Quantitative evaluation of vulnerability and resilience in Indian meteorological hazards using correlation, clustering, PCA, and random forest

Meteorological disasters in India impose multidimensional impacts that span human casualties and fiscal recovery gaps. This study applied an integrated framework combining correlation analysis, clustering, principal component analysis (PCA), composite indices, and Random Forest regression. Pearson’s correlation revealed strong positive associations between direct losses and recovery needs (r = 0.89) and weaker correlations between fatalities and funding gaps (r = − 0.21), highlighting a disconnect between human and financial impacts. K-means and fuzzy c-means clustering grouped events into three clusters, with Kerala floods and Himachal floods/landslides forming a high-vulnerability cluster. PCA reduced the dataset to two principal components, explaining 78% of total variance, with PC1 dominated by economic losses and PC2 by fatalities. The Vulnerability Index identified Kerala floods (VI = 1.66) and Himachal floods/landslides (VI = 1.67) as most vulnerable, while Cyclone Amphan (VI = 0.63) and Rajasthan drought (VI = 0.65) were least vulnerable. Resilience Scores ranged from 0.04 (Cyclone Michaung & floods) to 0.12 (Himachal floods/landslides), indicating significant variation in fiscal adequacy. Random Forest analysis showed direct losses (importance = 0.41) and indirect losses (importance = 0.33) as the strongest predictors of funding gaps, while insured losses contributed minimally (importance = 0.07). This study advances disaster risk science by quantifying the systemic disconnect between human and fiscal impacts. Given the small sample size (five events), findings should be interpreted as exploratory rather than definitive.