On the application and improvement of the frequency matching method to rainstorm forecasts

Based on the daily precipitation forecast data of European Centre for Medium Range Weather Forecasts and National Center for Environmental Forecasting from May to August in 2017, combining with observation data during the same time period, the frequency matching technique is used to correct the model bias of summer precipitation forecast in Jianghuai River basin, and improvement to the scheme is also discussed. The following four conclusions are drawn from this study. (1) The model has significant false alarm forecast of light rain and missed forecast for heavy rain and above. (2) By lowering the precipitation with low intensity and raising the precipitation with high intensity, the intensity and area forecast skills in different precipitation magnitudes are improved to some extent. The correction effect is more obvious at both ends of precipitation magnitude, the accuracy of light rain and torrential rain forecast is significantly improved after the bias correction. (3) The improvement effect of rainstorm area forecast depends on the precipitation type. Typical case analysis shows that: for the large area Meiyu front rainstorms with more accurate forecast of rain belt location, the rainstorm TS score is significantly improved after bias correction. For rainstorms in a small area at the periphery of the subtropical high and the Meiyu front rainstorm or typhoon rainstorm with inaccurate forecast of rain belt location, the improvement of rainstorm TS score is not obvious or even lower after the bias correction. (4) To solve the above problems, the improvement ideas of dynamic coefficient adjustment and model integration are put forward, i.e., for small area rainstorm with average rainfall of 5 mm or less, the coefficient should be appropriately increased, while for large area rainstorm with average rainfall of 15 mm or above, the coefficient should be lowered appropriately. Model integration can effectively improve the rainstorm TS score.