A Fire-Spotting Parameterization Coupled with the WRF-Fire Model

,

parameterization is assessed through a qualitative analysis of wildfires in Colorado.Uncertainties in fire ignition observations, used to initialize fires in the WRF-Fire model, often limit the ability to accurately model fire area, which in turn controls the firebrands' emission location.Limited spotting observations are also a challenge to an objective verification of the module skill.We expect that the most recent remote sensing products will improve the representation of surface properties and accuracy of ignition parameters for WRF-Fire, which will directly transfer to the fire-spotting module capability.Direct enhancements to the parameterization may be incorporated into the module as laboratory experiments and field campaigns provide data to improve our ability to model firebrands' initial properties (e.g.firebrand size and ejection height) and physical processes (burnout and terminal velocity).
The Fire-Spotting parameterization releases firebrands at gridpoint locations along the fire front, from grid points with high fire rate-of-spread and denser fuel loads.Firebrands are released at multiple heights, transported with the atmospheric flow, and consumed by combustion.Firebrands that are not entirely consumed and land on unburned grid points outside the active fire source area are accumulated in a 2-D field during regular intervals.
A fire-spotting likelihood field is computed using the ratio of landing firebrands per gridpoint to the total number of landing particles within the corresponding time interval of model outputs.The ratios are then scaled by a relative measure, function of fuel load and moisture content at the landing grid points.
Features in Fire-Spotting parameterization.
The parameterization runs in the atmospheric model inner domain using a large-eddy simulation (LES) to resolve boundary layer processes, including turbulence and fire-induced convection.The model codebase used in this study contains the most recent development implemented onto the Colorado Fire Prediction System (CO-FPS), including a 5th-order WENO level-set method for fire perimeter propagation, 40-category fuel type and load, dynamic fuel moisture, and the fire spotting likelihood parameterization.
The fire-spotting processes are integrated according to the inner model domain dimensions and timestep.Data arrays from WRF-Fire stored on the refined grid mesh are used with their refined resolution to calculate firebrand release points and rescaled to the atmospheric inner grid for spotting likelihood.

CAMERON PEAK FIRE SIMULATIONS
The Fire-Spotting parameterization is demonstrated with simulations of the Cameron Peak fire (https://inciweb.nwcg.gov/incident/6964/).The fire started on 08/13/2020 approx. 2 PM MDT on the Arapaho and Roosevelt National Forests, west of Chambers Lake, and has become the largest wildfire in Colorado's history (https://www.coloradoan.com/story/news/2020/11/17/cameron-peak-firecolorados-largest-wildfire-damages-homes-landscape/6212902002/).The fire was contained on 12/05/2020 with an estimated area of ~209 Acres._______________________Footnotes____________________ *Although the reference perimeter is timestamped at 22:30Z (white polygon), Suomi and NOAA-20 satellite fire products indicate that the spot fire east of the State Highway CO-14 was likely at its early stage by 20-21Z.

FIRE-SPOTTING SIMULATION EXPERIMENT _Firebrand Release______________________
Firebrands are released from the gridpoints along the fire front with the highest fire rate of spread and dry fuel load.

_Firebrand Landing______________________
Firebrands often land on neighboring gridpoints with active fire and are intentionally excluded in the spotting likelihood product.

FIRE-SPOTTING FORECAST ENSEMBLE
A set of simulations with varying ignition times was used as a 4-member ensemble to account for the uncertainty in the perimeter ignition time**.The ensemble members show multiple clusters with high fire spotting likelihood.All of the ensemble members indicate a high likelihood for spot fires in the region across the containment barriers (east of State Highway CO-14 and Cache La Poudre River. Relative fire spotting likelihood accumulated for a 6-hour period starting at 08/14 18Z (12 PM MDT) with the fire area representing the perimeter forecast by the end of the interval, at 08/14 21Z (6 PM MDT).The panel shows the four ensemble members with fire perimeter ignited at 0.5, 1.5, 3, and 4h from model initialization time(6:30, 7:30, 09, 10Z, resp.)_Fire-Spotting Likelihood___________________ _Firebrand Landing_______________________ _______________________Footnotes____________________ **The fire perimeters are created from mixed satellite imagery and often include additional aerial methods.Because the information pertains to a temporal window, uncertainties in the perimeters' timestamps are inherent.High-resolution satellite imagery is only available 2x day with observations 1-2h apart between different satellites, challenging the timestamping of spot fires and general wildfire forecast verification in the hourly time scale.Frequency and resolution of perimeters with reliable timestamps _________________________________________________ To improve resilience, preparedness, and response to wildfire disasters, researchers, stakeholders, and communities need a robust wildfire and spotting model.
Collaborations and partnerships are fundamental to overcome current challenges to develop a community model and verify wildfire forecasts.__________________Ongoing Research_______________ Incorporate canopy height to calculate firebrand release levels and deposit threshold Sensitivity tests for Emission processes: firebrand emission momentum, emission parameters (fire and fuel properties), number of particles Spotting likelihood parameters: fuel properties Merge the Fire Spotting parameterization into the WRF model public release _________________Acknowledgements_______________ Funding was provided by the State of Colorado through the Center of Excellence for Advanced Technology Aerial Firefighting, Division of Fire Prevention and Control, and by the National Science Foundation, which sponsors NCAR.Thanks to Brad Schmidt and Ben Miller (Center of Excellence for Advanced Technology Aerial Firefighting), Janice Coen (NCAR), and Representative Tracy Kraft-Tharp (Colorado General Assembly).