"Long-term records appear to consistently estimate 5–6 days change in leafing and flowering per °C," Elizabeth Wolkovich of the University of California San Diego told environmentalresearchweb. "This was a surprisingly consistent finding – even using a completely different dataset – and suggests that, for temperate regions, we may have a relatively accurate estimate of mean changes in plant leafing and flowering with warming."

But experiments predicted less than 1–2 days' advance per °C of temperature rise. On average, changes in flowering time suggested by the tests were 8.5 times too small. Meanwhile, leafing date advances needed to be four times larger to fit with observations.

"This suggests current estimates from experiments – which are sometimes used in models of future species, land and atmospheric changes – are highly uncertain and might be greatly underestimating what future ecosystems will look like," said Wolkovich.

Wolkovich and colleagues from the US, Canada, Sweden, UK and Switzerland carried out a meta-analysis of observations and experiments from four continents – North America, Europe, Asia and Australia – for 1634 plant species.

When the team analysed the 36 species for which both observational and experimental data were available, the experiments predicted a delay in flowering and little change in leafing with higher temperatures. The observations, in contrast, indicated an advance of more than 4.6  days per °C for both flowering and leafing.

Wolkovich believes experiments will be critical to help understand and predict species' responses to warming, as we do not have records in most places and are headed towards temperature increases far above historical records.

"My hope is that this work will further a conversation already underway about how best to design, fund and implement warming experiments – as well as how to continue and increase long-term monitoring," she said.

The researchers reckon the disparity between observations and experimental results could be due to complex interactions among multiple drivers in the wild, or because the experiments reduce solar irradiance and create drier soils, dampening the plants' response to manipulated warming. Scientists typically use techniques such as open-top chambers or above-canopy heaters to create artificial warming.

Wolkovich is calling for standardized temperature monitoring for experiments as well as more monitoring networks for collecting observations. "Many of the datasets used in the paper were never funded and have ended data collection, just when we need continuing data most," she said. It would also be useful to record snowmelt date, which affects plant phenology in many alpine and tundra habitats.

Now Wolkovich, Ben Cook of NASA Goddard and several other team members are "digging into the long-term observational data to improve how well we can predict the differences between species in how they respond to warming". In the current study, the mean response across species was quite consistent but the variation between species was huge, said Wolkovich, and so the team is working on understanding this.

"We are looking at why species that are annuals or bloom early in the spring tend to change much more with warming than perennial or later-blooming species and how year-round warming, including in the winter when many species rely on continued cold temperatures ('chilling') to trigger spring leafing and flowering, affects how much species shift their leafing and flowering with climate change," she said.

The scientists reported their work in Nature.