Development of a Python Framework (pyRadman) for QPE using radar and CML data at DWD

The daily life of everybody is affected by weather, especially by precipitation events. Climate projections indicate that the number and intensity of heavy rain events could increase in future. Therefore, the interest to improve QPE has increased rapidly, particularly for assurances, public infrastructure and flood risk management.

Currently, the QPE is calculated by using the RADOLAN algorithm of Deutscher Wetterdienst. This algorithm combines the data of 17 weather radars and roughly 1,200 rain gauges in Germany by adjusting the radar reflectivity to the precipitation amount measured at the ground. The adjustment process is done every 10 minutes using the hourly total of radar and rain gauge data. Due to the rain gauge data delivery the adjustment process is delayed by 25 minutes. For this reason, short convective precipitation events can only be observed insufficiently.   

Therefore, the RADOLAN algorithm has to be adapted to improve the QPE based on shorter data accumulation time and contemporary data delivery. One approach is to use almost real-time available data from the telecommunication network. Rainfall leads to attenuation of the signal level of commercial microwave links (CMLs). The path integral of the attenuation along one sender-receiver pair can be related to a certain precipitation amount. Germany is covered by several thousands (~130,000 in total) of CMLs, which can potentially be used to quantify rainfall events. Especially in urban areas the density of CMLs exceeds the density of meteorological networks rain gauges clearly. Therefore, it becomes possible to observe convective extreme weather events with higher temporal and spatial resolution.

Within the project HoWa-PRO, the Deutscher Wetterdienst (DWD) collaborates with University of Augsburg, Karlsruhe Institute of Technology in Garmisch-Partenkirchen, and Ericsson. One of the first tasks is to set up a continuous data flow from Ericsson to DWD. To investigate different combinations of data sources and adjustment intervals, a fast, flexible and expandable software framework for combining and processing this data has to be developed.      

We present first results of QPE after adjustment using different combinations of data e.g. CML+radar data and gauge+radar data. These results were analyzed and compared to show the potential of using opportunistic data from CMLs for radar adjustment.