Extreme precipitation and spatial dependencies
Extreme precipitation in Switzerland accounting for spatial dependence
High-resolution precipitation data sets usually cover an all too short time span to cover several heavy precipitation events with different spatial characteristics. Thanks to a new methodological approach, the Mobiliar Lab is able to generate such event data.
Seasonal weather forecasts are now operationally calculated by several weather centers. These forecasts extend beyond the forecast horizon of classical weather forecasts. They aim at estimating seasonal trends and answering questions like: Will the coming winter be exceptionally cold? Will the next summer be especially dry?
For Central Europe, the accuracy of such predictions is currently low. Nevertheless, the simulations calculated by the models are realistic. They represent one of the possible variants - or in technical jargon: a possible realization of the climate system. Seasonal weather forecasts are therefore an archive in which you can search for heavy precipitation events or other weather extremes. Data from this archive can supplement measurement series and provide additional information on extreme, and thus rare, events.
The Mobiliar Lab has tested this approach with data from the European Centre for Medium-Range Weather Forecasts for Switzerland. In total, the dataset covers around 9000 years of weather simulations. The grid width of this geodata is about 35 kilometers, the temporal resolution 6 hours. For a global weather model this spatial and temporal resolution is very high, but for hydrological applications in small-scale Switzerland it is too coarse. One of the challenges of the project was therefore to increase the resolution of the model. The Mobiliar Lab applied a statistical method for this purpose, which is also used in the context of climate change studies.
The experiments conducted in the context of the study have shown that this approach actually works for Switzerland. This is for three reasons: First, the spatial and temporal resolution can be increased with the chosen statistical method. Second, the simulations of the weather model reproduce the statistical properties of observed heavy precipitation events. And third, the data set contains extreme precipitation events with different spatial characteristics.
For small areas, however, the approach can only be applied to a limited extent. Artificially increasing the spatial and temporal resolution fails when it comes to reproducing high precipitation intensities on a small spatial scale, such as those measured by weather stations. However, if the focus is on catchment areas larger than 300 km2 , which is the area of the Canton of Schaffhausen, the precipitation scenarios provide realistic values.