Forecast Error Source Analysis of Heavy Rainfall Events in the Meiyu period of 2020

Based on the WRF (Weather Research and Forecasting) model, this study investigates the forecast error sources of eleven heavy precipitation events during the Meiyu period in 2020. Forecasts using the WRF model were generated with the initials of NCEP_FNL (Final Global Data Assimilation System) and TIGGE_EC (THORPEX Interactive Grand Global Ensemble from European Centre for Medium-Range Weather Forecasts) and were subsequently compared. Results indicated that the TIGGE_EC initially exhibited higher forecast skills, demonstrating that the uncertainty of the initial conditions heavily influences the forecast results. Next, the regions (sensitive areas) and variables (sensitive variables) where the uncertainty of the initial errors originated were explored. It was observed that the sensitive areas were concentrated in the upper reaches of the western side of the precipitation area. The proportions of the kinetic, effective potential and specific humidity energies in the total energy of the initial error indicated that the specific humidity energy occupied the smallest proportion; however, the sensitivity experiment demonstrated that the disturbance in a specific humidity field has the highest impact on the forecast. Therefore, seven rainstorm events were selected wherein the specific humidity field had the greatest impact on the forecasts and the water vapor source and path of the maximum accumulated precipitation point were tracked using the HYSPLIT backward trajectory model. Of the seven rainstorm events, six events showed the water vapor source channels moving upstream to the west of the precipitation area. Therefore, the upstream water vapor change in the west of the precipitation area was a valuable source for forecast errors in the heavy precipitation events during the Meiyu period in 2020, and its accurate description could help improve the rainfall forecasts.