Enhanced simulation of gross and net carbon fluxes in a managed Mediterranean forest by the use of multi-sensor data

Abstract

The current paper presents the last advancements introduced into a modelling strategy capable of simulating gross and net forest carbon (C) fluxes, i.e. gross and net primary and net ecosystem production (GPP, NPP and NEP, respectively). The simulation is performed by combining the outputs of a NDVI driven model, Modified C-Fix, and a bio-geochemical model, BIOME-BGC, taking into account the effects of forest disturbances. The proposed advancements are aimed at improving the model performance in managed Mediterranean forests and concern: i) the calibration of C-Fix GPP sensitivity to water stress; ii) the quantification of the green, woody and soil C pools which regulate the prediction of NPP and NEP. These two issues are addressed by the processing of additional remotely sensed datasets, i.e. low spatial resolution satellite imagery and high spatial resolution airborne laser scanner data. The original and modified model versions are tested in a Mediterranean pine forest which has been the subject of several investigations and where a new eddy covariance flux tower was installed at the end of 2012. This allows the assessment of the GPP and NEP estimates versus daily tower observations of eleven years (2013–2023), while mean stand NPP estimates are evaluated against measurements of current annual increments (CAI) taken in the pine forest. The results obtained support the capability of the proposed modifications to improve the model accounting for the major environmental factors which regulate the three C fluxes. The calibration of C-Fix, in particular, improves the reproduction of the high mean daily GPP observations consequent on the moderate ecosystem sensitivity to water stress (r² increases from 0.87 to 0.91, whilst RMSE and MBE decrease from 1.65 to 1.04 and from −1.37 to −0.56 g C m⁻² day⁻¹, respectively). The quantification of the forest C pools enables the consideration of stand aging, which is decisive for the correct simulation of the relatively low NPP and NEP observations. The assessment of the final CAI estimates, in fact, yields a high accuracy (r² = 0.653, RMSE = 1.38 m³ ha⁻¹ y⁻¹ and MBE = 0.42 m³ ha⁻¹ y⁻¹); the case is similar for the mean daily NEP estimates, which accurately reproduce the flux tower observations (r² = 0.669, RMSE = 0.91 g C m⁻² day⁻¹ and MBE = 0.11 g C m⁻² day⁻¹).