Fusing GEDI waveforms with spaceborne SAR data for improved forest structure estimation

Abstract:

The estimation and monitoring of forest vertical structure at large scales strongly rely on the use of remote sensing techniques, among which two have been identified as key technologies for forest structure estimates by the international ecosystem science community: lidar and synthetic aperture radar (SAR). Lidar remote sensing is often acknowledged as a “gold standard” tool that enables accurate measurements of forest vertical structure and estimates of biomass; but lidar has very limited data coverage worldwide because of its sampling nature. On the other hand, spaceborne SAR technique is able to provide forest maps at a high spatial resolution and global coverage; nonetheless, the retrieval of forest structure from a SAR map is complicated and less accurate, often requiring an increased observation space, for example, by acquiring interferometric pairs in different polarization channels or by collecting multiple-baseline images that are currently available only from airborne platforms.

In this study, we investigate the fusion of NASA’s Global Ecosystems Dynamics Investigation (GEDI) lidar data with data from the first spaceborne tandem SAR– TerraSAR-X/TanDEM-X (TDX) mission for improved forest structure estimation, which has the potential for mapping forest vertical structure and biomass at the global scale. The fusion framework is based on the ability and equivalence of lidar and SAR measurements to express the physical forest structure. Specifically, we use the physical distribution of vegetation elements reflected in a lidar waveform to simulate the vertical radar reflectivity profile. The radar reflectivity profile forms a Fourier pair with volume decorrelation observed by an interferometric SAR (InSAR) system. These pseudo vertical radar reflectivity profiles along GEDI tracks are assumed to be representative within each InSAR acquisition, which will be validated as the density of GEDI shots increase during mission time. This enables the derivation of forest height at a high spatial resolution from an interfermetric SAR (InSAR) coherence image acquired at single-baseline and single-polarization state. Using pseudo radar reflectivity profiles generated from GEDI waveforms at Gabon tropical rain forest, we derive forest height at 25 m spatial resolution from TDX data. These heights are filtered to remove potential underestimated pixels due to the saturation of X-band signal at dense tropical forest and aggregated afterwards to 100 m resolution. As the GEDI shot density and geolocation accuracy are increased in Gabon, a ground topography product will be used to improve the height estimates from TDX, resulting in a final height product of 25 m resolution. We use canopy heights from GEDI cross-over and LVIS for validation. This study on fusing GEDI waveforms with spaceborne SAR data has the potential of large-scale forest structure and biomass estimation.