Calibrations without raw data—A response to “Seasonal calibration of the end-cretaceous Chicxulub impact event”

Primary data

No primary isotopic data are openly shared, either with the article itself or through a linked online repository. The publication also lacks a Data Availability Statement (DePalma et al., 2021), despite such a statement being a requirement for publication in Scientific Reports.

Analytical facility

The facility where the stable isotopic analyses were carried out is not revealed, nor are the dates when the experiments were conducted. Curtis McKinney, who is stated to have carried out the analyses (DePalma et al., 2021), passed away in 2017 and cannot be consulted about the referred analytical work.

Methods

The isotopic section of the Methods declaration is brief and lacks a sufficiently detailed account of the techniques, sample weights, and adopted standards necessary for replicating the analyses. To adhere to best practices in reporting, the following methodological details should be provided:

• Description of the analytical facilities used and the specific location of the laboratory where the analyses were conducted;

• Estimation of sample weights (in milligrams or micrograms) after extraction;

• Description of sample treatment, including the acids and bases used, as well as the procedures followed;

• Identification of carbonate standards utilised for calibration purposes;

• Specification of the analytical precision achieved.

These methodological details are conventionally declared in published literature, as evidenced by studies cited (Schulte et al., 2010; During et al., 2022; During, Voeten & Ahlberg, 2022; Robson et al., 2016; DeNiro & Epstein, 1978; Hobson, 1999; Fry, 1981; Fry & Sherr, 1989; Finlay, 2001; Hobson, Piatt & Pitocchelli, 1994; Schell, Saupe & Haubenstock, 1989; Mizutani et al., 1990; LeBreton, Beamish & McKinley, 2004; Brand et al., 2014; Cullen et al., 2019; De Rooij et al., 2022; Coplen, 2011; Bryant et al., 1996; Vennemann et al., 2001; Hodgkins et al., 2020; Passey et al., 2005; Breitenbach & Bernasconi, 2011; Torres, Mix & Rugh, 2005).

Sampling density and amount of carbon

Stable isotope graphs in DePalma et al. (2021) imply that, in some cases, up to 43 samples must have been successfully obtained along an 800 µm transect. Since the drill bits used for sampling are conical in shape, drilling deeper to yield more material inevitably also widens the drill holes. Caution thus needs to be exercised as to avoid intersection of the drill holes and prevent contaminated spot sampling. A total of 43 individual samples yielded along an 800-µm-long transect corresponds with a maximum drill bit width of 18.60 µm, which equates to much less than half the diameter of a typical human hair. Only 10% by weight of hydroxyapatite consists of the structural carbonate (Torres, Mix & Rugh, 2005) that is available for stable isotope analysis. Since the amount of material required for reliable stable isotope analysis varies across the potentially available analytical techniques, the protocols used for obtaining sufficient material as well as for conducting isotopic analyses at the reported spatial resolution require explicit specification. Although the reported resolution may have theoretically been achievable prior to 2017, when the co-author declared responsible for the analyses passed away, it must always have involved a novel approach or modification from an existing protocol. The parameters involved require a declaration to ensure reproducibility, but such a specification is lacking here (DePalma et al., 2021).

Graphs in the article and in the Supplementary Information

Figure 2 of the DePalma et al. (2021) contains the stable isotope record (43 sampling spots) and osteohistology of specimen number FAU.DGS.ND.755.57.T, which is a sturgeon pectoral fin spine. However, this exact specimen number is also declared in the supplemental information of DePalma et al. (2021) associated with a different graph that only involves 35 sampling spots. That graph also includes a line dip without a marked data point, which may be a 36th sampling point, but this has not been clarified. Multiple inconsistencies are best explained as graphs (DePalma et al., 2021) that were produced by image-handling software. Since the article does not provide the original data (DePalma et al., 2021), these results cannot be corroborated or understood. The referred issues described below are also illustrated in Figs. 1–9 of this article.

According to the figures in DePalma et al. (2021), each sample location yielded hydroxyapatite samples for stable carbon and oxygen isotope analyses. The methods (DePalma et al., 2021) declare that these samples were analysed using a Gas Bench II linked to a Thermo Finnigan dual-inlet MAT 253 Stable Isotope Ratio Mass Spectrometer (DePalma et al., 2019). The analytical setup in this study (DePalma et al., 2021) mirrors the one utilised at the Vrije Universiteit Amsterdam (During et al., 2022). However, our experience (During et al., 2022) indicates that this configuration does not allow for the high-resolution sampling of such small samples (DePalma et al., 2021), unless additional measures are taken, such as temporarily cryo-focussing of the produced CO₂ with liquid nitrogen (During et al., 2022). Notably, such an approach was not described or declared in DePalma et al. (2021).

Carbonate, in this case as a component of the hydroxyapatite samples, undergoes a reaction with orthophosphoric acid (H₃PO₄), leading to the generation of CO₂ gas (Torres, Mix & Rugh, 2005). Subsequently, this sample gas is introduced into the mass spectrometer, where simultaneous measurements are taken for stable carbon and oxygen isotope ratios (Torres, Mix & Rugh, 2005). Consequently, a single measurement provides ratios between ¹²C and ¹³C, as well as between ¹⁶O and ¹⁸O, derived from the CO₂ gas within a single sample. The expected outcome is a data table capturing both carbon and oxygen isotopic ratios for each sampling point (Torres, Mix & Rugh, 2005).

Graphing software is then employed to generate two spot values, one for oxygen and one for carbon, precisely aligned on the same vertical line (Breitenbach & Bernasconi, 2011). However, it is important to recognize that analyses may face challenges due to various factors. Failures could stem from a breach of vacuum or low-amplitude measurements, compromising the reliability of both carbon and oxygen measurements. Alternatively, if the analysis of either carbon or oxygen fails, such as due to high inter-peak variation in the measurement of either element, the other element can still be considered reliable and plotted independently.

It is noteworthy that only ²/₃ of the oxygen from carbonate (CO₃²⁻) is converted into the CO₂ that can be measured by the isotope ratio mass spectrometer. Consequently, the oxygen isotope values are highly sensitive to conditions during acid digestion, particularly temperature. Interestingly, some samples omit both carbon and oxygen isotopic values as data points (DePalma et al., 2021). When isotopic values of one of the elements are omitted, it is advisable to at least provide a threshold value and a possible explanation for these failures.

In the graphs presented in the supplemental materials of DePalma et al. (2021) we observe multiple instances where only one of the two isotope measurements is displayed, and in many cases, the two spots are not perfectly vertically aligned (see Figs. 1–3 in this article). The misalignment observed in these graphs suggests that they may not be based on isotopic measurements from the same CO₂-molecule. Alternatively, they were not produced by graphing software. Conversely, in three places we encounter double and widely separated measurements of one of the two isotope ratios on the same vertical line. Double measurements of a single isotope, visualising both a minimum and maximum in their range, supposedly from a single sample point, as depicted in our Figs. 2 and 3, cannot be attributed to technical readouts. Furthermore, in the top right graph of the Supplemental Materials 9 of DePalma et al. (2021) (see our Fig. 4), the line graph exhibits two noticeable ‘dips’ without corresponding sampling points or error bars, and no explanation for these anomalies is provided. The error bars accompanying the carbon isotope records in Fig. 2 of the article (DePalma et al., 2021) (Fig. 5 in this article) differ fundamentally from all the error bars in the carbon isotope records in the supplemental information of DePalma et al. (2021), which seem to be duplicates of the oxygen isotope error bars. Despite the considerable difference in the vertical scale of the graphs, the error bars on both the oxygen and carbon plots are identical in length. Although the Methods section (DePalma et al., 2021), specifies a precision of ±0.3‰, this level of precision is not reflected in the graphs. The error bars are often not centred correctly and display inconsistencies, deviating from the uniform values and appearance expected of analytical software (Figs. 5 and 6 of this article). Moreover, as no carbonate or apatite standard has been specified, the origin of the error estimate remains unexplained. The first author of the article stated that these error bars represent icons (communicated in peer-review) that have been misinterpreted as error bars. This of course may explain their homogeneity, but it is recommended to plot actual error bars in graphs of published articles.

We interpret these characteristics to be inconsistent with direct outputs of analytical graphing software. Moreover, they cannot represent faithful reproductions of machine-generated analytical graphs, as they exhibit features that are inconsistent with copying errors. These features notably include the double measurements of an isotope and use of icons that resemble error bars but do not represent true error estimates. The pattern between the oxygen and carbon value maxima and minima across all referred specimens is extraordinarily consistent (Figs. 1–6 of this article). Achieving such consistency would necessitate perfect sampling of growth intervals without the mixing of bounding layers. Even under ideal conditions, sample heterogeneity as well as analytical uncertainty (Breitenbach & Bernasconi, 2011) will generally prevent reproduction of these exact values in the same individual—let alone in others. Notably, the stable isotope curves consistently exhibit equivalent maximum and minimum values each year in each individual fish, implying a consistent annual diet, a pattern not observed in any comparable analysis known to us.

The overlay image presented in SUP MAT 12 (DePalma et al., 2021) exemplifies the problem. It provides no scale on the x-axis, but verification against one of the source figures, SUP MAT 10 (DePalma et al., 2021) (Fig. 7 in this article), which has such scale bars, reveals features that require explanation. The curves for specimens. FAU.DGS.ND.755.38.T., FAU.DGS.ND.755.22.T and FAU.DGS.ND.755.11.T match almost perfectly when overlaid: over the course of the eight recorded growing seasons, it shows all three individuals depositing exactly 800 µm of bone. While matching sequences of “fat years” and “lean years” are plausible and can be expected, as they represent responses to the same ecosystem that hosted all the individuals, this perfect match of total amount of bone deposition is remarkable. Furthermore, the curves of FAU.DGS.ND.755.70.T appear at first sight not to match those of the three aforementioned specimens, but in fact do so perfectly when aligned to the left margin and stretched rightwards to exactly 150% of original length (Fig. 8 of this article). Similarly, the curves of FAU.DGS.ND.755.31.T fit perfectly onto those of FAU.DGS.ND.755.36.T when aligned to the left margin and shortened to exactly 70% of original length (Fig. 9 of this article). It seems improbable for these specimens to have been growing consistently at rates of 150% faster and 70% slower, necessitating an explanation to confirm the validity of the data.

Fish sizes

Figure 3 (DePalma et al., 2021) illustrates the sizes of “sub-yearling” fishes. In this figure, the number of measured fishes is unspecified, and a corresponding data table providing individual fish measurements is not included. Upon revisiting the supplemental information of DePalma et al. (2021) describing the locality (Torres, Mix & Rugh, 2005), a contradictory graph is shown, suggesting that fish sizes commence at 15 cm. However, in DePalma et al. (2019), raw data and a description of the measurement protocol are lacking.

DePalma et al. (2021) state that “The smallest Acipenseriformes at Tanis (<16 cm fork length) fall below the expected length of yearling extant acipenseriform taxa, and we interpret that they died during sub-yearling ontogeny (i.e., YOY).” This contradicts the depiction of fish sizes in the initial Tanis article (Lucchese & Mitra, 2003), where all fishes are described as being at least 15 cm in length. Additionally, the “Tanis fossil sizes” range indicated in Fig. 3 (DePalma et al., 2021) suggests that the fishes measure between 8–12 cm in length.

Moreover, there is a lack of explanation for the data points in Fig. 3 (DePalma et al., 2021), with several instances of fish with the same body size allocated to different ‘spawning seasons’. While Fig. 1 (DePalma et al., 2021) does include depictions of juvenile fishes, inconsistencies in the scale bars prevent accurate size reconstruction, raising concerns about the reliability of these size claims. For example, the width of section “E” in Fig. 1A is roughly one third of the width of the 3 cm scale bar in the same figure, which makes it impossible for “E” to have a valid 1 cm scale bar that spans less than half its own width. Additionally, the fishes depicted in Fig. 1 (DePalma et al., 2021) do not significantly differ in size from the other works describing fishes extracted from Tanis by Steve Niklas and Robert Coleman in 2010 (Hilton et al., 2023; Hilton & Grande, 2023), which do not mention any small fishes that could be considered hatchlings. During fieldwork in 2017, no sub-yearling fish was observed, and body sizes were not measured or recorded (personal observation M. During, 2017).

Conclusions of a spring death

The histological sections shown in DePalma et al. (2021) contain air bubbles, and the histology itself appears out of focus, indicating that they may have been photographed from the side of the protective glass cover rather than from the side of the sample. Consequently, it becomes challenging to properly assess the histology of the specimens, making it difficult to discern osteocyte distribution (Castanet, 1993; Davesne et al., 2020; De Buffrénil, Quilhac & Castanet, 2021). The approach selected by DePalma et al. (2021) instead focused on the “darker” and “lighter bands”. This represents a more basic approach that is sometimes used in age-at-death estimates of vertebrates (Brennan & Cailliet, 1989). By binarizing a histological section into either “lighter bands” (associated with autumn and winter) and “darker bands” (associated with spring and summer), one can indeed discriminate between death in autumn/winter or spring/summer. However, quantification of osteocyte distribution is more reliable and allows for distinction between the four individual seasons (During et al., 2022; Davesne et al., 2020). Furthermore, the low resolution of the histological sections and the modifications in the graphs do not meet the standards for publication in Scientific Reports.

While the stable carbon isotope record displays oscillations between summer maxima and winter minima, akin to findings in During et al. (2022), the final measurements of all specimens in DePalma et al. (2021) exhibit a maximum, suggesting fishes that perished not in spring but rather in late summer, similar to the previous maxima that are rapidly followed by a dip to the minima. DePalma et al. (2021) conclude that these fishes perished in late spring early summer, but irregularities in the methods and data presented appear sufficient to undermine confidence in either conclusion.