Extreme precipitation and flooding in Berlin under climate change, and its reduction by grey and blue-green measures

The effect of climate change on the statistical distribution of extreme precipitation in Berlin is assessed by analyzing output of a single-model set of regional climate scenario simulations at convection permitting resolution. The simulations were conducted by the German Weather Service (DWD) with the regional model COSMO-CLM driven by the global model MIROC5. Three 30-year periods were analysed: The historical period under observed greenhouse gas concentrations from 1971 to 2000 and two RCP8.5 scenario periods from 2031 to 2060 and from 2071 to 2100. The statistical rainfall distribution was evaluated by fitting a duration dependent General Extreme Value distribution to the annual maxima. For the historical period, the estimated 1-hour rainfall sum for a 100-year return level agrees well with the statistical values derived from station observations (RADOLAN). For the period 2031–2060 under RCP8.5 conditions the respective rainfall sum of the 1-hour 100-year event increases by 46 % and the strongest hourly intensity in all three simulated 30-year periods at a grid point (“the strongest event”) is 123 % stronger than the historical 100a event.

Using the statistical 100a and the strongest event as input, the impacts in terms of flooding characteristics are investigated by conducting simulations for the local flooding hotspot Gleimtunnel with the 2D surface flow model hms++ coupled to a 1D drainage model. Assuming a temporal rainfall distribution according to “Euler-2”, the 2031-2060 100a event results in a 51 % increase in the simulated maximum water depth compared to its counterpart for the historical period, a 43 % increase in maximum surface runoff, and a 33 % increase in the volume of combined sewer overflow. For the strongest event, the respective increases are 137 % (maximum water depth), 296 % (maximum surface runoff), and 74 % (combined sewer overflow). The drainage system (scaled to cope with 5-yearly events) strongly reduces flooding, especially at hotspots. Retrofitting all roof surfaces into retention roofs alone can reduce flood depths by about 20 % during a 100a rainfall event. As these values were obtained with hourly rainfall disaggregated to a 5 min resolution according to “Euler-2”, it is tested if direct output from the convective permitting simulation at the same frequency can lead to higher flood levels.