Landslide Hazard Mapping with Geospatial Foundation Models: Geographical Generalizability, Data Scarcity, and Band Adaptability

2025-11-06

AI Summary

Landslide mapping suffers from poor generalization across regions, sensors, and scarce-label scenarios, hindering timely disaster response. To address this, we propose a three-axis analytical framework—sensor, label, and domain—and systematically adapt the geospatial foundation model Prithvi-EO-2.0 for landslide mapping. Our approach leverages global remote sensing data for self-supervised pretraining and introduces an adaptable fine-tuning strategy to support multispectral and heterogeneous sensor inputs. Compared with task-specific models (e.g., U-Net, SegFormer), our method achieves consistently high accuracy and robustness across diverse geographical regions and data-limited settings. Experiments demonstrate significant resilience to spectral band discrepancies, extreme label scarcity, and cross-domain transfer. This work establishes a new paradigm for scalable, generalizable intelligent landslide risk monitoring.

Technology Category

• Computer Vision: Remote Sensing / Geospatial AI
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Landslides cause severe damage to lives, infrastructure, and the environment, making accurate and timely mapping essential for disaster preparedness and response. However, conventional deep learning models often struggle when applied across different sensors, regions, or under conditions of limited training data. To address these challenges, we present a three-axis analytical framework of sensor, label, and domain for adapting geospatial foundation models (GeoFMs), focusing on Prithvi-EO-2.0 for landslide mapping. Through a series of experiments, we show that it consistently outperforms task-specific CNNs (U-Net, U-Net++), vision transformers (Segformer, SwinV2-B), and other GeoFMs (TerraMind, SatMAE). The model, built on global pretraining, self-supervision, and adaptable fine-tuning, proved resilient to spectral variation, maintained accuracy under label scarcity, and generalized more reliably across diverse datasets and geographic settings. Alongside these strengths, we also highlight remaining challenges such as computational cost and the limited availability of reusable AI-ready training data for landslide research. Overall, our study positions GeoFMs as a step toward more robust and scalable approaches for landslide risk reduction and environmental monitoring.

Problem

• Addressing landslide mapping challenges across different sensors and regions
• Overcoming limited training data for accurate disaster prediction models
• Improving model adaptability to spectral variations and geographic diversity

Innovation

• Geospatial foundation models adapt across sensors and domains
• Global pretraining with self-supervision enables robust generalization
• Fine-tuning maintains accuracy under label scarcity conditions