Last year, Isabelle Larocque-Tobler and coauthors published a chironomid-inferred August air-temperature reconstruction for the 20th century from a Polish lake in Quaternary Science Reviews. The correlation between instrumental temperature records and the near-annual resolution reconstruction from Lake Żabińskie is remarkably good at 0.74. The mean absolute error is only 0.75°C, considerably lower than the cross-validation uncertainty in the Polish-Canadian calibration set (RMSEP = 2.3 °C).
Yesterday a corrigendum was published, reporting three problems with Larocque-Tobler et al (2015; hereafter LT15).
The chironomid stratigraphy was not consistent with the online data table
More precisely, LT15 used an incorrect version of the chironomid stratigraphy from Lake Żabińskie. The corrigendum reports several taxonomic and other changes between the published stratigraphy and the data now archived at the Palaeoclimate Data Centre (I think this is the first chironomid calibration set to be archived publicly). No explanation of how the wrong data came to be used is given.
The reconstruction obtained from the online data differed from the reconstruction published
More precisely, the reconstruction in LT15 was incorrect as there were several data processing errors including: misspelt taxa (which will then be excluded from the analysis); inclusion of lakes with low chironomid counts; and the inclusion of temperature as a species in the calibration set (this will make much less difference to the reconstruction than one might suppose).
The number of head capsules per sample used for the temperature reconstruction was below 50, the typical number used for temperature inferences (Heiri and Lotter, 2001, Larocque, 2001 and Quinlan and Smol, 2001)
This directly contradicts the claim in LT15 that
At least 50 head capsules (Heiri et al., 2001 and Larocque et al., 2001) were mounted. (Methods 3.1)
If a couple of levels in the stratigraphy had 48 chironomid head capsules rather than the requisite 50, it would matter not. Unfortunately, the counts are often far lower, ranging from 19 to 68.5. A third of counts are below 30 chironomids; only an eighth of are 50 or more. This matters because small counts are expected to have larger counting errors and hence larger reconstruction uncertainties.More importantly, readers must be able to rely upon claims made by the authors.
The corrigendum provides a new stratigraphy and a new August air-temperature reconstruction which is, like the old one, remarkably good.
LT15 has been cited eight times.