A Weighted Cellular Automata for Flood Modelling

Abstract

This study introduces a novel flood simulation framework, termed WeightedCA, which extends the CA2D model by incorporating weight-based transition rules for simulating flooding over surface terrains. The model is quasi-physical, leveraging simple shallow-water equations and a weighting system to expedite simulation processes. Through the application of this model to the Kerala state in India, known for its heavy monsoon rainfall and predominantly sea-level terrain, we demonstrate the model's capability in accurately predicting flood patterns. The WeightedCA model utilizes a Digital Elevation Model (DEM) as input and processes it through five distinct layers, including ground level, water columns, slope field, flow directions, and previous intercellular volume, to simulate water flow and accumulation dynamically. Transition rules are defined to update water heights in each cell, considering the mass conservation principle and flow directions based on the least resistance path. Additionally, the study explores flow accumulation matrices as a heuristic for large-scale water flow predictions, alongside metrics such as total mass, average flow rate, and fraction of cells flooded to evaluate the model's performance. A rainfall model simulating 30 days of Poisson-distributed rainfall patterns further complements the flood simulation. The case study of Kerala, with its complex topography and monsoon climate, provides a challenging yet insightful application for the WeightedCA model, showcasing its potential in flood risk management and planning. The results indicate the model's effectiveness in identifying flood-prone areas and evaluating flood mitigation strategies, albeit with limitations in resolution and predictive precision for large-scale applications. Future work will focus on refining the model through smaller catchment simulations and advanced rainfall models for improved accuracy in flood prediction and management strategies.