Complex systems don’t defy predictability – it’s embedded in the patterns our simple models can’t see
Imagine a wildfire spreading through a forest -flames spreading, intensity building. As evening approaches and humidity rises, the fire shifts. Flames give way to smoldering. Spread slows, sometimes stops. But the next morning, as the sun heats things up, the fire roars back to life.
Predicting these transitions is critical for forecasting where a multi-day fire will spread. Yet our most widely used fire prediction models can’t do it. The models assume fire spreads at a steady rate, while observations show otherwise.
To forecast multi-day fires, managers must arbitrarily specify when they think the fire will actually be active. As one fire engineer puts it: “What managers do now is kind of just make a guess.”
The problem isn’t that fires behave unpredictably. The problem lies in the assumptions that drive our models; assumptions that shape both predictions and solutions.
The model we’ve relied on for 50+ years was built by burning experimental fires under controlled conditions—measuring wind speed, fuel, slope—then recording how fast fire spread.
The model we’ve relied on for 50+ years was built by measuring how fast fire spreads controlled experimental conditions.
It works well for predicting fire behavior under those controlled conditions.
But real wildfires don’t work that way. They behave as dynamic systems. Its patterns are emergent – arising from interactions that can’t be predicted based on any single factor. We can’t know what the fire will do by measuring wind speed OR fuel load OR topography; we need to understand how they’re all influencing each other right now, in thislandscape, under these conditions, creating this particular fire.
Fire behavior depends not just on current external conditions, but on the fire’s own recent history; what it was just doing shapes what it does next.
And once crisis unfolds, understanding the system’s actual dynamics becomes the difference between resolution and escalation.
The model can’t capture this because it wasn’t built to. It measures inputs and outputs without accounting for the physical processes—heat transfer, fluid dynamics, chemical reactions—that actually cause fire to behave the way it does.
A pioneer team of fire scientists and engineers are now working together to develop models that treat fire as a dynamic system, capturing the transitions and behaviors that black-box approaches miss.
But for decades, we’ve worked around model limitations instead of questioning the assumptions the models were built on.
The fire doesn’t need us to control it. It needs us to understand it well enough to respond skillfully to what it’s actually doing.
The solution isn’t forcing simple models onto complex reality, but learning to read the patterns that complexity reveals.