Improved Regional Frequency Analysis of rainfall data

Rainfall are subject to local orography features and their intensities can be highly variable. In this context, identifying climatically coherent regions greatly help interpreting rainfall patterns improve the inference of return levels. practice, partitioning a region interest into homogeneous sub-regions is delicate statistical task, especially in regards modeling heavy features. work, our main goal propose study fast efficient clustering algorithm. Compared classical regional frequency analysis techniques, key aspect that algorithm does not rely on priori choice covariates. The proposed numerical scheme only based precipitation dataset at hand, including low, moderate rainfall. terms inference, approach builds easy-to-compute reliable probability weighted moments commonly used hydrology. While being compliance with extreme value theory, we do impose parametric form distributions and, neither thresholding nor block maxima steps required approach. By construction, method preserves scale invariance principle any analysis. performance assessed detailed experimental design extended Generalized Pareto distribution. Sensitivity number clusters carefully analyzed. We apply Switzerland daily measured 191 sites. found consistent outperforms normalized elevation coordinates as To complete Swiss rainfall, three models outputs. A comparison between local, semi-regional indicates relatively simple model two spatially varying parameter compete very well against complex models.