不同降水卫星数据反演降水量精度评价——以雅鲁藏布江流域为例

Accuracy Assessment for Two Satellite Precipitation Products:Case Studies in the Yarlung Zangbo River Basin

利用多种定量指标和分类指标,评估PERSIANN-CDR(Precipitation Estimation from Remotely Sensed Information using Artificial Neural Netw orks-Climate Data Record)和TRM M 3B42 V7 (Tropical Rainfall M easuring M ission 3B42 V7)两种降水卫星产品在雅鲁藏布江流域的反演精度,并首次在雅鲁藏布江流域使用降水量体积分类指标对卫星数据的适用性进行评价。结果表明:(1)降水卫星数据的偏差主要表现在微量和重度降水的偏差上,两种降水卫星数据总是高估了弱降水,低估了强降水,PERSIANN降水卫星要比TRM M降水卫星对降水数据偏离程度小;(2) PERSIANN-CDR降水卫星数据与地面实测数据的相关系数为0. 663,偏差为0. 845,TRMM 3B42 V7降水卫星数据与地面实测数据的相关系数为0. 666,偏差为0. 579。只考虑定量指标的评价体系,两个降水卫星数据的精度差异相对较小;(3) PERSIANN-CDR卫星数据在各站点的各分类指标数值范围均比TRMM 3B42 V7卫星数据的指标数值范围大,PERSIANN-CDR卫星数据对降水事件和降水量的反演精度要高于TRMM 3B42 V7卫星数据。考虑降水量分类指标的评价指标体系比单纯使用传统定量指标评价降水卫星数据更能有效地反映出降水卫星对资料稀缺的高寒地区地面降水特征的捕捉能力。

Precipitation plays a very important role in the hydrological cycle and energy balance in many basins.In the Qinghai-Tibet Plateau,good quality precipitation data are unavailable and not easily accessible.Satellite precipitation data are being extensively used for alternative data sources for the ungauged basin.The uncertainties of satellite products over a certain region are largely unknown for systematic and random error.Therefore,satellite products need extensive validation before using in water resources research and streamflow simulation.Based on daily precipitation data of meteorological stations in the Yarlung Zangbo River Basin during 1998-2015,the accuracy was assessed and compared between PERSIANN-CDR(Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks-Climate Data Record)and TRMM 3 b42 V7(Tropical Rainfall Measuring Mission 3B42 V7)satellite precipitation products against quantitative indices and classification indices.The volumetric classification indices including VHI(Volumetric Hit Index,VHI),VFAR(Volumetric False Alarm Ratio,VFAR),and VMI(Volumetric Miss Index,VMI)are used to evaluate precipitation products in the Yarlung Zongbo River Basin for the first time.The results show that:(1)The deviation of light and heavy precipitation is the main expression for the deviation of satellite products,and the satellite data always overestimate precipitation in areas where precipitation is small and underestimates precipitation in areas where precipitation is great.The deviation of PERSIANNT is smaller than the deviation of TRMM.(2)The correlation coefficient of PERSIANN precipitation satellite data is 0.663,and the deviation is 0.845.The correlation coefficient of TRMM precipitation satellite data is 0.666,and the deviation is 0.579.The accuracy for PERSIANN and TRMM satellite data are similar only by using quantitative indices.(3)Overall,the values of POD,VHI,FAR,VFAR,and VMI are 0.830,0.927,0.54,0.33,and 0.073 respectively for PERSIANN.The values of POD,VHI,FAR,VFAR,and VMI are 0.674,0.833,0.50,0.31,and 0.167 respectively for TRMM.PERSIANN satellite has better performance than TRMM satellite on each site against classification indices.The evaluation system considering volumetric indices provides additional information beyond the traditional quantitative indices that can be useful in evaluating satellite products in plateau alpine region.This study is very meaningful and it could provide useful information for water resource management,water allocations and hydrologic simulation in the Yarlung Zangbo River Basin.