In the last decades, the availability of multi-source, multi-scale, and multi-resolution remote sensing data and the consequent progress of processing techniques brought a significant positive impact for landslide detection. As a result, nowadays also public institutions dealing with geo-hazard management worldwide regularly use satellite data and products in landslide investigations. Due to the complexity of the phenomenon which might involve the displacement of massive rocks, soil, and both wet and dry vegetation from hillslopes, and the significant impact on the safety of the population and road infrastructure, the development of specific procedures for the rapid detection of landslides is extremely challenging. This is particularly important in the first phases of landslide risk management to evaluate the extension of the area involved by the movement and proceed with an initial delimitation of the so-called alarm zone, where preventive evacuation must be applied. In this study, the wide-ranging Tasseled Cap Transformation (TCT) is proposed as a not-sophisticated and rapid method able to detect different land-change features simultaneously by processing only two (pre- and post-event) Sentinel-2 images. The RGB color composite image obtained by stacking δTCTBrightness, δTCTGreenness, and δTCTWetness, as the intensity of red (R), green (G) and blue (B) was able to provide information on landscape changes supported by the physical nature of the TCT indices, by associating the colors to the physical characteristics of the changes. The method tested over the Pomarico site in Basilicata region (Southern Italy), was able to define the footprint of a landslide occurred in 2019 with a good accuracy (ACC = 0.95). Therefore, the proposed procedure is effective to detect landslides involving land-surface spectral changes (the majority), without the need of additional in-situ information. Furthermore, the free availability of the Sentinel-2 database and its frequent revisit times guarantee its global exportability.
Read full abstract