Adaptive water infrastructure planning for nonstationary hydrology.

The uncertainty of a changing climate raises challenges for water infrastructure planning and design. Not accounting for nonstationarity may result in under-designed structures that fail too frequently, or over-designed structures that are economically inefficient. This concern is magnified by uncertainty in the long-term frequency and magnitude of future extreme events. Planning strategies that allow adaptations over a structure's life could improve both reliability and economic efficiency. This study develops a method to inform adaptive water infrastructure planning with uncertain hydrologic and other forms of nonstationarity, applied to levee system planning. A stochastic dynamic programming model including a Markov process is developed for infrastructure planning with uncertain nonstationarity in flood frequency. Bayes' theorem is used to update peak flow probabilities conditioned on observed past peak flows and to update expected residual flood damages over time. A levee system planning problem with a numerical example from California illustrates the approach to derive optimal levee heights over time, and economic values of adapting to uncertain nonstationary flood risk. The projected range of probabilistic hydrology scenarios affects the optimal results, particularly in later planning stages as hydrology scenarios diverge with time. Adaptive planning strategies allowing more levee upgrades over time slightly lowers the overall cost and provides better flood protection than one-time construction under nonstationary hydrology for any climate in the example. Compared to a known future nonstationary hydrology, incorporating uncertain nonstationary climate results in higher levees being planned for observed severe hydrology scenarios in later stages. The overall present value cost with uncertain nonstationary climate depends on rates of change in peak flow distribution parameters in future hydrology scenarios. [ABSTRACT FROM AUTHOR]