Spectrally Transformed Hydroclimatic Covariates Improve Seasonal Flood Forecasting

Abstract

Reliable seasonal flood forecasting is vital for managing reservoirs and disaster response. This study investigates whether probabilistic forecasts of seasonal floods can be improved by integrating spectrally transformed hydroclimatic variables. We apply the Wavelet System Prediction (WASP) method to enhance climate covariates within a Generalized Extreme Value (GEV) model. Using streamflow observations from 649 European catchments, we compare forecasts using raw and spectrally transformed covariates. Results show that the transformation significantly improves forecast skill, measured by the Ranked Probability Skill Score (RPSS), especially at longer lead times. The most notable gains are observed in Northern and Western Europe, including the UK and Norway. The proposed hybrid WASP-GEV forecasting framework, integrating spectral transformation, significantly enhanced seasonal flood forecast skills with up to 3 months of lead time. These findings highlight the potential of advanced data transformation techniques to improve hydroclimatic extreme forecasts, benefiting water resource management in a changing climate.