复杂地形下降水的高空间分辨率插值方法研究  

Spatial interpolation of high-resolution daily precipitation over complex terrains

作　　者：郎立晨 唐诚[1] 高星[2] 李志慧[2] 吴锋[2] LANG Lichen;TANG Cheng;GAO Xing;LI Zhihui;WU Feng(Yantai Institute of Coastal Zone Research,CAS,Yantai 264003,Shandong,China;Institute of Geographic Sciences and Natural Resources Research,CAS,Beijing 100101,China;University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院烟台海岸带研究所,烟台264003 [2]中国科学院地理科学与资源研究所,北京100101 [3]中国科学院大学,北京100049

出　　处：《地理学报》2023年第1期101-120,共20页Acta Geographica Sinica

基　　金：国家自然科学基金项目(U2102208)。

摘　　要：复杂地形山区降水格局在多种地形要素的综合影响下呈现出显著异质性特征,弱化了降水站点观测资料的空间代表性,限制了遥感及再分析产品的适用性及传统插值方法的准确性。常用的PRISM降水插值算法通过提取并权重化地形要素,借助加权最小二乘算法对降水站点观测数据进行空间插值,被广泛应用于降水产品制备。本文针对PRISM算法对中小地形地貌刻画能力不足的问题,在解析影响复杂地形降水模式的地形要素的基础上,改进了PRISM的地形要素计算与权重化过程,同时,考虑到实际日降水量的随机性,将改进后的PRISM嵌入到“平均态日降水—比值”插值框架,构建了适用于复杂地形的日降水插值算法MPRISMR。随后,以具有复杂地形特征的元江流域为例,通过交叉验证及与ERA5-Land和TRMM_3B42降水产品的对比分析,发现该算法具有较高的准确性与可靠性。结果表明,在元江流域23个气象站点上,MPRISMR的插值结果与观测值的相关系数、相对偏差的中位数分别为0.72、0.98%,总体上优于ERA5-Land和TRMM_3B42日降水产品。另外,MPRISMR插值结果的精度随时间变化更小,更为稳定。最后,研究基于MPRISMR制备了空间分辨率约3 km的元江流域日尺度降水格网数据。本文可为复杂地形的陆面模式或流域水文水质模型提供高精度降水驱动场数据产品,从而支持流域可持续管理决策。Significant spatial heterogeneity of precipitation distribution can be observed over complex terrains.As a result,the spatial representativeness of ground-based precipitation observations,the applicability of remote sensing and reanalysis products,and the reliability of traditional interpolation methods are largely limited.PRISM(Parameter-elevation Regressions on Independent Slopes Model)has been widely used for precipitation interpolation.It interpolates the precipitation observation data by extracting and weighting topographic elements,and further introducing them into the regression of precipitation-elevation relationships through the weighted least square method.However,traditional PRISM is not capable of reflecting relatively small-scale topographic features.Thus,based on the analysis of impact mechanism of topographic factors on precipitation patterns over complex terrains,this study aimed to improve the calculation and weighting process of PRISM topographic elements,and integrate the modified PRISM into the"climatological-ratio"interpolation framework considering the random noise in actual daily precipitation observations and finally propose an interpolation framework,MPRISMR.Taking the Yuanjiang River Basin with typical complex terrains as an example,we verified the feasibility and reliability of MPRISMR for daily precipitation interpolation based on the"leave-one-out"cross-validation and by comparing with the ERA5-Land and TRMM_3B42 precipitation products.The results showed that over 23observation stations in the Yuanjiang River Basin,the performances of daily precipitation interpolation results based on MPPRISMR were better than those on the basis of both ERA5-Land and TRMM_3B42 precipitation products.Specifically,the verification evaluation index results showed that the medians of the Pearson’s correlation coefficient and relative bias of the MPRISMR results were 0.72、0.98%,respectively.Besides,compared with ERA5-Land and TRMM 3B42 precipitation products,the accuracy of MPRISMR precipitation interp

关 键 词：PRISM 复杂地形 降水插值 高分辨率降水产品 干热河谷 元江流域 

分 类 号：P426.6[天文地球—大气科学及气象学]