Assessing disaster resilience in mountain villages using an improved DPSIR-A framework and multi-model machine...

Area of Science:

• Environmental Science
• Geosciences
• Disaster Management

Background:

• Mountainous regions face escalating climate change and geological hazards.
• Assessing disaster resilience is vital for sustainable development in these areas.
• High-altitude, tourism-dependent ethnic villages require specialized resilience frameworks.

Purpose of the Study:

• To propose an integrated resilience assessment framework for high-altitude ethnic villages.
• To incorporate an Adaptability dimension, including traditional ecological knowledge and community learning.
• To address small-sample geological hazard data limitations using advanced machine learning.

Main Methods:

• Extended the DPSIR model with an Adaptability (A) dimension.
• Integrated SMOTE oversampling with a multi-model evaluation chain (IVM-SVM-RF).
• Employed SMOTE-enhanced Random Forest (S-RF) for improved small-sample generalization.

Main Results:

• The S-RF model achieved superior performance (AUC=0.753, Kappa=0.754), excelling in identifying low-resilience zones.
• Identified spatial resilience differentiation in Zhangzha Town, with high-resilience cores and scattered low-resilience peripheries.
• Determined key drivers including Fractional Vegetation Cover, Rainfall, Distance from Roads, Building Disaster Resistance, and Villagers' Disaster Awareness.

Conclusions:

• The proposed framework effectively assesses disaster resilience in small-sample, high-altitude village contexts.
• SMOTE augmentation significantly improves model accuracy for marginal low-resilience zones.
• Both physical and 'soft' resilience factors are critical for mitigating risks in mountainous communities.