资源环境科技发展态势分析平台

Estimating future hydrologic floods under nonstationary climate is a key challenge for flood management. Climate-informed approaches to long-term flood projection are an appealing alternative to traditional modeling chains. This work formalizes climate-informed approaches into a general methodology consisting of four steps: (1) selection of predictand representing extreme events, (2) identification of credible large-scale predictors that mechanistically control the occurrence and magnitude of the predictand, (3) development of a statistical model relating the predictors to the predictand, and (4) projection of the predictand by forcing the model with predictor projections. These four steps, developed from a review of the current literature, are demonstrated for multiple gages in the northwest Ohio River Basin in the United States Midwest as a case study. Floods are defined as annual maximum series events in January through April and are linked to geopotential height and soil moisture predictors in a Bayesian linear regression model. The projections generally show a slight decrease in future flood magnitude and demonstrate the transparency of the climate-informed approach. An initial step for more general application across the United States and remaining challenges associated with climate-informed flood projection are discussed.