Application and Verification of Convective Scale Ensemble Forecast for a Heavy Precipitation Event That Occurred in Eastern Southwest China

Abstract

This study aims to provide forecasters with valuable and practical insights into the effective application of convective-scale ensemble forecasts for precipitation prediction. Statistical verification and subjective analyses were conducted on the forecast performance during a heavy precipitation event in eastern Southwest China. The results indicate that different postprocessed deterministic forecast products each have distinct advantages and limitations that forecasters should consider. The ensemble mean forecast (EMF) has shown strengths in forecasting small magnitude precipitation (i.e., light rain, moderate rain, and heavy rain events), but it tends to smooth out information regarding extreme precipitation. The probability-matched EMF (PMEMF) outperforms the EMF for extreme precipitation predictions. In general, optimal ensemble quantile forecasts outperform the corresponding EMFs and PMEMFs, as well as most individual ensemble members, but notably, the optimal quantiles vary significantly across different cases. The ensemble forecast system is capable of predicting certain probabilities of heavy rainstorms and extraordinary rainstorm events as early as 4 days in advance. Based on the verification results, it is recommended that forecasters should remain cautious even when only a single or few ensemble members predict extremely heavy precipitation (or whether a certain probability of extreme precipitation exists, even if it is relatively low), thus helping to reduce decision-making errors. Furthermore, probabilistic forecasting should be more comprehensively and effectively applied in China.