Modeling Monthly Extreme Rainfall Using the Generalized Extreme Value Distribution for Flood Risk Assessment and Agricultural Planning in Upper Northeastern Thailand

Abstract

This study aimed to model monthly extreme rainfall in the upper northeastern region of Thailand using the Generalized Extreme Value (GEV) distribution with parameter estimation based on the Maximum Likelihood Estimation (MLE) method. The dataset comprised daily maximum rainfall records from 1984 to 2025 collected at six meteorological stations, namely Sakon Nakhon Agricultural Station, Sakon Nakhon, Nakhon Phanom, Nakhon Phanom Agricultural Station, Bueng Kan, and Mukdahan. Descriptive analysis indicated that extreme rainfall at all stations exhibited pronounced right-skewed distributions, which are appropriate for extreme value analysis. The goodness-of-fit assessment using the Kolmogorov–Smirnov test and parameter estimates revealed that most stations were well characterized by the GEV framework, predominantly following the Weibull-type distribution, while some stations exhibited Fréchet-type behavior. Return levels were estimated for recurrence intervals of 2, 5, 10, 50, and 100 years, showing a consistent increasing trend with longer return periods across all stations. The highest return levels were observed at the Nakhon Phanom Agricultural Station, with the 100-year return level reaching approximately 281.75 mm, whereas the lowest values were found at the Sakon Nakhon station. Diagnostic plots, including Probability Plots, QQ-Plots, Return Level Plots, and Density Plots, confirmed the strong performance of the GEV model in capturing the empirical behavior of extreme rainfall. The findings provide valuable insights for flood risk assessment, agricultural planning, and water resource management in the upper northeastern region of Thailand under increasing climatic variability.