Where the author or a verifier believed the claims file misread the paper, they asked, citing the paper, and a reader who had not seen the data answered from the paper alone, amending the claims file where it was wrong. The author then ran the claim as answered. Attempts before protocol pilot-11 also asked here for method details the claims file did not give.
How is the square root taken for starch, sugar and Asat when some values are negative? The claims file names readings S1 (untransformed) and S2 (square root after adding the constant that brings the minimum to zero) only for negative increments, but the dataset also has negative starch (77 tree-campaign rows), sugar (3) and Asat (5) values, and Table S1 marks starch, sugar and Asat as square-root transformed. Tested here with S1 and S2 applied to these three traits as well. The same question applies to every claim that uses the treatment or interaction test on starch, sugar or Asat (disc-oak-warming-no-carbon-change, fig4-beech-mr, fig4-oak-mr, fig4-beech-wmr, fig4-oak-wmr, fig4-wmr-same-as-mr, disc-nsc-unchanged-drier-soil, tables6-no-seasonal-variation and the claims on Supplemental Table S3).
Passage. Supplemental Tables S1 and S3, the footnote to the starred traits, "square root transformation has been used."
Why it matters. In beech, the W against C contrast of starch in 2020 has adjusted p = 0.041 under S2 and 0.099 under S1 (T1, yearly tree means), which alone decides whether fig4-beech-warming-none holds under every reading.
The reader’s answer. The paper does not say. Its only words on the transform are the footnote "square root transformation has been used" (Supplemental Tables S1, S3 and S6); it mentions no negative values and no constant. The treatment test header now applies S1 and S2 to negative starch, sugar and Asat values as to negative increments, and adds (S3) negative values dropped, as R's sqrt gives NaN for them. fig4-beech-warming-none, disc-oak-warming-no-carbon-change, disc-nsc-unchanged-drier-soil, the fig4 claims and tables6-no-seasonal-variation now name S1, S2 and S3; the interaction test uses the same readings. Claims and expected values unchanged.
Which years enter the interaction model (lme(trait ~ (warming + moisture reduction + year + interaction)^3, random = ~1|Chamber))? The claims file gives readings T1 (2019 to 2021) and T2 (2020 and 2021) for the treatment test but none for the interaction test. Tested here under both. The same question applies to every claim on Supplemental Tables S2 and S3 and to oak-mr-no-interaction.
Passage. Methods, Statistical analysis, paragraph 3, "a second linear mixed-effects model was conducted for both species using an additional species interaction effect (monocultures vs. mixtures) as a fixed effect"; the first model (paragraph 2) uses year 2019, 2020 and 2021.
Why it matters. The number of significant mixture against monoculture phenology contrasts changes with the years (8 under T1 and 3 under T2 with contrasts per year), and so can any claim that counts them.
The reader’s answer. The text does not say; paragraph 3 adds the interaction factor to the first model, which lists "year (2019, 2020, and 2021)", while the set-up says "The first year of measurements (2019) was not included in this study". The F and p pairs of Table S2 settle it per trait. Year terms have 1 numerator df for bud swelling and development duration (oak Y x I "9.37 (0.003)") and 2 df for onset of senescence, senescence duration and growing season length (beech onset Y x I "0.18 (0.837)"); all Table S3 traits have 2 df (beech diameter Y x I "3.08 (0.047)"). The interaction test header now names (Y1) bud swelling and development duration on 2020 and 2021 and the other traits on 2019 to 2021, used by the criteria, and (Y2) 2020 and 2021 for all, as a variant. Table S1 likewise fits three years (beech Asat W x Y "5.22 (0.006)"), which supports T1. Claims and expected values unchanged.
Which years and which F tests? The criterion leaves the years open (all of 2019 to 2021, or 2020 and 2021), and the claims file reads the F tests of the supplement as nlme's sequential anova, with marginal tests as a variant. Fitted on the deposited data, the marginal tests reproduce Supplemental Table S1 (oak AGWB increment exactly on yearly tree means, beech Asat within 0.25 of every F on campaign values), while the sequential tests differ by up to 68 in F. Should Table S6 (and the Table S1 terms the criteria report) be read as marginal tests?
Passage. Supplemental Table S6 footnote, the full model "lme(Measured traits ~ (warming + moisture reduction + Season)3, random = ~1|Chamber)"; Supplemental Table S1, F values per term.
Why it matters. With sequential tests the season term has p <= 0.05 for at least one of the three traits under all 24 readings tested (not reproduced); with marginal tests on 2020 and 2021 monocultures, pooled or oak, all three have p > 0.05 (the claim would hold under those readings).
The reader’s answer. The paper does not say. Table S6 gives only "Full model: lme(Measured traits ~ (warming + moisture reduction + Season)3, random = ~1|Chamber)" and calls the tests "ANOVA tests"; it links no code, so sequential or marginal is not settled, and Table S6 has no year or species column. The Methods tie Table S6 to the treatment test ("we averaged them per year before the analyses (Supplemental Table S6)"), which is on monocultures with years 2019 to 2021. The tables6-no-seasonal-variation criterion now names the years (y1 2019 to 2021, y2 2020 and 2021) and the F tests (F1 sequential, F2 marginal) as readings, with the outcome the one all readings give. That the marginal tests match Table S1 on the data is a finding of the attempt, which a reader cannot confirm from the paper. Claims and expected values unchanged.
Over which family are the Tukey-type mixture against monoculture contrasts of the interaction test adjusted? The claims file names readings I1 (within each treatment and year) and I2 (over 2020 and 2021) but no family. Tested here as (A1) each contrast alone, unadjusted, and (A2) single-step over the model's eight treatment and year contrasts (I1) or four treatment contrasts (I2), as the treatment test adjusts over the six pairs of a year. A second detail for this claim. Under S3 (negative increments dropped), is the mixture minus monoculture difference compared with 1.4 taken over the rows the model keeps, or over all trees? The same adjustment question applies to every claim on Supplemental Tables S2 and S3 and to oak-mr-no-interaction.
Passage. Methods, Statistical analysis, paragraph 3, "Tukey-type post-hoc tests were used to reveal significant differences between mixtures (multcomp function)."
Why it matters. tables3-oak-wmr-diameter holds only under Y1, S3 and A1 with the difference over the rows S3 keeps (+1.38, p 0.023). Under A2 the same contrast has p 0.16, and over all trees the difference is +1.69, outside 1.4 +- 0.14. The verifier of round 1 raised both points.
No reader has answered it yet.
The claims file reads Eq. 3 of Supplemental Methods S4 as [N x (FO x WO / FE - WO) x FE] x 100 with N = 2, applied to one species. With one species this is exactly twice the NBE of Eq. 2, so the complementarity effect cannot be tested apart from the NBE. Its reading of Eq. 4 makes the selection effect exactly 0 under B1 with L1, as the reference FE is the same for every tree of a group. Is that the paper's computation, or should CE and SE be computed over both species of the mixture, as Loreau and Hector (2001) define them? The same question applies to fig5-oak-agwb-nbe-warming-2020, fig5-oak-asat-nbe-wmr-2021 and figs2-beech-asat-selection-warming.
Passage. Supplemental Methods S4, Eqs. 2 to 4 (garbled in the extracted text, as the claims file notes), and the Results' "explained by complementarity" and "positive selection effect".
Why it matters. The "explained by complementarity" parts of three claims are not independent tests under the current reading, and figs2-beech-asat-selection-warming can be tested only under B2 or B1 with L2. The verifier of round 1 reported this as a misreading of the paper by the claims file. Under this reading NBE is also not CE + SE. For oak Asat under WMR in 2021, B2 with L2 gives NBE +194, CE +388 and SE -275 (p 0.0039), as the verifier of round 2 pointed out.
No reader has answered it yet.
The criterion defines growing season length as bud swelling to onset of senescence, as the Methods do. The dataset's Growth_length ("Growing season length", no definition given) never equals that difference (median 55 days longer), and it gives slopes of 4.46 and 5.65 days per degree, within 0.04 of the paper's 4.5 and 5.7. The criterion's definition gives 1.09 (beech, p 0.085) and 2.16. Which is to be tested, the deposited Growth_length or the difference of the two dates? The same applies to reading (b) of fig1-oak-steeper-advance.
Passage. Methods, Phenology, "For each tree, the growing season length was calculated as the number of days between the bud swelling stage and the onset of senescence"; Figure 1I and 1J.
Why it matters. The outcome of fig1ij-growing-season is reproduced with the deposited column and not reproduced with the criterion's definition. The verifier of round 1 raised it.
No reader has answered it yet.