From data to action in flood forecasting leveraging graph neural networks and digital twin visualization

Area of Science:

• Hydrology and Water Resource Management
• Artificial Intelligence in Environmental Science
• Geospatial Analysis and Disaster Management

Background:

• Hydrological systems exhibit complex nonlinear dynamics, making accurate flood forecasting challenging.
• Existing flood prediction methods often neglect spatial interconnections and lack robust visualization tools for water level assessment.
• Effective flood management requires advanced predictive models and decision-support systems.

Purpose of the Study:

• To develop an advanced graph neural network model (LocalFLoodNet) for enhanced water level prediction.
• To create a simulation prototype for visualizing flood impacts and supporting disaster prevention and policy-making.
• To provide a comprehensive framework for assessing flood impacts and evaluating preventive strategies in urban areas.

Main Methods:

• Implementation of a graph neural network (LocalFLoodNet) with a graph learning module to model hydrological network connectivity.
• Development of an interactive simulation prototype utilizing a digital twin for scenario analysis and visualization of predicted water levels.
• Application and validation of the model and prototype in the Greater Montreal Area (GMA), specifically Terrebonne, Quebec.

Main Results:

• LocalFLoodNet effectively captures interconnections within water systems for improved water level prediction.
• The simulation prototype provides valuable visual insights for decision-making in flood risk management.
• The study demonstrates a comprehensive approach to flood impact assessment and the evaluation of preventive measures.

Conclusions:

• The integration of graph neural networks and simulation prototypes offers a significant advancement in flood forecasting and water management.
• The developed tools enhance the ability to predict water levels and assess the effectiveness of flood mitigation strategies.
• This research establishes a benchmark for applying advanced AI and digital twin technologies to hydrological challenges in diverse geographic regions.