Impact of variability and anisotropy in the correlation decay distance for precipitation spatial interpolation in China

ABSTRACT: Correlation decay distance (CDD) plays a key role in the angular-distance weighting (ADW) interpolation technique, where it is used as the search radius to select correlated neighbors and to calculate relative weights. The purpose of this study was to assess any improvement obtained by using a regionally and seasonally variable CDD rather than a fixed CDD based on the entire mainland China daily precipitation dataset or the intermediate case of a seasonally invariant CDD within each region. We also assessed the influence of using anisotropic versus isotropic CDDs. We found that the CDD of daily precipitation in China varies spatially and seasonally, and it presents anisotropic behavior, as a result of topography and the predominant atmospheric circulation. In general, CDD is largest in winter and smallest in summer, except for limited regions such as the Tibetan plateau. From a cross-validation analysis, we found that taking account of spatial and seasonal variations in CDD generally improves ADW interpolation. Utilization of anisotropic CDDs increases the interpolation skill scores in regions with a dense monitoring network, significant elevation variation (southwestern China), or strongly anisotropic CDDs (Tibetan plateau).