This Fire! activity is a simple computer model that allows the user to investigate the spread of a fire through a forest of trees.  The propagation of the fire is affected by a single variable - the probability that a given tree will ignite, given that it's neighbor is on fire.

The model progresses through a sequence of steps, or iterations.  Each iteration, every green tree follows the same set of rules:

Look at the neighbor above, has that tree started burning?
        No: Stay green!
        Yes: Pick a random number between 0 and 1.  Is it less than the burn probability?
            No: Stay green!
            Yes: Ignite!
Look at the neighbor below, has that tree started burning?
        No: Stay green!
        Yes: Pick a random number between 0 and 1.  Is it less than the burn probability?
            No: Stay green!
            Yes: Ignite!
Look at the neighbor to the left, has that tree started burning?
        No: Stay green!
        Yes: Pick a random number between 0 and 1.  Is it less than the burn probability?
            No: Stay green!
            Yes: Ignite!
Look at the neighbor to the right, has that tree started burning?
        No: Stay green!
        Yes: Pick a random number between 0 and 1.  Is it less than the burn probability?
            No: Stay green!
            Yes: Ignite!

Once a tree has ignited, it stops checking it's neighbors and burns merrily until it is a charred trunk and branches.

Fire! is an example of a stochastic model - it uses a random variable to represent uncertain behaviors.  What type of things might this probability represent in reality? What 'real' factors would contribute to a high burn probability?  A low burn probability?

All computer models make some sort of assumptions and simplifications that limit how well the model can represent reality.  What aspects of Fire! are less than realistic?

Even given the simplifications, the behavior of the Fire! activity is chaotic - we cannot predict in advance whether a particular tree is burned or not, and small changes in the probability and the location of the initial burning can radically alter the final state of the forest. However, general patterns in the burning of the trees can be seen for particular probabilities. Even with this simple model we can generate a fairly large amount of data, and analyze the data to learn more about how this model behaves.

Developed by Dr. Garrett R. Love for
The Shodor Education Foundation, Inc.
Copyright © 2003
Last modified: Monday, 09-Jun-2003 07:56:44 EDT
Questions or comments about this page should be directed to glove@shodor.org