Side Event @ ESA Φ-lab PHInovation Summit

• Asbestos mapping from multispectral imagery: using Sentinel-2 spectral bands, texture features, and machine learning classifiers to identify legacy ACM rooftops in urban areas without requiring full hyperspectral acquisition

‍

• Methane and trace gas detection: leveraging SWIR bands (Sentinel-2) and coarse-resolution atmospheric retrievals (TROPOMI / Sentinel-5P) to localise point-source CH₄ emissions and bridge global atmospheric data with urban-scale monitoring

‍

• Generalisability across hazards: exploring whether analogous spectral and spatial proxy approaches extend to heavy metal contamination indicators, hydrocarbon spills, or industrial dust plumes

‍

• AI as the resolution bridge: how deep learning, transfer learning, and data fusion can compensate for spectral and spatial limitations when high-resolution hyperspectral data is unavailable

• Present the state of the art in low-resolution EO-based environmental hazard detection

‍

• Critically examine the trade-offs: spatial vs. spectral resolution, temporal revisit, model generalisation

‍

• Surface open challenges: sensor domain adaptation, scarce labelled data, false positive rates in complex urban scenes

‍

• Connect researchers, data providers, AI developers, and end-user communities (civil protection, urban planning, insurance, environmental agencies)