Multimodal AI fusion for infrastructure resilience: real-time urban analytics framework aligned with SDG-9

Area of Science:

• Artificial Intelligence
• Urban Planning
• Disaster Risk Reduction

Background:

• Urban flood risk management faces challenges due to limited human capacity, technical support, and planning integration.
• Existing models often fail to capture complex temporal and spatial urban data dynamics.

Purpose of the Study:

• To propose a multimodal AI fusion framework for advanced urban flood risk assessment.
• To develop a dynamic Resilience Scoring Index (RSI) for real-time anomaly detection and decision support.
• To evaluate the generalizability of the proposed framework across diverse urban data environments.

Main Methods:

• A hybrid AI model combining Long-Short Term Memory (LSTM) and Graph Neural Networks (GNN) to analyze temporal and spatial urban data.
• Integration of Edge-AI for immediate sensor data processing and decision dashboards for city insights.
• Evaluation across three distinct cities: Singapore, Chennai, and Rotterdam.

Main Results:

• The LSTM+GNN model demonstrated superior performance compared to ARIMA, Random Forest, and unimodal deep networks.
• A statistically significant improvement in F1 score (p < 0.05) was observed.
• The model showed robust performance with only minor degradation under noisy and incomplete data conditions.

Conclusions:

• The proposed AI framework offers a scalable, evidence-based solution for urban infrastructure planning and disaster risk reduction.
• This work supports Sustainable Development Goal 9 (SDG-9) by enhancing smart city resilience.
• The framework provides a replicable model for global smart city resilience initiatives.