四川省2013年夏季卫星降雨数据的对比研究

Comparison of Two Different Satellite Precipitation Data in Sichuan from May to August in 2013

选取2013年5 8月CMORPH version1.0、TRMM3B42 version6两种卫星降雨产品,分别在3 h,1天及1个月三种不同时间尺度上,结合同时段四川省157个地面站点观测的降雨数据,对比评估了两种卫星降水数据对降雨量和降雨事件探测的效果。对比研究中,根据站点的位置选择与之空间距离最近的卫星降雨格点作为对比对象,选用相关系数评估卫星数据的同步效果,选用偏差BIAS评估卫星降雨的系统误差和相关程度,选用均方根误差RMSE评估卫星数据的偏差。结果表明,3 h时间尺度上,两种降雨产品与地面站点的同步效果都很差,都不能很好地反映四川省降雨信息,但TRMM3B42 V6的系统误差比CMORPH小,与站点相关程度更高;在1天的时间尺度上,两种卫星降雨数据与站点数据的同步效果都有所增加,TRMM3B42 V6降雨产品与站点降雨量偏差更小,相关程度更高,表现出更强地一致性;在1个月的时间尺度上,TRMM3B42 V6对降雨量的探测表现出一定程度的低估,但同CMORPH产品相比,TRMM3B42 V6降雨产品仍表现出与站点降雨更高的一致性,而CMORPH对降雨量的探测则明显偏高,可能原因是由于CMORPH产品利用红外数据进行了移动矢量计算,而对流云对云顶亮度温度有较大影响,从而导致CMORPH探测的降雨量偏高。研究还选用了发生概率和部分分数及Heidke评分标准来确认卫星降雨产品能否反映降雨的发生,发现CMORPH在探测降雨事件发生时优于TRMM3B42 V6降雨产品。

This research chooses two satellites＇ precipitation products which are the newly CMORPH version1.0and TRMM3B42 V6,using these two products to estimate the precipitation and events of precipitation in Sichuan province from May to August 2013 under three different temporal resolution which are 3 hours,daily and monthly.Using gauge data＇s location to choose the corresponding nearest data of satellite products,this method is different from before those which to resample and interpolate the pixel data,in this way the error may decline.This research chooses correlation coefficient to study the effect synchronous of precipitation algorithms,BIAS to study precipitation algorithm＇s systematical error and degree of correlation,RMSE to study the deviation of precipitation algorithm.Finding that both of the two products have bad synchrony with gauge data and can＇t reflect the precipitation information under 3 hours temporal resolution,but TRMM3B42 V6 has less systematic error and corresponds with gauge data better than CMORPH.For both of the two products are origin from different satellite data,each of these initial data may has its own error,so the error may still exist after merging them to the new product;The initial satellite data consist geosynchronous data,the shortage of geosynchronous satellite data is the period,if the period is long or short enough,the satellite may get the wrong data,these two reasons may cause the error of the final precipitation products.Both of the two products＇ synchrony are increasing when the temporal resolution comes to daily,TRMM3B42 V6＇s bias is less than CMORPH,it corresponds with gauge data better than CMORPH and has better synchrony with gauge data.TRMM3B42 V6 underestimates the precipitation and CMORPH overestimates the precipitation under monthly temporal scale,and TRMM3B42 V6 has higher correlation with gauge data than CMORPH,because CMORPH has use infrared data to get motion vector,and convective cloud can influent the brightness temperature of the cloud,when it overestimates brightness temperature of the top cloud will cause precipitation＇s overestimation.As for estimating the events of precipitation,this research uses three coefficients to analysis this issue,and they are：probability of detection（PD）,frequency of hit（FH）and Heidke skill score（HSS）.Form these three coefficients can know that CMORPH is better than TRMM3B42 V6 at detecting the events of precipitation.