Researchers have depended increasingly on experiments that measure how plants respond to artificial warming in an attempt to understand how plants around the world will act in a warming climate.
But a new study says that such experiments are underestimating potential advances in the timing of flowering and leafing four to eightfold, when compared with natural observations.
As a result, species could change far more quickly than the experiments suggest, with major implications for water supplies, pollination of crops and ecosystems.
The comparison has been done by an interdisciplinary team from some 20 institutions in North America and Europe.
"Up to now, it's been assumed that experimental systems will respond the same as natural systems respond-but they don't," said coauthor Benjamin Cook, a climate modeler at the NASA Goddard Institute for Space Studies and Columbia University's Lamont-Doherty Earth Observatory.
"This suggests that predicted ecosystem changes-including continuing advances in the start of spring across much of the globe-may be far greater than current estimates based on data from experiments," Elizabeth Wolkovich, who led the team as a postdoctoral fellow at the University of California, San Diego, said.
The timing of annual plant and animal life events-the study of which is known as phenology - has emerged as perhaps the most consistent and visible gauge of nature's response to rising temperature.
Globally over the past century, land surfaces have warmed an average of about half a degree Celsius (1.25 degrees Fahrenheit), but some places, such as Alaska, are warming much more rapidly (there, about 1.8 degrees C, or over 3 degrees F).
As a result, long-term historical records show that many plant species are flowering and leafing out days, or even weeks, earlier over recent decades.
But because historical records are not available in many places and the future may bring ever-higher temperatures, many scientists are also trying to project by doing experiments in which they heat small field plots and measure the responses.
The researchers in the study created new global databases of plant phenology, pitting calculations from experiments versus those from long-term monitoring of natural records.
They included data from 50 different studies covering 1,643 species on four continents. Their analysis showed that experiments predicted every degree rise Celsius would advance plants' flowering and leafing from half a day to 1.6 days.
But in looking at actual observations in nature, they found advances four times faster for leafing-and over eight times faster for flowering. In sum, the natural records showed that phenological events advancing on average, five to six days per degree Celsius.
The finding was strikingly consistent across species and datasets. Wolkovich said this suggests that long-term records "are converging on a consistent average response," and that future plant and ecosystem responses to climate change may be much higher than estimated from experimental data alone.
A number of factors could explain the discrepancies, said the researchers. These could include effects of longer-term climate change, including shifts in plants' genes as they adjust to warming, which would not be mirrored by shorter-term experiments.
Or, it could be specific aspects of the experiments themselves, such as exactly how researchers manipulate temperatures and how accurately they measure them, they said.
For instance, experimenters have used a variety of methods to increase temperatures, including cables buried in the soil, small greenhouse-like structures and heat sources placed above plants.
"Some experiments get closer to nature than others," said Cook.
"We need to address this by improving experiments. In the meantime, we should pay more attention to nature, because it's giving us critical information. For effective policy and conservation plans, we really need to have accurate predictions for which species will respond, and how much," Cook added.
The study has been published in the journal Nature.