The humble bucket turns out to be at the bottom of a perplexing anomaly in the climate records for the twentieth century.

The time series of land and ocean temperature measurements, begun in 1860, shows a strange cooling of about 0.3 °C in the global mean temperature in 1945, relative to the 1961–90 average. The sharpness of the drop stands out even more if the signatures of internal climate variability, such as those associated with El Niño events, are filtered from the record.

This cooling at the end of the Second World War is one of several temperature drops in the record. But unlike others, such as the 1991 cooling caused by the eruption of Mount Pinatubo in the Philippines, it is limited to ocean temperatures and is not associated with any known climatic or geological phenomenon. The nuclear explosions in 1945 over Hiroshima and Nagasaki were ruled out as a possible cause because they are thought to have had no impact on global temperature. Other theories proposed as explanations for the cooling include a massive 1940s El Niño event that had somehow slipped attention, or that it was the result of sulphate aerosols from burning dirty coal. But neither of these was convincing.

A US–British team of climate scientists has now found a surprisingly simple explanation for the long-standing conundrum (page 646). It turns out that the mysterious drop is due to differences in the way that British and US ships' crews measured the sea surface temperature (SST) in the 1940s.

Only a few SST measurements were made during wartime, and almost exclusively by US ships. Then, in the summer of 1945, British ships resumed measurements. But whereas US crews had measured the temperature of the intake water used for cooling the ships' engines, British crews collected water in buckets from the sea for their measurements. When these uninsulated buckets were hauled from the ocean, the temperature probe would get a little colder as a result of the cooling effect of evaporation. US measurements, on the other hand, yielded slightly higher temperatures due to the warm engine-room environment.

The standard logbook entries made at the time contain no information about how the measurements were taken, so the cause was overlooked, says David Thompson, first author on the paper and an atmospheric scientist at Colorado State University in Fort Collins. As a result, the bias — which, although small, was large enough to produce the sharp drop in global mean temperature — was never adjusted for.

“The time series is one of the great climate records we have,” Thompson says. “During a sabbatical in Britain, I revisited work that I had started a long time ago, and it suddenly occurred to me that the mid-1940s cooling might not necessarily have physical causes.”

Thompson discovered the explanation after questioning maritime experts from different countries about the history of shipping, and searching the scientific literature and international databases for scattered bits of relevant information.

“We always thought the observed cooling was real,” says Phil Jones, a climate researcher at the University of East Anglia, Norwich, UK, who carried out the study with Thompson. “We did know that there were fewer measurements during the war than before and thereafter, but we simply made wrong assumptions on how and by whom the measurements were taken,” he says. “It is pretty clear now that the bias is instrumental.”

It is welcome news for climate modellers. The post-war temperature anomaly has been grossly outside the range of all computer-based climate reconstructions considered by the Intergovernmental Panel on Climate Change (IPCC), and it was prominently featured in the group's 2007 summary for policy-makers.

“The unusual up and down in SSTs in the 1940s stood out like a sore thumb in the past,” says Susan Solomon, a senior scientist at the US National Oceanic and Atmospheric Administration (NOAA) in Boulder, Colorado, and co-chair of the IPCC working group on the physical basis of climate change. “We couldn't explain it, so we showed all the fingers, sore thumb and all,” she says.

Climate researchers can now start setting the twentieth-century temperature record straight. The abrupt drop in 1945 will then probably disappear, but what the corrected time series will look like is not yet clear.

And further corrections may come. For example, the gradual shift since the 1970s from (warm-biased) ship-based measurements to (cold-biased) drifting buoys has probably led to a slight underestimate of SST warming, says Richard Reynolds of NOAA's National Climatic Data Center in Asheville, North Carolina. “More data problems will undoubtedly come along, if likely of smaller magnitude,” he says.

Climate researchers know from past experience that satellite sensors, radiosondes and ocean profilers are prone to bias. Land measurements are considered much more robust. Climate scientists should think about data quality more often, says Jones, so that there is no opportunity for incorrect data to sow seeds of doubt in people's minds about the reality of climate change.