Detecting and monitoring the Shallow Landslides in Agricultural Environments using Google Earth and Sentinel-2 images: two case studies

The detection and monitoring of shallow surface landslides in agricultural environments using remote sensing imagery present several critical challenges. These landslides are often very small, resulting in a limited number of pixels representing the landslide body. Moreover, their occurrence is frequently intermittent, as seasonal rainfall may trigger slope failures that are subsequently altered or erased by agricultural practices such as plowing, making multi-temporal analysis complex. In addition, because shallow landslides involve only a thin layer of soil, their spectral characteristics are often very similar to those of the surrounding terrain, further complicating their identification.

To overcome these limitations, a methodology originally developed for the detection of buried archaeological remains was adopted, as both applications face comparable detection constraints. The approach is based on the quantitative analysis of “tonal” differences between pixels corresponding to landslide-affected areas and those of the surrounding stable terrain. Several image-processing products were generated to enhance and measure these subtle spectral and tonal variations. This quantitative framework plays a key role in reducing subjectivity related to the experience of photo-interpreters and in limiting uncertainties associated with image processing and interpretation.

The spatial resolution of high-resolution imagery and Sentinel-2 data allowed the testing and validation of the proposed methodology, while the high temporal resolution of Sentinel-2 imagery enabled its application for monitoring shallow landslides over time. The integration of multi-temporal satellite data made it possible to observe changes related to landslide occurrence and surface modifications in agricultural landscapes.

Overall, the combined use of multiple image-processing products and the quantitative assessment of tonal differences proved effective in distinguishing areas affected by shallow landslides from stable surrounding areas. The results highlight the potential of this approach as a reliable tool for the detection and monitoring of shallow surface landslides in agricultural environments.