Fire Model Fidelity Is Bought With Run Time, From Milliseconds to Days
Fire behaviour models sit on a ladder of cost. A surface-fire spread calculation returns a rate of spread for one point in well under a second on a laptop. A landscape simulator growing a perimeter across terrain with hourly weather takes seconds to minutes. A coupled fire-atmosphere simulation, which lets the fire's own plume change the wind that drives it, takes hours on many processors, and a computational fluid dynamics model of a room fire can take days. Each step up buys physics the step below assumed away, and each step up costs roughly an order of magnitude more computing. The rule that makes this usable: write the deadline down before choosing the model. A six-hour run has a fidelity of zero for a decision that must be made in twenty minutes, and a model whose extra physics needs inputs nobody measured on this fire has bought you nothing either. The confusion this resolves is the assumption that more physics means a more accurate answer; it means a more accurate answer to a question you may not be able to ask in time or feed properly. After this Concept you can place a named fire model on that cost ladder and say whether its run time fits the decision in front of you.
This Concept is waiting for its first lesson!
Fire behaviour models sit on a ladder of cost. A surface-fire spread calculation returns a rate of spread for one point in well under a second on a laptop. A landscape simulator growing a perimeter across terrain with hourly weather takes seconds to minutes. A coupled fire-atmosphere simulation, which lets the fire's own plume change the wind that drives it, takes hours on many processors, and a computational fluid dynamics model of a room fire can take days. Each step up buys physics the step below assumed away, and each step up costs roughly an order of magnitude more computing. The rule that makes this usable: write the deadline down before choosing the model. A six-hour run has a fidelity of zero for a decision that must be made in twenty minutes, and a model whose extra physics needs inputs nobody measured on this fire has bought you nothing either. The confusion this resolves is the assumption that more physics means a more accurate answer; it means a more accurate answer to a question you may not be able to ask in time or feed properly. After this Concept you can place a named fire model on that cost ladder and say whether its run time fits the decision in front of you.
Are you a teacher? Sign in to start contributing.
Sign In