Climate Vulnerability Assessment

Overview

This project aims to systematically assess climate vulnerability at the block level across India by developing a composite index based on three key dimensions: Exposure, Sensitivity, and Adaptive Capacity. The methodology integrates data from multiple government and public repositories, processes it through robust statistical transformations, and applies Principal Component Analysis (PCA) to derive vulnerability sub-indices for different climate hazards, including floods, droughts, cyclones, and earthquakes.

Data Sources

The dataset combines publicly available records from:

• India Data Portal (IDP)
• ISB-provided block-level datasets
• Government sources such as census data, agricultural surveys, and institutional capacity reports

Methodology

1. Data Preprocessing

• Ratio Transformations: Many raw variables (e.g., bank branches, households, crop areas) were converted into ratios to enable meaningful comparisons, such as:
  • bank_branches_per_10k_population = total_branches / (total_population / 10000)
  • irrigated_fraction = irrigated_area / total_geographic_area
• Imputation: Missing values were handled through a combination of government-sourced data and district-wise median imputation to ensure consistency.
• Index Creation: We designed new composite indices, such as the Crop Vulnerability Index, which integrates rainfed agriculture and crop diversity to assess agricultural risk exposure.
• Variable Inversion: Certain variables (e.g., travel time to major cities) were inverted so that all indicators align with a "higher is better" interpretation for easier downstream analysis.
• Outlier Handling: Extreme values in counts (e.g., total irrigated land, number of bank branches) were identified and scaled using a Robust Scaler, which normalizes data based on the median and interquartile range (IQR), reducing sensitivity to outliers.

2. PCA-Based Sub-Index Computation

• Principal Component Analysis (PCA) was applied separately to each climate hazard category (Floods, Droughts, Cyclones, Earthquakes) to extract meaningful patterns and reduce dimensionality.
• Weighting Mechanism: The sub-indices were constructed by multiplying principal components with their explained variance percentages and summing them to produce a composite score.

3. Final Vulnerability Score

• The final Climate Vulnerability Index (CVI) for each hazard is computed as:

  Vulnerability = (Exposure * Sensitivity) / Adaptive Capacity

• This allows policymakers and researchers to identify high-risk blocks that require targeted interventions.

Future Improvements

• Integration of Real-Time Data: Incorporating satellite-based rainfall and soil moisture data for dynamic vulnerability assessment.
• More Hazard Categories: Extending the index to include heatwaves, landslides, and other climate-driven risks.
• Machine Learning Approaches: Exploring supervised models for validation against historical disaster impacts.

License

This project is licensed under the MIT License. See LICENSE for details.