Campus, NOAA study validates wildfire smoke forecasting model

photo of orange sky from wildfire smoke
Lisi Ludwig/Senior Staff
A study conducted by UC Berkeley researchers in collaboration with the National Oceanic and Atmospheric Association evaluated a model used for forecasting the movement of smoke in the 2018 Camp Fire.

Related Posts

In collaboration with the National Oceanic and Atmospheric Association, or NOAA, and other research institutions, UC Berkeley researchers evaluated NOAA’s High-Resolution Rapid Refresh wildfire smoke model, or HRRR-Smoke, that forecasted concentration and movement of smoke during the 2018 Camp Fire.

Campus civil and environmental engineering professor Tina Chow said the Camp Fire is an ideal model for evaluation as it was a singular, regional burn, releasing smoke from one area and affecting the entire Bay Area. The Camp Fire remains the most destructive and deadliest wildfire in California’s recorded history, resulting in 85 casualties and the destruction of over 18,000 structures, according to CAL FIRE.

The U.S. air quality index surpassed 200 a week after its ignition, and the “very unhealthy” air quality status prompted campus Chancellor Carol Christ to announce a cancellation of classes on Nov. 15, 2018 that later extended until Nov. 20.

“There was just that one fire going on,” Chow said. “It’s easy to say that all of the smoke was coming from the camp fire over that entire region.”

Chow said she constantly checked air quality forecasts to evaluate smoke and wind conditions during the fire. In the midst of her search, she came across the then-experimental HRRR-Smoke website, prompting her to reach out to NOAA modelers for collaboration on the project.

The HRRR-Smoke model became operational December 2020 and is considered as an official U.S. forecasting tool, according to Eric James, a co-author and research associate within the NOAA Global Systems Laboratory and senior research associate at the Cooperative Institute for Research in Environmental Sciences at the University of Colorado Boulder.

“This study is informative in the sense that it helps other researchers understand the strengths and weaknesses of the smoke model that we were using,” said evaluation co-author and campus alumnus Alexander Young. 

The recent study published in the Bulletin of the American Meteorological Society in June is one of the first most in-depth validations of the HRRR-Smoke model, Chow said.

Researchers compared outputs of the HRRR-Smoke model to various data sources, such as forecast sensors from the Environmental Protection Agency sensory network, satellite imagery and observations of atmospheric conditions to see whether the concentration of smoke at a given location matched, according to Young.

“The model does perform quite well and can help predict the location of smoke plume and its concentration with a certain amount of lead time, for example, a day in advance,” Young said.

However, one of the limitations of the model is that the result showed an underpredicted amount of smoke during the second week of the Camp Fire. James and Chow said the huge amount of smoke over the fire during the second week inhibited the satellites ability to detect the fire.

James said this limitation shows the importance of high quality satellite detections and motivates future work for satellites to improve their algorithms during smoky periods.

Young said this study will allow scientists and the modelers to use its data to improve future smoke forecasting and to mitigate potential health effects.

“We are working hard to develop the new system which is going to be called the Rapid Refresh Forecast System and it’s going to be essentially a replacement for the HRRR,” James said.

Apart from the current parameters in HRRR-Smoke, the new model will include dust blowing which enables prediction in reduced visibility due to dust in the deserts, James added.

The Rapid Refresh Forecast System aims to streamline overlapping models and reduce maintenance and code management. He added it will be able to forecast fire, smoke, severe thunderstorms, precipitation and other atmospheric weather conditions.

“If you just look at the weather forecast, every time they collect data about the current conditions in the weather, they’re updating the weather model,” Chow said. “Now we’re just saying we should do the same thing for smoke.”

Winnie Lau is the lead environment and climate reporter. Contact her at [email protected], and follow her on Twitter at @winniewy_lau.

Correction(s):
A previous version of the photo caption incorrectly stated that the NOAA smoke forecasting model forecasted the movement of the 2018 Camp Fire. In fact, the NOAA model forecasted the movement of smoke in the 2018 Camp Fire.
A previous version of this article said Chow constantly checked “air condition programs” to evaluate “air and wind direction conditions” during the fire. In fact, Chow checked air quality forecasts to evaluate smoke and wind conditions.
A previous version of this article said the data sources include “forecast sensors” and “the Environmental Protection Agency sensory network”. In fact, data sources include air quality sensors from the Environmental Protection Agency sensor network.
A previous version of this article stated that Young said this study will allow scientists and the modelers to use its data to improve future smoke forecasting and potential health effects. In fact, this study will help to mitigate potential health effects.