中高纬度地区TRMM卫星降雨数据的精度评价

卫星降雨产品在降雨空间格局分析中扮演着重要角色，其中TRMM 3B42（3B42）是重要的产品之一，具有适宜的空间和时间尺度，已成为生态、气候、水文模型的重要驱动因子.当前对3B42 V7新版本数据的精度和误差认识仍然有限，尤其缺少中高纬度地区的数据精度评价研究，制约其在生态、气候和水文等领域的应用.本文基于内蒙古自治区1998—2012年3B42 V7产品和53个气象站点数据，评估其在中高纬度地区的精度，揭示了各精度评价指标的空间分布特征及其影响因子.总体精度评价结果表明： 3B42在年、月、日尺度上均高估降雨量，且误差随时间尺度增大而增加，日降雨量的平均误差和平均绝对误差分别为-0.06和0.88 mm；降雨事件侦测能力表现良好，公正预兆评分达到0.23，但对降雨事件总频率存在微弱高估.数据精度受海拔和多年平均降雨量影响，随海拔升高，降雨绝对误差减小，降雨事件探测能力减弱，多年平均降雨量对数据精度的影响则相反.

Accuracy evaluation of the TRMM satellite-based precipitation data over the mid-high latitudes.

Satellite-based precipitation product plays a significant role in analyzing spatial pattern of precipitation. TRMM 3B42 (3B42), as one of important precipitation products, is a key forcing factor for ecological, climate and hydrological models with its sufficient spatial and temporal scales. At present, there is still limited knowledge of accuracy and error structure of new version 3B42 V7, especially the lack of evaluating its accuracy in regions of mid-high latitudes, which restricts its application in the field of ecology, climate and hydrology. Based on 3B42 V7 product and data obtained from 53 weather stations in the period of 1998-2012 over Inner Mongolia, the paper evaluated the accuracy of 3B42 and revealed the spatial distribution of accuracy evaluation indices and its influencing factors. The result of overall accuracy evaluation indicated that there was overestimation of the amount of precipitation by 3B42 at daily, monthly and annual timescales. Error of 3B42 increased with the increasing timescale, and mean error and mean absolute error of daily precipitation were just -0.06 and 0.88 mm, respectively. It showed good performance for detecting the occurrence of rain events with equitable threat score (ETS) of 0.23, but slightly overestimated the frequency of rainy events. Moreover, altitude and average annual precipitation had impacts on data accuracy. The absolute error of precipitation decreased and yet the ability of detecting rain events became weak with the increasing altitude, while the effect of average annual precipitation on data accuracy was opposite.