Hybrid Quantum CNNs spot volcanic clouds in weather-satellite imagery

A new paper on arXiv (2608.00072) from researchers Federica Torrisi and colleagues explores using hybrid quantum convolutional neural networks (QCNNs) to detect volcanic clouds in geostationary satellite multispectral imagery. The work addresses a long-standing challenge in remote sensing: volcanic clouds look similar to meteorological clouds, have variable eruption signatures, and geostationary sensors like SEVIRI offer coarse spectral sampling. By integrating quantum computational layers into a classical convolutional framework, the team built two QCNN variants using 2 and 4 qubits.

These models were trained to classify SEVIRI images containing volcanic clouds composed of ash, SO2, or mixed components, alongside non-volcanic backgrounds. The hybrid quantum-classical architectures were compared against purely classical CNNs to evaluate whether quantum phenomena such as superposition and entanglement can improve feature extraction for Earth Observation data. The paper highlights potential applications in aviation safety, real-time eruption response, hazard assessment, and climate impact evaluation. While the results are preliminary, this work is part of a growing trend exploring quantum machine learning for practical remote sensing tasks, potentially enabling faster and more accurate global monitoring of volcanic emissions.