Multi-Temporal Pixel Trajectories of SAR Backscatter and Coherence in Tropical Forests

Forest cover dynamics and disturbance can be tracked using a pixel based time-series analysis of multi-temporal Interferometric Synthetic Aperture Radar (InSAR) backscatter and coherence data. In particular, derived features from pixel trajectories in time can be a powerful tool to map changes in tropical forest, where deforestation and forest degradation occur driven by a series of processes such as fire, selective logging, subsistence agriculture and complete clearance of forest due to large scale deforestation. The research presents results from tropical forest environments in Cameroon, Republic of Congo and Indonesia.  Several SAR data with different frequency and resolution were tested including ENVISAT ASAR, ALOS PALSAR and TanDEM-X. Furthermore, the analysis was undertaken on both TanDEM-X backscatter and coherence at HH polarization. Multi-temporal coherence was employed due to its sensitivity to the upper canopy volume, which causes decorrelation as a function of the amount of vegetation (e.g. disturbance event).

A pixel trajectory is defined as a set of values of all resolution elements (backscatter or coherence) at the same row and column position in the stack of images. The stack is generated by multi-resolution analysis (MRA) at a number of spatial resolutions, enabling analysis in the combined time and space domains. Analysis of the trajectories over an area by means of a set of parameters (features) that characterize its time evolution can give insight on the nature and changes of landcover. The following set of trajectory features was computed: running ratios with respect to a baseline year, linear fitting (trend), coefficient of determination (goodness of fit), dispersion around trend, maximum change relative to mean (swing) and statistics of first derivative (variance, kurtosis). These features are designed to detect in each pixel trajectory the presence of a linear trend, the stationary of the distribution around the linear regression, the occurrence of intermittent events, and the dynamic range of the changes.