基于Triple-collocation方法的多源降水产品评估

多源降水产品在过去十几年来得到了广泛的应用,然而在数据缺失区其可靠性检验的难度较大。采用Triple-Collocation(TC)方法可以在未知真值的情况下,对三组相互独立的产品进行交叉评估。基于TC方法,综合评估了卫星遥感产品、再分析产品与地面实测数据集在中国区域的不确定性。研究所选空间分辨率为0.25°,时间分辨率为日尺度,数据选用时段为2013—2015年。首先,在中国南方对TC方法进行可靠性验证。通过与传统方法所得结果(相关性系数CC与均方根误差RMSE)对比,相对偏差仅为4.5%与3.0%,而后将方法应用于整个中国大陆区域(除台湾地区),并重点关注青藏高原数据缺失区。结果显示:在卫星产品中,IMERG整体表现最优;在青藏高原无台站网格点上,IMERG与ERA-interim所得平均CC分别为0.63与0.61,略优于地面实测数据集(0.58),这表明卫星遥感产品与再分析产品在青藏高原区域具有一定的优势和应用潜力;实测数据在台站稀疏区域的高不确定性也得到了进一步验证。研究成果证明了TC方法在中国大陆区域的适用性和可靠性,为数据缺失区的降水产品评估提供新的思路和方法。

Triple-collocation method-based assessment on multisource precipitation product

It is already more than ten years since the multisource precipitation products are widely used，whereas the verifications of their reliabilities are more difficult in the regions without any relevant data. With Triple-collocation ( TC) method，cross-assessment can be made among 3 set mutually independent products under the condition of unknowing the true values. Base on TC method，the uncertainties of the satellite remote sensing product，the re-analyzed product and the ground measured data set within the area of China are comprehensively assessed herein; for which 0. 25° is selected for the spatial resolution，Ldiurnal scale is selected for the temporal resolution and the selected time interval for the data is 2013—2015. At first，the reliability of TC method is verified in the South China. The relative errors of the result ( the correlation coefficient CC and the mean squared error ＲMSE) are only 4. 5% and 3. 0% through comparing with those obtained from the conventional method，and then the method is applied to the whole area of China ( except Taiwan region) ，especially focuses on the region of the Tibetan Plateau without any relevant data. The result exhibits that the overall performance of IMEＲG is the best within all the satellite products and the mean CCs obtained from IMEＲG and EＲA-interim on the grid points without any stations in the Tibetan Plateau are 0. 63 and 0. 61 respectively，which are slightly better than that from the ground measure data set ( 0. 58) . It is indicated that both the satellite remote sensing product and the re-analyzed product have certain superiorities and application potentials for the region of the Tibetan Plateau，while the high uncertainty of the measured data within those regions with few stations is further verified as well. The applicability and reliability of TC method applied to the area of China are demonstrated by the study result，thus new idea and method are provided for the assessment to be made on the multisource precipitation products for those regions without any relevant data.