Integrating Innovative Observation Capabilities into Coupled Wildfire Models
This project will develop the technology needed for integrating emerging wildfire observation capabilities into coupled atmosphere-fire forecast modeling. Using advanced cloud computing and data assimilation techniques, this work will produce more accurate initial conditions for wildfire models and provide new opportunities for forecast verification. These improvements will enable more accurate forecasts of phenomena like fire spread and smoke dispersion—forecasts that can be generated in the field by wildfire managers.
Accurate wildfire and smoke modeling is a complex – but essential – component of successful wildfire management strategies. Largely dependent on initial burn conditions, reliable wildfire and smoke models require not only well-instrumented field campaigns, but also information systems capable of synthesizing in-situ, airborne, and spaceborne observations into a single, accessible science product. These advanced information systems improve fire incident responders’ ability to model both flame geometry and fireline dynamics.
This project will develop strategies for the cost-effective use of cloud computing and high-performance computing to obtain the lowest possible latency. Physical models alone are unlikely to meet latency requirements especially for providing probabilistic information from ensemble modeling, so machine learning will be explored to accelerate the physical model and data assimilation components of the coupled modeling system. The machine learning development will be applicable to numerous modeling systems, and to data assimilation and ensemble modeling for fire and smoke forecasting in general.
- High-efficiency cloud computing will request and process direct broadcast data to deliver low latency VIIRS active fire products in less than 12 minutes.
- First-of-its-kind data assimilation will incorporate 3D Doppler lidar and radar measurements into traditional wildfire models, dramatically improving the accuracy of fire spread forecasts.
- A flexible post-processing system will combat information overload by letting wildfire managers control which variables will be processed and how that data will be visualized.
Kyle Hilburn is an atmospheric scientist, specializing in remote sensing, artificial intelligence, and data assimilation. He is the project’s principal investigator. He received his PhD in atmospheric science from Colorado State University (CSU) in 2023, and he has been a Research Scientist at the National Oceanic and Atmospheric Administration (NOAA) Cooperative Institute for Research in the Atmosphere (CIRA) at CSU since 2016. Kyle is assisted by co-investigators Jan Mandel at University of Colorado Denver, Adam Kochanski at San Jose State University, Assad Oberai at University of Southern California, Louis Nguyen at NASA Langley Research Center, Andrew Michaelis at NASA Ames Research Center, and Derek Mallia at University of Utah.