Transient scattering effects in Sentinel 1 ground deformation analyses of the rural UK

Abstract

Interferometric analysis of Synthetic Aperture Radar (SAR) data is widely performed to measure deformation of the ground and infrastructure. Through analysis of SAR interferograms calculated for diverse windows in time, ‘Multi-Temporal InSAR’ (MTI) analysis techniques using current imaging satellites can regularly measure certain features to millimetric precision, over regions spanning hundreds of kilometres. The accuracy of MTI measurements is drastically reduced when measuring surface features in rural areas, as the impermanent geometry and radiometric properties of vegetation and soil (which are semi-transparent at typical SAR frequencies) scatterers cause decorrelation of the interferometric phase over relatively short periods of time. While this decorrelation has historically been modelled as a zero-mean noise, more recent literature has demonstrated mechanisms whereby this model does not hold. Phenomena affecting the dielectric properties of soil and vegetation lead to gradual changes in scattering and subsequent interferometric phase variation, and the spatial and temporal correlation of these phenomena can significantly deformation measurements. This work presents a set of approaches for systematically identifying and evidencing these transient scattering processes. Distinct patterns in non-zero mean phase are identified and highlighted using a priori maps of vegetation type. These distinct patterns are associated with transient scattering behaviour of the soil and vegetation under study, and this work demonstrates that these patterns are pervasive in time and occur across a range of rural land covers and types of vegetation. This work has significant implications for MTI deformation measurements in vegetated environments. Where present, significant transient scattering phases necessitate modifications to existing approaches and the interpretations of their results. A novel method has been presented to isolate distinct patterns in transient scattering behaviour using K-means clustering of correlations found in the frequency domain. Valuable improvements to deformation measurements are subsequently demonstrated.