Extreme precipitation has serious implications for environment, humans, and economy, as it can lead to various natural events and hazards, such as floods, mudslides, dam bursting, water pollution, and erosion. Accurate estimates and evaluations of extreme precipitation are essential to avoid potential damages and losses. This is particularly important for risk management and the development of hydraulic systems as well as their operational business, as it is crucial to precisely estimate the frequency and intensity of extreme precipitation events. However, these estimates have been associated with large uncertainties, as the frequencies of very rare, sometimes even unobserved intensities, have to be determined. One of the major challenges is the limited availability of data, including both long-term time series and widespread observation stations. In Germany, daily precipitation sums have been measured comprehensively since the mid-20th century. However, extremely rare events, occurring only once every 100 years or less, may not have been recorded within this observation period. Furthermore, assessments of extreme precipitation are often conducted at an annual resolution, although it is also realizable for seasonal or monthly data. Additional information about the seasonality of extreme precipitation can be valuable for sectors such as agriculture and tourism, enabling seasonally adapted risk management. Additionally, anthropogenic greenhouse gas emissions and, thus, rising air temperatures are influencing the intensity of precipitation. So far, long-term changes in extreme precipitation have not been considered in the planning of hydraulic systems and risk assessment. However, for suitable climate adaptation in the future, this will become necessary. The aim of this work is to develop an extreme value model that 1) resolves the seasonal cycle of extreme precipitation, providing additional information, 2) utilizes existing data points more efficiently to improve the accuracy of estimates, called return values, and reduce their uncertainty, and 3) models long-term changes due to natural variability and anthropogenic climate change, building an opportunity for future climate adaptation measures. Here, a nonstationary generalized extreme value (GEV) distribution with flexible distribution parameters is used and adapted to monthly maxima of daily precipitation sums from observation stations in Germany. In this thesis, a seasonal model is developed to capture the seasonal cycle of extreme precipitation, using harmonic functions for flexible parameters instead of modelling each month separately with a stationary GEV. This approach improves the accuracy of both monthly and annual return level estimates and provides insights into how extreme precipitation changes throughout the year. This information can be useful for different stakeholders, such as in emergency and evacuation planning for disaster protection, or the implementation of protective measures for young plants in agriculture. Additionally, a spatial-seasonal model is introduced, which uses available data points more efficiently by employing orthogonal polynomials to simultaneously incorporate data from all stations, and model the spatial variations of extreme precipitation. The application to different example regions demonstrates improved accuracy in return level estimates, the capability to derive return levels at ungauged sites, and the ability to assess the occurrence frequency of precipitation events. The investigation also incorporates long-term trends to examine changes in seasonality due to climate change and climate variability. The evaluation of return levels indicates evidence that anthropogenic climate change has a detectable impact on the seasonality of extreme precipitation in Germany, while type and magnitude of changes vary in space. The introduced method can be applied to data from other regions and other durations, such as hourly precipitation sums. It provides additional temporal and spatial information, and more reliable results. Moreover, it integrates climate change-induced alterations into the calculation of return levels, thereby offering a valuable approach for shaping future climate adaptation strategies.
