Under changing environmental conditions, the inclusion of non-stationarity in risk assessment of hydroclimatic extremes provides reliable estimates of quantiles or design levels; however, such assessments are usually limited to univariate analysis. A multivariate non-stationary framework is capable of considering the interdependencies between multiple characteristics of such extreme events. This study offers a new non-stationary multivariate risk assessment framework based on the concepts of Expected Number of Events (ENE) and Expected Waiting Time (EWT) that have worked well in the recent past for return level estimates of univariate extremes under non-stationarity. The proposed framework is first demonstrated on synthetically generated multi-attribute datasets to prove its effectiveness for both increasing and decreasing trends in extremes. Further, real-world applicability of the approach is presented for multivariate non-stationary drought risk assessment in India, where dynamic copulas are used to capture the joint transient behaviour of drought severity and duration. The design severity and duration for any given return period increases (decreases) for positive (negative) trend in drought attributes, implying underestimation (overestimation) of risk by the stationarity approach. Globally, changes in characteristics of hydroclimatic extreme events are increasingly reported, and such changes are likely to be exacerbated by climate change. The proposed framework considers time-varying nature of multiple characteristics for risk assessment of such extremes, and can aide in engineering design and environment management.
Citation
Sahana, V.; Mondal, A. 2025. Multivariate return period and risk analysis of non-stationary extremes: dynamic drought severity-duration-frequency in India. Stochastic Environmental Research and Risk Assessment, 21p. (Online first). doi: https://doi.org/10.1007/s00477-025-03090-1