Global climate changes account for the main trends of conodont diversity but not for their final demise

Abstract

Free download link for 50 days : https://authors.elsevier.com/a/1bsT13HcE1cwvI

Conodonts, one of the longest-lived early groups of vertebrates, have a very complete fossil record ranging from the late Cambrian to the end of the Triassic and persisted through many global climatic and biotic events. In this paper, we analyse a large dataset harvested from the Paleobiology Database to compute global diversity curves at the generic level and explore patterns of conodont paleogeographic distribution. Our results partly confirm the most prominent findings of earlier studies including the occurrence of an Ordovician acme, a Permian nadir and a short-lived Triassic recovery. Major peaks of origination were found in the Early Ordovician and Early Triassic, while major extinctions occurred in the Upper Ordovician and Pennsylvanian. Paleogeographical extent of conodonts was impacted by i) the position of paleo-continents (notably impacting the latitudinal gradient of diversity), ii) the available continental shelf area and iii) ice sheets expansion. Diversity trends were mostly impacted by transitions between hothouse and icehouse ages, with major glaciations and associated marine regressions co-occurring with major extinctions. The influence of global sea level was less marked than that of temperature. However, the final demise of conodonts at the end of the Triassic did not coincide with either a major glaciation or marine regression. This supports the view that extinction of the group was mostly due to biotic factors such as competition with ‘Mesozoic’ taxa.

Full text

UNCORRECTED PROOF
Global and Planetary Change xxx (xxxx) xxx-xxx
Contents lists available at ScienceDirect
Global and Planetary Change
journal homepage: http://ees.elsevier.com
Global climate changes account for the main trends of conodont diversity but not for
their \nal demise
Samuel Ginot a,b,⁎, Nicolas Goudemand b
a)*69);619- ,+636/1- ,-: @,96:@:;-4-: );<9-3: -; 5;09671:-: –'$  $% ' – 9<- $)70)-3 <*61: );14-5;: )9>15   69-3  *6<3-=)9, ,<  56=-4*9-  
(133-<9*)55- 9)5+-
b5:;1;<; ,- -56418<- 65+;1655-33- ,- @65 –'$  % @65  $%  $ ' – )=-5<- &65@ )951-9  @65 9)5+-
ARTICLE INFO
-@>69,:
Past climate
Court Jester
Abiotic drivers
Conodonta
Mass extinction
ABSTRACT
Conodonts, one of the longest-lived early groups of vertebrates, have a very complete fossil record ranging from
the late Cambrian to the end of the Triassic and persisted through many global climatic and biotic events. In this
paper, we analyse a large dataset harvested from the Paleobiology Database to compute global diversity curves
at the generic level and explore patterns of conodont paleogeographic distribution. Our results partly con\rm the
most prominent \ndings of earlier studies including the occurrence of an Ordovician acme, a Permian nadir and
a short-lived Triassic recovery. Major peaks of origination were found in the Early Ordovician and Early Triassic,
while major extinctions occurred in the Upper Ordovician and Pennsylvanian. Paleogeographical extent of con-
odonts was impacted by i) the position of paleo-continents (notably impacting the latitudinal gradient of diver-
sity), ii) the available continental shelf area and iii) ice sheets expansion. Diversity trends were mostly impacted
by transitions between hothouse and icehouse ages, with major glaciations and associated marine regressions
co-occurring with major extinctions. The in]uence of global sea level was less marked than that of temperature.
However, the \nal demise of conodonts at the end of the Triassic did not coincide with either a major glaciation
or marine regression. This supports the view that extinction of the group was mostly due to biotic factors such as
competition with ‘Mesozoic’taxa.
1. Introduction
Biodiversity is impacted by numerous interconnected factors, and
disentangling their effects is a challenge for biologists and paleontolo-
gists alike. In this regard, the relative importance of biotic and abiotic
factors in the variation of global biodiversity through geological times
remains an open debate, often referred to as the Red Queen vs. Court
Jester debate. Biotic factors encompass aspects intrinsic to the living or-
ganisms, as well as interactions between organisms, such as competi-
tion; while abiotic factors include extrinsic aspects, often related to cli-
matic or geological changes (Benton, 2009). At high taxonomic lev-
els and broad timescales, abiotic factors may play a larger role (Ben-
ton, 2009). Yet, it remains dif\cult to test biotic effects in the fossil
record, where the interactions between taxa are hardly accessible, and
the true null hypothesis, namely that diversity follows a random walk
through time is generally ignored. Simulations and modeling, which are
employed more and more frequently, are a way to remedy to this. For in
stance, they allow to assess how trophic relationships in paleo-commu-
nities may affect the stability of the considered ecosystem to perturba-
tions, for instance whether the extinction of a few taxa or functional
groups may lead to cascading effects and eventually to the collapse of
the ecosystem (Roopnarine et al., 2007); or to determine how and
when within a clade history biotic interactions are most likely to con-
trol the dynamics of speciation, extinction and diversity (Aguilée et al.,
2018;Hofmann et al., 2019).
Another currently outstanding challenge is to anticipate the evolu-
tion of biodiversity to the current anthropogenic crisis, based on how
biodiversity reacted in the past to global changes analogous to the ones
Earth is experiencing today (Payne and Clapham, 2012). How are
different parts of the biosphere differentially affected by critical abi-
otic factors? Studies of the fossil record in correlation with major cli-
matic and geological variables remain the most direct way to address
this challenge. Historically, marine biodiversity has been the focus of
such studies, essentially because the fossil record of marine organisms is
⁎Corresponding author at: Laboratoire d'Ecologie des Hydrosystemes Naturels et Anthropises –UMR5023, CNRS, UCBL –3-6 rue Raphael Dubois, Batiments Darwin C & Forel, 43
boulevard du 11 novembre 1918, 69622 Villeurbanne, France.
4)13 ),,9-::-: samuel.ginot@univ-lyon1.fr (S. Ginot); nicolas.goudemand@ens-lyon.fr (N. Goudemand)
https://doi.org/10.1016/j.gloplacha.2020.103325
Received 14 January 2020; Received in revised form 17 September 2020; Accepted 17 September 2020
Available online xxx
0921-8181/© 2020.

UNCORRECTED PROOF
%156;)5, 6<,-4)5, 36*)3)5,"3)5-;)9@0)5/- ??? ???? ??????
more complete and better resolved than that of terrestrial organisms
(e.g. Sepkoski, 1981).
 656,65; ,1=-9:1;@ ;096</0 ;14-
Despite its long and global record, the clade of conodonts remains
understudied in terms of macroevolution. Conodonts form an early and
diverse group of marine vertebrates, most probably associated to extant
cyclostomes (lampreys and hag\shes; Miyashita et al., 2019). Eco-
logically, conodonts are generally considered to have been small active
swimmers, mostly nektonic, primary consumers (Purnell, 2001). This
broad picture is however probably not true for all conodont species, con-
sidering their high morphological disparity and the variety of sedimen-
tological facies in which their remains are found (Purnell and Jones,
2012;Ginot and Goudemand, 2019). Through their long fossil record
(over 300 My, from Cambrian to the end of the Triassic; Dong et al.,
2004,Zhen et al., 2018,Tanner et al., 2004), conodonts as a group
have faced some of the most prominent events in the history of Earth,
including four out of \ve mass extinction crises (end Ordovician, Late
Devonian, end Permian, and end Triassic), and they survived all, ex-
cept the last, eventually disappearing at the end of the Triassic. As such,
they constitute a good model organism to study how the diversity of
small, nektonic organisms may be impacted by major environmental
changes. The responses of conodont diversity to environmental changes
have been generally studied at the scale of individual events only, often
with the biostratigraphic objective of de\ning biozones (e.g. Orchard,
2007), rather than as a whole. Conodont diversity trends have not re-
cently been tackled over large timeframes (Clark, 1983, 1987;Sweet,
1988)–although conodont data was included in much broader studies
(Friedman and Sallan, 2012;Whalen and Briggs, 2018)–which
is now made possible and more exhaustive by the advent of global fos-
sil occurrence databases (i.e. the Paleobiology Database), and more reli-
able thanks to new methods for the quanti\cation of diversity, origina-
tion and extinction (e.g. Alroy et al., 2008;Alroy, 2010b). Here, we
aim not only at updating the previously published trends of conodonts
diversity, but also at statistically testing their potential control by some
abiotic variables (temperature, sea-level, paleogeography) for which we
collected quantitative data from the literature. Other abiotic parameters,
such as water pH, bathymetry or paleo-currents, may also play a role
but because of a lack of available, appropriate, quantitative data, those
were not included here.
The rise of the ‘Paleozoic Fauna’identi\ed by Sepkoski (1981)
started in the Ordovician, forming what we now call the Great Ordovi-
cian Biodiversi\cation Event (GOBE). Although abiotic factors such as
a cooling of sea water temperatures have been proposed to be linked
to the GOBE, there is still no consensus about the causes of this event
(Servais and Harper, 2018). The ‘plankton revolution’of the Early
Ordovician, which represents the \rst major biodiversity event of the
GOBE, may be linked to increasing sea level, oceanographic changes
and oxygenation. The consequent increase in planktonic resources may
have brought the diversi\cation of other forms. Although the \rst con-
odonts arose during the Cambrian (Dong et al., 2004), their radiation
is mostly part of the GOBE, and may be related to both biotic (plankton
diversi\cation) and abiotic factors. The Ordovician diversi\cation was
interrupted by the end Ordovician extinction, the \rst of the ‘Big Five’
extinctions. This two-pulse extinction is generally attributed to a major
glaciation associated with a regression (\rst pulse), followed by a trans-
gression associated with anoxia (second pulse). Although short-spanned
(~1 Ma), it caused important habitat loss and impacted conodonts,
among others, at both pulses (Harper et al., 2014). However, the du-
ration of this glaciation is debated, and may extend to 10 Ma, which
would have put conodonts and the biosphere through an even greater
challenge. The next major climatic episode may have been the Middle
Devonian ‘super-greenhouse’, although the nature of this event is now
challenged (Joachimski et al., 2009): the Early Devonian was already
characterized by warm temperatures, and it now appears the Middle De-
vonian was in fact characterized by a global cooling –without glaciation
–followed by a global warming that led seawater temperatures to reach
two maxima around the Frasnian/ Fammenian boundary (Joachimski
et al., 2009). The two Kellwasser environmental events occurred be-
fore, and in-between those maxima and correspond to cooler and anoxic
episodes. These events are known to have affected both conodont diver-
sity and morphology (Girard and Feist, 1996;Balter et al., 2008).
From the end of the Fammenian, and throughout the Carboniferous and
Permian, a period known as the Late Paleozoic Ice Age, the global cli-
mate was mostly cold and associated with glaciations and low sea lev-
els, (Montañez and Poulsen, 2013). According to Buggisch et al.
(2008), the tipping point between the Devonian greenhouse and the
Late Paleozoic Ice Age was reached in the Mississippian. Two cycles
of glaciation are recorded, the \rst peaking (coldest temperatures) at
the transition between the Mississippian and Pennsylvanian and the sec-
ond around the Carboniferous / Permian boundary (Buggisch et al.,
2008;Montañez and Poulsen, 2013). These glacial events were ac-
companied by important and rapid eustatic variations (e.g. Joachimski
et al., 2006;Barrick et al., 2013;Montañez and Poulsen, 2013;
Bahrami et al., 2014). During this ice age (Pennsylvanian), inverte-
brates showed low rates of origination and extinction and diversity was
low (Stanley and Powell, 2003;Powell, 2005;Alroy et al., 2008).
Similar trends were reported by Clark (1983) for conodonts, with a
large drop in origination and extinction rates and in the number of gen-
era between Mississipian and Pennsylvanian.
There is evidence for episodic glacial deposits until the end of the
Middle Permian, and deglaciation in the upper Permian (~260 Ma;
Montañez and Poulsen, 2013). Global temperatures increased and
peaked at the P/T boundary, linked with the volcanic activity of the
Siberian Traps, causing global greenhouse warming, ocean acidi\cation,
associated with episodes of anoxia or euxinia (Sun et al., 2012;Ro-
mano et al., 2013;van de Schootbrugge and Wignall, 2016).
Throughout the Permian only a handful of conodont genera remained,
several of which crossed the P/T boundary (Clark, 1983;Orchard,
2007), and thrived in the Early Triassic, in stark contrast with the ma-
jority of other organisms at the time (e.g. Brayard et al., 2017, and
references therein). The aftermath of the P/T crisis is marked by gen-
erally unstable conditions, several cooling and warming episodes oc-
curring in the Early Triassic (Goudemand et al., 2019). The transi-
tion between Early and Middle Triassic is marked by a global cooling
and the appearance of worldwide monsoon events, among which the
Carnian Pluvial event was the largest (Preto et al., 2010;Sun et al.,
2012). These conditions were maintained through the Middle Triassic
and Late Triassic, during which climate appears to have been fairly sta-
ble (Preto et al., 2010). Conodont diversity apparently progressively
declined throughout the Middle and Late Triassic (e.g. de Renzi et al.,
1996;Hallam, 2002;Martínez-Pérez et al., 2013;van de Schoot-
brugge and Wignall, 2016) and generally had high extinction rates
during this interval, especially at the end of the Norian, which was pre-
viously considered to correspond to their \nal extinction (Tanner et
al., 2004). Indeed the Rhaetian conodonts are represented by a handful
of species only, the last ones eventually going extinct near the Triassic
/ Jurassic boundary. The slow decline of conodont diversity throughout
the Triassic may appear surprising considering the relatively stable con-
ditions of the Triassic, as well as the fact that the explanations for the
P/T and T/J boundary crises are more or less convergent (Lucas and
Tanner, 2004;Mazza et al., 2010), but would \t the null hypothesis
of diversity following a random walk.
2

UNCORRECTED PROOF
%156;)5, 6<,-4)5, 36*)3)5,"3)5-;)9@0)5/- ??? ???? ??????
 "9-=16<: :;<,1-: )5, +<99-5; 796*3-4);1+:
Considering their status of early vertebrates, their amazing re-
silience, and their fairly anti-climactic disappearance, the global diver-
sity of conodonts and its drivers have received surprisingly little atten-
tion since the end of the eighties. Meanwhile, fossil occurrence and pa-
leoenvironmental data have grown enormously. Clark (1983, 1987)
already recognized two diversity maxima in the Ordovician and Devon-
ian, and a ‘last hurrah’in the Triassic. He further observed that origina-
tion and extinction peaked in the Ordovician, Mississippian and Trias-
sic, with the latter two intervals corresponding to higher extinction than
origination rates. Clark, however, did not comment at length on the pos-
sible drivers of conodont's diversity ]uctuations. Sweet (1988) mostly
agreed with the pattern Clark described, although he split the Ordovi-
cian diversity peak into two ‘long-term cycles’. Sweet went on to suggest
that the \nal extinction of conodonts was due to a sea level drop. On the
contrary, De Renzi et al. (1996) showed that conodonts declined pro-
gressively from the Middle through the Late Triassic, and therefore fa-
vored competition as the main player in the disappearance of conodonts;
an idea recently supported also by Martínez-Pérez et al. (2013). In
other words, the Court Jester vs. Red Queen debate applies also to the
extinction of conodonts. In our opinion, the abundant conodont data at
hand therefore constitute a unique opportunity to understand how abi-
otic factors –including large catastrophes –may drive the diversity of
nektonic animals, but also possibly how biotic interactions may be in-
strumental in explaining the extinction of such a large taxonomic group.
In this study, we investigate whether the picture Clark (1987) and
Sweet (1988) drew still holds today, using updated dataset and meth-
ods. Using current knowledge on paleoenvironment, we will also aim at
identifying the main abiotic drivers of conodont diversity and associated
paleogeographical patterns, and test whether those drivers may or may
not account for the \nal demise of conodonts.
2. Material and methods
 );):-;
The data used in this study was obtained from the Paleobiology Data-
base. The occurrences were downloaded from the Paleobiology Database
(https://paleobiodb.org) on 26 June 2018, selecting by taxonomy, with
group name = ‘Conodonta’. The resulting dataset included 23,520 oc-
currences, with associated generic or speci\c accepted names, time in-
tervals, localities and paleo-coordinates computed by the GPlates model
implemented in the Paleobiology Database (Wright et al., 2013).
The references on which the dataset is based cumulate a total of 357
unique \rst authors (secondary bibliography available as supplemen-
tary material), the main contributors (> 500 occurrences) being Zhang
( = 1236), Barrick ( = 1167), Männik ( = 964), Ji ( = 904),
Suttner ( = 871), Farrell ( = 832), Bultynck ( = 743), Klapper
( = 714), Aboussalam ( = 602), Gouwy ( = 507). The main enter-
ers (> 1000 occurrences) of the data into the Paleobiology Database
were M. Krause ( = 5735), S. Gouwy ( = 3547), E. Jarochowska
( = 3010), P. Nätscher ( = 1363), J. Sessa ( = 1169), M. Foote
( = 1149), and P. Novack-Gottshall ( = 1086). All subsequent data
manipulation and analyses were run in R (R Core Team, 2018).
 1=-9:1;@ 691/15);165 )5, -?;15+;165
From this data, diversity was estimated for genera at the series
level. First, the dataset was trimmed of single occurrences whose tem-
poral resolution is not constrained to one series: their low stratigraphic
resolution makes them useless and possibly would distort the results.
Therefore the sample was reduced to = 19,737 occurrences. The
genus name was used even for occurrences identi\ed at the speci\c
level. A total of 265 unique genera were recorded in the dataset, af-
ter checking for duplicates due to spelling mistakes. From this subset
was produced a generic presence / absence matrix. Generic diversity
estimates were computed using Alroy's ‘shareholder quorum sampling’
(SQS; Alroy, 2010a, 2010b), running 1000 iterations while randomly
resampling the presence / absence matrix with replacement. The \nal
diversity estimate was the average across these iterations, and the as-
sociated standard deviations were also computed. SQS estimates for the
Terreneuvian, Cambrian Series 2 and Lopingian could not be computed
properly due to the very low generic diversity, and are therefore re-
placed by raw generic counts. A similar analysis was run for occurrences
for which the resolution was at the stage level ( = 15,279). Because
the resulting diversity curve was for a large part discontinuous, it is
shown as supplementary material (SM1). Additionally, ‘single-interval’
diversity was computed (at the series level) as the number of genera
present in only one series. From the same subset, extinction and orig-
ination rates were computed using the ‘three-timer’formulas of Alroy
(2010b). The use of SQS diversity estimates, and three-timer formulas
aims at avoiding several biases present in most paleontological studies
of diversity. These include the edge and Signor-Lipps effects by which
the diversity arti\cially drops before and rises after boundaries (espe-
cially across mass extinctions), but not the Pull of the Recent effect, since
conodonts do not have a fossil record in the Recent. The Lagerstatte ef-
fect, and unique taxa due to taxonomic identi\cation dif\culties in con-
odonts, are taken into account by calculating SQS and independently
computing single-interval generic counts. Finally, poly-cohort contour
graphs (Brayard et al., 2009) were produced, which allow a different
representation of diversity, extinction and origination trends. The trends
revealed by these graphs were similar to results from the SQS, origina-
tion and extinction curves, therefore they are only presented as supple-
mentary material (SM2, SM3).
 656,65; -=63<;165)9@ .)<5):
To re\ne the analysis of conodont diversity, we investigated the ex-
istence of several ‘evolutionary faunas’based on the temporal range of
the various genera, excluding the ‘single-interval’genera. First, a Mul-
tiple Correspondence Analysis (MCA) was run on the generic presence
/ absence matrix at the series level, using the ‘MCA’function imple-
mented in R package ‘FactoMineR’(Lê et al., 2008). Hierarchical clus-
tering (function ‘HCPC’of ‘FactoMineR’) was then used on the produced
multivariate space to form clusters and assign genera to them, to de\ne
‘evolutionary faunas’. Clusters were de\ned arbitrarily, based on two
criterias: inertia gain and non-overlapping of the groups.
Patterns of diversity were assessed for each evolutionary fauna, with-
out using SQS, but only raw generic counts, excluding single-interval
taxa. Scotese's (2016) paleomaps were used to look at the paleobio-
geographical patterns of the different evolutionary faunas, by displaying
the presence / absence of each group as different colors in the cells (see
below).
 -5-91+ +6<5;: )5, %#% ,1=-9:1;@
Excluding single-interval taxa, which are biased by taxonomic dif\-
culties and Lagerstatte effects, raw generic counts at the series level were
computed. The relationship between the decimal logarithm of these
counts and that of the SQS values was tested by a linear regression. The
Terreneuvian, Cambrian Series 2 and Lopingian were excluded of this
model, since their SQS values could not be computed.
3

UNCORRECTED PROOF
%156;)5, 6<,-4)5, 36*)3)5,"3)5-;)9@0)5/- ??? ???? ??????
 ")3-6*16/-6/9)701+)3 7);;-95:
To analyse paleobiogeographical patterns, we used the same ap-
proach as Kocsis et al. (2018). Scotese's ‘PALEOMAPS PaleoAtlas’was
downloaded and the paleomaps data for the Cambrian to Triassic pe-
riod were imported in R (Scotese, 2016). For each map in the atlas,
we de\ned an age range, corresponding to the stage represented by the
map. Using R package ‘icosa’(Kocsis, unpublished), a 2D penta-hexag-
onal grid was created, to be projected on the maps, with ‘cells’(faces)
of approximately equal surfaces and an average side (‘edge’) length of
740.74 km. These cells were used as a coarse resolution unit for our pa-
leogeographical investigations. Conodont occurrences whose range was
entirely included within the age range of each map ( = 12,448) were
automatically assigned to the corresponding cells, based on their paleo-
latitudes and paleolongitudes, using function ‘locate’of package ‘icosa’.
The rest of the occurrences, which could not be restricted to the same
resolution as the paleomaps, were excluded in this part, to avoid placing
on the maps occurrences that may not have existed at the time repre-
sented by the paleomap.
The cells including areas of continental platforms and/or coast were
manually counted for each stage as a proxy for the theoretically avail-
able area for conodonts. The number of cells was also computed for each
stage, as a proxy of conodont geographical spread and occupation. The
link between theoretically available area and occupied area was tested
by a linear regression, out of which the residual variation was extracted
to check for potential correspondence with other abiotic events. The
maximal and minimal latitude occupied by conodonts were obtained as
the latitudes of the center of the northernmost and southernmost occu-
pied cells.
 *16;1+ =)91)*3-: ;9-5,:
Sea surface temperature (SST) data were obtained from Song et al.
(2019), and sea-level data were extracted from Hannisdal and Peters
(2011) using the WebPlotDigitizer application (https://apps.automeris.
io/wpd/). Data for the latter ultimately derived from Haq and Schutter
(2008) and Haq et al. (1987). After extracting this data, we averaged
it at the series level, to correlate it with our diversity, origination and ex-
tinction estimates. Those relationships were tested using linear models.
The global mean annual temperature curve from Scotese (2015) was
also extracted using WebPlotDigitizer and added for comparison with
SST.
 );1;<,15)3 /9),1-5;:
Finally, latitudinal gradients were investigated at the series level,
by computing generic counts in bins of 10° latitude. No SQS was used
here, the gradients represent raw generic counts. Latitudinal gradients
of diversity were also computed for each evolutionary fauna as a whole
(rather than by series), either as raw generic counts, or as proportion of
their respective generic diversity. Bootstrapping of occurrence was used
(10,000 iterations) to test for significant differences between these gra-
dients across the three evolutionary faunas.
3. Results
 656,65; /-5-91+ ,1=-9:1;@ 691/15);165 )5, -?;15+;165
Conodont SQS diversity (Fig. 1A) shows three conspicuous peaks:
throughout Ordovician, in the Early Devonian, and in the Early Trias-
sic, matching with peaks of single-interval taxa diversity. Diversity is
at its highest throughout the Ordovician. Major decreases occur across
the Ordovician –Silurian boundary, Early –Middle Devonian boundary,
and Carboniferous –Permian boundary. Peaks of conodont origination
rates (Fig. 1B) are observed during the Early Ordovician, Early Sil-
urian (Llandovery), Late Carboniferous (Pennsylvanian) and Early Trias-
sic. Extinction peaks (Fig. 1B) are observed in the Late Ordovician, the
Late Devonian (smaller peak) and Pennsylvanian (at which time it over-
comes the synchronous origination peak).
There is a significant negative linear relationship between average
SST and extinction rate (adjusted R2= 0.265, "= 0.017, df = 1, 16),
and a positive relationship between average sea-level and SQS diversity
(adjusted R2= 0.195, "= 0.029, df = 1, 18) at the series level. Other
relationships between SST or sea-level and diversity, extinction and orig-
ination were non-significant ("> 0.05).
 =63<;165)9@ .)<5):
The multiple correspondence analysis, followed by hierarchical clus-
tering, revealed three clusters, or evolutionary faunas, based on tempo-
ral distribution of the genera (Fig. 2). The \rst (48 genera) is mostly
restricted to the Ordovician, the second (37 genera) also starts diver-
sifying in the Early Ordovician, but reaches its acme during the Sil-
urian, and starts decaying progressively from the Early Devonian on
(Fig. 3). The latest evolutionary fauna (48 genera) rises in the Early De-
vonian, reaches its acme in the Late Devonian, and then progressively
goes down through the Carboniferous, followed by a steeper decay dur-
ing the Permian. This evolutionary fauna recovers and constitutes most
of the diversity of the Triassic period. Following the classi\cation pro-
posed by Donoghue et al. (2008), the stem Prioniodontida, Balog-
nathidae, stem Ozarkodinida, as well as several Prioniodinina are for the
most part included in the early evolutionary fauna (Fig. 3C). The in-
termediate evolutionary fauna contains stem Ozarkodinina, stem Polyg-
nathacea, as well as some Prioniodinina (!<36,<:). The late evolution-
ary fauna contains some Prioniodinina, and most derived Ozarkodinina
(Polygnathacea and the unamed superfamily containing ‘gondolellids’).
 %#% ,1=-9:1;@ -:;14);-: )5, /-5-91+ +6<5;:
The SQS diversity estimates (Fig. 1A) and total of non-single-inter-
val genera (Fig. 3A) are significantly correlated at the series level, ex-
cluding the Terreneuvian, Series 2 and Lopingian (R2= 0.4, "< 0.01,
df = 1, 17; Fig. 4). Discrepancies between SQS and generic counts are
frequent. For the Series 3, Furongian, Guadalupian, and Early Triassic,
generic counts are lower than expected. For the Middle Devonian, Late
Devonian, Wenlock, Ludlow, and Llandovery counts are higher than ex-
pected. The most notable discrepancies are the Middle and Late De-
vonian, and the Guadalupian, during which SQS estimates and generic
counts show opposite trends (Fig. 3A).
 ")3-6/-6/9)701+)3 7);;-95: 6. +656,65; ,1=-9:1;@
Geographically, the \rst occurrences of conodonts are rather spread
out (Fig. 1E, 5A). The number of cells occupied by conodonts sees
an important rise in the Late Cambrian, which is sustained through
the Ordovician despite a short term drop at the Cambrian –Ordovi-
cian boundary (Fig. 1C-D). The Hirnantian glaciation clearly reduces
the geographical distribution of conodonts, both in its latitudinal ex-
tremes and number of occupied cells (Fig. 1C-E, 5C-D). The extent of
conodonts then increases again from the Wenlock into the Early Devon-
ian (Lochkovian), but is restricted during the rest of the Early Devon-
ian (Fig. 1C-E, 5F-G). A short-lived spread is seen in the Eifelian (Mid-
dle Devonian), followed by a continued decrease during the remaining
of the Devonian and the Mississippian (Fig. 1C-E, 5H, 6A-B). The geo-
graphical spread of conodonts remains fairly stable through the Missis-
sippian and Pennsylvanian transition (Fig. 1C, E), although the avail-
able space becomes more and more restricted (Fig. 1C, D), it decreases
4

UNCORRECTED PROOF
%156;)5, 6<,-4)5, 36*)3)5,"3)5-;)9@0)5/- ??? ???? ??????
5

UNCORRECTED PROOF
%156;)5, 6<,-4)5, 36*)3)5,"3)5-;)9@0)5/- ??? ???? ??????
Fig. 1.
Fig. 2. First two axes of the Multiple Correspondance Analysis (MCA) run on the generic time series presence / absence matrix, excluding single-series taxa. Hierarchical clustering was
used on this MCA, and three clusters were retained based on the inertia gain and non-overlap of the groups. Abbreviations represent the series in which genera were present. Csrl:
Cisuralian, ErlD: Early Devonian, ErlO: Early Ordovician, ErlT: Early Triassic, Frng: Furongian, Gdlp: Guadalupian, Ldlw: Ludlow, Llandovery: Llnd, Lpng: Lopingian, LtDv: Late Devonian,
LtOr: Late Ordovician, LtTr: Late Triassic, MddD: Middle Devonian, MddO: Middle Ordovician, Mddt: Middle Triassic, Msss: Mississippian, Pnns: Pennsylvanian, Prdl: Pridoli, Srs2: Series
2, Srs3: Series 3, Trrn: Terreneuvian, Wnlc: Wenlock. The \rst cluster (magenta triangles) forms the ‘Early Evolutionary Fauna’. The second cluster (yellow circles) forms the ‘Intermediate
Evolutionary Fauna’. The third cluster (black squares) forms the ‘Late Evolutionary Fauna’. See Fig. 3 for more details on these clusters. (For interpretation of the references to colour in
this \gure legend, the reader is referred to the web version of this article.)
again at the end of the Carboniferous, with conodonts being restricted
to equatorial latitudes (Fig. 1E). An all-time low is reached during the
middle of the Cisuralian (Early Permian; Fig. 1C-E, Fig. 6D-E), asso-
ciated with the near-total extinction of conodonts (Fig. 1A), and de-
spite ice sheets pulling back. A marked spread can be seen at the end
of the Permian, notably towards the extreme latitudes (Fig. 1C-E, Fig.
6F-G). The extent of conodonts remains high in the Early Triassic, then
goes back down for a short time in the Middle Triassic (Fig. 1C-E, Fig.
6H), before re-increasing in the Late Triassic. The number of occupied
cells is significantly correlated to the number of theoretically available
cells (R2= 0.15, P < 0.01, df = 1, 44), however with a large amount
of residual variation (Fig. 1C-D), the latter being apparently synchro-
nous with the sea level curve until the Early Carboniferous (Fig. 1F).
 );1;<,15)3 /9),1-5;:
Latitudinal gradients of generic diversity (LGGD) at the series level
can be interpreted only from the Early Ordovician on. In the Early Or-
dovician, the diversity is fairly spread out between −50 and 50°, with
multiple modes, two being found around 40° and −30°, and a third
just north of the equator (Fig. 7). Two LGGD modes are also seen in
the Middle Ordovician, one just south of the equator, the other around
25°. During the Late Ordovician the LGGD ]attens, with conodont di-
versity spreading southward, only the northernmost latitudes being de-
void of conodonts. Following the Hirnantian, the southern occurrences
disappear (Fig. 5D) and a fairly classical, mostly unimodal latitudinal
gradient is established, which is maintained until the Mississipian (Fig.
7). One particularity of this latitudinal gradient is that its mode is not
6
Synthetic view of diversity and paleobio-geographical patterns through the Phanerozoic. Alternating gray and white areas delimit the boundaries between series. A) Black line: average
SQS generic diversity ± std. dev. over 1000 iterations (gray outline). Blue dashed line: count of single-series genera (i.e. genera which occur in only one series). B) Origination (red) and
extinction (black) rates, computed using the 3-timer approach of Alroy (2014). C) Blue line and area: geographical area theoretically available to conodonts, as the number of cells of the
paleomap including continental shelf or coast. Black line and area: geographical area where conodonts occur, computed as the number of cells including at least one occurrence of con-
odont (see Material and Methods). Available and occupied space are not at the same scale. D) Residual variation of a linear regression of number of cells occupied against number of cells
available, i.e. variation in the occuped area after removing the effect of continental shelf / coastal area. E) Median northern (red) and median southern (blue) paleolatitudes of conodont
occurrences, dashed lines are the maximal and minimal paleolatitudes occupied by conodonts, computed as the middle latitude of the northern-most and southern-most occupied cells. F)
Phanerozoic global sea surface temperature curve, from Song et al. (2019; red curve and points, dashed red line is the series-averaged cruve), global mean temperature mod-
i\ed from Scotese (2015); black line), and global sea level curve, modi\ed from Haq and Schutter (2008; blue line and area). (For interpretation of the references to colour in
this \gure legend, the reader is referred to the web version of this article.)

UNCORRECTED PROOF
%156;)5, 6<,-4)5, 36*)3)5,"3)5-;)9@0)5/- ??? ???? ??????
Fig. 3. Diversity trends of the three ‘evolutionary faunas’.A) Raw generic count (excluding single-series taxa) by series for genera assigned to the different ‘evolutionary faunas’(full
lines), and the sum of these curves (dashed line). Gray outline shows the SQS diversity estimate std. dev. (see Fig. 1A) for comparison. B) Percentage represented by the faunas in the total
diversity in each series. The ‘Early Fauna’dominates through Cambrian and Ordovician, but is replaced by the ‘Intermediate Fauna’at the start of the Silurian. Finally the ‘Late fauna’
becomes dominant in the Middle Devonian, although remnants of earlier faunas survive until the Triassic. C) Tree modi\ed from Donoghue et al. (2008), with genera highlighted
according to their evolutionary fauna. Genera not highlighted could not be included in the analysis (i.e. single-series taxa).
Fig. 4. Plot of raw generic counts (excluding single-series taxa) against average SQS diver-
sity estimate, by series. Black line shows the linear regression line (R2= 0.40, "< 0.01),
dashed red lines represent 95% con\dence intervals. Named points are those outside of
the 95% C.I., i.e. series in which the raw counts do not follow the predicted relationship
with the SQS estimates. (For interpretation of the references to colour in this \gure legend,
the reader is referred to the web version of this article.)
equatorial, but located in the Southern Tropics. The whole diversity
of conodonts in this time frame seems to concentrate in the South-
ern Hemisphere. In the Pennsylvanian, the distribution seems to shift
slightly northward, but conodont diversity is drastically reduced dur-
ing the Permian (Fig. 6D-E). In the Early Triassic, a latitudinal gradient
is re-established but ]atter than previously, with a small mode in the
Northern Hemisphere. Diversity is again hit during the Middle Triassic,
before a more marked latitudinal gradient is established in the Late Tri-
assic, with a mostly northern diversity and a single mode in the North-
ern Tropics.
4. Discussion
 )+;69: 147)+;15/ +656,65; ,1=-9:1;@ )5, /-6/9)701+)3 7);;-95:
The large conodont diversi\cation after the Cambrian –Ordovician
boundary occurs within a transgressive cycle, peaking in the Middle
Ordovician (Fig. 1F), combined with hot but decreasing temperatures.
It constitutes a part of the ‘Great Ordovician Biodiversi\cation Event’
(GOBE, e.g. Harper et al., 2015), as conodonts may have been among
the \rst groups to colonize the water column. The increase in avail-
able shelf area may have enabled the origination of many conodont taxa
(Fig. 1B-C, Harper et al., 2015). However, it also appears that, at that
time, some conodonts colonized open sea environments (Fig. 5B), pos-
sibly via the increase in diversity and abundance of plankton, and the
likely establishment of open marine food chains (Harper et al., 2015;
Servais et al., 2008). Either factors, or their combination, may explain
the dramatic increase in conodont diversity. The level of conodont di-
versity reached during the Ordovician was not equaled at any later time
(Fig. 1A).
The Hirnantian / end-Ordovician extinction brought an end to the
acme of conodont diversity. It was synchronous with a large drop in
seawater temperature and sea level, and with ice sheets spreading (Fig.
5C-D, Harper et al., 2014). The conodonts' geographical distribu-
tion was greatly reduced at that time, especially in southern localities
that became covered by ice (Fig. 1E, Fig. 5D). This extinction mainly
affected the early evolutionary fauna (Fig. 3), while the intermedi-
ate evolutionary fauna remained diverse across the boundary, and di-
versi\ed in the Early Silurian. The global diversity plummeted due to
the loss of early fauna genera, but the intermediate fauna rised. At a
\ner temporal resolution, the drop in conodont diversity appears as a
two-step process, with one extinction event at the Katian –Hirnantian
boundary, and a second during the Hirnantian (Harper et al., 2014).
It should be noted that the early fauna taxa that disappeared shared
their geographical and temperature distribution with the earliest inter
7

UNCORRECTED PROOF
%156;)5, 6<,-4)5, 36*)3)5,"3)5-;)9@0)5/- ??? ???? ??????
Fig. 5. Paleobiogeographic evolution of conodont diversity from end-Cambrian to Late Devonian. Paleomaps from Scotese (2015) are divided into a grid of 362 pentagonal and hexago-
nal ‘cells’(faces) with approximately equal area and ‘side’(edge) length. Average side length is 740.74 km. Conodont occurrences were automatically assigned to the corresponding cells,
based on their paleolatitudes and paleolongitudes from the Paleobiology Database. Each cell occupied by a colored circle includes at least one occurrence. Multi-colored circles denote the
presence of conodont genera belonging to the different ‘Evolutionary Faunas’described herein. Note that the proportions of each colour are entirely arbitrary and ,6 56; relate in any way
to the proportions of local generic diversity. Circles were plotted at the center of the cells and their position do not imply that the occurrences are actually found together at a \ner scale
(i.e. within a cell).
mediate fauna taxa (Fig. 5C-E), suggesting that there was no fundamen-
tal difference in the temperature tolerance of the two groups. Therefore
the reason why some taxa survived and others not may instead be re-
lated to the presence of refugias, or to the in]uence of temperature on
the preys of the conodonts. The recovery fauna, which started to diver-
sify at the end of the Hirnantian may correspond to the early radiation
of the intermediate evolutionary fauna. The turnover in conodont diver-
sity may therefore be explained roughly by the disappearance of early
fauna taxa due to the drop in temperatures and loss of available habi-
tat (South Pole glaciation and sea level drop), leaving open ecological
niches for intermediate fauna taxa to diversify later on. Direct detrimen-
tal competition between the two faunas does not seem plausible, as both
were found in the same geographical areas earlier on (Fig. 5A-D). The
Early Silurian peak of origination rate (Fig. 1B) re]ects the large in-
crease in diversity of this intermediate fauna (Fig. 3A). Although it may
have started in the Hirnantian (Harper et al., 2014), most of the di-
versi\cation coincides with a global increase of temperatures, associated
with ice sheets melting and a sea level rise (Fig. 1F).
In the Ordovician, conodonts constituted most of the nekton's diver-
sity. The subsequent increase in nekton diversity during the Silurian /
Early Devonian (Whalen and Briggs, 2018;Klug et al., 2010) may
have played a role in the limitation of the conodont's recovery via ei-
ther predation pressures (the earliest known direct evidence of preda
tion on conodonts dates from the Late Devonian; see Zatoń and
Rakociński, 2014,Zatoń et al., 2017), and/or competition for eco-
logical spaces. This scenario, if real, may converge with that of the \nal
extinction of conodonts (see below).
The peak of conodont diversity in the Early Devonian occurred
within a temperature decreasing trend (with large variations) and co-
incided roughly with a lowstand (switch from a regressive to a trans-
gressive trend) in the middle Early Devonian (Fig. 1A, F). This series
displayed stable standing diversity in the intermediate fauna taxa, while
several late fauna taxa appeared (Fig. 3A). Extinction and origination
rates were almost equal, because several intermediate fauna taxa dis-
appeared by the end of the Early Devonian while the late evolutionary
fauna was rising (Fig. 1B). Although there was an increase of occupied
areas in the earliest Devonian, it was short lived (Fig. 1C-D), and was
not re]ected in the latitudinal extremes of the distribution, which re-
mained stable (Fig. 1E). This geographical pattern may be explained
by the position of the paleo-continents, which started to assemble. Ar-
eas occupied only by intermediate fauna taxa were either emptied or
became shared with late fauna taxa (Fig. 5F-H). This, combined with
the fact that intermediate evolutionary fauna taxa generally constitute
groups that are basal to late fauna taxa along the tree of Donoghue
et al. (2008), could mean that some species of the intermediate fauna
went through an episode of speciation at that time and gave rise to the
8

UNCORRECTED PROOF
%156;)5, 6<,-4)5, 36*)3)5,"3)5-;)9@0)5/- ??? ???? ??????
Fig. 6. Paleobiogeographic evolution of conodont diversity from Late Devonian to Middle Triassic. Paleomaps from Scotese (2015) are divided into a grid of 362 pentagonal and hexago-
nal ‘cells’(faces) with approximately equal area and ‘side’(edge) length. Average side length is 740.74 km. Conodont occurrences were automatically assigned to the corresponding cells,
based on their paleolatitudes and paleolongitudes from the Paleobiology Database. Each cell occupied by a colored circle includes at least one occurrence. Multi-colored circles denote the
presence of conodont genera belonging to the different ‘Evolutionary Faunas’described herein. Note that the proportions of each colour are entirely arbitrary and ,6 56; relate in any way
to the proportions of local generic diversity. Circles were plotted at the center of the cells and their position do not imply that the occurrences are actually found together at a \ner scale
(i.e. within a cell).
late fauna diversity, possibly linked with an increase in sea level (Fig.
3C).
The drop in diversity observed in the middle Devonian (Fig. 1A) is
mostly explained by extinctions of intermediate fauna and single-inter-
val taxa (Fig. 1A, 3A), which are not compensated by originations of
late fauna taxa. This drop in global diversity happens despite the coeval
sea-level high and may be related to the mid-Devonian hothouse (Fig.
1F), which may have negatively impacted the intermediate fauna taxa,
while favoring late fauna taxa. The loss of diversity is not re]ected by
a notable extinction peak (Fig. 1B), possibly because single-series gen-
era are not taken into account in the computation of the extinction rate
(Alroy, 2010b). The small peak of extinction rate in the Late Devonian,
despite an increase of SQS diversity (Fig. 1A-B), may also be due to the
loss of single-interval taxa and intermediate fauna taxa. Both intermedi-
ate and late fauna generic diversity go down across the Devonian –Car-
boniferous boundary although not re]ected by SQS diversity estimates.
Extinction rate culminates in the late Carboniferous, while some of
the lowest temperatures and lowest sea level of the entire studied pe-
riod are reached (Fig. 1B, F). Repeated or continuous glaciation(s) and
regression(s) marked the Carboniferous (Fielding et al., 2008). Those
clearly had a large impact on conodont diversity, as well as on their ge-
ographical extent, which was reduced to equatorial latitudes by the end
of the Carboniferous (Fig. 1C-E, 6C-D). Following the large extinc-
tion peak of the Pennsylvanian, conodont diversity was reduced to only
a handful of genera (Fig. 1A, 3A). Conodont diversity remained ex-
tremely low during the Early and Middle Permian, with conodonts being
restricted to a few refugia (Fig. 6E). In the Late Permian (Lopingian)
conodont diversity was still very low (Fig. 1A), but the geographical
extent of conodonts started again to increase, possibly re]ecting the in-
creasing temperatures and the concurrent melting of the ice sheets (Fig.
1C-F, 6F).
Finally, diversity rose again after the P-T boundary, with a peak of
origination and large numbers of single-interval taxa, mostly late fauna
taxa (Fig. 1A-B, 3A-B). This occurred while the sea level was very low,
and the temperatures very high (Fig. 1F). At a higher resolution, the
widely ]uctuating climate of the Early Triassic interval (e.g. Goude-
mand et al., 2019) may explain why so many taxa of this interval are
single-interval only, and therefore produced a peak of origination. The
fact that conodonts started to colonize new areas already in the Late Per-
mian may have driven allopatric speciations following the extreme con-
ditions of the P-T crisis.
It is clear from the paleomaps in Figs. 5 and 6 that conodonts were
more abundant along coastlines and on continental shelves. Not sur-
prisingly, the paleogeographical spread of conodonts is significantly cor-
related to the extent of continental shelves (Fig. 1C). Yet, conodonts
9

UNCORRECTED PROOF
%156;)5, 6<,-4)5, 36*)3)5,"3)5-;)9@0)5/- ??? ???? ??????
Fig. 7.
were not restricted to these areas, as evidenced by the open sea loca-
tions during the Ordovician and Devonian (e.g. Fig. 5B, G-H). Residual
variation in the geographical distribution of conodonts (Fig. 1D) corre-
sponds with the sea level curve, and the ice ages of the end-Ordovician
and Pennsylvanian clearly brought restrictions to the extension of con-
odonts, especially in their latitudinal extremes (Fig. 1E). Overall, our
data suggest that conodonts were not ecologically restricted to coastlines
or warm waters, but that they could not adapt to very cold / glacial wa-
ters. Extremely hot waters may also have been lethal to conodonts, but
in general, increased temperatures seems to have favored conodont di-
versi\cation.
It appears that the positions of paleo-continents had an in]uence on
the evolution of the latitudinal gradient of conodont diversity (Fig. 7).
Throughout the Ordovician, the LGGD is rather symmetrical between
the Northern and Southern Hemispheres. Conodonts were then present
in the northern extent of the Gondwanian continental shelf (Fig. 5B-C),
around Laurentia and Baltica, and further North in the open ocean. From
the Early Silurian (Llandovery) to the Carboniferous, the LGGD became
restricted mostly to the Southern Hemisphere, and was generally uni-
modal. This is in accordance with the southern location of most coast-
lines and continental shelves from the Llandovery until the Early Car-
boniferous. Northern movements of Pangea during the Carboniferous,
combined with the appearance of large South Pole ice sheets produced
a northward shift of the conodonts, and the LGGD became almost cen-
tered on the equator (Fig. 6D, Fig. 7). When conodont diversity recov-
ered in the Early Triassic, Pangea extended from North to South Pole,
with most continental shelves in the Northern Hemisphere (Fig. 6G-H).
Consequently the LGGD moved to the Northern Tropics, where it was
maintained until the end of the Triassic (Fig. 7). This interpretation is
in accordance with studies of both past and present latitudinal diversity
gradients that demonstrated a close link with the position of continental
shelves (e.g. Powell, 2009;Chaudhary et al., 2016).
Several of the major biases of the fossil record may affect our results.
Notably, as we showed that most conodonts occurrences are found in
continental shelves environments, their diversity and paleogeographic
patterns might be related to the quantity of preserved sedimentary rock
(Peters, 2005). The correlation between SQS diversity and sea-level,
which is also known to impact the amount of sedimentary rock, may
be an artifact of the preservation bias, although we would favor a com-
mon-cause hypothesis (Peters, 2005). Sampling effort bias is also prob-
ably playing a role in the paleogeographic patterns observed, as con-
odont workers have mostly explored facies expected to yield numerous
fossils. Some paleo-environments (e.g. hypersaline) are undersampled,
but can yield surprising diversity (Jarochowska et al., 2017). Finally,
although multi-element taxonomy has been adopted by conodont work-
ers for several decades now (Bergstrom, 1966), it is not always ap-
plicable, which may lead to taxonomic oversplitting and overestimations
of diversity. Notably, very few complete apparatuses are known for the
oldest conodont genera and the topological schemes used for establish-
ing homologies between elements vary between earlier and later taxa
(Donoghue et al., 2008).
Keeping these caveats and biases in mind, it appears that, indeed,
abiotic factors –temperature and sea-level, as well as paleocontinent
positions –in]uenced conodont diversity, as it is clear that the largest
10
Left panel. Paleolatitudinal gradients of conodont diversity through the Paleozoic, com-
puted as raw generic counts for each series, binned by 10° of latitude. Each generic diver-
sity point is plotted against the corresponding mid-bin value. The right panel shows the
extent of epicontinental seas, computed as the total number of grid cells including coast-
lines. The \gure as a whole follows stratigraphic order in relative terms, but does not rep-
resent absolute ages.

UNCORRECTED PROOF
%156;)5, 6<,-4)5, 36*)3)5,"3)5-;)9@0)5/- ??? ???? ??????
peaks of origination and extinction and diversity are synchronous with
peaks or lows in the global temperature and/or sea-level curve. When
testing these relationships, however, only a negative link between tem-
perature and extinction and a positive link between sea-level and diver-
sity were significant. Furthermore, both correlations were driven by ex-
treme points. It is interesting to note that it was not a necessary con-
dition to combine both sea-level and temperature to positively impact
conodont diversity. For example, the Ordovician radiation corresponded
with a large transgression and high temperatures, the Early Triassic orig-
ination and diversity peak was concomitant with a very low sea level
and hot and variable temperatures, while the Early Devonian diversity
peak (which was more limited) was synchronous with variable but not
particularly extreme sea-level and temperatures (Fig. 1A, F). This shows
that either factors can be self-suf\cient to have an in]uence. Addition-
ally, our proxy of geographical areas available to conodonts (number
of cells including coast or continental shelf) was significantly related to
occupied area (Fig. 1C), suggesting a partial explanation of how sea
level trends impacted the geographical spread of conodonts (Figs. 5-6).
Conodont mobility may have played a role in maintaining diversity in
times of low sea level and glaciation. Our results suggest that the main
events in the history of conodont diversity were related to extreme abi-
otic conditions. Yet, much variation in diversity, extinction and (in par-
ticular) origination remains unexplained, and could be due instead to
biotic factors. Notably, the \rst origination peak of the Early Ordovi-
cian, the Early Devonian peak of diversity and the \nal extinction of
conodonts require closer investigation, in particular regarding the inter-
action of conodonts with other marine groups.
 647)91:65 >1;0 79-=16<: :;<,1-: )5, 9-4)92: 65 ;0- A5)3 -?;15+;165 6.
+656,65;:
Our results regarding global conodont diversity trends are mostly
in accordance with previously published curves (Clark, 1983, 1987;
Sweet, 1988). Most of the major events of the group's history are rec-
ognized, in particular the Ordovician acme, followed by a major drop in
the Silurian. The Permian minimum, and the Early Triassic ‘last hurrah’
of diversity are also corroborated. On the other hand, we note discrep-
ancies for the Devonian and Carboniferous. In particular, previous stud-
ies described diversity and origination as low in the Early Devonian and
high in the Late Devonian. Conversely, our results show a peak in diver-
sity in the Early Devonian, followed by a drop in the Middle Devonian
and a very limited diversi\cation in the Late Devonian (Fig. 1A). Like-
wise, previous studies suggested important diversity drops during the
Mississipian and Pennsylvanian, while our results suggest on the con-
trary that diversity was fairly high and stable throughout the Carbonif-
erous, with a major extinction between the Pennsylvanian and the Per-
mian (Fig. 1A). These discrepancies may be due to several factors: i) the
geographical extent of the dataset (e.g. Sweet, 1988 focused on North
American data), ii) the taxonomic rank used (speci\c level in Sweet,
1988, versus generic level in Clark, 1983, 1987, and our study), iii)
the sheer increase in size of the dataset, or iv) methods to assess diver-
sity and rates. Here, we chose to use advances notably promoted by Al-
roy (2010a, 2010b) to limit the impact of several biases: for example,
Clark (1983) used a range-through counting method (although not de-
\ned as such), which is subject to edge and Signor-Lipps effects (the Pull
of the Recent having no impact on conodonts; Alroy, 2010b) meaning
that diversity estimates will tend to artefactually drop before and rise af-
ter boundaries, notably at mass extinction events. Despite these biases,
these older studies did manage to reveal the most prominent elements
also highlighted in the present work (Ordovician acme, Permian low and
Triassic short-lived recovery), suggesting that these trends are real, and
robust to any kind of biases, as well as across taxonomic levels.
Clark (1983) suggested that the \nal extinction of conodonts may
be linked to a drop in sea level, and was followed in this interpreta-
tion by Sweet (1988). However, sea-level was low throughout the Tri-
assic, and its variations are not agreed upon. Furthermore, recent stud-
ies suggest that it may have been rising, rather than dropping, during
the Late Triassic (e.g. Van der Meer et al., 2017). Even if the regres-
sive trend actually took place, its age was likely younger than the start
of conodonts' decline (Middle Triassic; Martínez-Pérez et al., 2013).
Furthermore our results suggest that regressive trends do not necessarily
correlate with major conodont extinctions. Global temperatures in the
Late Triassic were fairly similar to those in the Early and Middle Trias-
sic, and the world was ice-free at that time, as was generally the case
in times when conodonts were thriving. Based on our results, consid-
ering the absence of glaciation and a late drop in sea level and CAMP
province in]uence, we may conclude that two of the most important
abiotic factors (temperature and sea-level) cannot be considered as suf\-
cient to explain the extinction of conodonts. Other abiotic factors, which
are more dif\cult to assess, such as water salinity or pH may have played
a role. Alternatively, and in agreement with De Renzi et al. (1996)
and Martínez-Pérez et al. (2013), biotic factors such as predation by,
or competition with groups of the ‘Modern Fauna’(Sepkoski, 1981,
Hu et al., 2010,Brayard et al., 2017), e.g. Neopterygian \shes and
Neoselachian sharks that started radiating in the Middle Triassic (Cuny
and Benton, 1999;Xu et al., 2013), may have driven the last con-
odonts to extinction.
5. Conclusion
Our analysis con\rms the in]uence of abiotic factors on conodont di-
versity at a large scale. Notably, despite their mobility, these early ver-
tebrates were strongly impacted by sea-level variations as well as major
glaciations, which restricted the extent of their favored coastal habitat
and probably limited the possibility of allopatric speciation. Biases of the
fossil record should however be kept in mind as potentially confounding
or correlated factors. These abiotic factors do not however explain par-
ticular conodont events such as their Ordovician radiation or their \nal
demise at the end of the Triassic. Instead it is likely that biotic factors
played a prominent role in the extinction of conodonts. Despite the fact
that biotic interactions arguably take place locally in time and space,
the sum of their effects may ultimately emerge as large-scale patterns,
leading for example to the extinction of a highly successful group like
conodonts, which had thrived in oceans for more than 300 Ma.
Declaration of Competing Interest
The authors have no con]ict of interest to declare.
Acknowledgements
This work was supported by a French Agence Nationale de la
Recherche @Raction grant (ACHN project EvoDevOdonto
ANR-14-ACHN-0010)/.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.gloplacha.2020.103325.
References
Aguilée, R., Gascuel, F., Lambert, A., Ferriere, R., 2018. Clade diversi\cation dynamics
and the biotic and abiotic controls of speciation and extinction rates. Nat. Commun. 9
(1), 3013.
Alroy, J., 2010. The shifting balance of diversity among major marine animal groups.
Science 329 (5996), 1191–1194.
11

UNCORRECTED PROOF
%156;)5, 6<,-4)5, 36*)3)5,"3)5-;)9@0)5/- ??? ???? ??????
Alroy, J., 2010. Fair sampling of taxonomic richness and unbiased estimation of
origination and extinction rates. Paleontol. Soc. Pap. 16, 55–80.
J. Alroy M. Aberhan D. J. Bottjer M. Foote F. T. Fürsich P. J. Harries ... M. A. Kosnik
Phanerozoic trends in the global diversity of marine invertebratesSci-
ence32158852008 97100
Bahrami, A., Boncheva, I., Königshof, P., Yazdi, M., Khan-Abadi, A.E., 2014. Conodonts of
the Mississippian/Pennsylvanian boundary interval in Central Iran. J. Asian Earth Sci.
92, 187–200.
Balter, V., Renaud, S., Girard, C., Joachimski, M.M., 2008. Record of climate-driven
morphological changes in 376 Ma Devonian fossils. Geology 36 (11), 907–910.
Barrick, J.E., Lambert, L.L., Heckel, P.H., Rosscoe, S.J., Boardman, D.R., 2013.
Midcontinent Pennsylvanian conodont zonation. Stratigraphy 10 (1–2), 55–72.
Benton, M.J., 2009. The Red Queen and the Court Jester: species diversity and the role of
biotic and abiotic factors through time. Science 323 (5915), 728–732.
Bergstrom, S.M., 1966. Conodonts from the Lexington Limestone (Middle Ordovician) of
Kentucky, and its lateral equivalents in Ohio and Indiana. Bull. Am. Paleontol. 50
(229), 271–441.
A. Brayard G. Escarguel H. Bucher C. Monnet T. Brühwiler N. Goudemand ... J. Guex Good
genes and good luck: ammonoid diversity and the end-Permian mass extinctionSci-
ence32559442009 11181121
A. Brayard L. J.Krumenacker J. P. Botting J. F.Jenks K. G. Bylund E. Fara ... S. Charbon-
nier Unexpected Early Triassic marine ecosystem and the rise of the Modern evolu-
tionary faunaScience Advances322017 e1602159
Buggisch, W., Joachimski, M.M., Sevastopulo, G., Morrow, J.R., 2008. Mississippian
δ13Ccarb and conodont apatite δ18O records—their relation to the late Palaeozoic
Glaciation. Palaeogeogr. Palaeoclimatol. Palaeoecol. 268 (3–4), 273–292.
Chaudhary, C., Saeedi, H., Costello, M.J., 2016. Bimodality of latitudinal gradients in
marine species richness. Trends Ecol. Evol. 31 (9), 670–676.
Clark, D.L., 1983. Extinction of conodonts. J. Paleontol. 652–661.
Clark, D.L., 1987. Phylum Conodonta. In: Boardman, R.S., Cheetham, A.H., Rowell, A.J.
(Eds.), Fossil Invertebrates. Blackwell Scienti\e Publications, Oxford, pp. 636–662.
Core Team, R., 2018. R: A language and environment for statistical computing. In: R
Foundation for Statistical Computing Vienna, Austria. URL. https://www.R-project.
org/.
Cuny, G., Benton, M.J., 1999. Early radiation of the neoselachian sharks in Western
Europe. Geobios 32 (2), 193–204.
de Renzi, M., Budurov, K., Sudar, M., 1996. The extinction of conodonts - in terms of
discrete elements - at the Triassic-Jurassic boundary. J. Iberian Geol. Int. Publ. Earth
Sci. 20, 347–366.
Dong, X., Repetski, J.E., Bergström, S.M., 2004. Conodont biostratigraphy of the middle
Cambrian through lowermost Ordovician in Hunan, South China. Acta Geologica
Sinica- English Edition 78 (6), 1185–1206.
Donoghue, P.C., Purnell, M.A., Aldridge, R.J., Zhang, S., 2008. The interrelationships of
‘complex’conodonts (Vertebrata). J. Syst. Palaeontol. 6 (2), 119–153.
Fielding, C.R., Frank, T.D., Isbell, J.L., 2008. The late Paleozoic ice age—a review of
current understanding and synthesis of global climate patterns. Geol. Soc. Am. Spec.
Pap. 441, 343–354.
Friedman, M., Sallan, L.C., 2012. Five hundred million years of extinction and recovery:
a Phanerozoic survey of large-scale diversity patterns in \shes. Palaeontology 55 (4),
707–742.
Ginot, S., Goudemand, N., 2019. Conodont size, trophic level, and the evolution of
platform elements. Paleobiology 45 (3), 458–468.
Girard, C., Feist, R., 1996. Eustatic trends in conodont diversity across the
Frasnian–Famennian boundary in the stratotype area, Montagne Noire, Southern
France. Lethaia 29 (4), 329–337.
Goudemand, N., Romano, C., Leu, M., Bucher, H., Trotter, J.A., Williams, I.S., 2019.
Dynamic interplay between climate and marine biodiversity upheavals during the
early Triassic Smithian-Spathian biotic crisis. Earth Sci. Rev..
Hallam, A., 2002. How catastrophic was the end-Triassic mass extinction? Lethaia 35 (2),
147–157.
Hannisdal, B., Peters, S.E., 2011. Phanerozoic Earth system evolution and marine
biodiversity. Science 334 (6059), 1121–1124.
Haq, B.U., Schutter, S.R., 2008. A chronology of Paleozoic Sea-level changes. Science 322
(5898), 64–68.
Haq, B.U., Hardenbol, J.A.N., Vail, P.R., 1987. Chronology of ]uctuating sea levels since
the Triassic. Science 235 (4793), 1156–1167.
Harper, D.A., Hammarlund, E.U., Rasmussen, C.M., 2014. End Ordovician extinctions: a
coincidence of causes. Gondwana Res. 25 (4), 1294–1307.
Harper, D.A., Zhan, R.B., Jin, J., 2015. The Great Ordovician Biodiversi\cation Event:
reviewing two decades of research on diversity’s big bang illustrated by mainly
brachiopod data. Palaeoworld 24 (1–2), 75–85.
Hofmann, R., Tietje, M., Aberhan, M., 2019. Diversity partitioning in Phanerozoic benthic
marine communities. Proc. Natl. Acad. Sci. 116 (1), 79–83.
S. X. Hu Q. Y. Zhang Z. Q. Chen C. Y. Zhou T. Lü T. Xie ... M. J. Benton The Luoping
biota: exceptional preservation, and new evidence on the Triassic recovery from
end-Permian mass extinctionProceedings of the Royal Society B: Biological Sci-
ences27817162010 22742282
Jarochowska, E., Viira, V., Einasto, R., Nawrot, R., Bremer, O., Männik, P., Munnecke, A.,
2017. Conodonts in Silurian hypersaline environments: Specialized and unexpectedly
diverse. Geology 45 (1), 3–6.
Joachimski, M.M., von Bitter, P.H., Buggisch, W., 2006. Constraints on Pennsylvanian
glacioeustatic sea-level changes using oxygen isotopes of conodont apatite. Geology
34 (4), 277–280.
M. M. Joachimski S. Breisig W. Buggisch J. A. Talent R. Mawson M. Gereke ... K. Weddige
Devonian climate and reef evolution: insights from oxygen isotopes in apatiteEarth
and Planetary Science Letters2843‐42009 599609
C. Klug B. Kröger W. Kiessling G. L. Mullins T. Servais J. Frýda ... S. Turner The Devonian
nekton revolutionLethaia4342010 465477
Kocsis, Á.T., Reddin, C.J., Kiessling, W., 2018. The biogeographical imprint of mass
extinctions. Proc. R. Soc. B Biol. Sci. 285 (1878), 20180232.
Lê, S., Josse, J., Husson, F., 2008. FactoMineR: an R package for multivariate analysis. J.
Stat. Softw. 25 (1), 1–18.
Lucas, S.G., Tanner, L.H., 2004. Late Triassic extinction events. Albertiana 31, 31–40.
Martínez-Pérez, C., Plasencia, P., Cascales-Miñana, B., Mazza, M., Botella, H., 2013. New
insights into the diversity dynamics of Triassic conodonts. Hist. Biol. 26 (5), 591–602.
Mazza, M., Furin, S., Spötl, C., Rigo, M., 2010. Generic turnovers of Carnian/Norian
conodonts: Climatic control or competition? Palaeogeogr. Palaeoclimatol. Palaeoecol.
290 (1–4), 120–137.
T. Miyashita M. I. Coates R. Farrar P. Larson P. L. Manning R. A. Wogelius ... P. J. Currie
Hag\sh from the Cretaceous Tethys Sea and a reconciliation of the morphological–
molecular con]ict in early vertebrate phylogenyProceedings of the National Academy
of Sciences11662019 21462151
Montañez, I.P., Poulsen, C.J., 2013. The late Paleozoic ice age: an evolving paradigm.
Annu. Rev. Earth Planet. Sci. 41, 629–656.
Orchard, M.J., 2007. Conodont diversity and evolution through the latest Permian and
early Triassic upheavals. Palaeogeogr. Palaeoclimatol. Palaeoecol. 252 (1–2), 93–117.
Payne, J.L., Clapham, M.E., 2012. End-Permian mass extinction in the oceans: an ancient
analog for the twenty-\rst century? Annu. Rev. Earth Planet. Sci. 40, 89–111.
Peters, S.E., 2005. Geologic constraints on the macroevolutionary history of marine
animals. Proc. Natl. Acad. Sci. 102 (35), 12326–12331.
Powell, M.G., 2005. Climatic basis for sluggish macroevolution during the late Paleozoic
ice age. Geology 33 (5), 381–384.
Powell, M.G., 2009. The latitudinal diversity gradient of brachiopods over the past 530
million years. J. Geol. 117 (6), 585–594.
Preto, N., Kustatscher, E., Wignall, P.B., 2010. Triassic climates—state of the art and
perspectives. Palaeogeogr. Palaeoclimatol. Palaeoecol. 290 (1–4), 1–10.
Purnell, M.A., 2001. Scenarios, selection, and the ecology of early vertebrates. In: Ahlberg,
P.E. (Ed.), Major events in early vertebrate evolution: Palaeontology, Phylogeny,
Genetics and Development. Systematics Association Species Volume Series, n°61.
Taylor and Francis, London, pp. 187–208.
Purnell, M.A., Jones, D., 2012. Quantitative analysis of conodont tooth wear and damage
as a test of ecological and functional hypotheses. Paleobiology 38 (4), 605–626.
C. Romano N. Goudemand T. W. Vennemann D. Ware E. Schneebeli-Hermann P.
A. Hochuli ... H. Bucher Climatic and biotic upheavals following the end-Permian mass
extinctionNature Geoscience612013 57
Roopnarine, P.D., Angielczyk, K.D., Wang, S.C., Hertog, R., 2007. Trophic network models
explain instability of early Triassic terrestrial communities. Proc. R. Soc. B Biol. Sci.
274 (1622), 2077–2086.
Scotese, C.R., 2015. Some thoughts on global climate change: the transition from icehouse
to hothouse. PALEOMAP Project, 21a 1 (2).
Scotese, C.R., 2016. Tutorial: PALEOMAP PaleoAtlas for GPlates and the PaleoData Plotter
Program.
Sepkoski, J.J., 1981. A factor analytic description of the Phanerozoic marine fossil record.
Paleobiology 7 (1), 36–53.
Servais, T., Harper, D.A., 2018. The Great Ordovician Biodiversi\cation Event (GOBE):
de\nition, concept and duration. Lethaia 51 (2), 151–164.
Servais, T., Lehnert, O., Li, J.U.N., Mullins, G.L., Munnecke, A., Nuetzel, A., Vecoli, M.,
2008. The Ordovician Biodiversi\cation: revolution in the oceanic trophic chain.
Lethaia 41 (2), 99–109.
Song, H., Wignall, P.B., Song, H., Dai, X., Chu, D., 2019. Seawater temperature and
dissolved oxygen over the past 500 million years. J. Earth Sci. 30 (2), 236–243.
Stanley, S.M., Powell, M.G., 2003. Depressed rates of origination and extinction during the
late Paleozoic ice age: a new state for the global marine ecosystem. Geology 31 (10),
877–880.
Y. Sun M. M. Joachimski P. B. Wignall C. Yan Y. Chen H. Jiang ... X. Lai Lethally hot tem-
peratures during the Early Triassic greenhouseScience33861052012 366370
Sweet, W.C., 1988. The Conodonta: morphology, taxonomy, paleoecology, and
evolutionary history of a long-extinct animal phylum (Vol. 10, pp. 1212). Clarendon
Press, Oxford.
Tanner, L.H., Lucas, S.G., Chapman, M.G., 2004. Assessing the record and causes of late
Triassic extinctions. Earth Sci. Rev. 65 (1–2), 103–139.
van de Schootbrugge, B., Wignall, P.B., 2016. A tale of two extinctions: converging end-
Permian and end-Triassic scenarios. Geol. Mag. 153 (2), 332–354.
Van der Meer, D.G., van Saparoea, A.V.D.B., Van Hinsbergen, D.J.J., Van de Weg, R.M.B.,
Godderis, Y., Le Hir, G., Donnadieu, Y., 2017. Reconstructing \rst-order changes in sea
level during the Phanerozoic and Neoproterozoic using strontium isotopes. Gondwana
Res. 44, 22–34.
Whalen, C.D., Briggs, D.E., 2018. The Palaeozoic colonization of the water column and the
rise of global nekton. Proc. R. Soc. B Biol. Sci. 285 (1883), 20180883.
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).
Xu, G.H., Zhao, L.J., Gao, K.Q., Wu, F.X., 2013. A new stem-neopterygian \sh from
the Middle Triassic of China shows the earliest over-water gliding strategy of the
vertebrates. Proc. R. Soc. B Biol. Sci. 280 (1750), 20122261.
12

UNCORRECTED PROOF
%156;)5, 6<,-4)5, 36*)3)5,"3)5-;)9@0)5/- ??? ???? ??????
Zatoń, M., Rakociński, M., 2014. Coprolite evidence for carnivorous predation in a late
Devonian pelagic environment of southern Laurussia. Palaeogeogr. Palaeoclimatol.
Palaeoecol. 394, 1–11.
Zatoń, M., Broda, K., Qvarnström, M., Niedźwiedzki, G., Ahlberg, P.E., 2017. The \rst
direct evidence of a late Devonian coelacanth \sh feeding on conodont animals. Sci.
Nat. 104 (3–4), 26.
Zhen, Y.Y., Percival, I.G., Woo, J., Park, T.Y.S., 2018. Latest Cambrian–Earliest Ordovician
Conodonts and Microbrachiopods from Northern Victoria Land. Handler Ridge
revisited. Palaeoworld, Antarctica.
13