An improved MSG SEVIRI wet-dry product based on a convolutional neural network

Adequate spatial coverage of precipitation measurements is not available for large regions. Many countries are only equipped with networks of meteorological stations to measure precipitation based on sparse point measurements. Attenuation data from commercial microwave links (CML) allow precipitation estimates over existing networks, such as the cellular network. However, the processing of the data requires a distinction between wet and dry. Here, satellite data in countries of the Global South play a significant role, as they can be used in place of conventional reference data to distinguish between dry periods and precipitation events.

Deep Learning (DL) can be used to develop a dry indicator based on geostationary satellite data that can be applied for dry-wet classification in CML processing. A convolutional neural network is used to process visual and infrared cloud information from geostationary satellites and to create a dry indicator. Satellite derived products exist, such as the NWC SAF products PC and PC-Ph, and are utilized as baseline products. The baseline products and the DL based dry indicator developed in this work (DL product) are evaluated with radar and station data in Germany.

The evaluation shows that the developed DL product improves the performance at day and especially at nighttime. Limitations in detecting the correct rain field area is reduced by the DL product. In total the DL product improves the Matthews Correlation Coefficient value by about 0.05 compared to the PC-Ph product.