Volume 27, Issue 6 pp. 704-713

RESEARCH PAPER

Full Access

Marine invertebrate migrations trace climate change over 450 million years

Correction(s) for this article

Carl J. Reddin,

Corresponding Author

Carl J. Reddin

[email protected]

orcid.org/0000-0001-5930-1164

GeoZentrum Nordbayern, Department of Geography and Geosciences, Universität Erlangen-Nürnberg, Erlangen, Germany

Correspondence Carl J. Reddin, GeoZentrum Nordbayern, Department of Geography and Geosciences, Universität Erlangen-Nürnberg, Loewenichstraße 28, 91054 Erlangen, Germany. Email: [email protected]Search for more papers by this author

Ádám T. Kocsis,

Ádám T. Kocsis

orcid.org/0000-0002-9028-665X

GeoZentrum Nordbayern, Department of Geography and Geosciences, Universität Erlangen-Nürnberg, Erlangen, Germany

MTA-MTM-ELTE Research Group for Paleontology, Budapest, Hungary

Search for more papers by this author

Wolfgang Kiessling,

Wolfgang Kiessling

orcid.org/0000-0002-1088-2014

GeoZentrum Nordbayern, Department of Geography and Geosciences, Universität Erlangen-Nürnberg, Erlangen, Germany

Search for more papers by this author

Carl J. Reddin,

Corresponding Author

Carl J. Reddin

[email protected]

orcid.org/0000-0001-5930-1164

GeoZentrum Nordbayern, Department of Geography and Geosciences, Universität Erlangen-Nürnberg, Erlangen, Germany

Ádám T. Kocsis,

Ádám T. Kocsis

orcid.org/0000-0002-9028-665X

GeoZentrum Nordbayern, Department of Geography and Geosciences, Universität Erlangen-Nürnberg, Erlangen, Germany

MTA-MTM-ELTE Research Group for Paleontology, Budapest, Hungary

Search for more papers by this author

Wolfgang Kiessling,

Wolfgang Kiessling

orcid.org/0000-0002-1088-2014

GeoZentrum Nordbayern, Department of Geography and Geosciences, Universität Erlangen-Nürnberg, Erlangen, Germany

Search for more papers by this author

First published: 03 April 2018

https://doi.org/10.1111/geb.12732

Citations: 26

Funding information: VolkswagenStiftung; Deutsche Forschungsgemeinschaft, Grant/Award Number: KI 806/16-1 and KO 5382/1-1

Share a link

Email
Wechat
Bluesky

Abstract

Aim

Poleward migration is a clear response of marine organisms to current global warming but the generality and geographical uniformity of this response are unclear. Marine fossils are expected to record the range shift responses of taxa and ecosystems to past climate change. However, unequal sampling (natural and human) in time and space biases the fossil record, restricting previous studies of ancient migrations to individual taxa and events. We expect that temporal changes in the latitudinal distribution of surviving taxa will reveal range shifts to trace global climate change.

Location

Global.

Time period

Post-Cambrian Phanerozoic aeon.

Major taxa studied

Well-fossilized marine benthic invertebrates comprising stony corals, bivalves, gastropods, brachiopods, trilobites and calcifying sponges.

Methods

We track deviations in the latitudinal distribution of range centres of age boundary crossing taxa from the expected distribution, and compare responses across latitudes. We build deviation time series, spanning hundreds of million years, from fossil occurrences and test correlations with seawater temperature estimates derived from stable oxygen isotopes of fossils.

Results

Seawater temperature and latitudinal deviations from sampling are positively correlated over the post-Cambrian Phanerozoic. Simulations suggest that sampling patterns are highly unlikely to drive this putative signal of range shifts. Systematically accounting for known sampling issues strengthens this correlation, so that climate is capable of explaining nearly a third of the variance in ancient latitudinal range shifts. The relationship is stronger in low latitude taxa than higher latitude taxa, and in warm ages than cool ages.

Main conclusions

Latitudinal range shifts occurred in concert with climate change throughout the post-Cambrian Phanerozoic. Low latitude taxa show the clearest climate-migration signal through time, corroborating predictions of their shift in a warming future.

1 INTRODUCTION

The movement of species to previously colder areas may be the most dependable response of organisms to climate warming (Chen, Hill, Ohlemüller, Roy, & Thomas, 2011; Parmesan & Yohe, 2003; Poloczanska et al., 2013). While fine-scale range shifts responding to regional climate change are not always directed poleward (Burrows et al., 2014), in the oceans, climate-induced latitudinal shifts (‘migrations’ henceforth) are already dramatic (García Molinos et al., 2016; Sorte, Williams, & Carlton, 2010). Poloczanska et al. (2013) observed an average rate of 72 km/decade at which poleward range edges expand, although the movement of centroids is only half of that, mostly because trailing edge contractions are still moderate. Nevertheless, these migrations indicate that even modest changes in global sea surface temperature (0.07 °C/decade; Burrows et al., 2011) can have dramatic effects on the latitudinal distribution of marine animals. Marine ectotherms are expected to track coherently their thermal limits (Sunday, Bates, & Dulvy, 2012), especially because their thermal niches are generally conserved over macro-evolutionary time-scales (Saupe et al., 2014) and there are hard limits to adaptation (Storch, Menzel, Frickenhaus, & Pörtner, 2014). However, rates of migrations vary among species and higher taxa such that communities are unlikely to migrate cohesively, changing community composition (Graham et al., 1996; Valentine & Jablonski, 1993; Walther et al., 2002). Consequently, this redistribution of biodiversity is expected to alter ecosystem functioning and, thereby, human well-being (Pecl et al., 2017). Therefore, the separation of general and contingent aspects of biodiversity responses to climate warming is a key goal for ecologists, enabling confident prediction of future developments.

The current expected magnitude, and perhaps even rate, of warming may not be without equal in the Phanerozoic aeon (Kemp, Eichenseer, & Kiessling, 2015; Zachos, Dickens, & Zeebe, 2008; but see Foster, Royer, Lunt, Olsen, & Beerling, 2017). Organisms have endured global cooling intervals lasting several millions of years (Royer, Berner, Montañez, Tabor, & Beerling, 2004) and shorter intervals of excessive tropical heat (Song et al., 2015; Sun et al., 2012), both forced partly by atmospheric CO₂ (Foster et al., 2017; Royer, 2006). Tropical taxa may be the most responsive to long-term climate warming because high development rates enable rapid responses (Walters, Blanckenhorn, & Berger, 2012). They also often live close to maximum thermal limits set by phylogeny (typically < 35 °C, never exceeding 45 °C for metazoans; Song et al., 2015; Storch et al., 2014; Walther et al., 2002). In response to climatic stressors, studies show examples of fossil taxa ostensibly migrating by latitude both for near time (e.g. Greenstein & Pandolfi, 2008; Kiessling, Simpson, Beck, Mewis, & Pandolfi, 2012; Roy, 2001) and deep time (e.g. Blois & Hadly, 2009; Martin-Garin, Lathuilière, & Geister, 2012; Thibault & Gardin, 2010). However, the consistency of global biodiversity responses to warming events is unproven. The fossil record is unique in its capacity to record how organisms responded to such ‘natural experiments’ of the past. Yet, numerous issues exist regarding the temporal and palaeo-geographical precision of fossil occurrences, which can distort evidence for global ecological patterns in deep time (e.g. Clapham & James, 2012). Spatio-temporal biases in sampling are the main driver of patterns in the fossil record and need handling correctly in order to identify ecological patterns (Vilhena & Smith, 2013).

Here we analyse the link between net latitudinal range shifts of marine invertebrates from geological age-to-age and seawater temperature. This scale emphasizes long-term net migration in response to long-term climatic processes. The effect on range shifts due to processes operating at finer scales, such as Milanković cycles, are thus averaged out. We expect latitudinal range shifts, either of individual fossil taxa or else averaged over many, to track global climate change, which we quantify independently. We expect this relationship to be clearest in low latitude taxa during relatively warm ages, as these taxa often live close to metazoan thermal maxima and warm global climate intervals may have forced many taxa to vacate the tropics. Three factors anticipated to affect the clarity of range shifts are the median occurrence latitude, number of sampled taxa and extinction rate. The first concerns which taxa are used to characterize the proxy for range shifts, while the second and third concern the observed number of taxa that could shift their ranges.

2 MATERIALS AND METHODS

2.1 The database

We downloaded occurrence data of fossil marine invertebrates on 8 February 2017 from the most comprehensive dataset of fossil metazoans currently available to science, the publicly-accessible Paleobiology Database (https://paleobiodb.org/). This used the standard download protocol except it excluded genera identified with some degree of uncertainty (‘aff.’, ‘cf.’, ‘?’). These data have their palaeo-coordinates calculated based on the GPlates plate tectonic model (Wright, Zahirovic, Müller, & Seton, 2013). Fossil collections were binned by their central age estimate to one of 94 geological ages (Gradstein, Ogg, Schmitz, & Ogg, 2012), averaging 6.12 Ma in duration. Collections with age uncertainties greater than the longest age duration (> 19.5 Ma) were removed. Similar to comparable studies of ancient geographical ranges (Foote, 2014), we assess range shift patterns at the genus level. This taxonomic level is a compromise between uncertainty in the species level taxonomy of fossils and data loss at coarser taxonomic resolutions (Valentine, 1974). Furthermore, species-level information is inconsistent and sparse in the Paleobiology Database and species longevities are typically below our temporal resolution. Duplicate genus occurrences per collection were removed to make collections more comparable (i.e. between collections with species level identified or not identified).

We included marine benthic invertebrate classes with good fossil records owing to their heavily calcified body parts, including stony corals, bivalves, gastropods, brachiopods, trilobites and sponges such as Stromatoporoidea. The Cambrian and older periods were removed because of sparse temporal and geographic coverage of both occurrence and climate proxy records. After filtering, 458,000 occurrences in 76,000 collections were available for analyses.

2.2 Sampling and taxa location shifts

We initially conceived various candidate proxies for ancient taxon range shifts. Initially, these included tracing age-to-age changes in surviving genus (i.e. that occurred in the previous age and the focal age) palaeolatitude, by range edges or centre, corrected for changes in sampling (occurrence) patterns (details in Supporting Information Appendix S1). Briefly, we concluded that data for specific genera may be too noisy to track across ages. Although range shifts are expressed at the individual genus level, a net range shift approach is required to overcome sampling inconsistencies and random noise. Therefore, we decided to average their range centres per age, rather than the age-to-age changes in them, maintaining our focus on survivors. The time series variance of median fossil genus range centres or edges is always dominated (R² > .9) by sampling patterns (human and geological; Figure 1a and Supporting Information Appendix S2; also see Clapham & James, 2012). Therefore, we accepted sampling as the dominant driver of range shift patterns (Supporting Information Appendix S3 for further justification; cf. Powell, 2009; Vilhena & Smith, 2013). All candidate proxies for fossil range shifts were assessed by backward selection, based on Akaike information criterion (AIC) values, with median occurrence latitude, temperature and shifts in occurrence median latitudes as predictor variables. The occurrence distribution and individual distributions of fossil genera per age were typically skewed and multimodal so we decided a priori to use the median rather than the mean. Throughout most of the Phanerozoic, the northern palaeo-hemisphere is the best sampled (Figure 1; Vilhena & Smith, 2013) and thus we test our hypotheses on this hemisphere.

Details are in the caption following the image — **Figure 1**
Open in figure viewer PowerPoint

The palaeolatitudinal distribution of fossil occurrences (a) through the post-Cambrian Phanerozoic and (b) a schematic of how this is used to calculate the ‘latitudinal deviation’. (a) Each occurrence (n = 469,000) is shown as a black dot by its central age estimate. Global (grey line) and per hemisphere (black lines) trends are shown by median occurrence palaeolatitudes binned to geological ages and plotted at their mid-points. Blue lines show the median latitudinal range centre of genera per hemisphere (blue line minus black line gives the genus latitudinal deviation from sampling focus). Timescale after Gradstein et al. (2012); geological periods (shaded) are abbreviated along the x axis: O = Ordovician; S = Silurian; D = Devonian; C = Carboniferous; P = Permian; Tr = Triassic; J = Jurassic; K = Cretaceous; Pg = Palaeogene; N = Neogene. A hypothetical time slice, i, from (a) is shown in (b) for a single hemisphere to illustrate calculation of median and other latitudinal deviation quantiles. Note latitude is on the y axis in (a) and x axis in (b). In blue is a hypothetical distribution of genus range centres, a, of which any quantile can be calculated, the median being the .5-quantile. In grey is a hypothetical distribution of occurrences, b. For whichever quantile chosen, *a – b* gives the deviation of observed genus range centres from the ‘expected’ range centre, termed the ‘latitudinal deviation’

Global temperature was selected as an explanatory variable only for one candidate proxy (Supporting Information Appendix S2). This was the deviation of a_i, the median of genus latitudinal range centres in time slice i, from b_i, the median latitude of occurrences in time slice i (also termed sampling ‘centre’ or ‘focus’; Figure 1b). Here, b represents the null-expected median range centre of genera. Thereby, a_i – b_i is henceforth termed the ‘latitudinal deviation’ and a time series is built. This proxy is equivalent to the residuals from a regression. To maximize this proxy's effectiveness, we are interested in contrasting sampling bias (via b, the occurrence distribution) against ecological patterns (via a, the median genus range centres), so that the latitudinal deviation (a – b) represents bias-corrected ecological signal. b should therefore aim to represent the spatial biases as best possible, which we hypothesize is by the following steps. Supporting Information Appendix S2 discusses further how a latitudinal deviation time series describes genus range shifts from age to age.

To calculate genus range centres, a, we used only those genera that were sampled in two consecutive ages (bin-to-bin survivors, which were thus able to migrate). Furthermore, in an attempt to homogenize geographical sampling coverage between consecutive ages (i.e. sampling effort), we used matching cells to calculate a (Kiessling & Kocsis, 2016). Genus range centres were thus calculated from occurrences in only those geographical cells that were sampled in both consecutive time intervals. Grid cells were equal-area hexagons constructed using the R package icosa (Kocsis, 2017), with a mean cell area of 794,000 km². Genus range centres were calculated separately for the Northern and Southern Hemispheres, to minimize inter-hemispherical differences (e.g. sampling coverage, genus ranges, regional climatic phenomena, continental shelf availability; Figure 1a). The Southern Hemisphere values were useful to supplement intervals when the Northern Hemisphere is poorly sampled (see below). To calculate sampling latitude, b, we used all marine fossils in our dataset, including stratigraphical singletons and without matched cells. Median sampling latitude was uncorrelated with climate (always p > .2; Supporting Information Appendix S4).

To assess differences in range shift responses between lower and higher latitude genera, we investigated different quantiles of the distribution of survivor latitudinal range centres per age, in addition to the median (quantiles in steps of .1, Figure 1b). These were used to calculate quantile latitudinal deviations following the previous approach, by adjusting the genus range centre quantile by the appropriate quantile of the sampling distribution for a time slice. For example, the .2-quantile of sampling latitudes was subtracted from the .2-quantile of genus latitudes to provide the .2-quantile latitudinal deviation (a_.2-quantile – b_.2-quantile). Time series of each quantile could thus be correlated with temperature. We modelled the behaviour of these quantile deviations (Supporting Information Appendix S5) and found that their correlations with independent climate data can be compared without problem. We also directly investigated sampling focal latitude as a source of error on latitudinal deviations. Standardizing variances of both occurrence time series and range centre time series, for each quantile separately, had little effect on quantile latitudinal deviations and their correlations with temperature (Supporting Information Appendix S5).

Inspection of Figure 1 suggests that the signal of climate-induced range shifts might be refined further by addressing particular sampling biases (listed 1–3). The global sampling focus drifts through time with the tectonic movement of the well-sampled modern-day Europe and North America, into the Southern Hemisphere prior to the Permian (Figure 1, grey line). Therefore (1) The Northern Hemisphere number of surviving genera was sometimes low (< 100, Supporting Information Figure S1). Switching the focal hemisphere, that is, replacing deviations in one hemisphere with those of the other, was systematically and automatically explored by setting thresholds for a minimum number of surviving genera, from one to 200. Thus, ages with surviving genera numbers below this threshold would switch to the other hemisphere to calculate deviations. The effect of this hemisphere switching was quantified for both hemispheres separately. (2) Across the post-Cambrian Phanerozoic, the median sampling latitude was 27.8° but it distinctly rose from the late Triassic onwards (Supporting Information Figure S2, absolute medians before late Triassic = 20.2°, after = 34.5°). We expected this rise to weaken the signal of climate induced range shifts, following our hypothesis that low latitude taxa should show clearer migratory responses to climate. Therefore, to maintain a low latitude sampling focus, we made a time series integrating the median with the strongest-related lower latitude quantile, using the former in ages when sampling focus was tropical (median sampling latitude < 23.4°) and the latter when the median sampling latitude was subtropical or higher (> 23.4°). This decreased the median palaeolatitude of genus occurrences to 19.9°. (3) We expected higher extinction rates to decrease the signal of climate-induced range shifts. We therefore used a weighted correlation test (see Statistical analysis). There is less than 1% chance that the climate-deviation correlation strengths resulting from the above refinements were caused by random entry of new collections (or < 6% chance for Northern Hemisphere only; Supporting Information Appendix S4).

2.3 Climate proxy data

To reconstruct temperature change over time, we use arguably the best available proxy for ‘deep time’: the δ¹⁸O values of well-preserved calcareous shells (but see Ridgwell, 2005; Royer et al., 2004). We downloaded the whole surface water (mixed layers < 300 m deep) oxygen isotope dataset of Veizer and Prokoph (2015). Oxygen stable isotope measurements from well-preserved calcareous fossils are traditionally used to reconstruct seawater temperatures but the oxygen isotopic composition of seawater needs to be known (Veizer & Prokoph, 2015). To adjust for the long-term trend in oxygen isotopic composition of seawater, we follow Veizer and Prokoph (2015) and detrended the time series using the equation: δ¹⁸O_pw(‰) = −0.00003‰ t² + 0.0046‰ t, with pw being Phanerozoic seawater in standard mean ocean water (SMOW) and t being age in Ma. For any given age, spatial variation in the oxygen isotopic composition of seawater, such as that attributable to salinity and pH variation (Grossman, 2012; Ridgwell, 2005), is expected to contribute mostly to variance rather than mean estimates (Veizer & Prokoph, 2015). Variation attributable to ice volume is relatively minor and relates to climate (Veizer & Prokoph, 2015). Tropical (including subtropical) records dominate the Veizer and Prokoph (2015) dataset (75% of all records). To optimize the representativeness of these data to the habitat of genera in our dataset, we only use data from tropical and subtropical palaeolatitudes. However, during the whole Jurassic Period to the Barremian age, when temperate isotope records dominate (Supporting Information Figure S3), we averaged temperate, subtropical and tropical records to ensure there was sufficient data over which to average. We binned isotope data to geological ages and computed their medians. The Induan age is not represented by any samples in this dataset so its temperature was interpolated between neighbouring age medians. We translated δ¹⁸O into degrees Celsius in all cases, using the transfer function in Visser, Thunell, and Stott (2003; recommended in Veizer & Prokoph, 2015), representing low latitude upper layer seawater.

2.4 Statistical analysis

Unless otherwise noted, hypothesis testing is based on nonparametric statistics such as Spearman's rho. Autocorrelation of time series data was removed using generalized differencing (McKinney & Oyen, 1989) [Δx = x_i – r_A(x_i_–1), where i is the time interval and r_A the lag 1 autocorrelation coefficient]. We employed a moderate smoothing of latitudinal deviations and temperature using a symmetrical moving window of three consecutive bins for analysis, unless otherwise noted. This approach dampens the influence of various sources of error, including sample temporal accuracy, global representativeness and coverage of sampling distribution, and emphasizes long-term changes. Bootstrapped 95% confidence intervals (CIs) around the latitudinal deviation were calculated by resampling with replacement, for a from the list of genera range centres per age and for b from the occurrence latitudes per age. We performed 10,000 resamples and extracted the .025 and .975 quantiles from the distribution of a–b. The 95% CI time series are smoothed, as above. Ages that were more influential to the correlation tend to have narrower CIs (Supporting Information Appendix S2). Weighting of the correlation (package wCorr) was by inverse per capita extinction rate, uncorrected for age duration (Foote, 2000; smoothed, as for the other data), with all other details remaining the same as before. To investigate the relationship between sampling focal latitude and genus latitudinal deviation, we compare least squares regression slopes where residuals are approximately normal. This used the Z-statistic calculation, Z = b₁ – b₂/√( $urn:x-wiley:1466822X:media:geb12732:geb12732-math-0001$ + $urn:x-wiley:1466822X:media:geb12732:geb12732-math-0002$ ), where b₁ is the regression slope of sample set 1 and SEb₁ is the standard error of that slope (Cohen, Cohen, West, & Aiken, 2003). All analyses were performed in the R statistical computing environment (R Development Core Team, 2017).

3 RESULTS

3.1 Latitudinal deviations as climate-induced range shifts

Genus median latitudinal deviations of the better-sampled northern palaeo-hemisphere are positively correlated with seawater temperature across the post-Cambrian Phanerozoic (rho = .30, p = .01; Figure 2). This supports our hypothesis of linked climate change and latitudinal range shifts, although further validation of the potential signal is desirable to support, or deny, its existence. We anticipated three factors to affect clarity of the signal: (a) systematically supplementing the Northern Hemisphere latitudinal deviations with occurrences from the Southern Hemisphere when the number of sampled surviving genera became low vastly increased the correlation to rho = .48 (95% CIs = .29–.64, optimum threshold = 108 genera, Figure 3). The Southern Hemisphere remained uncorrelated with climate (Figure 3), even when restricted to the well-sampled Palaeozoic (rho = .22, p = .2). Additionally, (b) when the Northern Hemisphere median occurrence palaeolatitude vacated the tropics (> 23.4°; Supporting Information Figure S2), this vacation was corrected for by integrating the .2-quantile deviation, transcending the Northern Hemisphere correlation strength of either the median or .2-quantile alone with climate (rho = .40, 95% CIs = .19–.57). However, combining these two steps, (a) and (b), led to no further improvement in the correlation strength (rho = .40, 95% CIs = .19–.57, Figure 4). Finally, (c) weighting the correlation from step (a) (the highest so far) by age inverse extinction rate increased the correlation strength between seawater temperature and genus latitudinal deviation further (rho = .56, 95% CIs = .39–.70). The substantial increase in the correlation strength achieved by considering these a priori factors supports the validity of the latitudinal deviations as revealing ecological signal. Exploring the contribution of individual taxa to the correlation between temperature and latitudinal deviation suggests brachiopods to be particularly important (Supporting Information Appendix S6).

3.2 Low versus high latitude genus responses to climate change

Genus latitudinal distribution quantiles vary systematically in their latitudinal focus such that the .1- and .2-quantiles tend to represent palaeotropical taxa, having median range centres of 12.5° and 16.2°, respectively. Higher quantiles represent higher latitude taxa, with median range centres up to 41.4° for the .9-quantile. Latitudinal deviation quantiles generally show an increase in strength of relationship with climate from higher to lower quantiles (rho = –.78, n = 9, p = .02; Figure 5; quantiles without hemisphere switching). 95% CIs around the correlation coefficient overlap zero for higher latitude taxa with range centres > 24.5°, although .8- and .9-quantiles appear particularly susceptible to random noise (Supporting Information Appendix S5).

The latitudinal focus of occurrences changes over time, which permits an alternative test of differences in magnitude of latitudinal deviation between low and high latitude taxa. The magnitude of latitudinal deviation is significantly and negatively related to sampling latitude (rho = –.34, p = .002, unsmoothed data, Figure 6), such that sampling foci at lower latitudes more consistently show larger and positive (i.e. poleward) genus latitudinal deviations. When warm and cool ages are analysed separately, warm ages contributed more to the negative relationship between focal latitude and observed deviation, with significantly different warm and cool age regression slopes (Z-statistic = −2.65, p = .004; Figure 6a). The effects of sampling latitude and temperature on latitudinal deviation are largely independent (Figure 6b).

4 DISCUSSION

Latitudinal deviations for marine invertebrate genera consistently trace global changes in seawater temperature throughout the post-Cambrian Phanerozoic. Global temperature explains up to 31.4% of the variance in putative range shifts (by rho squared, the variance shared between ranked variables), which is even more than observed for recent marine organisms (11% of variance, Poloczanska et al., 2013). This is strikingly high given the dating issues in both fossil occurrence and climate proxy data. Simulations of sampling biases (Supporting Information Appendix S4) suggest that this correlation is highly unlikely to emerge from sampling. Thus, we interpret genus median latitudinal deviations as indicating genuine range shift trends of fossil marine taxa. Low latitude taxa exhibit especially predictable migratory responses to climate warming. Fossils thus have great potential to elucidate ancient climate change and to help anticipate the effects of modern climate warming.

Perhaps our clearest result is the reliability of lower latitude taxa to respond to climate warming by poleward range shifts. In particular, the disproportionate strength of correlation of the .1-quantile with climate (versus the absence of correlation of even the .6-quantile) is striking, because the fringe quantiles (.1, .8 and .9, generally right-skewed distributions) are expected to be more biased by sampling constraints (edges) and therefore statistically noisier (Supporting Information Figure S4). Climate-range shift agreement is increased when sampling focal latitude is constrained to lower latitudes, here using a median–.2-quantile hybrid of latitudinal deviations. Tropical warming may then force regional range shifts (Walther et al., 2002) and even tropical extirpation of some species (e.g. Kiessling et al., 2012), which may live close to their phylogenetic maximum thermal limits (Song et al., 2015; Storch et al., 2014). Perhaps most importantly, temperature isotherms locally move at different speeds, or ‘climate velocities’, across the Earth's surface, and are expected to be higher at low latitudes, especially in marine habitats (Burrows et al., 2011). Poleward ‘pushes’ of taxon ranges may have been common in the past, when latitudinal diversity peaks (of a hemisphere) often did not fall upon the equator (Powell, 2009), and tropical species may be well adapted to track ‘pushes’ (Walters et al., 2012). Still, current migration patterns are more prevalent at the leading (poleward) edge than at the trailing edge (Poloczanska et al., 2013), suggesting a prevalence of species ‘pull’ forces arising from new climate opportunities. Marine species track closely their thermal niches (Sunday et al., 2012), such that ‘push’ forces may become just as strong as warming advances. Our study thus suggests that the most reliable impact of climate warming on species distributions still lies ahead.

We found no evidence for consistent climate-induced range shifts of higher latitude taxa, but we suggest that this is only because the signal is less clear, not because they are less affected by climate (Peck, Webb, & Bailey, 2004; cf. Powell, Moore, & Smith, 2015). High latitude genera also appeared to have no greater sensitivity to long-term cooling than lower latitude taxa. Perhaps this is because our seawater habitat temperature proxy focusses on low latitudes (Supporting Information Figure S3). Higher sampling latitudes do show equatorward migrations but this tendency is expected by chance (non-directed random migrations), given greater habitat area toward the equator (e.g. Jablonski et al., 2013). Higher latitude genera are far more volatile than lower latitude taxa (Supporting Information Figure S4), with some volatility attributable to real migrations (Powell et al., 2015) and some to proximity to the occurrence distribution edges. Higher latitude taxa endure seasonal fluctuation extremes, while regional climate changes, such as caused by the disruption of ocean currents or ice formation, may complicate migratory responses (Tingley, Koo, Moritz, Rush, & Beissinger, 2012). Therefore, higher latitude taxa may change latitude for many reasons other than changes in average global climate.

Short-term climate change related extinction events, where taxa are unable to track their thermal niche, may partly explain a weak observed climate–migration link. Weighting the time series by inverse extinction rate raised the climate–migration correlation strength. Temporal as well as regional extinctions may add noise to the data, with the lower number of surviving genera more influenced by random fluctuations. Extinction rates may also vary by latitude, depending on temporal interval (e.g. Kiessling & Aberhan, 2007; Powell et al., 2015). A modelling study suggests that extinction may be more likely a response by temperate species compared with tropical ones, as development rates may offset risks associated with being a thermal specialist (Walters et al., 2012). Consequently, extinctions may add another element obscuring climate-induced migrations of higher latitude taxa. The big five ancient mass extinctions of Raup and Sepkoski (1982) did not have a consistent effect on overall latitudinal migrations, perhaps because their hypothesized causes were diverse and their temporal durations far briefer than our temporal scale.

Averaging across our age-to-age absolute migrations over the 450 million years gives between 3–9 × 10⁻⁵ km/decade, depending on the quantile, which is five orders of magnitude lower than the average 31 km/decade observed over the last 60 years (Poloczanska et al., 2013). However, distances given by latitudinal deviations are not expected to be large. Firstly, latitudinal deviation is a relative measure of distance from the sampling centre to the median of genus range centres (Supporting Information Appendix S2). Secondly, genus range shifts integrate the somewhat chaotic dynamics of potentially many constituent species ranges. Thirdly, longer duration studies systematically underestimate rates. There is a negative power-law relationship between the timespan of observation and the observed rate of change both in evolutionary patterns and in rates of climate change (Gingerich, 2009; Kemp et al., 2015). This relationship is due to transient switchbacks in climatic trends for which there is ample evidence in palaeoclimates, even within intervals considered relatively stable (e.g. mid-Cretaceous; Wilson & Norris, 2001). Therefore, the potential of prehistoric species to change latitudinal positions over timeframes comparable with recent studies is critically underestimated.

The most prominent agreements between the time series of migrations and seawater temperature were around the cool intervals of the end-Ordovician and the Permo-Carboniferous (Royer et al., 2004), both with subsequent poleward pulses (hemisphere switching time series). Strong warming in the Early Triassic epoch (Sun et al., 2012; Song et al., 2015) drove poleward migrations of all Northern Hemisphere lower latitude quantiles (.1–.5), both in age-to-age and multi-age (smoothed) series. However, large changes in sampling distribution may weaken the ability of taxa deviations to describe real migrations, such as for the Southern Hemisphere Mesozoic and Cenozoic eras. We suggest that some parts of the fossil record (e.g. Early to Mid-Permian and mid-Cretaceous warming) are more reliable for detecting migration patterns than others (Clapham & James, 2012, for migration patterns of the Southern Hemisphere Permian; Ridgwell, 2005, for Phanerozoic intervals where oxygen isotopes become less reliable). In particular, mass extinction events severely decrease the number of surviving genera available to inform analyses of migration (e.g. Supporting Information Figure S1) and accounting for extinction rate over time slightly strengthened our correlation. Our analyses show how accepting that sampling patterns dominate the fossil record may still allow discernment of ecological signal, although this is not the only null hypothesis (Powell, 2009). Nevertheless, for patterns in the fossil record, the hypothesis of sampling bias must remain the first culprit to disprove.

Additionally, we detected a consistent poleward direction in accumulated latitudinal deviations, as would be expected if ‘out of the tropics’ dynamics (OTT, Jablonski, Kaustuv, & Valentine, 2006) were consistent over the Phanerozoic. The OTT hypothesis predicts a poleward accumulation of range edges by positing a general evolutionary tendency of latitudinal expansion of geographical range without loss of tropical occupancy (Jablonski et al., 2006). The OTT hypothesis has only been verified for the Cenozoic until now (Jablonski et al., 2013). Unfortunately, our results could not be separated from biases associated with sampling.

Poleward migration of low latitude marine invertebrate taxa during climate warming appears to be a general response over taxa, continental configuration and climatic scenarios. Therefore, our results corroborate predictions of a climate-induced net poleward migration of tropical marine benthic species in the foreseeable warming future (Burrows et al., 2011). New adaptations of physiological tolerance to heat may be rare, even over millions of years (e.g. Saupe et al., 2014). Therefore, taxa unable to migrate at the pace set by local climate velocities risk extinction. Our results suggest that the most dependable ecological consequences of global climate warming are yet to come, as current shifts in species equatorward edges are modest in relation to their poleward edges (Poloczanska et al., 2013). Consistent warming could eventually result in hemisphere diversity peaks vacating the tropics (Burrows et al., 2014; García Molinos et al., 2016; Kiessling et al., 2012). Migratory responses of higher latitude taxa may be more complicated to predict, partly because of extinctions (Walters et al., 2012). Incomplete fossil preservation and our methodology mean that common species, which perform the greater share of ecosystem functions (Gaston, 2010), drive our described patterns. Common species' distributions often respond similarly across taxa (Reddin, Bothwell, & Lennon, 2015), driving richness patterns. Thus, their redistribution will have repercussions on human well-being (Pecl et al., 2017).

ACKNOWLEDGMENTS

We are grateful for comments by Matthew Powell, Matthew E. Clapham and an anonymous referee that improved our manuscript. Michael Burrows, David Schoeman and Kenneth de Baets also provided valuable suggestions. This work was funded by the VolkswagenStiftung and grants from the Deutsche Forschungsgemeinschaft (DFG: KI 806/16-1 and KO 5382/1-1) and is embedded in the Research Unit TERSANE (FOR 2332: Temperature-related stressors as a unifying principle in ancient extinctions). This is Paleobiology Database publication 303.

DATA ACCESSIBILITY

Raw occurrence data, R code for calculating latitudinal deviations, and our own latitudinal deviation and temperature time series are available in the Dryad Digital Repository https://doi.org/10.5061/dryad.r20d2s4.

Supporting Information

REFERENCES

Blois, J. L., & Hadly, E. A. (2009). Mammalian response to Cenozoic climatic change. Annual Review of Earth and Planetary Sciences, 37(1), 181–208.
10.1146/annurev.earth.031208.100055
CAS Web of Science® Google Scholar
Burrows, M. T., Schoeman, D. S., Buckley, L. B., Moore, P., Poloczanska, E. S., Brander, K. M., … Richardson, A. J. (2011). The pace of shifting climate in marine and terrestrial ecosystems. Science, 334(6056), 652–655.
10.1126/science.1210288
CAS PubMed Web of Science® Google Scholar
Burrows, M. T., Schoeman, D. S., Richardson, A. J., Molinos, J. G., Hoffmann, A., Buckley, L. B., … Poloczanska, E. S. (2014). Geographical limits to species-range shifts are suggested by climate velocity. Nature, 507(7493), 492–495.
10.1038/nature12976
CAS PubMed Web of Science® Google Scholar
Chen, I., Hill, J. K., Ohlemüller, R., Roy, D. B., & Thomas, C. D. (2011). Rapid range shifts of species associated with high levels of climate warming. Science, 333(6045), 1024–1026.
10.1126/science.1206432
CAS PubMed Web of Science® Google Scholar
Clapham, M. E., & James, N. P. (2012). Regional-scale marine faunal change in eastern Australia during Permian climate fluctuations and its relationship to local community restructuring. Palaios, 27(9), 627–635.
10.2110/palo.2012.p12-003r
Web of Science® Google Scholar
Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/correlation analysis for the behavioral sciences ( 3rd ed., pp. 46–47). Mahwah, NJ: Lawrence Erlbaum Associates.
Google Scholar
Foote, M. (2000). Origination and extinction components of taxonomic diversity: General problems. Paleobiology, 26, 74–102.
10.1666/0094-8373(2000)26[74:OAECOT]2.0.CO;2
Web of Science® Google Scholar
Foote, M. (2014). Environmental controls on geographic range size in marine animal genera. Paleobiology, 40(03), 440–458.
10.1666/13056
Web of Science® Google Scholar
Foster, G. L., Royer, D. L., Lunt, D. J., Olsen, P. E., & Beerling, D. J. (2017). Future climate forcing potentially without precedent in the last 420 million years. Nature Communications, 8, 14845.
10.1038/ncomms14845
CAS PubMed Web of Science® Google Scholar
García Molinos, J., Halpern, B. S., Schoeman, D. S., Brown, C. J., Kiessling, W., Moore, P. J., … Burrows, M. T. (2016). Climate velocity and the future global redistribution of marine biodiversity. Nature Climate Change, 6(1), 83–88.
10.1038/nclimate2769
Web of Science® Google Scholar
Gaston, K. J. (2010). Valuing common species. Science, 327(5962), 154–155.
10.1126/science.1182818
CAS PubMed Web of Science® Google Scholar
Gingerich, P. (2009). Rates of evolution. Annual Review of Ecology, Evolution, and Systematics, 40(1), 657–675.
10.1146/annurev.ecolsys.39.110707.173457
Web of Science® Google Scholar
Gradstein, F. M., Ogg, J. G., Schmitz, M., & Ogg, G. (2012). The geologic time scale 2012 ( 1st ed.). Amsterdam, The Netherlands: Elsevier.
Google Scholar
Graham, R. W., Lundelius, E. L., Graham, M. A., Schroeder, E. K., Toomey, R. S., Anderson, E., … Wilson, M. C. (1996). Spatial response of mammals to late Quaternary environmental fluctuations. Science, 272(5268), 1601–1606.
10.1126/science.272.5268.1601
CAS PubMed Web of Science® Google Scholar
Greenstein, B. J., & Pandolfi, J. M. (2008). Escaping the heat: Range shifts of reef coral taxa in coastal Western Australia. Global Change Biology, 14(3), 513–528.
10.1111/j.1365-2486.2007.01506.x
Web of Science® Google Scholar
Grossman, E. L. (2012). Applying oxygen isotope paleothermometry in deep time. In L. C. Ivany & B. T. Huber (Eds.), Reconstructing earth's deep-time climate - The state of the art in 2012 (Vol. 18, pp. 39–67). Boulder, CO: Paleontological Society Papers.
Google Scholar
Jablonski, D., Belanger, C. L., Berke, S. K., Huang, S., Krug, A. Z., Roy, K., … Valentine, J. W. (2013). Out of the tropics, but how? Fossils, bridge species, and thermal ranges in the dynamics of the marine latitudinal diversity gradient. Proceedings of the National Academy of Sciences USA, 110(26), 10487–10494.
10.1073/pnas.1308997110
CAS PubMed Web of Science® Google Scholar
Jablonski, D., Kaustuv, R., & Valentine, J. W. (2006). Out of the tropics: Evolutionary diversity gradient. Science, 314(5796), 102–106.
CAS Web of Science® Google Scholar
Kemp, D. B., Eichenseer, K., & Kiessling, W. (2015). Maximum rates of climate change are systematically underestimated in the geological record. Nature Communications, 6(1), 8890.
10.1038/ncomms9890
CAS PubMed Web of Science® Google Scholar
Kiessling, W., & Aberhan, M. (2007). Environmental determinants of marine benthic biodiversity dynamics through Triassic-Jurassic times. Paleobiology, 33, 414–434.
10.1666/06069.1
Web of Science® Google Scholar
Kiessling, W., & Kocsis, Á. T. (2016). Adding fossil occupancy trajectories to the assessment of modern extinction risk. Biology Letters, 12(10), 20150813.
10.1098/rsbl.2015.0813
PubMed Web of Science® Google Scholar
Kiessling, W., Simpson, C., Beck, B., Mewis, H., & Pandolfi, J. M. (2012). Equatorial decline of reef corals during the last Pleistocene interglacial. Proceedings of the National Academy of Sciences USA, 109(52), 21378–21383.
10.1073/pnas.1214037110
CAS PubMed Web of Science® Google Scholar
Kocsis, Á. T. (2017). icosa: Global triangular and penta-hexagonal grids based on tessellated icosahedra. EGU General Assembly Conference Abstracts, 18, 17430.
Google Scholar
Martin-Garin, B., Lathuilière, B., & Geister, J. (2012). The shifting biogeography of reef corals during the Oxfordian (Late Jurassic). A climatic control? Palaeogeography, Palaeoclimatology, Palaeoecology, 365–366, 136–153.
10.1016/j.palaeo.2012.09.022
Web of Science® Google Scholar
McKinney, M. L., & Oyen, C. W. (1989). Causation and nonrandomness in biological and geological time series: Temperature as a proximal control of extinction and diversity. Palaios, 4(1), 3–15.
10.2307/3514729
Google Scholar
Parmesan, C., & Yohe, G. (2003). A globally coherent fingerprint of climate change impacts across natural systems. Nature, 421(6918), 37–42.
10.1038/nature01286
CAS PubMed Web of Science® Google Scholar
Peck, L. S., Webb, K. E., & Bailey, D. M. (2004). Extreme sensitivity of biological function to temperature in Antarctic marine species. Functional Ecology, 18(5), 625–630.
10.1111/j.0269-8463.2004.00903.x
Web of Science® Google Scholar
Pecl, G. T., Araújo, M. B., Bell, J. D., Blanchard, J., Bonebrake, T. C., Chen, I.-C., … Williams, S. E. (2017). Biodiversity redistribution under climate change: Impacts on ecosystems and human well-being. Science, 355(6332), eaai9214.
10.1126/science.aai9214
PubMed Web of Science® Google Scholar
Poloczanska, E. S., Brown, C. J., Sydeman, W. J., Kiessling, W., Schoeman, D. S., Moore, P. J., … Richardson, A. J. (2013). Global imprint of climate change on marine life. Nature Climate Change, 3, 919–925.
10.1038/nclimate1958
Web of Science® Google Scholar
Powell, M. G. (2009). The latitudinal diversity gradient of brachiopods over the past 530 million years. The Journal of Geology, 117(6), 585–594.
10.1086/605777
Web of Science® Google Scholar
Powell, M. G., Moore, B. R., & Smith, T. J. (2015). Origination, extinction, invasion, and extirpation components of the brachiopod latitudinal biodiversity gradient through the Phanerozoic Eon. Paleobiology, 41(02), 330–341.
10.1017/pab.2014.20
Web of Science® Google Scholar
R Development Core Team. (2017). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing.
Google Scholar
Raup, D. M., & Sepkoski, J. J. (1982). Mass extinctions in the marine fossil record. Science, 215(4539), 1501–1503.
10.1126/science.215.4539.1501
CAS PubMed Web of Science® Google Scholar
Reddin, C. J., Bothwell, J. H., & Lennon, J. J. (2015). Between-taxon matching of common and rare species richness patterns. Global Ecology and Biogeography, 24(12), 1476–1486.
10.1111/geb.12372
Web of Science® Google Scholar
Ridgwell, A. (2005). A mid mesozoic revolution in the regulation of ocean chemistry. Marine Geology, 217(3–4), 339–357.
10.1016/j.margeo.2004.10.036
CAS Web of Science® Google Scholar
Roy, K. (2001). Analyzing temporal trends in regional diversity: A biogeographic perspective. Paleobiology, 27(4), 631–645.
10.1666/0094-8373(2001)027<0631:ATTIRD>2.0.CO;2
Web of Science® Google Scholar
Royer, D. L. (2006). CO2-forced climate thresholds during the Phanerozoic. Geochimica et Cosmochimica Acta, 70(23), 5665–5675.
10.1016/j.gca.2005.11.031
CAS Web of Science® Google Scholar
Royer, D. L., Berner, R. A., Montañez, I. P., Tabor, N. J., & Beerling, D. J. (2004). CO₂ as a primary driver of Phanerozoic climate. Geological Society of America Today, 14(3), 4–10.
Google Scholar
Saupe, E. E., Hendricks, J. R., Portell, R. W., Dowsett, H. J., Haywood, A., Hunter, S. J., & Lieberman, B. S. (2014). Macroevolutionary consequences of profound climate change on niche evolution in marine molluscs over the past three million years. Proceedings of the Royal Society B: Biological Sciences, 281(1795), 20141995.
10.1098/rspb.2014.1995
PubMed Web of Science® Google Scholar
Song, H., Wignall, P. B., Chu, D., Tong, J., Sun, Y., Song, H., … Tian, L. (2015). Anoxia/high temperature double whammy during the Permian-Triassic marine crisis and its aftermath. Scientific Reports, 4(1), 4132.
10.1038/srep04132
Web of Science® Google Scholar
Sorte, C. J. B., Williams, S. L., & Carlton, J. T. (2010). Marine range shifts and species introductions: Comparative spread rates and community impacts. Global Ecology and Biogeography, 19(3), 303–316.
10.1111/j.1466-8238.2009.00519.x
Web of Science® Google Scholar
Storch, D., Menzel, L., Frickenhaus, S., & Pörtner, H. O. (2014). Climate sensitivity across marine domains of life: Limits to evolutionary adaptation shape species interactions. Global Change Biology, 20(10), 3059–3067.
10.1111/gcb.12645
PubMed Web of Science® Google Scholar
Sun, Y., Joachimski, M. M., Wignall, P. B., Yan, C., Chen, Y., Jiang, H., … Lai, X. (2012). Lethally hot temperatures during the Early Triassic greenhouse. Science, 338(6105), 366–370.
10.1126/science.1224126
CAS PubMed Web of Science® Google Scholar
Sunday, J. M., Bates, A. E., & Dulvy, N. K. (2012). Thermal tolerance and the global redistribution of animals. Nature Climate Change, 2(9), 686–690.
10.1038/nclimate1539
Web of Science® Google Scholar
Thibault, N., & Gardin, S. (2010). The calcareous nannofossil response to the end-Cretaceous warm event in the Tropical Pacific. Palaeogeography, Palaeoclimatology, Palaeoecology, 291(3–4), 239–252.
10.1016/j.palaeo.2010.02.036
Web of Science® Google Scholar
Tingley, M. W., Koo, M. S., Moritz, C., Rush, A. C., & Beissinger, S. R. (2012). The push and pull of climate change causes heterogeneous shifts in avian elevational ranges. Global Change Biology, 18(11), 3279–3290.
10.1111/j.1365-2486.2012.02784.x
Web of Science® Google Scholar
Valentine, J. W. (1974). Temporal bias in extinctions among taxonomic categories. Journal of Paleontology, 48, 549–552.
Web of Science® Google Scholar
Valentine, J. W., & Jablonski, D. (1993). Fossil communities: Compositional variation at many time scales. In R. E. Ricklefs & D. Schluter (Eds.), Species diversity in ecological communities (pp. 341–349). IL: University of Chicago Press.
Google Scholar
Veizer, J., & Prokoph, A. (2015). Temperatures and oxygen isotopic composition of Phanerozoic oceans. Earth-Science Reviews, 146, 92–104.
10.1016/j.earscirev.2015.03.008
CAS Web of Science® Google Scholar
Vilhena, D. A., & Smith, A. B. (2013). Spatial bias in the marine fossil record. PLoS One, 8(10), e74470.
10.1371/journal.pone.0074470
CAS PubMed Web of Science® Google Scholar
Visser, K., Thunell, R., & Stott, L. (2003). Magnitude and timing of temperature change in the Indo-Pacific warm pool during deglaciation. Nature, 421(6919), 152–155.
10.1038/nature01297
CAS PubMed Web of Science® Google Scholar
Walters, R. J., Blanckenhorn, W. U., & Berger, D. (2012). Forecasting extinction risk of ectotherms under climate warming: An evolutionary perspective. Functional Ecology, 26(6), 1324–1338.
10.1111/j.1365-2435.2012.02045.x
Web of Science® Google Scholar
Walther, G., Post, E., Convey, P., Menzel, A., Parmesan, C., Beebee, T. J. C., … Bairlein, F. (2002). Ecological responses to recent climate change. Nature, 416(6879), 389–395.
10.1038/416389a
CAS PubMed Web of Science® Google Scholar
Wilson, P. A., & Norris, R. D. (2001). Warm tropical ocean surface and global anoxia during the mid-Cretaceous period. Nature, 412(6845), 425–429.
10.1038/35086553
CAS PubMed Web of Science® Google Scholar
Wright, N., Zahirovic, S., Müller, R. D., & Seton, M. (2013). Towards community-driven paleogeographic reconstructions: Integrating open-access paleogeographic and paleobiology data with plate tectonics. Biogeosciences, 10(3), 1529–1541.
10.5194/bg-10-1529-2013
Web of Science® Google Scholar
Zachos, J. C., Dickens, G. R., & Zeebe, R. E. (2008). An early Cenozoic perspective on greenhouse warming and carbon-cycle dynamics. Nature, 451(7176), 279–283.
10.1038/nature06588
CAS PubMed Web of Science® Google Scholar

BIOSKETCHES

Carl J. Reddin is a quantitative marine ecologist with an interest in macroecological methods and their use to extend our understanding of temporal and spatial biodiversity patterns.

Ádám T. Kocsis is a postdoctoral research fellow with interests including global scale biogeographical and ecological processes, species distribution and occupancy modelling, and diversity dynamics.

Wolfgang Kiessling is professor and chair of the Paleoevironments and Paleobiology Programme at GeoZentrum Nordbayern. His research focusses on reef evolution, mass extinctions, and biodiversity dynamics at multiple scales.

Citing Literature

Volume27, Issue6

June 2018

Pages 704-713

Filename	Description
geb12732-sup-0001-suppinfo.docx30.9 KB	Supporting Information
geb12732-sup-0002-suppinfoFigs.pdf1.4 MB	Supporting Information Figures

Marine invertebrate migrations trace climate change over 450 million years

Correction(s) for this article

Marine invertebrate migrations trace climate change over 450 million years

Abstract