Forest Canopy Height Estimation from TerraSARX/TanDEMX InSAR DEM

Estimating forest height in large region are challenging forest investigation and forest remote sensing. This paper investigated the possibility of estimating forest canopy height from TerraSARX/TanDEMX polarimetric and interferometric data and topographic map in the forest of Mengla in Yunnan, China. The TerraSARX/TanDEMX data was interfered to produce four polarimetric digital surface model (DSM), which contained forest canopy height. Meanwhile, the digital terrain model (DTM) was extracted from the topographic map. After horizontally coregistering DTM with each polarimetric DSM, the relative height models were separately computed from DTM and each polarimetric DSM as RDSM and RDTM by referring to the same flat nonforest plot. Then forest canopy height was estimated from the difference between RDTM and each polarimetric RDSM, and it finally produced four forest canopy height estimations. Forest inventory height was used to validate the reasonability of these estimated canopy heights. The result showed that the elevation datum of DTM was lower than that of DSM, and a flat agriculture plot could be used as the reference surface to extract RDSM and RDTM. Comparison to the forest inventory record showed that, on the plot scale, there were some significant positive correlation between the estimated canopy heights and the inventory height, and the estimated height value was reasonable. It was also found that the estimation on planted forest was a bit better than on natural forest. The correlation between the estimated canopy height and the inventory record was higher on the certain slope surface where natural forest and planted forest were mainly distributed. And there was no significant difference among the estimated heights from the four polarization channels. The result suggests that DEM difference between RDSM from TerraSARX/TanDEMX data and RDTM from topographic map could provide an effective tool to measure forest canopy height in large areas.