A Probabilistic Approach for InSAR Time-Series Postprocessing

Monitoring the kinematic behavior of enormous amounts of points and objects anywhere on Earth is now feasible on a weekly basis using radar interferometry from Earth-orbiting satellites. An increasing number of satellite missions are capable of delivering data that can be used to monitor geophysical processes, mining and construction activities, public infrastructure, or even individual buildings. The parameters estimated from these data are used to better understand various natural hazards, improve public safety, or enhance asset management activities. Yet, the mathematical estimation of kinematic parameters from interferometric data is an ill-posed problem as there is no unique solution, and small changes in the data may lead to significantly different parameter estimates. This problem results in multiple possible outcomes given the same data, hampering public acceptance, particularly in critical conditions. Here, we propose a method to address this problem in a probabilistic way, which is based on multiple hypotheses testing. We demonstrate that it is possible to systematically evaluate competing kinematic models in order to find an optimal model and to assign likelihoods to the results. Using the B-method of testing, a numerically efficient implementation is achieved, which is able to evaluate hundreds of competing models per point. Our approach will not solve the nonuniqueness problem of interferometric synthetic aperture radar (InSAR), but it will allow users to critically evaluate (conflicting) results, avoid overinterpretation, and thereby consolidate InSAR as a geodetic technique.