Retrieval of Total Suspended Matter in Inland Water Bodies via EnMAP and PRISMA Spaceborne Hyperspectral Imagery: A Dual Neural Network Approach Trained on Decadal In-Situ Spectroscopy

Total Suspended Matter (TSM) is a primary indicator of water quality. Yet, its ecological impact is dictated by its composition— specifically the ratio between Suspended Particulate Inorganic Matter (SPIM) and Suspended Particulate Organic Matter (SPOM). In Mediterranean reservoirs, where agricultural runoff and seasonal algal blooms create complex optical mixtures, retrieving these components separately is essential for distinguishing between mineral sediment transport and organic productivity. Traditional multispectral sensors struggle to decouple the distinct scattering and absorption signatures of organic and inorganic fractions. This research motivates the use of spaceborne imaging spectroscopy, leveraging the high spectral resolution of EnMAP and PRISMA to perform a simultaneous retrieval of both fractions through a dual Neural Network (Dual-NN) inversion strategy. The networks were trained using a historical dataset spanning two distinct periods (2001–2007 and 2017–2019) consisting of 163 in situ samples collected during pan-Iberian campaigns, and independently evaluated with validation data and satellite matchups (n=36 for PRISMA and n=19 for EnMAP) between 2023 and 2024. The models demonstrated a high capacity to simultaneously resolve mineral sediment transport versus algal organic productivity. Finally, spatial application over continental water bodies of the Iberian Peninsula confirmed the physical-optical coherence of the hyperspectral inversions, paving the way for operational monitoring in future missions like ESA CHIME.