Rain detection using commercial microwave link data and k-means clustering

Rain monitoring is crucial for preventing natural disaster damage and for agriculture. Commercial Microwave Link (CML) data has been used to predict rain events, especially when Rain Gauges (RGs) or radars are scarce. In this work, we modified a classic variance approach with k-means clustering for rainfall classification. For this, CML data was applied and validated using RG data. The results showed that the Unsupervised Learning (UL) approach is sufficient for classification. The proposed approach achieved a precision of 82% for a 1 hour and 15 minutes window. The classical is limited because it needs a priori RG data.