Terrafuse leverages physics-enabled AI models to help organisations understand climate-related risk at the hyperlocal level.
In October 2017, wildfires devastated Northern California, burning 245,000 acres and 8,900 buildings. They were the costliest wildfires ever, including USD $11 billion in insured losses. With so much on the line, government agencies, insurance companies and the general public need a way to accurately forecast and understand wildfire risk at any given location.
Terrafuse is building technology infrastructure on Microsoft Azure to rapidly forecast wildfire risk at the hyperlocal level, starting in areas affected by the 2017 California wildfires. By combining historical fire data with existing physical simulations and real-time satellite observations, Terrafuse is building sophisticated fire risk models that will be made available via APIs and graphical interfaces to anyone interested in mitigating the effects of wildfires.
Assessing wildfire risk with machine learning
Terrafuse uses machine learning to forecast climate-related risks. Terrafuse leverages historical wildfire data, numerical simulations and satellite imagery on Microsoft Azure to model wildfire risk for any location. Wildfire forecast information is made available via APIs and graphical tools.
How Terrafuse works
A fire-prone area is identified. Terrafuse uses the cloud to aggregate historical fire spread data, fire simulations and real-time satellite observations of rainfall, wind, soil and moisture. Physics-informed machine learning creates models so governments, insurance companies and the public can access and explore fire risk.