8 240 JEFFREY T. PAYNE ET AL. Table II Increasing storage allocations for environmental flow requirements Current Alt. 1 Alt. 2 Alt. 3 Alt. 4 Alt. 5 operations Storage allocated (%) Allocation volume 2.3/ / / / / /41 (maf/bcm) ducing reservoir evacuation duration by three intervals: [zero, 2-weeks, 1-month]. The 1-month interval allows each reservoir to refill one month earlier; the 2-week option is approximated by allowing reservoirs to refill by half in their final flood evacuation month, and zero represents current refill schedules. Changes in the Seasonality of Firm Energy Demands. Because warmer winter temperatures could cause a decrease in winter power demand, the second adaptive strategy allows for greater conservation of winter inflows for spring refill needs. This set of alternatives reveals the sensitivity of operations to the reallocation of winter hydropower generation (i.e., firm power) to the summer (when its market value in California is high). To this end, firm power demands were reallocated from peak winter to low summer months using arbitrary percentages of seasonal demand: [5, 10, 15%]. Increasing the Storage Allocation for Instream Targets. The third adaptive strategy increases reservoir storage allocations for meeting monthly varying regulatory instream flow targets for fish. Current allocations make up only about 7% of the basin s total storage and are relatively small given the seasonal release requirements (hence current operations rely more heavily upon system inflows to meet demands). Warming-induced summer inflow reductions increase both the frequency and quantity of target flow shortfalls. The alternative operating strategy addresses shortfalls at the two lower Columbia River BiOP fish flow targets (at Mc- Nary and Bonneville Dams) by increasing reservoir storage allocations (Table II: volumes given in million acre-feet, maf, and billion cubic meters, bcm). Storage augmentation in the Snake River for the Lower Granite Dam BiOP target was not investigated because existing federal storage at Dworshak Dam is already heavily drafted in the current operations and affords little flexibility. Combination Alternative Policy. This alternative combines all three of the above operations changes, using an iterative process designed to restore the instream flow deficits at McNary Dam to values comparable to those achievable under the current climate and storage allocation, while maintaining hydropower revenues and current levels of flood mitigation. Operations in each of the three PCM periods (and for the RCM scenario) were calibrated independently. The

9 THE EFFECTS OF CLIMATE CHANGE ON THE COLUMBIA RIVER BASIN 241 iterative selection of operational changes focused on: (1) storage allocations reducing the McNary Dam average annual instream flow deficit to control climate levels; and (2) flood evacuations and system refill timing changes to maximize hydropower revenues, subject to maintaining the firm power capacity produced using current operations, and to limiting releases at The Dalles to avoid exceeding Portland-Vancouver area damage thresholds. 3. Results 3.1. CLIMATE CHANGE SUMMARY Figure 2a shows the downscaled, PNW-average annual mean precipitation and temperature time series for the BAU climate scenarios and their average, and the long-term averages for the PCM and RCM control climate scenarios, as well as the observed average. The control run averages were slightly warmer than the observed basin averages (reflecting warming in the last half-century), while the observed and control climate averages for precipitation were roughly equivalent. Although the BAU climate simulations were initially cooler than the control run, they were about 0.5, 1.3 and 2.1 C warmer than the control climate in Periods 1 3, respectively. The warming trend was not pronounced relative to interannual variability until Period 2. The control and BAU precipitation time series appear stationary and have similar means, although decadal and interannual variability created distinctive features such as the low precipitation in Period 1 and at the end of Period 3, and increased oscillatory behavior about the mean starting just before Period 2. Figure 2b shows that downscaled PNW-averaged temperature and precipitation were similar for the RCM and PCM runs from an annual average perspective. Figure 2c, however, shows that the monthly average climate changes over the CRB (roughly the eastern 2/3 of the PNW domain, upstream of The Dalles, OR, for which the subsequent water resources analysis was conducted), were quite different in some cases for RCM and PCM. For PCM, temperature increased progressively for Periods 1 3 in all months, and was slightly greater in winter and summer than in spring and fall; while precipitation changes were less structured, but by Periods 2 and 3, spring increases and summer decreases began to be pronounced. Winter season precipitation changes for PCM Periods 1 3 were 3, +5 and +1, respectively, while the RCM BAU winter had a 3% drop in precipitation. The RCM BAU climate was markedly warmer in late-winter and spring than for PCM even for Period 3, and fell between Periods 2 and 3 for other months. RCM winter precipitation decreased relative to the RCM control run by 3%, while the largest percentage changes (and also differences between PCM and RCM control climate precipitation) occurred in the low precipitation summer months. The differences between the downscaled RCM and PCM climate results are attributable in part to different periods for which BAU and control climate were

10 242 JEFFREY T. PAYNE ET AL. Figure 2. (a) Downscaled PCM BAU climate PNW-average annual total precipitation and average temperature, compared with long-term averages from the PCM and RCM control climates and observations ( ); (b) comparison of downscaled RCM and PCM BAU-averaged climate variables (legend from (a)); and (c) CRB-average monthly total precipitation and average temperature: PCM BAU Period average changes relative to PCM control climate ( Per-1 to Per-3 ), RCM BAU average changes relative to RCM control climate ( RCM BAU/CTRL ), and RCM control climate difference from PCM control climate ( RCM CTRL/CTRL ).

11 THE EFFECTS OF CLIMATE CHANGE ON THE COLUMBIA RIVER BASIN 243 analyzed, but were also a direct consequence of RCM s finer resolution of the land surface and RCM s atmospheric physics. RCM s higher spatial resolution allowed it to represent better winter and spring snow cover variations that greatly raise the land surface albedo over snow covered areas, modifying the surface energy balance and creating a positive feedback for temperature changes that diminish or expand the snow cover (Leung and Ghan, 1999; Giorgi et al., 1997) SNOWPACK CHANGES Spring snow water equivalent (SWE) was markedly reduced by the temperature changes of the BAU scenarios, particularly for the RCM simulations (Figure 3). For PCM, BAU climate SWE decreased progressively relative to the control run for Periods 1 3, but in Period 2, SWE was only slightly lower than in Period 1 because increased warming was offset by a relative winter-spring precipitation increase. The RCM control run was slightly warmer than the PCM control run, and had slightly lower April 1 SWE, relative to which, the RCM BAU SWE decrease was nearly as large as the PCM Period 3 decrease (relative to the PCM control SWE). RCM s BAU SWE decrease relative to the PCM control climate baseline was essentially equivalent to PCM s Period 3 decrease. RCM s greater SWE sensitivity was consistent with RCM s larger spring warming, compared to PCM Period 3 (Figure 2c), and the slight winter precipitation decreases. The changes in SWE did not decrease the April 1 snow extent greatly except in PCM s Period 3 and RCM s BAU period ( ), a result that contrasts with previous findings (e.g., Miles et al., 2000; Mote et al., 1999; Hamlet and Lettenmaier, 1999) of a substantial decline in snow covered area and average SWE. Reasons for the contrast include the earlier studies use of GCM scenarios with higher temperature sensitivities than the PCM BAU scenarios, differences in the baselines used for comparison (e.g., control as opposed to historical climate), and differences in averaging domains. The PCM BAU scenario SWEs changed more (by 22, 23 and 39%) relative to the average SWE from the cooler historical climate of as opposed to the control climate (note temperatures in Figure 2a). Also, PCM-based SWE changes for the west side of the Cascade Mountains, which were not included in the CRB averages, were larger: 17, 28 and 51% of control climate SWE and 36, 45 and 63% of SWE RUNOFF AND STREAMFLOW CHANGES Two effects of climate change on runoff volume changes and seasonality changes are traceable largely to changes in basin-averaged precipitation and temperature, respectively. Averaged over the CRB, average annual runoff for PCM, shown in Figure 4a, changed little: by 5, 0 and 3 percent for Periods 1 3, respectively. This result is consistent with the fairly stationary annual precipitation time series of Figure 2a. Spatially, however, larger decreases occurred in the mountainous headwater regions in Idaho and just north of the U.S. border, while there were

12 244 JEFFREY T. PAYNE ET AL. Figures 3 and 4.

13 THE EFFECTS OF CLIMATE CHANGE ON THE COLUMBIA RIVER BASIN 245 increases (by Periods 2 and 3) in the Canadian headwaters, as well as in the portion of the basin west of the Cascade Mountain range (i.e., west of the routing network of Figure 1). The RCM BAU climate changes relative to the RCM control climate exhibited the same spatial pattern, but the runoff decrease was exaggerated relative to the PCM decreases, and the headwater runoff increase was more spatially confined. The RCM control run, however, had increased runoff relative to the PCM control run, particularly in the mountainous regions. The control run differences in annual runoff in the two models likely resulted from the higher late summer and fall precipitation of the RCM run (Figure 2b,c), which sustained more baseflow than in the PCM control scenario, an effect that is evident in the streamflow hydrograph of Figure 4b. The Columbia River seasonal streamflow hydrograph at The Dalles, OR (which integrates the effects to most of the CRB domain) shows seasonality change (increased winter/spring flows and decreased summer flows) in PCM Periods 2 and 3, relative to the PCM control run, which reflects the progressively increasing temperatures and earlier snowmelt. In Period 1, summer flow decreased on average, but there was no apparent seasonality shift, as the associated temperature increase was slight. The RCM control run peak flows, on average, were similar to those of the PCM control run, but the RCM fall and winter flows were relatively elevated. The RCM BAU climate streamflow, however, showed severe peak and late summer flow reductions compared to even PCM BAU Period 3 flows, which were consistent with the severity of reductions in SWE (Figure 3) WATER RESOURCES SYSTEM EFFECTS Effects under Current Operations System performance reliabilities (Figure 5) were somewhat sensitive to the modest PCM BAU streamflow changes, but they were most sensitive to the RCM scenario, which projected the greatest streamflow timing shifts. The number of spring flood control exceedances at The Dalles of the 12,700 m 3 /s (450,000 cfs) index threshold decreased for all scenarios, although flooding sometimes occurred earlier (February and March) with seasonal hydrograph shifts. In autumn (the primary month for Facing page: Figure 3. Simulated April 1 snow water equivalent (SWE) with SWE CRB-average change (percent) for the PCM control climate and averaged BAU climate ensemble for Periods 1 3 (changes with respect to PCM control climate); for the RCM control climate (difference with respect to PCM control climate); and for the RCM BAU climate (changes with respect to RCM control climate). Figure 4. (a) Change/difference in average annual runoff for the PCM BAU climate Periods 1 3, and for the RCM control and BAU climates. Percentages are for the CRB average, and refer to the scenario pairings given above each figure. (b) Mean monthly streamflow hydrograph for the Columbia River at The Dalles, OR, for the PCM BAU ensemble average climate and RCM BAU climate, and the PCM and RCM control climate scenarios.

14 246 JEFFREY T. PAYNE ET AL. Figure 5. The effect of PCM and RCM climate change projections on CRB reservoir system reliabilities, as compared to the PCM control climate and operations scenario. shortfalls) firm power reliability also diminished as seasonality shifts progressed. BAU climate sustainable firm power (i.e., altered safe yield) decreased to 93, 95, 93 and 84% of the control scenario for the three PCM periods and the RCM scenario, respectively. Hydropower revenues, which are more dependent on flow volume than seasonality, were less sensitive, decreasing to 96, 101, 98 and 94% of the control climate levels, respectively. Climate change affected the McNary Dam flow target s cumulative annual shortfalls more than the average reliability (April through August) (Figure 5). The BAU climate summer flow volume reductions increased the annual flow deficits for all three PCM periods and the RCM scenario (by +30, +20, +40 and +76% relative to the control climate, respectively). Likewise, deficits for the Middle Snake River plain increased for all three PCM periods and were most severe for the RCM scenario. The reliability of the recreational storage level at Lake Roosevelt (which is maintained until September) depends mostly on summer runoff, and was also most sensitive to the RCM scenario flow changes (Figure 5) Effects of Changing Flood Evacuation and Refill Requirements Lower flood evacuation requirements and earlier refill schedules produced minimal benefits for instream flow targets because their combined effect compromised winter flood control, and because of the limitations on storage allocations during summer months. Using earlier refill (Figure 6) to capture more spring inflow redistributed shortfalls throughout the summer, causing lower target reliabilities. Figure 7 shows the sensitivity to earlier refill of hydropower and flood con-

15 THE EFFECTS OF CLIMATE CHANGE ON THE COLUMBIA RIVER BASIN 247 Figure 6. Monthly effects of earlier refill on McNary environmental target (deficit and reliability). The Control bar represents current climate conditions under current operations. The Current Operations bar presents the climate change ensemble period operated under current reservoir policies. The two-week and one-month bars present the specified climate change ensemble period, operated with an earlier (specified) refill date.

16 248 JEFFREY T. PAYNE ET AL. Figure 7. Sensitivity of system refill timing experiments. Change in Annual Revenue and % Firm Power Sustainable are relative to values produced under the control climate with current operations. trol. Mean annual flood damages increased due to the higher reservoir levels in early spring, but were dwarfed monetarily by the accompanying mean annual hydropower revenues. Furthermore, firm power increased by 0, 2 and 3% for Periods 1 3, respectively. Generally, this indicates that earlier refill dates were the most effective policy for meeting spring refill goals without compromising flood control throughout the winter (despite having a limited benefit for instream flow target reliability). Tradeoff curves are not shown for the RCM scenario, but the results were similar in character.

17 THE EFFECTS OF CLIMATE CHANGE ON THE COLUMBIA RIVER BASIN 249 Table III Sensitivity of operations to reallocations of firm power demand from winter to summer months Hydropower Sustainable Flood McNary McNary revenues a firm power a damages b deficits a reliability c (%) (%) ($M) (%) (%) Period 1 Current operations % Reallocation Period 2 Current operations % Reallocation Period 3 Current operations % Reallocation a As percent of the control climate with current operations scenario. b Annual average over period of analysis. c Percent of months where target is met entirely Effects of Changing the Seasonality of Firm Energy Demands Warmer winters should reduce winter power demand, decrease winter hydropower drafts and increase carryover storages. Reallocating seasonal firm power demands, however, changed system operations (Table III) little for BAU Periods 1 3, primarily because (1) firm power allocations for streamflows control drafts only in the lowest hydrologic years, and (2) flood control evacuation requirements continue to mandate similar winter releases from storage despite the lowered winter hydropower demand Effects of Increasing the Storage Allocation for Instream Targets The inability of earlier storage refill to diminish shortfalls for the McNary Dam instream target suggests that greater access to reservoir storage is necessary to maintain current levels of flow augmentation. Increasing the storage allocation to 10.2 billion m 3 (8.3 million acre-feet) from the current 2.8 billion m 3 (2.3 million acre feet) allocation restored the annual deficit at McNary Dam to the control climate value for all BAU periods (Table IV; Figure 8). PCM Period 3 and the RCM scenario required the greatest storage allocations (6.0 and 8.3 million acre-feet, respectively). The target reliability, however, was not correspondingly improved because ColSim distributes allocations across the season instead of meeting the entire demand in the first few months. Increasing reservoir storage for instream flows

18 250 JEFFREY T. PAYNE ET AL. Table IV Characteristics of the combination alternative Current Period 1 Period 2 Period 3 RCM operations Sustainable Firm Power emand (average million MWh/month) Allocation of storage for environmental flows: billion m 3 (million acre feet) (2.3) (4.3) (4.3) (6.0) (8.3) Timing of refill Current 2-weeks 2-weeks 2-weeks 2-weeks timing early early early early Assigned reallocation of firm power 0% 0% 5% 10% 10% demand (winter to summer) Flood evacuation (as a percent of current 100% 100% 100% 100% 100% COE requirements) also strongly reduced firm power (Figure 9), a tradeoff that held for Periods 1 3 and the RCM scenario (not shown, but evident from Table IV), and overshadowed effects on flood damages and power revenues. Using storage to maintain the Mc- Nary Dam instream target was the most effective policy for preserving their present performance levels, but conflicted with hydropower operations requiring storage to meet autumn energy targets. The tradeoffs shown in Figure 9 summarize the starkest result of this study: maintaining the performance of the reservoir system with respect to (summer season total) instream flow targets would result in substantial (10 20%, depending on the future time period) reductions in hydropower production as represented by firm power. And, even with these reductions in firm power, late summer minimum flows would still be lower than at present Effects of Combining Alternative Polices The combination alternative (detailed in Table IV) was designed to maintain current levels of instream flow support and hydropower revenues without compromising flood control. Generally, streamflow seasonality shifts caused both autumn firm power and the McNary Dam instream targets to rely more on storage allocations than on inflows, hence, as Figure 10 shows, limiting McNary Dam annual deficits to control climate levels carried the tradeoff of reducing firm power (all BAU periods). Although the actual flood risk would better be estimated via a detailed routing study, our analysis suggests that a slight increase in flood risk (or operational changes that reduce flood evacuation targets) may offer a cost ef-

19 THE EFFECTS OF CLIMATE CHANGE ON THE COLUMBIA RIVER BASIN 251 Figure 8. Effects of greater storage allocations for environmental flows upon the McNary Dam target by month. Control bar represents the control climate with current operations. Current operations bar represents the specified climate change analysis period with current operations and a 2.8 billion m 3 (2.30 million acre-foot) storage allocation for instream augmentation. Numeric bars represent the modified volume designation for instream targets, in million acre-feet. [Allocations in metric units are 4.1, 5.3, 7.4, 10.2, and 41 billion m 3, respectively].

20 252 JEFFREY T. PAYNE ET AL. Figure 9. Sensitivity to an increased instream target allocation in the upper Columbia R. basin. fective (although perhaps politically difficult) strategy for increasing hydropower revenues. 4. Conclusions Statistically downscaled PCM and RCM future climate scenarios were compared with corresponding current climate control runs, and the associated simulations of CRB hydrology and water resources were evaluated for future PCM Periods 1

21 THE EFFECTS OF CLIMATE CHANGE ON THE COLUMBIA RIVER BASIN 253 Figure 10. Combined effects of changes to flood evacuation and refill timing, firm demand distribution, and instream storage allocation.

22 254 JEFFREY T. PAYNE ET AL. 3 ( , and ) and the RCM period The climate change scenarios suggest that a moderate, progressive warming reaching 2.1 C in Period 3 will produce a gradual shift toward diminished snowpacks and earlier snowmelt runoff, accompanied by reduced summer and fall low flows. As in previous studies, this is the dominant predicted signature of global warming on the hydrology of PNW streams over the next century. Because precipitation changes in the downscaled climate scenarios were secondary to temperature effects, annual average runoff volume changes (relative to control climate runoff) were predicted to be small generally less than 5%. Although annual average temperatures and precipitation in the future scenarios were broadly similar for RCM and PCM after statistical downscaling, RCM s greater spring warming and larger decreases in winter precipitation suggest that climate change effects would be accelerated relative to those projected by the statistically downscaled PCM scenarios. The RCM midcentury snow water equivalent and runoff decreases and runoff timing shifts were at least as severe as those found in PCM Period 3. From the perspective of water resource operations, the projected hydrologic changes would have the greatest effect from spring to autumn, when the reservoir system refills and is intended to maintain storage until the winter reservoir drawdown for flood control and hydropower production. Increasing winter inflows associated with seasonality shifts would necessitate the continuation of present flood control policies despite the decreased ability of the system to replenish current evacuations in the spring. Lower summer streamflows would exacerbate the reduced reservoir refill by increasing drafts for instream flow targets. The lower resulting storage at the end of summer would reduce the ability of the system to meet present firm power production (hydropower safe yield ) targets during the winter, before reservoir storage begins to be restored by winter precipitation. Hydropower revenues were predicted to be relatively unaffected, however, because annual streamflow volume changes were generally small, and fall and early winter generation reductions would be compensated by increases in late winter and spring. The starkest result of this study is an evolving tradeoff between reservoir releases to maintain instream flows for fish, and hydropower production. In order to maintain performance of the reservoir system with respect to instream flow targets developed under the NMFS Biological Opinion associated with ESA listing of Columbia River salmonids, substantial (10 20%, depending on the future time period) reductions in firm hydropower would be required. Even with these reductions in firm power, late summer minimum flows would still be lower than at present. Although the monthly time step used in this study makes it impossible to state explicitly the changes in flood risk and/or damages that would accompany the projected hydrologic changes, the opportunity costs associated with maintaining the same general flood control policy appear to be much higher than the associated benefits. This study suggests that the reconsideration of flood control needs and values in the context of instream flow and hydropower production considerations posed by a warmer PNW climate would be highly desirable.

23 THE EFFECTS OF CLIMATE CHANGE ON THE COLUMBIA RIVER BASIN 255 Acknowledgements The U.S. Department of Energy s Office of Science (BER) Accelerated Climate Prediction Initiative provided funding for this research, which was performed at the University of Washington under Grant No AQO. This publication was also supported by the Joint Institute for the Study of the Atmosphere and Ocean (JISAO) under NOAA Cooperative Agreement #NA17RJ1232, Contribution #922. References BPA (Bonneville Power Administration), U.S. Army Corps of Engineers, and U.S. Bureau of Reclamation: 2001, The Columbia River System: The Inside Story, 2nd edn., Report DOE/BPA Published for the Columbia River System Review by the COE and USBR. Christensen, N. S., Wood, A. W., Lettenmaier, D. P., and Palmer, R. N.: 2004, Effects of Climate Change on the Hydrology and Water Resources of the Colorado River Basin, Clim. Change 62, COE NPD (U.S. Army Corps of Engineers, North Pacific Division): 1991, Columbia River and Tributaries Review Study: Review of Flood Control, COE Report. Dai, A., Washington, W. M., Meehl, G. A., Bettge T. W., and Strand, W. G.: 2004, The ACPI Climate Change Simulations, Clim. Change 62, Giorgi, F., Hurrell, J. W., Marinucci, M. R., and Beniston, M.: 1997, Elevation Dependency of the Surface Climate Signal: A Model Study, J. Climate 10, Gleick, P. H. and Chaleki, E. L.: 1999, The Impacts of Climatic Changes for Water Resources of the Colorado and Sacramento-San Joaquin River Basins, J. Amer. W. Res. Assn. 35 (6), Hamlet, A. F., Huppert, D., and Lettenmaier, D. P.: 2002, Economic Value of Long-Lead Streamflow Forecasts for Columbia River Hydropower, J. Water Plan. & Mgmt. ASCE128 (2), Hamlet, A. F. and Lettenmaier, D. P.: 1999, Effects of Climate Change on Hydrology and Water Resources in the Columbia River Basin, J. Amer. W. Res. Assn. 35 (6), Lettenmaier, D. P., Brettman, K. L., Vail, L. W., Yabusaki, S. B., and Scott, M. J.: 1992, Sensitivity of Pacific Northwest Water Resources to Global Warming, Northwest Envir. J. 8 (2), Lettenmaier, D. P. and Gan, T.-Y.: 1990, Hydrologic Sensitivities of the Sacramento-San Joaquin River Basin, California, to Climate Change, Water Resour. Res. 26, Lettenmaier, D. P. and Sheer, D. P.: 1991, Climatic Sensitivity of California Water Resources, J. Water Res. Plan. & Mgmt. 117 (1), Lettenmaier, D. P., Wood, A. W., Palmer, R. N., Wood, E. F., and Stakhiv, E. Z.: 1999, Water Resources Implications of Global Warming: A U.S. Regional Perspective, Clim. Change 43 (3), Leung, L. R. and Ghan, S. J.: 1998, Parameterizing Subgrid Orographic Precipitation and Surface Cover in Climate Models, Mon. Wea. Rev. 126 (12), Leung, L. R. and Ghan, S. J.: 1999, Pacific Northwest Climate Sensitivity Simulated by a Regional Climate Model Driven by a GCM. Part II:2 CO(2) Simulations, J. Climate 12 (7), Leung, L. R., Qian, Y., and Bian, X.: 2003, Hydroclimate of the Western United States Based on Observations and Regional Climate Simulation of Part I: Seasonal Statistics, J. Climate 16 (12), Leung, L. R., Qian, Y., Bian, X., Washington, W. M., Han, J., and Roads, J. O.: 2004, Mid-Century Ensemble Regional Climate Change Scenarios for the Western United States, Clim. Change 62,

24 256 JEFFREY T. PAYNE ET AL. Liang, X., Lettenmaier, D. P., Wood, E. F., and Burges, S. J.: 1994, A Simple Hydrologically Based Model of Land Surface Water and Energy Fluxes for General Circulation Models, J. Geophys. Res. 99 (D7), Martin, K.: 2001, Hydro Recommendations to Counter Global Warming in the Pacific Northwest: CRITFC s Natural-Peaking Hydrograph, Memo to University of Washington s Climate Impacts Group, Oct. Maurer, E. P., Lettenmaier, D. P., and Roads, J. O.: 1999, Water Balance of the Mississippi River Basin from a Macroscale Hydrologic Model and NCEP/NCAR Reanalysis, EOS, Transactions of the American Geophysical Union 80 (46), F Maurer, E. P., Wood, A. W., Adam, J. C., Lettenmaier, D. P., and Nijssen, B.: 2002, A Long-Term Hydrologically-Based Data Set of Land Surface Fluxes and States for the Conterminous United States, J. Climate 15, McCabe, G. J. and Wolock, D. M.: 1999, General-Circulation-Model Simulations of Future Snowpack in the Western United States, J. Amer. Water Res. Assn. 35 (6), Miles, E. L., Snover, A. K., Hamlet, A. F., Callahan, B., and Fluharty, D.: 2000, Pacific Northwest Regional Assessment: The Impacts of Climate Variability and Climate Change on the Water Resources of the Columbia River Basin, J. Amer. Water Res. Assn. 36 (2), Mote P. (ed.) and co-authors: 1999, Impacts of Climate Variability and Change: Pacific Northwest, Regional Report for the PNW to the National Assessment of the Impacts of Climate Variability and Change, JISAO Climate Impacts Group, University of Washington, Nov. Nijssen, B., Lettenmaier, D. P., Liang, X., Wetzel, S. W., and Wood, E. F.: 1997, Streamflow Simulation for Continental-Scale River Basins, Water Resour. Res. 33 (4), Nijssen, B., O Donnell, G. M., Hamlet, A. F., and Lettenmaier, D. P.: 2001, Hydrologic Sensitivity of Global Rivers to Climate Change, Clim. Change 50 (1 2), NMFS (National Marine Fisheries Service): 1995, Final 1995 FCRPS Biological Opinion, NOAA Publication. NMFS: 2000, Final 2000 FCRPS Biological Opinion, NOAA Publication. Pierce, D. W., Barnett, T. P., Tokmakian, R., Semtner, A., Maltrud, M., Lysnc, J., and Craig, A.: 2004, The ACPI Project, Element 1: Initializing a Coupled Climate Model from Observed Conditions, Clim. Change 62, VanRheenen, N. T., Wood, A. W., Palmer, R. N., and Lettenmaier, D. P.: 2004, Potential Implications of PCM Climate Change Scenarios for California Hydrology and Water Resources, Clim. Change 62, Washington, W. M., Weatherly, J. W., Meehl, G. A., Semtner, A. J., Bettge, T. W., Craig, A. P., Strand, W. G., Arblaster, J., Wayland, V. B., James, R., and Zhang, Y.: 2000, Parallel Climate Model (PCM) Control and Transient Simulations, Clim. Dyn. 16 (10 11), Wood, A. W., Leung, L.-Y., Sridhar, V., and Lettenmaier, D. P.: 2004, Downscaling Climate Model Surface Precipitation and Temperature: A Comparison of Methods, Clim. Change 62, Wood, A. W., Maurer, E. P., Kumar, A., and D. P. Lettenmaier: 2002, Long Range Experimental Hydrologic Forecasting for the Eastern U.S., J. Geophys. Res. 107 (D20), 4429, doi: /2001jd (Received 21 June 2002; in revised form 13 August 2003)