The largest single adjustment climate scientists make to historical global temperatures

Mark Roulston, Senior Data Scientist •

Blog posts

It is tempting to think that data quality is more of an issue in social science, where messy human beings are involved, than when dealing with measurements of physical quantities from sensors. But unfortunately this is not the case—as this overview on historical sea temperature measurement shows.

Since the eighteenth-century ships have routinely taken measurements of the temperature of the ocean. Historically this was done by lowering a bucket into the ocean, hauling it back up, and putting a thermometer in the water collected. Copious amounts of data on sea temperatures have been obtained this way: The International Comprehensive Ocean-Atmosphere Data Set (ICOADS) contains almost 40 million of these measurements taken from ships’ logbooks between 1784 and 1949.

From the 1920s some steamships began taking sea temperatures using samples from the water taken in to cool their engines. Unlike the water collected in buckets, this water had no opportunity to exchange heat with the surrounding air, and tended to be slightly cooler than bucket water. This method of taking temperatures was particularly common aboard American ships, and during the Second World War the U.S. fleet contributed a large fraction of the global temperature data. After the war, the proportion of bucket measurement data one again increased.

Over the last quarter of a century floating buoys have replaced ships as the primary source of sea temperature data, with over three million measurements taken every year. The buoy measurements are about 0.1°C cooler than the ship measurements they have superseded.

There is no shortage of historical data on sea surface temperatures but if you want to understand how these temperatures have changed over the last hundred years you have to deal with the differences in measurements taken using different methods.

Having more measurements of any type will not, on its own, solve this problem. Failing to deal with it could lead you to draw erroneous conclusions about how climate has changed during the twentieth century. A large amount of effort has gone into determining the appropriate adjustments that should be made to the data to allow robust comparisons to be made between different time periods. This work even includes considering whether a bucket used to collect sea water was made of canvas or wood.

Whether bucket measurements or buoy measurements are of “higher quality” is not really the issue, but the fact that they are different is crucial.

Accounting for how sea temperature measurements were taken is the largest single adjustment that climate scientists make to historical global temperature data—and it actually reduces estimates of warming since 1880 by over a third, relative to the unadjusted data.