H020-08
Opportunistic Rainfall Nowcasting with Commercial Microwave Link Data

Monday, 7 December 2020: 17:58
Virtual
Ruben Olaf Imhoff1,2, Aart Overeem1,3, Claudia Brauer1, Hidde Leijnse3, Albrecht Weerts1,2 and Remko Uijlenhoet1, (1)Wageningen University and Research, Hydrology and Quantitative Water Management Group, Wageningen, Netherlands, (2)Deltares, Operational Water Management, Delft, Netherlands, (3)Royal Netherlands Meteorological Institute, De Bilt, Netherlands
Recorded Presentation
Abstract:
For forecast horizons of three hours or less, rainfall nowcasts can play an important role in forecasting the timing and location of rainfall for early warning. Nowcasts are commonly constructed with radar-based quantitative precipitation estimates (QPE), but the use of alternative sources can be of interest in the absence of, or complementary to, weather radars. A promising option is the use of signal level data from the roughly four million commercial microwave links (CMLs) worldwide. These are close to the ground radio connections used in cellular telecommunication networks. Rain-induced attenuation and, subsequently, path-averaged rainfall intensity can be retrieved from the signal’s attenuation between transmitter and receiver.

As a proof of concept, we demonstrate the use of country-wide rainfall maps from CML QPE for rainfall nowcasting in the Netherlands. We created probabilistic nowcasts with CML and radar QPE by employing the pySTEPS nowcasting algorithm in a probabilistic sense (20 ensemble members) for twelve summer days in 2011.

Generally, CML nowcasts compare well to the radar rainfall nowcasts. Estimated locations and shapes of the rainfall fields are, however, not as accurate in the CML QPE as in the radar QPE, which also affects the quality of the nowcasts. However, CML rainfall volumes are closer to the observed amounts than the non-bias-corrected radar QPE, which results in better high-intensity rainfall forecasts in real time (e.g. up to 50% higher CSI for a 5 mm h-1 intensity).

We identify two limitations for nowcasting with CML data. First, the effective measurement resolution of the CML data is usually coarser than the 1 km2 on which the data is projected, which requires considerable upscaling of the forecasts to scales that can be larger than the application scale. Second, the limited CML data coverage in some parts of the country affects the rainfall advection derivation in the nowcasting algorithm. The use of motion fields from numerical weather prediction models or satellites can potentially overcome this issue. Hence, we see potential for rainfall nowcasting with CML data, especially in urban areas with a high CML density.

<< Previous Abstract | Next Abstract