This research presents a comprehensive canopy height map of French Guiana at 10 m spatial resolution, employing a data fusion approach integrating optical (Sentinel-2), radar (Sentinel-1 and ALOS), and ancillary data sources. The primary objective is to leverage a U-Net neural network model, trained and validated using Global Ecosystem Dynamics Investigation (GEDI) data as reference canopy height. We aim at understanding how canopy height prediction models can be improved through the integration of relevant remote sensing and environmental descriptors related to canopy structure. The accuracies of the generated canopy height maps are assessed against high-resolution airborne LiDAR (ALS) acquisitions conducted by the French National Forest Office. We observe that enriching input data with height above nearest drainage (HAND) as well as forest landscape information yielded improved accuracies for the prediction models. Moreover, accounting for GEDI database uncertainties, through filtering of usable waveforms and correction of geolocation errors, also resulted in a performance gain for canopy height estimation using a U-Net model.

Lidar (Light Detection and Ranging) has become an essential part of the remote sensing toolbox used for biosphere monitoring. In particular, Lidar provides the opportunity to map forest leaf area with unprecedented accuracy, while leaf area has remained an important source of uncertainty affecting models of gas exchanges between the vegetation and the atmosphere. Unmanned Aerial Vehicles (UAV) are easy to mobilize and therefore allow frequent revisits to track the response of vegetation to climate change. However, miniature sensors embarked on UAVs usually provide point clouds of limited density, which are further affected by a strong decrease in density from top to bottom of the canopy due to progressively stronger occlusion. In such a context, discriminating leaf points from wood points presents a significant challenge due in particular to strong class imbalance and spatially irregular sampling intensity. Here we introduce a neural network model based on the Pointnet ++ architecture which makes use of point geometry only (excluding any spectral information). To cope with local data sparsity, we propose an innovative sampling scheme which strives to preserve local important geometric information. We also propose a loss function adapted to the severe class imbalance. We show that our model outperforms state-of-the-art alternatives on UAV point clouds. We discuss future possible improvements, particularly regarding much denser point clouds acquired from below the canopy.

Monitoring the functioning of tropical forests is central to assessing both their biological diversity and their role in global biogeochemical cycles. Technological opportunities offered by improved spatial, spectral and temporal resolutions of space-borne sensors, but also by drone (UAV)-based passive and active (LiDAR) systems provide new research avenues. We present early results from the French Guiana phenological observatory, highlighting the potential of these tools for upscaling leaf level ecophysiological measurements and plant architecture and traits to landscape-level eddy flux observations and global vegetation models.