SafeLand FP7 Deliverable Archive – Hazard Modelling Science
Deliverable D1.5 of the SafeLand FP7 project evaluates statistical and empirical approaches for predicting precipitation-induced landslides in Europe.
The study was coordinated with contributions from the International Institute for Geohazards (ICG) and other European research partners.
Main Objective
The primary objective of D1.5 is to identify rainfall thresholds associated with landslide triggering and to evaluate models suitable for early warning system design.
The deliverable supports hazard prediction at local and regional scales by analysing meteorological and hydrological variables related to slope failure events.
Rainfall-Based Landslide Triggering Models
The document examines several modelling approaches for precipitation-driven landslide prediction.
- Intensity–Duration (I–D) threshold models
- Antecedent precipitation accumulation models
- I–A–D combined hydrological models
- Hydrological proxy systems such as FLaIR
- Machine learning approaches including neural networks
Short-duration high-intensity storms are typically associated with debris flows, while longer precipitation accumulation periods are more relevant for soil slope failures.
Geological and Environmental Influencing Factors
Lithology and Soil Properties
- Permeability differences between coarse and fine materials
- Soil depth and structural composition
- Drainage behaviour of geological layers
Low permeability soils may require longer antecedent precipitation periods to reach critical pore pressure levels.
Hydrological Conditions
- Groundwater level variation
- Antecedent moisture accumulation (typically 1–46 days)
- Snowmelt runoff contribution in alpine regions
- Freeze–thaw cycle effects on rock stability
Practical Application in Early Warning Systems
The deliverable proposes multi-stage warning frameworks using probability thresholds.
- Green – Normal conditions
- Yellow – Increased vigilance
- Orange – High risk monitoring
- Red – Emergency response activation
Threshold calibration is important to balance false alarms and missed events.
Performance evaluation commonly uses ROC and AUC statistical metrics.
European Case Study Data
Model testing was conducted using datasets from several European regions including Switzerland, France, Italy, and Norway.
Examples include:
- La Frasse – ARX and neural predictive models
- Barcelonnette – Antecedent precipitation optimization
- Norwegian coastal and inland debris flow regions
- Italian landslide-prone volcanic and mountainous zones
Technical Keywords
- Intensity–duration rainfall thresholds
- Antecedent precipitation analysis
- I–A–D hydrological models
- FLaIR system
- ROC performance evaluation
- Early warning probability thresholds
- Snowmelt integration
- Soil moisture proxy modelling
- Neural ARX prediction models
- Debris flow triggering dynamics
Role Within SafeLand Research Framework
Deliverable D1.5 contributes to Work Area 1 hazard science development by improving understanding of precipitation-triggered slope instability processes.
The modelling approaches discussed are relevant for climate adaptation planning, hazard zoning, and early warning system calibration.
Archive Integration Notes
- Provide PDF download or embed viewer where copyright permits.
- Add metadata tags: SafeLand FP7, D1.5, 2012, precipitation landslide modelling.
- Link internally to hazard mechanism pages and monitoring clusters.
- Ensure accessibility compliance for figures and tables.
Precipitation-induced landslides are controlled by the interaction between meteorological forcing, soil hydraulic behaviour, and geological structure. Statistical and empirical models support risk estimation but are not deterministic prediction tools.