Estimating flood quantiles using a novel climate-informed-seasonal mixing approach

Código

I-EBHE0079

Autores

Abinesh Ganapathy, Bruno Merz, Sergiy Vorogushyn, Ankit Agarwal

Tema

WG 1.10: Hydrologic Design - Solutions & Communication

Resumo

Given widespread changes in catchment and climate, researchers have challenged the stationarity assumption in the traditional flood frequency analysis. One of the limitations of the stationary assumptions is its non-inclusion of changes in extremes associated with future climatic conditions. This can be overcome by incorporating climate covariates into the estimation of flood probability through the non-stationary Climate-Informed Flood Frequency Analysis (CIFFA). The CIFFA methodology comprises 1) selection of predictands (usually seasonal maxima), 2) identification of suitable predictors (large-scale climate indices), and 3) derivation of a statistical link between predictands and predictors. However, the applicability of CIFFA is limited to the gauges where flood extremes occur only during the dominating season. In many gauges, flood extremes occur in multiple seasons and are triggered by different flood-generating mechanisms. To address this, we develop and test a novel non-stationary Climate-Informed-Seasonal-Mixing approach across various European basins. In the proposed approach, we condition the location parameter of the seasonal peak distribution (boreal seasons) on the selected covariates. The selection of best climate covariates for each season among a set of predictors is based on widely applicable information criterion, which computes the log posterior predictive density while penalizing the overfitting. The approach is flexible enough to include the traditional stationary model as the best model if it has the least WAIC value among the competent models. After estimating the seasonal distribution parameters, the annual flood quantiles are derived by multiplicatively mixing all the seasonal distributions. Furthermore, the performance of the proposed approach is demonstrated by splitting the entire period into calibration and validation, fitting the model based only on calibration samples, and then evaluating it for the validation period. The projected quantiles during the validation period are compared with a benchmark model (a traditional model fitted solely with validation samples). Our results suggest that for many gauges, the flood quantiles estimated by the proposed Climate-Informed-Seasonal-Mixing approach align with the baseline estimates where the traditional approaches fall short.