How does internal climate variability propagate to the catchment-characteristic flood types? Insights from a hydrological large ensemble

Due to various characteristics of a catchment area, such as topography, size and shape, soil and climate, different flood types prevail. These types are categorized according to the flood-generating processes, which are rainfall over short to long durations and the influence of snow dynamics. Often, a flood typology is devised based on observational time series or single hydrological simulations. However, the influence of internal climate variability on flood types is not well understood and quantified yet.

Here, we apply a unique hydrological large ensemble over a central European domain featuring the catchments of the upper Danube and the southern parts of the Main and Elbe catchments. The driving climate stems from a 50-member high-resolution single model initial-condition large ensemble (SMILE), the CRCM5-LE at 12 km resolution. SMILEs are driven by the same external forcing and apply a single climate model – hence, the variability within the SMILE can be interpreted as a model representation of internal climate variability. After a bias adjustment, the CRCM5-LE is used to drive the physically-based hydrological model WaSiM at 3-hourly temporal resolution and 500 m spatial resolution yielding a 50-member hydrological SMILE for 98 river gauges in the study area.

For a 60-year historic period (1961 – 2020) we differentiate between five types of rain-induced and snowmelt-affected floods via a statistical flood typology analysing the hydrographs and the hydrometeorological drivers. The flood types feature short-rain floods, floods driven by medium- to long-duration rainfall, long-rain floods with frequent multiple peaks, rain-on-snow floods, and snowmelt-dominated floods.     

We show that the frequency and intensity of the different flood types largely varies between the 50 simulations indicating a strong influence of internal climate variability on the flood typology. The highest degree of variability over all catchments is found for short-duration rainfall floods. The sensitivity of the flood typology to climate variability also varies greatly over the 98 catchments. We see a tendency for a higher sensitivity of smaller catchments to internal climate variability. Further, reservoirs and lakes are found to lower the effect of climate variability on the flood types due to their buffering behaviour.

We follow that using a single time series (may it be observational or simulated) might lead to a strong under- or overestimation of flood peaks per flood type, miss the catchment-specific flood typology, and induce an under- or overdetection of trends in the flood types. Hence, we suggest the application of SMILEs for the determination of flood peak return levels and the robust trend detection of certain flood types in order to incorporate the uncertainties of internal climate variability.