Present and future return values of heavy precipitation over complex terrain – the Black Forest region in Germany as an example

The Black Forest region in southern Germany is an example of complex mountainous terrain with steep slopes, narrow valleys, high variability of land use and strong convective activity. Population and trade reside mainly in valleys prone to flooding due to heavy precipitation which can cause severe damage due to rapidly rising water levels and short response times. The area coincides in large parts with the COPS campaign area presented in the parallel “Conference on Mountain Meteorology” and links of the work presented here to COPS will be discussed briefly. The high spatial and temporal variability, especially of heavy summer precipitation, in this region has been established in detailed analyses of time series obtained from a dense network of climate stations beginning in the first decades of the last century. The vital question, how such events and patterns will develop under a changing climate, has been addressed in an integrated research project from which a selection of results is presented. As stated in the Contribution of WG I to the 4^th IPCC report, the regional projections for mean and heavy precipitation in central Europe are quite uncertain, ranging from decrease to increase. One reason for this is certainly the coarse spatial resolution of the global models employed, which cannot resolve the observed small-scale variability over complex terrain. This scale problem can be overcome by using dynamical downscaling with high resolution regional climate models (RCMs). In this study we used the results of present day (1971-2000) climate and a future scenario (A1B, 2011-2040) simulations with two different RCMs, namely COSMO-CLM at 18 km and REMO at 10 km horizontal resolution. Although a weaker climate signal is to be expected, we chose the relatively near future time horizon for its relevance for planning and mitigation purposes.

 

The talk discusses the evaluation of the two RCMs in terms of precipitation statistics of the past 30 years on the one hand and on the comparison of the present and future statistics on the other. The evaluation was done using gridded 24 h precipitation fields derived from climate network observations. Such an evaluation is necessary to gain sufficient confidence in the model results for reliable conclusions from future projections to be drawn. The ability of the RCMs to realistically simulate the regional variability of precipitation distributions and precipitation return values (RVs) at various return periods (e.g. 1, 10 and 100 years) is assessed and the uncertainty of the RVs is estimated. A comparison of the distributions and RVs between the present (1971-2000) and the projection based on the A1B scenario for 2011-2040 follows.

 

To calculate the RVs, the observed time series of the 24 h precipitation of the 30-years time period 1971-2000 at each grid point was fitted to several test distributions, including the generalized Pareto distribution and the Kappa distribution, using the L-moments method. We found that generally the Kappa distribution gives the best agreement with the observed distribution. Our results show that the observed spatial pattern of the RVs can be reproduced reasonably well by the models for the summer and the winter period. Quantitatively, the RVs are overestimated by the models, especially in winter, one possible reason being biases in the driving global model data. Averaged over the investigation area, an insignificant increase of the RVs of a few percent between present and future is predicted, which is in the range of the global model predictions. The small scale variability, however, is quite high with increases and decreases of the RVs in the order of several 10 percent, underlining the importance and added value of high resolution downscaling.

 

Finally, we give an outlook on our next steps. These include simulations at resolutions of a few kilometres, ensemble simulations with Bayesian model averaging and possible applications to other regions.