Comparative Analysis of Precipitation Forecasts in China Between Two Initialization Times of the ECMWF Integrated Forecasting System (IFS)

Abstract

Diagnosing the precipitation forecast biases in numerical weather prediction (NWP) is an important step and method to improve the accuracy of precipitation forecasts.This study utilizes deterministic precipitation forecast data from the ECMWF Integrated Forecasting System (IFS), along with precipitation observations from 2415 meteorological stations in China and ERA5 data, to analyze the characteristics and possible causes of precipitation forecast biases in the IFS over China. Methods include, such as Root Mean Square Error (RMSE), Mean Error (ME), Frequency Bias (FBias), forecast verification scores, and Integrated Water Vapor Transport (IVT) diagnostics. The main conclusions are as follows: (1) The precipitation forecast from the IFS exhibits a significantly high Fbias with a pronounced wet bias. The ME shows larger positive deviations under lower daily average precipitation, transitioning to negative errors as daily precipitation increases. The RMSE of precipitation demonstrates distinct spatiotemporal variations: higher values are observed during summer half-year compared to winter, and in southern regions versus northern areas. The influence of underlying surface topography on RMSE is clearly evident. (2)The RMSE and ME increase with increasing forecast lead time. The RMSE for 1200 UTC precipitation forecasts performs better than for 0000 UTC beyond 48 hours. (3) The Threat Score (TS) and Equitable Threat Score (ETS) scores for summer half-year precipitation forecasts are higher than in the winter half-year, with 1200 UTC initial forecast times being higher than 0000 UTC, indicating that the true forecast capability of 1200 UTC initial forecast times is better than for 0000 UTC. (4) The forecasts initialized at 0000 UTC exhibit overestimated southerly wind components and stronger vertical velocity, with larger standard deviations in both vertical motion and IVT. In contrast, the forecasts initialized at 1200 UTC show amplified easterly wind biases yet demonstrate smaller IVT discrepancies. These systematic differences in dynamical and moisture variables may be important reasons for the differences in precipitation forecasts between the two initialization times.