Identification of dry and rainy periods using telecommunication microwave links

Microwave links from telecommunication networks (MWL) provide rainfall data at high spatial (<1 km 2 ) and temporal (1-5 min) resolutions, but currently no standardized approach exists to process MWL signals. In this study, we therefore discuss three novel methods to classify every measurement of the signal strength as either belonging to wet or dry periods: i) a moving window algorithm, ii) a statistical classification algorithm using random forests and iii) an algorithm based on a Gaussian factor graph, which is a rather novel signal-processing method. We find that the first algorithm is straight forward to apply, but shows a mixed performance. The random forests deliver minimal classification errors of precipitation, but require an extensive dataset. The last algorithm is very elegant and can best cope with low quality MWL data, due to the internal dynamic system model. As expected, the results are very much dependent on the training procedure and data of the algorithms. We find that the quantization of the MWL signal is a major influence factor and conclude that MWL should be selected carefully.