In July of 2015, California’s forests began to crumple. Parched from more than three years of severe drought, trees died in droves, transforming entire swaths of the Sierra Nevada from vibrant green to dull, lifeless red.
Within months, millions of trees had been lost statewide—a massive die-off whose seemingly abrupt onset left researchers like Mukesh Kumar shaken. “It really caught our eye,” recalls the University of Alabama ecohydrologist. “We couldn’t understand...why the trees died so suddenly.”
But as Kumar and his colleagues would discover, California’s trees had sounded a subtle death knell long before they breathed their last.
Now, by analyzing satellite images that track forest health over time, Kumar and his team have found a way to tap into some of the signals that portend arboreal death. Their technique, published today in the journal Nature Climate Change, can forecast mortality six to 19 months in advance—buying time, perhaps, for forest managers to intervene.
“Drought-induced tree die-off is becoming increasingly common and troubling,” says Aaron Lien, an environmental policy expert at the University of Arizona who was not involved in the study. “This is a great first step to get us to the point where we can respond to climate change impacts on forests in a more proactive way.”
As global temperatures climb, more and more trees are wilting under the heat. The stakes of these die-offs are high: Every compromised forest reduces Earth’s ability to store carbon, cycle water, brace soil against erosion, and support countless other forms of life. But the complexity of arboreal ecosystems makes climate-driven tree mortality “really challenging to predict,” says Mallory Barnes, an ecohydrologist and quantitative ecologist at Indiana University who was not involved in the study. “We don’t fully understand how or why it happens. We don’t even always know when a tree is finally ‘dead.’”
Previous models have tried to disentangle these relationships by amassing data on the cause itself—climate stress—or by studying the physiology of deteriorating plants. But both strategies have their limits, especially as both climate and ecosystems shift over time, Barnes says.
Kumar compares the process to forecasting how a human population might fare in the wake of hardship, like a natural disaster or an epidemic. A multitude of factors goes into determining who succumbs or survives: It’s not enough to just know the traits of the responsible pathogen, and there’s too much variation between people to reasonably account for all the ways an infection or its symptoms might manifest.
A better predictor of survival is a population’s resilience—that is, how quickly and effectively its members are able to recover from a small disturbance, says study author Yanlan Liu, who developed the new method as a part of her graduate thesis under Kumar’s supervision. For forests, that might include a brief temperature spike or bout of low soil moisture.
And luckily, tree tenacity is something ecologists already know how to monitor, says Liu, now a researcher at Stanford University.
Every 16 days, orbiting satellites snap aerial images of the world’s forests. The greener plants are, the more near-infrared wavelengths of light they reflect back into space—something that can be translated into what’s called a vegetation index.
It’s normal for greenery to fluctuate as trees dip in and out of the growth season, weather the occasional heat wave, or fight off an attack from a hungry cavalry of bugs, Liu says. Healthy trees bounce back from perturbations fast. But when forests start to look the same from month to month or year to year, that’s a clear sign of decline—something Kumar calls “a critical slowdown of a natural system.”
This so-called self-similarity becomes apparent when the vegetative index spits out consistent values over a three-month period or more, Liu says. That can happen when forests get brown and then stay brown—an obvious harbinger of death—but also when there’s a lot of greenery, which might not seem alarming at first glance, she says.
So Liu designed an algorithm to scour satellite data for these patterns. When the researchers tested their model with more than 16 years of images of California’s forests, they found that the drought-induced die-off that began in the summer of 2015 hadn’t been so sudden after all. Even while things looked lush and green from above, the forests’ resilience had begun to wane in certain regions. Most of these early warning signals appeared between six and 19 months in advance—but some popped up as early as October 2012, less than a year after the devastating dry spell had begun.
Liu and Kumar have yet to test their prediction strategy in other locations or against other environmental stressors like insect infestations or low nutrient availability. They also both acknowledge that the model has its limits. For one, its accuracy drops off as the lead time increases. Satellite data also isn’t foolproof: Images can be obscured when there’s excess cloud cover or snow, and they don’t capture what’s going on in a forest’s understory.
Even within a forest that appears to be on the brink, some tree species will survive better than others, Barnes points out. That means at-risk ecosystems may need to be evaluated on a case-by-case basis.
Still, “this is an exciting and cool way of looking at tree mortality,” Barnes says. “And this lead time is longer than most of the models we have now.” With enough advance notice, forest managers might be able to implement mitigation tactics to reduce mortality, she says. (Warning signals aside, she adds, it’s never too early to bolster a forest’s baseline resilience by boosting species diversity.)
Once trees are at risk, however, a few months’ head start might not be enough to get a moribund forest to reverse course, Lien says. “It’s enough time to prepare for some potential impacts like increased fire risk,” he says, “but it’s a difficult window to get action going on the ground.”
Extending the lead time will probably require bringing other sources of data into the model, Liu says—something that’s becoming increasingly urgent as climate change ups the intensity and frequency of extreme droughts. “To better predict mortality, it will be important to understand the mechanisms [of tree death],” she says. “We need to know the reasons behind low resilience.”
For now, the team’s approach underscores the need to preserve the world’s natural resources while we still can, Lien says. “Forest die-off is a big deal,” he says. “It’s not just about seeing a change in a landscape we’re used to. It goes beyond the forest...and we should be paying attention.”