Introduction
Firebrands or embers are burning pieces of materials (e.g., branches, bark, building materials) generated at a fire source and carried with the wind and convection. Fire spotting occurs when firebrands are lofted into the air, land on unburned areas, and ignite new fires \cite{FernandezPello2017}. Spot fires are considered short-range within a few hundred meters from the source fire, or long-range, with reports of spotting distances as high as 35 km \cite{storey2021}. Short-range spotting accelerates the fire rate of spread by expanding the fire perimeters beyond the fire front, whereas long distance spotting can ignite new fires several kilometers downwind, possibly in areas beyond containment boundaries. Fires driven by high wind speed combined with low relative humidity and flammable vegetation often result in high fire intensities, rapid growth rates, and showers of embers that can start new fires. Intense spotting increases danger to fire crews, affects fire behavior predictability, and challenges emergency response and containment efforts.
Embers play a significant role in spreading wildfires and are important especially in the wildland-urban interface (WUI; \citealt{Manzello2020}). When a fire reaches urban zones, it spreads through two primary mechanisms: (a) radiant and convective heat from adjacent structures; and (b) ember accumulation on flammable vegetation (e.g., lawns, mulch beds) or building components (e.g., leaf filled gutters, dry vents, porches, fences). In an urban setting, embers are the leading cause of home ignitions \cite{Manzello2020}.
When firebrands are not represented in fire behavior simulations, the predicted fire area can be considerably smaller than the observed. This is because the absence of spotting can underestimate the simulated fire rate of spread, while containment barriers depicted in the model fuels, such as streams and infrastructure, can prevent the simulated fire from spreading. \citealt{decastro2022a} showed that some fire areas can be significantly different depending on the location of the ignition relative to nearby barriers. In the case of the 416 Fire (Colorado, U.S.), when ignition points were placed west of US 550, the fire area was significantly reduced because the fire spread towards the east was contained by the road.
Fire behavior simulations are intrinsically dependent on the model’s fire fuels \citealt{Andrews2018,Ervilha2017}. In addition to accurate depiction of burnable fuels \cite{decastro2022b}, fuel barriers can also be misrepresented and affect simulation results. When it comes to fuel barriers (i.e., no fuel), accuracy is generally determined by the dataset spatial resolution combined with the spatial scale of containment features (water bodies, roads and other infrastructures). The continuity of fuel barriers depends on whether the dataset resolution can accurately represent the existing continuities. \citet{Frediani2023} showed that when barriers appear as discontinuities in the fuels dataset, the simulated fire is able to spread. However, when barriers of large spatial scale, such as multiple-lane highways, are accurately depicted as continuous, the simulated fire cannot spread without a spotting capability.
The approach to modeling spotting depends on the fire behavior model to which it is combined. Fire behavior processes are frequently parameterized because resolving combustion and chemical processes is computationally expensive. The fidelity of the fire behavior and associated spotting parameterization (if present) also weighs on the computational costs and consequently determine their suitability for certain applications. FIRETEC \cite{Linn_2002} and WFDS \cite{Mell_2007} are examples of models using more comprehensive representations of fire behavior processes; however, their computational requirements make them unsuitable for operations. Consequently, the models used in fast-computation applications are simplified parameterizations mainly based on the Rothermel parameterization or cellular automata. These models include FARSITE \cite{finney1998farsite}, which is now incorporated into FlamMap \cite{FlamMap}, Prometheus \cite{tymstra2010development}, BehavePlus \cite{Andrews2014}, and QUIC-Fire \cite{Linn2020}. These fire behavior models typically use atmospheric fields (winds, temperature, humidity) from numerical weather prediction models or wind simulators \cite{Sanjuan2014}, such that the fire responds to the atmosphere but the atmosphere does not respond to the fire \cite{Mell2018}.
Operational models typically used in fire management applications do not account for interactions between the fire and atmosphere, and their fire spread can be inconsistent with fire spreading characteristics \cite{ager2011integrating}. Conversely, models that couple atmosphere and fire (such as WRF-Fire, \citealt{coen2013}, WRF-SFire, \citealt{mandel2014recent}, FOREFIRE, \citealt{filippi2010}, and ARPS/DEVS-Fire, \citealt{xue2012}) have the ability to resolve winds in complex terrain and model fire induced phenomena in the atmosphere (e.g., \citealt{Roberts2023}), which becomes more relevant with increasing fire scale and threat to communities adjacent to the WUI.
In the current literature, the effects of firebrand spotting in fire behavior simulations can be modeled with more computationally expensive computational fluid dynamics (CFD) models, such as the spotting implementation by \citealt{koo2007using} in FIRETEC and the coupled stochastic parametric model of firebrand transport by \citealt{tohidi2017stochastic}, which provide high fidelity results comparable to wind-tunnel results. Yet, to be feasible for fire operation applications, firebrand spotting effects are vastly based on empirical models, typically based on spotting distance and probability of ignition (\citealt{mcarthur1967,tarifa1967,tarifa1965,albini1979,albini1983,ellis2015,woycheese1998}), or incorporated into a cellular automata fire-spread framework \cite{Alexandridis2011,Freire2019}. Firebrands from urban structures specifically can be modeled by the implementations of \citealt{himoto2021generalization} and the SWUIFT model by \citealt{masoudvaziri2021streamlined}.
The Firebrand Spotting parameterization described in this article was initially developed for the WRF-ARW model version starting at V4.4 \cite{Frediani2023,frediani2022,frediani2022fire}. The parameterization uses a Lagrangian particle transport framework to simulate firebrand advection, originating in active fire points determined by WRF-Fire’s fire behavior model, which is based on Rothermel’s rate of spread model \cite{rothermel1972mathematical} coupled with a level-set algorithm to propagate surface fire (Munoz-Esparza et al. 2018).
WRF-Fire is an extension of the Numerical Weather Prediction (NWP) community model WRF-ARW and serves a distinct role amongst fire behavior models for its broadscale use in weather research and forecasting applications. Fires affect vertical convection and interfere with cloud formation and precipitation \cite{Andreae2004,Bowman2009,frediani2008}. Thus, fire simulation capabilities are an important component for simulating regional weather, and a mesoscale weather model without it is incomplete. With climate trends indicating fires are becoming larger and more frequent \cite{mcginnis2023}, fire and weather cannot be treated as independent phenomena in local and regional scales. The next generation of NWP models is being built based on the successes of the current ones. Hence, building upon the existing strengths, overcoming limitations, and understanding the forecast skill of current coupled models are critical pieces to the success of the upcoming generation of NWP modeling systems.
The firebrand spotting parameterization in WRF-Fire provides a more realistic representation of the fire physics, as it allows fire spots to accelerate the fire spread and initiate fires across barriers. Its overarching goal is to converge research relevant to fire-weather at local and regional scales by addressing an important fire behavior modeling limitation. This work is a novel step towards enhancing the numerical weather and fire prediction in WRF-Fire, and is a stepping stone to future advancements and potentially operational applications.
In this article, we use idealized simulations to describe the parameterization implementation, and demonstrate its application with a simulation experiment of the Marshall Fire. The article is divided into five sections: (1) Introduction; (2) Implementation of the Firebrand Spotting Parameterization, where we describe the parameterization components and associated methods; (3) Idealized Simulations, where we demonstrate the parameterization components through various idealized numerical experiments; (4) Marshall Fire Simulations, where we apply the parameterization to forecast the fire spread and quantify its accuracy; and (5) Conclusions.
Implementation of the Firebrand Spotting Parameterization
The Firebrand Spotting parameterization was developed for WRF-Fire in WRF-ARW versions starting at V4.4, with fire spot ignitions implemented in the upcoming V4.6.x. This subsection describes the most recent parameterization implementation. The source code is written in FORTRAN 2003 following the WRF model coding standards. It consists of two distinct modules, one containing the main firebrand spotting code and associated routines, and another, for specific Message Passing Interface (MPI) routines that were not part of the WRF model. Both modules are included within the WRF physics source code. The firebrand processes modeled by the parameterization are firebrand generation, transport, burnout, landing and ignition. In this Lagrangian framework, firebrands are modeled individually and may ignite fire spots at landing when ignition conditions are met. The main aspects of the parameterization are illustrated in Figure 1 and described in detail in the following subsections.