TanDEM-X digital surface models in boreal forest above-ground biomass change detection

Abstract

Satellite images provide spatially explicit information on forest change covering wide areas. In this study, bistatic TanDEM-X (TDX) synthetic aperture radar (SAR) satellite data were used to derive digital surface models (DSMs) of forest areas using SAR interferometry (InSAR). The capability of change features derived from bi-temporal InSAR DSMs to detect forest height (90th percentile of canopy height distribution, H90) and density variations was investigated. Moreover, changes in the forest above-ground biomass (AGB) were estimated from height changes between two InSAR DSMs. Bi-temporal airborne laser scanning (ALS) data, aerial orthoimages and an ALS-based AGB change map from a study area in Southern Finland were used as references. The results indicate that the InSAR height change of a forested area correlates more with vegetation density change than with height change. The correlation between the InSAR mean height change and the height change feature from ALS was 0.76 at stand level. Correspondingly, the correlation between the InSAR mean height change and the ALS penetration rate change was 0.89. The AGB changes predicted based on InSAR height change agreed well with the reference data; the root-mean-square error (RMSE) was 20.7 Mg/ha (18.5% of the mean biomass in 2012) at stand level and 27.4 Mg/ha (27.0%) for 16 × 16 m grid cells. The results show that TDX DSMs can be used to detect biomass changes of different orders of magnitude, e.g. due to logging and thinning.