At CLIMATIG, we continuously refine our climate intelligence platform to provide asset owners with comprehensive and actionable insights into physical climate risks. Today, we are excited to announce the addition of landslide hazard, a significant enhancement to our suite of climate risk assessments.
Why landslides matter for climate risk assessment
Landslides pose severe threats to infrastructure, real estate, and human safety, particularly in areas with steep terrain, heavy precipitation, and unstable geological conditions. With climate change increasing the frequency and intensity of extreme precipitation events, the risk of landslides is expected to rise, making it essential for asset owners to assess and mitigate this hazard.
How we built the CLIMATIG landslide model
Developing an accurate and scalable landslide susceptibility model required a machine-learning-driven approach that integrates diverse geospatial and climate data. Our methodology involves:
- Geospatial feature analysis: We processed high-resolution digital elevation models (DEM) to derive key terrain attributes such as slope, aspect, curvature, and elevation.
- Land cover, soil properties, and hydrology: Using remote sensing and soil databases, we incorporated information on vegetation cover, soil properties, and proximity to hydrological networks.
- Climate and precipitation triggers: To assess time-dependent landslide susceptibility component, we analyzed historical precipitation records and extreme precipitation indices, correlating them with observed landslide occurrences.
- Machine learning classification: We trained multiple models to classify landslide-prone areas using real-world landslide inventories across Europe. Among the trained classifiers, Random Forest (RF) was selected as the best one. Using 70/30% training/test split, we achieved an accuracy score of 0.78 and AUC (area under ROC curve) of 0.86 for Landslide Susceptibility (LS) model, while for the Precipitation-Triggered Landslide Hazard (PTLH) model, we used 80/20% training/test split and obtained accuracy score of 0.88 and AUC of 0.87.
- Probability-based risk scoring: The final landslide risk score combines both prediction probabilities of time-independent LS model and time-independent PTLH model into a normalized risk score.
What this means for CLIMATIG users
With this new hazard module, users can now:
- Assess landslide susceptibility at a high resolution for any asset location and throughout the whole period 1991-2100 for two climate scenarios (RCP 4.5 and RCP 8.5).
- Incorporate precipitation-driven landslide risk into climate adaptation planning.
- Generate reports with clear, science-backed risk scores to inform decision-making.
The CLIMATIG landslide risk score is now live on the CLIMATIG platform, accessible via our web application and API. This addition reinforces our mission to provide data-driven climate intelligence that helps businesses, insurers, and policymakers build resilience against evolving climate risks.
Want to learn more?
Explore the CLIMATIG platform or get in touch with our team to see how our new landslide risk assessment can support your climate adaptation strategy.