In 2023, Canada lost over 18 million hectares of forest to wildfire. That's an area roughly the size of North Dakota, burned in a single season. Smoke from those fires blanketed the eastern United States for weeks, turning skies orange from New York to Atlanta. It was the kind of event that makes people ask: is something different happening?
The answer is yes. But the fire isn't really the story. The story is what made the fire possible, and why we didn't see it coming.
The map we're still using
Most people, including many who manage forests professionally, still carry a mental model of forests that was built in the 20th century. It goes something like this: forests are stable. Trees grow. Sometimes there's a fire or a pest outbreak, and then the forest recovers. Our job is to suppress the fire, spray the pest, and get back to baseline.
This model was never quite right, but for a long time it was close enough. The climate was relatively stable. Disturbances were episodic and localized. Recovery timelines were measured in decades, which felt manageable.
That world is gone.
When stresses compound
Today's forests face something qualitatively different from historical disturbance patterns. It's not just that individual stresses are getting worse. It's that they're compounding.
A forest weakened by three consecutive years of drought doesn't just suffer reduced growth. It becomes vulnerable to bark beetle infestation. Beetle-killed trees become fuel. When fire arrives (and in a hotter, drier climate, fire always arrives) it burns hotter and more completely than it would have in a healthy stand. The soil is sterilized. The seed bank is destroyed. The species that once regenerated after fire can't establish in the new climate conditions. What comes back, if anything comes back, isn't the same forest.
This isn't a disturbance-and-recovery cycle. It's a state change. And we're seeing it play out across the American West, the boreal forests of Canada, the Mediterranean, and increasingly in the southeastern United States.
If you live in a city that depends on mountain snowpack filtered through forested watersheds, this is your problem. Hundreds of millions of people do. When those forests shift state, the watershed chemistry changes, filtration capacity drops, and water infrastructure built around historical flow patterns stops working the way it was designed to. Forest health is water security, and most water utilities aren't tracking it.
Forests die in slow motion
Here's what makes this especially dangerous: forests are slow. A tree stressed today might not die for three to five years. A forest transitioning from carbon sink to carbon source doesn't announce it with a press release. The canopy stays green long after the system underneath has crossed a threshold.
This means our feedback loops are broken. By the time mortality is visible, by the time the satellite image or the aerial survey shows red and gray where green used to be, the damage was done years ago. We're always looking at the past and calling it the present.
The world's forests absorb roughly 30% of annual CO₂ emissions. That number is treated as stable in most climate accounting. It isn't. As forests become stressed, that absorption declines. In some regions, forests have already flipped from net carbon sinks to net carbon sources, accelerating the very climate change that's driving the stress. But because the flip happens slowly, and because the canopy stays green while it's happening, it doesn't show up in the models until after the fact.
The same lag shows up everywhere you look. Insurance markets price wildfire risk using historical fire data, but historical data doesn't capture the fuel loading happening right now in stands that haven't burned yet. Power utilities plan vegetation management on fixed schedules, but tree mortality doesn't follow schedules. Rural communities watch their local economies erode as forest productivity declines, and by the time anyone notices, the window for intervention has passed.
Forest management decisions made today won't show meaningful results for 20 to 30 years. We're steering a ship with a decades-long lag between the wheel and the rudder, in waters that are changing faster than our charts can track.
What forecasting could change
We forecast weather seven days out and nobody thinks twice about it. We don't forecast forest health at all. There's no operational system that tells a forest manager: this stand is accumulating stress in a pattern that historically precedes beetle outbreak, and here's your window to intervene.
We have better tools than ever to build one. Satellite imagery at multiple spectral bands can detect physiological stress before it's visible to the eye. LiDAR reveals structural changes in the canopy. Machine learning can find patterns in decades of climate and disturbance data that no human analyst could track across millions of acres. The raw capability exists to move from reactive management to something that looks more like operational forecasting.
Building that kind of system is technically possible. It's not easy, because ecological systems are far more complex than atmospheric dynamics in some important ways. But it's possible. The question is whether we'll build it before the forests we're trying to protect have already crossed thresholds we can't reverse.
What this series is about
Over the coming weeks, I'm going to write about forest health: what it means, how it's measured, what threatens it, and what we can do about it. I'll cover the science, the technology, and the human dimensions.
My perspective comes from building systems that try to forecast ecological disturbance, combining physics-based models with machine learning to see what's coming before it's visible. I'll share what I've learned, what I've gotten wrong, and what I think the path forward looks like.
Let's start paying attention before the canopy turns gray.