The authors of a paper published in Science have retracted their article following the discovery of calculation errors.
The article,“Drought sensitivity in mesic forests heightens their vulnerability to climate change” by Robert Heilmayr of the University of California, Santa Barbara and colleagues found that in drier areas, trees are less sensitive to drought and in hotter regions with a wet climate, tree growth is expected to decrease.
It has been cited once, according to Clarivate’s Web of Science. Since its publication in December, the article has been downloaded 4,641 times, posted by 154 X users, and written about by 20 news outlets and press release sites.
In January, a group led by Stefan Klesse of the Swiss Federal Institute for Forest, Snow and Landscape Research informed the authors of errors in their R script, which was used to characterize weather and climate in these drought-affected regions. When the authors reran their script, the statistical significance of some conclusions, and test results changed – prompting them to retract the work.
“We believe the journal has handled this process incredibly well,” Heilmayr told Retraction Watch. “They encouraged us to work with Klesse et al. to get to the bottom of their concerns, and then followed our recommendation to retract the paper. Although we are disappointed, we were happy we were able to correct the scientific record.”
Klesse and his colleagues did not respond to our request for comment. [See an update at the end of this post.]
Four papers have been retracted from Science since the beginning of last year. The first was a similar case, in which researchers discovered an error and the authors retracted. Two involved the high-profile case of former Stanford president Marc Tessier-Lavigne, and the other was from another high-profile case at the Dana Farber Cancer Institute.
A spokesperson for Science said “It continues to encourage us to see examples like this where a data user lets the authors know about an error and the authors work quickly to correct the record.”
Update, 1800 UTC, 6/10/24: Klesse said our request for comment landed in his spam folder, and told us:
The main message of the paper stood against everything that I (and the greater tree-ring community, there were certainly more people puzzled by the results than just the four of us who reached out to the authors) was used to know from working with basically the same dataset for years. If this had been published in a “lesser” journal than Science or similar I might have just shrugged and not cared about it too much. But published in such a lighthouse journal, I really tried to understand how they could have gotten their results. Did WE always do something wrong, or is there some mistake in THEIR data/code? And after some days of digging, I stumbled over some rather big calculation mistakes, that changed a lot on my/our end in the “expected” direction (confirmed by using other independent climate datasets and target climate variables). Basically, the review process (the reviewers to be specific) failed the authors and the journal – and I am not saying that the reviewers should have read their R code, we can’t expect that. But thanks to the open data policy of Science, we were able to pinpoint the mistake and correct the record. Too bad Science doesn’t publish their review reports.
Like Retraction Watch? You can make a tax-deductible contribution to support our work, subscribe to our free daily digest or paid weekly update, follow us on Twitter, like us on Facebook, or add us to your RSS reader. If you find a retraction that’s not in The Retraction Watch Database, you can let us know here. For comments or feedback, email us at team@retractionwatch.com.
Looks like this is an innocent mistake. Wonder if how science community would handle this
Doing the right thing.
Kudos.
Help me out – I’m not an ecologist. Their claim was that drought-adapted plants are less affected by drought than plants adapted to non-drought conditions? And that was among the very best research submitted for publication from across the planet?
And the statistical significance of some results changed? In what direction did the results change? I’d just like to know in what way this entirely innocent error occurred. Did the error go against their hypothesis, or did it just happen to serve as the whole basis for their efforts?
One more thing: do you deserve kudos when you back down after you’ve been caught? “Yes, officer, I don’t belong in this bank in the middle of the night with burglary tools. Can I go now?”
“Being caught” somewhat implies intent, and your little scenario after does a lot more than *somewhat* imply it. Now, sure, a lot of what is seen here on Retraction Watch is indeed intentionally deceiving behavior–but not all of it. Genuine mistakes and errors DO happen, too.
That the authors of the paper in question were receptive in regards to concerns of errors, willingly re-running the script, and retracting when those errors did turn out to indeed be present does strongly imply there was no intent to deceive–that’s not the behavior generally seen in authors committing wilful scientific fraud, as can be seen from many, many posts on this site.
Rather than your analogy, this certainly appears (based on the information present and/or implied in the post itself) to be closer to an “Sorry, I thought this door led to the restroom, not the restaurant’s kitchen. I’ll get out of your way, my apologies.”
Which, yes, should not be deserving of kudos but simply be the norm when such an error occurs–but when the restaurant has to deal with heaps of folks accidentally and/or deliberately planting their behinds in the kitchen on a daily basis, while at best sticking their fingers in their ears and loudly singing “lalalalala can’t hear you”, damned if a “whoops, this isn’t where I meant to go, sorry for the bother” doesn’t look near-saintly in comparison.
In this case what seems to have happened, if you look at what the retraction notice in the journal says, is that the authors choose 1 method of calculating “evapotranspiration”, and this method resulted in statistically significant relationships. If they had instead choose another equally plausible method of calculating “evapotranspiration”, then many of the statistically significant relationships disappear. The way that they mislead is in presenting results of only the 1 option and not the others. There are in fact more than 2 ways of calculating “evapotranspiration”, and what the authors SHOULD have done is write a paper comparing the approaches and noting that there are differences. However, that paper would not make it into Science.