Derivation of Probabilistic Depth–Area–Frequency Curves Based on Spatial Adjacency Using the Generalized Extreme Value Distribution in South Korea

In recent decades, climate change has intensified extreme rainfall events and expanded their spatial extent, highlighting the need for area-based design rainfall estimation in regional flood control planning. Conventional Depth–Area–Frequency (DAF) curves rely on point rainfall observations combined with empirical Area Reduction Factors (ARFs), which limits their ability to represent actual area-averaged rainfall and spatially connected rainfall structures. This study develops a probabilistic DAF framework that explicitly accounts for spatial adjacency and area-averaged rainfall characteristics. Using 30 years of rainfall observations from the Automated Synoptic Observing System (ASOS) across South Korea, spatially connected area combinations were constructed through adjacency analysis, and representative area sets were selected using the Latin Hypercube Sampling technique. Area-averaged annual maximum rainfall was then derived for each area scale, and multiple probability distributions were applied to characterize extreme rainfall behavior. Goodness-of-fit evaluations indicate that the Generalized Extreme Value (GEV) distribution most appropriately describes area-based extreme rainfall across different spatial scales. Based on the selected GEV distribution, probabilistic DAF curves corresponding to various return periods were derived. The proposed framework eliminates reliance on empirical ARFs and provides a physically consistent and probabilistically rigorous approach for estimating design rainfall, thereby improving the reliability of regional and national-scale flood control and hydrologic design applications.