Non-stationary low-flow frequency analysis with Mixture Generalized Extreme Value (GEV) distribution

Abstract

Extreme low flow is a critical component of the river flow regime, posing significant risks for water management by impacting water availability and quality. Addressing these challenges requires accurate information on design low flow corresponding to specific non-exceedance probabilities. Traditional low-flow frequency analysis assumes stationarity and process homogeneity; however, these assumptions become questionable under the influence of climate change and varying generation processes for low flows, such as in seasonal snow climate that the annual extreme series will be a mixture of both summer and winter low-flow events. The study aims to extend regional low flow frequency analysis to non-stationary conditions and account for seasonal variation for a better statistical description of extreme events.

First, we analyse temporal trends in the study area separately for annual minimum winter and summer series and investigate whether they can be related to temperature increase or other climate trends. Then, we apply modelling concepts to extend the mixed distribution model of Laaha (2023) to non-stationary conditions using a conditional Generalized Extreme Value (GEV) distribution. This allows us to consider the detected trends in low flow frequency analysis. The results of the study provide a new perspective on low flow processes and impact chains in river systems.

Keywords: Non-stationary frequency analysis, low flow, drought, climate change, seasonality

Reference

Laaha, G. (2023). A mixed distribution approach for low-flow frequency analysis – Part 1: Concept, performance, and effect of seasonality. Hydrol. Earth Syst. Sci., 27(3), 689-701. https://doi.org/10.5194/hess-27-689-2023