Deep learning based retrieval of biomass and forest height using SAR images.

Aboveground forest biomass has fundamental role in carbon cycling, climate processes, and global warming. Information on the spatial distribution of biomass is a prerequisite for understanding and managing the processes involved in the carbon cycle, and supports international policies for climate change mitigation and adaptation. Several remote sensing techniques were used to map biomass, and yet, biomass cannot be estimated without uncertainty. One of the uncertainties is related to the missing element of the vertical structure of the forest in the remote sensing images. Unlike optical images, the signal from SAR can penetrate the tree canopy and sense the ground beneath the foliage. Therefore, SAR images comprise the information about the structure of the forest and thus are a useful source of information to address the uncertainty in biomass mapping. This proposal aims to apply Deep Learning to SAR images to resolve the uncertainty and estimate biomass in tropical forest regions.