Abstract. As the global community intensifies efforts to combat climate change, insights on the influence of management on forest carbon stocks and fluxes are becoming invaluable for establishing sustainable forest management practices. However, accurately and efficiently monitoring carbon stocks remains technologically challenging. In this study, we aim to (1) assess the effect of forest management on carbon stock by comparing unconfounded pairs of managed and unmanaged forests in the National Park Brabantse Wouden (Flanders, Belgium) and (2) leverage the complementary strengths of optical, light detection and ranging (lidar), and synthetic aperture radar (SAR) remote sensing technologies to improve overall accuracy and scalability in carbon stock estimation. Remote sensing data from Sentinel-2, Sentinel-1, and a canopy height product derived from the Global Ecosystem Dynamics Investigation (GEDI) mission and Sentinel-2 were used as predictors in a generalized additive model (GAM) to estimate carbon stock. The combination of Sentinel-1 and Sentinel-2 significantly improved model accuracy (R2=0.73, RMSE=59.21 t ha−1, MAE=50.29 t ha−1) compared to a model using only Sentinel-2 indices (R2=0.56, RMSE=99.44 t ha−1, MAE=91.40 t ha−1). The addition of canopy height estimates did not affect the model fit. While field assessment exhibited higher carbon stocks in unmanaged stands compared to managed ones, this difference was not detectable using a remote sensing model that incorporated Sentinel-2, Sentinel-1, and/or GEDI-derived variables. Potential explanations for this discrepancy include signal saturation and the need for more training data.
Winckel et al. (Thu,) studied this question.