Measuring rainfall using cell phone links: classification of wet and dry periods using satellites

Abstract

Commercial cellular communication networks can be used for detecting precipitation by looking at the attenuation of the electromagnetic signals transmitted between antennas from microwave links. This attenuation can be translated into a path-averaged rainfall intensity. As the received link signal also fluctuates during dry period, a reference level is needed to represent dry weather in order to calculate the amount of attenuation due to rainfall. This requires a method to separate wet and dry periods. One classification methodology for wet and dry spells is the ”link approach” in which a 15-min interval is labeled wet if the mutual decrease in minimum received powers of nearby links in the same interval exceeds two thresholds. Another methodology is to use ground-based radar rainfall intensities to identify wet and dry spells, called the ”radar approach”. Conditions can be such that these methods can not be used. Satellites can be a good alternative for wet-dry classification (called the ”satellite approach”). Three Meteosat Second Generation products were tested for the Netherlands: Precipitating Clouds, Convective Rainfall Rate and Cloud Physical Properties. The satellite products were first analyzed separately both visually and quantitatively over a period of four days and compared to gauge-adjusted radar data. The Convective Rainfall Rate product proved to be of limited use for wet-dry classification, so only the Precipitating Clouds and Cloud Physical Properties products were tested as a wet-dry classification for rainfall maps calculated from link data without wet-dry classification or filter for several days. The Precipitating Clouds and Cloud Physical Properties products were combined to use as a wet-dry classification, which improved the results. Finally, link-based country-wide rainfall maps for a 12-day validation set were derived for the Netherlands employing the different wet-dry classification methodologies (link approach, radar approach, satellite approach) with and without use of a filter to remove malfunctioning links. Those maps were compared and validated against gauge-adjusted radar data, considered as ground-truth, showing that the satellite approach could be a good alternative to the link and radar approaches when using a filter.