Flood Risk Visualization in Complex Virtual Worlds

How Our Project Will Contribute To The Growth Of The Decentralized AI Platform

Extreme weather events are increasingly common: heavy rains, tropical storms & hurricanes flooding riverine & coastal regions, all resulting in loss of life, property destruction & economic pressures on all levels.

Combining robust riverine flood datasets with real-time immersive 3D first-person simulation can become a robust business for SNET, applicable across national boundaries. This will help grow the AI platform with new markets with policy-makers, government & emergency response. 

Company Name (if applicable)

Xyris XR

Description of the new services.

We will develop a flood prediction model of select locations based on Google's Flood Hub strategy of topological & hydrological model training. We will build an API service to access this model data.

Training a machine learning model combining hydrology, inundation models, and topographic map layers has proven to compensate for areas of scarce data points while actually improving accuracy (see links).

We will compare our model with Google Flood Hub's forecasting outputs to determine our accuracy. Of course, a model benefits from more iterations, and we expect to continue improving the model throughout and after this project. 

We will connect API data from our flood model to our 3D environment simulating terrain, water sources, civic infrastructure, traffic, citizenry and first responders, providing an effective multidimensional first-person immersive experience.

With this combination we can demonstrate the value of presenting flood information in an experiential first-person context.

Names of the existing services

We'd like to utilize Photrek's Risk-Aware Assessment and SNET's Time Series Forcasting to analyse the incoming data, and potentially provide data visualization within the Unity UI as an adjunct to the experiential environment.

We will also utilize Carbix's ChatGPT and Speech to text APIs to make conversation possible with in-world avatars enacting the roles of citizens and responders.

We are also very interested in new services being built during Round 3. We're aware of other SingularityNET services coming online that will ease our access to datasets and to combine SingularityNET services. We'll use these too as they become available.

The core problem we are aiming to solve

Flooding due to climate change is having an increasingly significant impact on lives, housing, infrastructure & the economy. 
In the last three years, in Canada alone, tropical storms and hurricanes, along with exceptional levels of precipitation, have created catastrophic flooding on both coasts.
Map data coverage is often insufficient in areas most at risk due to lack of budget in smaller civic areas. Communicating these changing conditions to long-term residents & communicating strategies to volunteer responders are additional barriers to safety.

We need a way to visually portray climate-change-driven flooding events, based on data, to overcome climate skepticsm & promote life and property-saving policy changes.

AI modelling of flood prediction helps overcome scarce data points & reduces complexity of physical simulations. Pairing model data with a first-person interactive demonstration of flood consequences & strategies is highly convincing to stakeholders who may not have the science training needed to visualize the data.

Our specific solution to this problem

1. Train our own Flood Prediction model based on existing datasets and public API services.

2. Build an API which can return riverine flood prediction data from our model.

3. Within Unity, demonstrate effects of flood waters on a generated city's infrastructure:        

     - Overwhelmed levees and berms         
     - Overwhelmed sewage channels         
     - Eroded home foundations of varying quality (dirt, stone, cinder block, poured concrete, concrete posts).         
     - Interrupted services (electrical, cellular, gas, transit)         
     - Boyant objects and debris         
     - Catastrophic failure of civic structures

4. Simulate crowd behaviour of citizenry and behaviour of first responders in Unity3D.

5. Select avatars to be given conversational abiilty VIA Carbix's ChatGPT SNET API services.
     - Additionally, script an avatar lesson on emergency preparedness for residents in flood-prone areas.
     - Provide emergency evacuation routes for residents & NGOs, based on policy & data, for our select locations.
     - Compare established evacuation routes with our model's prediction of flooded locations.

Project details

Introduction:

At the time of writing, both coasts of North America is receiving a potential historic high precipitation event impacting power infrastructure and neccessitating evacuations. Canada's East coast is seeing record snowfall of up to four feet, and snowdrifts of up to ten feet, over only a few days creating subsequent flood risks from impending snowmelt and service interruptions. 
In Cape Breton, snowfall drifts are heavy enough to destroy windows, exposing occupants to frigid temperatures. Neither can first responders get through the snowfilled streets to help.
When the accumulated snow melts, these may overwhelm infrastructure with not designed for this volume of water.

In an era marked by the escalating frequency of extreme weather events such as wildfires, heavy rains, and tropical storms, the imperative for advanced flood prediction capabilities has become more pronounced. These events pose threats to coastal & riverine regions, leading to the loss of life, property destruction, and economic strains.

The integration of artificial intelligence services for modeling flood threats, coupled with immersive applications for education and training in these emergency situations, is crucial.
Our proposal focuses on developing a flood prediction model by incorporating historical precipitation with flooding datasets, then representing the data in our interactive real-time 3D Unity environment.

Project Overview:

We propose the development of a flood prediction model grounded in existing datasets made accessible through an API service on the SingularityNET marketplace. To augment user understanding, we will leverage Unity3D to visualize the flood data in a first-person simulation depicting civic infrastructure, traffic, citizens, and emergency responders.

Utilizing SingularityNET Marketplace Services:

To bolster the accuracy and effectiveness of our flood prediction model, we plan to integrate Photrek's Risk-Aware Assessment to check model accuracy. We will also compare our model outcomes with Google's Flood Hub alert JSON, which contains limited flood forecasting data. 

Additionally, we intend to incorporate Carbix's ChatGPT and Speech-to-Text APIs to enable conversations with avatars within the simulated world, allowing avatars to assume the roles of citizens and responders responding to user questions.

Addressing Climate Change Impacts:

The devastating impact of climate change-induced flooding on lives, housing, infrastructure, and the economy necessitates urgent action. Recent catastrophic events in Canada, including tropical storms, hurricanes, and historic flooding with shoreline loss, underline the urgency of our project. 

We aim to visually depict climate-change-driven flooding events based on modelling data to counter climate skepticism and advocate for life and property-saving policy changes.

Project Objectives:

1. Train a proprietary Flood Prediction model using existing datasets and services.

2. Develop an API providing riverine flood prediction data for select locations or sets of topographic and hydrological map layers.

3. Showcase flood effects on infrastructure within Unity3D, including portrayal of overwhelmed levees, berms, sewage channels, eroded home foundations, and disruption of civic services.

4. Simulate crowd behavior and first responder movement within the flood area.

5. Implement contextual conversational abilities for avatars using Carbix's ChatGPT API services.

Resources and Capabilities:

With an extensive library of Unity assets, including terrain generated from map layer data, city generation tools, the Kitbash3D model library, customizable avatar design software, behavioral graphs, fluid simulation tools, and experience in this medium, we are well-equipped to efficiently construct detailed Minimum Viable Products (MVPs). 
Our capabilities also extend to sound design, text-to-speech, conversational avatars, and Virtual Reality (VR) builds.

This robust toolkit and experience positions us to deliver a cutting-edge compelling and realistic user experience.

The integration of our flood prediction model and real-time in-game 3D simulation technology provides invaluable benefits to various stakeholders including policy makers, residents, and first responders. With enhanced predictive capabilities, we can refine our understanding of flooding impacts, facilitate research and foster advancements in flood mitigation strategies.

Benefits:

Policy makers stand to gain significant advantages from our proposal. The real-time 3D environment allows for communicating complex flood scenarios in the first person, provides shared experience in the simulation, allows the design of challenging situations without personal risk, while aiding policy makers testing informed decisions about first response strategies, land-use planning, extreme weather risks, infrastructure development, and disaster preparedness.

This immersive approach enables us to comprehend the severity and chaos of climate change-induced flooding, then formulate policies that are not only effective but also responsive to the evolving nature of these threats.

For the public, the benefits of our project extend to education of flood risks & being prepared for extended periods without electricity, heat and clean water, with potentially the need for basic first aid training. 

This heightened awareness empowers individuals to make informed decisions, evacuate when necessary, and actively participate in community resilience efforts. 

Moreover, the economic advantages are evident as well, with potential for reduced property damage and loss of life, leading to significant cost savings for both individuals and the broader economy. 

Finally, emergency response teams could benefit from our flood prediction model by gaining access to data and immersive simulations enhancing their preparedness conversations and decision-making during flood events, ultimately improving their effectiveness in safeguarding communities and mitigating the impact of disasters.

Conclusion:

In conclusion, our proposal to develop a flood prediction model, coupled with a highly interactive real-time 3D simulation, represents a strategic and worthwhile investment. The integration of AI-driven flood prediction and Unity3D simulations not only addresses the urgent need for improved forecasting in the face of escalating climate change-induced flooding but also offers a multi-faceted approach benefitting scientists, policy makers, the public, the economy, and emergency response teams.

The public benefits from increased awareness and education on flood risks, empowering communities to take proactive measures.

From an economic perspective, our proposal offers significant cost savings by visualising policy improvements that could lead to reduced property damage and enhanced protection of life.

In essence, this investment is more than a technological endeavor; it is a commitment to resilience, informed decision-making, and the safeguarding of lives and property.

Competition and USPs

Google Flood Hub provides limited (but highly accurate) flood prediction data. ArcGIS provides flood mapping. We distinguish our model as open source, and usable within a real-time environment like Unity3D. 

We believe showing how effective a first-person demonstration of flood data is, and how accessible the model is via the SingularityNet marketplace is, without requiring an expensive software license, and aiding communities in flood risk locations, will help us to succeed in the market.

Needed resources

Collaboration with Singularity community leaders flood emergency protocols, in probability assessment and in model training are most welcome. I have external local contacts in mapping and geology who have offered access to their data and expertise as well.

Existing resources

We have an extensive library of Unity assets including map layer transformers used to generate terrain, roads, building massing models, rivers and lakes into a real-time format.

City generation tools, the entire Kitbash3D model library of buildings and objects, avatar design software, behavioural graphs, animal models and behavioural simulators, fluid simulation, weather simulation, ocean sims, physics sims, vehicles and vehicle controllers, time of day effects, localized weather, geosynced night/day cycles and more also allow us to quickly & efficiently build immersive MVPs.

Additionally I'm equipped for professional sound design, text-to-speech, conversational avatars, and VR builds. 

Collaboration with Singularity community leaders in probability assessment and modelling are our most powerful resource. 

Open Source Licensing

Commercial Unity assets, datasets under other ownership, and real time Unity original project files & assets cannot be included in any open source commitment because of software license restrictions. 

The API code to access our trained model and the model code shall be made open source at the project conclusion.

Links and references

Flood Forcasting
https://sites.research.google/floodforecasting/
https://blog.google/technology/ai/flood-forecasts-india-bangladesh/
https://blog.research.google/2020/09/the-technology-behind-our-recent.html

Dataset candidates:
https://shorturl.at/gsuZ8
https://crisisresponse.google/forecasting-and-alerts/

AI & Flood Forcast:
https://arxiv.org/abs/2307.16104Floodhub
https://sites.research.google/floods/l/0/0/3

Additional videos

Miro board: https://miro.com/app/board/uXjVNxtJpNg=/?share_link_id=992813906157

Carbix MVSE Highlight Reel (avatars & immersive installations)
https://www.xyris.ca/carbix-mvse/

Quick city visualization with a flooding event, and Unity map layer import screens:
https://shorturl.at/duJO0

Ocean simulation example (science fiction environment with Kitbash3D models):
https://youtu.be/jxQwmg7sLms?si=SwrmrVFk5E4Wm3Jn
Nightime city simulation:
https://youtu.be/kHhwkuJHJWo?si=CI0Ft4v2YsBVy3nM

Revenue Sharing Model

IP Sharing Percentage: