Landslide Hazard Assessment Based on Sentinel1 in a Quarry in West Java, Indonesia

Disaster Background

On May 30, 2025, a major landslide occurred at the Gunung Kuda quarry on the edge of Cipanas village, Dukupuntang district, Cirebon administrative region, West Java, Indonesia. As of June 3, 2025, it had caused 21 deaths.

The landslide occurred at approximately 10:00 AM local time in the Gunung Kuda Hill C mining area, where workers were conducting mining operations. The collapsed earth and rocks buried dozens of workers, three excavators, and six trucks. Search and rescue operations were suspended on the evening of May 30th due to insufficient light and the risk of further landslides that occurred multiple times that night.

Figure 1. Google Earth image of the quarry landslide site, coordinates [-6.7754, 108.4022].

Figure 2. Drone footage of the landslide site at the quarry.

Causes of disaster

The Geological Department stated that the mining area is located in a high-risk area with a landslide probability exceeding 50%. Bambang Tirto Mulyono, Director of the West Java Department of Energy and Mineral Resources, pointed out that the main cause was improper mining methods, specifically digging from the bottom of a cliff, which weakened the soil structure. Since February 2025, repeated warnings from the Department of Energy and Mineral Resources and police cordons had been ignored by the mine management. Therefore, the West Java Provincial Government revoked the mining permit, which was scheduled to expire in October 2025, and permanently closed the site. Ono Surono, Vice Chairman of the West Java Provincial Council, called for a comprehensive assessment of mining activities in the province, emphasizing the health risks of water pollution caused by mining to nearby residents.

Besides the loss of life, the landslide has exacerbated the environmental damage in the area, contaminating water sources and threatening safety with the risk of further landslides. The local economy has also been impacted, as many residents rely on mining for their livelihoods. West Java Governor Dedi Mulyadi had warned of the dangers of the mining area and expressed regret over the previous lack of preventative measures.

Disaster assessment methods and procedures

To more intuitively assess the impact range and severity of landslide disasters, we developed a corresponding automated processing toolkit based on Sentinel-1 ground range detected (GRD) images before and after the landslide, using the Multivariate Alteration Detector (MAD), and identified and assessed landslide areas and risk areas.

The MAD algorithm is a statistical change detection method suitable for processing multi-temporal and multi-band remote sensing images. It features high robustness, orthogonality, and automation, enabling it to handle image data of different modalities while avoiding information redundancy. Furthermore, it automatically determines change thresholds through statistical analysis, reducing subjective intervention. It is applicable to complex geological and surface cover conditions in landslide areas. The specific process is as follows:

 I. Data Preprocessing  

To ensure the quality and analytical accuracy of Sentinel-1 GRD data, the following preprocessing workflow is adopted:

1. Polarization selection

Sentinel-1 data provides two polarization modes: VV and VH. Experimental analysis shows that VH polarization better reflects surface roughness and vegetation change characteristics in the target area, particularly excelling in distinguishing between material stockpiling areas, construction areas, and vegetation clearing areas. Therefore, this project primarily uses VH polarization data for change detection.

VV polarization data are used as an auxiliary tool for verification and supplementary analysis.

2. Noise Reduction Processing

SAR images are often affected by salt-and-pepper noise, which degrades image quality. This project uses a 3x3 window Lee filter to denoise VH polarimetric images, effectively suppressing noise while preserving ground feature edges and texture information.

The signal-to-noise ratio of the denoised image is significantly improved, making it suitable for subsequent change detection and feature extraction.

3. Radiometric correction and topographic correction

Radiometric correction is performed to eliminate the influence of sensors and the atmosphere, ensuring that the image grayscale values ​​reflect the true backscattering characteristics of the Earth's surface.

Topographic correction is performed using a digital elevation model (DEM) to eliminate the influence of topographic relief on the geometric distortion of SAR images and improve the geographic accuracy of the images.

4. Image registration

Accurate registration of the two images before and after the landslide was performed to ensure pixel-level alignment, providing a reliable basis for subsequent change detection.

II. Change Detection Methods

 Figure 3. Roadmap of change detection technology based on sentinel1  

Figure 4. Flowchart toolbox for change detection based on MAD algorithm

Disaster assessment results

Figure 5. Surface change detection results based on Sentinel-1

As shown in the image above, the landslide area is mainly located east of the quarry, running northeast to southwest, with a length of approximately 290 meters, a maximum width of about 80 meters, and an area of ​​approximately 13,744 square meters. The southernmost point of the landslide area is less than 100 meters from the main road, requiring close monitoring of subsequent rainfall and potential secondary disaster risks.

The risk areas are those where there are currently no landslides, but where surface cover has changed, mainly located in the north and southwest of the mine. The northern risk area is another mining area with a significant elevation difference, and like the landslide area, it is mined from the bottom, thus posing a higher risk. The southwestern area contains several water bodies, which are less than 200 meters from nearby farmland, requiring close monitoring of water pollution to prevent its spread to nearby farmland.

There are several quarries in the Cirebon area of ​​West Java province. It is recommended to conduct continuous surface deformation monitoring in similar high-risk areas to predict risks based on trends. Simultaneously, monitor forest cover in the mining area to assess the risk of landslides caused by increased vegetation destruction. Furthermore, conduct hyperspectral remote sensing-based monitoring of suspended matter and transparency in nearby rivers, lakes, and reservoirs to ensure the safety of water sources around the mining area.

Reference source:

  https://eos.org/thelandslideblog/gunung-kuda-1  

  https://unsia.ac.id/tragedi-longsor-gunung-kuda-cirebon-dan-pentingnya-menjaga-alam/  

  https://www.youtube.com/watch?v=iuNiXXcL0f4  

  https://www.youtube.com/watch?v=uNwfefzmabs