Cheyenne supports active wildfire research as fires burn

When wildfires multiply and grow quickly, endangering lives and property as they rage through forests and grasslands, they stretch fire and rescue resources to their limits. As fires increase in size and intensity, their power to modify local weather conditions increases as well. Running coupled fire-atmosphere models on Cheyenne is allowing researchers not only to study how fires create their own weather but also to forecast such events to inform firefighters about potential changes in weather conditions and smoke pollution induced by raging fires.

Smoke plume in the background from a prescribed, experimental burn in Fishlake National Forest, Utah. Foreground, truck-mounted mobile Ka-band polarized Doppler radar.

Truck-mounted mobile Ka-band polarized Doppler radar at Annabella Reservoir Prescribed/Experimental Burn. A part of the Fire and Smoke Model Evaluation Experiment (FASMEE), Fishlake National Forest. Utah, November 5, 2020. Photo by Adam Kochanski. FASMEE was informed by pre-fire analysis and simulation on Cheyenne, described in Adam K. Kochanski, Aimé Fournier, and Jan Mandel: "Experimental Design of a Prescribed Burn Instrumentation." Atmosphere, 9(8), Article number 296, 2018. doi:10.3390/atmos9080296

A team of scientists from several universities – including the University of Colorado Denver, San Jose State University, and the University of Utah – has been doing just that during this year’s damaging fires that have raged in California and Colorado. Their work is funded by the National Science Foundation Prediction of and Resilience against Extreme Events (PREEVENTS) program, the NASA Earth Science Disasters Program, and the USDA Forest Service Pacific Northwest (PNW) Research Station Fire and Smoke Model Evaluation Experiment (FASMEE).

Jan Mandel, professor of mathematical and statistical sciences at CU Denver, said the team has been using the WRF-SFIRE model to run coupled fire-atmosphere forecasts based on a workflow initiated and developed by Adam Kochanski, assistant professor of meteorology and climate science at San Jose State University. Mandel said the resulting forecasts, which were shared as animated maps on the project’s web portal and the NASA Disasters Mapping Portal, improved situational awareness by providing information about fire growth as well as local weather conditions, fuel moisture, and smoke concentrations. Being able to execute the model and share operational forecasts with responders was critical to present the new modeling system to practitioners such as the U.S. Forest Service and California National Guard as well as to receive feedback that led to numerous system improvements. Based on the users’ feedback, a surface visibility forecast was added to the system to support evacuations, and smoke at altitude levels was added to support aircraft operations near active fires.

At the peak of fire activity in Colorado and California, the team used nearly 215,000 Cheyenne core-hours in September and 212,000 in October. Team members also used Casper cluster GPUs for some of their work.

As of early November 2020, the system was still running, supporting prescribed burn operations in Fishlake National Forest in Utah as part of the FASMEE led by Roger Ottmar from the PNW Research Station. High-resolution forecasts of local weather, fire progression, plume evolution, and smoke dispersion run by the San Jose State University team helped make final decisions about the burn and assisted in experimental planning. Being able to predict local weather conditions and smoke predictions helped safely allocate research assets and collect critical data for model development and validation.

The fires and studies go on. Mandel said development and scientific runs will proceed through the winter and simulations of new fires will continue into 2021. “We are now focusing on testing and validation after a recent upgrade of WRF-SFIRE to a fork of WRF 4.2. Aside from scientific runs, we provide daily forecasts for selected wildfires with fire progression and smoke predictions, as well as hourly fuel moisture nowcasts for the Western U.S., running continuously.”

This work is funded by the NSF PREEVENTS program (PI Steven Krueger, University of Utah), the NASA Earth Science Disasters Program (PI Kyle Hilburn, Colorado State University), FASMEE and the Utah Division of Air Quality (PI Adam Kochanski, San Jose State University, Wildfire Interdisciplinary Research Center). Members of the team include James Haley, graduate student at the University of Colorado Denver; Derek Mallia, research assistant professor of atmospheric sciences at the University of Utah; and Kochanski, assistant professor of meteorology and climate science, and postdoc Angel Farguell at San Jose State’s Wildfire Interdisciplinary Research Center.

A screenshot of the web portal with an animation of fire progression and surface smoke forecast for the 2020 Creek Fire in California

A screenshot of the web portal with an animation of fire progression and surface smoke forecast for the 2020 Creek Fire. Photo by Jan Mandel. Simulation by Angel Farguell based on a workflow developed by Adam Kochanski and the system described in Jan Mandel, Martin Vejmelka, Adam K. Kochanski, Angel Farguell, James D. Haley, Derek V. Mallia, and Kyle Hilburn: "An Interactive Data-Driven HPC System for Forecasting Weather, Wildland Fire, and Smoke," 2019 IEEE/ACM HPC for Urgent Decision Making (UrgentHPC), Denver, CO, USA, pages 35–44. IEEE, 2019.