Analysis Of Snowfall Detection Capability Of GPM DPR

Snowfall is an important freshwater resource and an important part of the global climate system.It is closely related to people's life and production.It is of great significance to study the spatial distribution,frequency and intensity of global snowfall.Following the PR(Precipitation Radar)on the TRMM(Tropical Rainfall Measuring Mission)satellite,the GPM(Global Precipitation Measurement)core observatory carries the first spaceborne dual-frequency precipitation radar(DPR),operating at Ku and Ka bands(13 and 35 GHz,respectively).Its ability to detect snowfall has attracted much attention.This Paper analyses a snow index(SI)algorithm to identify surface precipitation phases based on GPM-DPR detection taking three processes that occurred in China as examples,including snowfall,summer mixed rainfall and winter stratiform rainfall.The algorithm for calculating SI is improved.And then,the snowfall estimates from DPR_NS,DPR_MS and DPR_HS are evaluated using the observation data from 2168 automatic stations in China for the period of November-April of 2015 until 2016.Finally,a large-scale rain-snow process in Anhui on January 3,2018 is comprehensively analyzed using GPM-DPR and GMI(GPM Microwave Imager).The main conclusions are as follows:(1)SI is calculated using three ingredients of DFR_m slope,maximum value of reflectivity at Ku-band,as well as storm top height(STH).It is found that if the auxiliary information(0°C isotherm height,etc.)is not used,it is easy to identify summer rainfall and snowfall using the original SI formula,but difficult to identify winter rainfall and snowfall.Therefore,the quality control method of reflectivity factor is improved and the slope of DFR_m vertical profile is calculated by the least square method.The results show that the identification results of the improved method are in good agreement with the ground automatic station observation data.(2)The ground validation of DPR retrivel snowfall rate shows that the false alarm rate(FAR)of DPR is higher than 50%.The probability of detection(POD)is very low(about8%-10%),because the snowfall echo is seriously affected by DPR sensitivity and ground clutter as relatively low echo and storm top height.The POD of NS and HS are relatively higher and the FAR of HS is the lowest.In addition,the snowfall rate estimates of DPR_NS,DPR_MS and DPR_HS tend to underestimation.The Bias and Root Mean Square Error(RMSE)of DPR_NS estimates are the smallest.The Bias of DPR_MS is the largest,but the correlation of MS is best.DPR can effectively detect heavy snowfall with an hourly snowfall which is more than 1.6 mm/h,while the ability to detect small to medium snow is weak.The POD of DPR_NS,MS and HS for different intensity snowfall is different.DPR_HS has a strong ability to detect light snow.(3)A comprehensive analysis of the large-scale rain-snow process in Anhui on January 3,2018 using various data shows that DPR observations can well shows the horizontal and vertical distribution characteristics of the precipitation.The surface snowfall echo at Ku-band is flaky.The echo intensity is low and uniform.The echo in vertical structure gradually increases from top to bottom.The strong echo center is near the ground.The echo intensity of Ka_HS is weaker than that of KuPR and Ka_MS.Compared with uncorrected radar reflectivity factors of KuPR,Ka_MS and Ka_HS,it is found that the STH of Ka_HS is higher than KuPR and Ka_MS because of higher sensitivity.STH for the corrected radar reflectivity factors of KuPR,Ka_MS and Ka_HS is similar.The Advantage detecting snowfall by Ka_HS is not obvious.Compared with the observation results of ground automatic station,the snowfall rate estimates of DPR_NS are underestimated.GMI overestimates the snowfall rates for surface weak snowfall and underestimates the snowfall rates for the strong snowfall.Compared with DPR_NS retrieval results,GMI detected snowfall in Bozhou undetected by DPR.The retrieved snowfall rates by GMI are about 0.8 mm/h higher than DPR_NS estimations.The error of GMI in detecting snowfall is less than that of DPR_NS for this case.