Open Access

Nico Blettner

• Rainfall strongly influences the availability of water on the land surface, and hence, its quantification is utterly relevant for addressing a variety of social, economic, and environmental matters. Quantification via traditional rainfall measuring devices has its limitations and can be supported by opportunistic sensors like commercial microwave links (CMLs), which theoretically enable rainfall estimation on large spatial scales due to their vast global abundance. However, estimation across organizational (e.g., national) boundaries is challenging due to heterogeneous CML data sets with customized rainfall retrieval methods. Moreover, common interpolation techniques have shortcomings in using path-averaged CML observations for spatial rainfall reconstruction. These challenges of CML-based transboundary rainfall estimation have been addressed in this thesis by generating rainfall maps of hourly temporal resolution, which were evaluated using a weather radar reference. Two large CMLRainfall strongly influences the availability of water on the land surface, and hence, its quantification is utterly relevant for addressing a variety of social, economic, and environmental matters. Quantification via traditional rainfall measuring devices has its limitations and can be supported by opportunistic sensors like commercial microwave links (CMLs), which theoretically enable rainfall estimation on large spatial scales due to their vast global abundance. However, estimation across organizational (e.g., national) boundaries is challenging due to heterogeneous CML data sets with customized rainfall retrieval methods. Moreover, common interpolation techniques have shortcomings in using path-averaged CML observations for spatial rainfall reconstruction. These challenges of CML-based transboundary rainfall estimation have been addressed in this thesis by generating rainfall maps of hourly temporal resolution, which were evaluated using a weather radar reference. Two large CML data sets from Germany and the Czech Republic with distinctly different network characteristics were combined and processed jointly via established and extended algorithms to generate transboundary rainfall maps. Beyond that, the German CML data set was combined with a countrywide network of rain gauges to generate rainfall maps via a stochastic reconstruction approach called Random Mixing (RM). The quality of these maps was analyzed considering an alternative standard Kriging approach and an object-based validation scheme named eSAL, which quantifies errors in structure, amplitude, and location. The computational complexity of RM was examined and reduced. It was found that the German and Czech CML data sets could be processed jointly to generate consistent transboundary rainfall maps once issues of limited data quality were identified and addressed by appropriate universal algorithms. The strong influence of partly hardware-dependent data quality issues could be demonstrated. Furthermore, stochastic reconstruction via RM proved to enable the generation of rainfall maps with accurate pattern representation. Despite a general underestimation and relatively high computational complexity, the method had clear advantages over the Kriging approach as indicated in particular by significantly lower structure errors and by providing probabilistic ensemble solutions. The results yield evidence for the capabilities of generating high-quality CML-based rainfall maps on large spatial scales, even across political borders, and hence, they contribute to better utilize the potential of CMLs as widespread rainfall sensors worldwide.…