TITLE:
Best Model Selection for Extreme Values Analysis Using the Peaks-Over-Threshold Method and the Annual Maximum Method: Case Study of Bamako, Mali
AUTHORS:
Alikalifa Sanogo, Roland Songotu Kabange, Prince Appiah Owusu, Ousmane Mohamed Maїga, Lassina Dit Papa Koné
KEYWORDS:
Annual Maximum (AM) Series, Best Distribution Models, Extreme Values Analysis and Peaks-Over-Threshold (POT) Method
JOURNAL NAME:
Journal of Geoscience and Environment Protection,
Vol.13 No.12,
December
5,
2025
ABSTRACT: Heavy rains and floods are long considered critical societal concerns due to adverse effects on society, environment, and economies. The best appropriate identification for rainfall distribution is equally of significant concern to society and planners due to its application in the hydrological and water resources management sectors, and agricultural planning. Two methods for extreme values theory namely the annual maximum (AM) series method and the peaks-over-threshold (POT) method are generally employed for extreme values analyses. This study therefore compared the results of both methods for ARC2 daily rainfall data at Bamako-Senou station for the period 1991 - 2021. Five (05) commonly used distribution functions, namely the Normal, Log-Normal (LN), Gumbel type I, Gamma and Pearson type 3 (P3) distributions were used to fit the AM data. The method of moments (MOM) and the method of maximum likelihood estimation (MLE) were employed for parameters estimation in AM analyses. The generalized Pareto (GP) distribution was also used to fit the peaks over threshold (POT) method. The results indicated that the P3 distribution gave better result than other distributions when parameters were estimated with the MLE. The LN distribution was also best fit distribution to AM series when parameters were estimated with the MOM. The P3 distribution gave higher quantile estimates than other distributions. The POT method gave better results for quantiles estimation than the AM series. It is recommended that further study should include various truncation levels and tests for the choice of an optimal threshold in the POT method.