Wild Fire Forecasting

To further the goal of developing short-range forecasts of weather & wildfire behavior with models steered by real-time weather data, fire-mapping images, and sensor streams, new data assimilation methods have been developed to assimilate fire mapper data into numerical models.

In 2007, Janice and colleagues will develop verification metrics to assess forecasted fire/weather model results in comparison to satellite and aircraft data and release a beta version of WRF-Fire for community user evaluation.

While there are several tools available for simulating fire behavior ranging from simple equations to cumbersome complex computational fluid dynamics/chemical reaction numerical models, none address the interconnected weather-fire system as a forecasting problem.  Scientists at NCAR, in collaboration with university, state and federal agency researchers are trying to do just that: Create a fire model that takes wind and humidity information from the atmospheric model, calculates how the fire would behave and grow, and provide releases of heat, water vapor, and smoke back into the atmospheric model.  Looking at wildfire as a forecasting problem rather than just a modeling problem means not only simulating the process, but developing objective verification techniques and measures of skill, including techniques like data assimilation and ensemble forecasting, that weave data and information about model and data errors into the running model.

FY06 Accomplishments

J. Coen is studying how multiple fires interact by modifying the atmospheric dynamics in their vicinity.  This is an important, but entirely unstudied, topic for fire crews setting backfires. If a backfire of the right intensity is set close to a wildfire, it is drawn back into the main fire, creating a fire break that stops the wildfire; if the backfire is too intense or set too far away, it becomes another uncontrollable fire.  Work is also focused on improving short-range weather forecasting techniques by focusing on variables such as wind speed and direction and use of proper scales (100s of m grid spacing), all of which are critical for understanding fire behavior. Another area of research relates to development of objective verification techniques for determining how 'good' a forecast of the speed and direction of spread, intensity, and area burned by a fire is and for understanding the cause of errors in a fire forecast. Work has also focused on building a real-time fire/atmosphere modeling system steered by data assimilation.  This involves real-time modeling of weather and fire growth using images of fire location and weather sensors developed at Rochester Institute of Technology, as well as work with University of Colorado researchers to develop novel data assimilation techniques for assimilating out of order data and non-Gaussian distributed errors

FY07 Plans

Verification metrics will be developed to assess forecasted fire/weather model results in comparison to satellite and aircraft data. The WRF-Fire model is currently available to university researchers and operational weather and fire management personnel for use in their daily forecasting activities.  As part of an agreement with partners at NIST and NCEP, NCAR will be adapting WRF-Fire to be part of the NCEP suite of WRF forecast products.  A beta version will be released in upcoming months for community user evaluation.