The publisher of Environmental Research Letters today took the bizarre position that expecting consistency between models and observations is an “error”.
The publisher stated that the rejected Bengtsson manuscript (which, as I understand it) had discussed the important problem of the discrepancy between models and observations had “contained errors”.
But what were the supposed “errors”? Bengtsson’s “error” appears to be the idea that models should be consistent with observations, an idea that the reviewer disputed.
The reviewer stated that IPCC ranges in AR4 and AR5 are “not directly comparable to observation based intervals”:
One cannot and should not simply interpret the IPCCs ranges for AR4 or 5 as confidence intervals or pdfs and hence they are not directly comparable to observation based intervals (as e.g. in Otto et al).
Later he re-iterated that “no consistency was to be expected in the first place”:
I have rated the potential impact in the field as high, but I have to emphasise that this would be a strongly negative impact, as it does not clarify anything but puts up the (false) claim of some big inconsistency, where no consistency was to be expected in the first place.
The reviewer summarized his concern in terms of media issues:
Summarising, the simplistic comparison of ranges from AR4, AR5, and Otto et al, combined with the statement they are inconsistent is less then helpful, actually it is harmful as it opens the door for oversimplified claims of “errors” and worse from the climate sceptics media side
Thus, the “error” (according to the publisher) seems to be nothing more than Bengtsson’s expectation that models be consistent with observations. Surely, even in climate science, this expectation cannot be seriously described as an “error”.
This is not to say that specific comparisons cannot be flawed e.g. around the edges, one may take issue on whether the geographical coverage of models corresponds to geographical coverage of observations (pace Cowtan-Way). But such craftsmanship issues hardly impugn the purpose of comparing models to observations – an exercise which was widely valued by climate scientists prior to the hiatus. It is ludicrous to suggest that the expectation of consistency is an “error” because “no consistency was to be expected in the first place”.
The publisher also said that the article “did not provide a significant advancement in the field”. However, most academic articles do not constitute “significant advancements” in their field and they still get published.
However, when AR5 was released, I noted that there was negligible literature available to AR5 about the discrepancy between models and observations, leaving IPCC in a very awkward position when it came for assessment. I noted that journal gatekeeping had contributed to this dilemma – both Lucia and coauthors and Ross and I had had submissions about model-observation discrepancies rejected. The Bengtsson rejection seems entirely in keeping with these earlier rejections.
Given the failure of the publisher to show any “error” other than the expectation that models be consistent with observations, I think that readers are entirely justified in concluding that the article was rejected not because it “contained errors”, but for the reason stated in the reviewers’ summary: because it was perceived to be “harmful… and worse from the climate sceptics’ media side”.