Model for estmating non-consistency hydrological drought dynamic risks based on Vine Copula

Hydrological drought has posed serious threats to water security in the arid and semi-arid regions. Particularly, the surface water resources are already scarce in the typical oasis areas in the Hotan River Basin of the southern Xinjiang Uygur Autonomous Region in China. However, the water demands have expanded to cause a non-consistent pattern in the hydrological regime of the basin under climate change. The conventional consistency assumption on the hydrological drought is also limited in the large-scale production of sustainable agriculture. Previous research has also focused on the non-consistent drought indices and drought features under changing conditions. It is often required to accurately assess the dynamic drought risk for agricultural planning and water resources. In this study, a dynamic risk assessment framework was developed for the non-consistent hydrological drought in the Hotan River Basin. The runoff series was selected to examine the non-consistent features using statistical testing. Empirical evidence was then provided for the impacts of environmental changes on the hydrological process. The Generalized Additive Models for Location, Scale, and Shape (GAMLSS) framework was employed to construct a non-consistent standardized runoff index (NSRI). Distribution parameters were explicitly incorporated with the time-varying influences of the climatic factors and anthropogenic activities. Consistent drought indices were avoided to validate the superiority of the framework. The NSRI was systematically compared against SRI using drought characteristic analysis and historical drought events. Better performance was achieved in capturing the drought severity, in agreement with actual disaster occurrences. Multiple drought features were integrated with the occurrence probabilities under non-consistent conditions using the Vine Copula method. The high-dimensional joint distributions were decomposed into a series of conditional bivariate copulas, effectively reducing the complex parameter estimation and the complex dependencies among different drought attributes. The drought risk was quantified as more actionable information using continuous, time-varying metrics rather than a static value. The performance was validated using well-documented historical drought events, indicating reliable early warnings. Some insights were obtained. Firstly, there was a significant variation in the hydrological regime, where the non-consistent features of the runoff series were attributable to both climatic shifts and human interventions. Secondly, the superior performance of the NSRI was achieved in capturing the drought events, compared with the SRI. More accurate characterization of drought severity resulted in better agreement with the documented drought occurrences. Thirdly, the risk assessment indicates that the basin is currently subjected to moderate drought risk levels, where both major tributaries share similar risk profiles. The effectiveness of the model was validated to identify the high-risk periods, according to the typical drought events in 1991-1992. Timely warnings were provided for both scientific and practical applications. Non-consistent hydrological drought analysis was integrated with the GAMLSS index with Vine Copula-based dynamic risk assessment in a unified framework. The model can be readily adapted to similar regions under hydrological conditions. The finding can provide valuable support for drought monitoring, agricultural irrigation, and risk management in the oasis agricultural regions. Non-consistent hydrological drought can also offer a robust framework to enhance the drought resilience in the water-stressed regions under environmental conditions in sustainable agriculture.