Akademik Veri Yönetim Sistemi
WATER RESOURCES MANAGEMENT, cilt.40, sa.1, 2026 (SCI-Expanded, Scopus)
This study evaluates 50-year precipitation estimates derived from stationary (S) and non-stationary (NS) models developed using different covariates and probability distributions, with model selection based on the Akaike Information Criterion (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:AIC$$\end{document}) and Bayesian Information Criterion (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:BIC$$\end{document}). Annual maximum precipitation (AMP) data from 53 meteorological stations (MSs) across southern and central T & uuml;rkiye are analyzed. Four probability distributions (gamma, Generalized Extreme Value (GEV), Gumbel (Gu), and log-normal) and five covariates (time, maximum temperature, North Atlantic Oscillation (NAO) index, the number of days with temperatures exceeding the long-term average, and meteorological drought magnitude) are considered. Both S and NS models with up to two covariates are developed for location and scale parameters. Results show that GEV and Gu distributions are most frequently selected based on \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:AIC$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:BIC$$\end{document}, respectively. Kendall Rank Correlation Coefficient analysis reveals inconsistencies in the top 30 models ranked by \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:AIC$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:BIC$$\end{document}, emphasizing the importance of performance criterion. While NS models generally improve predictive performance, they require more complex formulations that often involve multiple covariates in the location or scale parameters which increase computational demand and may lead to unrealistically high 50-year precipitation estimates.