CAIG: Unifying Landslide Domain Knowledge and AI to Understand Landslide Causality

Landslides are a significant natural hazard that recurrently threaten lives and infrastructure worldwide, with their frequency and impact intensifying due to climate change. The complexity of landslide triggers, marked by nonlinear processes across different scales and times, along with the scarcity and limitations of available data, poses critical challenges in understanding landslide causality.

Existing approaches, whether physics-based or data-driven, only partially capture landslide mechanisms due to the difficulty of navigating multidisciplinary process drivers. This award supports research to understand landslide causality by combining advanced artificial intelligence (AI) and machine learning (ML) technologies with established geoscientific domain knowledge.

This research will enhance landslide situational awareness and preparedness for communities and guide informed decision-making among stakeholders. The research will also inform broader efforts to integrate AI/ML approaches into earth surface modeling, such as floods and soil erosion. Research outcomes will feed into educational experiences for K-12, undergraduate, and graduate students from diverse backgrounds.

This project will pursue four research thrusts: (1) establishing and annotating datasets and benchmarks for landslide causality analysis; (2) discovering new technologies for building AI-physics hybrid models by unifying landslide domain knowledge with AI models to explore hydro-eco-geomorphic synergy; (3) explaining AI-physics hybrid models and analyzing key factors that contribute to landslide causality; and (4) evaluating the effectiveness of the proposed algorithms across scales, times, and ecoregions. By leveraging the expertise of the investigators in geosciences, geotechnical engineering, machine learning, and computer vision, this research will lead to new advances in scientific understanding of landslide mechanisms, enhancing the integration of geoscience insights with AI and ML models, and developing effective strategies for overcoming the technical challenges associated with applying AI models to handle scarce, noisy, and limited geoscience data.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.