Analysis of characteristic bias in diurnal precipitation variation forecasts and possible reasons in a regional forecast system over Beijing area

In order to investigate the characteristic bias in diurnal precipitation variation forecast and the possible reasons in the Rapid Update Cycle Forecast System (BJ-RUCv2.0) in Beijing area, auto-weather station and radiosonde observations were used to evaluate the 3 km resolution forecast of BJ-RUCv2.0 over Beijing region during June to August in 2012 and 2015. Regional features and propagation characteristics of the bias in diurnal precipitation variation forecast were studied, and the responsive mechanism of bias in local circulation forecast was revealed. Research results show that nighttime precipitation was missing in most cycles of precipitation forecast of BJ-RUCv2.0 in the plain area of Beijing. Specifically, rainfall in the model forecast in western mountainous area was overestimated; it was difficult for precipitation to develop in the plain area in BJ-RUCv2.0; precipitation forecasted by the model was too much in mountainous area in the evening but too low in plain area at midnight. The local circulation forecast bias and its responsive mechanism demonstrated that an intense temperature gradient between mountainous and plain area occurred earlier than normal due to the warm and dry forecast bias during the daytime near the surface layer in plain area and moist bias in the middle level in mountainous area, which caused more precipitation in the western mountainous area. The warm bias strengthened after sunset and northerly winds prevailed over plain area, which resulted in a northward shift of the topographic convergence between mountainous and plain area and subsequently affected the moving of rain belt from mountain to plain area. There existed a continuous dry bias in plain area, which prohibited the development of rainfall in plain area and caused the missing forecast of precipitation. Soil moisture of GFS, which was used to provide initial condition for cold start of BJ-RUCv2.0, is noticeably drier than observations, which might contribute to the warm and dry bias in the surface layer.