Sentinel-1 Data Help Capture Pre-failure Signatures of Slope Instability – Toward Forecasting of the Temporal Occurrence of Landslides with the Aid of Multi-temporal Interferometry

Wasowski, Janusz (1); Bovenga, Fabio (2); Nutricato, Raffaele (3); Nitti, Davide Oscar (3); Chiaradia, Maria Teresa (4), 2017, Sentinel-1 Data Help Capture Pre-failure Signatures of Slope Instability – Toward Forecasting of the Temporal Occurrence of Landslides with the Aid of Multi-temporal Interferometry, FRINGE2017, 10th International Workshop on "Advances in the Science and Applications of SAR Interferometry and Sentinel-1 InSAR", pp. 94–94, Helsinki (Finlandia), 05-09/06/2017,

The regularity and higher frequency of acquisitions of Sentinel-1A/B (S-1) with respect to earlier ESA's satellite C-band sensors (ERS1/2, ENVISAT) represent clear advantages for users of multi-temporal interferometry (MTI) products. The utility of the IW acquisition mode of S-1 for regional scale slope instability detection through MTI has already been demonstrated, e.g., via studies of landslide-prone areas in Italy. In this work, we explore the potential of S-1 data for local (site-specific) scale landslide monitoring and capturing pre-failure signs of slope instability. This is done by using examples of two unstable slopes from different environmental settings and MTI through the Persistent and Distributed Scatterers processing capability of the SPINUA algorithm. The first case regards a hilltop town in the Apennine Mts., whose stability is threatened by a large (~600 x 300 m2), slow moving deep landslide. We processed over 50 S-1A images acquired since October 2014. The comparison of the MTI results with those based on ERS and ENVISAT imagery shows that a much higher number of radar targets is obtained from S-1A data (e.g., from ~2 to 5 times higher, respectively on the landslide and in the overall area of interest, including also the town and peri-urban areas). With more targets, we can better depict the spatial movement pattern and local velocity gradients, which is important for geotechnical assessment. Furthermore, the lateral boundaries of the landslide can now be delimited in more detail, overcoming the mapping uncertainties typical in cases of very slow, deep landslides affecting urbanized areas. This offers invaluable information for local inhabitants/property owners and for engineering scale hazard assessment. Importantly, the MTI from S-1A data also revealed an accelerating trend with a nearly doubled velocity of the displacements with respect to those in the earlier period covered by ERS-ENVISAT data. The higher frequency of S-1A acquisitions (about 30/year in this case) helped highlighting the non-linearity of surface deformations within the faster displacement phase, whose timing was consistent with the increase in landslide movements detected through subsurface inclinometer monitoring and field observations. The latter demonstrated that this faster movement phase coincided with (or was preceded by) a failure of the landslide toe, which occurred in the inhabited area. The second case represents an example of a retrospective investigation of a huge (over 2 km long, few tens of m deep) landslide, which occurred in 2016 in an important open-cast coal mine in central Europe. The apparently sudden failure disrupted the mining operations, destroyed in part the mining machinery and resulted in high economic losses. In this case, we exploited over 60 S-1A/B images acquired since November 2015. Despite the presence of spatial gaps in information (due to intensive surface disturbance by mining operations), the MTI results provided a good overview of the ground instability/stability condition within and outside the active mine. Furthermore, it was shown that the 2016 slope failure was preceded by very slow (generally 1-3 cm/yr) creep-like deformations, already detectable in 2014. Although it would not have been simple to issue a short-term warning of the impeding failure based on the displacement time series, the MTI results showed that the slope had been in the critical instability state some months prior to the landslide event. Furthermore, the spatio-temporal mapping of coherence changes in the unstable area indicated a sharp coherence loss in the last few weeks before the slope collapse. The above examples demonstrate that by securing long-term, regular, high-frequency acquisitions over wide-areas, the Sentinel-1 mission facilitates a more effective use of MTI in slope hazard assessment. We note further improvement thanks to the availability of S-1B data (e.g., more frequent measurements to forewarn potential instability hazards). This has practical impacts on landslide monitoring activities and will aid future research on slope failure forecasting. Thanks to this and free imagery, the site-specific investigations relying on MTI will become even more feasible and cost-effective for non-scientific users (e.g., engineering geology/geotechnical consulting) and commercial services (e.g., Rheticus®).

Data from