Edmund Meredith (Germany) 1; Henning Rust (Germany) 1; Uwe Ulbrich (Germany) 1; Paula Lorza (Germany) 2; Marc Scheibel (Germany) 2
1 - Freie Universität Berlin; 2 - Wupperverband
Hydrological modelling requires high-resolution climate data [O(1 km)] at the catchment scale because of small-scale water system related structures. The spatial scales of available numerical simulation results for climate change and decadal predictions are thus too coarse for direct application. Practitioners thus require a workflow which provides them with suitable input data for the region of their responsibility.
A particular challenge are extreme events, as the observations are typically sparse, and climate change both on the decadal and the centennial scale may produce events of an intensity not observed before. Here, dynamical downscaling to 2 km resolution is applied in order to generate physically possible cases from the large scale conditions.
Such dynamical downscaling is, however, extremely computationally expensive, so that multi-year downscaling is often infeasible.
Within the framework of the H2020 project BINGO (Bringing INnovation in onGOing water management www.projectbingo.eu), a flexible classification algorithm was developed for the identification of days with enhanced likelihood of extreme local rainfall events, so that these days can be selectively downscaled to high resolution, thus drastically reducing computational expense. Large-scale weather patterns associated with extreme precipitation are identified using reanalysis. Days with similar circulations are classified as potential extreme days (PEDs), subject to additional discrimination with local-scale parameters affecting the occurrence of extremes.
For the identified PEDs, high-resolution, convection-resolving downscaling is performed and the changing risk for such events is assessed, including a comparison with observed historical data.
Our method can be used to inexpensively create high-resolution climate model data of extreme precipitation events, for use by hydrologists modelling the impacts of extreme precipitation – both past and future – on their own catchments. The inherent flexibility of the method allows transferability to other catchments and application across diverse climate data sets.
Results are used for an improved targeting of investments for disaster risk reduction, with adequate cost-benefit relationship. Knowledge of the potential small-scale impacts under specific large-scale weather conditions also helps to develop suitable contingency plans. The methodology can operationally be applied to forthcoming predictions.