Operational Forecasting with Real-Time Databases

Abstract

An integrated hydrometeorological system for flow forecasting has been designed and implemented for real-time use in an operational environment at the U.S. Army Corps of Engineers Rock Island District, Illinois. The system operates routinely with data collected by real-time data-collection platforms and stored in a real-time database. Forecasts are based on coupled hydrological and meteorological models, and a state estimator that updates system states in real time using all current data. Five soil-water models are used to estimate rainfall abstractions: a generic API model, the modified Sacramento soil moisture accounting model, and three soil-water options in the HEC-1 model. Initial sensitivity analysis of the short-term system forecasts with respect to the soil-water model and the state-updating components indicates robust system behavior in short-term real-time flow prediction. In addition, sensitivity analysis shows that when updating is performed using all available operational data, the differences in results among soil-water models are small. Thus, selection of a soil-water model to estimate hydrologic abstractions may be based on user familiarity with a certain model and the estimation of its parameters.