Using telecommunication microwave links to identify faulty rain gauge in urban areas

For urban hydrology, rainfall intensity must be monitored at a high temporal and spatial resolution, which is often achieved using networks of rain gauge. The strong spatial and temporal variability of rainfall makes it difficult to spot faulty rain gauges, in particular when their density is limited. It has been recently demonstrated that measurements from telecommunication microwave links can be used to monitor rainfall. Taking advantage of the high density of microwave links in urban areas, we propose a method to identify faulty rain gauges using neighboring telecommunication microwave links. The method is based on the co-fluctuation between rain gauge and microwave link measurements when rainfall occurs. Depending on the frequency of the microwave link, the relation between the specific attenuation and the rain rate is non-linear. In addition, because of the spatial variability of rainfall, of the difference in sampling volume between the two types of instruments, as well as the uncertainty in rain rate values for a rain gauge and in the estimation of the attenuation solely due to rain for a microwave link, a perfect match between rain gauge and microwave link rain rate estimates can not be expected. To keep the method simple and fast, we focus on the comparison of dry and rainy occurrences seen by a rain gauge and the neighboring microwave links, by means of daily contingency tables. In this way it is not necessary to convert link measurements into rain rate estimates, which limits the sources of uncertainty. The method is tested on data from 14 microwave links and 14 rain gauges in the Zürich area in Switzerland, and correctly detects faulty rain gauges. The main limitation is the fact that biases affecting rain gauges data only during rainy time steps cannot be detected. But this type of error can however be assumed to be rare.