"Advanced Analysis Of InSAR Displacement Time Series For Hazard Monitoring

Multi-temporal SAR interferometry (MTInSAR), by providing both mean displacement maps and displacement time series over coherent objectson the Earth's surface, allows analysing wide areas, identifying ground displacements, and studying the phenomenon evolution on long timescales. This technique has also been proven to be very useful for detecting and monitoring instabilities affecting both terrain slopes and man-made objects. In this contest, an automatic and reliable characterization of MTInSAR displacements trends is of particular relevance as pivotalfor the detection of warning signals related to pre-failure of natural and artificial structures. Warning signals are typically characterised by highrates and non-linear kinematics, so reliable monitoring and early warning require a detailed analysis of the displacement time series looking forspecific trends. However, this detailed analysis is often hindered by the large number of coherent targets (up to millions) required to beinspected by expert users to recognize different signal components and also possible artifacts affecting the MTInSAR products, such as, forinstance, those related to phase unwrapping errors. This work concerns the development of methods able to fully exploit the content of MTInSAR products, by automatically identifying relevantchanges in displacement time series and to classify the targets on the ground according to their kinematic regime. We introduced a newstatistical test based on the Fisher distribution with the aim of evaluating the reliability of a parametric displacement model fit with a determinedstatistical confidence [1]. We also proposed a new set of rules based on the statistical characterization of displacement time series, whichallows different polynomial approximations for MTInSAR time series to be ranked. The method was applied to model warning signals. Moreover,in order to measure the degree of regularity of a given time series, an innovative index was introduced based on the fuzzy entropy, whichbasically evaluates the gain in information by comparing signal segments of different lengths [2]. This fuzzy entropy index, without postulatingany a priori model, allows highlighting time series which show interesting trends, including strong non linearities, jumps related to phaseunwrapping errors, and the so-called partially coherent scatterers. The work introduces the theoretical formulation of these two selection procedures and show their performances as evaluated by simulating timeseries with different characteristics in terms of kinematic (stepwise linear with different breakpoints and velocities), level of noise, signal lengthand temporal sampling. The proposed procedures were also experimented on real MTInSAR datasets. We show results obtained by processingboth Sentinel-1 and COSMO-SkyMed datasets acquired over Southern Italian Apennine (Basilicata region), in an area where several landslidesoccurred in the recent past [3]. The MTInSAR displacement time series were analysed by using the proposed methods, searching for nonlineartrends that are possibly related to relevant ground instabilities and, in particular, to potential early warning signals for the landslide events. Theindex based on the fuzzy entropy was able to recognize coherent targets affected by phase unwrapping errors, which should be corrected toprovide reliable displacement time series to be further analyzed. The procedure based on the Fisher distribution was used for classifying targetsaccording to the optimal degree of a polynomial function describing the displacement trend. This allowed to select targets showing nonlineardisplacement trends related to the several ground and structure instabilities. Specifically, the work presents an example of slope pre-failure monitoring on