Impacts of topography for forest height estimation

Abstract

Forest height is one of the important ecological elements and its estimation has been developed recent years. Forest height estimation is usually obtained through the scattering centers between canopy and ground using Pol-InSAR data. There are three classes of method for forest height estimation, phase-difference between interferograms, coherence optimization based on coherent / incoherent scattering models and phase-coherence inversion method. However, most of these methods are carried out over relatively flat earth surface, that's mean there are only a little slope or no slope in test site for forest height estimation. When these methods are used for forest height estimation in mountainous region, there may be big errors. In order to analyze the impact of topography for forest height estimation, simulated and analysis experiments with zero temporal decoherence were launched. Impacts of slopes in slant range ground and azimuth ground on backscattering coefficient, scattering mechanism, interferometric coherence and forest height were studied through simulated Pol-InSAR data sets. Impact of topography on forest height estimation is investigated through simulation. A couple of images were simulated using ESAR configuration in Culai test site with DEM obtained by Lidar. The other couple of images were also simulated using ESAR configuration with flat earth surface hypothesis. Several forest height estimation methods were used in the two experiments. The result shows that it is able to estimate forest height over complex topography using simulated Pol-InSAR data by phase-difference between interferograms, coherence optimization and phase-coherence inversion method. Varied topography makes the accuracy of forest height estimation lower than that of the hypothetical subdued topography. It is concluded that topography affects surface phase fixation and therefore affects the accuracy of forest height estimation using phase information. Ground range slope affects the backscattering coefficients and estimated forest height. Meanwhile, azimuth ground slope has little effect on these fields. Both of increasing ground range slope and azimuth slope will cause a little degressive coherence coefficients and large residual phases in ground range and flight direction, and neither of them leads to scattering mechanism variation as slopes in ground range and azimuth increases from 5% to 40%.