Northwest Territories Crown Fire Experiment and Infrared Analysis

In collaboration with the U.S. and Canadian Forest Services, Terry Clark and Larry Radke (ATD) participated in the Northwest Territories Crown Fire Experiment during June and July. Three controlled burns took place during the experiment. The burn of Plot 6 on 9 June was particularly successful, as wind conditions, observing position and data recording were all near optimal. High time and spatial resolution IR data were obtained by NCAR using an Inframetrics Thermacam digital IR camera mounted on a 50 ft U.S.F.S. tower. Grayscale data was recorded on a SVHS/NTSC videotape for post-processing.

Subsequent to the experiment Don Middleton (SCD) converted the analog SVHS data to one-byte binary files for digital image processing. Preliminary analysis by Clark and colleagues, during his visit to Monash University, shows that meaningful winds and sensible heat fluxes can be derived using the binary data. Using ideas taken from image processing literature, it was found that assuming gradient flow produced the most realistic results. This assumption is equivalent to assuming the convective motions in the fire move in the direction of the strongest temperature gradients, which is perhaps similar to what one would expect for motion resulting from an explosion. Assuming structures move with the flow as in, say, TREC analysis, showed no particular promise for these data. The data to date show updraft speeds of 50 to 60 m/s. Collaborations with scientist at the Canadian Forestry Service (CAFS) at Sault Ste. Marie in Ontario Canada were also established. The CAFS also obtained IR data of the Plot 6 burn using an IR line scanner. Clark and colleagues agreed to process their data in a similar fashion to extract winds from their side-viewing data source.

Development of a New Fire Model

In work to improve the coupled atmosphere-fire model, Clark and Janice Coen developed and applied an improved fire model that uses four Lagrangian tracers per fuel cell to define the geometry of a burning region. The motion of these tracers depends upon the speed and direction of the local wind as well as on the fire line shape. Using the time, the ignition time, the burnout time of fuels, and the mass of unburned fuels remaining in the fire area of the fuel grid, the model estimates the sensible and latent heat fluxes for the grid. This scheme has numerous advantages over the previous fire model, including: 1) varying portions of any fuel cell can be burning at any time; 2) patches of unburned forest can remain after a fire's passage, leaving streaks as in some observed fires; 3) the model allows ground fires without canopy fires and vice versa; and 4) spotting (i.e., ignition of new fires by lofted burning embers) can be treated, provided that we adequately represent firebrands.

This development was assisted by collaboration with Paul Ginoux (ACD) who simulated small grass fires, where the fire spread rate is quite small, the fuel load relatively light, and the fire line very narrow. This work tested the model in a different regime of fire behavior than much of the previous work examined, and greatly improved the model and confidence in its ability to represent a wide range of fire behavior.

Sensitivity Test on Shear and Slope

Coen and Clark began sensitivity tests with the new fire model to examine the effect of environmental conditions on fire line dynamics. Fire managers' experience suggests that negative vertical wind shear often accompanies blowup fires (i.e., fires that rapidly increase their energy output in a short time and are accompanied by extreme, unpredictable behavior such as fire whirls and spotting by lofted burning embers). So, in an extension of their previous work, Coen and Clark began to try to understand the degree to which negative vertical wind shear, perhaps from nocturnal drainage winds, microburst outflows, or gust fronts, might lead to fire blowups. This work showed that strong negative wind shear can have a dramatic effect on wildfire behavior, producing strong fire whirls, rapid intensification of dynamics at the fire line, erratic spread rates, and strong updrafts capable of lofting burning embers away from the primary fire line. Their simulations also suggest that a fire might modify the winds in its environment so much that this modified local environment drives the fire behavior.

In related experiments, Coen and Clark examined how sloped terrain affects the fire spread rates, and compared their results to fire spread rates and the fire intensity predicted from BEHAVE, an empirically-derived algorithm widely used by fire managers. Slope changes the propagation speed by exposing the fuel ahead of the fire to additional convective heating and radiation, so that fires burn more rapidly uphill; the steeper the slope, the faster the fire spreads. Experiments showed that the simulated and empirically-predicted intensities of the fire compared extremely well, and the modeled fire spread rates were within a range of predicted fire spread rates that depended how one characterized the fuel. Since the numerical model considers primarily only convective heat transfer and a very simple radiation treatment, the increase in spread rate over the hills is not completely understood. (Also see the visualizations done by Don Middleton on the SCD Visualization Gallery web page, including A Numerical Model of a Forest Fire)

NCAR | UCAR | NSF | NCAR FY96 ASR | NCAR FY97 ASR