Merging statistical and dynamical forecasting models to improve seasonal streamflow forecasts in eastern Australian catchments

Seasonal streamflow forecasts are valuable for planning and allocation of water resources. Statistical forecasting methods generally use predictors related to catchment wetness at the start of a forecast period and climate over the forecast period. The predictors include antecedent streamflow and rainfall to represent catchment wetness and lagged climate indices to represent the climate. The predictors are selected on the basis of their best predictive performance in cross validation over historical data. In this study, we test two strategies aimed at improving the seasonal streamflow forecasting capabilities. The first strategy involves combining multiple candidate models using the Bayesian Model Averaging (BMA) approach. The BMA approach combines strength of candidate models and accounts for model uncertainty. The second strategy is to incorporate rainfall forecasts from a dynamical climate model into the existing forecasting system. The dynamical models simulate physical processes and capture concurrent relations between SST anomalies and climate at the forecast region. We test both strategies in 22 catchments located in eastern Australia. Our results show that combining multiple statistical forecast models generally yields more skilful forecasts than the 'best' model. The addition of rainfall forecasts from a dynamical climate model only marginally improves the streamflow forecasts.