Content uploaded by Catalina Pimiento
Author content
All content in this area was uploaded by Catalina Pimiento on Sep 26, 2017
Content may be subject to copyright.
Articles
DOI: 10.1038/s41559-017-0223-6
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
The Pliocene marine megafauna extinction and its
impact on functional diversity
Catalina Pimiento 1,2,3,4*, John N. Griffin 4, Christopher F. Clements5, Daniele Silvestro6,7,
Sara Varela3, Mark D. Uhen 8 and Carlos Jaramillo 2
The end of the Pliocene marked the beginning of a period of great climatic variability and sea-level oscillations. Here, based on
a new analysis of the fossil record, we identify a previously unrecognized extinction event among marine megafauna (mammals,
seabirds, turtles and sharks) during this time, with extinction rates three times higher than in the rest of the Cenozoic, and with
36% of Pliocene genera failing to survive into the Pleistocene. To gauge the potential consequences of this event for ecosystem
functioning, we evaluate its impacts on functional diversity, focusing on the 86% of the megafauna genera that are associated with
coastal habitats. Seven (14%) coastal functional entities (unique trait combinations) disappeared, along with 17% of functional
richness (volume of the functional space). The origination of new genera during the Pleistocene created new functional entities
and contributed to a functional shift of 21%, but minimally compensated for the functional space lost. Reconstructions show that
from the late Pliocene onwards, the global area of the neritic zone significantly diminished and exhibited amplified fluctuations.
We hypothesize that the abrupt loss of productive coastal habitats, potentially acting alongside oceanographic alterations, was a
key extinction driver. The importance of area loss is supported by model analyses showing that animals with high energy require-
ments (homeotherms) were more susceptible to extinction. The extinction event we uncover here demonstrates that marine
megafauna were more vulnerable to global environmental changes in the recent geological past than previously thought.
In the Anthropocene, rapid environmental change and the resultant
loss of habitat pose a major threat to marine fauna1,2. Throughout
geological time, habitat loss caused by sea-level changes has been
widely associated with extinction events3. After the last mass extinc-
tion at the Cretaceous/Palaeogene boundary and throughout the
past 66 million years, the largest global sea-level changes occurred
mainly during the Pliocene and Pleistocene epochs (herein, the
Plio− Pleistocene; from 5.33 to 0.01 million years ago (Ma)), with
multiple large eustatic oscillations that were amplified after the
onset of the Northern Hemisphere glaciation in the late Pliocene4–7.
Although it has been proposed that global cooling and sea-
level fluctuations in the Plio− Pleistocene were responsible for the
regional extinction of marine invertebrates8, it has been assumed
that global marine biodiversity was generally resistant to these
environmental changes3,9. Individual examples of faunal turnover
and extinctions of large marine vertebrates (collectively known as
‘marine megafauna’, which includes, but is not limited to marine
mammals, seabirds, turtles, sharks and rays10,11) have been observed
around this period. These include a substantial drop in cetacean12–14
(but see ref. 15) and penguin diversity16,17, the extinction of dugon-
gids in the Western Atlantic and Mediterranean regions18–20, the loss
of the largest shark that ever lived (Carcharocles megalodon)21,22 and
extinctions of sea turtles (for example, Psephophorus, a leatherback
turtle)23. However, it remains unclear whether these megafauna
losses were simply conspicuous background extinctions or formed
part of a global marine extinction event resulting from the envi-
ronmental changes of the Plio− Pleistocene8,24. Evaluating the extent
and consequences of the marine megafauna extinctions is relevant
because these organisms play fundamental roles in ecosystems25–27
and because modern megafauna assemblages were established dur-
ing the Pleistocene (for example, ref. 28; Supplementary Fig.1).
Historically, studies of marine extinctions have focused almost
exclusively on taxonomic loss (for example, species, genera and
family; but see ref. 29). While this taxonomic perspective quanti-
fies the loss of diversity sensu stricto (for example, ref. 30), it ignores
the ecological contributions of these species to ecosystems. Linking
taxonomic identity with ecological roles can be used to assess the
selectivity of extinctions24,31–35, to evaluate shifts in the structure of
communities after an extinction event32 and to gauge the potential
implications for ecosystem functioning36. This ‘functional diversity’
approach (reviewed in ref. 32) consists of quantifying the distribu-
tion of species in a multidimensional functional space defined by
species traits (that is, the intrinsic characteristics of species that
directly influence their ecological role32). The few studies that have
used this or similar approaches have focused specifically on the
ecological consequences of the extinction of invertebrates24,33–35 (but
see ref. 37). These organisms have important ecological roles, but are
usually small in size, occupy low trophic levels and tend to be highly
speciose. Conversely, marine vertebrates include the largest organ-
isms on Earth, occupy a variety of trophic roles, are relatively species
poor and are accordingly less likely to be ecologically redundant38.
Moreover, they are often wide ranging and are known to struc-
ture modern food webs from the top down25. The goal of linking
the extinctions of large animals with consequences for ecosystem
1 Paleontological Institute and Museum, University of Zurich, 8006 Zurich, Switzerland. 2 Smithsonian Tropical Research Institute, PO Box 2072,
Balboa, Panama. 3 Museum für Naturkunde, Leibniz Institute for Evolution and Biodiversity Science, Invalidenstrasse 43, 10115 Berlin, Germany.
4 Department of Biosciences, Swansea University, Wallace Building, Singleton Park, Swansea SA2 8PP, UK. 5 Department of Evolutionary Biology
and Environmental Studies, University of Zurich, 8057 Zurich, Switzerland. 6 Department of Biological and Environmental Sciences, University of
Gothenburg and Gothenburg Global Biodiversity Centre, 405 30 Gothenburg, Sweden. 7 Department of Computational Biology, University of Lausanne,
1011 Lausanne, Switzerland. 8 Department of Atmospheric, Oceanic, and Earth Sciences, George Mason University, Fairfax, VA 22030, USA.
*e-mail: catalina.pimientohernandez@pim.uzh.ch
NATURE ECOLOGY & EVOLUTION | www.nature.com/natecolevol
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved. © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Articles NaTuRe eCology & evoluTioN
functioning is particularly relevant today, as large-bodied marine
species are the most vulnerable to current human impacts2.
Here, we evaluate the severity of the extinction of marine mega-
fauna during the Pliocene, and examine the potential causes and
consequences of this event. We first assess whether the Pliocene
extinction rates were higher than those of the rest of the Cenozoic,
and examine the proportional loss of genera from the Pliocene to
the Pleistocene. Then, we quantify the differences in functional
diversity between the Pliocene and Pleistocene coastal megafauna
assemblages, to assess the potential effects of extinctions on eco-
system functioning. Finally, we evaluate the possible drivers of
extinction by estimating the habitat loss due to eustatic variations,
and by modelling traits as predictors of extinction. The results of
this research provide a broader understanding of the state and vul-
nerability of the marine megafauna in the recent geological past,
and forewarn of the likely sensitivity of megafauna biodiversity as
anthropogenic climate change accelerates and brings massive per-
turbations to coastal ecosystems39–41.
Results and discussion
The extinction event. We estimated the expected number of extinc-
tion events per genus per Myr during the Cenozoic, while account-
ing for preservation biases and dating uncertainties using a Bayesian
framework42 (Supplementary Tables1–3 and Supplementary Fig.2).
Marine megafauna present significantly elevated extinction rates
in the Pliocene (Fig.1a), with a threefold increase relative to the
rest of the Cenozoic, and with the highest rates occurring in the
late Pliocene, specifically between 3.8 and 2.4 Ma (Supplementary
Fig.3). Per-clade analyses reveal that all groups of marine megafauna
follow this trend except sea birds, which present higher extinction
rates in the Paleocene (Supplementary Fig.4). Conversely, we did
not find evidence of changes in origination rates during the entire
Cenozoic (Supplementary Fig. 3). Until now, disappearances of
Pliocene marine megafauna species were thought to represent iso-
lated examples within a broader assemblage that remained largely
intact (for example, ref. 3, but see ref. 43). Our results show that these
extinctions, which peaked in the late Pliocene, were part of a hith-
erto unrecognized global loss of marine megafauna biodiversity.
Closer examination of the Pliocene megafauna fossil record
reveals the proportional losses of genera (see Supplementary Table4
for numbers of genera and samples). We found that 36% of Pliocene
genera were extirpated (that is, not present in the Pleistocene). In
line with previous studies3,44, marine mammals present the high-
est proportional extinction, losing 55% of their generic diversity
(for example, the aquatic sloth Thalassocnus and the beluga-
like odontocete Bohaskaia). Seabirds lost 35% of their generic
diversity (for example, the penguin Inguza), sea turtles 43% (for
example, Syllomus and Psephophorus) and sharks 9% (for example,
Carcharocles) (see Fig.1b and Supplementary Table 4). New gen-
era also evolved: 25% of the Pleistocene genera were new (that is,
they were not reported in the Pliocene), including 38% of mammals
(for example, the polar bear Ursus) and 41% of seabirds (for exam-
ple, the storm petrel Oceanodroma and the penguin Megadyptes)
(Supplementary Table 4). Nevertheless, in line with the elevated
extinction relative to origination rates, generic diversity of global
megafauna suffered a net decline of 15% between the Pliocene
and Pleistocene. Furthermore, we found that most of the Plio−
Pleistocene marine megafauna (86%; Supplementary Table4) were
associated with coastal habitats (that is, the neritic zone at a depth of
less than 200 m), where the absolute loss of genera was greater (see
Fig.1b and Supplementary Dataset1). However, since this finding
could be biased by differential fossil preservation and/or sampling,
it should be interpreted with caution.
Impacts on functional diversity. To assess the potential effects of the
Pliocene extinction on ecosystem functioning, we performed trait-
based analyses following the methods described by Mouillot et al.32
for genera associated with coastal environments. Accordingly, we
assigned traits to the Plio− Pleistocene coastal megafauna (184
genera; Supplementary Table 4) to (1) determine the functional
entities (groups with unique trait combinations, herein FEs) and
(2) construct a functional space32. The coastal megafauna data-
set includes 146 Pliocene genera from 711 occurrences and 129
Pleistocene genera from 858 occurrences (Supplementary Table4).
We found that 55 (38%) coastal Pliocene genera went extinct
(Fig.2c), resulting in the loss of seven (14%) FEs (Fig.2a) along with
17% of functional richness (Fig.2e; the volume of functional space
after accounting for sample size differences (see Supplementary
Fig.5 and the Methods)). The post-extinction Pleistocene assem-
blage hosted 38 new genera (29%), reducing the net taxonomic loss
of coastal habitats to 12% (Fig.2d). The evolution of these new gen-
era resulted in the addition of four FEs (9%; Fig.2b) and the net
loss of three FEs. However, these new FEs, which were exclusively
occupied by mammals, minimally compensated for the functional
richness lost (by 1%), leaving a net functional richness loss of 16%.
Furthermore, the loss and gain of Plio− Pleistocene FEs drove a func-
tional shift (non-overlap of functional volume32) of 21% (Fig.2f).
We next investigated the interaction between extinction and the
functional structure of the megafauna assemblage. The functional
structure of the Pliocene assemblage ultimately rendered it sensitive
All
Mammals
Seabirds
Turtles
Sharks
% of genera extinct
0
10
20
30
40
b
50
a
Palaeocene
Eocene
Oligocene
Miocene
Pliocene
Pleistocene
0.05
0.10
0.15
Extinction rate
Epoch
CoastalOceanic
Figure 1 | Elevated extinction rates of marine megafauna in the late Pliocene. a, Extinction rates within epochs. The extinction rates in the Pliocene are
significantly higher than in any other epoch in the Cenozoic: they are 2.2-fold higher than in the Miocene, 60% higher than in the Pleistocene and threefold
higher than the average Cenozoic rate (n = 11,241 global occurrences). Boxes show the central 50% posterior credible intervals with error bars indicating
the 95% credible intervals. b, Proportional extinction of the Pliocene megafauna. Species associated with coastal environments (strictly coastal, coastal–
terrestrial or coastal–oceanic) represent 86% of the megafauna. Strictly oceanic species represent the remaining 14% of the megafauna.
NATURE ECOLOGY & EVOLUTION | www.nature.com/natecolevol
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved. © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Articles
NaTuRe eCology & evoluTioN
in the face of taxonomic extinctions: although it had an average of
three genera per FE (functional redundancy sensu45), they were con-
centrated within specific FEs (over-redundancy45), leaving over half
with only a single genus (functional vulnerability45) (Supplementary
Fig.6 and Supplementary Table6). All lost and gained FEs (except
one) contained a single genus (Fig.2a,b, see legend), suggesting that
low-redundancy FEs largely drove the changes in functional space.
The net losses of functional richness and the functional shift were
greater than expected given the mean background extinction rate over
the Cenozoic (22 genera; see the Methods) and the new Pleistocene
FEs (Supplementary Fig.7a,b). However, these functional changes
were no different than would be expected given the 55 genera lost
(Supplementary Fig.7c,d) and the functional structure of the assem-
blage, indicating that the loss of genera per se from the functionally
vulnerable Pliocene assemblage, rather than the observed pattern
of genera loss, determined the functional changes. Ultimately, the
Pleistocene assemblage was left with a greater proportion of single
genus FEs (80%)—that is, a greater functional vulnerability—than the
pre-extinction Pliocene assemblage (59%) (Supplementary Table6).
In light of the increasing volume of literature linking functional
diversity to ecosystem functioning46–49, it follows that the contribu-
tions of megafauna to marine ecosystems may have been diminished
(loss of functional richness), altered (functional shift) and rendered
less resistant to subsequent extinctions (increased functional vulner-
ability) after the Pliocene extinction event.
A common finding among the handful of previous studies that
have used a multi-trait-based approach in this context is that losses
in global functional diversity are negligible after an extinction event,
even in the face of mass extinctions, when greater than 70% of gen-
era were lost33–35. Our detection of a larger, although still modest
(16–21%), functional diversity change, despite lower taxonomic loss
(38%), is probably because most previous studies have focused on
benthic, smaller-bodied, more speciose invertebrate assemblages,
while ours focuses on large vertebrates. Ecosystems hold fewer large
than small species50, thus among large-bodied species there is likely to
be less scope for functional insurance provided by redundant species,
making functional diversity among large animals more sensitive in the
face of extinction51,52. This conclusion is supported by the high lev-
els of functional vulnerability among Plio− Pleistocene coastal mega-
fauna (Supplementary Table6), the singular roles megafauna taxa are
thought to play in modern systems26 and the accumulating evidence of
ecosystem consequences induced by their declines25.
−0.5
0.0
0.5 −0.5 0.0 0.5 −0.5 0.00.5
−0.5 0.0 0.5
−0.5
0.0
0.5
−0.5 0.00.5
−0.5
0.0
0.5
−0.5
0.0
0.5
a
cd
ef
b
A1 A1
A3 A3
A2A4
A2 A4
Pliocene (pre-extinction) Pleistocene (post-extinction)
0
20
40
60
80
100
Shift
Functional richness
0
20
40
60
80
100
120
140
Genera
Post
Pre
Post
Pre
Post
Pre
Post
Pre
∆ = 38%
∆ = 12% ∆ = 17% ∆ = 16%
Figure 2 | Changes in coastal marine megafauna functional diversity
from the Pliocene (pre-extinction) to the Pleistocene (post-extinction).
a,b, Functional space plotted using the first four axes (A1–A4) from a
principal coordinate analysis and the empirical data (not accounting for
differences in sample size). The coloured dashed line denotes the the convex
hull and the dots represent the Functional Entities (FEs): Pliocene = 49 FEs;
Pleistocene = 46 FEs. The filled dots denote FEs that changed (those
that were either extirpated or originated), whereas the open dots denote
unchanged FEs (winners). Note that since multiple genera can occupy a
single FE, the loss or gain of genera does not necessarily result in the lost or
gain of a FE. FE codes can be found in Supplementary Table5. Refer to the
Methods section ‘Functional traits and functional taxonomic units’ for details
on the differences between taxonomic levels. a, Pliocene space showing FEs
that went extinct and their taxonomic affiliations. Dark blue = FE 25, one
genus (Carcharocles, Lamniformes); light blue = FE 36, one genus (Parotodus,
Lamniformes); red = FE 27, two genera (Cetotherium and Nannocetus,
Mysticeti); green = FE 49, one species (Herpetocetus morrow, Mysticeti);
light grey = FE 50, one genus (Nanosiren, Sirenia); yellow = FE 52, one genus
(Thalassocnus, Xenarthra); and dark grey = FE 46, one genus (Psephophorus,
Testudines). b, Pleistocene space showing the new FEs and their taxonomic
affiliations. Pink = FE 47, one genus (Mirounga, Pinnipedia); green = FE 13,
one species (Orcinus orca, Odontoceti); blue = FE 31, one genus (Proterozetes,
Pinnipedia); and red = FE 3, one genus (Ursus, Carnivora). c,d, Taxonomic
richness (number of genera) loss after the extinction event. c, Raw genus
loss (not including the new genera that originated in the Pleistocene).
d, Net genus loss (including the new genera that originated in the
Pleistocene). e,f, Functional richness (functional space volume32) loss and
shift. The vertical lines are error bars resulting from the 1,000 permutations
of the resampled data (see the Methods). e, Raw functional richness.
f, Net functional richness.
Neritic area (million km
2
)
15
20
25
30
35
Time (Ma)
0
12345
Late
Pliocene
PleistocenePliocene
Figure 3 | Reduction of neritic areas as a putative extinction driver. Global
extent of neritic areas based on eustatic levels reported by deBoer et al.5.
The horizontal lines represent mean values for the Pliocene and Pleistocene.
NATURE ECOLOGY & EVOLUTION | www.nature.com/natecolevol
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved. © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Articles NaTuRe eCology & evoluTioN
Extinction mechanisms. It has been widely stated in the literature
that the onset of the Northern Hemisphere glaciation at the end of
the Pliocene resulted in an abrupt transition to a new climatic state
dominated by colder and more variable temperatures, and large sea-
level oscillations4–7,12. However, to our knowledge, there have been
no quantitative and global assessments of consequent changes in the
extent of coastal habitats during this time (but see ref. 53 for specific
regions). We therefore reconstructed the global extent of neritic
areas based on eustatic variations4,5 during the Plio− Pleistocene
and found that as the sea level regressed, neritic areas dropped pre-
cipitously during the late Pliocene. This abrupt change coincides
in time with the highest extinction rates found (Supplementary
Fig.3). After this sudden drop in coastal habitat availability, large-
area oscillations took place: there was a 250% increase in the coef-
ficient of variation from the Pliocene (0.07) to the Pleistocene
(0.17). Additionally, the total neritic area available was significantly
reduced from 79.1 million km2 in the Pliocene to 57.9 million km2
in the Pleistocene, representing a 27% reduction in the mean area
(t-test: P < 0.001; Fig.3 and Supplementary Fig.8).
Numerous studies have identified a regional invertebrate extinc-
tion during the Plio− Pleistocene and attributed this to climatic
changes (mainly temperature)8,54–56, but only one region-specific
study has implicated habitat loss associated with sea-level changes
in extinctions57. Here, we document a global-scale reduction in
coastal habitat availability that abruptly started in the late Pliocene
and hypothesize that this, probably acting alongside oceanographic
alterations such as changes in productivity and ocean circulation (for
example, refs 24,58–61) was the extinction driver for the Pliocene marine
megafauna. Some genera may have only succumbed to repeated sea-
level oscillations or when habitat loss coincided with other extinc-
tion drivers (for example, prey availability and/or competition)62,63,
which may explain the continuation of elevated extinction rates
in the Pleistocene. Similar mechanisms might be responsible for
the previously noted decline of some megafauna groups in the late
Miocene16,43, although such losses were not comparable in magni-
tude to the Pliocene losses documented here (Fig.1a).
To assess extinction selectivity, we modelled traits as predictors
of survivorship using generalized linear models. Thermoregulation
is the trait that best predicts extinctions in the Pliocene (Fig.4a),
with endotherms and mesotherms (homeotherms; that is, those able
to regulate, at least to some degree, their internal temperatures64,65)
having significantly higher chances of going extinct than their poiki-
lothermic counterparts (Fig.4b and Supplementary Tables7 and 8).
Large body size—a trait associated with extinction risk in the
Anthropocene3,66—does not predict extinction risk, nor affect the
explanatory power of thermoregulation. Although we found a taxo-
nomic signal in extinction probabilities, with mammals and sharks
presenting significant differences (the grey part of Supplementary
Tables7 and 8), the signal of thermoregulation was independent and
held when ‘class’ (Aves, Chondrichthyes, Mammalia and Reptilia)
was controlled as a fixed or random factor in the generalized linear
models (Supplementary Table9). Notably, the homeotherms that
became extinct were not exclusively endothermic mammals or sea-
birds: three of the five mesotherms (two shark genera and a turtle
genus) were also lost.
Unlike poikilotherms, homeotherms are buffered against exter-
nal temperature changes, but require greater resources to sustain
higher metabolic demands64,65,67,68. Homeotherms should there-
fore show greater extinction susceptibility in the face of declining
habitat and associated resource availability69, as our results show.
In contrast, if temperature fluctuations or overall cooling had
directly driven this extinction, the opposite result would have been
expected (that is, a greater susceptibility of poikilotherms compared
with homeotherms). Feeding plasticity, as grey whales seem to have
exhibited during the late Pleistocene, and possibly even across the
Pliocene–Pleistocene boundary70, may have aided the survival of
some homeothermic genera in the face of habitat loss. Overall, the
greater susceptibility of energy-demanding homeotherms supports
our hypothesis that the abrupt reduction of neritic areas was a key
driver of the marine megafauna extinction. Whether thermoregula-
tion covaries with other traits (for example, those associated with
extinction risk)71, and the effects of such correlations in determin-
ing the selectivity of the Pliocene marine megafauna extinction, are
beyond the scope of this study but should be further explored.
Conclusions
Here, we report an extinction and consequent erosion of functional
diversity of marine megafauna during the Pliocene. We propose that
these extinctions were driven by habitat loss produced by sea-level
oscillations, probably acting alongside other oceanographic alterations
such as changes in productivity and ocean circulation, in addition to
biotic drivers such as prey availability and/or competition. Since the
modern marine megafauna became established in the Pleistocene
(Supplementary Fig. 1), this event shaped the Earth’s present-day
assemblages of these large ecosystem-structuring organisms (for exam-
ple, refs 25,27,32,72). The discovery of this extinction event reveals that the
biodiversity and functional contributions of marine megafauna were
more sensitive to environmental changes in the recent geological past
than hitherto assumed. Today, and historically, over-exploitation has
been considered the chief threat to marine megafauna27. Our study
cautions that as anthropogenic climate change accelerates and triggers
regime shifts in coastal ecosystems39–41 the potential consequences for
marine megafauna should not be underestimated.
Methods
Cenozoic dataset. We downloaded all the records of marine megafauna for
the Cenozoic (that is, the past 66 Ma) from the Paleobiology Database (https://
paleobiodb.org; last search: November 2016). e Paleobiology Database follows
Guild
0
20
40
Independent eects (%)
60
80
ThermoregulationVertical
position
Habitat
a b
Maxium size Endotherm
0
0.5
1.0
Mesotherm Poikilotherm
*
Extinction probability
Figure 4 | Thermoregulation explains the susceptibility of genera to the Pliocene megafauna extinction. a, Hierarchical partitioning output based on
generalized linear models showing the proportion of explained deviance in extinction probabilities that can be attributed to each trait. In the full model, traits
collectively explained 20% of deviance in extinction probabilities. b, Extinction probabilities among the thermoregulation categories (see Supplementary Table8).
The vertical lines denote error bars and the asterisk denotes statistical significance compared with both of the other categories according to Tukey tests.
NATURE ECOLOGY & EVOLUTION | www.nature.com/natecolevol
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved. © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Articles
NaTuRe eCology & evoluTioN
the most recent geological timescale of Gradstein et al.73. In the absence of a
formal, size-based denition of ‘marine megafauna’, we included all the genera of
the groups of animals that contain the largest marine vertebrates (that is, marine
mammals, seabirds, sea turtles, and sharks and rays10). We focused on the genus
level because generic assignments have greater consistency across dierent research
groups, and because it is more robust to taxonomic error than the species level. All
taxonomic identications were evaluated and corrected. Dubious and equivocal
records were excluded from our analyses. Accordingly, we used 11,241 occurrences
(Supplementary Table1). Details on the search criteria and data assessment can be
found in theSupplementary Information. Furthermore, all references supporting
the occurrences can be found in Supplementary Dataset2.
Extinction rates. We estimated the extinction and origination rates of marine
megafauna for the entire Cenozoic using PyRate74. This program implements
Bayesian algorithms to analyse all available fossil occurrences (identified to genus
level in this case) while accounting for preservation biases and dating uncertainties.
Accordingly, three main sets of parameters were simultaneously estimated: (1)
the preservation rates quantifying the expected number of fossil occurrences per
sampled lineage per time unit (1 Myr); (2) the origination and extinction times for
each genus, which probably extend beyond the observed temporal range between
first and last appearances; and (3) the origination and extinction rates (expected
number of origination and extinction events per lineage per Myr) and their
temporal variation42. We estimated origination and extinction times assuming a
time-variable Poisson preservation model, and used them to infer origination and
extinction rates within epochs using a time-variable birth–death model where
the rates are estimated as independent parameters in each predefined time frame
(Supplementary Tables1–3). To reduce the risk of over-parameterization, we used
half-Cauchy priors on the origination and extinction rates with scale parameters
estimated from the data using hyper-priors75. We ran 2,000,000 Markov chain
Monte Carlo iterations under this model and summarized the posterior extinction
rates in box plots for each epoch, except for the Holocene, as the temporal and
taxonomic resolution of our data was insufficient to reliably estimate extinction in
such a short time frame. We considered extinction rates as significantly different
between subsequent epochs when 0 fell outside the 95% credible interval of their
difference, based on all posterior samples. We ran these analyses on the full dataset
of all megafauna groups first, and then repeated them for each group, namely
marine mammals, sea birds, sea turtles, and sharks and rays. We ran additional
analyses to assess more precisely the timing of origination and extinction rate
changes, using birth–death models in which the times of shift were not fixed, but
estimated as time-continuous parameters42. We tested models with a different
number of rate shifts and combined the results from each model using Bayesian
model averaging; that is, after resampling their posterior samples proportionally to
the respective relative probabilities76. We then summarized the marginal extinction
rates through time within 0.1 Myr time bins as mean and 95% credible intervals.
More details can be found in theSupplementary Methods.
Plio−Pleistocene dataset. We selected all marine megafauna genera occurring
in the Pliocene (5.3–2.6 Ma) and Pleistocene (2.6–0.01)77. In total, we gathered
1,763 global occurrences. Most of the data were not dated to the ‘stage’ level.
Accordingly, we used geological epochs as our interval unit; for example, whenever
a genus was reported in the Pliocene, the Zanclean or the Piacenzian, it was
assigned to the Pliocene. Since the Plio− Pleistocene is our interval of interest,
we performed a second evaluation process for this subset of data in which the
taxonomic assignments and age of each record were assessed following a procedure
described previously22 (Supplementary Methods). We followed the most recent
age for the Pliocene− Pleistocene boundary, at 2.58 Ma (ref. 77). Accordingly, all
Gelasian records were treated as Pleistocene occurrences. More details can be
found in theSupplementary Methods. It is worth noting that even though it has
been proposed that marine vertebrates from the Pleistocene are poorly known
(for example, ref. 15), we were able to gather 906 occurrences from the Pleistocene,
which is comparable with the 857 records gathered from the Pliocene.
Proportional extinction. In total, we compiled data for 215 Plio− Pleistocene
genera. Of these, 177 occurred in the Pliocene and 151 in the Pleistocene
(Supplementary Table4). Sixty-one genera occurred only in the Pliocene and 37
only in the Pleistocene. Based on these numbers, we calculated the proportion of
genera that were extirpated from the Pliocene, and the proportion that originated
in the Pleistocene. We did this for all megafauna, and for each individual group.
Finally, we calculated the net loss of genera as the percentage of Pliocene genera
that were lost passing into the Pleistocene.
Functional traits and functional taxonomic units. We assigned five ordered
categorical functional traits to the Pliocene and Pleistocene marine megafauna:
guild (most frequent diet in adults); body size (maximum total length); vertical
position (most frequent vertical position when they fed); habitat (typical zone
where they occurred); and thermoregulation capability (whether they were
endotherms, mesotherms or poikilotherms). Traits are inferred properties of
individual organisms known to directly influence their ecological role32,45. More
details on how traits were coded can be found in theSupplementary Methods.
Our Plio− Pleistocene occurrences dataset had a generic taxonomic resolution
(see above), which facilitated the assignment of traits given that most genera have
modern analogues on which we can base our assessments. Traits were assigned
using authoritative taxon-specific texts, online databases and expert assessments
based on both extant relatives and fossil records (references are provided in
theSupplementary Methods). Whenever we found a genus consisting of multiple
known species with different trait values, we treated them independently to
assign traits (see specific cases in theSupplementary Methods). As a result, our
traits were assigned mostly (95%) to genera, and sub-divisions of certain genera
according to shared traits. These functional taxonomic units are, in our opinion,
the lowest taxonomic resolution for systematically assigning functional traits to
fossil marine vertebrates. Given that functional taxonomic units correspond to
genera in 95% of cases, we still refer to them as ‘genera’ in the main text and figures
for consistency and simplicity. Traits were assigned to each occurrence of each
functional taxonomic unit. Whenever there was not enough information to assign
traits, we disregarded such an occurrence in our analyses (see specific cases in
theSupplementary Methods, which represent around 12% of the total number of
occurrences gathered from the Paleobiology Database).
Functional entities, redundancy, over-redundancy and vulnerability. For our
trait-based analyses we focused on genera that are associated with coastal habitats
(that is, strictly coastal, coastal–terrestrial and coastal–oceanic genera). These
genera represent 86% of the megafauna (Supplementary Table4) and had 1,569
global occurrences in our dataset. Based on the trait assignments, we calculated
the number of possible unique trait combinations, or FEs32,45. Pliocene and
Pleistocene marine megafauna fill 8% of the total number of FEs (that is, 53 out
of 648 FEs). Genera were assigned to FEs independent of taxonomy. Based on the
number of FEs and their corresponding genera, we then calculated the functional
redundancy (FR: the mean genera per FE), over-redundancy (FOR:the percentage
of genera that filled FEs above the mean level of functional redundancy, that is, the
overrepresentation of some FEs) and vulnerably (FV:the percentage of FEs with
onlyone genus45, representing a potential decrease in functional diversity following
taxonomic loss).
Functional space. We used the methods of Mouillot et al.32 to create the functional
space based on the FEs calculated above. We used the R package FD78 to create the
distance matrix (using the function ‘gowdis’) and to retrieve axes of the principal
coordinate analysis (using the function ‘dbFD’). Using the ‘quality_funct_space’
R function79, we determined that our data are best represented using four
dimensions, or principal coordinate analysis axes (Supplementary Fig.9). We then
used the ‘FDChange’ function to calculate the functional richness (the percentage
of the total volume occupied in the functional space) and shift (non-overlap of
functional volume)32. Given the multidimensional nature of the functional diversity
analyses, the four axes used to represent the functional space of Plio− Pleistocene
marine megafauna are correlated with multiple trait combinations (Supplementary
Fig.10) and, therefore, it is not possible to associate portions of the functional
space to single traits, nor to pinpoint changing segments of the space.
Resampling simulations. We tested the effects of sample size in the calculation of
functional diversity indices by randomly resampling each community (Pliocene
and Pleistocene) without replacement, bootstrapping the data 1,000 times over
20 evenly spaced bins from 10 to 711 (711 being the lowest sample size for
coastal habitats found in the Pliocene) using the R function ‘sample’. We found
variation in functional indices due to sample size. We standardized the Pleistocene
communities to 711 occurrences and recalculated the functional diversity
indices based on this resampled community running 1,000 permutations (with
replacement)80. Finally, we tested for significant differences between the Pliocene
and the Pleistocene using a Wilcoxon test.
Comparative simulations. We investigated whether the changes in functional
diversity during the Pliocene were significantly higher than those expected under
background extinction rates among genera associated with coastal habitats. To
do so, we calculated the mean extinction rate for the Cenozoic (except for the
Pliocene), as described in the ‘Extinction rates’ section above. Then, we estimated
the number of genera that would have been lost under background rates using
Foote’s boundary-crossing method81 as in equation (1).
Δ=− ∕+ ∕ln NNNtER (( )) (1)
bt bt b
where Nbt is the genera that crossed the bottom and the top of the interval (that
is, they were sampled in both the Pliocene and the Pleistocene); Nb is the genera
that only crossed the bottom of the inter val (that is, they were sampled only in the
Pliocene); and Δ t is the length of the interval in millions of years (that is, 2.75 Myr
following the timescale of Gradstein et al.)73. We then solved for Nb in equation (1),
replacing the extinction rate (ER) with 0.05 (the mean extinction rate during the
Cenozoic). Based on this, 22 genera would have been lost in the Pliocene under
mean background conditions (whereas 55 were actually lost in coastal habitats).
Accordingly, running 1,000 permutations, we simulated a Pleistocene subset in
our Plio− Pleistocene dataset by randomly removing 22 Pliocene genera and
calculating the delta (Pliocene functional richness – Pleistocene functional
richness) and the functional shift (see 'Functional space’ section of the Methods).
NATURE ECOLOGY & EVOLUTION | www.nature.com/natecolevol
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved. © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Articles NaTuRe eCology & evoluTioN
We then compared the observed delta and shift with the distributions of these
metrics under background conditions. Additionally, we investigated whether the
loss of functional diversity was greater than expected given the number of taxa
lost. This was achieved by randomly removing 55 Pliocene genera (the number of
coastal genera lost) and calculating the delta and shift. We tested the significance of
these comparisons by running 1,000 permutations. Although the above simulations
did not account for the modest sample size differences between the Pliocene and
Pleistocene, the post-extinction (Pleistocene) assemblage had more samples and
thus the tests are conservative.
Environmental reconstructions. We calculated the global extent of the neritic
areas available during the Pliocene and Pleistocene, given the sea-level changes of
these time periods. We regarded as the neritic area the section of the ocean with a
water depth of a maximum of 200 m. To estimate the global extent of these areas,
we used the current land topography and ocean bathymetry ETOPO1 Global
Relief Model82. We selected the ocean cells within the neritic zone and quantified
their area using the function ‘area’ from the R package raster83, which takes into
account the latitudinal decrease of the projected map cells as a consequence of
the globe curvature. To calculate changes in the global extent of the neritic areas
across time, we used two independent measures of sea-level change during the past
5.3 Myr (refs 4,5), applying a temporal resolution of 100,000 years; the choice of
measure had no bearing on the qualitative patterns of sea-level change. The model
based on deBoer et al.5 is presented in the main text. Global sea-level changes
were calculated using oxygen isotope variation4 and an ice-sheet model forced by
benthic δ
18O (ref. 5). Based on these data, we assessed the temporal changes in the
global extent of neritic areas available, and in temperature, in the Pliocene (n = 27)
versus Pleistocene (n = 26), calculated the coefficient of variation for each epoch,
and tested for significant differences in the mean values using a t-test. Both time
bins (Pliocene and Pleistocene) have similar temporal extents (2.7 and 2.6 Myr,
respectively), allowing direct comparisons.
Generalized linear models. We evaluated the effects of traits on extinction
probabilities by modelling survivorship (status: extinct or not extinct) in response
to genus traits. We initially used a generalized linear model with binomial error
and a logit link to simultaneously assess the effects of all traits (that is, glm (status
≈ trait1 + trait2… )). Furthermore, we used a metric of pseudo R2 (1 – (residual
deviance/null deviance)) to assess its explanatory power. This model was then
re-run, first by adding taxonomic identity (for instance, ‘class’: Mammalia, Aves,
and so on) as a fixed effect (for instance, glm (status ≈ trait1 + trait2… + class))
to account for its influence on extinction probabilities; and second, by adding
class as a random effect using a generalized linear mixed model in the R package
lme4 (ref. 84) (for instance, glmer (status ≈ trait1 + trait2… (1|class)) to control for
the potential non-independence of the extinction probabilities of species within
each class. Furthermore, in case the explanatory power of a trait was contingent
on the inclusion or exclusion of other traits in the model, we used a hierarchical
partitioning approach to run all possible single and multiple traits as additive
extinction predictors and to partition the proportional independent effects of each
trait using the R package hier.part85. Finally, we modelled extinction probability
as a function of thermoregulation (the most explanatory trait) and elucidated
differences among categories using Tukey tests.
Code availability. The code used to infer the origination, extinction and
preservation rates is available at https://github.com/dsilvestro/PyRate.
Data availability. The authors declare that all data supporting the findings of this
study are available within the paper and itsSupplementary Information files.
Received: 8 December 2016; Accepted: 23 May 2017;
Published: xx xx xxxx
References
1. Brook, B. W., Sodhi, N. S. & Bradshaw, C. J. A. Synergies among extinction
drivers under global change. Trends Ecol. Evol. 23, 453–460 (2008).
2. McCauley, D. J. et al. Marine defaunation: animal loss in the global ocean.
Science 347, 1255641 (2015).
3. Harnik, P. G. et al. Extinctions in ancient and modern seas. Trends Ecol. Evol.
27, 608–617 (2012).
4. Miller, K. G. et al. e phanerozoic record of global sea-level change. Science
310, 1293–1298 (2005).
5. De Boer, B., van de Wal, R. S. W., Bintanja, R., Lourens, L. J. & Tuenter, E.
Cenozoic global ice-volume and temperature simulations with 1-D ice-sheet
models forced by benthic delta O-18 records. Ann. Glaciol. 51, 23–33 (2010).
6. Zachos, J., Pagani, M., Sloan, L., omas, E. & Billups, K. Trends, rhythms, and
aberrations in global climate 65 Ma to present. Science 292, 686–693 (2001).
7. Lisiecki, L. E. & Raymo, M. E. Plio–Pleistocene climate evolution: trends and
transitions in glacial cycle dynamics. Quat. Sci. Rev 26, 56–69 (2007).
8. Van Woesik, R. et al. Hosts of the plio-pleistocene past reect modern-day
coral vulnerability. Proc. R. Soc. Lond. B Biol. Sci. 279, 2448–2456 (2012).
9. Valentine, J. W. & Jablonski, D. Biotic eects of sea-level change: the
pleistocene test. J. Geophys. Res. Solid Earth 96, 6873–6878 (1991).
10. Lewison, R. L., Crowder, L. B., Read, A. J. & Freeman, S. A. Understanding
impacts of sheries bycatch on marine megafauna. Trends Ecol. Evol. 19,
598–604 (2004).
11. McClain, C. R. et al. Sizing ocean giants: patterns of intraspecic size
variation in marine megafauna. PeerJ 3, e715 (2015).
12. Marx, F. G. & Uhen, M. D. Climate, critters, and cetaceans: Cenozoic drivers
of the evolution of modern whales. Science 327, 993–996 (2010).
13. Pyenson, N. D. & Sponberg, S. N. Reconstructing body size in extinct crown
Cetacea (Neoceti) using allometry, phylogenetic methods and tests from the
fossil record. J. Mamm. Evol. 18, 269–288 (2011).
14. Uhen, M. D. & Pyenson, N. D. Diversity estimates, biases, and historiographic
eects: resolving cetacean diversity in the tertiary. Palaeontol. Electronica 10,
1–22 (2007).
15. Boessenecker, R. W. Pleistocene survival of an archaic dwarf baleen whale
(Mysticeti: Cetotheriidae). Naturwissenschaen 100, 365–71 (2013).
16. Ando, T. & Fordyce, R. E. Evolutionary drivers for ightless, wing-propelled
divers in the Northern and Southern hemispheres. Palaeogeogr.
Palaeoclimatol. Palaeoecol. 400, 50–61 (2014).
17. Olson, S. L. An early pliocene marine avifauna from Duinefontein Cape
Province South Africa. Ann. S. Afr. Mus. 95, 147–164 (1985).
18. Sorbi, S., Domning, D. P., Vaiani, S. C. & Bianucci, G. Metaxytherium
subapenninum (Bruno, 1839) (Mammalia, Dugongidae), the latest sirenian of
the mediterranean basin. J. Vert. Paleontol. 32, 686–707 (2012).
19. Velez-Juarbe, J., Domning, D. P. & Pyenson, N. D. Iterative evolution of
sympatric seacow (Dugongidae, Sirenia) assemblages during the past similar
to 26 million years. PloS ONE 7, e31294 (2012).
20. Domning, D. P. Sirenians, seagrasses, and Cenozoic ecological change in the
Caribbean. Palaeogeogr. Palaeoclimatol. Palaeoecol. 166, 27–50 (2001).
21. Pimiento, C. & Balk, M. A. Body-size trends of the extinct giant shark
Carcharocles megalodon: a deep-time perspective on marine apex predators.
Paleobiology 41, 479–490 (2015).
22. Pimiento, C. & Clements, C. F. When did Carcharocles megalodon become
extinct? A new analysis of the fossil record. PloS ONE 9, e111086 (2014).
23. Dodd, C. K. & Morgan, G. S. Fossil sea-turtles from the early Pliocene Bone
Valley Formation, Central Florida. J. Herpetol. 26, 1–8 (1992).
24. Leonard-Pingel, J. S., Jackson, J. B. C. & O’Dea, A. Changes in bivalve
functional and assemblage ecology in response to environmental change in
the Caribbean Neogene. Paleobiology 38, 509–524 (2012).
25. Estes, J. A. et al. Trophic downgrading of planet Earth. Science 333,
301–306 (2011).
26. Doughty, C. E. et al. Global nutrient transport in a world of giants. Proc. Natl
Acad. Sci. USA 113, 868–873 (2016).
27. Malhi, Y. et al. Megafauna and ecosystem function from the Pleistocene to
the Anthropocene. Proc. Natl Acad. Sci. USA 113, 838–846 (2016).
28. Boessenecker, R. W. A new marine vertebrate assemblage from the Late
Neogene Purisima Formation in Central California, part II: pinnipeds and
cetaceans. Geodiversitas 35, 815–940 (2012).
29. Villeger, S., Novack-Gottshall, P. M. & Mouillot, D. e multidimensionality
of the niche reveals functional diversity changes in benthic marine biotas
across geological time. Ecol. Lett. 14, 561–568 (2011).
30. Raup, D. M. & Sepkoski, J. J. Jr Mass extinctions in the marine fossil record.
Science 215, 1501–1503 (1982).
31. Bambach, R. K., Bush, A. M. & Erwin, D. H. Autecology and the lling of
ecospace: key metazoan radiations. Palaeontology 50, 1–22 (2007).
32. Mouillot, D., Graham, N. A. J., Villeger, S., Mason, N. W. H. & Bellwood, D. R.
A functional approach reveals community responses to disturbances. Trends
Ecol. Evol. 28, 167–177 (2013).
33. Aberhan, M. & Kiessling, W. Persistent ecological shis in marine molluscan
assemblages across the end-Cretaceous mass extinction. Proc. Natl Acad. Sci.
USA 112, 7207–7212 (2015).
34. Foster, W. J. & Twitchett, R. J. Functional diversity of marine ecosystems aer
the Late Permian mass extinction event. Nat. Geosci. 7, 233–238 (2014).
35. Dineen, A. A., Fraiser, M. L. & Sheehan, P. M. Quantifying functional
diversity in pre- and post-extinction paleocommunities: a test of ecological
restructuring aer the end-Permian mass extinction. Earth Sci. Rev. 136,
339–349 (2014).
36. Duy, J. E., Lefcheck, J. S., Stuart-Smith, R. D., Navarrete, S. A. & Edgar, G. J.
Biodiversity enhances reef sh biomass and resistance to climate change.
Proc. Natl Acad. Sci. USA 113, 6230–6235 (2016).
37. Dick, D. G. & Maxwell, E. E. e evolution and extinction of the ichthyosaurs
from the perspective of quantitative ecospace modelling. Biol. Lett. 11,
20150339 (2015).
38. Lawton, J. H. What do species do in ecosystems. Oikos 71, 367–374 (1994).
39. Altieri, A. H. et al. Tropical dead zones and mass mortalities on coral reefs.
Proc. Natl Acad. Sci. USA 114, 3660–3665 (2017).
40. Hughes, T. P. et al. Global warming and recurrent mass bleaching of corals.
Nature 543, 373–377 (2017).
NATURE ECOLOGY & EVOLUTION | www.nature.com/natecolevol
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved. © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Articles
NaTuRe eCology & evoluTioN
41. Wernberg, T. et al. Climate-driven regime shi of a temperate marine
ecosystem. Science 353, 169–172 (2016).
42. Silvestro, D., Schnitzler, J., Liow, L. H., Antonelli, A. & Salamin, N. Bayesian
estimation of speciation and extinction from incomplete fossil occurrence
data. Syst. Biol. 63, 349–367 (2014).
43. Villafaña, J. A. & Rivadeneira, M. M. Rise and fall in diversity of Neogene
marine vertebrates on the temperate Pacic coast of South America.
Paleobiology 40, 659–674 (2014).
44. Finnegan, S. et al. Paleontological baselines for evaluating extinction risk in
the modern oceans. Science 348, 567–570 (2015).
45. Mouillot, D. et al. Functional over-redundancy and high functional
vulnerability in global sh faunas on tropical reefs. Proc. Natl Acad. Sci. USA
111, 13757–13762 (2014).
46. Coux, C., Rader, R., Bartomeus, I. & M., T. J. Linking species functional roles
to their network roles. Ecol. Lett. 19, 762–770 (2016).
47. Lefcheck, J. S. & Duy, J. E. Multitrophic functional diversity predicts
ecosystem functioning in experimental assemblages of estuarine consumers.
Ecology 96, 2973–2983 (2015).
48. Dehling, D. M., Jordano, P., Schaefer, H. M., Boehning-Gaese, K. &
Schleuning, M. Morphology predicts species’ functional roles and their
degree of specialization in plant–frugivore interactions. Proc. R. Soc. Lond. B
Biol. Sci. 283, 20152444 (2016).
49. Cadotte, M. W., Carscadden, K. & Mirotchnick, N. Beyond species: functional
diversity and the maintenance of ecological processes and services. J. Appl.
Ecol. 48, 1079–1087 (2011).
50. Kozlowski, J. & Gawelczyk, A. T. Why are species’ body size distributions
usually skewed to the right? Funct. Ecol. 16, 419–432 (2002).
51. Duy, J. E. Biodiversity loss, trophic skew and ecosystem functioning. Ecol.
Lett. 6, 680–687 (2003).
52. Duy, J. E. Biodiversity and ecosystem function: the consumer connection.
Oikos 99, 201–219 (2002).
53. Holland, S. M. Sea level change and the area of shallow-marine habitat:
implications for marine biodiversity. Paleobiology 38, 205–217 (2012).
54. Jackson, J. B. C., Jung, P., Coates, A. G. & Collins, L. S. Diversity and
extinction of tropical American mollusks and emergence of the isthmus of
Panama. Science 260, 1624–1626 (1993).
55. Budd, A. F., Johnson, K. G. & Stemann, T. A. in Evolution and Environment
in Tropical America (eds Jackson, B. C., Budd, A. F. & Coates, A. G.) 168–204
(Univ. Chicago Press, 1996).
56. Allmon, W. D., Emslie, S. D., Jones, D. S. & Morgan, G. S. Late Neogene
oceanographic change along Florida’s west coast: evidence and mechanisms.
J. Geol. 104, 143–162 (1996).
57. Klaus, J. S. et al. Rise and fall of Pliocene free-living corals in the Caribbean.
Geology 39, 375–378 (2011).
58. Todd, J. A. et al. e ecology of extinction: molluscan feeding and faunal
turnover in the Caribbean Neogene. Proc. R. Soc. Lond. B Biol. Sci. 269,
571–577 (2002).
59. O’Dea, A. et al. Environmental change preceded Caribbean extinction by 2
million years. Proc. Natl Acad. Sci. USA 104, 5501–5506 (2007).
60. Smith, J. T. & Jackson, J. B. C. Ecology of extreme faunal turnover of tropical
American scallops. Paleobiology 35, 77–93 (2009).
61. Valenzuela-Toro, A. M., Gutstein, C. S., Varas-Malca, R. M., Suarez, M. E. &
Pyenson, N. D. Pinniped turnover in the South Pacic Ocean: new evidence
from the Plio− Pleistocene of the Atacama Desert, Chile. J. Vert. Paleontol. 33,
216–223 (2013).
62. Metcalf, J. L. et al. Synergistic roles of climate warming and human
occupation in Patagonian megafaunal extinctions during the last deglaciation.
Sci. Adv. 2, e1501682 (2016).
63. Pimiento, C. et al. Geographical distribution patterns of Carcharocles
megalodon over time reveal clues about extinction mechanisms. J. Biogeogr.
43, 1645–1655 (2016).
64. McNab, B. K. Resources and energetics determined dinosaur maximal size.
Proc. Natl Acad. Sci. USA 106, 12184–12188 (2009).
65. Grady, J. M., Enquist, B. J., Dettweiler-Robinson, E., Wright, N. A. &
Smith, F. A. Evidence for mesothermy in dinosaurs. Science 344,
1268–1272 (2014).
66. Payne, J. L., Bush, A. M., Heim, N. A., Knope, M. L. & McCauley, D. J.
Ecological selectivity of the emerging mass extinction in the oceans. Science
14, aaf2416 (2016).
67. Boyles, J. G., Seebacher, F., Smit, B. & McKechnie, A. E. Adaptive
thermoregulation in endotherms may alter responses to climate change.
Integr. Comp. Biol. 51, 676–690 (2011).
68. McNab, B. K. Energetics, body size, and the limits to endothermy. J. Zool.
199, 1–29 (1983).
69. Wright, D. H. Species-energy theory: an extension of species-area theory.
Oikos 1, 496–506 (1983).
70. Pyenson, N. D. & Lindberg, D. R. What happened to grey whales
during the Pleistocene? e ecological impact of sea level change on
benthic feeding areas in the North Pacic Ocean. PloS ONE 6,
e21295 (2011).
71. Davidson, A. D. et al. Drivers and hotspots of extinction risk in marine
mammals. Proc. Natl Acad. Sci. USA 109, 3395–3400 (2012).
72. Wilmers, C. C., Estes, J. A., Edwards, M., Laidre, K. L. & Konar, B. Do trophic
cascades aect the storage and ux of atmospheric carbon? An analysis of sea
otters and kelp forests. Front Ecol. Environ. 10, 409–415 (2012).
73. Gradstein, F. M., Ogg, G. & Schmitz, M. e GeologicTime Scale
2012(Elsevier, 2012).
74. Silvestro, D., Salamin, N. & Schnitzler, J. PyRate: a new program to estimate
speciation and extinction rates from incomplete fossil data. Methods Ecol.
Evol. 5, 1126–1131 (2014).
75. Silvestro, D., Cascales-Miñana, B., Bacon, C. D. & Antonelli, A. Revisiting the
origin and diversication of vascular plants through a comprehensive
Bayesian analysis of the fossil record. New Phytol. 207, 425–436 (2015).
76. Hoeting, J. A., Madigan, D., Raery, A. E. & Volinsky, C. T. Bayesian model
averaging: a tutorial. Stat. Sci. 14, 382–417 (1999).
77. Gibbard, P. L., Head, M. J. & Walker, M. J. Formal ratication of the
Quaternary system/period and the Pleistocene series/epoch with a base at
2.58 Ma. J. Quat. Sci 25, 96–102 (2010).
78. Laliberté, E. & Shipley, B. FD: measuring functional diversity from multiple
traits, and other tools for functional ecology. R package version 1.0-11 (2011).
79. Maire, E., Grenouillet, G., Brosse, S. & Villeger, S. How many dimensions are
needed to accurately assess functional diversity? A pragmatic approach
for assessing the quality of functional spaces. Glob. Ecol. Biogeogr. 24,
728–740 (2015).
80. Kowalewski, M. I. & Novack-Gottshall, P. H. in Quantitative Methods in
Paleobiology(eds Alroy, J. & Hunt, G.) 19–54 (e Paleontological
Society, 2010).
81. Foote, M. Origination and extinction components of taxonomic diversity:
general problems. Paleobiology 26, 74–102 (2000).
82. Amante, C. & Eakins, B. W. 2009 ETOPO1 1 Arc-Minute Global Relief Model:
Procedures, Data Sources and Analysis(NOAA Technical Memorandum
NESDIS NGDC-24, National Geophysical Data Center,accessed June 2017).
83. Hijmans, R. J. et al. Package ‘raster’. R package version 2.5-8 (2015).
84. Bates, D. & Martin, M. Fitting linear mixed-eects models using lme4. J. Stat.
Sow. 67, 1–48 (2015).
85. Walsh, C., Mac Nally, R. & Walsh, M. C. e hier. part package. R package
version 1.4-0 (2003).
Acknowledgements
We thank M. Sánchez-Villagra for his support during the development of this
research, S. Villegér, A. Antonelli, F. Leprieur, J. Lefcheck and L. Gamfeldt for their
valuable suggestions, C. Ricotta and K. Boersma for their assistance with the use
of R functions, B. Mcnab and M. Balk for their insights on thermoregulation, and
J. Velez-Juarbe for his support assigning traits to marine mammals. We are grateful
for the constructive comments provided by P. Novack-Gottshall, which significantly
improved this work. PyRate analyses were run at the high-performance computing
centre Vital-IT of the Swiss Institute of Bioinformatics (Lausanne, Switzerland).
C.P. was supported by a Forschungskredit postdoctoral fellowship from the University
of Zurich (FK-15-105), J.N.G. was supported by a European Union Marie Curie Career
Integration Grant (FP7 MC CIG 61893), D.S. was funded by the Swedish Research
Council (2015-04748) and S.V. was first supported by the Universidad de Alcalá
postdoctoral programme, and then by the Alexander von Humboldt Foundation
and the Federal Ministry for Education and Research (Germany). This is the
Paleobiology Database publication number 284.
Author contributions
C.P., J.N.G. and C.J. designed the research, C.P., J.N.G. and M.D.U. performed the
research, C.P., J.N.G., C.F.C., D.S., S.V. and M.D.U. analysed the data, C.P. and J.N.G.
wrote the paper, and C.F.C., S.V., D.S., M.D.U. and C.J. improved the final manuscript.
Competing interests
The authors declare no competing financial interests.
Additional information
Supplementary information is available for this paper at doi:10.1038/s41559-017-0223-6.
Reprints and permissions information is available at www.nature.com/reprints.
Correspondence and requests for materials should be addressed to C.P.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
NATURE ECOLOGY & EVOLUTION | www.nature.com/natecolevol