Multi-Model Integration Technology for Probabilistic Forecasting of Short-Time Severe Rainfall
Article
Figures
Metrics
Preview PDF
Reference
Related
Cited by
Materials
Abstract:
Based on the grided precipitation analysis data from April to August in 2018, and precipitation forecast data of four regional high-resolution models, percentile frequency matching is applied to calibrate the hourly precipitation frequency of each model. Based on the verification of key areas, the technology of dynamic weighted multi-model integration shows its application in probabilistic forecasting of short-time severe rainfall. The results demonstrate that percentile frequency matching displays positive effects in heavy rain precipitation, which can reduce the systematic forecast bias of the model and improve the accuracy of precipitation forecasting. In the nowcasting leadtime, the precipitation information of the upstream key areas is suggestive of the precipitation forecast in the downstream area. The dynamic weighted integration provides more skillful and stable performance in probabilistic forecasting than the equally weighted integration. The case also illustrates that the new technology performs well on the short-time severe precipitation forecasting and early warning within the nowcasting lead time.