Estimation of future rainfall extreme values by temperature-dependent disaggregation of climate model data

Abstract

Rainfall time series with high temporal resolution play a crucial role in various hydrological fields, such as urban hydrology, flood risk management and soil erosion. Understanding the future changes in rainfall extreme values is essential for these applications. Since climate models typically offer daily resolution only, statistical downscaling in time seems a relevant and computationally effective solution. The micro-canonical cascade model conserves the daily rainfall amounts exactly, and having all model parameters expressed as physical interpretable probabilities avoids assumptions about future rainfall changes. Taking into account that short-duration rainfall extreme values are linked with high temperatures, the micro-canonical cascade model is further developed in this study. As the introduction of the temperature dependency increases the number of cascade model parameters, several modifications for parameter reduction are tested for 45 locations across Germany. To ensure spatial coherence with the climate model data, a composite product of radar and rain gauges with the same resolution was used for the estimation of the cascade model parameters. For the climate change analysis the core ensemble of the German Weather Service, which comprises six combinations of global and regional climate models, is applied for both RCP4.5 and RCP8.5 scenarios. For parameter reduction two approaches were analysed: (i) the reduction via position-dependent probabilities and (ii) parameter reduction via scale independency. A combination of both approaches led to a reduction in the number of model parameters (48 parameters instead of 144 in the reference model) with only a minor effect on the disaggregation results. The introduction of the temperature dependency improves the disaggregation results, particularly regarding rainfall extreme values and is therefore important to consider for future studies. For the disaggregated rainfall time series of climate scenarios, an intensification of the rainfall extreme values is observed. Analyses of rainfall extreme values for different return periods for a rainfall duration of 5 min and 1 h indicate an increase of 5 %–10 % in the near-term future (2021–2050) and 15 %–25 % in the long-term future (2071–2100) compared to the control period (1971–2000).