Flood Prediction & Disaster Response Simulation

Turning Environmental Research into an Interactive XR Simulation

Project Overview

This project explored how immersive technology, environmental simulation, and AI could be combined to create a practical platform for understanding and preparing for flooding.

Funded by SingularityNET, the project began as an investigation into how flood behaviour, mitigation infrastructure, and environmental data could be represented inside an interactive 3D environment. The objective was not simply to visualize a flood, but to create a flexible simulation in which users could examine the causes and consequences of flooding, investigate mitigation strategies, and explore how predictive models could eventually inform the experience.

This project demonstrates our ability to take a complex scientific and operational problem, research its underlying systems, construct an interactive simulation, integrate external modelling technologies, add a virtual population and conversational avatars, and turn the result into an approachable XR experience.

Flood Prediiction walkthrough:
https://www.youtube.com/live/lg0UEGlDQrY?si=cvAvRxi5IXiGiMQn

1. Defining the Simulation Environment

The initial concept called for a riverside suburban environment representing a real-world community exposed to flooding.

A fully generated city proved unnecessarily complex for the research objectives and difficult to integrate with models and assets from other sources. We therefore narrowed the scope and designed a more focused, editable test environment in which individual flood-management systems could be clearly demonstrated.

Using procedural terrain and environmental tools, we constructed a suburban setting containing:

• Residential buildings

• Roads and pathways

• Vegetation

• A riverside floodplain

• Drainage infrastructure

• Flood-control structures

• Commercial infrastructure

• Underground utility and water-management systems

• Non-interactive avatars to populate the location.

• Conversational LLM-driven avatars providing emergent training scenarios.

This provided a controlled environment in which individual components could be modified and tested.

2. Creating the Terrain and Floodplain

The terrain was developed specifically for the simulation rather than treated as a background landscape.

We shaped the land surrounding the river to create a low-lying floodable area and introduced different elevations and surfaces that would influence the movement and accumulation of water.

A river asset was incorporated primarily as an environmental reference, while additional fluid emitters were introduced to create the actual flooding behaviour within the test environment. Following that year's flood events as a guide, we simulated the breakage of a natural dam and subsequent flooding downhill to the suburban environment.

To depict more generalized flooding over a wide area, we added simple mechanisms of water height and bouyant objects.

This distinction allowed us to experiment with water movement independently from the visual representation of the river itself.

3. Modelling Flood-Mitigation Infrastructure

A major objective was to demonstrate how engineered and natural infrastructure can affect flood behaviour.

The environment incorporated examples of:

• Berms and levee structures

• Vegetated areas

• Barriers

• Groynes

• Detention structures

• Drainage channels

• Underground piping

• Water-treatment pathways

The structures were intentionally made visible and accessible so that users could understand their function.

For example, a groyne was introduced to demonstrate how structures extending into a river can influence flow. A detention structure demonstrated how runoff can be temporarily contained and controlled before entering a larger drainage network.

Rather than presenting these systems as diagrams, the simulation allowed them to exist as physical objects within the environment.

Additionally, we added hidden barrier objects to simulate slowing of water over a cultivate area of riverside featuring rocks, bushes and grass, compared to water movement over a flat area of grass.

4. Making Hidden Infrastructure Visible

Some of the most important systems in flood management are normally underground.

To make these systems understandable within an XR environment, sections of the residential and commercial infrastructure were constructed below the surface and exposed for examination as a large underground observation area.

One example demonstrated separation between wastewater and floodwater systems. A partial barrier could redirect lighter rainwater guided into river or ocean-facing piping from heavier wastewater that's guided through an alternative pipe leading toward a treatment facility.

A separate drainage and detention example showed how water entering a private or commercial drainage system could be collected in a containment structure before being released into the public drainage network.

These structures were presented almost as an infrastructure laboratory: users could move through the environment and examine systems that would normally be inaccessible.

5. Building an Interactive User Experience

The simulation was designed as an experience rather than a passive visualization.

We developed navigation and teleportation functionality allowing users to move between important locations in the environment. Labels were added to identify flood-control structures and explain their purpose.

A controllable orbit-style camera was refined to support a user's navigation and inspection of the environment as a familiar game-style environment.

The result allowed users to move from the river to the residential area, examine mitigation structures, inspect underground systems, and observe the effects of simulated water movement from different positions.

This interaction was particularly important to the research because it allowed the simulation to function as both a demonstration environment and a potential training platform.

6. Testing Fluid Behaviour

Fluid simulation was introduced to test how water could interact with the terrain and infrastructure.

Multiple emitters were used to create different sources of water, allowing us to experiment with flooding in specific areas rather than relying on a single global effect.

The system was tested against the hilly terrain, pathways, barriers, drainage structures, and riverside floodplain.

This phase established the technical foundation for more sophisticated flood scenarios in which environmental conditions, real-world weather and infrastructure could influence the resulting simulation.

7. Integrating Environmental Modelling and AI

The project then moved beyond visual simulation toward predictive modelling.

An ocean and flood-level modelling API was incorporated as part of the research into connecting external environmental data with the interactive environment.

The objective was to create a pathway by which environmental measurements and predictive models could influence the virtual world.

Research included investigation of river discharge, flood behaviour, stormwater infrastructure, and flood-control systems. River discharge is particularly important because understanding the relationship between channel geometry, water velocity, and flow provides a foundation for modelling when rivers may exceed their capacity and inundate surrounding areas.

The next stage of research considered incorporating location-specific river-depth data from Teknaf, Bangladesh to improve the representation of river flow and discharge within the model.

The project also explored the possibility of transferring the methodology to other geographic locations, demonstrating that the simulation architecture could potentially become a reusable framework rather than a single-purpose environment.

8. Preparing the Platform for Training

Although the project focused on flood modelling, the environment was designed with a broader application in mind.

The same technology can support emergency-response training by combining environmental simulation with instructional scenarios.

A flood condition can become the starting point for a training exercise.

An AI instructor can introduce the situation.

The trainee can investigate the environment, make decisions, communicate with other participants, and respond to changing conditions.

The simulation can then be repeated with different environmental parameters or scenarios.

Our predictive model canbe expanded to other flood-prone conditions, utilize map data, and integrate fire risk models.

This creates a direct connection between our environmental modelling research and our XR training platform, TrainVR Pro.

Outcomes

The project successfully established a working prototype demonstrating how environmental simulation, 3D visualization, interactive infrastructure, and external modelling technologies can be combined within an XR environment.

Key outcomes included:

• A purpose-built riverside suburban simulation environment

• Procedurally constructed terrain and floodplain

• Interactive flood-water simulation

• Multiple water emitters for scenario testing

• Examples of flood mitigation infrastructure

• Underground drainage and water-management systems

• Interactive navigation and location-based exploration

• Environmental and flood-level modelling integration

• Research into connecting real-world river data with simulation

• A foundation for AI-assisted environmental prediction

• A reusable approach that can be extended to other locations and disaster scenarios

• Conversational avatars for emergent training scenarios.

More importantly, the project demonstrated a capability: the ability to transform a complex environmental problem into an interactive digital laboratory.

Extending the Capability

Flood prediction is one application of a broader technical approach.

The same methodology can be applied to wildfire propagation, storm systems, industrial hazards, emergency evacuation, infrastructure failure, search and rescue, and other complex scenarios where physical conditions interact with human decisions.

Our role is to bring these systems together.

We can research the underlying problem, construct the 3D environment, integrate simulation technologies and external data, develop interactive interfaces, and connect the resulting environment to training and AI systems.

The result is more than a visualization.

It is a platform for simulation, experimentation, training, and decision-making.

From Research to Operational Experience

This project demonstrates how we approach emerging technology: start with a real problem, research the systems behind it, build a focused working environment, test the underlying technology, integrate external models, and develop a pathway toward a practical application.

The technology can evolve as better environmental data, predictive models, AI systems, and simulation technologies become available.

The virtual environment becomes the interface through which all of those systems can be understood.

We build immersive environments that make complex systems visible, testable, and actionable.