Rain detection is hydrologically relevant because it defines rainfall occurrence, event onset, dry spell duration, and the spatial footprint of rainfall. These characteristics are inherently difficult to capture accurately at sub-hourly scale, and the challenge is amplified in data-scarce regions where conventional observations are sparse, making rainfall timing and volume quantification unreliable. Observations from geostationary satellites and networks of commercial microwave links (CMLs) can help in this regard. This paper proposes a convolutional neural network (CNN) algorithm that combines CML total loss with two SEVIRI products at 15 min resolution: Cloud Physical Properties (CPP) and Precipitating Cloud (PC). The satellite products are mapped to each CML path and used with the CML total loss time series. The models are evaluated on independent CML datasets from Germany and Czech Republic with 3692 and 1445 CMLs, respectively. The novelty of the proposed framework lies in integrating CML total-loss time windows with path-matched CPP and PC information within a single CNN, evaluating how the resulting wet–dry classification affects the rainfall volume retained, missed, or falsely attributed to dry periods, and testing transfer between two independent national CML networks.</p> <p>In general, models combining SEVIRI and CML information outperform the CML-only model and fixed-threshold SEVIRI benchmark. The CML-only model retains 53 % of the reference rainfall volume and misses 35 %. Adding SEVIRI information reduces missed rainfall: the model using total loss and PC retains 81 %, the model using total loss and CPP retains 85 %, and the model using total loss with both CPP and PC retains 89 %, with only 8 % missed rainfall. The increase in retained rainfall volume with both CPP and PC directly improves rainfall accumulation for subsequent hydrological analyses. The gain is strongest around 1.0–2.5 mm h<sup>−1</sup>, while falsely detected rainfall remains low. Satellite-informed models also show more stable performance when transferred between Germany and the Czech Republic.