Frequency analysis based on Peaks-Over-Threshold approach for GPM IMERG precipitation product

Abstract

The accurate estimation of extreme precipitation frequency relies heavily on the quality of precipitation data. Conventional rain gauges provide in-situ observations, but they are susceptible to errors and are sparse or even absent in many parts of the world. Satellite-based precipitation products developed in recent years provide high-quality precipitation estimates on a regional and global scale. Before being used in such applications, it is vital to reduce errors by investigating their accuracy. In this study, we aimed to evaluate the application of the Integrated Multi-satellitE Retrievals for GPM (IMERG) product in the frequency analysis of extreme precipitation over Turkey. Initially, we evaluated the performance of the GPM IMERG over nine selected stations and deployed a bias correction procedure. Then, using the Generalized Pareto Distribution (GPD), we estimated 2-, 5-, 10-, 20-, 50-, and 100-year return levels. Implementation of the bias correction method yielded significant improvements in the results. The agreement of the GPM IMERG product with in-situ observations was between 0.24 and 0.65 on the Kling-Gupta Efficiency (KGE) metric and increased to the range of 0.54 to 0.75 after bias correction. Moreover, in the Mean Absolute Error (MAE) metric, the range of values changed from 5.30 to 1.39 mm/day and 5.80 to 1.10 mm/day. Furthermore, it noticeably reduced the error for most of the return periods. The error percentages between in-situ observations and bias-corrected IMERG products decreased below 25% at all stations for short return periods. Notably, at Ilgaz station, bias correction indicated an improvement of over 50% in error percentages for all return periods. The results demonstrated that GPD is a good solution, particularly for dealing with short time series, and suggest that IMERG can be a valuable alternative source for analyzing precipitation frequency over Turkey.