Robustness of precipitation and streamflow extremes in the surrogate world of seasonal forecasts

Water and disaster risk management require accurate information about hydrometeorological extremes. Estimation of rare events by extreme value analysis is hampered by short observational records. Probabilistic seasonal forecasts allow assessing the uncertainty in the estimation of extremes. From meteorological seasonal reforecasts and therewith driven hydrological simulations, we create hundred- to thousand-year-long surrogate timeseries across Europe. We identify independent samples based on the assessment of the forecast skill, and extract precipitation and streamflow extremes to explore the impact of sample size on return period estimations. The analysis clearly demonstrates the large uncertainty in long return period estimates with typical available samples of only few decades. The uncertainty is reduced at 100-year samples, and stabilizes at very low uncertainty around 500 years. We discuss the benefits and limitations of this method, and how it can be applied to study climate change and multiple extremes.