Open Access

Severin Kaspar

• The European Alps are particularly vulnerable to the effects of climate change, as evidenced by an above-average increase in surface air temperature in recent decades. Besides the thermal changes impacts on the hydrological cycle can also be expected. Alpine water resources act as an important factor not only for the ecosystem but also for humanity, in the form of water security considering irrigation, hydrological energy production or winter tourism. Consequences are manifold but yet often not sufficiently investigated. This study is intended to give insight on modifications in the locally available water resources at the Zugspitze and Hoher Sonnblick area deducted from observed and simulated meteorological time series. Target variables include temperature and precipitation at both stations as well as relative humidity and wind speed at Zugspitze. Future scenarios are provided by Earth System Model projections. To bridge the gap between the coarse spacial resolution of climate models,The European Alps are particularly vulnerable to the effects of climate change, as evidenced by an above-average increase in surface air temperature in recent decades. Besides the thermal changes impacts on the hydrological cycle can also be expected. Alpine water resources act as an important factor not only for the ecosystem but also for humanity, in the form of water security considering irrigation, hydrological energy production or winter tourism. Consequences are manifold but yet often not sufficiently investigated. This study is intended to give insight on modifications in the locally available water resources at the Zugspitze and Hoher Sonnblick area deducted from observed and simulated meteorological time series. Target variables include temperature and precipitation at both stations as well as relative humidity and wind speed at Zugspitze. Future scenarios are provided by Earth System Model projections. To bridge the gap between the coarse spacial resolution of climate models, a statistical downscaling framework is developed to derive the corresponding local impact. The statistical transfer function between the large and local-scale is gained from non-linear methods, namely Artificial Neural Networks, Cluster Analysis and a novel approach combining both methods. Statistical downscaling models are calibrated and evaluated on a daily basis considering large-scale reanalysis datasets and local-scale observations. A main goal of this study is to find the best predictor setup with Artificial Neural Networks, whereby in climate studies rarely used sensitivity studies, in particular the Partial Derivative method, produced the most reliable results. Usually best modelling performance in validation is obtained by Artificial Neural Networks. Future changes are investigated by transferring the statistical downscaling models on Earth System Model datasets. Hereby, the historical period is compared to time periods within future scenarios RCP 4.5 and RCP 8.5 and analysed for differences. Considering precipitation, high increases are found at both stations in winter, suggesting a future shift in the annual precipitation distribution. With low uncertainties, temperature is simulated to increase successively, with most intense warming taking place in summer and winter. Simultaneously, relative humidity at Zugspitze is found to continuously decrease. A weak trend in wind speed at Zugspitze is projected by the statistical downscaling models, showing a significant increase not before the end of the 21st century. The resulting time series of the downscaling framework are applied in a statistical modelling application based on Artificial Neural Networks targeting monthly measurements of snow depths at both stations. In the annual average especially the RCP 8.5 based simulations expect partly drastically reduced snow depths of up to 50 %.…