Firefighters fighting a forest fire (Photo: Roland Plett, pixabay.com)

NSF NCAR scientists combine two forecast models for significantly greater accuracy

The spread of forest fires can now be predicted more accurately thanks to a new system developed by experts at the National Centre for Atmospheric Research (https://ncar.ucar.edu) (NSF NCAR). Endangered areas can thus be evacuated in good time and fire-fighting teams can be strategically and skilfully positioned. The experts used computers to simulate the devastating forest fires that ravaged the Hawaiian town of Lahaina last year.

Prediction within minutes

“Our simulation shows that in the not too distant future we will be able to predict the spread of fires within minutes if we know where and when they start. In the future, our approach can serve as a basis for understanding how extreme weather conditions affect fire behaviour in different types of built environments. Ultimately, we can then better protect vulnerable communities,” said NSF-NCAR researcher Timothy Juliano.
In recent years, fast-spreading wildfires have devastated communities as far away as Superior and Louisville in Colorado, Talent and Phoenix in Oregon, Paradise in California, and Gatlinburg and Pigeon Forge in Tennessee. The Lahaina Fire of 8 and 9 August 2023 was particularly tragic, killing 100 people and destroying over 2,200 buildings. In each of these cases, the flames spread from forests to residential areas and shopping centres in extreme wind conditions.

Combination of two models

To get a detailed picture of the Lahaina fire, the team combined two computer models with different capabilities. One, the NSF-NCAR-based Weather Research and Forecasting (WRF) model, was used by the team to simulate the downburst storm that broke out on the day of the fire, generating gusts of up to 130 kilometres per hour. The high-resolution model was able to show the turbulent wind currents around Lahaina and a “hydraulic jump” – an event in which winds flowing downhill rise abruptly when they collide with winds flowing in the other direction, triggering chaotic and strong air movements.
The researchers have fed the wind fields from the WRF simulation into the second model “Streamlined Wildland-Urban Interface Fire Tracing”. This allows the spread of flames in a built-up area to be simulated by recording how extreme heat and rising embers set structures on fire. The simulation results generally corresponded well with witness reports and video recordings of the spread of fire. “This shows the potential for developing a decision support system for active forest fires,” says Juliano.