Tropical-Forest Profiles and Biomass from TanDEM-X, Single-Baseline Interferometric SAR: InSAR Performance at Higher Frequencies and Bandwidths

Abstract:

The principal method for remotely sensing forest biomass, particularly high-biomass tropical forests, is to measure vertical structural properties of the forest and relate them to biomass. Interferometric SAR (InSAR) and lidar are the two principal technologies applied to this task. InSAR profile information is constrained in the traditional, look-averaged analysis, because it measures the vertical Fourier transform of the radar power at one and only one vertical frequency specified by the baseline. Lidar produces a full profile, including all Fourier frequencies—all vertical scales of fluctuation. In TanDEM-X data over tropical forests at the Tapajos National Forest, Brazil, we show the results of potentially improving InSAR’s Fourier coverage. This is done by estimating many Fourier frequencies with a single baseline, based on the assumption that sampling of the phase height of small (~1.5 m x 2.5 m) looks of TanDEM-X is equivalent to sampling the vertical structure of the forest; a spatial ergodicity. We show a similarity between the distribution of InSAR look-phase-heights (LPH) and lidar and field profiles over 0.25-ha areas. Using Fourier transforms of the histogram of LPH over 0.25 ha areas, biomass estimation improved by about 17% over using InSAR coherence and mean forest height, making it competitive with some lidar results. The RMS of biomasses estimated about field biomass was 48 Mg/ha, with biomasses as high as 430 Mg/ha, and an average of 174 Mg/ha. Perhaps LPH distributions bear a similarity to lidar and field profiles because shorter wavelengths (~3 cm) can penetrate holes in the canopy to scatter off a sample of the vegetation with each look. This “hole mechanism” favors higher frequencies, which are able to penetrate smaller holes. Because they increase the number of look samples, higher bandwidths are also preferred.