Large eddy simulation of atypical wildland fire spread on leeward slopes

The WRF-Fire coupled atmosphere–fire modelling system was used to investigate atypical wildland fire spread on steep leeward slopes through a series of idealised numerical simulations. The simulations are used to investigate both the leeward flow characteristics, such as flow separation, and the fire spread from an ignition region at the base of the leeward slope. The fire spreadwas considered under varying fuel type andwith atmosphere-fire coupling both enabled and disabled.When atmosphere–fire coupling is enabled and there is a high fuelmass density, the fire spread closely resembles that expected during fire channelling. Specifically, the fire spread is initially dominated by upslope spread to the mountain ridge line at an average rate of 2.0 kmh , followed by predominantly lateral spread close to the ridge line at a maximum rate of 3.6 kmh . The intermittent rapid lateral spread occurs when updraft–downdraft interfaces, which are associated with strongly circulating horizontal winds at the mid-flame height, move across the fire perimeter close to the ridge line. The updraft–downdraft interfaces are formed due to an interaction between the strong pyro-convection and the terrainmodified winds. Through these results, a new physical explanation of fire channelling is proposed. Received 10 May 2012, accepted 12 December 2012, published online 25 March 2013 Introduction A wildland fire is capable of exhibiting highly complex behaviour in response to multi-scale interactions between the fire and the local fire environment, namely the fuel, weather and topography. It has previously been identified that terrainmodified atmospheric conditions, particularly in complex terrain, can significantly affect fire spread and behaviour (Sharples 2009; Sharples et al. 2010a). This paper investigates a fire spread phenomenon referred to as ‘fire channelling’, in which atmosphere–terrain–fire interactions are believed to play an important role. McRae (2004) first noted this phenomenon, which he referred to as ‘lee-slope channelling’, through the presence of atypical fire spread patterns in multispectral linescan data from the Canberra 2003 bushfires. A study by Sharples et al. (2010b) investigated fire channelling at the laboratory scale, through a series of combustion tunnel experiments, and confirmed the incidence of atypical lateral fire spread across a leeward slope, apparently driven by an interaction between the wind, the terrain and the fire. Previous studies have identified several important distinguishing features of fire channelling (Sharples andMcRae 2011; Sharples et al. 2011, 2012). These features include rapid lateral fire spread across the top of a steep leeward slope in a direction approximately perpendicular to the synoptic wind conditions. The upwind edge of the lateral spread is constrained by a major break in topographic slope, such as a mountain ridge line. There is an extension of the active flaming zone downwind of the synoptic flow, possibly through a process such as spotting. Additional features include darker smoke and vigorous convection associated with the laterally advancing flanks of the fire. The rapid lateral fire spread indicates that fire channelling may pose a significant danger to fire fighter and civilian safety. Sharples et al. (2012) have previously determined that several environmental conditions are necessary for fire channelling. The leeward slope angle of the mountain should be greater than ,258 and the topographic aspect and synoptic wind direction should be within ,408 of each other. The synoptic wind speed should also be greater than ,25–30 kmh , which allows for flow separation in the lee of the mountain. This study postulated that fire channelling occurs due to an interaction between the fire and a lee rotor, which is formed through leeward flow separation. It was further conjectured that the lateral fire spread is driven by thermal expansion of the air within the lee rotor as heat is added to it from the fire, with the resulting lateral atmospheric flow effectively following a path of least resistance. CSIRO PUBLISHING International Journal of Wildland Fire 2013, 22, 599–614 http://dx.doi.org/10.1071/WF12072 Journal compilation IAWF 2013 www.publish.csiro.au/journals/ijwf An important step in this study is to use a numerical weather prediction model to investigate the nature of the atmospheric flow in the lee of a mountain. Several studies have previously investigated turbulent flow for environmental conditions and atmospheric scales similar to that considered in this paper (Schär and Durran 1997; Allen and Brown 2002; Doyle and Durran 2002, 2007; Ding et al. 2003; Pathirana et al. 2003; Hertenstein and Kuettner 2005; Sheridan and Vosper 2005; Ayotte 2008; Katurji et al. 2011). Many of these studies used a large eddy simulation model to investigate the nature of the atmospheric flow. Based on the results of these studies, it is evident that the nature of the leeward atmospheric flow is closely associated with many environmental conditions, including the atmospheric stability, surface roughness, upstream wind conditions and the geometric properties of the terrain. Another important step in this study is to simulate the fire spread across the leeward slope of a mountain using a coupled atmosphere–fire model. Over the past two decades, several studies have used coupled atmosphere–fire modelling to investigate fire spread, fire behaviour and atmosphere–fire interactions. Previous studies focussing on fire spread across flat terrain have been able to reproduce several physically realistic fire spread features (Heilman and Fast 1992; Clark et al. 1996a, 1996b, 2004; Cunningham et al. 2005; Cunningham and Linn 2007;Mell et al. 2007; Sun et al. 2009). Thesemodelled features include the parabolic fire shape that typically develops under the influence of light uniform winds. Previous studies focussing on fire spread across non-flat terrain have similarly yielded useful results (Heilman 1992; Linn et al. 2002, 2007; Clark et al. 2004; Coen 2005). The aim of this paper is to perform a series of idealised numerical simulations that allow for an evaluation of the fire channelling hypothesis forwarded by Sharples et al. (2012) and facilitate a detailed analysis of the physical mechanisms responsible for driving the lateral fire spread associated with fire channelling. The numerical modelling system used to perform these numerical simulations is described in the next section. The results of the two(2-D) and three-dimensional (3-D) atmospheric simulations of flow over a mountain are presented in the following two sections. The results of the simulations of fire spread across the leeward slope of a mountain are presented in the subsequent section. A summary of the study is presented in conjunction with several conclusions in the final section. Numerical models Atmospheric model The atmospheric model used in this study is version 3.3 of the Weather Research and Forecasting model (WRF) (Skamarock et al. 2008). It is used in a highly idealised large eddy simulation (LES) configuration that is well suited to studying turbulent atmospheric flow on length scales of tens to hundreds of metres. The model explicitly resolves the large-scale atmospheric eddies, whereas the effects of subgrid-scale motions on the resolved turbulence are modelled using a subfilter-scale stress model. The model uses fully compressible nonhydrostatic equations with a mass-based terrain-following coordinate system. The model is used in both a 2-D and 3-D configuration. Themodel domain is configured to capture the turbulent flow in the lee of a mountain at high resolution. In both the 2-D and 3-D simulations, the west–east dimension (x-axis) has an extent of 30 km. In the 2-D simulations, the south–north dimension (y-axis) is three grid points wide with cyclic boundary conditions, whereas in the 3-D simulations it has an extent of 5 km. The horizontal grid spacing in both the 2-D and 3-D simulations is 50m. The model top in each simulation is initially set to 10 km, with an initial vertical grid spacing of 50m. However, due to the use of terrain-following sigma coordinates, the model top, and therefore also the vertical grid spacing, varies in time with the atmospheric pressure. However, the model top descends no more than a few hundred metres in any simulation and the vertical grid spacing is therefore always in the range of 47–50m. The model grid cells are therefore approximately isotropic throughout the duration of each simulation. The atmospheric model has a computational domain of 600 3 200 (x,y,z) and 600 100 200 (x,y,z) grid points in the respective 2-D and 3-D simulations. This model domain setup is similar to that considered in previous WRF LES studies (Mirocha et al. 2010; Kirkil et al. 2012). The lateral boundary conditions are specified using a onedimensional input sounding. The surface pressure is 1000 hPa and the surface moisture mixing ratio is zero. The vertical profiles of the vapour mixing ratio, horizontal wind conditions and potential temperature are also specified. The vapour mixing ratio and y-axis wind velocity are zero at all heights. The horizontal wind conditions at the western lateral boundary are set as a temporally uniform incomingwesterly wind field, which varies with height according to: UðzÞ 1⁄4 20 z 200 2 z 200