Various semi-empirical models for linking PolInSAR data (polarimetric synthetic aperture radar interferometry) to canopy height of vegetation exist. However, only single-polarized data were used during the TanDEM-X mission in order to create a global digital elevation model (DEM). Therefore, simplifications of the semi-empirical models have to be applied to use the PolInSAR models for canopy height estimation with single-polarized TanDEM-X data. We extracted the volume coherence from TanDEM-X acquisitions and used a linear as well as a sinc model for the estimation of canopy height, which are based on the semi-empirical Random Volume over Ground model (RVoG). Both, the linear as well as the sinc model, were applied in temperate forests of Germany and boreal forests of Canada. The estimated canopy height was validated with LiDAR based canopy height models. In general, the sinc model resulted in higher coefficients of determination R2 from 0.08 to 0.64 and lower root mean squared errors (RMSE) between 4.8m and 12.5m compared to the linear model with R2 values between 0.08 and 0.62 (RMSE=5.4m to 13.5 m). Higher accuracies were generally achieved in winter and with higher height of ambiguity.
|Number of pages||13|
|Journal||International Journal of Applied Earth Observation and Geoinformation|
|Publication status||Published - Oct 2019|
- Vegetation height
- Temperate forests
- Boreal forests