Temperature drop due to measurement error

A longstanding puzzle in Earth’s climate record has just been solved. Between 1940 and 1970 — right in the middle of a steady, post-Industrial warming trend — global mean temperatures mysteriously dropped, with the most precipitous drop in 1945. Experts have debated a handful of explanations, including an increase in anthropogenic sulfate emissions that blocked sunlight and anomalous atmosphere-ocean interactions. But a new study suggests that the story is much simpler: An abrupt change in how scientists measured temperatures above the ocean, not the climate itself, explains the slump.

The change came in 1945, says David Thompson of Colorado State University in Colorado Springs, lead author of the study published May 29 in Nature. That year, the source of sea-surface temperature data — which are used alongside land-surface temperature data to determine global mean temperatures — dramatically shifted. Since the late 19th century, a varying roster of nations, including the United States, United Kingdom, Russia, Germany and Japan, contributed to the sea-surface temperature database. The countries use different measurement protocols with slightly different intrinsic biases. With so many nations participating, however, these biases cancelled each other out. But in the years surrounding World War II, that wasn’t the case, Thompson and colleagues wrote.

From 1942 to August 1945, 80 percent of observations came from U.S. ships, which took their readings of sea-surface temperatures as seawater entered the engine cooling system. At the war’s end, ships from the United Kingdom resumed data collection as well. Within a year, half the observations in the record were British, with only 30 percent from U.S. ships. British sailors collected sea-surface temperature readings by dropping uninsulated buckets from the deck, reeling them up full of water, and measuring the temperature. While the engine room method tends to inflate measurements a bit because the water is near the hot ship engine, the bucket method tends to deflate them as the water cools when contacting ambient air, the team wrote. So that baffling mid-century downturn, they say, is an artifact of instrumentation.

In creating the global-mean temperature time series, researchers had corrected for the bias through 1941, but not through later decades. Previously published versions of this time series show a diffuse stretch of cooling between 1940 and 1970, not a sharp discontinuity at 1945. But these figures include two sources of “natural noise”: the El Niño/Southern Oscillation and Northern Hemisphere land-sea interactions, according to Thompson and his colleagues. So they decided to take another look at this time series, filtering out the effects of the noise, and bringing the temperature drop into sharper focus.

The team’s analysis highlights five additional drops in the record, but each aligns with a massive volcanic eruption. No such sun-shading event marked 1945. The atomic bombings of Hiroshima and Nagasaki likely had little effect on global temperatures: Models suggest that such explosions are predicted to cause a decline of less than 0.03 degrees Celsius. In the six months following August 1945, however, the recorded drop was 10 times that large.

If the lowly bucket is indeed at the bottom of this mystery, one wonders why scientists hadn’t seen it before, says Chris Forest, a member of the Joint Program on the Science and Policy of Global Change at MIT in Cambridge, Mass., who commented on the research in the same issue of Nature.For one reason, he says, “they’d already made corrections for measurement bias prior to 1942.  They hadn’t realized that would still be a big issue after World War II.” In particular, he says, “they overlooked the mixture of countries doing the measuring, and 1945 is the one point in time when that’s critical.”

Though this may seem like a big issue, the researchers suggest that it’s not. Although the temperature series will need adjustments and climate models that incorporated mid-century data will need to be recalibrated, “the adjustments are unlikely to significantly affect estimates of century-long trends in global-mean temperatures,” they conclude, because the data before and after this period are still sound.

Although the changes "could be another source of skepticism for those inclined to be skeptical” about climate models, Forest says, it won’t change most scientists’ verdict on the last hundred years: The world is warming. More scrupulous study only underscores their confidence. Forest, for one, is excited. “We’re going to continue to see updates in the next decades as we come up with more clever ways to analyze the data,” he says. “That’s the way the science moves forward.”

Rachel Carr
Geotimes contributing writer

Links:
Videocast: Ocean temperature gaffe, Geotimes online, June 9, 2008

Back to top