Estimation of rainfall fields using commercial microwave communication networks of variable density

The use of commercial microwave radio networks which are a part of cellular communication infrastructure for mapping of the near-the-ground rainfall is challenging for many reasons: the network geometry in space is irregular, the distribution of links by frequencies and polarizations is inhomogeneous, and measurements of rain-induced attenuation are distorted by quantization. A non-linear tomographic model over a variable density grid is formulated, and its applicability and performance limits are studied by means of a simulated experiment using a model of a real microwave network. It is shown that the proposed technique is capable to accurately measure integrated near-the-ground rainfall amounts over the area of 3200 km2 with a bias smaller than 10%. In urban area, where the density of microwave links is high, the average correlation in space between the simulated model and reconstructed rainfall fields reaches 0.89 over the variable density grid with average cell size of 5.7 km2 and 0.74 when interpolated into the rectangular grid with pixel size 0.775 × 0.775 km2, for the quantization interval of 0.1 dB.