Reddin CJ, Kocsis Á, Kießling W (2020)

**Publication Type:** Journal article, Erratum

**Publication year:** 2020

**Book Volume:** 29

**Pages Range:** 1280-1282

**Journal Issue:** 7

**DOI:** 10.1111/geb.13114

JUSTIFICATION The original results to Reddin, Kocsis, and Kiessling (2018) included an error in the code that intended to match geographical grid cells between neighbouring ages (“matched-cells” hereafter), via a nonsensical call to the subset() function. After correcting this coding mistake, we checked these results against those acquired with the matched-cells step omitted entirely. The results held only when the matched-cells step was omitted. This invalidated our original belief that matched-cells would help “homogenize geographical sampling coverage between consecutive ages”. The restriction to matched-cells results in a median loss of 23% of survivor species compared to without matched-cells (median = 118 survivor species per age to median = 153 without matched-cells). Additionally, we suspect that the two spatial corrections for sampling variation interfere with one another, that is, the use of (a) matched-cells and (b) subtraction of the sampling focus from the median genus range centre (to calculate the latitudinal deviation). We therefore reproduce the analyses without limiting data to matched-cells and present the updated results, including the five changed results figures, here. In summary, the changes to the results are minor and the scientific conclusions remain as originally stated. UPDATED VERSIONS OF KEY FIGURES All results hold following the omission of the matched-cells step, albeit with slightly weaker strength, with the central correlation decreasing from rho = 0.30 (p =.01) to rho = 0.24 (p =.034; Figure 1). Following the order of presentation in the original paper, the central correlation rises, as before, with (a) hemisphere-switching, but to rho = 0.39 (95% CIs = 0.19–0.56, optimum threshold = 126; Figure 2) instead of the original rho = 0.48 (95% CIs = 0.29–0.64, optimum threshold = 108 genera). The central correlation for the Southern Hemisphere remained insignificant (rho = 0.08, p =.6). If (b) the median latitudinal deviation was supplemented with the 0.2-quantile latitudinal deviation, when the Northern Hemisphere median occurrence palaeolatitude vacated the tropics, a substantial improvement to the central correlation is no longer observed (small increase to Rho = 0.26, 95% CIs = 0.04–0.46, rather than the original increase to rho = 0.40, 95% Cis = 0.19–0.57; see next paragraph). However, as in the original version, combining both hemisphere-switching and integrating the 0.2-quantile deviation (steps a and b), raised the central correlation to rho = 0.35 (95% CIs = 0.14–0.53), but did not improve upon using hemisphere-switching alone. (c) Weighting the correlation from the hemisphere-switching approach by the age inverse extinction rate resulted in the highest correlation strength, at rho = 0.41 (95% CIs = 0.2–0.58; Figure 3) instead of the original rho = 0.56 (95% Cis = 0.39–0.70). 1 Figure (Figure presented.) Latitudinal range shifts of Northern Hemisphere marine genera (black lines) are correlated with low latitude mean shallow seawater temperature (red lines, right-hand y axis). Temperature calculated from fossil shell oxygen isotopes following Veizer and Prokoph (2015). Grey shading around the thick black line shows bootstrapped 95% confidence intervals. For both variables, thick lines show a moving window of three consecutive bins. Thin lines show the unsmoothed bin-to-bin changes. This is the updated original Figure 2. Contributing genera median range centre = 23.6°. There is now no gap in the latitudinal deviation time series for the Northern Hemisphere. Latitudinal deviation mean is 0.60° (absolute mean = 1.93°). Geological periods are lightly shaded and abbreviated as O = Ordovician; S = Silurian; D = Devonian; C = Carboniferous; P = Permian; Tr = Triassic; J = Jurassic; K = Cretaceous; Pg = Palaeogene; N = Neogene 2 Figure (Figure presented.) The effect of changing a threshold minimum number of genera in “hemisphere switching” on the correlation of latitudinal deviation with climate. When the minimum threshold is reached, latitudinal deviations are sourced from the opposite hemisphere. Dotted lines show the correlation without any hemisphere switching, for northern and southern palaeo-hemispheres separately. Filled circles are p <.05. Serial autocorrelation in times series removed by generalized differencing. This is the updated original Figure 3. Optimum threshold of sampled genera = 126 3 Figure (Figure presented.) Time series of latitudinal range shifts of marine genera (black lines) improved by hemisphere switching alongside low latitude shallow seawater temperature (red lines, right-hand y axis). Details as in Figure 1. This is the updated original Figure 4. Contributing genera median range centre = 27.6°. Latitudinal deviation mean is 0.38° (absolute mean = 1.89°). All other descriptions as in Figure 1. O = Ordovician; S = Silurian; D = Devonian; C = Carboniferous; P = Permian; Tr = Triassic; J = Jurassic; K = Cretaceous; Pg = Palaeogene; N = Neogene The latitudinal deviation quantiles still show an increase in strength of relationship with climate from higher to lower quantiles (Figure 4, rho = −0.58, n = 9, p =.11; down from the original rho = –0.78, n = 9, p =.02), but the trend is only significant when the sampling focus vacated the tropics, approximately after the Carnian age (rho = −0.83, n = 9, p =.008; Figure 4; also see original Figure 1 at ~ 230 myr). This is presumably because the highest and lowest quantiles are more affected by noise and only become useful when the median sampling focus is no longer opportunely positioned, that is, when the sampling centre vacates the tropics. Prior to the Norian age (Triassic), the correlation between latitudinal deviation and climate with hemisphere switching (i.e. step (a), above) is significant even without smoothing (rho = 0.34, 95% Cis = 0.04–0.58, ages n = 43). The slopes of warm and cool ages remain significantly different to each other (Z-statistic = −3.23, p =.0006; Figure 5). 4 Figure (Figure presented.) Correlation strength between shallow seawater temperature and genus latitudinal range shifts decreases from more equatorward (0.1-quantile, median palaeolatitude = 12.7°) to poleward genera (0.9-quantile, median palaeolatitude = 39.0°; quantile steps of 0.1). Solid symbols show Spearman's rho coefficients that 95% confidence intervals (dashed lines) overlap zero. Serial autocorrelation in times series removed by generalized differencing. This is the updated original Figure 5. The latitudinal means for the coloured lines are the mean sampling latitudes. The strongest correlation over all ages is with the 0.5-quantile (i.e. median, given in main text) but when the quantile time series are split, the post-Carnian 0.2-quantile has the strongest correlation (rho = 0.35, p =.036) 5 Figure (Figure presented.) Poleward migrations are more likely to be recorded by lower latitude sampling foci, (a) especially during warm ages (red line and points, age seawater temperature > Phanerozoic mean, slope = −0.20, p =.0005, n = 33). Each point represents one geological age. Blue line and points represent cool ages (age seawater temperature < Phanerozoic mean, slope = −0.02, n.s., n = 45). (b) Termplots of the independent effects of the predictor variables in a multiple regression including both temperature and sampling latitude as predictor variables for latitudinal deviations. Points are the partial residuals. This is the updated original Figure 6 The interpretation of these results is unchanged. Thanks are owed to Gwen S. Antell for noticing this error. The code was updated in the Dryad repository on January 10, 2020. The authors apologise for this error.

Carl James Reddin
Lehrstuhl für Paläoumwelt
Ádám Kocsis
Lehrstuhl für Paläoumwelt
Wolfgang Kießling
Lehrstuhl für Paläoumwelt

**APA:**

Reddin, C.J., Kocsis, Á., & Kießling, W. (2020). Corrigendum to: Marine invertebrate migrations trace climate change over 450 million years (Global Ecology and Biogeography, (2018), 27, 6, (704-713), 10.1111/geb.12732). *Global Ecology and Biogeography*, *29*(7), 1280-1282. https://dx.doi.org/10.1111/geb.13114

**MLA:**

Reddin, Carl James, Ádám Kocsis, and Wolfgang Kießling. "Corrigendum to: Marine invertebrate migrations trace climate change over 450 million years (Global Ecology and Biogeography, (2018), 27, 6, (704-713), 10.1111/geb.12732)." *Global Ecology and Biogeography* 29.7 (2020): 1280-1282.

**BibTeX:** Download