Trait similarity in reef fish faunas across the

world’s oceans
Matthew McLeana,1, Rick D. Stuart-Smithb, Sébastien Villégerc, Arnaud Auberd, Graham J. Edgarb,
M. Aaron MacNeila,e, Nicolas Loiseauc, Fabien Leprieurc,f, and David Mouillotc,f
a
Department of Biology, Dalhousie University, Halifax, NS, Canada, B3H 4R2; bInstitute for Marine and Antarctic Studies, University of Tasmania, Hobart,
TAS 7001, Australia; cMARBEC, Univ Montpellier, CNRS, Institut Français de Recherche pour l’Exploitation de la Mer, IRD, 34095 Montpellier, France; dInstitut
Français de Recherche pour l’Exploitation de la Mer, Unité Halieutique de Manche et mer du Nord, 62321 Boulogne-sur-Mer, France; eOcean Frontier
Institute, Dalhousie University, Halifax, NS, Canada, B3H 4R2; and fInstitut Universitaire de France, 75005 Paris, France

Edited by Nils Chr. Stenseth, University of Oslo, Oslo, Norway, and approved February 10, 2021 (received for review June 15, 2020)

Species’ traits, rather than taxonomic identities, determine com- spanning eight marine realms in both temperate and tropical
munity assembly and ecosystem functioning, yet biogeographic oceans, we examined global biogeographic patterns in reef fish
Downloaded from https://www.pnas.org by UNIVERSIDADE FEDERAL DO ESTADO DO RIO DE JANEIRO on June 19, 2025 from IP address 200.130.54.39.

patterns have been far less studied for traits. While both environ- trait composition and evaluated whether trait composition is
mental conditions and evolutionary history shape trait biogeogra- shaped primarily by the environment, taxonomic relatedness, or
phy, their relative contributions are largely unknown for most evolutionary history. Specifically, we examined patterns of spe-
organisms. Here, we explore the global biogeography of reef fish cies distribution in trait space relative to environmental condi-
traits for 2,786 species from 89 ecoregions spanning eight marine tions and taxonomic and phylogenetic composition.
realms with contrasting environmental conditions and evolution- We explicitly considered two spatial scales—marine realms
ary histories. Across realms, we found a common structure in the and marine ecoregions—for which we examined distinct ques-
distribution of species traits despite a 10-fold gradient in species tions. Marine realms are large areas of ocean basins for which
richness, with a defined “backbone” of 21 trait combinations “biotas are internally coherent at higher taxonomic levels, as a
shared by all realms globally, both temperate and tropical. Across result of a shared and unique evolutionary history” (18) and are

ECOLOGY
ecoregions, assemblages under similar environmental conditions
shaped by major differences in environmental regimes and his-
had similar trait compositions despite hosting drastically different
torical or broadscale isolation. Given their distinct biotas, envi-
species pools from separate evolutionary lineages. Thus, despite
ronmental regimes, species richness, and evolutionary histories,
being separated by thousands of kilometers and millions of years
we began by examining how realms differed in trait composition
of evolution, similar environments host similar trait compositions
in reef fish assemblages worldwide. Our findings suggest that
and redundancy. For instance, we examined whether temperate
similar trait-based management strategies can be applied among
and tropical oceans contained similar trait diversity and trait
regions with distinct species pools, potentially improving conser- proportions and whether certain trait combinations existed in all
vation outcomes across diverse jurisdictions. realms. We next examined ecoregion-scale patterns to assess re-
gional variation in trait composition and redundancy and whether
|
biogeography community assembly | functional ecology | trait composition varied predictably within realms. Ecoregions are
|
macroecology phylogenetics smaller areas of “relatively homogeneous species composition,” (18)
which are shaped by distinct oceanographic and environmental
conditions. We hypothesized that regional species assemblages
B iogeographic patterns reflect how past and current environ-
mental conditions along with evolutionary history have shaped
biodiversity, ecosystem functioning, and the ecosystem services Significance
human societies depend on. While biogeography has historically
focused on species composition (1), species traits (morphological, Biogeography has focused extensively on species identities, yet
physiological, or behavioral features of organisms) are increasingly global patterns in species traits (morphological, physiological,
recognized as the main drivers of community assembly and eco- or behavioral features) are not well known, including whether
system functioning (2–5). Trait composition has been shown to they are shaped by modern environmental conditions or by
mediate species’ interactions (6), shape ecological niches (7), de- shared evolutionary history. Our global analysis of nearly 3,000
termine species’ responses to environmental fluctuations (8), and reef fish species found a consistent variety of traits across ocean
govern species’ influences on key ecological processes like nutrient basins worldwide, including a backbone of 21 trait combinations
common to all oceans. At the regional scale, we found that fish
cycling (9, 10) and biomass production (11).
assemblages in similar environments had similar trait composi-
Large-scale variation in trait composition is shaped by biotic
tions despite being separated by up to 100 degrees of latitude
(e.g., trophic interactions) and abiotic constraints (e.g., environmental
and hosting different species with distinct evolutionary histories.
filtering) under the influence of evolutionary and biogeographic
Thus, environmental conditions have likely shaped global pat-
processes (i.e., speciation, extinction, and immigration), yet global
terns in reef fish traits regardless of geography, species identity,
patterns and processes in trait biogeography are poorly known for
or evolutionary history.
most organisms. In marine ecosystems, how environmental con-
straints and evolutionary history have shaped species diversity has Author contributions: M.M., R.D.S.-S., S.V., A.A., G.J.E., M.A.M., N.L., F.L., and D.M. de-
been broadly debated (12, 13), and recent studies suggest that both signed research; M.M. performed research and analyzed data; and M.M., R.D.S.-S., S.V.,
adaptation to regional climate (14) and geographical expansion of A.A., G.J.E., M.A.M., N.L., F.L., and D.M. wrote the paper.

clades (15) have influenced global patterns in marine biodiversity. The authors declare no competing interest.
On shallow rocky and coral reefs, large-scale environmental gradi- This article is a PNAS Direct Submission.
ents and evolutionary history have led to ocean basins with vastly Published under the PNAS license.
different species richness and composition (13, 16), yet the distri- 1
To whom correspondence may be addressed. Email: [email protected].
bution and drivers of associated trait composition are still unclear. This article contains supporting information online at https://www.pnas.org/lookup/suppl/
Here, using data collected through the Reef Life Survey doi:10.1073/pnas.2012318118/-/DCSupplemental.
(RLS) (17) (http://www.reeflifesurvey.com/) from 89 ecoregions Published March 15, 2021.

PNAS 2021 Vol. 118 No. 12 e2012318118 https://doi.org/10.1073/pnas.2012318118 | 1 of 10

 would show similar trait composition under similar environ- has limited the number of viable functional entities that can persist
mental conditions worldwide regardless of species composition in all ocean basins.
or shared evolutionary history.
Latitudinal Similarity in Reef Fish Traits. To evaluate similarity in
Results species assemblage trait composition at the regional scale, we
A Global “Backbone” of Reef Fish Traits. To quantify the distribu- examined trait space patterns across 89 ecoregions globally. We
tion of reef fish traits, we first created a multidimensional trait generated probability densities of PCoA 1 and PCoA 2 scores
space of 2,786 reef fish species (using principal coordinates (i.e., species distributions on the first and second axes of trait
analysis, PCoA) according to five categorical traits. Trait variation space) for regional species pools and calculated the centroid
among species was primarily explained by water-column position position (i.e., mean value along each axis, SI Appendix, Fig. S4)
(benthic, demersal, site-attached pelagic, or mobile pelagic), for each ecoregion, as well as species richness and the number of
gregariousness (solitary, paring, or schooling), and active period functional entities. Across ecoregions, we found that PCoA 1
(diurnal or nocturnal). Along the first axis, there was a clear centroids had a clear unimodal relationship with latitude, being
gradient from benthic to pelagic species, and solitary and noc- similar in temperate regions in both hemispheres and converging
turnal species were distinguished from schooling and diurnal toward the equator (Fig. 2), highlighting similar trait composi-
species (Fig. 1 A and C). Species also varied largely in diet, with tions at similar latitudes worldwide. Linear models of PCoA 1
Downloaded from https://www.pnas.org by UNIVERSIDADE FEDERAL DO ESTADO DO RIO DE JANEIRO on June 19, 2025 from IP address 200.130.54.39.

invertivores being distinct from planktivores and piscivores along centroids using the absolute value of latitude revealed that in-
the first axis. Along the second axis, species primarily varied in cluding hemisphere as a predictor added no explanatory power
body size, with a clear gradient from small to large species. (latitude: R2 = 0.327; latitude + hemisphere: R2 = 0.332), further
Across five tropical and three temperate marine realms, we demonstrating latitudinal similarity regardless of hemisphere.
found a common pattern in the distribution of species in trait Reef fish assemblages in temperate ecoregions were distinguished
space despite a 10-fold gradient in species richness (Fig. 1D). In from those in tropical ecoregions along the first axis of trait space
all realms, species clustered in the same area of trait space and (Figs. 1 and 2), with temperate ecoregions more characterized by
occupied similar ranges of the overall space (Fig. 1D). The pro- benthic, solitary, nocturnal, and invertivore species and tropical
portions of different trait categories were also strikingly similar ecoregions more characterized by demersal, pairing, diurnal, and
across realms (SI Appendix, Fig. S1). Thus, reef fish trait compo- grazer species (SI Appendix, Fig. S5 and Table S2). PCoA 2 cen-
sition appears remarkably consistent worldwide when considered troids also showed a unimodal relationship with latitude, indi-
at the scale of marine realms, despite major differences in species cating that body sizes were larger at higher latitudes and that
number and identity. Across realms, species richness ranged from temperate regions were more characterized by piscivores (Fig. 2C
6 to 54% of the observed global pool, whereas the number of and SI Appendix, Fig. S5).
functional entities (i.e., unique trait combinations) ranged from 24 We further found that both species richness and the number of
to 77%, indicating that species-poor realms maintain dispropor- functional entities were lowest at temperate latitudes and in-
tionately high trait diversity (SI Appendix, Table S1). Mouillot creased toward the equator; however, species richness increased
et al. (19) previously found that the breadth of reef fish traits was 2.5 times faster than the number of functional entities (2.8 spe-
consistent across the tropics, with a ninefold gradient in species
cies versus 1.1 functional entities per degree latitude; SI Ap-
richness corresponding to only a threefold gradient in the number
pendix, Fig. S6). Thus, as species richness increased, an average
of functional entities. Here, we extend those patterns to temperate
of 2.5 species were packed into existing functional entities before
oceans, showing astonishing similarity between reef fish faunas
a new functional entity emerged. Together, these results indicate
worldwide despite drastic differences in species richness, envi-
that as species richness increases in the tropics, species are packed
ronmental regimes, and habitat types (i.e., coral versus rocky
into existing functional entities, increasing redundancy rather than
substrates). Together, these findings show that key ecological
functions are retained, even in species-poor realms, through high extending the breadth of trait space (19, 21).
differentiation and limited redundancy that capture the overall To determine if trait similarity resulted from shared species
breadth of reef fish traits (19). composition (taxonomy) or evolutionary history (phylogeny), we
We next identified functional entities shared by all marine examined dissimilarity patterns between realms and between
realms, both temperate and tropical (Fig. 1B). A total of 21 entities ecoregions. We defined trait dissimilarity as the Euclidean dis-
(of 356 in the global species pool) define a common “backbone” tance between species-pool centroids in the first four dimensions
(20) in trait space filled by reef fishes from all ocean basins. These of trait space and defined taxonomic and phylogenetic dissimi-
21 core entities encompass a broad range of trait values, in- larity as the turnover component of pairwise Jaccard dissimilarity
cluding a variety of body sizes, behaviors, and diets (SI Ap- (22). While both taxonomic and phylogenetic dissimilarity are
pendix, Fig. S2). In particular, the most common backbone naturally bounded between 0 and 1, trait dissimilarity (distance
traits were invertivorous diet, demersal water-column position, between ecoregion centroids) has no intrinsic upper limit. We
diurnal active period, solitary behavior, and small to medium therefore used a null model approach to define the maximum
body size, most frequently represented by families such as expected trait dissimilarity between realms under random assem-
wrasses, sea basses, groupers, damselfishes, and gobies. We also bly (see Materials and Methods). For consistency, this approach
identified specific backbones for temperate and tropical realms, was also used for both taxonomic and phylogenetic dissimilarity.
which revealed consistent ecological structuring within tem- We found that marine realms were highly similar in trait com-
perate (42 entities) and tropical (45 entities) oceans (Fig. 1B position but highly dissimilar in taxonomic and phylogenetic
and SI Appendix, Fig. S2). However, null models revealed that composition, suggesting that common trait distributions are not a
the numbers of shared functional entities within the global and result of shared species or evolutionary histories (Fig. 3). We also
tropical backbones were lower than expected by random (SI found that when ecoregions became farther apart in latitude, they
Appendix, Fig. S3). This suggests that similar trait distributions became more dissimilar in taxonomic and phylogenetic composi-
across realms were not due to identical functional entities but tion. However, trait composition was not related to latitudinal
rather similar trait proportions. Hence, similar overall trait com- distance since both geographically near and far ecoregions had
position can arise from a variety of functional entities, depending similar trait compositions. Thus, ecoregions separated by up to 100
on their relative frequencies, particularly when functional entities degrees of latitude displayed equivalent trait compositions despite
share multiple traits in common. Furthermore, this result indicates hosting radically different species pools from separate evolution-
that intense filtering by environmental and ecological constraints ary lineages (Fig. 4 and SI Appendix, Fig. S7).

2 of 10 | PNAS McLean et al.

https://doi.org/10.1073/pnas.2012318118 Trait similarity in reef fish faunas across the world’s oceans
 A Global Species Pool B
Global 'Backbone' Temperate 'Backbone' Tropical 'Backbone'
PCoA2 (15% of Variation)
0.5
0
−0.5

Nb. Realms = 8 Nb. Realms = 3 Nb. Realms = 5

Nb. Sp = 2786 Nb. FE = 356

−0.5 0 0.5
PCoA1 (24% of Variation)
C Diet Body Size Water Column Gregariousness Active Period
Downloaded from https://www.pnas.org by UNIVERSIDADE FEDERAL DO ESTADO DO RIO DE JANEIRO on June 19, 2025 from IP address 200.130.54.39.

PI 80+ PM SC DI
IN 50−80 PS PA NO
CO 30−50 DE SO
PL
15−30 BE
OM
MI 7−15
GR 0−7

Central Indo-Pacific Temperate Australasia Eastern Indo-Pacific Western Indo-Pacific

D

ECOLOGY
Nb. Sp = 1498 Nb. FE = 273 Nb. Sp = 935 Nb. FE = 213 Nb. Sp = 593 Nb. FE = 147 Nb. Sp = 548 Nb. FE = 147

Tropical Eastern Pacific Tropical Atlantic Temperate North Pacific Temperate North Atlantic

Nb. Sp = 244 Nb. FE = 106 Nb. Sp = 227 Nb. FE = 98 Nb. Sp = 190 Nb. FE = 84 Nb. Sp = 154 Nb. FE = 89

Fig. 1. Trait space for the global pool of 2,786 reef fish species and for each of the eight marine realms. (A) The first two axes of trait space for the global
species pool showing 50, 75, and 95% contours of species density, along with marginal distributions of PCoA 1 and PCoA 2 scores. (B) The common
“backbone” of 21 functional entities shared by all realms globally, as well as the temperate and tropical backbones. (C) Convex hulls showing the distribution
of trait categories within the trait space for each of the five categorical traits. Diet: PI, piscivore; IN, invertivore; CO, corallivore; PL, planktivore; OM, omnivore;
MI, microphage; and GR, grazer. Water column: PM, pelagic mobile; PS, pelagic site-attached; DE, demersal; and BE, benthic. Gregariousness: SC, schooling;
PA, pairing; and SO, solitary. Active period: DI, diurnal; and NO, nocturnal. (D) The first two axes of trait space for each of the eight marine realms showing 50,
75, and 95% contours of species density, along with marginal distributions of PCoA 1 and PCoA 2 scores. The total number of observed species and functional
entities for the global species pool and for each marine realm are indicated in the bottom left and bottom right corners of each plot; realms are ordered by
descending number of species. While all species are plotted, species belonging to the same functional entity (i.e., having identical trait values) are
superimposed.

Environment Shapes Trait Composition. Debate over the relative range, nitrate, phosphate, and net primary productivity (see Ma-
influences of contemporary environmental conditions and shared terials and Methods). We found that environmental dissimilarity
evolutionary history has focused extensively on species composi- was the primary driver of trait dissimilarity (effect size = 0.60 ±
tion rather than trait composition (13, 16). Therefore, to deter- 0.01 and R2 = 0.56), while taxonomic (effect size = 0.09 ± 0.02 and
mine if trait similarity was more strongly driven by environmental R2 = 0.20) and phylogenetic dissimilarity (effect size = 0.11 ± 0.02
conditions than by taxonomic or phylogenetic composition, we and R2 = 0.24) added little explanatory power (all variables to-
evaluated the relationship between trait dissimilarity and taxo- gether R2 = 0.58; Fig. 4).
nomic, phylogenetic, and environmental dissimilarity using multi- Trait similarity among geographically distant assemblages
ple regression of distance matrices (MRM) (23). We defined inhabiting similar environmental conditions could arise through
environmental dissimilarity as the Euclidean distance between both environmental filtering—in which regional trait composition is
ecoregions according to mean sea-surface temperature (SST), SST filtered from the realm pool—and evolutionary convergence—in

McLean et al. PNAS | 3 of 10

Trait similarity in reef fish faunas across the world’s oceans https://doi.org/10.1073/pnas.2012318118
 A Ecoregions in Trait Space B PCoA 1 Centroids C PCoA 2 Centroids

0.15
−0.2 −0.1 0.0 0.1 0.2 0.3

−0.1 0.0 0.1 0.2 0.3 0.4

Northern Hemisphere

Absolute Value of Latitude (°)

Southern Hemisphere 50
PCoA 2 Centroids

PCoA 1 Centroids

PCoA 2 Centroids
Galapagos
40 Islands Cocos
Island

0.05
30

20

−0.05
10

−0.1 0.0 0.1 0.2 0.3 0.4 −60 −40 −20 0 20 40 60 −60 −40 −20 0 20 40 60
PCoA 1 Centroids Latitude (°) Latitude (°)
D
>0.20
80
Downloaded from https://www.pnas.org by UNIVERSIDADE FEDERAL DO ESTADO DO RIO DE JANEIRO on June 19, 2025 from IP address 200.130.54.39.

60

0.10
40

PCoA 1 Centroids
0.05
Latitude (°)
20
0

0.00
−20

−0.05
−40
−60

−0.10
−150 −100 −50 0 50 100 150
Longitude (°)

Fig. 2. Latitudinal patterns in trait space centroids for reef fish assemblages from 89 global ecoregions. (A) The centroid position of each ecoregion in trait
space colored by the absolute value of latitude per ecoregion. (B) PCoA 1 centroid of each ecoregion plotted against latitude. (C) PCoA 2 centroid of each
ecoregion plotted against latitude. The Galapagos Islands and Cocos Island are indicated as outliers. In B and C, red lines indicate the best fits from second-
order polynomial regressions, which are shown for visualization. (D) Global map of the 89 ecoregions colored by PCoA 1 centroid values.

which species with distinct trait combinations independently evolve Fig. S9), whereas the relationship between trait dissimilarity and
in response to similar environmental pressures (24, 25). Both pro- both taxonomic and phylogenetic dissimilarity did not differ from
cesses have likely contributed to contemporary patterns in trait neutral expectation (observed taxonomic dissimilarity: R2 = 0.06,
composition yet disentangling their relative effects is challenging mean of R2 null distribution = 0.08, 95% = 0.13; observed phylo-
and contentious (25). Therefore, to assess the likelihood of con- genetic dissimilarity: R2 = 0.08, mean of R2 null distribution = 0.12,
vergent evolution in shaping spatial trait similarity among assem- 95% = 0.21; SI Appendix, Fig. S9).
blages, we tested whether the observed relationship between trait
dissimilarity and taxonomic, phylogenetic, and environmental dis- Discussion
similarity was different from expected under neutral evolution by Our results show that similar trait compositions have arisen
simulating trait dissimilarity according to a Brownian model of among reef fish faunas worldwide in response to contemporary
evolution. Following Mazel et al. (25), we simulated species’ posi- environmental conditions rather than shared evolutionary history
tions on the first axis of trait space, recalculated trait dissimilarity (14, 26–28). Across marine realms, we found a common pattern
between ecoregions, recalculated the relationship between trait in the distribution of reef fish species in trait space, indicating a
dissimilarity and taxonomic, phylogenetic, and environmental dis- wide range of shared functional roles despite major differences
similarity, and compared null expectations to the observed rela- in species richness and composition, evolutionary history, and
tionship. We also calculated the magnitude of difference in environmental regimes (19). For instance, the distribution of
observed versus expected trait dissimilarity between ecoregions species in trait space was highly similar between the Central Indo-
[i.e., standardized effect size (SES), termed “phylogenetically stan- Pacific and Temperate Northern Pacific despite fundamentally
dardized trait distance” in Mazel et al. (25)]. We found a positive, different habitats (i.e., coral versus rocky substrates) and a near
albeit weak, relationship (slope = 0.036; SI Appendix, Fig. S8) be- 10-fold difference in species richness. These results suggest that
tween SES and environmental dissimilarity, with 9% of ecoregion shallow reefs have been shaped by consistent environmental and
pairs sharing more similar trait compositions than expected under ecological pressures, hosting a similar range of trait combinations
neutral evolution. More importantly, we found that the observed at different levels of redundancy, with a common backbone of
relationship between trait dissimilarity and environmental dissimi- functional entities worldwide. This is a key discovery highlighting
larity was stronger (R2 = 0.45) than expected under neutral evolu- that universal ecological roles exist on shallow reefs globally and
tion (mean of R2 null distribution = 0.21, 95% = 0.43; SI Appendix, have been filled by a diversity of species and evolutionary lineages.

4 of 10 | PNAS McLean et al.

https://doi.org/10.1073/pnas.2012318118 Trait similarity in reef fish faunas across the world’s oceans
 A Trait Dissimilarity Taxonomic Dissimilarity Phylogenetic Dissimilarity
CIP Color Key CIP Color Key CIP Color Key
Similar Dissimilar Similar Dissimilar Similar Dissimilar
TAA 0 0.25 TAA 0 1 TAA 0 0.8
EIP EIP EIP
WIP WIP WIP
TEP TEP TEP
TAT TAT TAT
TNP TNP TNP
TNA TNA TNA

TAA

TEP

TNP
WIP

TAA
TAA

TAT

TEP

TNP
TEP

TNP

WIP
WIP

CIP

EIP

TNA

TAT
TAT

CIP

EIP

TNA
CIP

EIP

TNA

B
Central Indo-Pacific Temperate Australasia Eastern Indo-Pacific Western Indo-Pacific
Downloaded from https://www.pnas.org by UNIVERSIDADE FEDERAL DO ESTADO DO RIO DE JANEIRO on June 19, 2025 from IP address 200.130.54.39.

ECOLOGY
Tropical Eastern Pacific Tropical Atlantic Temperate Northern Pacific Temperate Northern Atlantic

Fig. 3. Heatmaps of trait, taxonomic, and phylogenetic dissimilarity across marine realms, along with phylogenetic trees for species pools from each realm.
(A) Heatmaps of trait, taxonomic, and phylogenetic dissimilarity across realms; blue indicates high similarity whereas red indicates high dissimilarity. CIP,
Central Indo-Pacific; TAA, Temperate Australasia; EIP, Eastern Indo-Pacific; WIP, Western Indo-Pacific; TEP, Tropical Eastern Pacific; TAT, Tropical Atlantic; TNP,
Temperate Northern Pacific; and TNA, Temperate Northern Atlantic. Trait dissimilarity was calculated as the Euclidean distance between realm centroids in
the first four dimensions of trait space. Taxonomic and phylogenetic dissimilarity were calculated as the turnover component of pairwise Jaccard dissimilarity.
(B) Phylogenetic trees for each of the eight marine realms; species present in each realm are colored, whereas absent species are gray.

Several hypotheses have attempted to explain global patterns result from evolutionary convergence, environmental filtering, or
in reef fish diversity, with debate over the importance of evolu- a combination of both. Only 9% of ecoregion pairs showed greater
tionary history versus contemporary environmental variation. trait similarity than expected by neutral evolution (SI Appendix,
For instance, the center-of-origin hypothesis suggests that the Fig. S8), providing little evidence for evolutionary convergence in
greater diversity of reef fishes in the Indo-Australian Archipel- shaping trait similarity among distantly related reef fish assem-
ago arises from higher rates of speciation (ref. 16, but see ref. 29), blages. However, the observed relationship between trait dissimi-
while the “climate stability” and “time for speciation” hypotheses larity and environmental dissimilarity was stronger than expected
suggest that higher diversity results from greater environmental under neutral evolution, providing evidence for environmental
stability during the Quaternary (30) and older colonization events, filtering (SI Appendix, Fig. S9). Similarly, Mazel et al. (25) found
providing more time for in situ diversification (31). Other hy- that trait similarity between most mammal assemblages did not
potheses suggest that reef fish diversity is driven by contempo- differ from neutral expectation, concluding that evolutionary
rary gradients in productivity and energy availability (32). While convergence may be an important process shaping trait similarity
existing evidence supports a mixture of evolutionary and envi- in certain regions only (e.g., Australia). Overall, our results suggest
ronmental mechanisms in shaping reef fish species diversity (13), that trait composition arises independently of taxonomic diversity
there is little knowledge on how these patterns and processes and is primarily driven by contemporary environmental filtering that
extend to trait composition. operates similarly across oceans and evolutionary lineages (33).
Across ecoregions, we found that reef fish assemblages dis- This study not only reveals the importance of the environment
played similar trait compositions in similar environments, with in shaping trait assembly in reef fishes but also implies that trait
little influence of shared evolutionary history. This pattern could composition could be substantially modified by future environmental

McLean et al. PNAS | 5 of 10

Trait similarity in reef fish faunas across the world’s oceans https://doi.org/10.1073/pnas.2012318118
 0.30

0.30

0.30
A Effect Size = 0.09 B Effect Size = 0.11 C Effect Size = 0.60
100

Latitudinal Distance (°)

Trait Dissimilarity

Trait Dissimilarity
Trait Dissimilarity
80
0.20

0.20

0.20
60

40
0.10

0.10

0.10
20
0.00

0.00

0.00
0.3 0.5 0.7 0.9 0.2 0.4 0.6 0.8 0 2 4 6 8
Taxonomic Dissimilarity Phylogenetic Dissimilarity Environmental Dissimilarity
Fig. 4. Trait dissimilarity in relation to taxonomic, phylogenetic, and environmental dissimilarity between ecoregions. (A–C) Environmental dissimilarity had
Downloaded from https://www.pnas.org by UNIVERSIDADE FEDERAL DO ESTADO DO RIO DE JANEIRO on June 19, 2025 from IP address 200.130.54.39.

the strongest influence on trait dissimilarity between ecoregions; although trait dissimilarity increased with taxonomic and phylogenetic dissimilarity, many
latitudinally distant regions had similar trait compositions despite highly dissimilar taxonomic and phylogenetic compositions. Points in all panels are pairwise
comparisons between ecoregions and are colored by latitudinal distance between ecoregions. Lines are best fits from simple linear regressions.

changes. Environmentally driven trait similarity supports a “periodic trait compositions even when species pools have identical trait spaces. For
table of niches” (27), which hypothesizes that predictable community instance, substantial variation exists in species evenness between temperate
configurations should arise in response to similar environmental and tropical realms, which has strong implications for trait distribution
among individuals (39). All indices and metrics throughout the manuscript
conditions through common responses to environmental filtering.
were therefore calculated using presence/absence data; however, sensitivity
Therefore, the predictability of reef fish trait composition could be tests were performed using species abundances (see Sensitivity of Results to
key for anticipating future changes in ecosystem functioning and Occurrence versus Abundance Data).
ecosystem services under climate change and human development
(34, 35). Our results also imply that trait-based resource man- Traits, Trait Space, and Common Trait Backbones. We compiled five categorical
agement strategies can be applied among regions with distinct traits for each of the 2,786 species in the data set, which were chosen to
species pools (36). Thus, trait-based approaches could improve describe life history, trophic ecology, habitat preference, and behavior. While
conservation outcomes across diverse jurisdictions by identifying we recognize that these traits only represent a small portion of the true
trait compositions associated with high-functioning reefs and fa- multidimensional species niche, they are recognized proxies of fish functions
vorable ecosystem states (37). Given the link between environ- (5) and have been used to examine reef fish biogeographic patterns (19, 38,
39) and impacts on ecosystem functioning (40). Diet characterized species’
mental conditions and reef fish trait composition, future climate
main food items and was coded using seven trophic categories: grazer, mi-
change will profoundly alter the ecological roles sustained by reef crophage (feeding on cyanobacteria and microorganisms), planktivore,
fishes, threatening the current livelihoods of millions of people omnivore, corallivore, invertivore, and piscivore. Microphage was chosen in
(38). Uncovering links between reef fish traits and ecosystem lieu of “scraping herbivore” or “excavator” based on recent reviews by
resilience will be key to maintaining functional reefs in an era Clements et al. (41) and Bellwood et al. (42). Water-column position was
dominated by human impacts. coded using four ordered categories: benthic (sedentary on the bottom),
demersal (swimming near the bottom), pelagic site attached (swimming off
Materials and Methods the bottom within a reef), and pelagic mobile (swimming off the bottom
Fish Survey Methods. Fish data were collected through the RLS, a standardized between reefs). Active period was coded as either diurnal (active during the
monitoring program that surveys fish communities on coral and rocky reefs day) or nocturnal (active during the night). Gregariousness was coded using
worldwide (www.reeflifesurvey.com). RLS was chosen for this study because three ordered categories: solitary, pairing (or sometimes forming small
1) all surveys follow a standardized protocol, and patterns between differ- schools), and schooling. Body size was coded using six ordered categories:
ent biogeographic regions are readily comparable; 2) RLS is conducted 0 to 7 cm, 7.1 to 15 cm, 15.1 to 30 cm, 30.1 to 50 cm, 50.1 to 80 cm, and
uniquely on shallow reef habitats; and 3) RLS has extensive spatial coverage, >80 cm. Trait data came from Stuart-Smith et al. (39), the majority of which
including both temperate and tropical regions. RLS is comprised of teams of were derived from FishBase and expert knowledge. The multidimensional
professional scientists and trained recreational divers who complete un- trait space was created by applying PCoA to a Gower similarity matrix of the
derwater visual censes of fish communities using 50 × 10 m belt transects. On species × traits table and was primarily examined by plotting the first two
each transect, divers record the species identity, abundance, and estimated Principal Coordinate axes (43), which cumulatively explained 39% of the
size of all fishes encountered within 5 m of either side of the transect. For total variance. Trait space was additionally examined using the third and
full details refer to Edgar and Stuart-Smith (17). For this study, fish data were fourth Principal Coordinates axes which together explained 25% of the total
compiled from 10,913 transects (conducted between 2006 and 2018), variance (SI Appendix, Figs. S10 and S11). Thus, the first four axes of trait
encompassing 2,786 taxa from 89 marine ecoregions spanning 10 marine space accounted for 64% of overall variation, and because a test of mean
realms (two realms were not included in realm-level analyses, see Traits, absolute deviation (44) identified four axes as the optimal number for de-
Trait Space, and Common Trait Backbones). Of the 2,786 observed taxa, scribing trait space, additional axes were not examined.
2,582 were recorded at species level, with the remaining recorded primarily While we analyzed trait biogeography by considering single trait values
at genus level (162 at genus level and 43 at family level); however, for for each species, it should be noted that trait–environment relationships
simplicity we refer to all taxa as species throughout the paper. Ecoregions operate at the scale of individuals, and substantial trait variation may exist
and realms were defined according to the Marine Ecoregions of the World within species (45). Traits may also be highly plastic and can evolve in re-
categorization (18), which was also used to classify ecoregions as temperate sponse to environmental or ecological pressures (46). In turn, this intraspe-
or tropical. For this study, we considered species occurrences and did not cific variability can influence ecosystem functioning (47). Here, we were
integrate abundance or biomass. Species occurrences were chosen to ex- unable to integrate intraspecific trait variability due to insufficient data;
amine trait composition and redundancy within different realms and bio- however, by using broad categorical traits (e.g., diet, water-column position,
geographic regions and to identify how these patterns are shaped by and size bin) rather than highly specific continuous traits (e.g., length at ma-
environmental conditions and evolutionary history. Species occurrences re- turity and fecundity), we capture the ecological profile of adult individuals.
veal changes in trait composition due to differences in the existence or re- To evaluate reef fish trait composition globally, we began by examining
dundancy (number of species) of functional entities (i.e., unique trait the distribution of species within trait space for three temperate and five
combinations), whereas species abundances can lead to drastically different tropical marine realms. Temperate South America and Temperate Southern

6 of 10 | PNAS McLean et al.

https://doi.org/10.1073/pnas.2012318118 Trait similarity in reef fish faunas across the world’s oceans
 Africa were not included in realm-level analyses due to low and patchy the distribution of sampling effort (number of transects per ecoregion) was
sampling effort, as few sites and ecoregions have been sampled within these highly skewed, with few ecoregions containing disproportionate numbers of
realms. For instance, only 25 species have been observed in Temperate transects (SI Appendix, Table S3). Transects were randomly sampled with
Southern Africa in comparison to 154 species in Temperate North Atlantic, replacement because some ecoregions had fewer than six transects; how-
the most species-poor realm included. Thus, we were not able to confidently ever, we tested the robustness of these choices (see below).
assess trait-space patterns for Temperate South America and Temperate To help visualize latitudinal similarity in ecoregion centroids, we used
Southern Africa. However, the individual ecoregions sampled within these second-order polynomial linear regressions. However, the purpose of these
realms were included in ecoregion-level analyses, which were separate from regression models was primarily visualization—providing a trend line to
realm-level analyses. We calculated 50, 75, and 95% density contours for highlight the unimodal latitudinal pattern in ecoregion centroids (and ad-
each realm based on the position and density of species in trait space using ditionally for a null model, see Null Model for the Influence of Species
the R package emdbook (48) function HPDregionplot. We then generated Richness). Yet, we do note that the relationship between PCoA 1 centroids
probability density plots of PCoA 1 and PCoA 2 scores for all species in each and latitude was substantially better fit by a second order polynomial re-
realm to further examine the distribution of species along each of the trait- gression (R2 = 0.46 and Akaike Information Criterion [AIC] = −254.8) than by
space axes. To examine which traits were most responsible for trait-space a simple linear regression (R2 = 0.007 and AIC = −203.4). To provide quan-
structure and to examine how species were distributed according to their titative evidence for latitudinal similarity in trait composition, we also used
trait combinations, we plotted convex hulls of each trait category within the simple linear regressions of PCoA 1 centroids in functional of the absolute
trait space. value of latitude and examined whether including hemisphere as a predictor
Functional entities refer to unique combinations of trait categories, that is, improved explanatory power.
Downloaded from https://www.pnas.org by UNIVERSIDADE FEDERAL DO ESTADO DO RIO DE JANEIRO on June 19, 2025 from IP address 200.130.54.39.

the number of functional entities for a given species pool is the total number To test the robustness of the number of transects and procedure used to
of combinations of the different trait categories present (19). Species in the calculate ecoregion centroids, we recalculated PCoA 1 centroids and reex-
same functional entity also occupy identical positions in trait space. Here, amined latitudinal similarity patterns by randomly selecting 1, 5, 10, 15, and
the 2,786 species and five traits considered led to 1,008 possible trait com- 20 transects per ecoregion instead of 6. This was done first with replacement
binations, of which 356 were realized in the global species pool. Common and no deletion, second with replacement but after deleting all ecoregions
trait “backbones” were calculated by identifying functional entities that with less than the selected number of transects, and third without re-
were shared by assemblages in 1) all realms, 2) temperate realms only, and 3) placement after deleting all ecoregions with less than the selected number of
tropical realms only. Temperate Australasia is considered a temperate realm transects. Latitudinal trait similarity was robust to all choices except deleting
but borders the Great Barrier Reef along the east coast of Australia and has ecoregions with fewer than 20 transects, because this removed over 40%
been extensively sampled by the RLS program, leading to high species (38 of 89) of ecoregions (SI Appendix, Fig. S12).
richness and similarities to Indo-Pacific realms. We therefore examined how

ECOLOGY
temperate and tropical backbones changed if we reclassified Temperate Magnitude of Variation in PCoA Centroids across Ecoregions. As visible in
Australasia as a tropical realm or if we removed it entirely. In both cases, Fig. 2A, distances between ecoregion centroids were much smaller than the
there was little change, as reclassifying Temperate Australasia as tropical extent of the overall trait space, implying that variation across ecoregions
reduced the tropical backbone from 45 to 43 functional entities and re- was limited. However, the extent of the trait space is defined by the maxi-
moving it from temperate realms expanded the temperate backbone from mum distance between the most dissimilar individual functional entities,
42 to 44 entities. which is an inappropriate context for evaluating variation among regional
species assemblages. To determine the expected maximum distance be-
Null Models. Several null models were used throughout this study to assess the tween ecoregion centroids, we randomly shuffled species composition
sensitivity of results and to identify whether results could have arisen because among ecoregions while maintaining initial species richness and species
of chance (random assembly) or from confounding patterns like latitudinal occurrence frequency and calculated the maximum distance between ecor-
gradients in species richness. These null models all use the approach of egions in the first four dimensions of trait space. We repeated this process
randomizing species composition across assemblages (realms or ecoregions) 999 times to build a null distribution of expected maximum distances be-
while maintaining observed species richness and species occurrence fre- tween ecoregion centroids. The maximum null value was 0.60, substantially
quency, using the “independent swap” algorithm (49) in the R package lower than the maximum distance between any two functional entities
picante (50). We chose to randomize species composition among assem- (1.24). The actual observed value of maximum distance between ecoregion
blages rather than randomize species-trait values to avoid creating species centroids (0.550) was greater than 95% of the null distribution (95 null =
with ecologically unfeasible trait combinations. 0.549), demonstrating that ecoregions displayed substantial variation in trait
composition that was greater than expected under random assembly (SI
Null Models for Trait Backbones. To determine whether the number of shared Appendix, Fig. S13).
functional entities comprising global, temperate, and tropical backbones was
different from expected by chance, we randomly shuffled species composi- Contribution of Traits to Latitudinal Patterns in PCoA Centroids. To examine
tion between realms while maintaining initial species richness and species the contribution of each trait to latitudinal patterns in PCoA 1 centroids, and
occurrence frequency and calculated the number of shared functional en- to examine the distinction between temperate and tropical ecoregions, we
tities. We repeated this process 999 times to build null distributions of shared calculated average trait values per ecoregion as the proportions of different
functional entities and compared the actual observed numbers of functional trait categories present in each species pool using the R package FD (51)
entities to the null distributions. Both the global and tropical backbones function functcomp. We then calculated Pearson correlations between PCoA
contained fewer shared functional entities (global = 21 and tropical = 45) centroid values and trait values for each ecoregion, and we used multiple
than 95% of the null distribution (global null 5% = 26 and tropical null 5% = factor analysis (MFA) to examine associations between traits and ecoregions.
50), while the temperate backbone contained fewer entities than 90% of MFA is a modified PCoA that can account for categorical data and therefore
the null distribution but not fewer than 95% (temperate = 42 and tem- give equal weighting to traits with different numbers of categories, for
perate null 5% = 40; SI Appendix, Fig. S3). example, diet, which has seven categories, is not overweighted. Associations
between traits and temperate versus tropical ecoregions were assessed by
PCoA Centroids, Species Diversity, and Functional Entities across Ecoregions. the contributions of trait values to the first and second axes of the MFA
PCoA centroids refer to the central positions of species pools within trait analysis (SI Appendix, Fig. S5). MFA was conducted using the R package
space and reflect the mean trait values of the observed species. PCoA cen- FactoMineR (52) function MFA.
troids were therefore calculated for each ecoregion as the mean positions of
observed species pools on each of the PCoA axes (SI Appendix, Fig. S4). To Trait Dissimilarity between Marine Realms. In this study, trait dissimilarity was
calculate PCoA centroids and functional entities for each ecoregion while calculated as the Euclidean distance between species-pool centroids in the
accounting for uneven sampling effort across regions, we randomly selected first four dimensions of trait space, which is not bounded by an upper limit
six transects per ecoregion to generate species pools, calculated metrics for of 1. To define the expected upper limit of trait dissimilarity between marine
the resulting species pools, repeated this process 99 times with replacement, realms, we randomly shuffled the species composition of each realm while
and calculated mean values for the 99 repetitions. For PCoA centroids, in maintaining initial species richness and species occurrence frequency and
each of the 99 repetitions, the ecoregion centroid value was calculated as recalculated the Euclidean distance between realms. We repeated this pro-
the mean centroid of the six transects. We chose six transects because it was cess 999 times and calculated the maximum trait dissimilarity between
the most common number of transects performed within any ecoregion, realms across the 999 repetitions. The observed maximum value (0.26) was
that is, the mode. The mode was chosen in lieu of mean or median because then used to define the upper limit of trait dissimilarity between realms in

McLean et al. PNAS | 7 of 10

Trait similarity in reef fish faunas across the world’s oceans https://doi.org/10.1073/pnas.2012318118
 Fig. 3A. This procedure was chosen to provide a conservative estimate of species’ positions on the first axis only because PCoA axes are perfectly un-
maximum dissimilarity between realms, rather than defining maximum correlated by construction and simulating multiple PCoA axes under Brow-
dissimilarity between two species in the trait space (the theoretical maxi- nian motion would induce strong correlations, leading to bias in the neutral
mum distance between assemblages that would each have only one of these data (see ref. 25).
two extreme species). For consistency, this procedure was also applied to
taxonomic and phylogenetic dissimilarity, which resulted in a maximum Sensitivity of Results to the Choice of Traits. We tested whether latitudinal
taxonomic dissimilarity of 1 and a maximum phylogenetic dissimilarity similarity in PCoA 1 centroids was robust to the choice and number of traits by
of 0.91. deleting each trait, one at a time, recalculating PCoA 1 centroids for each
ecoregion, and reexamining latitudinal patterns. Latitudinal similarity was
Influences of Taxonomic, Phylogenetic, and Environmental Dissimilarity. To again robust to all choices (and in some cases became more pronounced),
evaluate taxonomic and phylogenetic dissimilarity, we use the additive indicating that the results were not dictated by any single trait but rather the
partitioning of pairwise Jaccard dissimilarity proposed by Baselga (22) for overall composition of traits (SI Appendix, Fig. S15).
taxonomy and adapted by Leprieur et al. (53) for phylogeny, which de-
composes beta diversity into nestedness and turnover components. Because Sensitivity of Results to Extreme PCoA Centroid Values. We tested whether
we were interested in knowing whether similarity in trait composition latitudinal similarity in PCoA 1 centroids as well as the relationship between
resulted from similarity in species composition or evolutionary lineages, we trait dissimilarity and environmental dissimilarity were robust to removing
focused on taxonomic and phylogenetic turnover—dissimilarity that occurs extreme centroid values—highly positive PCoA 1 centroids at high latitudes
when species or lineages present in one region are absent in another region (i.e., ecoregions with the highest leverage). We sequentially removed each
Downloaded from https://www.pnas.org by UNIVERSIDADE FEDERAL DO ESTADO DO RIO DE JANEIRO on June 19, 2025 from IP address 200.130.54.39.

but are replaced by other species or lineages not found in the first region of the top 15 ecoregions with the highest PCoA 1 centroid values and
independently of the difference in species richness. As with PCoA centroids, reexamined latitudinal patterns for the remaining PCoA 1 centroids, as well
we randomly selected six transects per ecoregion to generate species pools, the relationship between trait dissimilarity and environmental dissimilarity.
calculated taxonomic and phylogenetic turnover for the resulting species However, only 13 ecoregions were removed for assessing the relationship
pools, repeated this process 99 times, and calculated mean values for the 99 between trait dissimilarity and environmental dissimilarity, because Malvi-
repetitions. Phylogenetic trees were derived from the R package fishtree nas/Falklands and Channels/Fjords of Southern Chile were not included in
(54), and taxonomic and phylogenetic turnover were calculated using the R the original MRM analysis (see Influences of Taxonomic, Phylogenetic, and
package betapart (55). For this analysis, two ecoregions with only five spe- Environmental Dissimilarity). Latitudinal similarity in PCoA 1 centroids
cies present in the phylogeny (Malvinas/Falklands and Channels/Fjords of (i.e., unimodal relationship between centroids and latitude) was again ro-
Southern Chile) were removed. Environmental dissimilarity between ecor- bust to removing ecoregions with the most extreme values; however, the
egions was calculated according to SST mean and range, nitrate, phosphate, unimodal pattern became less pronounced as ecoregions were removed (SI
and net primary productivity. Following extensive discussion among coau- Appendix, Fig. S16). The strong linear relationship between trait dissimilarity
thors, these environmental variables were chosen based on existing
and environmental dissimilarity was highly robust and remained clear even
knowledge of reef fish functional responses to environmental gradients (56).
with 13 ecoregions removed. (SI Appendix, Fig. S17).
However, we tested the sensitivity of results to the choice of environmental
variables, including adding pH and salinity (see Sensitivity of Results to the
Sensitivity of Results to the Choice of Environmental Variables. We tested
Choice of Environmental Variables). Data for mean SST and SST range were
whether the results of the MRM analysis were robust to the choice of envi-
obtained from the National Oceanic and Atmospheric Administration Coral
ronmental variables used to calculate environmental dissimilarity. We removed
Reef Watch (57). Data for nitrate and phosphate were obtained from the
each environmental variable, one at a time, recalculated environmental dis-
Bio-ORACLE environmental data set (58). Net primary productivity data were
similarity, and reran the MRM analysis. We also tested the effect of adding pH
derived from the standard Vertically Generalized Production Model (59). We
and salinity, both individually and together. Salinity data were obtained from
calculated mean values of environmental variables per ecoregion and
the MULTIOBS_GLO_PHY_REP_015_002 database (62) available through Coper-
standardized them (using the R function scale) before calculating Euclidean
nicus Marine Environment Monitoring Service (https://marine.copernicus.eu/
distances between all pairs of ecoregions. As for realms, we calculated trait
services-portfolio/access-to-products/?option=com_csw&view=details&product_
dissimilarity between ecoregions as the Euclidean distance between ecor-
id=MULTIOBS_GLO_PHY_REP_015_002), and pH data were derived from the
egion centroids in the first four dimensions of trait space. We then used
Norwegian Earth System Model forced ocean simulation (63). Again, the re-
MRM [R package ecodist (60) function MRM (23)] to quantify the relative
influences of taxonomic, phylogenetic, and environmental dissimilarity on sults were robust to all combinations of environmental variables, with mean
trait dissimilarity. Because trait dissimilarity and environmental dissimilarity effect sizes (across all models) of 0.60 (±0.03) for environmental dissimi-
were not bounded between 0 and 1, all distance matrices were normalized larity, 0.13 (±0.04) for phylogenetic dissimilarity, and 0.09 (±0.04) for
between 0 and 1 prior to analysis. Because the MRM function does not allow taxonomic dissimilarity (SI Appendix, Table S4).
direct model assessment, we assessed model quality and assumptions (e.g.,
residual normality) by performing a standard multiple linear regression with Sensitivity of Results to Family-Level Taxonomic Dissimilarity. Because not all
distance matrices as vectors, which provides equivalent results to the MRM taxa included in this study were identified to species level, and because
function (SI Appendix, Fig. S14). different species in the same family may have similar traits, we tested
whether results of the MRM analysis were robust to calculating taxonomic
Brownian Motion Null Models. To compare observed results to those expected dissimilarity at the family level rather than species level. Thus, we recalculated
under neutral evolution, we simulated trait dissimilarity using Brownian taxonomic dissimilarity between ecoregions and reran the MRM analysis
motion. A Brownian motion model was first fit to the observed phylogeny modeling trait dissimilarity in function of family-level taxonomic dissimilarity,
and the corresponding sigma value was used for the simulations. Following phylogenetic dissimilarity, and environmental dissimilarity. Again, results
Mazel et al. (25), we simulated species’ positions on the first axis of the trait were robust with effect sizes of 0.54 for environmental dissimilarity, 0.11 for
space (PCoA 1) using the R package geiger (61). We repeated this process phylogenetic dissimilarity, and 0.09 for taxonomic dissimilarity.
999 times and, in each repetition, we recalculated trait dissimilarity (distance
between ecoregion centroids) and reexamined the relationship between Null Model for the Influence of Species Richness. We tested whether lat-
trait dissimilarity and taxonomic, phylogenetic, and environmental dissimi- itudinal similarity in PCoA centroids could have resulted from latitudinal
larity to develop null distributions of R2 values between variables. We then patterns in species richness using two null models where we randomly
compared observed R2 values, also calculated using trait dissimilarity on the shuffled species composition among ecoregions while maintaining initial
first PCoA axis only, to the null distributions to determine whether the ob- species richness and species occurrence frequency. We therefore shuffled
served relationship was stronger than neutral expectations. We also calcu- species composition, recalculated PCoA centroid values for each ecoregion,
lated the magnitude of difference between observed and expected trait and calculated the strength (R2 value) of the relationship between ran-
dissimilarity [i.e., SES, termed “phylogenetically standardized trait distance” domized PCoA 1 centroid values and latitude using 1) second-order poly-
in Mazel et al. (25)] as observed dissimilarity—mean expected dissimilarity/ nomial regression and 2) simple linear regression with the absolute value of
SD of expected dissimilarities—and examined the linear relationship be- latitude. We repeated this process 999 times to build null distributions of R2
tween SES and environmental dissimilarity. Ecoregion pairs were considered values and compared observed R2 values to the null distributions. R2 values
more similar or less similar than expected if observed dissimilarity was less from both observed models were greater than 95% of the values in the null
than or greater than 97.5% of simulated dissimilarity values (two tails used distributions, indicating that latitudinal trait similarity was not a result of
to allow for both “convergent” and “divergent” pairs). We simulated latitudinal patterns in species richness alone (SI Appendix, Fig. S18).

8 of 10 | PNAS McLean et al.

https://doi.org/10.1073/pnas.2012318118 Trait similarity in reef fish faunas across the world’s oceans
 Sensitivity of Results to Occurrence versus Abundance Data. In this study, we MRM analysis using functional beta diversity from Chao et al. (64) instead of
used species occurrences rather than species abundances (see Fish Survey distances between ecoregion centroids in trait space, where functional beta
Methods). To determine whether results were robust to this choice, we diversity is calculated as a generalized version of pairwise Rao’s quadratic
recalculated PCoA 1 centroids for all ecoregions as abundance-weighted entropy between assemblages. In agreement with the original MRM analysis,
centroids and tested the correlation between the original centroid values and individual regressions, environmental dissimilarity had the highest effect
and the abundance-weighted centroid values. We then reran the MRM size (0.31) and individual R2 value (0.28), while taxonomic and phylogenetic
analysis in which trait dissimilarity was calculated as the Euclidean distance dissimilarity had weaker effect sizes (taxonomic: −0.012; phylogenetic: −0.002)
between abundance-weighted ecoregion centroids in the first four dimen- and R2 values (taxonomic: 0.04; phylogenetic: 0.06) and added little explana-
sions of trait space, and taxonomic turnover was calculated with species tory power (R2 full model: 0.281).
abundances using Bray–Curtis dissimilarity. The Pearson’s correlation value
between the original (occurrence-based) PCoA 1 centroid values and the
Data Availability. RLS data are freely available and can be accessed at https://
abundance-weighted PCoA 1 centroid values was 0.97, demonstrating that
reeflifesurvey.com/. Reef fish trait data are freely available and can be
trait similarity was robust to both species’ occurrences and abundances. Ef-
accessed at http://www.fishbase.org. Reef fish phylogenies are freely avail-
fect sizes from the MRM analysis, and R2 values from individual regressions,
able through the R package fishtree (42). All environmental data came from
using species abundances were nearly identical to those of the original
freely available public sources (see Materials and Methods).
occurrence-based analysis (Original analysis: environmental dissimilarity:
effect size = 0.60 and R2 = 0.56; taxonomic dissimilarity: effect size = 0.09
and R2 = 0.20; phylogenetic dissimilarity: effect size = 0.11 and R2 = 0.24. ACKNOWLEDGMENTS. We thank all RLS divers worldwide whose data
collection efforts made this study possible. This research used the National
Downloaded from https://www.pnas.org by UNIVERSIDADE FEDERAL DO ESTADO DO RIO DE JANEIRO on June 19, 2025 from IP address 200.130.54.39.

Abundance-based analysis: environmental dissimilarity: effect size = 0.51

Collaborative Research Infrastructure Strategy-enabled Integrated Marine
and R2 = 0.56; taxonomic dissimilarity: effect size = 0.003 and R2 = 0.12;
Observing System infrastructure for database support and storage. This
phylogenetic dissimilarity: effect size = 0.15 and R2 = 0.28), again demon-
research was funded through the 2017–2018 Belmont Forum and BiodivERsA
strating that results were robust to both occurrence and abundance data. REEF-FUTURES project under the BiodivScen ERA-Net COFUND program
along with the French National Research Agency, the Natural Sciences and
Sensitivity of Results to the Trait Dissimilarity Metric. We examined whether Engineering Research Council (Grant No. RGPBB/525590), the Canada Re-
results were robust to how trait dissimilarity was quantified by rerunning the search Chairs Program, and the Ocean Frontier Institute.

1. C. Violle, P. B. Reich, S. W. Pacala, B. J. Enquist, J. Kattge, The emergence and promise 24. G. H. Orians, “Convergent evolution at the community level” in Coevolution,
of functional biogeography. Proc. Natl. Acad. Sci. U.S.A. 111, 13690–13696 (2014). D. J. Futuyama, M. Slatkin, Eds. (Sinauer Associates, 1983), pp. 431–458.

ECOLOGY
2. S. Díaz, M. Cabido, Vive la différence: Plant functional diversity matters to ecosystem 25. F. Mazel et al., The geography of ecological niche evolution in mammals. Curr. Biol.
processes. Trends Ecol. Evol. 16, 646–655 (2001). 27, 1369–1374 (2017).
3. F. de Bello et al., Towards an assessment of multiple ecosystem processes and services 26. L. M. Bower, K. O. Winemiller, Fish assemblage convergence along stream environ-
via functional traits. Biodivers. Conserv. 19, 2873–2893 (2010). mental gradients: An intercontinental analysis. Ecography 42, 1691–1702 (2019).
4. M. F. Rosenfield, S. C. Müller, Plant traits rather than species richness explain eco- 27. K. O. Winemiller, D. B. Fitzgerald, L. M. Bower, E. R. Pianka, Functional traits, con-
logical processes in subtropical forests. Ecosystems (N. Y.) 23, 52–66 (2020). vergent evolution, and periodic tables of niches. Ecol. Lett. 18, 737–751 (2015).
5. S. Villéger, S. Brosse, M. Mouchet, D. Mouillot, M. J. Vanni, Functional ecology of fish: 28. R. E. Drenovsky, A. Khasanova, J. J. James, Trait convergence and plasticity among
Current approaches and future challenges. Aquat. Sci. 79, 783–801 (2017). native and invasive species in resource-poor environments. Am. J. Bot. 99, 629–639
6. C. Fortunel, R. Valencia, S. J. Wright, N. C. Garwood, N. J. B. Kraft, Functional trait (2012).
differences influence neighbourhood interactions in a hyperdiverse Amazonian for- 29. D. L. Rabosky et al., An inverse latitudinal gradient in speciation rate for marine
est. Ecol. Lett. 19, 1062–1070 (2016). fishes. Nature 559, 392–395 (2018).
7. N. J. B. Kraft, R. Valencia, D. D. Ackerly, Functional traits and niche-based tree com- 30. L. Pellissier et al., Quaternary coral reef refugia preserved fish diversity. Science 344,
munity assembly in an Amazonian forest. Science 322, 580–582 (2008). 1016–1019 (2014).
8. M. McLean et al., Trait structure and redundancy determine sensitivity to disturbance 31. E. C. Miller, K. T. Hayashi, D. Song, J. J. Wiens, Explaining the ocean’s richest biodi-
in marine fish communities. Glob. Change Biol. 25, 3424–3437 (2019). versity hotspot and global patterns of fish diversity. Proc. Biol. Sci. 285, 20181314
9. J. E. Allgeier, C. A. Layman, P. J. Mumby, A. D. Rosemond, Consistent nutrient storage (2018).
and supply mediated by diverse fish communities in coral reef ecosystems. Glob. 32. D. J. Currie et al., Predictions and tests of climate‐based hypotheses of broad‐scale
Change Biol. 20, 2459–2472 (2014). variation in taxonomic richness. Ecol. Lett. 7, 1121–1134 (2004).
10. A. Norkko, A. Villnäs, J. Norkko, S. Valanko, C. Pilditch, Size matters: Implications of 33. D. Bellwood, P. Wainwright, C. Fulton, A. Hoey, Assembly rules and functional groups
the loss of large individuals for ecosystem function. Sci. Rep. 3, 2646 (2013). at global biogeographical scales. Funct. Ecol. 16, 557–562 (2002).
11. C. Roscher et al., Using plant functional traits to explain diversity-productivity rela- 34. M. McLean et al., Fish communities diverge in species but converge in traits over three
tionships. PLoS One 7, e36760 (2012). decades of warming. Glob. Change Biol. 25, 3972–3984 (2019).
12. D. P. Tittensor et al., Global patterns and predictors of marine biodiversity across taxa. 35. I. Trindade-Santos, F. Moyes, A. E. Magurran, Global change in the functional diversity
Nature 466, 1098–1101 (2010). of marine fisheries exploitation over the past 65 years. Proc. Biol. Sci. 287, 20200889
13. T. Gaboriau et al., Ecological constraints coupled with deep-time habitat dynamics (2020).
predict the latitudinal diversity gradient in reef fishes. Proc. Biol. Sci. 286, 20191506 36. L. A. K. Barnett, N. S. Jacobsen, J. T. Thorson, J. M. Cope, Realizing the potential of
(2019). trait-based approaches to advance fisheries science. Fish Fish. 20, 1034–1050 (2019).
14. G. J. Edgar et al., Abundance and local-scale processes contribute to multi-phyla 37. A. J. Woodhead, C. C. Hicks, A. V. Norström, G. J. Williams, N. A. J. Graham, Coral reef
gradients in global marine diversity. Sci. Adv. 3, e1700419 (2017). ecosystem services in the Anthropocene. Funct. Ecol. 33, 1023–1034 (2019).
15. D. Jablonski et al., Out of the tropics, but how? Fossils, bridge species, and thermal 38. J. E. Cinner et al., Meeting fisheries, ecosystem function, and biodiversity goals in a
ranges in the dynamics of the marine latitudinal diversity gradient. Proc. Natl. Acad. human-dominated world. Science 368, 307–311 (2020).
Sci. U.S.A. 110, 10487–10494 (2013). 39. R. D. Stuart-Smith et al., Integrating abundance and functional traits reveals new
16. C. Mora, “Large-scale patterns and processes in reef fish richness” in Ecology of Fishes global hotspots of fish diversity. Nature 501, 539–542 (2013).
on Coral Reefs, C. Mora, Ed. (Cambridge University Press, 2015), pp. 88–96. 40. F. Micheli et al., High vulnerability of ecosystem function and services to diversity loss
17. G. J. Edgar, R. D. Stuart-Smith, Systematic global assessment of reef fish communities in Caribbean coral reefs. Biol. Conserv. 171, 186–194 (2014).
by the Reef Life Survey program. Sci. Data 1, 140007 (2014). 41. K. D. Clements, D. P. German, J. Piché, A. Tribollet, J. H. Choat, Integrating ecological
18. M. D. Spalding et al., Marine ecoregions of the world: A bioregionalization of coastal roles and trophic diversification on coral reefs: Multiple lines of evidence identify
and shelf areas. Bioscience 57, 573–583 (2007). parrotfishes as microphages. Biol. J. Linn. Soc. Lond. 120, 729–751 (2017).
19. D. Mouillot et al., Functional over-redundancy and high functional vulnerability in 42. D. R. Bellwood, R. P. Streit, S. J. Brandl, S. B. Tebbett, The meaning of the term
global fish faunas on tropical reefs. Proc. Natl. Acad. Sci. U.S.A. 111, 13757–13762 ‘function’in ecology: A coral reef perspective. Funct. Ecol. 33, 948–961 (2019).
(2014). 43. D. Mouillot, N. A. J. Graham, S. Villéger, N. W. H. Mason, D. R. Bellwood, A functional
20. B. Bramon Mora, D. Gravel, L. J. Gilarranz, T. Poisot, D. B. Stouffer, Identifying a approach reveals community responses to disturbances. Trends Ecol. Evol. 28, 167–177
common backbone of interactions underlying food webs from different ecosystems. (2013).
Nat. Commun. 9, 2603 (2018). 44. E. Maire, G. Grenouillet, S. Brosse, S. Villéger, How many dimensions are needed to
21. A. L. Pigot, C. H. Trisos, J. A. Tobias, Functional traits reveal the expansion and packing accurately assess functional diversity? A pragmatic approach for assessing the quality
of ecological niche space underlying an elevational diversity gradient in passerine of functional spaces. Glob. Ecol. Biogeogr. 24, 728–740 (2015).
birds. Proc. Biol. Sci. 283, 20152013 (2016). 45. J. E. Allgeier, T. C. Adam, D. E. Burkepile, The importance of individual and species-
22. A. Baselga, Partitioning the turnover and nestedness components of beta diversity. level traits for trophic niches among herbivorous coral reef fishes. Proc. Biol. Sci. 284,
Glob. Ecol. Biogeogr. 19, 134–143 (2010). 20170307 (2017).
23. J. W. Lichstein, Multiple regression on distance matrices: A multivariate spatial 46. J. A. Sheridan, D. Bickford, Shrinking body size as an ecological response to climate
analysis tool. Plant Ecol. 188, 117–131 (2007). change. Nat. Clim. Chang. 1, 401–406 (2011).

McLean et al. PNAS | 9 of 10

Trait similarity in reef fish faunas across the world’s oceans https://doi.org/10.1073/pnas.2012318118
 47. S. Des Roches et al., The ecological importance of intraspecific variation. Nat. Ecol. 56. J. E. Duffy, J. S. Lefcheck, R. D. Stuart-Smith, S. A. Navarrete, G. J. Edgar, Biodiversity
Evol. 2, 57–64 (2018). enhances reef fish biomass and resistance to climate change. Proc. Natl. Acad. Sci.
48. B. M. Bolker, Ecological Models and Data in R (Princeton University Press, 2008). U.S.A. 113, 6230–6235 (2016).
49. N. J. Gotelli, Null Model analysis of species co-occurrence patterns. Ecology 81, 57. G. Liu et al., Reef-scale thermal stress monitoring of coral ecosystems: New 5-km
2606–2621 (2000). global products from NOAA coral reef Watch. Remote Sens. 6, 11579–11606 (2014).
50. S. W. Kembel et al., Picante: R tools for integrating phylogenies and ecology. Bio- 58. L. Tyberghein et al., Bio-ORACLE: A global environmental dataset for marine species
informatics 26, 1463–1464 (2010). distribution modelling. Glob. Ecol. Biogeogr. 21, 272–281 (2012).
51. E. Laliberté, P. Legendre, B. Shipley, M. E. Laliberté, Package ‘FD (Meas. Funct. Divers. 59. M. J. Behrenfeld, P. G. Falkowski, Photosynthetic rates derived from satellite-based
chlorophyll concentration. Limnol. Oceanogr. 42, 1–20 (1997).
Mult. Traits Tools Funct. Ecol, 2014).
60. M. W. Pennell et al., Geiger v2.0: An expanded suite of methods for fitting macro-
52. F. Husson, J. Josse, S. Le, J. Mazet, M. F. Husson, Package ‘FactoMineR.’ Package
evolutionary models to phylogenetic trees. Bioinformatics 30, 2216–2218 (2014).
FactorMineR (2020).
61. S. C. Goslee, D. L. Urban, The ecodist package for dissimilarity-based analysis of
53. F. Leprieur et al., Quantifying phylogenetic beta diversity: Distinguishing between
ecological data. J. Stat. Softw. 22, 1–19 (2007).
‘true’ turnover of lineages and phylogenetic diversity gradients. PLoS One 7, e42760
62. B. B. Nardelli, A novel approach for the high-resolution interpolation of in situ sea
(2012). surface salinity. J. Atmos. Ocean. Technol. 29, 867–879 (2012).
54. J. Chang, D. L. Rabosky, S. A. Smith, M. E. Alfaro, An r package and online resource for 63. Ø. Seland et al., The Norwegian Earth System model, NorESM2 - Evaluation of
macroevolutionary studies using the ray-finned fish tree of life. Methods Ecol. Evol. theCMIP6 DECK and historical simulations. Geosci Model Dev Discuss 2020, 1–68
10, 1118–1124 (2019). (2020).
55. A. Baselga, C. D. L. Orme, betapart: An R package for the study of beta diversity. 64. A. Chao et al., An attribute-diversity approach to functional diversity, functional beta
Methods Ecol. Evol. 3, 808–812 (2012). diversity, and related (dis)similarity measures. Ecol. Monogr. 89, e01343 (2019).
Downloaded from https://www.pnas.org by UNIVERSIDADE FEDERAL DO ESTADO DO RIO DE JANEIRO on June 19, 2025 from IP address 200.130.54.39.

10 of 10 | PNAS McLean et al.

https://doi.org/10.1073/pnas.2012318118 Trait similarity in reef fish faunas across the world’s oceans