High-resolution urban precipitation monitoring in a tropical megacity using Commercial Microwave Links (CMLs)

Accurately monitoring urban precipitation is notoriously difficult, even more so in many parts of Africa where the coverage of ground-based weather radars is usually sparse. The built-up environment near radars may result in underestimation due to beam blockage and local overestimation due to clutter. In addition, due to the strongly varying urban canopy and terrain, rainfall measurements from gauges are often only representative for an even smaller area than in rural terrain. At the same time, accurate and timely precipitation measurements and, consecutively, forecasts are crucial because the response time of urban catchments is very fast and costs associated with potential damage are high.

Coincidentally, urban areas with high population density can generally be associated with high mobile phone traffic, which in turn requires a high density of cell phone towers to transmit the signals. The wireless connection between two cell phone towers, more commonly referred to as a Commercial Microwave Link (CML), is known to be (partially) attenuated by rainfall. Deriving the magnitude of the attenuation from the transmitted and received signal levels stored operationally in the network management system of mobile network operators, one can estimate the average rainfall intensity over the path. 

In this study we use a dense network of several thousand CMLs in the megacity of Lagos, Nigeria, to estimate path-averaged rainfall intensities. We employ the open-source R package RAINLINK to process the 15-minute CML data into path-averaged rainfall intensities, and, where available, evaluate these against dedicated rain gauge measurements and satellite precipitation retrievals. In addition, based on the CML-derived rainfall intensities, 2D rainfall maps are created using a geostatistical interpolation method. This study highlights both the benefits and complexities of using a very dense network of CMLs with, predominantly, a sub kilometer path length and limited reference rainfall data.