1,3,4- 1University of Silesia, Faculty of Natural Sciences, Institute of Earth Sciences, ul. Będzińska 60, 41-200 Sosnowiec, Poland (j.godziek23@gmail.com)
- 2Department of Forest Ecology, The Silva Tarouca Research Institute, Lidicka 25/27, 602 00 Brno, Czech Republic
- 3University of Silesia, International Environmental Doctoral School, ul. Będzińska 60, 41-200 Sosnowiec, Poland
- 4Institute of Earth Sciences, Vietnam Academy of Science and Technology, Hanoi, Vietnam
Multiple landslides triggered by heavy rain, associated with debris flows and flash floods are major geohazard in the mountainous areas of Northern Vietnam, resulting in lost of life and property. Mapping landslides immediately after their occurrence remains crucial for providing a better understanding of their causes , the course of their formation, and the influence they exert on both nature and human.
We analyzed the effects of several landslide events that occurred between 2020 and 2024 in Northern Vietnam. We aimed to develop a fully automated geospatially integrated software workflow for rapid and accurate mapping of landslide and debris flows in the subtropical zone. The method we applied was Change Vector Analysis (CVA), which is based on detecting changes betweeen two images (pre- and post-event) by emploing two metrics: magnitude, referring to the amount of change between pixels, and direction, describing the type of change. As input data, we used the Sentinel 2A optical imagery with a spatial resolution of 10 m. For each landslide event we analyzed a separate area, where its geomorphic effects were the most robust. As the exact dates of landslide events varied for each study area, we downloaded pre- and post-event image pairs for each area with different acquisition dates and low cloudiness (below 10%). Due to the mountainous terrain and the potentially disruptive influence of atmospheric correction, we decided to use L1C data. For validation, we used the landslide vectorization polygons. For each study area, we generated random points labeled as “landslide” or “no landslide” based on the landslide polygons. Then, we performed CVA parameter tuning for each area and selected the CVA variant most effective at landslide delineation. We integrated the entire workflow into R script. The results indicate that simple data analysis methods such as CVA can be efficient for landslide mapping. Despite the cloudiness limitation, optical Sentinel-2 data can be applied in the subtropical zone to map the landslides and debris flows.
The study has been supported by the Polish National Science Centre (project no 2023/49/B/ST10/02879).
How to cite: Godziek, J., Pawlik, Ł., and Hieu, T. T.: The Northern Vietnam landslide events mapped with Change Vector Analysis, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14135, https://doi.org/10.5194/egusphere-egu26-14135, 2026.
