Estimating canopy height and aboveground biomass in tropical mangrove restoration areas through multisource remote sensing

Mangrove restoration activities have expanded in the last decades after the incorporation of blue carbon ecosystems in the global carbon budget. In this study, Synthetic Aperture Radar (SAR) polarimetric and interferometric features along with spectral bands and indices were extracted from C-band SAR Sentinel-1, L-band SAR SAOCOM-1 and optical multispectral Sentinel-2 to estimate the forest structure of mangrove restoration areas in a tropical region. A 15-m spatial resolution reference dataset of canopy height and aboveground biomass was generated with regression models between field observations and UAV-LiDAR metrics. Through the use of a random forest algorithm and spatial cross-validation, we found that models combining all three data sources (SAOCOM-1, Sentinel-1, Sentinel-2) significantly outperformed single- or dual-source models. From 62 assessed variables, SAOCOM-1 interferometric coherences and Copernicus DEM elevation had the highest importance, followed by Sentinel-2 and SAR polarimetric variables. The performance evaluation of the final model with reduced variables resulted in R 2 = 0.58 and RMSE = 2.3 m for canopy height, while for aboveground biomass R 2 = 0.55 and RMSE = 56.4 Mg ha −1 . To assess the transferability potential of the final model, the model performance was assessed in another study area where only field data is available that were not used for training the models. This assessment showed that mangrove canopy heights and aboveground biomass under restoration could be better estimated than non-restoration mangroves in the untrained study area. Our study on forest structure estimation and mapping provides an essential reference for monitoring the progress of mangrove restoration programs. • Canopy height and biomass models with satellite predictors and UAV-LiDAR estimators. • UAV-LiDAR reveals forest structure variety in mangrove restoration area. • Forest structure complexity increases over two decades of mangrove restoration. • Forest structure complexity impacts signal saturation and model estimation accuracy. • Model suitability for monitoring mangrove restoration programs.