A broad-scale probabilistic model of forest fires, EMBYR, has been develope
d to simulate the effects of large fires burning through heterogeneous land
scapes. Fire ignition and spread are simulated on a gridded landscape by (1
) examining each burning site at each time step, (2) independently evaluati
ng the probability of spread to eight neighbors based on fuel type, fuel mo
isture, wind speed and direction, and (3) distributing firebrands to downwi
nd sites, where the probability of ignition of new fires is a function of f
uel type and moisture conditions. Low values for the probability of spread,
I, produce a dendritic burn pattern resembling a slow, meandering fire, wh
ereas higher values of I produce solid patterns similar to a rapidly moving
, intensely burning fire. I had to be greater than a critical value, i(c),
estimated to lie between 0.250 and 0.251, to have a 50% chance of propagati
ng across the landscape by adjacent spread alone. The rate of spread of fir
e at I = 0.30 was nearly four times faster when firebrands were included in
the simulations, and nearly eight times faster in the presence of moderate
wind. Given the importance of firebrands in projecting fire spread, there
is a need for better empirical information on fire spotting. A set of model
parameters was developed to represent the weather conditions and fuel type
s on the subalpine plateau of Yellowstone National Park, WY, USA. Simulatio
n experiments were performed to reveal relationships between fire and lands
cape-scale heterogeneity of fuels. In addition, EMBYR was used to explore f
ire patterns in the subalpine plateau by simulating four scenarios of weath
er and fuel conditions. The results of repeated simulations were compared b
y evaluating risk (the cumulative frequency distribution of the area burned
) as a function of the change in weather conditions. Estimates of risk summ
arized the high degree of variability experienced in natural systems, the d
ifficulty of predicting fire behavior when conditions are near critical thr
esholds, a quantification of uncertainties concerning future weather condit
ions, and useful tool for assessing potential wildfire effects. (C) 2000 El
sevier Science B.V. All rights reserved.