<p>A critical research area regarding wildfire modelling often overlooked is the task of finding where a wildfire started and how long that wildfire burned. A literature review revealed that there are no automated methods with the goal of estimating the location of ignition points and propagation duration of a wildfire from a burn scar. This paper describes a novel method called the Wildfire Source Genetic Algorithm (WSGA) which can estimate the ignition points and the propagation duration of a wildfire, given the wildfire’s burn scar and environmental conditions used as input for a forward running wildfire simulator. This paper uses twenty input burn scars generated from a wildfire simulator to validate the WSGA, as their ignition points and propagation durations are known. The WSGA generates sets of ignition points and propagation durations, which are then simulated and compared to the input burn scar. The ignition points and propagation durations of the WSGA seeded burn scars that most closely resemble the input burn scar are iteratively modified using a genetic algorithm to seed wildfires that more closely resemble the input burn scar. The ignition points and propagation duration of the best fitting WSGA seeded burn scar is compared to the inputted burn scar by evaluating two measures of error developed in this paper called the relative distance error and relative simulation duration error. The WSGA had a relative distance error of 0 to 1.25 times the diameter of the inputted burn scar. Lower errors were associated with larger wildfires. The WSGA had a relative simulation duration error of 0.0006 to 0.49 times the propagation duration of the input burn scar.</p>