Evaluating Downscaled Climate Models for Hydrological Applications in the Mississippi River Basin

Precipitation projections are routinely used to drive hydrological models for water resources planning and management. Because simulated river discharge can hide critical biases in the precipitation forcings, users must carefully select hydroclimatic datasets before accepting the results of their hydrological routing models. In this analysis, six five-year downscaled historical climate simulations of precipitation over the Mississippi River basin (MRB) are characterized using two performance metrics. The first is a standard tabulation of mean percent error, whereas the second weights the percent error calculation by the discharge contribution of the associated MRB sub-basin. While the “discharge-weighted” percent error did not alter any conclusions regarding the relative performance of the six example simulations, it did yield larger (smaller) error characterizations for simulations whose greatest biases occurred over important (nominal) regions of runoff production within the MRB. To better highlight the conceptual utility of this approach, four hypothetical precipitation bias fields for the MRB are presented and again characterized by both metrics. In these examples, the “most accurate” precipitation field depends on the discharge-weighting process, with comparable MRB-mean percent errors yielding significantly different discharge-weighted biases. These results are immediately useful for the flurry of scientific activity directed at developing future MR discharge projections. However, any user of precipitation data for hydrological modeling may consider adapting this method to identify the most promising hydroclimate data to drive future river-discharge simulations over macro-river basins.