The United States has entered a new era of increasing wildfire frequency and intensity, which has culminated in several devastating wildfire seasons over the past decade. There is a significant need for wildland fire management decisions that are based on a multifaceted analysis of risks and benefits associated with wildfires and prescribed burns.
Accurate wildfire and smoke modeling is extremely sensitive to the initial conditions: how dry are the fuels, where are the strongest winds, and where is fire burning? How and where these elements of the combustion triangle come together has a dramatic influence on the subsequent fire spread and smoke production.
Current observational capabilities lack the coverage, resolution, and timeliness to produce the firefighter-scale forecasts that are needed to significantly improve wildfire management. However, improved observations alone will not lead to improved forecasts because there is a significant technological challenge in connecting observations with models.
This project synthesizes innovative observation capabilities, including mobile Doppler RADAR observations, the northern California Doppler wind LIDAR network, extremely high-resolution hyperspectral wind fields, and low latency satellite fire detections to develop the technology needed for utilizing these observations in coupled atmosphere-fire forecasting.
The forecasting system uses machine learning and data assimilation to integrate new observations in the open source coupled physical model WRF-SFIRE using the WRFx workflow management system. A combination of cloud computing and high-performance computing strategies are employed to provide forecast data and visualizations to users with the lowest possible latency.