Taking the Earth's temperature requires more than lots of thermometers. It also relies on surveying tools, satellites—and confidence in statistical models used to put the numbers together.
Still, those thermometer readings do matter. So a team of scientists recently tackled the job of cleaning up and organizing 1.6 billion temperature readings from two centuries and nearly 40,000 land-based locations.
That required removing duplicate numbers, tossing out clearly erroneous records—such as temperatures above 5,000 degrees Fahrenheit—and fixing Celsius readings mislabeled as Fahrenheit. The process took nearly two years and has yielded a data set the project's leader says can be analyzed in a matter of hours.
"Before us, there was a huge barrier to entry" in the field of analyzing temperature numbers, says Richard Muller, scientific director of the Berkeley Earth Surface Temperature team and a physicist at the University of California, Berkeley.
Many scientists are giving the Berkeley Earth team kudos for creating the unified database.
Plaudits are coming even from some scientists who dispute the conclusions of the Berkeley Earth team itself. The team analyzed its own temperature records and concluded that the Earth's land has warmed by an average of about 0.9 degree Celsius (1.6 degrees Fahrenheit) since the mid-1950s. That is in line with some earlier efforts that used smaller sets of temperature readings.
Some scientists reaching into the new database, while generally agreeing the Earth has been warming, disagree with details of the team's findings. For instance, Berkeley Earth's trend graphs, which reflect some modeling, show temperatures continuing to rise since the late 1990s. But the raw numbers show no definitive evidence of an increase in that time. The group also hasn't made use of satellite-derived temperature readings. These show a smaller increase. The difference may reflect that some land-based weather stations aren't well maintained.
"I'm inclined to give [satellite] data more weight than reconstructions from surface-station data," says Stephen McIntyre, a Canadian mathematician who writes about climate, often critically of studies that find warming, at his website Climate Audit. Satellites show about half the amount of warming as that of land-based readings in the past three decades, when the relevant data were collected from space, he says.
Such disputes demonstrate the statistical and uncertain nature of tracking global temperature. Even with tens of thousands of weather stations, most of the Earth's surface isn't monitored. Some stations are more reliable than others. Calculating a global average temperature requires extrapolating from these readings to the whole globe, adjusting for data lapses and suspect stations. And no two groups do this identically.
The Berkeley Earth team uses a statistical tool it calls a scalpel to cut out data when there is a gap. It uses a process called Kriging to fill in gaps in data, borrowed from geoscientists and surveyors, who used it, for example, to estimate the elevation at point B that is between A and C, points where altitude is known. And it weights data more heavily from stations that are reliable than that from those that produce erratic numbers.
The result is a statistical model of what the temperature was at any given moment at any given location. Actual thermometer readings, when they are available, are used only as fodder for the statistical model.
The Berkeley scientists "have a very complicated model," says William Briggs, a member of the probability and statistics committee of the American Meteorology Society. "They reported on the setting of one of those dials" in the model. "That is not the actual temperature."
Calculating a global temperature is necessary to track climate trends because, as your TV meteorologist might warn, local conditions can differ. Much of the U.S. and Northern Europe has cooled in the last 70 years, Berkeley Earth found. So did one-third of all weather stations world-wide, while two-thirds warmed. The project cites this as evidence of overall warming; skeptics aren't convinced because it depends how concentrated those warming sites are. If they happen to be bunched up while the cooling sites are in sparsely measured areas, then more places could be cooling.
The risk with models is that scientists can become enslaved to one they have chosen, says Mr. McIntyre. "The best antidote is for authors to make all their data available at the time of publication together with scrupulous documentation," he says, crediting Berkeley Earth with attempting to do this.
Any statistical model produces results with some level of uncertainty. The Berkeley Earth project is no different. That uncertainty is large enough to dwarf some trends in temperature. For instance, fluctuations in the land temperature for the past 13 years make it extremely difficult to say whether the Earth has been continuing to warm during that time.
This possible halting of the temperature rise led to a dispute between members of the Berkeley Earth team. Judith Curry, Mr. Muller's co-author and a professor of earth and atmospheric sciences at the Georgia Institute of Technology, told a reporter for the Daily Mail she questioned Mr. Muller's claim, which he published in an opinion column in The Wall Street Journal, that "you should not be a skeptic, at least not any longer." She said that if the global temperature has flattened out, that would raise new questions, and scientific skepticism would remain warranted. Asked for further comment, she referred questions to Mr. Muller, and on her blog she said that after a 90-minute talk with him, "there isn't much that we disagree on."
This sort of messy hashing-out of the global climate record is happening in the open because the Berkeley Earth team chose to release its data, and its papers, before undergoing peer review by scientific journals. Already some feedback has led to updates and corrections to the research. Berkeley Earth plans other work, including adding ocean temperature trends to the land records and fixing errors in its database.
"Some people mistakenly think peer review means secret review by anonymous referees at journals," Mr. Muller says. "We're getting wonderful peer review, from McIntyre, from Briggs, from other people. That's the process of science."