A Copula-based Dynamic Bayesian Network for analyzing spatiotemporal runoff encounter patterns in water source areas...

The multiscale spatiotemporal characteristics of runoff encounter patterns are crucial for optimizing water diversion and improving scheduling reliability of inter-basin water transfer projects. In this study, a Copula-Dynamic Bayesian Network (CDBN) model is developed to analyze runoff encounter patterns. Monthly runoff states are represented as DBN nodes, while Copula functions are used to derive conditional runoff-state probabilities. The water source regions of the Eastern Route of South-to-North Water Diversion Project in China are selected as the study area. The results show clear temporal persistence, with relatively high wet-to-wet and dry-to-dry transition probabilities and lower same-lake dry-wet reversal probabilities, indicating the need for in-advance water storage or flood-control preparation under persistent dry or wet conditions. Spatially, adjacent lakes show generally synchronous wet-dry responses, with dry-dry and wet-wet probabilities in most months. Under synchronous dry risks, supplementary water supply from other available lake sources or the Yangtze River should be considered when necessary. Spatiotemporal transition analysis reveals dry and wet states tending to persist from an upstream lake in the current month to the downstream lake in the following month, among which the Hongze and Luoma Lake show the strongest state persistence. The proposed model provides a probabilistic foundation for water resource allocation and operation in inter-basin water transfer projects under changing environmental conditions.