Quantifying Flood Impacts on Ecosystem Carbon Dynamics Using Remote Sensing and Machine Learning in the Climate-Stressed Landscape of Emilia-Romagna

Flood events, intensified by climate change, pose significant threats to both human settlements and ecological systems. This study presents an integrated approach to evaluate flood impacts on ecosystem carbon dynamics using remote sensing and machine learning techniques. The case of the Emilia-Romagna region in Italy is presented, which experienced intense flooding in 2023. To understand flood-induced changes in the short term, we quantified the differences in net primary productivity (NPP) and above-ground biomass (AGB) before and after flood events. Short-term analysis of NPP and AGB revealed substantial localized losses within flood-affected areas. NPP showed a net deficit of 7.0 × 103 g C yr−1, and AGB a net deficit of 0.5 × 103 Mg C. While the wider region gained NPP (6.7 × 105 g C yr−1), it suffered a major AGB loss (3.3 × 105 Mg C), indicating widespread biomass decline beyond the flood zone. Long-term ecological assessment using the Remote Sensing Ecological Index (RSEI) showed accelerating degradation, with the “Fair” ecological class shrinking from 90% in 2014 to just over 50% in 2024, and the “Poor” class expanding. “Good” and “Very Good” classes nearly disappeared after 2019. High-hazard flood zones were found to contain 9.0 × 106 Mg C in AGB and 1.1 × 107 Mg C in soil organic carbon, highlighting the vulnerability of carbon stocks. This study underscores the importance of integrating flood modeling with ecosystem monitoring to inform climate-adaptive land management and carbon conservation strategies. It represents a clear, quantifiable carbon loss that should be factored into regional carbon budgets and post-flood ecosystem assessments.