Abstract Author: Shraddhanand Shukla, F. Munoz-Arriola, T. Bohn, A.C. Steinemann, and D.P. Lettenmaier
Abstract Title: Assessment of ESP based Drought Prediction Skill
Abstract: Drought is a slowly developing natural phenomenon which often results in large economic losses. The repeated occurrence of drought over different regions in the U.S. has motivated the development of risk based drought management approaches. A key consideration in applied agricultural and hydrologic drought prediction is that much of the potential skill, particularly at lead times up to a few months, derives from knowledge of hydrologic initial conditions – primarily soil moisture and snow water equivalent. Improved (and more objective) nowcasting of droughts is possible using gridded observation fields (primarily of precipitation and temperature) to force macroscale hydrological models, whereas improved drought forecasts can be made by utilizing the nowcasts as initial conditions for ensemble prediction methods like that ESP (ensemble streamflow prediction) method of the National Weather Service’s Office of Hydrologic Prediction. This approach also allows quantification of forecast skill via uncertainty attribution, which is largely associated with uncertainty in the nowcast for short lead forecasts, and weather and climate prediction uncertainty at longer lead times. The University of Washington Surface Water Monitor (www.hydro.washington.edu/forecast/monitor) is a web based hydrologic monitoring and prediction system which uses a land surface model forced with observed precipitation and temperature to simulate the current status of soil moisture and runoff conditions across the country. In this work we explore how well this system can be utilized for drought prediction at time scale varying from weeks to a season. Utilizing an ESP-based drought prediction approach, we have assessed skill for identification and forecasting of drought events in 1988 (Central and Eastern U.S.) 1999 (Eastern U.S.) 2002 (Western U.S. and Great Plains) 2005 (Western U.S.), 2006 (Southwest U.S.) 2007 (Southeast and Western U.S.). We compare the monthly to seasonal ESP forecasts with model-based reconstructions of soil moisture and observed streamflow during those drought events. We assess variations in prediction skill strictly associated with the initial hydrologic conditions, seasonally and across continental United States.