Nature Climate Change | Climate Feedback

AMS2010: Data gaps and errors may have masked warming

Olive Heffernan

New analyses provide preliminary evidence that temperature data from the UK Met office may under-estimate recent warming. That’s the conclusion of a talk given here today by Chris Folland of the Met Office Hadley Centre. Folland says that there is a very good chance that there has been more warming over land and over the ocean in the past decade than suggested by conventional data sets, but he says that the issues with land and ocean data are entirely unrelated.

For land, the problem of underestimating warming stems from data gaps in the average monthly temperature data set of the Met Office Hadley Centre, known as HadCruT3. Temperatures over the past decade were recently re-analyzed using a european climate model by Adrian Simmons of the European Centre for Medium-Range Weather Forecasts in Reading, UK and colleagues, and are soon to be published in the Journal of Geophysical Research [subscription]. Simmons and colleagues compared air temperature and humidity data collected over the past decade by the Hadley Centre with re-analyzed data for the same period. Average warming over land was larger for the fully sampled re-analyzed data than for the HadCRUT3 temperature data. The difference between the data sets is particularly notable for northeast Canada, Greenland and nothern parts of Asia, areas which are warming particularly rapidly.

Recent observations support the idea that warming is, in fact, greater than the temperature record would suggest, says Folland. In the Arctic, for example, summer sea ice is disappearing more quickly than expected based on the HadCRUT3 temperature data. But Folland also cautions that the results are preliminary and that the re-analysis, which only goes back twenty years, should now be extended back seventy or eighty years.

For the ocean data, it’s a different issue. John Kennedy of the Met Office and colleagues previously reported in Nature [subscription] that changes in the methods used to collect sea surface temperature (SST) data at the end of World War II caused problems in comparing pre- and post-war data. Now they have a new analysis (yet to be published) suggesting that smaller changes in data collection methods since the end of the war could also be significant.

Over the past 20 years, the primary source of SST data has changed from ships to ocean buoys. Because ships warm the water during data collection, there has been a drop in recorded SSTs since bouys,which are more accurate, became the main data source. So what could appear to be a relative cooling trend in SSTs over the past decade may actually just due to changes in errors in the data. Scientists are confident that the buoy data are more accurate because they compare favourably with reliable satellite data.

Although scientists have previously suspected that recent warming has been more pronounced than records suggest, these new studies provide much stronger evidence that that’s likely to be true. It is still quite possible that there has been some slowing of global warming in the last decade, adds Folland, but its likely to be significantly less than previously thought.


  1. Report this comment

    John Robertson said:

    This isn’t news – people have been discussing this possible bias in temperature readings (intake of cooling water to ships) since at least 2008. The question is – as stated by Geoff Larson in June 2008 “Surely if they had followed a scientific approach, and I assume they did, the transition from ship- to buoy-derived SST’s would have been accompanied by strict comparison & control between the 2 methods, to iron out any biases, during the transition period.” see:

    Others there point out “I know two old navy engineers who pooh-pooh this claim. The induction ports were located far way from the machinery.” It should be simple to find out the truth by studying the plans for the ships to see where the induction ports and the thermometers were located.