What if the missing heat has been there all along? In 2013, the Intergovernmental Panel on Climate Change (IPCC) flagged an odd phenomenon: Atmospheric temperature data collected over the past few decades suggested that global warming had slowed down beginning around 1998. Global warming skeptics crowed, and scientists delved into the global climate system to find out where the missing heat had gone. But a new analysis suggests that the real culprits are the data themselves. When better corrections for various sources of bias are applied to the data, the authors say, the so-called global warming hiatus vanishes—and in fact, they argue, global warming may have sped up.

That won't startle some scientists, who say the “hiatus” was always a misnomer. “There is no hiatus or pause,” says climate scientist Michael Mann of Pennsylvania State University, University Park, who prefers the term “temporary slowdown.” But he and others do think something has changed since the late 1990s: Perhaps the deep waters of the Pacific and Atlantic oceans are storing more heat, or volcanic eruptions and pollution have been shading the planet and offsetting the warming. What's more, they note, 1998 was a particularly strong (and hot) El Niño year—not an ideal starting point for determining a subsequent trend.

But the temperature data themselves—collected by a variety of techniques from land and sea—have also been a source of concern, says Thomas Karl, director of the National Oceanic and Atmospheric Administration's National Climatic Data Center in Asheville, North Carolina, and the lead author on the new paper published online this week in Science (http://scim.ag/TKarl). Climate scientists have worked for years to improve corrections for bias in the data. “It's an ongoing activity,” Karl says.

Creating a single, self-consistent, long-term record of sea surface temperature (SST) has proven especially tricky. For much of the past 2 centuries, ocean temperatures were measured from ships, by means of a bucket thrown over the side. Different fleets used different measurement techniques and, over time, various types of buckets—first wooden ones, then specially designed canvas ones. Eventually, buckets gave way to ship engine intake measurements, taken when water was brought in to cool the machinery. And by the end of the 20th century, far more accurate buoy measurements took over. Each technique required different corrections.

Another challenge was incorporating land-based readings from thousands of new measurement stations in regions that have long had scant coverage, particularly Asia, South America, and Africa. New data from these regions have been amassed over the past 5 years as part of the International Surface Temperature Initiative, which released its first report just last year.

In their paper, Karl's team sums up the combined effect of additional land temperature stations, corrected commercial ship temperature data, and corrected ship-to-buoy calibrations. The group estimates that the world warmed at a rate of 0.086°C per decade between 1998 and 2012—more than twice the IPCC's estimate of about 0.039°C per decade. The new estimate, the researchers note, is much closer to the rate of 0.113°C per decade estimated for 1950 to 1999. And for the period from 2000 to 2014, the new analysis suggests a warming rate of 0.116°C per decade—slightly higher than the 20th century rate. “What you see is that the slowdown just goes away,” Karl says.

And that's without including the elephant in the room: Arctic warming. A 2014 paper in the Quarterly Journal of the Royal Meteorological Society highlighted how the scarcity of temperature data from the Arctic, which is warming twice as fast as the rest of the planet, has produced a significant “cool” bias in the global trends, especially since 1997.

“The post-1998 period is really difficult, partially because of Arctic warming and partly because of the change in SST measurements,” says Kevin Cowtan, a computational scientist at the University of York in the United Kingdom, who co-authored the 2014 paper. “The fact that it's caused problems is completely understandable, if unfortunate.”

To estimate how Arctic warming might alter global trends, Karl's team used a nonlinear technique to fill in the data gaps for the polar region. Including the Arctic, they found, would add between 0.02°C and 0.03°C of warming per decade. Karl notes that this is just an estimate, however, and wasn't included in the paper's final reanalysis of recent warming.

Not everyone agrees that the 21st century slowdown is entirely a data artifact. Mann notes that there is “very clear” evidence of a slowdown in large-scale warming in the tropical Pacific; in a previous paper, he and others linked it to a natural decades-long climate pattern that brought about La Niña–like cooler conditions in the past decade (Science, 27 February, p. 988). “The tropical Pacific definitely warmed less over that time period than climate models had predicted,” Mann says.

Cowtan agrees, adding that there are a lot of lingering uncertainties in the data, particularly in the Arctic, as well as in some of the shipboard corrections during the last century. “My feeling is, they've got the right answer—but not for quite the right reasons,” Cowtan says. “My guess is there's a little bit too much ocean warming [in their calculations], and not enough from the Arctic.”

Karl says his team is planning ways to address the Arctic temperature issue next. He also says research into the slowdown has spurred important insights that help clarify the global climate system. “Global temperatures might have been even warmer than we're reporting had some of these other factors not come into play,” Karl says. “And once these things play out, we may find we're warming at an even more rapid rate than we saw at the end of the last century.”