Application of a Physically Informed Neural Network for the recovery of vertical greenhouse gas profiles in the Mediterranean Basin

During March 2025, three intrusions of Saharan dust affected southern Italy, with observable effects on atmospheric composition and, in particular, on greenhouse gases. A recent study conducted by the Institute of Methodologies for Environmental Analysis of the National Research Council of Italy (CNR-IMAA) documented these events through integrated in situ and remote sensing observations. Significant variations in CH₄ and CO₂ concentrations were detected in correspondence with the dust transport episodes. In this work, we propose an approach based on Physics-Informed Neural Networks (PINNs) to retrieve the vertical profile of CH₄. The results are evaluated against high-precision ground-based measurements from CNR-IMAA, in order to assess the model’s predictive accuracy and its sensitivity to atmospheric variations associated with the presence of mineral aerosols.