CLIMATE-RESILIENT CIVIL INFRASTRUCTURE: AN INTEGRATED DATA-DRIVEN FRAMEWORK FOR STRUCTURAL RISK ASSESSMENT AND SUSTAINABLE URBAN DEVELOPMENT

Abstract

Climate change intensifies flooding, extreme heat, and rainfall variability, creating interconnected challenges for infrastructure safety and sustainable urban development. This study develops an integrated framework combining infrastructure condition assessment, climate hazard mapping, machine learning, and spatial planning to identify vulnerable assets and prioritize resilience investments. The dataset comprises 12,000 inspection records representing 1,200 infrastructure assets, including 600 buildings, 350 road segments, and 250 bridges, monitored between 2019 and 2025. A total of 36 variables capture material deterioration, structural age, maintenance history, drainage capacity, surface temperature, rainfall intensity, flood exposure, population density, and socioeconomic vulnerability. Data preparation involves missing value treatment, outlier screening, appropriate feature standardization, and geographic harmonization within a geographic information system. Structural risk categories are established using inspection criteria and engineering assessments. Earlier observations support model training, while later observations from geographically separated assets provide validation and testing to reduce temporal and spatial information leakage. Random Forest and Extreme Gradient Boosting models classify infrastructure risk, and SHAP analysis explains the contributions of individual predictors. A multicriteria decision module integrates predicted risk, service criticality, intervention cost, and environmental impact to prioritize adaptation measures. Results show that Extreme Gradient Boosting achieves 92.4% classification accuracy, 90.8% macro precision, 89.7% macro recall, and 90.2% macro F1, outperforming Random Forest, which achieves 87.6% accuracy. Flood exposure, deterioration severity, drainage inadequacy, and maintenance delay emerge as influential risk predictors. Spatial assessment identifies 21.5% of infrastructure assets as high risk, with vulnerability concentrated in densely populated floodplain areas. Under a constrained investment budget, scenario analysis indicates that targeted drainage upgrades, structural rehabilitation, and reflective surface treatments could reduce the composite risk index by 24.8% and estimated annual disruption by 18.6% relative to baseline conditions. These findings demonstrate the value of integrating interpretable predictive modelling with engineering assessment and spatial investment planning. The framework supports transparent prioritization of infrastructure interventions while connecting structural safety with environmental sustainability and service continuity. Further validation across different urban settings and climate scenarios can strengthen its transferability and support equitable, evidence-based adaptation planning for resilient civil infrastructure under evolving climatic pressures.