Fusion of Sentinel-1 and Sentinel-2 satellite imagery to rapidly detect landslides through Google Earth Engine

Landslides constitute a significant geohazard causing human losses and significantly affecting the economy worldwide. Earth Observation and the exploitation of the freely available Copernicus datasets, such as the Sentinel-1 and Sentinel-2 satellite images, can assist in the systematic monitoring of landslides overcoming the restrictions arising from in situ measurements. This study shows how the Google Earth Engine (GEE) platform can be utilised for the rapid mapping of landslides and effectively integrate both passive and active satellite data to enhance the results’ reliability. The GEE is a cloud computing platform designed to store and process huge datasets for scientific analysis and visualization of geospatial datasets where open-source images are acquired by several satellites. 

For this study, Ground Range Detection (GRD) Sentinel-1 and multispectral Sentinel-2 satellite data were utilised for a time period between 2016 and 2021. Multitemporal SAR change detection was conducted to identify potential landslides using GRD Sentinel-1 satellite images. Moreover, the Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Soil Moisture Index (SMI) and Bare Soil Index (BSI) indices were used for the multispectral data. Multi-temporal image composites were created for the two periods. Furthermore, for all image collections, the calculated spectral indices were added as new bands to all images, and the maximum value for each pixel of the vegetation indices was taken. Following, the difference image for each spectral index was created based on two methods, i.e., the first method was based on subtracting the two time periods, and the second one on subtracting each year from the total average for the time period from 2016 until 2021. The possible events were then masked using the thresholding technique based on the trial-and-error procedure where the analyst adjusts manually the thresholds and evaluates the resulting image until satisfied. Based on the results derived from the abovementioned processing, the use of the second method, i.e., subtracting each year from the average, based on the NDVI spectral index provides better results. The proposed methodology was tested in Paphos city in Cyprus because of the occurrence of numerous landslide events in this area, based on the landslide inventory provided by the Geological Survey Department of the Ministry of Agriculture, Rural Development and Environment. The results of this study were validated using high-resolution images from Google Earth in combination with the data from the Geological Survey department. 

Acknowledgements 

The authors acknowledge the 'EXCELSIOR': ERATOSTHENES: Excellence Research Centre for Earth Surveillance and Space-Based Monitoring of the Environment H2020 Widespread Teaming project (www.excelsior2020.eu). The 'EXCELSIOR' project has received funding from the European Union's Horizon 2020 research and innovation programme under Grant Agreement No 857510, from the Government of the Republic of Cyprus through the Directorate General for the European Programmes, Coordination and Development and the Cyprus University of Technology. The authors would also like to thank the Geological Survey Department of the Ministry of Agriculture, Rural Development and Environment for the provision of the landslide inventory.