Region-of-influence approach to a frequency analysis of heavy precipitation in Slovakia

Abstract

Abstract. The L-moment-based regionalization approach developed by Hosking and Wallis (1997) is a frequently used tool in regional frequency modeling of heavy precipitation events. The method consists of the delineation of homogeneous pooling groups with a fixed structure, which may, however, lead to undesirable step-like changes in growth curves and design value estimates in the case of a transition from one pooling group to another. Unlike the standard methodology, the region-of-influence (ROI) approach does not make use of groups of sites (regions) with a fixed structure; instead, each site has its own "region", i.e. a group of sites that are sufficiently similar to the site of interest. The aim of the study is to develop a version of the ROI approach, which was originally proposed in order to overcome inconsistencies involved in flood frequency analysis, for the modeling of probabilities of heavy precipitation amounts. Various settings of the distance metric and pooled weighting factors are evaluated, and a comparison with the standard regional frequency analysis over the area of Slovakia is performed. The advantages of the ROI approach are assessed by means of simulation studies. It is demonstrated that almost any setting of parameters of the ROI method yields estimates of growth curves and design values at individual sites that are superior to the standard regional and at-site estimates.