Satellite radar interferometry (InSAR) techniques can monitor the ground deformation with millimeter precision. With Time-series InSAR (TInSAR) methodology, the ground deformation time series can be derived from InSAR observations. One of the important ways to analyze the InSAR deformation time series is to parameterize the InSAR deformation time series with deformation models.
The previous ways of modelling InSAR deformation time series are usually point-wise, i.e. they focus on the deformation models of single InSAR measurement points. The deformation model of each point is either assumed to be a linear function of time, or is selected from the predefined alternative models. The point-wise modeling methodologies can well interpret the deformation behavior of each point, but is limited on modeling the spatial deformation patterns.
In this study, we design and implement methodologies to model the spatio-temporal deformation patterns, based on given spatial smoothness information of the deformation. We introduce a work flow to digest the spatial smoothness information from external sources, and use the information to improve the functional and stochastic model. We also propose a model selection methodology based on hypothesis testing to select the the most probable spatio-temporal deformation model from given potential models. The spatio-temporal deformation modeling methodology is applied to the simulated data, as well as the real InSAR measurements. We apply the spatio-temporal deformation methodology to study the deformation in a hydrocarbon production field in California, and successfully detect the instantaneous uplifting and subsiding events. Based on the simulation and real case study, we conclude that given proper contextual information, spatio-temporal deformation modeling is able to derive the deformation model in both temporal and spatial domain, and has a good performance on parameterizing the non-linear deformation behavior in the temporal domain.