Storm-Based Geospatial Data for Predicting Flood Risk in Chennai Using the Enhanced Grid Cellular Automata Algorithm (EGC2A)

Abstract

Background: Urban flooding in coastal cities like Chennai has intensified due to climate change and increased cyclonic activity in the Bay of Bengal. The 2015 Chennai floods demonstrated the critical need for accurate prediction systems to mitigate disaster impacts. Current models face challenges in balancing computational efficiency, spatial accuracy, and interpretability for urban planning applications.

Methods: This study develops an Enhanced Grid Cellular Automata Algorithm (EGC2A) that integrates mean meteorological parameters (rainfall: 16.11–17.72 mm, wind speed: 4.94–6.31 kmph) with elevation-aware flood propagation rules. The framework processes a  grid (54 km ) of Chennai, grouping cells into 20 neighborhoods through geospatial proximity analysis. Validation employs historical flood data and comparison against CNN-based approaches, with visualization through Folium maps and TikZ diagrams.

Results: The model achieves 89.2% accuracy (F1-score: 0.87) in flood prediction, identifying high-risk zones including T. Nagar (65.2% affected) and Velachery (58.7%). Quantitative analysis reveals 65% of Chennai falls under "Fully Affected" classification within 5 km of coastline. The system reduces false positives by 22% compared to deep learning benchmarks while maintaining computational efficiency (<2 sec/simulation cycle).

Concluding Remarks: EGC2A provides a robust, interpretable solution for urban flood prediction, validated through Monte Carlo simulations ( 5% error margin). The framework’s geospatial outputs enable actionable insights for disaster management, with potential applications in other coastal cities. Future work will integrate IoT sensors and hybrid machine learning to enhance real-time forecasting capabilities.