NOAA’s Wildfire Smoke Forecasting Model Scored a Win with the Camp Fire

The Camp Fire in Paradise, California, was the deadliest and most destructive wildfire in California's history, and the most expensive natural disaster in the world in 2018. The wildfire claimed 88 lives and blanketed millions of people in Northern California with dense smoke for two weeks. 

The Camp Fire is just one of many major wildfires that have choked the summer skies across the United States in recent years, increasing the demand for accurate air quality forecasting. In a new study published in the Bulletin of the American Meteorological Society, researchers demonstrated that NOAA’s experimental HRRR-Smoke model accurately predicted the general movement and concentration of the Camp Fire’s smoke. 

“HRRR-Smoke is a powerful tool for forecasting such extreme smoke pollution events,” said paper co-author Ravan Ahmadov, a CIRES scientist in NOAA's Global Systems Laboratory (GSL) who helped lead the development of HRRR-Smoke. “The model accurately captured the spread of dense smoke, an impressive achievement given wind conditions that drove the smoke downvalley toward the Bay Area, over the Sierras into the Central Valley, and across the Central Valley to the coastal mountain range.”

The HRRR-Smoke model—actually a module within NOAA’s High-Resolution Rapid Refresh (HRRR) short–term weather model—predicts the transport of wildfire smoke and the smoke’s impact on visibility and weather over the United States. Using the most recent weather data and information about fires detected by polar-orbiting satellites during the previous 24 hours, HRRR-Smoke calculates a fire’s size and couples that information with weather simulations from HRRR to produce forecasts of near-surface smoke and smoke aloft. The model was developed by GSL and transitioned to the NOAA National Weather Service operational suite of models in 2020. 

Using the 2018 Camp Fire as a validation case, the researchers compared model forecasts with data from local weather stations and air quality sensors. They determined that HRRR-Smoke accurately predicted the intensification of smoke pollution when a high-pressure system trapped extreme amounts of smoke close to the surface. The model was able to provide smoke forecasts at a resolution of 3 km—or at the level of individual neighborhoods.

animation from the HRRR-Smoke model on 11/11/2018, showing how thick smoke was predicted to blanket parts of northern California during the Camp fire. Credit: noaa's Global Systems Laboratory

“The model did a very good job,” said Tina Katopodes Chow, a professor of civil and environmental engineering at the University of California Berkeley and the study’s lead author. “You could use it for overall predictions of how long the smoke will last over a certain area.” 

The study also identified aspects of the model that could be improved. For example, during the second week, the model underperformed—perhaps because of thick smoke obscuring the view of satellites that pick up fire’s heat signatures to initialize the model. Presently, HRRR-Smoke relies on relatively infrequent satellite observations of fire radiative power. The researchers said they may be able to tap into NOAA’s new GOES-18 satellite, a geostationary satellite which could provide much more frequent observations. 

Researchers in GSL have applied what they learned from this case and others like it to improve NOAA’s next-generation smoke model, a module of the Rapid Refresh Forecast System. Additional validation and model improvements will translate into more accurate prediction of wildfire or prescribed burn smoke events, ensuring community health and safety. 

“With wildfires now creating large-scale smoke events regularly affecting many people in the western United States, HRRR-Smoke is an essential tool for providing real-time operational support for weather and air quality forecasters,” said co-author Eric James, a CIRES researcher in GSL.

Other co-authors of the study include researchers from NOAA’s Center for Satellite Applications and Research, the I.M. Systems Group, the Universities Space Research Association, NASA Goddard Space Flight Center, the Federal University of São João del-Rei in Brazil and the University of Maryland.