Spatial Deviation Analysis of Heavy Precipitation Forcast of Large Scale Numerical Weather Prediction in Eastern Northwest China

Quantitatively examining the spatial deviation of precipitation forecast in numerical models is important to improve the model forecast performance and enhance the ability of forecasters correcting the model forecast products subjectively.Based on encrypted precipitation observations and 24-hour precipitation data forecasted by the ECMWF， GRAPES-GFS large-scale numerical models with 36-hour limit from May to September of 2019 -2020 over the eastern Northwest China， 51 heavy precipitation targets are identified by using the Contiguous Rain Area technique.According to the features of 500 hPa circulation system， these targets are classified into low trough type， southwest flow type， and two high shear type.The spatial deviation， including displacement， intensity， and pattern deviation， of heavy precipitation forecast in the two models are analyzed quantitatively.The results are as follows： （1） The pattern deviation of precipitation forecast in ECMWF and GRAPES-GFS account for the largest proportion of the total deviation， which reach 65.99% and 63.72%， respectively， corresponding with 21.55% and 20.31% displacement deviation， and 12.47% and 15.97% intensity deviation.Pattern deviation are mainly related to the activities of the small-and medium-scale convective systems in the warm zone.（2） The heavy precipitation areas in two models shift more northwestward compared with observations.Forecast deviation largely depend on the scales of precipitation areas.The precipitation forecast of medium-β scale heavy precipitation areas in ECMWF is better than GRAPES-GFS， and the medium-α scale rainfall areas are overestimated in two models.The deviation of maximum precipitation and mean rainfall intensity forecasted by ECMWF are less than 6%， while are more than 33% in GRAPES-GFS.（3） The maximum precipitation in low trough type is underestimated by ECMWF， and the area and mean rainfall intensity vary with the precipitation scale.For southwest flow type， ECMWF's forecast indicates that the area and mean rainfall intensity are slightly larger， while the maximum precipitation is weaker than observations.The rainfall area in two high shear type forecasted by ECMWF is most easy to underestimated and even missed， but reasonable results can be seen in mean rainfall intensity and maximum precipitation.The area， average rainfall intensity， and maximum precipitation in the two high shear type and southwest flow type are underestimated by GRAPES-GFS.（4） For the displacement deviation in ECMWF， these are northwestward shift for low trough type and southwest flow type and northward shift for high shear type.However， in GRAPES-GFS， there are northwestward shift for low trough type and westward shift for southwest flow type.The displacement deviation are not obvious in two high shear type.