Characteristics and Short-Range Transport of Firebrands Generated by a Pile Fire

One major risk from wildfires is fire spotting–embers transported some distance by the wind starting a secondary fire elsewhere. Due to the complexity of the coupled dynamics between atmosphere, turbulent-buoyant plume and the embers themselves, fire spotting is highly stochastic. While transport models exist, they are often simplified, e.g. by assuming spherical or regular cylindrical shapes for firebrands or by only considering two-dimensional or steady-state representations of the fire plume and atmospheric flow. CFD has been an invaluable tool for investigating fire spotting, but there is a relative lack of experimental data to back up both the results of simulations and the assumptions used in simplified models. To this end, we present an experiment using Particle Tracking Velocimetry (PTV) on a hand-drawn pile fire to determine velocity and acceleration statistics of embers near the source flame. We collected settling embers around the pile fire to assess shape, size and ember concentration downwind of the experiment.

We performed our experiment at the University of California, Berkeley Blodgett Research Forest in the foothills of the Sierra Nevada mountains of California. One of the many techniques used by forest managers to thin fuels are pile burns. In this preliminary study, a pile fire provided a convenient, controlled setup for our experiment of ember transport within a forest canopy. The pile was drawn from a thinning of small trees, ponderosa pine (Pinus ponderosa) and Douglas fir (Pseudotsuga menziesii) with approximately 75/25 split between the two. The fuel was mostly fine with some coarse woody fuels. Fine fuels included 0.6–8 cm diameter pieces. 12-18 cm diameter fuels were placed on top of the piles. Prior
to the burn, we placed aluminum trays filled with water radially around the pile from 1.5 to 10m away from the pile edge and at spacing of 1 m from each other. The trays collected embers as they settled, allowing for characterization of size, shape, density and relative concentration with distance from the fire. We burned the pile at night to better visualize the glowing embers and captured images with a GoPro Hero 9, 5m from the pile, recording at 1080p and 120 Hz. To track embers between frames, we use a PTV algorithm based off the 4th-best-estimate method to obtain time-resolved ember positions. Particle velocities and accelerations are determined from the particle position between successive frames using convolution with the first and second derivatives of a Gaussian kernel. These measurements are used to estimate quantities like the mean ember ejection velocity from the pile fire as well as r.m.s. values. We use image segmentation to measure flame properties over the course of the experiment, like flame height, fractality and shape to investigate the relationships between the flame itself and the firebrand ejection velocities, directions and number density. To characterize ember size and shape, we first dilute our collected samples from the tray so individuals can be identified under a microscope. Different populations of embers can be observed, ranging from micro to sub-millimeter and a smaller number of larger ones. Also using image segmentation, we identify individual firebrands and calculate their 2-D area and aspect ratio. We record enough samples to obtain statistically converged distributions of 2-D ember size for each population. We also use a pycnometer which allows us to accurately measure the volume of a sample of embers, which is then combined with measurements of the sample mass to derive a bulk ember density.

The results of this experimental study will be valuable for validating and benchmarking computational models of firebrand origin and transport.