Improving statistical streamflow predictions using a water balance model for operational seasonal forecasts

In response to the need for a seasonal water availability prediction service to provide valuable Information for water allocation decisions, water markets, reservoir operation, environmental flow management and drought response strategies, the Bureau of Meteorology initiated a Seasonal Streamflow Forecasting (SSF) service in December 2010. This service delivers probabilistic forecasts of streamflow volume for the next three months at monitoring stations or total inflows into major water supply storages. The forecasts inform a range of water managers and users in their planning and management decisions, and to ensure security of water supply, including the setting of supply restrictions. The service currently uses a statistical prediction system based on a Bayesian joint probability model (BJP version 1} to provide reliable seasonal streamflow forecasts at 70 locations. The model uses relationships between forecast streamflows and initial catchment conditions (antecedent monthly streamflow and rainfall), El Nino-Southern Oscillation (ENSO) indices and other climate indicators. The prediction system includes a verification system .that measures skill and reliability of the forecasts. A dynamical monthly water partition and balance model (WAPABA) has been developed and integrated Into the statistical forecasting system to improve its forecast performance. The dynamics! model simplifies the predictor selection process, produces consistent predictors, moderates the errors of the worst forecasts by using only the best climate index in the current model, and most importantly, takes into account the important influence of catchment soil moisture storage when forecasting streamflow. This improved BJP model (version 2) is now being integrated as part of the Bureaus water availability forecasting system (WAFARi). This paper presents the major features of the improved modelling approach audits application to the SSF operational service.