1. General
Category
SDG 11: Sustainable Cities and Communities
SDG 13: Climate Action
SDG 15: Life on Land
Category
Computer Software
2. Project Details
Company or Institution
Fion Technologies Inc.
Project
Wildfire Spread Prediction
General description of the AI solution
Fion Technologies is a wildfire intelligence startup based in the San Francisco Bay Area. Fion leverages geospatial data and deep learning models to help customers understand wildfire behavior and make better decisions when dealing with them that save lives and protect properties.
Fion’s intelligence platform identifies areas at risk for wildfires, detects wildfire perimeters once they start using satellite imagery, and forecasts where active wildfires will spread over a 24 hour period.
Website
Organisation
Fion Technologies Inc.
3. Aspects
Excellence and Scientific Quality: Please detail the improvements made by the nominee or the nominees’ team or yourself if your applying for the award, and why they have been a success.
Currently, wildfire spread prediction worldwide is done using open source software called Farsite. A set of physics equations from 1975 called Rothermel’s Equations make up the foundation for Farsite. Wildfires are inherently random, physics equations are decidedly not – for any set of given inputs, there are a range of pre-defined outputs. Farsite does not work. Experts spend hours calibrating it by hand and know exactly what output it is going to give, so end up relying on their gut instinct. Machine learning must be used.
Until Fion, very few people in the fire space had the domain knowledge and insight to utilize machine learning for wildfire spread prediction. Fion’s deep learning model is built using a convolutional neural U-net architecture and trained on over 45 data layers from 10,000+ fires between 2012-2018. The model is production ready and has been used by the largest insurance companies in the world. Forbes has written about Fion and we have spoken at local town halls in San Mateo County with CalFire and our local state senator.
Scaling of impact to SDGs: Please detail how many citizens/communities and/or researchers/businesses this has had or can have a positive impact on, including particular groups where applicable and to what extent.
Over the past few years, wildfires went from an afterthought to rapidly affecting every country in the world. A large part of this is due to climate change. The heating of the earth leads to droughts, soil erosion, etc. which all contribute to wildfires getting larger and more frequent annually. This part cannot easily be controlled and will require many solutions. The part that can be controlled, is what we do once a fire has ignited. Currently, the forecasting capabilities are broken. In the US, a select group of fire behavior analysts effectively use their gut instinct to predict fire behavior and deploy troops to combat it and take mitigative actions. This is where Fion comes in. By providing accurate, 24 hour forecasts of fires, we allow governments to take decisive actions to save homes, businesses, and land before fires become uncontrollable. We are a welcome tool for state and federal government officials because we make their life easier and help them help their communities from burning down.
Scaling of AI solution: Please detail what proof of concept or implementations can you show now in terms of its efficacy and how the solution can be scaled to provide a global impact ad how realistic that scaling is.
Evidence for impact: We have been used by the largest insurance companies in the world and a state government in the US uses us. Data quality was the largest problem for us, but partnerships with several satellite companies and insurance companies solved that problem for us.
Scalability and sustainability of AI solution: By removing the need for calibration, we are making our government end users much more likely to use our solution and seek out other AI solutions.
Customer and end user: Fion is an API solution. Companies and governments use us as a primitive to build additional, more sophisticated tools (ie damage estimation) on top of our model.
Impact: We have been used by 2 of the largest insurance companies in the world, and are in use by a state government in the US. We do not store or collect any data on users.
Ethical aspect: Please detail the way the solution addresses any of the main ethical aspects, including trustworthiness, bias, gender issues, etc.
Our technology can only be used for good. Whether it is predicting the spread of a prescribed fire or a real wildfire, we are only able to be used for good and have stringent screening processes for our enterprise and government customers. Our model is fully compliant with US and international laws, adheres to ethical principles and values, and works worldwide given that we have sufficient data (which we can either generate or a customer can provide us). We have a diverse team representing a variety of races, genders, and cultural backgrounds.