Multivariate Copula Based Hydrological Drought modeling: A case study of Jiabharali sub-basin of Brahmaputra River, India

The growing apprehension of the vulnerable climate changes, viz-a-viz, frequent occurrence of water-related disasters, such as, hydrological drought in a river basin, assists in directing our all-out efforts to tackle various challenges posed by this creeping disaster in managing the precious water resources for the overall regional sustainability. This study uses a multivariate copula-based modeling framework to examine the characteristics and risks of occurrence of hydrological drought in a sub-basin one of the mightiest Rivers of the globe, i.e., the Brahmaputra River basin, India. The multi-time streamflow drought index (3, 6, 9, and 12 months), one of the most popular hydrological drought indices, was calculated by using the monthly discharge data from 2000 to 2023 obtained from the Bhalukpong gauge-discharge site located at the outlet of the Jiabharali sub-basin of the Brahmaputra River basin. The hydrological drought characteristics, such as, drought-severity, duration, and inter-arrival time were also investigated in different time scales. In this study, the maximum drought severity in the SDI-12 (SDI-9) time scale of magnitude about 91.2 (88.8) lasted for 60 (57) months in the Jiabharali sub-basin flowing through the northeastern state of Arunachal Pradesh, India. Different copula families, for example, Elliptical and Archimedean, were used to build joint probability distributions for the multivariate drought analysis. The study revealed that the Log-normal (Weibull) distribution observed to be the best fit in case of the univariate marginal distribution modeling for drought-severity and duration (drought-interval) in shorter time scales (longer time scales) based on two selection criteria, i.e., AIC and BIC. In contrast to examining dependence structure between the drought variables, strong positive correlations were observed between severity and duration, while weak or statistically insignificant correlations were found in association with the interval. The best reliable copula for capturing symmetric and heavy-tailed dependencies was found to be the t-copula from the Elliptical family, particularly for longer time scales (SDI6–SDI12). Multivariate joint return period analysis revealed that extreme joint drought events are infrequent but potentially catastrophic, emphasizing the importance of incorporating joint probability assessments in drought risk planning. The conditional probability between severity and duration revealed that shorter time scales are more likely to exhibit prolonged drought periods even when severity is relatively low. Such information is useful for planning the hydrological drought preparedness as integrated approach helps us in better understanding of the challenging behaviour of drought-characteristics and promotes more knowledgeable approaches to efficient water management in the region.