Forest Vertical Structure of Sierra Nevada Observed by the First Spaceborne Tandem SAR Mission

Abstract:

DLR’s TerraSAR-X (TSX, 2007 launched) /TanDEM-X (TDX, 2010 launched) (abbreviated as TDM) mission is the first and sole Spaceborne tandem SAR mission, allowing for consistent and continuous mappings of global forest structure without any temporal decorrelation. This study tested performance of three single-polarized (HH) TDM acquisitions for forest structure estimation over a coniferous forest of Sierra Nevada. Complex interferometric coherences were calculated and applied in a widely used scattering model – the Random Volume over Ground (RVoG) model to retrieve the forest structural properties. Scattering phase height (SPH) was derived from the interferometric phase for an evaluation of the average scattering height of this forest and the overall penetration capability of TDM X-band signal over this area. Since only single-polarized TDM data are available, in order to keep the RVoG inversion balanced and solvable, either one of the following two strategies can be applied: 1) simulating the ground phase using a digital terrain model (DTM) from Lidar, and 2) assuming a fixed extinction coefficient value for the entire study area when a Lidar DTM is not used. This study firstly used a 5-m Lidar DTM to estimate the SPH and RVoG forest height. All SPH of the three TDM acquisitions have demonstrated strong correlations to the Lidar RH100 (canopy top height) with r² of 0.75~0.77, bias of -11.60~-12.58m (i.e. the mean height difference between SPH and canopy top height; this bias can be associated to the average penetration depth) and RMSE of 12.79~13.73m at 90m. An r² of 0.79~0.81, bias of 0.37~1.97m and RMSE of 5.11~5.56m was observed between all the RVoG forest height and Lidar RH100. The study also tested the second method by fixing the extinction value, i.e. without applying a Lidar DTM for the ground phase. Despite the large variety of forest density and height over this area, the derived RVoG forest height has a surprisingly good correlation to the Lidar RH100 with r² of 0.45~0.55, bias of 1.01~1.61m and RMSE of 7.65~8.85m for all TDM acquisitions. These results are promising as they demonstrate the great potential of the TDM mission for very large-scale forest structure estimation in the future.