<![CDATA[Landslides are a recurrent hazard in Bogor Regency, where steep volcanic terrain, high rainfall, varied lithology, land-use changes and active faults contribute to slope instability. This study presents the first regency-wide landslide susceptibility model using Logistic Regression supported by field validation. A dataset of 220 landslide occurrences from 2017 to 2022 and multiple geospatial factors including rainfall, slope, lithology, landcover, and NDVI was analyzed using a 70:30 train–test split to generate coefficient weights, probability surfaces and a binary susceptibility map derived from ROC-AUC thresholds. Landcover shows the strongest positive influence on landslide occurrence, whereas NDVI has the strongest negative effect, reflecting the stabilizing role of vegetation. Fault proximity exhibits near-zero influence, likely due to inactive structures or limited spatial resolution. The model achieved 82 percent accuracy with an AUC of 0.86. Susceptibility clustering near historical data suggests possible inventory bias. Improving model reliability will require more evenly distributed landslide data and UAV-based mapping to detect vegetation-covered past landslides.]]>
Copyrights © 2024
