Combining Commercial Microwave Link and Rain Gauge Observations to Estimate Countrywide Precipitation: A Stochastic Reconstruction and Pattern Analysis Approach.

Accurate spatiotemporal precipitation quantification is a crucial prerequisite for hydrological analyses. The optimal reconstruction of the spatial distribution, that is, the rainfall patterns, is particularly challenging. In this study, we reconstructed spatial rainfall on a countrywide scale for Germany by combining commercial microwave link and rain gauge observations for a better representation of the variability and spatial structure of rainfall. We further developed and applied the Random‐Mixing‐Whittaker‐Shannon method, enabling the stochastic reconstruction of ensembles of spatial fields via linear combinations of unconditional random fields. The pattern of rainfall objects is evaluated by three performance characteristics, that is, ensemble Structure‐, Amplitude‐, and Location‐error. Precipitation estimates obtained are in good agreement with the gauge‐adjusted weather radar product RADOLAN‐RW of the German Weather Service (DWD) which was used as a reference. Compared to reconstructions by Ordinary Kriging, Random Mixing showed clear advantages in the pattern representation via a five times smaller median structure error. Plain Language Summary: Rainfall is commonly measured by dedicated sensors such as rain gauges or weather radars. Commercial microwave links (CMLs), which have the primary purpose of signal forwarding within cellular networks, can be used for rainfall measurements too. The signal, which is transmitted from one antenna to another, is being attenuated if it rains along the path. From the amount of attenuation an average rain rate can be retrieved. For many hydrological applications, it is of major interest to estimate area‐wide rainfall (i.e., rainfall maps) while observations provide only scattered information. In this study, we used the local information from almost 1,000 rain gauges and the information along the paths of 3,900 CMLs distributed over Germany to reconstruct rainfall maps. We did this by applying a method of stochastic simulation (called Random Mixing) which we compared to a more common method of estimation (Ordinary Kriging). To evaluate the quality of the obtained maps, we compared them to rainfall information from weather radars. We found that the general agreement is high, and that maps reconstructed by Random Mixing have particular advantages in representing the spatial structure, that is, the shape of rainfall cells. Key Points: Geostatistical Random Mixing simulation now capable of countrywide spatial rainfall interpolationVariability assessment via commercial microwave link path consideration and ensemble estimationRealistic rainfall pattern representation quantified by ensemble Structure‐, Amplitude‐, and Location‐error metrics [ABSTRACT FROM AUTHOR]