The effect of a canopy on a buoyant plume
1. Background
In Australia one of the most common natural hazards is bushfires. Small fires, often
started by lightening strikes, burning in largely inaccessible bushland can quickly spread
rapidly due to strong winds and impact on properties and towns. Predicting the behaviour of
these fires has long been based on experimental and observational results. With advances in
computational technology over the last decade it is becoming possible to simulate the
behaviours of bushfires from first principles.These simulations give great insight into the
mechanisms and rate of bushfire spread.
Large Eddy Simulation (LES) has in recent decades emerged as the tool of choice for
simulating flows in engineering applications and the atmospheric boundary layer. More
recently, LES has been applied to modelling wildfires (eg. Mell et al. [2007], or Cunningham
et al. [2005]). Cunningham et al. [2005] studied the behaviour of a wildfire plume as the
shearing wind velocity increased.Cunningham et al. [2005] the forest canopy by a uniform
aerodynamic drag term applied only in the first grid cell of the simulation.
Real forest canopies are often characterised by leaf-area index (LAI), that is, fraction
occupied by foliage in a given frontal area. In real forests the LAI varies in space. The drag
term used torepresent the canopy is
FD = CDA(z)|u||u|
CD is a constant drag (measured to be ~0.15) coefficient, u is the wind velocity, and A(z) is
theoccupied frontal area, ie. the leaf area index. Several measurements of A(z) exist in the
literature(eg. Amiro [1990]). The magnitude of A(z) is of order 3 and a commonly adopted
model of the profile is a Gaussian.
During a fire, burning embers can become entrained in the plume, transported several
kilometres, and start a new spotfire or impact on houses and infrastructure. Thurston et al.
[2014]studied the transport of embers by a plume and the distribution of where the embers
fell. They found the final distribution was significantly modified by the properties of the in-
plume turbulence.
2. Proposed study
For the sake of simplicity and computational tractability, the plumes in Cunningham et al.
[2005]and Thurston et al. [2014] are modelled as heat sources using the Boussinesq
approximation. We will also adopt this simplification. The magnitude of the fire is
characterised by the intensity ofthe heat source, denoted Q and measured in W/m 3. In this
study we will _x the intensity to1.7 x109 W/m3, similar to Cunningham et al. [2005] to
represent a moderate wildfire.
We wish to examine the effect of canopy properties, such as the functional form of A
(z), andthe magnitude max(A(z)), on the plume dynamics, in particular on the in-plume
turbulence. Wewill conduct simulations using the open source package Fire Dynamics
Simulator developed byNIST [McGrattan et al., 2013].
We will conduct a parametric study. Initially we will not consider the effect of a
shearing wind.The anticipated simulations are detailed in Table 1. We anticipate the
simulations will be conductedin a domain of physical size Lx = 1500m x Ly = 1500m x Lz =
1500m, with a resolution ofnx = 150 x ny = 150 x nz= 150, although the resolution and
convergence will have to be verified.
We will compare the change in quantities such as the mean velocities and
temperatures andtheir fluctuations across the plume as the canopy properties vary. We expect
the simulations willbe within reach of a standard workstation.
