Search papers, labs, and topics across Lattice.
This study introduces GeoPhysAdapter, a method that enhances cross-domain landslide mapping by leveraging vision foundation models while addressing the challenges of uncertain geographic context. By integrating spatial guidance and event-timing forcing, the approach effectively reduces high-confidence false alarms in landslide detection, particularly in regions with limited annotations. The results demonstrate a significant improvement in error reduction鈥攗p to 23.99%鈥攚hen adapting at the candidate landslide body level compared to pixel-level adaptation, showcasing the method's potential for more accurate emergency response and risk assessment.
Cross-domain landslide mapping can achieve a 23.99% reduction in false positives by adapting at the candidate body level, revealing the limitations of pixel-level approaches.
Newly triggered landslides rarely carry immediate annotations, so cross-domain transferability determines the value of landslide mapping for emergency response and regional risk assessment. Vision foundation models have strengthened representational transfer, yet on unseen regions, events, and data sources they still generate high-confidence false alarms. Terrain, material, and rainfall triggering can constrain such errors, but their supports are local, regional, and event-scale, so that resampling onto a 10~m grid misaligns them with the segmentation decision unit and compounds the uncertain geographic context problem (UGCoP). We propose GeoPhysAdapter, which anchors on a frozen vision foundation model, restricts terrain, material, and triggering to dense spatial guidance, regional modulation, and event-timing forcing, and applies bounded adaptation at two decision units, the pixel and the candidate landslide body, reverting exactly to the visual prediction where support is insufficient. On an event-isolated PILD dataset of four public sources, 55 global landslide events, and 7,890 test samples, 70.3% of cross-domain false-positive mass lies in near-pure spurious bodies of median equivalent diameter 207m, matching coarse-prior support rather than the pixel. Pixel-level adaptation removes a net 507,817 erroneous pixels and reduces error by 7.76%, whereas raising the decision unit to the candidate body, under identical samples, anchor, and baseline, increases error reduction to 23.99%, approximately 3.1 times the pixel-level effect, improves IoU by 0.031 (14.2% relative), and corrects 9.92 pixels per pixel harmed. The data and code are publicly available at: https://github.com/Liu-Zhihang/geophysadapter.
