Coupled Weather-Fire Modeling

Decision Support For Managing Wildland Fires

To fight wildland fires, decision makers need reliable, accurate, frequently updated, readily accessible, geo-referenced information about current and predicted weather and fire behavior. With this information, decision makers can better determine how a fire is behaving now and might behave in the future. Reliable information about the potential for a fire to spread rapidly and behave erratically is essential for saving life and property.


On October 14, 2020, the East Troublesome Wildfire ignited. It burnt over 190,000 acres over the course of 48 days, and nearly quadrupled in size between October 20 and 23. This visualization shows a Weather Research & Forecasting (WRF) simulation of the East Troublesome Wildfire from the perspective of Granby, Colorado. It uses a physics module (WRF-Fire) that incorporates an AI driven model for combustable plant matter known as Beetle Kill.


Currently, operational systems that predict how wildland fires move and behave are not coupled interactively to numerical weather prediction (NWP) models. These systems often rely on wind fields that lack critical details in space and time. Those are details essential to accurately predicting fire spread when winds change rapidly due to storm outflows, density currents, frontal passages, complex terrain, and other factors. Furthermore, large wildfires generate their own powerful updrafts and intense local winds, which drive flames quickly across the landscape. Large wildfires also generate vast, thick smoke plumes that cast shadows and change temperatures over large areas, while lofted particles and moisture can form pyrocumulus clouds. All of these phenomena can be predicted only by coupled models.

To fill this gap, scientists and engineers in RAL extended the functionality of the Weather Research and Forecasting (WRF) NWP model. New developments aim to improve the fire-spread model, investigate alternative fuel models and fuel-moisture data, allowing users to fine-tune fuel moisture in simulations, and predict where new spot-fires are likely to ignite. These developments are being included in the community WRF-Fire model.  This modeling system is being extensively evaluated and improved, based on what we're learning from fires observed in Colorado and other parts of the United States.

Our model can simulate thousands of individual embers and predict the locations where spotting is more likely

Our WRF-Fire model can simulate thousands of individual embers and predict the locations where spotting is more likely


During windy and dry weather, particularly when the surrounding vegetation is also dry, embers can travel farther and over fire barriers, such as water streams and roads, and result in rapid fire growth. Intense spotting increases danger to firefighters, affects our ability to predict fire spread, and challenges efforts to suppress it. In an urban setting, embers are the leading cause of home ignitions.

The WRF Model with extended functionality predicts spotting likelihood by simulating embers lofted into the atmosphere from locations where the fire is intense, carried with the wind, and landing on the surface near or far ahead downwind. The model takes into account the number of embers landing at each location, and that location's soil and vegetation conditions.


  • CO-FPS
  • ATEC

Representative Projects

  • LEAP-HI Project aims to develop a unique system for centralizing detailed assessments of wildland fire risk, alerting residents and firefighters days to weeks in advance of the potential for a major fire, RAL's assignment in this effort is combining satellite imagery of land surfaces with highly detailed weather forecasts. These data will be fed into WRF-Fire, which will help identify areas most at risk from blazes. For more about the LEAP-HI project, see our feature article.



Branko Kosovic

Director, Weather Systems Assessment Program


Jason Knievel

Deputy Director, RAL NSAP