High-resolution fully distributed hydrological modelling of flash floods based on convection-permitting regional climate model data: An integrated modelling framework

Spatial resolution is a key factor in the modelling of convective rainfall extremes and their environmental impacts under current and future climate. Rapid developments in the field of high-performance computing have advanced dynamical downscaling of climate simulations to convection-permitting scale. Such high-resolution regional climate models hold great potential for improved modelling of convective processes through refined depiction of land surface properties and solving of the vertical momentum equation. However, these simulations currently operate on scales (~ 3km) still too coarse to serve as direct input for hydrological modelling of flash floods in fast responding catchments with diverse land use/ land cover (LULC). We investigate the added value of such uncorrected convection-permitting regional climate model (CPRCM) data for hydrological impact modelling in a catchment of medium topographic complexity in Germany and suggest an outline for an integrated modelling framework for very high-resolution simulation of hydrometeorological extremes.

The study compares reanalysis-driven hourly precipitation simulations from the non-hydrostatic model ICON-CLM 2.6.4 at 3 km resolution (ICON3km) and its nest model ICON-CLM 2.6.4 with parametrised convection at 11 km (ICON11km) to adjusted radar data upscaled to respective resolution over a study area of 13,210 km² embedded between the Leipzig Lowlands and the Elster/ Ore Mountains in East Central Germany. While ICON3km alleviated the drizzle bias, it strongly overestimated heavy precipitation both in intensity and frequency. As a result, discharge computed using the distributed, physically based hydrological model WaSiM for the enclosed small to medium-sized catchments (107 to 529 km²) of the Weiße Elster river basin showed a strong positive bias when simulated based on uncorrected ICON3km data. The results suggest a necessity of bias correction of the CPRCM data before use in flash flood modelling.

In fast responding catchments with diverse LULC, hydrological impact simulations require meteorological data on an even finer scale than provided by common CPRCM setups. We suggest an integrated modelling framework for rural catchments, combining statistically downscaled CPRCM data and fully distributed hydrological models. An adequate representation of cultivated steep catchment slopes is implemented by high-resolution parametrisation of surface, vegetation and soil properties, as gained from freely available remote sensing and cadastral data. Key hydrological processes, such as Hortonian overland flow and saturation, are accounted for through process-based representation in an open-source modelling environment. The framework is envisioned to be applied i.a. for local flood hazard assessment and for the study of drivers of runoff dynamics under current and future climatic conditions. Furthermore, it is to be employed for the assessment of the effectiveness of selected agricultural runoff countermeasures under different climate change scenarios.