Research On Precipitation Nowcast By Blending The Radar Extrapolation And Numerical Weather Prediction Model Products

Precipitation nowcast has been a realistic difficult problem.Although radar extrapolation is mature,it cannot accurately predict the trend of precipitation system.Numerical weather prediction assimilating Doppler weather radar data to improve the initial field is not perfect yet,it has its own "spin-up" problem.Therefore,the blending of the two prediction methods is an important direction for the research of precipitation nowcast in recent years.This paper intends to research from three aspects: radar extrapolation,numerical weather prediction and both of them.A new blending method that can be used for operational forecast has been proposed,providing reference value for precipitation nowcast.The main contents and conclusions are as follows:(1)The optical flow method is applied to four different types of precipitation processes in Guangdong Province.The results show that the optical flow method can forecast the large-scale stable precipitation system or typhoon with rotational motion characteristics for up to 2 hours.The forecast for strong convective system is at least one-hour lead,but the prediction of intensity is still to be improved.This is because the optical flow algorithm just considers the change of the echo motion vector and obtain the position of the echo accurately,ignoring the development of convection in the vertical direction.Quantitative analysis was performed using the forecast scoring method.It shows that the accuracy of all cases forecast is decreasing with the increase of forecast duration.(2)The mesoscale numerical weather prediction model called WRF and its three-dimensional variational assimilation system were selected to predict the four cases described above.The results show that: the numerical forecast is more accurate in the area of precipitation and can grasp the changing trend of the precipitation system.However,there are errors in precipitation intensity and position,the rainfall estimating generally too large.Comparing the results of numerical weather prediction and radar extrapolation,it can be seen that the numerical forecast performs better than radar extrapolation on whether it rains.For different precipitation thresholds,the deviation of radar extrapolation is relatively stable as a whole,while with the increase of precipitation threshold,the larger the deviation of numerical weather prediction is.As the model cannot accurately forecast precipitation intensity,further correction techniques for numerical forecast intensity errors need to be considered.(3)The blending of radar extrapolation and numerical weather prediction is a feasible way to improve precipitation nowcast.A new dynamic weight blending algorithm called harmony search was developed.Two precipitation processes in Guangdong Province have been selected to analyze and examine.Compared with the Hong Kong observatory's hyperbolic tangent weight blending algorithm,the experimental results show that the effect of the blending is better than radar extrapolation or numerical weather prediction to a certain extent.The harmony search is closer to observations,which provides a certain reference value for the future research on the development of precipitation nowcast technology.