2003 Volume 81 Issue 4 Pages 679-696
This study explores the impacts of three different ensemble configurations on the predictability of typhoon-related precipitation. These ensemble configurations are constructed by; (1) using a single model starting with physically initialized ECMWF analyses assimilating five different precipitation estimates and with an ECMWF analysis itself (multi-analysis, MA), (2) using a single model with six different convection schemes and initialized with ECMWF analyses (multi-convection, MC), and (3) six operational forecast models initialized with the centers’ own analyses (multi-model, MM). Ensemble precipitation forecasts are verified with help of satellite-derived precipitation estimates. Typhoon-related precipitation systems are best predicted by the MM configuration, from both deterministic and probabilistic viewpoints. The single model ensemble setups (MA and MC) contain more biases than the MM setup. A weighed ensemble, the so-called superensemble (SE), technique shows a slight increase in forecast skills, compared to the bias-corrected ensemble.