Structural and Climatic Controls on Forest Carbon and Water Fluxes across the Italian Alps

Estimating carbon and water fluxes across mountain landscapes is challenging due to environmental heterogeneity. In the Italian Alps, complex topography and fragmented land use further complicates carbon accounting. The European Green Deal recognizes forests as climate allies, requiring accurate, spatially explicit assessments of their carbon and water exchange. To address this pressing need, we first simulated carbon and water fluxes of the Italian National Forest Inventory plots using a process-based ecosystem model. These simulations, together with predictors derived from remote sensing and a hydroclimatic model, were used to train four machine learning algorithms. We developed an integrated framework combining the process-based model with these data-driven methods to generate 30m resolution maps of Gross Primary Productivity (GPP), Net Primary Productivity (NPP), and Evapotranspiration (ET) across the Italian Alps for 2020. Among the tested models, Random Forest consistently achieved the highest predictive performance. Biomass predictors mainly explained GPP and NPP, climate controlled ET. The results (R² = 0.40, 0.28, and 0.27 for GPP, NPP, and ET, respectively) provide valuable insights into the complexity and spatial heterogeneity of Alpine forest ecosystems, highlighting the complementary information captured by integrated modelling approaches. Our results demonstrate the potential of integrated modelling to generate spatially explicit estimates of forest carbon and water fluxes, offering additional perspectives on ecosystem functioning beyond those captured by individual observational datasets. The resulting maps provide unprecedented spatial detail for GPP, NPP, and ET across Italian Alpine forests, supporting ecosystem monitoring, carbon budget assessments, and forest management strategies for climate adaptation and mitigation.