Physics-Informed Feature Engineering 1D-CNN for Multilayer Cloud Detection from Geostationary Satellites

Multilayer cloud detection from active--passive observation is vital for numerical weather prediction. In this study, channel selections derived from threshold-based algorithms are embedded as feature-engineering priors into a 1D-CNN, and machine learning (ML) is used to learn latent physical relationships to simplify physical retrievals for operational deployment. The results show that the 1D-CNN achieves a multilayer-cloud probability of detection (POD{mul}) of 0.620 and a false alarm rate (FAR{mul}) of 0.240, outperforming the conventional threshold algorithm (POD{mul} = 0.558, FAR{mul} = 0.369). These results demonstrate that prior physical knowledge derived from radiative transfer theory can serve as an effective feature-engineering prior. Further experiments show that ML-revealed physical mechanisms can also enhance traditional algorithms. Replacing AGRI channel 12 (C12, centered at 10.8 μm) with channel 13 (C13, centered at 12.0 μm) increased POD{mul} from 0.558 to 0.609 without materially affecting FAR{mul}. However, for AHI, substituting the 11.2 μm channel with the 12.3 μm channel yielded negligible improvement. In addition to spectral response function (SRF) mismatches, a primary contributing factor is the channels' on-orbit radiometric stability. Hence, physics-informed machine-learning methods appear promising for advancing remote-sensing AI, while sensor-specific characteristics must be considered during operational transfer.