- May 14, 2018
Zeke Hausfather, Research Scientist at Berkeley Earth, writes this comment to dispute NAS’s rendition of the debate between John Bates and Tom Karl about the reproducibility of National Oceanic and Atmospheric Administration (NOAA) climate data. The NAS welcomes free discussion of the irreproducibility crisis and climate science, and is delighted to publish Dr. Hausfather’s critique. We will also publish a reply to Dr. Hausfather by Irreproducibility Crisis co-author David Randall.
The recent claim in NAS’s Irreproducibility Report that the data used by climate scientist Tom Karl and colleagues in their 2015 paper in Science "disappeared" or is not publicly available is simply not true. All the data used by Karl et al was archived at the time of the paper's publication here: ftp://ftp.ncdc.noaa.gov/pub/data/scpub201506/
I know this quite well, because I led a team of researchers that evaluated NOAA’s updates to their ocean temperature record in response to the controversy after the publication of the 2015 paper. In our 2017 paper, published in the journal Science Advances, we compared the old NOAA record and the new NOAA record to independent instrumentally homogenous records created from buoys, satellite radiometers, and Argo floats. We found that the new NOAA record agrees quite well with all of these, while the old NOAA record shows much less warming.
Statements that Karl et al did not release their data are a misinterpretation of claims from former NOAA researcher John Bates, who was unhappy about the specific technical details of the data archiving procedures associated with the Karl et al paper. Specifically, Bates was concerned that the experimental land record used in the paper had not gone through the procedures associated with the production of an operational climate product at NOAA.
However, the new record presented by Karl and colleagues was experimental, and was not intended to be nor communicated as the new official "operational" NOAA land temperature record at the time of publication. Karl and colleagues decided to use all 35,000 available land temperature stations rather than just the 7,000 or so at use at the time, which provides a more accurate estimate of land temperatures, particularly in regions like the Arctic with poor coverage. They released the data from all of these stations so that researchers could replicate their work. This year, after additional development, the record first presented by Karl and colleagues will officially replace NOAA's current operational land temperature record with the launch of the Global Historical Climatological Network version 4.
When they published their paper, the authors released temperature anomalies, spatially gridded data land and ocean data, and the land station data associated with their analysis. They put all of this up on NOAA’s FTP site in early June 2015. As someone who works on and develops surface temperature records, the data they provided was sufficient for me to examine their analysis in detail and see how it compared to other groups. Citing this work as an example of irreproducible research is deeply misleading, as the authors released all the information necessary for reproduction by independent third-parties, reproduction that has subsequently been undertaken and that has independently validated their results.
For more detail on this and other criticisms of the Karl et al paper, see this article I wrote for Carbon Brief: https://www.carbonbrief.org/factcheck-mail-sundays-astonishing-evidence-global-temperature-rise.
Image Credit: Berkely Earth/Zeke Hausfather