Estimating Forest Biomass From TerraSAR-X Stripmap Radargrammetry

Abstract

Radargrammetry has a potential for forest inventories, based on the relationship between the canopy height model (CHM) and the forest variables such as biomass. The objective of this study is to describe the relationship between above-ground biomass and a stripmap TerraSAR-X radargrammetry CHM, with emphasis on accuracy and straightness of the relationship. The study was carried out in a spruce forest in south Norway, comprising biomass data from 145 plots of 250 m 2 within 21 selected stands. Above-ground biomass for the plots ranged from 0 to 338 t/ha. We derived a digital surface model (DSM) from six TerraSAR-X stripmap acquisitions by automatic stereo matching. We subtracted a digital terrain model (DTM) from the DSM and obtained a CHM. We assigned the nearest 10 m × 10 m pixel to each field plot. The height of the CHM increased linearly with biomass with 15 t/ha/m. The rmse values were 23 t/ha (18%) at the stand level and 58 t/ha (44%) at the plot level. The tendency of curvilinearity was so weak that it could hardly be distinguished from a straight linear relationship. The straightness of the relationship may enable monitoring of biomass changes without an external DTM as input. A comparison between radargrammetry and interferometric synthetic aperture radar (InSAR) showed that the relationship between the biomass and their respective CHMs was almost identical in terms of parameter estimates. The strength of the relationship was higher with InSAR. By combining ascending and descending pairs followed by editing, the performance of radargrammetry was equally good as with InSAR.