**4. Discussion**

It can be seen from the results that the link-experienced rain attenuation could be much higher than the ITU model [28]-predicted attenuation value calculated using the rain gauge records. One possible reason was the location of the rain gauge. The weather station used in this study was located on the rooftop of a business building. A non-rooftop weather station operated by SMHI near the measurement area is available, but the data was not collected very frequently, making comparison with the microwave link derived rain rate at 15 minutes interval di fficult. The daily rainfall amounts recorded by the SMHI rain gauge were 39 and 9 mm on 7 and 11 June, respectively, while the recorded daily rainfall values by the rain gauge used for this study was 17.4 and 7 mm, respectively, during these two days. Compression of atmospheric streamlines will produce significant Bernoulli lifting and cause under catch of rainfall in the rain gauge located on the rooftop. The actual rainfall along the links was greater than the rain gauge measurements, which was also indicated by the bias calculation results. The positive bias values show that the rainfall estimated based on the attenuation of the mmWave link signals is higher than the rain rate recorded by the rain gauge. As expected, the mmWave links detected more rainfall than the rooftop rain gauge, which underestimated the rainfall amount.

The widely deployed microwave links from existing cellular networks have become installation-free facilities, and they can be particularly useful for applications in urban hydrology and in supporting monitoring in flood-prone urban areas. Rain gauges have high accuracy but the measurement data is collected at the point scale. Because of the temporal and spatial variability of rainfall, especially in the heavy rain and flooding events, the changes of rain rate occur at very short time intervals. Even in dense urban areas, the density of rain gauges is often not su fficient to capture significant variation in observed rainfall. Radars and satellites can monitor over much wider areas, but the estimates from those sources are less accurate at near-ground levels, and the temporal and spatial resolution is not su fficient for flood monitoring purposes. The measurements presented here records a received signal level measurement every 30 seconds, producing 10 rain rate estimates within 5 minutes, which is usually the time resolution of weather radars. Microwave backhaul links can measure near-ground rain rates more accurately with high time resolution, which could assist in flash flood warnings [43–45], as the wireless networks exist over large regions of land, including complex topographies, where traditional monitoring equipment cannot be easily installed. Therefore, this technology could also be very useful in cities that lack monitoring of rainfall by radar. As more mmWave links are expected to be widely deployed in cities, those densely deployed links can all be used for rainfall monitoring, and the amount of rainfall estimation data will be impressively large. These links can be further processed using standard interpolation methods to create rainfall maps, or combined with other existing monitoring networks based on radars and rain gauges to produce high temporal and spatial resolution rainfall maps [46].
