Regional estimates of gross primary production applying the Process-Based Model 3D-CMCC-FEM vs. Remote-Sensing multiple datasets

Process-based Forest Models (PBFMs) offer the possibility to capture important spatial andtemporal patterns of carbon fluxes and stocks in forests. Yet, their predictive capacity should bedemonstrated not only at the stand-level but also in the context of broad spatial and temporalheterogeneity. We apply a stand scale PBFM (3D-CMCC-FEM) in a spatially explicit manner at 1km resolution in southern Italy. We developed a methodology to initialize the model thatincludes information derived from the integration of Remote Sensing (RS) and the NationalForest Inventory (NFI) data and regional forest maps to characterize structural features of themain forest species. Gross primary production (GPP) is simulated over 2005–2019 period andthe model predictive capability of the model in simulating GPP is evaluated both aggregated asat species-level through multiple independent data sources based on different nature RS-based products. We show that the model is able to reproduce most of the spatial (~2800 km2)and temporal (32 years in total) patterns of the observed GPP at both seasonal, annual andinterannual time scales, even at the species-level. These promising results open the possibilityof confindently applying the 3D-CMCC-FEM to investigate the forests’ behaviour under climateand environmental variability over large areas across highly variable ecological and bio-geographical heterogeneity of the Mediterranean region