Examination of the Value of QPF for Streamflow Forecasting

Abstract:

Forecasting how much water will flow in a stream or river requires mathematical models that convert observations of rainfall into runoff and streamflow. These models account for how much of the rain can be absorbed by the ground and how fast the water is moving on the landscape and in the river channel. The National Weather Service uses such models to forecasts streamflow and floods everywhere in the nation. These models forecast streamflow few hours or few days ahead, depending on the size of the basin. The smaller the basin the faster the streamflow may change making the forecasts more difficult. To extend that forecasts lead time (horizon), and thus give users more time to make decisions, the models can use the forecasted future rainfall. The problem is, as we all know, that weather forecasts, and in particular rainfall forecasts, are highly uncertain. In this study we will explore to which degree the errors in rainfall forecasts affect the streamflow forecasts. Are larger rivers able to “tolerate” error in the rainfall forecasts better than the small stream and river? How much better would the streamflow forecasts be if we knew the future rainfall perfectly? Addressing these questions for the entire country requires analysis of very large past data sets. We expect to gain significant insights into the forecasting skills by the National Weather Service models, and its dependence on basin size, lead time, accuracy of the rainfall forecasts, effectiveness of various corrective schemes, and more. These insights will guide development of the next generation models and establish targets for the rainfall forecasting performance.