Evaluation Of The Convective-Permitting Precipitation Forecasts In China During 2016 Summer Season And Analysis Of Precipitation Forecast Error

High resolution model is the development trend of numerical weather prediction.However,the lack of understanding of high resolution model deviation and its causes seriously hinders the improvement of high resolution model.In this paper,we will study how to use high resolution observations to reveal the main deviations of high resolution model.On the basis of deviation recognition,an objective evaluation scheme based on combination of neighborhood and different scales is proposed.It is pointed out that the reason for the low score of high resolution model for small scale systems in traditional evaluation is that the influence of intensity and position error is not taken into account The cause of the position deviation of the afternoon precipitation forecast along the coast of South China is further analyzed.It is pointed out that the deviation of the sea-land wind forecast is the main reason for the deviation of the local wind direction forecast and the precipitation forecast.The main contents of this paper are as follows:Firstly,the precipitation forecast of 4 km and 12 km WRF model in summer season test of Nanjing University in 2016 is evaluated by using the 5 km hourly merged precipitation products with high spatial and temporal resolution in the whole country The evaluation shows that the 12 km and 4 km WRF models can well simulate the spatial distribution and diurnal variation of precipitation.Compared with the 4 km WRF model,the precipitation fine structure predicted by the 4 km WRF model is closer to the observation,and the peak value in the afternoon is also closer to the observation.12 km WRF simulation underestimates precipitation intensity,especially afternoon peak,mainly because the model with resolution 12 km relies on cumulus parameterization scheme to resolve most afternoon convective precipitation,while Grell-3 cumulus convection parameterization can not predict afternoon convection very well.The 4km WRF forecast improves the intensity forecast,but in the coastal area of South China,the intensity of precipitation forecast is stronger and the location of precipitation forecast is inclined to the coastline.The results of objective and subjective evaluation are consistent.The ETS score of 4 km WRF forecast is better than 12 km WRF,especially for heavy rainfall forecast(precipitation intensity over 25 mm accumulated for 6 hours).It is found that the scale-decomposition objective score of 4km WRF is lower than that of 12km WRF,which is mainly due to the deviation of location and intensity of the small-scale convection in 4km WRF prediction,resulting in "double penalty"phenomenon of observed-but-not-forecast and forecast-but-not-observed.In this paper,the verification method based on neighborhood method and scale separation method is proposed.On the basis of scale separation,considering the position deviation of prediction,and using the idea of neighborhood evaluation for reference,the FSS scoring of different scales based on neighborhood is calculated.On this basis,considering the intensity location of the prediction,the percentile threshold method is introduced to further improve the objecti-ve evaluation results.Using the improved evaluation method,the precipitation forecast of Meiyu season in 2016 is evaluated.The results show that the obj ective score of 4km WRF forecast in small scale(space scale is less than 100km)is higher than that of 12km WRF forecast,which is consistent with the subjective evaluation results.Finally,the reasons for the position deviation of the afternoon precipitation forecast in South China are analyzed.The wind speed and direction predicted by 1000 hPa,925 hPa and 850 hPa are evaluated by using four sounding stations in Guangdong Province.At the beginning of the forecast,the deviation of wind speed and direction between observation and model forecast field is small.The 18-hour(14 LST),24-hour(20 LST)predictions are analyzed.At 14LST,the wind direction predicted by the model is obviously deviated from the observed wind direction to the coastline.At this time,the deviation between the precipitation forecast and the observation is large.At 20 LST,the deviation between the wind direction predicted by the model and the sounding wind direction decreases,and the position of precipitation forecast is consistent with the observation.It shows that the deviation of wind direction prediction may be related to the deviation of precipitation forecast position.Further analysis shows that 4km WRF can better predict afternoon convection over mountainous areas.The main deviation is the precipitation process related to the sea-land breeze circulation.Due to the deviation of the model in simulating the thermal difference between land and sea,the deviation in simulating the sea-land breeze is the main reason for the deviation in local wind direction forecast and precipitation forecast.