A GIS based spatially-explicit sensitivity and uncertainty analysis approach for multi-criteria decision analysis

Area of Science:

• Geosciences
• Geographic Information Systems (GIS)
• Environmental Risk Assessment

Background:

• Geographic Information Systems (GIS) multicriteria decision analysis (MCDA) is crucial for landslide susceptibility mapping and hazard preparedness.
• Uncertainties are inherent in MCDA techniques, impacting the reliability of landslide susceptibility maps.
• Assessing and quantifying these uncertainties is vital for improving predictive models.

Purpose of the Study:

• To systematically analyze and quantify uncertainties in landslide susceptibility maps produced by GIS-MCDA techniques.
• To compare the performance and uncertainty levels of Analytical Hierarchical Process (AHP) and Ordered Weighted Averaging (OWA) methods.
• To develop and apply a spatially-explicit approach integrating Dempster-Shafer Theory (DST) for uncertainty assessment.

Main Methods:

• Weights for landslide susceptibility factors were computed using MCDA techniques (AHP and OWA) within a GIS environment.
• Uncertainty and sensitivity analyses were performed using Monte Carlo Simulation and Global Sensitivity Analysis on the computed weights.
• Dempster-Shafer Theory (DST) and landslide inventory data were used for validation and comparison of the generated susceptibility maps.

Main Results:

• Analytical Hierarchical Process (AHP) demonstrated superior performance in predicting landslide susceptibility compared to Ordered Weighted Averaging (OWA).
• Ordered Weighted Averaging (OWA) produced landslide susceptibility maps with lower uncertainty than those generated by AHP.
• The integrated uncertainty-sensitivity analysis approach effectively decomposed and attributed uncertainty to model criteria weights, enhancing model accuracy.

Conclusions:

• Integrated uncertainty and sensitivity analysis is essential for improving the accuracy of GIS-based MCDA landslide susceptibility models.
• Understanding the sources of uncertainty, particularly from criteria weights, is key to refining hazard prediction and land-use planning.
• The study highlights the trade-offs between accuracy and uncertainty when selecting MCDA techniques for landslide susceptibility mapping.