The Experiment of Hybrid Ensemble Forecast Approach in Short-range Forecast for South China Rainstorm

Hybrid ensemble forecast approach composed of high-resolution and low-resolution model is tested with multi-initial conditions and multi-physics ensemble prediction system (EPS) based on ARPS and WRF. Three kinds of ensemble prediction experiments (ARPS model ensemble, WRF model ensemble and ARPS-WRF multi-model ensemble) are designed to comparatively analyze the precipitation effect between the hybrid ensemble forecast approach and traditional ensemble forecast approach based on the heavy rainfall event occurred in Southern China on 8 May 2013. ARPS model ensemble improves local precipitation simulation over southern part of Guangdong Province. The center intensity of southern ensemble mean reaches 150 mm, close to the observation, but its location shifts northeast a little. The distribution of ensemble spread is similar to that of ensemble mean. The probabilistic forecast area of the approach is close to the observation in terms of the torrential rain and downpour forecast. The Threat Score (TS) show that the greatest improvements for order of magnitude precipitation are obtained by the hybrid ensemble approach. The Brier Score (BS) also shows that the improvement on moderate rain, heavy rain and torrential rain is obvious, but the improvement on downpour and excessive storm is limited for the approach. WRF model ensemble has a better performance on precipitation simulation over northern part of Guangdong Province, but the hybrid approach is limited because of the large frequency of false alarms and misses. ARPS-WRF multi-models obviously improve precipitation simulation which has a better performance than those of traditional ensemble forecast approach on precipitation area and intensity. The center intensity of southern precipitation reaches 70 mm, less than the observation and it is caused by the weaker WRF members forecast. The distribution of ensemble spread is similar to that of ensemble mean. The probabilistic forecast of torrential rain and downpour in southern area is best up to 40%. The TS shows that the greatest improvements for order of magnitude precipitation are obtained by the hybrid ensemble approach, especially for heavy rain, torrential rain and downpour. The BS also shows that the improvement on moderate rain, heavy rain and torrential rain is obvious. ARPS-WRF multi-models have better performance than ARPS model on the forecast of downpour and excessive storm. Therefore, the hybrid ensemble forecast approach achieves high-resolution ensemble forecast system and improves the precipitation simulation combining low-resolution ensemble run with single high-resolution control model run. Meanwhile, a reference to high-resolution ensemble forecast system is provided for operational weather prediction centers.