Nonstationary multivariate analysis for compound extremes

Compound extremes occur when two or more hydrologic variables exhibit extremal behaviour simultaneously. The occurrence probabilities and impacts of compound extremes tend to be different than typical extreme event analysis used in hydrologic designs. Analysing and characterizing compound extremes is challenging because of joint consideration of multiple variables. Analysing compound extremes in the presence of non-stationarity makes it even more daunting. This study identifies compound extremes using multivariate analysis with precipitation and temperature serving as the key hydrologic drivers. A most conservative design is introduced for risk analysis. The proposed multivariate risk concept can be applied to study impacts of time dependency on compound extremes, and is demonstrated over the Godavari basin, India, using data from a recent period (1977 to 2020) during the peak monsoon season. The method is used to estimate the spatio-temporal risk of compound extreme events and would be of use to planners and resource managers.