Analysis of Temporal Smoothing for Separation of Deformation and Atmospheric InSAR Time Series Signals
Holley, R.J.¹; Wadge, G.²; Larkin, H.E.^1
1NPA Satellite Mapping, UNITED KINGDOM; ²NCEO, University of Reading, UNITED KINGDOM

InSAR processing methods commonly utilise stacks of many SAR scenes to produce measurements of ground deformation over time. Traditionally these varied methods have been loosely divided into multi-master distributed scatterer approaches such as Small Baseline Subsets (SBAS), and single-master point scatterer approaches under the collective name of Persistent Scatterer InSAR (PSI). These techniques were developed to mitigate a number of factors which impact the quality and accuracy of single-pair InSAR approaches, for example decreasing coherence over long time periods and at high baselines, and the influence of atmospheric artefacts associated with individual SAR scenes. In recent years this division has become blurred, with an increasing number of workflows combining information from point and distributed scatterers, and using hybrid approaches combining both single- and multi-master elements.

Assumptions of linear motion have also become less of a constraint, with an increasing ability to capture variations in motion through time. However this depends on the successful estimation and removal of atmospheric artefacts, which can still present a challenge, particularly in cases where the deformation signal of interest has similar spatial and or temporal characteristics to the atmosphere. This estimation commonly involves temporal smoothing of some form. However, retrieval of an accurate deformation time series depends on the choice of smoothing criteria which control partitioning of non-linear signals into deformation versus atmospheric phase screen (APS) components.

The choice of optimal smoothing criteria is clearly key to retrieval of an accurate and representative deformation time series. Rather than a qualitative estimation of these parameters, this work compares APS fields with water vapour data from MERIS and numerical weather models to evaluate the optimal smoothing for the APS model, and therefore retrieval of a best estimate of non-linear deformation. Mount Etna provides an excellent case study for this work, as it is highly susceptible to strong atmospheric effects, and has a well-characterised history of non-linear motion associated with the eruptive cycle.