Estimation of subcanopy topography based on single-baseline TanDEM-X InSAR data

In forest areas, when using interferometric synthetic aperture radar (InSAR) technique to extract subcanopy topography, a key problem is how to reduce the impact of the forest-scattering process on InSAR height measurements. In this paper, we propose a new methodology for subcanopy topography extraction by making full use of single-baseline TanDEM-X InSAR data. The interferometric phase is used to estimate a high-resolution and high-precision digital elevation model (DEM), and the coherence magnitude is used to extract the scattering phase center (SPC) height associated with canopy scattering. Allowing for the difference in the sensitivity of TanDEM-X bistatic interferometry information for forests with different densities, two scenarios are involved: (1) For pixels characterized by dense forest, the X-band SAR signal cannot penetrate through the forest canopy layer. In this case, an existing scattering model is used to estimate the InSAR penetration depth from the coherence magnitude. Note that when the penetration depth is small compared with the height of ambiguity (HoA), the former can be recognized as the difference between the InSAR DEM and the actual surface height. Some sparse forest heights (or SPC heights) measured by the Ice, Cloud, and land Elevation Satellite-2 (ICESAT-2) system are then used to help determine the linear relationship between penetration depth and sparse forest height (or SPC height). Finally the SPC can be estimated and removed from the TanDEM-X InSAR DEM. (2) For pixels characterized by sparse vegetation, the X-band likely penetrates the vegetation volume layer to ground level. For this scenario, a coherence threshold is set to identify these pixels, and the corresponding SPC heights can be considered zero where the subcanopy topography is approximately equal to the InSAR-generated DEM. The proposed method was validated by combining TanDEM-X coregistered single look slant range complex (CoSSC) data from two test sites with different forest types with ICESAT-2 points. The results are as follows. For the Krycklan test site, compared with the validation data, the RMSE varied from 7.76 m for the InSAR DEM to 3.22 m for the reconstructed subcanopy topography, which is an improvement of 58.5%. For the Oyan test site, subcanopy topography achieved an RMSE of 7.26 m compared with the RMSE of 27.12 m before SPC height correction, which represents an improvement of 80.1%. In addition, we investigated the impact of the topographic slope and the necessity of using the coherence threshold. For inconsistencies between TanDEM-X bistatic InSAR data and ICESAT-2 measured forest height over time, the fitting between penetration depth and SPC height calculated by ICESAT-2 subcanopy topographic elevation and InSAR-generated DEM can be used. These investigations confirm the feasibility and reliability of the proposed method.