On a slice of 1,392-year-old redwood trunk at Big Basin Redwoods State Park, generations of children traced tree rings through time. Small fingers passed markers for the Aztecs (1300s), the Magna Carta (1215), the Mayans (600s), then paused at the center: “544—Tree Sprouted, Byzantine Empire (Emperor Justinian).” After the August 2020 fires, those markers were found in a pile of ash.
The CZU Lightning Complex fires burned hot. Douglas firs burned to a crisp, but an estimated 90 to 98 percent of redwoods survived, according to Sempervirens Fund, which helped establish Big Basin back in 1902. A year later, with fire season already underway, Big Basin is deciding how to build back. Nearby, Sempervirens Fund is doubling down on redwoods.
In February 2021, armed with a shovel, a hoedad, and Ikea bags full of 6-inch redwood seedlings, land steward Ian Rowbotham hiked through charred hills near Big Basin. He gazed up through dead branches. He later told me that he thought to himself, “I’ll definitely need more sunscreen this summer.” He planted carefully, thinking ahead 200 years to design a less crowded, more drought-tolerant forest.
But two centuries? Redwoods can live two millennia. Will this forest still be here in 4021 CE? Can Rowbotham’s seedlings become the ancient giants of the future? The short answer is probably not. The long answer pushes the boundaries of today’s best climate models—and it gets weird fast.
“We are moving into conditions we don’t have any analog for,” says Anthony Ambrose, who studies ancient trees with the Marmot Society. He considered Rowbotham’s seedlings. “I wouldn’t be surprised if they were completely wiped out, except maybe for pockets in drainages and on north-facing slopes,” he told me, thinking ahead 2,000 years.
Redwood tree rings contain human history; they may also hint at the trees’ own future. Miguel Fernandez, a scientist at biodiversity nonprofit NatureServe, recently used past climate to predict how redwoods may shift next. Eventually, he told me, data from tree rings could help clarify those predictions further.*
Millions of years ago, redwoods circled the globe. By the time humans appeared, they had settled into their current range: a temperate, foggy stretch of coast from southern Oregon to central California. Remarkably resilient, redwoods mostly don’t mind heat, fire, and pests. But they do need lots of moisture.
In a hotter, dryer California, Fernandez’s model shows redwoods losing almost 80 percent of their land. In the short term, hot and dry is a safe bet. But long-term, Northern California’s climate is harder to predict. More humidity, and the range could expand.
With long life spans and a narrow habitat, redwoods occupy something of a climate modeling blind spot. Even today’s fastest supercomputers are too slow to simulate 2,000 years at a relevant scale.
That’s not to say today’s climate models aren’t impressive. The August 2021 report from the U.N.’s Intergovernmental Panel on Climate Change summarized what these models can now tell us with certainty: Humans are causing climate change, now. We’ve already locked in some changes for decades and even millennia. To stabilize, we need to reduce net carbon emissions to zero, preferably in a few decades.
To reach these conclusions—and get a 20,000-foot view of how the planet will respond—researchers build simulated worlds, set starting assumptions, like a “keep it in the ground” approach to oil and gas versus extensive new drilling, and press play.
Models account for sun glinting off the tops of clouds, pollen billowing from prehistoric forests, and the wobble of the earth as it spins. Each uses slightly different math to represent processes on land, ice, sea, and sky, so modelers compare with other models, recent weather data, and the deep past to check their work.
The most robust of these simulated worlds are called general circulation models, or GCMs. They can be 60 layers deep from the top of the atmosphere to the ocean’s inky depths. Complex equations represent processes within and between layers, connecting melting glaciers to ocean currents, car exhaust to the behavior of clouds.
For the first time, the IPCC released an interactive atlas so nonscientists can explore the futures these models predict. If you obsessed over “flatten the curve” graphics early in the pandemic, this may be a fun, soothing way to prepare for the apocalypse.
Try looking closer, though, and you will be disappointed. GCMs still run at Pac-Man-scale resolution; zooming in worldwide won’t be possible for decades. “For any factor-of-10 increase in resolution, you need to factor in a 10,000 percent increase in computer performance,” says Tapio Schneider, a climate modeler at Caltech. “That, ultimately, is the killer.” Most GCM “pixels” are 60 miles wide at best. The redwoods’ range is only 30 miles wide—so in these models, it vanishes. California’s mountains flatten; the Coast Ranges, responsible for the redwoods’ current micro-climate, can be 68 and breezy on the Pacific side and a bone-dry 110 inland. GCMs average those extremes.
Mountains aren’t the only things that disappear. So do rainstorms and eddies, as well as cities and even some countries—Monaco, the Maldives—frustrating policymakers.
To help, regional modelers are building zoomed-in maps one “pixel” at a time. They start with big-picture assumptions from a GCM, add local weather and topography a pixel at a time, then stitch continents back together from the bottom up. In recent years, regional modelers reached a 1-kilometer resolution, bringing California into clear focus.
But while regional climate models give more detail, they do so with less certainty. GCMs aren’t perfect, and regional models inherit their biases. Local data can also be incomplete: We have more weather stations in the foothills than on mountain peaks, for example. The net result is that regional models are not entirely reliable; two regional climate models in the same location could predict opposite trends for rainfall. While they can help policymakers consider different scenarios, they are not designed to explore millennium-scale changes.
Simulating the life span of a redwood on a standard GCM takes millions of hours of computing time. So far, kilometer-scale regional models have only simulated decades; kilometer-scale GCM prototypes can only simulate a few months.
Instead, to explore the deep past or far future, modelers turn to Earth system models of intermediate complexity, or EMICs. These minimalist models simplify to run fast, using a 2D atmosphere or ignoring vegetation completely.
At Oregon State University, Peter Clark recently used an EMIC to explore sea level rise over the next 10,000 years. California is planning for 3.5 feet of sea level rise by 2050. By 4021, Clark predicts seas 100 to 400 feet higher, depending on how much we lower emissions.
“It could be worse, it could be better,” Clark says. “That’s the main point in the next few decades. It’s up to us.” In either scenario, the takeaway for redwoods is similar: Coasts erode. Fertile riverbeds flood. Redwoods retreat uphill, halting at the snow line.
In the worst case, chunks of California really do fall off the map; the Central Valley becomes an inland sea. For the redwoods, though, the big difference between “do our best” and “emissions as usual” is not just about flooding, but about fog—and maybe an ice age.
The Pacific Ocean’s iconic fog rolls into coastal valleys like clockwork, giving many redwoods much of their moisture. Redwoods trap fog in a vast net of leaves—more than any other species—hydrating themselves and the ecosystems below.
In addition to watering redwoods, “marine layer” fog shades the earth. Losing it could increase global temperatures 8 to 10 degrees Celsius. At Caltech, Schneider found that somewhere between our best mitigation efforts and the nightmare scenarios, coastal fog could disappear completely.
“That would be terrible,” Schneider said. “Let’s not get there.” He doesn’t think we will; the tipping point would require around three times our current atmospheric CO2.
Still, it’s a good reminder that modeling wild cards remain. One big source of uncertainty is clouds—how much different types of clouds amplify warming and how cloud cover might change in a changing climate, to be specific. In fact, cloud uncertainties are shaking our most reliable climate orthodoxy: the relationship between CO2 and temperature, which has seemed stable since the 1970s. That ratio helps scientists calculate how much carbon we can burn without overshooting Paris Agreement goals.
This year’s IPCC report is more confident in the ratio than ever: It guesses 3 degrees warming every time atmospheric CO2 doubles. Thanks to unexpected cloud behavior in a few outlier GCMs, though, the IPCC can no longer rule out a ratio of 5 degrees or more.
That might mean our total emissions “budget” needs to shrink fast. Or it could mean cloud models have some bugs. The size of raindrops, the lifetime of clouds, the line between rain and snow are all under investigation.
For redwoods, there’s one silver lining to humanity’s bumbling: We may have helped them dodge nature’s slowest bullet. While most climate modelers focus on global warming, the International Atomic Energy Agency is working on a less pressing existential threat: glaciers.
New glaciers would threaten both redwoods and nuclear waste. To keep radioactive sludge politely buried, IAEA scientist Mike Thorne created a 200,000-year model to project the next ice age. Without the Industrial Revolution, it should have been due soon—relatively speaking. “I’d guess 50,000 years after present, but I wouldn’t rule out 23,000,” Thorne says.
By burning fossil fuels, though, humans likely pushed it off. “100,000 years is quite a good bet,” Thorne says. “In the highest [emissions] scenarios you can even push it beyond that: 200,000 or even 680,000 years after present.”
Meaning: If redwoods survive the next 2,000 years, they may have a better shot at the next 200,000. Maybe the descendants of Rowbotham’s trees will thank us.
Correction, Aug. 23, 2021: This article originally misstated that Miguel Fernandez predicted redwood habitat changes using tree ring data. He used past climate data and suggested that tree ring data be used to improve such predictions in the future.