Deutsch
 
Datenschutzhinweis Impressum
  DetailsucheBrowse

Datensatz

DATENSATZ AKTIONENEXPORT

Freigegeben

Poster

Improving landslide mapping by multi-temporal SAR data analysis and Deep Learning

Urheber*innen
/persons/resource/wandi

Wang,  Wandi
1.4 Remote Sensing, 1.0 Geodesy, Departments, GFZ Publication Database, Deutsches GeoForschungsZentrum;

/persons/resource/motagh

Motagh,  M.
1.4 Remote Sensing, 1.0 Geodesy, Departments, GFZ Publication Database, Deutsches GeoForschungsZentrum;

/persons/resource/magda88

Stefanova Vassileva,  M.
1.4 Remote Sensing, 1.0 Geodesy, Departments, GFZ Publication Database, Deutsches GeoForschungsZentrum;

Externe Ressourcen
Es sind keine externen Ressourcen hinterlegt
Volltexte (frei zugänglich)
Es sind keine frei zugänglichen Volltexte in GFZpublic verfügbar
Ergänzendes Material (frei zugänglich)
Es sind keine frei zugänglichen Ergänzenden Materialien verfügbar
Zitation

Wang, W., Motagh, M., Stefanova Vassileva, M. (2022): Improving landslide mapping by multi-temporal SAR data analysis and Deep Learning. - Poster presented at the Living Planet Symposium 2022 (Bonn, Germany 2022)


Zitierlink: https://gfzpublic.gfz-potsdam.de/pubman/item/item_5014168
Zusammenfassung
Landslides are a major type of natural hazard that cause serious economic losses, casualties, and damages to buildings, and critical infrastructures in mountainous regions around the world. The advent of satellite remote sensing brought about a revolution in the field of landslide investigation. Both optical and Synthetic Aperture Radar (SAR) satellite data are increasingly being used in detecting, monitoring, and assessing landslide hazards. The optical images reflect the abundant spectral information and geometric shape of ground objects, which can be used for surface change detection in landslide areas. However, methods using optical imagery cannot reliably support near real-time landslide change detection because cloud-free images covering landslide areas may not be readily available before and during a given landslide event. Active measurements from SAR systems offer new opportunities to support systematic mapping and monitoring of landslides over extensive regions independent of weather and sunlight conditions. Variation in SAR amplitude and coherence allows to detect structural and surface changes related to landslides. Moreover, interferometric measurements using the InSAR technique can be used to detect pre-failure and post-failure motions, which are key for hazard and risk assessment in landslide prone regions. The availability of free and open data from Earth observation programmers such as Copernicus has further improved our capability in using satellite observations for multi-scale analysis of landslide occurrence and evolution. In this study, we propose a methodology for identifying landslides that occur in vegetated regions using dual-pol Sentinel-1 SAR data and machine learning. Both amplitude and phase information from Sentinel-1 SAR data are used as input parameters for deep learning models to detect landslide changes and segment them automatically. Several examples of big landslides in China, Kyrgyzstan, and Iran are presented for which high quality external datasets and our own field observations allow detailed ground truthing needed for validation of the results.