Originally a guest post September 25, 2019 at Climate Etc.
Readers may recall that last autumn I wrote several article critiquing the Resplandy et al. (2018) ocean heat uptake study in Nature, which was based on measured changes in the O2/N2 ratio (δO2/N2) and CO2 atmospheric concentration. These were combined to produce an estimate (ΔAPOObs) of changes in atmospheric potential oxygen since 1991, from which they isolated a component (ΔAPOClimate) that can be used to estimate the change in ocean heat content. In four articles, here and here, here, and here, I set out why I thought the trend in ΔAPOClimate – and hence their ocean heat uptake estimate – was overstated, and its uncertainty greatly understated, essentially because of errors in their statistical methodology. The bulk of my criticisms were largely accepted by the authors of the study. However, it was evident from their related Realclimate article that in their submitted correction they had also made a change in an unconnected assumption, with the effect of offsetting much of the reduction in their ocean heat uptake estimate that correcting their statistical errors would have caused.
Nearly ten months have passed since then, without Nature publishing the authors’ correction.
However, Ruth Dixon has just spotted that the Resplandy et al. paper has today been retracted, at Nature’s request. This article at Retraction Watch covers the story. The Retraction Notice by the authors at Nature reads:
Shortly after publication, arising from comments from Nicholas Lewis, we realized that our reported uncertainties were underestimated owing to our treatment of certain systematic errors as random errors. In addition, we became aware of several smaller issues in our analysis of uncertainty. Although correcting these issues did not substantially change the central estimate of ocean warming, it led to a roughly fourfold increase in uncertainties, significantly weakening implications for an upward revision of ocean warming and climate sensitivity. Because of these weaker implications, the Nature editors asked for a Retraction, which we accept. Despite the revised uncertainties, our method remains valid and provides an estimate of ocean warming that is independent of the ocean data underpinning other approaches. The revised paper, with corrected uncertainties, will be submitted to another journal. The Retraction will contain a link to the new publication, if and when it is published.
I believe that this saga, as well as showing how ineffective journal peer review tends to be in spotting problematic issues in papers, illustrates the need for a much closer involvement of statisticians in climate science research. That was a point also made in one of the articles highlighted in Judith’s latest Week in Review post: Climate science needs professional statisticians [link].
Nicholas Lewis 25 September 2019