Optimizing Reservoir Operations to Adapt to 21st Century Expectations of Climate and Social Change in the Willamette River Basin, Oregon

Reservoir systems in the western US are managed to serve two main competing purposes: to reduce flooding during the winter and spring, and to provide water supply for multiple uses during the summer. Because the storage capacity of a reservoir cannot be used for both flood damage reduction and water storage at the same time, these two uses are traded off as the reservoir fills during the transition from the wet to the dry season. Climate change, population growth, and development in the western US may exacerbate dry season water scarcity and increase winter flood risk, creating a need to critically evaluate the status quo for reservoir operations. Focusing on a system of thirteen reservoirs (the Willamette Project) in the Willamette River Basin, Oregon, we estimated the values of reservoir management for reducing expected damages in the floodplain and for storing water for recreation in and around the reservoirs. We then used these values in a dynamic program to estimate the optimal fill path over the winter-spring transition period for both historical conditions and future scenarios of climate and social change. The value of stored water for summertime reservoir recreation was estimated based on the response of recreational use day counts to variation in water levels at nine of the reservoirs over the period 2001 to 2011. Visitor days were found to decline by as much as 2% per foot of drop in water level below full pool. The implied value of water to recreational users varied from $0.10 to $78 per acre-foot per month, depending on the reservoir. This range of values is comparable to prior estimates of the value of reservoir recreation in other parts of the western US, and is also similar to the value of water to irrigated agriculture in the Willamette River Basin, estimated in other studies. Because water cannot be used for recreation and irrigation at the same time, these results suggest that management of the Willamette reservoirs may benefit society by releasing stored water in summer for downstream needs from some reservoirs, while maintaining full pool for recreation in others, as prescribed currently. The expected value of flood damage reduction was estimated based on the probability of floods, flood inundation depth, and reservoir capacity, for three scenarios of future development in the Willamette River Basin, which were simulated by the Willamette Water 2100 land transition model over the period from present to 2100 using mid-range projections of future climate (MIROC5) and assumption about future population growth. Estimates of expected flood damage reduction varied as a function of flood risk over the course of the winter and spring, as well as the rate of population growth and associated development in the floodplain. At a weekly time scale the expected benefits of flood damage reduction ranged from a high in mid-January of $304 million to $1,284 million (depending on the rate of population growth), to near zero at the end of May. Finally, these analyses were combined in a dynamic programming approach to evaluate the optimal rate of fill for the Willamette Project reservoirs, at a weekly time step over the period from January to the end of May. This was done by treating the system of thirteen reservoirs as a single reservoir. The dynamic program found the fill path that minimized the sum of flood damages and foregone recreation benefits, subject to the constraint of available water inflows to the reservoir system. The estimate optimal fill path depended on the rate of decline of expected flood damages from mid-January to May. Anticipated future increases in winter flood risk and reduced spring streamflow, associated with climate change, shifted the optimal fill path to begin earlier and to fill more slowly, compared to the optimal fill path under historical conditions. The model confirms the intuitive result that the greater the value of stored water the earlier the optimal date to begin filling. Conversely, the greater the expected value of flood damage reduction the later the initiation of fill. Despite uncertainties in the estimated values of expected flood damage reduction and stored water for recreation, as well as the limitations of the dynamic program in modelling the coordinated management of multiple reservoirs, the approach and findings of this analysis contribute to our understanding of how reservoir management may need to adapt to future changes in water supply and demand