Correlation of multi-platform SAR data for multi-temporal slope instability analysis of the Benevento Province

A-DInSAR techniques allow to identify and map new deformation processes and to update inventories of existing landslide phenomena over large time periods and at different spatial scales. Multi-temporal data derived from both PSInSAR and ISBAS techniques were combined to map the state of activity of slow-moving landslides affecting several urban areas of the Benevento Province (Campania region, south Italy). The proposed method was performed at two different scale, provincial and local, and it was structured into three phases: (i) grid-based approach; (ii) multi-platform approach; (iii) state of activity matrix approach. Two vector grids with uniform reference cell size were generated at two different resolutions, 20x20 m and 100x100 m. These grids allowed the homogenization of ERS 1/2 (1992-2000), ENVISAT (2002-2010) and RADARSAT (2003-2007) PS and Sentinel-1 (2017-2020) ISBAS data in the same vector format. A statistic value calculation was executed from the velocity along LOS of each dataset first on the 20 m cell grid and then from this to the 100 m cell grid. In order not to overestimate the deformation velocity, for the 100 m cell grid a Weighted Average Velocity (V_WA) was computed, which takes into account both the statistically calculated mean value and the area actually covered by the cells. With the multi-platform approach the resulting V_WA maps were analysed individually and also compared to each other. This step leads to a single scale velocity representation, which allows a better multi-temporal observation of the movements affecting the Benevento Province. From the results obtained an activity threshold of ± 3 mm/y was also established and a preliminary stability code was executed for each cell to discriminate stable, unstable and no data areas. These classes were used to construct four two-factor matrices by combining pairs of temporally consecutive satellite data (e.g. ERS-ENVISAT); four activity maps were then obtained. Although this step already provided results for the identification of potential hotspots, in order to achieve a complete deformation overview of the province, two three-factor matrices were processed. Two “historical” state of activity maps for the entire time-span considered (1992-2020) were thus generated. The examination of both V_WA and state of activity maps at 100 m cell grid and their comparison with pre-existing landslide maps available for the Benevento Province allowed to identify specific hotspots interested by currently active deformation processes, corresponding to built-up areas and infrastructures. For representative case studies a detailed analysis of the PS distribution and deformation trends was carried out, also including their correlation with rainfall events. The V_WA and state of activity maps were produced at 20x20 m resolution and made it possible to reconstruct the deformation history of each case.

The methodology applied demonstrates how the availability of multi-temporal satellite data allows interpretation at different spatial scale. The results achieved can be conceived as proper management tool for the assessment of slow-moving landslides, enabling the study of deformation processes, in terms of state and distribution of activity.