Digging deeper: Authors retract soil paper so “the error we made does not propagate”

The authors of a 2018 paper on how much carbon soil can store have retracted the work after concluding that their analysis was fatally flawed. 

The article, “Soil carbon stocks are underestimated in mountainous regions,” appeared in the journal Geoderma. Its authors are affiliated with the French National Institute for Agricultural Research.

According to the abstract of the paper: 

Current estimates of soil organic carbon (SOC) stocks are calculated by multiplying the SOC density times the planimetric area of an individual cell. In the reality, the land surface is not always in horizontal planes, but sometimes in tilted planes, especially in mountainous regions. The differences between a horizontal plane and a tilted plane are controlled by slope, so we investigated the effect of inclination on the SOC stocks calculation using HWSD dataset in mountainous regions including the Alps, the Andes, the Plateau of Tibet and the Rocky Mountains. Our results showed that inclination effect strongly influenced SOC stocks calculation in mountainous regions and previous SOC stocks estimates were underestimated. SOC stocks increased between 4.04% and 15.00% when 90 m resolution elevation data was used for accounting the inclination effect, which was much higher than that of using 1 km resolution elevation (0.90% to 5.00%). Therefore, we suggest that it is necessary to consider the inclination effect in the calculation of SOC stocks in mountainous regions.  

However, the authors, led by Dominque Arrouays, learned that their analysis might be on shaky ground. As the retraction notice states: 

In this paper, we concluded that the Soil Organic Carbon (SOC) stocks were underestimated in mountainous regions, mainly due to the slope effect.

…

We think that our conclusions are biased and that this paper should be retracted. Actually, the hypothesis that SOC stocks estimates are sensitive to slope effect is wrong. It makes implicitly the hypothesis that soil is a kind of mantle characterized by a given thickness and that what we measure is this thickness perpendicular to the soil surface.

In practice, the observations contained in the databases should have been made vertically, therefore the slope does not change the volume at all. As shown in Fig. 1, if we consider a square pixel of N m2, (with N = length × width), the “Voxel” down to h cm is always N × h m3 (with h = height) no matter how the shape of the surface is (the only rare exception is overhanging areas).

Consequently, the conclusions of our paper are wrong. The global stocks under mountainous areas are not underestimated, and the estimates of SOC stocks realised using the HWSD (Hiederer and Köchy, 2012) are not biased due to a slope effect. We fully apologize for this misconception.

We retract this paper, in order that it is not used further by others to erroneously correct the SOC stock calculations, and that the error we made does not propagate into other studies.

More generally, this error also raises questions about the definitions of soil depth and soil or horizon thickness. Indeed, soil horizons are often parallel to the soil surface and one may consider that the thickness should be measured in a strictly vertical way, or perpendicularly to the soil surface.

‘[U]nfortunate that it was overlooked’

Jan Willem van Groenigen, chair of the editors in chief of Geoderma, told us that he received an email from a reader about the now-retracted paper in late January 2019 — nearly a year after the article appeared online: 

After collegial discussion between the authors and the reader, the authors recognized the flaw in their approach and asked for a retraction – the handling Editor-in-Chief and I decided to honour this request and publish the retraction note. 

Given the fundamental flaw with the paper, we wondered if it might have been picked up during peer review. Van Groenigen acknowledged so: 

Yes, of course – this is the type of error that ideally gets detected at the review stage and it was unfortunate that it was overlooked. As far as I know, Geoderma never had to retract an article before and I would very much have preferred to keep it that way. But honest mistakes are sometimes being made and we felt that the only way to correct this was with a retraction.

Van Groenigen requested that Elsevier, the journal’s publisher, retract the paper in May 2019. He said the paper was retracted on ScienceDirect, Elsevier’s online platform, on June 10. The retraction notice — linked to the original article — did not appear until January 20 of this year.

Van Groenigen told us:

According to the Elsevier staff this notice is largely intended for the print version of the journal (although few people read in print anymore of course) and therefore is subject to the schedules of printed issues.

Readers of this blog may recall that several years ago, Geoderma confronted an alleged citation-rigging scheme by one of its editorial board members. 

Like Retraction Watch? You can make a tax-deductible contribution to support our work, follow us on Twitter, like us on Facebook, add us to your RSS reader, or subscribe to our daily digest. If you find a retraction that’s not in our database, you can let us know here. For comments or feedback, email us at [email protected].