This is an explanation of the saved program, not validated wildfire prediction or emergency guidance. Each candidate neighbor receives an ignition probability from wind, moisture, slope, direction, and local fuel. At the saved wind twenty two, moisture eight and slope fifteen, the base factor is point two nine zero four. For an illustrative fuel density point eight and local moisture eight percent, an aligned neighbor has multiplier two point eight and probability point seven two eight five five five five two. An opposite neighbor is floored at point two, giving point zero five two zero three nine six eight. The bars use five hundred pixels per probability unit: about three hundred sixty four point three and twenty six. A random draw below that threshold ignites the next state. These are individual attempt thresholds, not final burned fractions or measured forecast accuracy. The telemetry panel does not derive every number from its fire geometry. Burned area adds burning and burned cells, multiplies by point two five hectares, then rounds to one decimal. Forty cells give ten hectares and eighty give twenty. At ten pixels per hectare those bars span one hundred and two hundred pixels. Its displayed rate of spread instead comes from eight point five plus point two two times wind, plus point one two times slope, minus point one five times moisture. The saved inputs give thirteen point nine after rounding, distinct from the initial fourteen point two fixture. The so called perimeter overlap increases with ticks and gains point zero five whenever any firebreak or retardant cell exists, capped at point nine six. It is not a calculated intersection between predicted and observed perimeters. No neural network inference appears in this saved simulation engine. Containment is a source rule: no burning cells, or more than fifteen break cells and fewer than three burning cells. Thus an illustrative two burning cells with fifteen breaks fails the second branch; sixteen passes. Twenty pixels per break yields three hundred and three hundred twenty pixels. Retardant cells count as breaks in that test. Painting firebreak or retardant changes a three by three neighborhood, clipped at boundaries; ignition changes only the central cell. The export contains environment parameters and the canonical telemetry object, not the complete grid, random seed or ember trajectories. Offline this saved source depends on unavailable p five before later controls bind, so that behavior must be checked separately. The examples explain executable conditions, not safe containment, field acreage, operational forecasts or validated aerial guidance.