Analysis Of Multi-Temporal SAR Interferometry Time Series For Warning Signal Detection

The work proposes a new set of rules based on statistical characterization of displacement time series, which allows, under certain constraints, recognising automatically different kinematic classes, and estimating the relevant parameters useful for target characterization in time. We introduce a new statistical test based on the Fisher distribution aimed at evaluating the reliability of a displacement model with a certain statistical confidence. We also studied the reliability of other tests already introduced in literature and used for comparing two different models, namely the Akaike Information Criterion, the Bayesian Information Criterion, and the Fisher test. A performance analysis has been carried out by simulating piecewise linear displacement time series with different characteristics in terms of kinematic, level of noise, and signal length. The displacement estimations performed by using the different tests have been compared. Finally, a procedure for selecting the optimum displacement model has been defined according to the output of the performance analysis. The proposed procedure has been also tested by using displacement time series derived from Sentinel-1 datasets.