Integrating weather and climate predictions for seamless hydrologic ensemble forecasting: A case study in the Yalong River basin

Abstract

Despite the tremendous improvement made in numerical weather and climate models over the recent years, the forecasts generated by those models still cannot be used directly for hydrological forecasting. A post-processor like the Ensemble Pre-Processor (EPP) developed by U.S. National Weather Service must be used to remove various biases and to extract useful predictive information from those forecasts. In this paper, we investigate how different designs of canonical events in the EPP can help post-process precipitation forecasts from the Global Ensemble Forecast System (GEFS) and Climate Forecast System Version 2 (CFSv2). The use of canonical events allow those products to be linked seamlessly and then the post-processed ensemble precipitation forecasts can be generated using the Schaake Shuffle procedure. We used the post-processed ensemble precipitation forecasts to drive a distributed hydrological model to obtain ensemble streamflow forecasts and evaluated those forecasts against the observed streamflow. We found that the careful design of canonical events can help extract more useful information, especially when up-to-date observed precipitation is used to setup the canonical events. We also found that streamflow forecasts using post-processed precipitation forecasts have longer lead times and higher accuracy than streamflow forecasts made by traditional Extend Streamflow Prediction (ESP) and the forecasts based on original GEFS and CFSv2 precipitation forecasts.