A Single Fire Perimeter Forecast Hides Uncertainty a Probability Map Shows
A deterministic run takes one set of inputs and draws one perimeter. It looks authoritative precisely because it shows no doubt, and readers treat the line as a boundary between burned and unburned. It is one sample from a range of futures. The alternative is to run the model many times across the plausible range of the inputs you are least sure of — wind speed and direction, ignition time and place, fuel moisture — and report, for each cell of the landscape, the fraction of runs in which it burned. The output stops being a line and becomes a field of values between zero and one. The rule for choosing: if the inputs are uncertain by more than the amount the decision cares about, an ensemble is the honest product; if they are not, a single run is fine and cheaper. The confusion resolved is that a probability map is not a vaguer version of a perimeter, it is a different and more complete answer, and a perimeter drawn from the ensemble mean is often a shape that no individual run produced. After this Concept you can decide whether a decision needs one perimeter or a probability field, and say what a lone perimeter is concealing.
This Concept is waiting for its first lesson!
A deterministic run takes one set of inputs and draws one perimeter. It looks authoritative precisely because it shows no doubt, and readers treat the line as a boundary between burned and unburned. It is one sample from a range of futures. The alternative is to run the model many times across the plausible range of the inputs you are least sure of — wind speed and direction, ignition time and place, fuel moisture — and report, for each cell of the landscape, the fraction of runs in which it burned. The output stops being a line and becomes a field of values between zero and one. The rule for choosing: if the inputs are uncertain by more than the amount the decision cares about, an ensemble is the honest product; if they are not, a single run is fine and cheaper. The confusion resolved is that a probability map is not a vaguer version of a perimeter, it is a different and more complete answer, and a perimeter drawn from the ensemble mean is often a shape that no individual run produced. After this Concept you can decide whether a decision needs one perimeter or a probability field, and say what a lone perimeter is concealing.
Are you a teacher? Sign in to start contributing.
Sign In