pronounced, possibly because of the smaller
number of available samples and wider range
of sampled depths (figs. S9 and S10). Although
V9 OTUs may aggregate taxa with distinct
biogeographies ( 10 , 13 ), this should weaken
biogeographic differences across planktonic
groups rather than generate spurious ones.
In addition, we found consistent patterns with
the alternative marker genes 18S-V4 rDNA
andpsbO, which suggests that our results are
robust to variations in taxonomic resolution
and coverage (see supplementary text S2). The
notable differences that we observed across
planktonic groups suggest that accounting for
their specificities is crucial to understanding
their biogeography (Fig. 2 and fig. S5).
We investigated how biogeographic differ-
encesamongmajorgroupsrelatetotheirdi-
versity (number of OTUs), mean body size,
and dominant ecological function, the latter of
which we coarsely categorized as phototroph,
phagotroph, metazoan, or parasite (table S1
and materials and methods). We found that
the strength of biogeographic structure (group
position on the first axis of variation; Fig. 2A)
is strongly correlated to diversity [Spearman
correlation coefficient (rS) = 0.76 below 2000
OTUs; Fig. 3A]. This may be partly explained
by the increased smoothing of stochastic var-
iations and the larger amount of statistical in-
formation that is available in more-diverse
groups, yet we found that the correlation be-
tween short-distance spatial autocorrelation
and original group diversity persists after rare-
faction to the same number of OTUs (fig. S11).
This suggests that the generation and main-
tenance of eukaryotic plankton diversity are
influenced by biogeographic structure, per-
haps, in relation to endemism. Position on
the first axis is also weakly anticorrelated to
group mean body size (rS=−0.45,P= 10−^4 ;
fig. S12A) and, after controlling for diversity,
is lower for metazoans than for phototrophs
[analysis of covariance (ANCOVA) two-sided
ttest:P= 0.04; fig. S12D]. This agrees with the
expected impact of turbulent stirring, which
has been shown to generate patchier distribu-
tions in larger plankton ( 20 , 21 ). By contrast,
the nature of biogeographic patterns (group
position on the second axis; Fig. 2A) is strongly
correlated to group mean body size (rS= 0.61;
Fig. 3B) and ecological function [analysis of
variance (ANOVA)Ftest:P= 10−^8 ;Fig.3C]and
only weakly correlated to diversity (rS= 0.20,
P= 0.1); fig. S12E). Although mean body size
covaries with ecological function (phagotrophs
are larger than phototrophs on average, and
metazoans are substantially larger than other
plankton types; fig. S13), the positive relation-
ship between group position on the second axis
and mean body size still holds within ecolog-
ical categories (ANCOVAFtest:P=0.006;fig.
S14). Conversely, ecological function signifi-
cantly influences group position on the second
axis even after accounting for body size differ-
ences (ANCOVAFtest:P=0.03).Thesetrends
hold when directly considering measures of
biogeographic structure (such as short-distance
spatial autocorrelation and scale of biogeo-
graphic organization) rather than position on
the two axes of variation (figs. S15 and S16) and
when accounting for within-group variations
in body size (fig. S17). In conclusion, diverse
groups are more biogeographically structured,
with large-bodied heterotrophs (metazoans such
as copepods and tunicates) displaying a basin-based structure at the scale of oceanic basins or
larger and small-bodied phototrophs (such as
Mamiellophyceae and Haptophyta) display-
ing a latitude-based structure at a smaller
spatial scale.
A global biogeography that matches oceanic
basins suggests that communities are ho-
mogenized by ocean circulation at the basin
scale ( 13 ) but have little ability to disperse be-
tween basins, either because of the compara-
tively limited connectivity by currents ( 22 ) or
because of environmental barriers. Conversely, a
biogeography that matches latitude and is
symmetric with respect to the equator suggests
a higher coupling to environmental variations
(that are partly structured by latitude) and low
between-basin dispersal limitation. To test these
hypotheses, we used minimum transport times
between pairs of stations, which were previ-
ously computed from a global ocean circulation
model ( 13 , 23 ), and yearly averaged local
environmental conditions. We transformed
the matrix of minimum transport times into
spatial patterns at different scales through
eigenvector decomposition and thus obtained
a set of so-called Moran eigenvector maps
(hereafter simply referred to as“Moran maps”;
see materials and methods). Each map repre-
sents the hypothetical geographic patterns
expected for plankton with temporal varia-
tion along currents matching a specific spatial
scale (figs. S18 and S19). We estimated local
abiotic conditions by using yearly averaged
values of temperature, nutrient concentration,
and oxygen availability [World Ocean Atlas
2013 ( 24 ); see materials and methods]. Be-
cause biotic interactions (predation, competi-
tion, and parasitic and mutualistic symbiosis)SCIENCEscience.org 29 OCTOBER 2021•VOL 374 ISSUE 6567 597
S=0.76
P=2e–12S=
==
=–0.2–0.10.00.10.21e+02 1e+03 1e+04 1e+05
Diversity (#OTUs), log scaleBiogeographic axis 1
S=0.61
–0.2 P==6e–08–0.10.00.10.230 100 300
Body size (μm), log scaleBiogeographic axis 2
ANOVA:
P=1e–08
–0.2–0.10.00.10.2PhototrophsPhagotrophsMetazoansParasitesPoriferaABCCollodariaDiplonemidaFig. 3. Relationship between biogeography and diversity, mean body
size, and ecological function across major eukaryotic plankton groups.
(A) Group position along the first axis of biogeographic variation, which is indicative
of the strength of biogeographic structure (see Fig. 2A) and increases sharply
with diversity up to about 2000 OTUs (rS, two-sidedttest) but not beyond (as
exemplified by Diplonemida and Collodaria, two of the most diverse groups).
(B) Group position along the second axis increases with mean body size, which
indicates that large-bodied plankton are organized at larger spatial scale and
according to oceanic basins rather than latitude. (C) Positions along the second
axis were binned into four broad ecological categories (N= 16, 28, 12, and 12,
respectively; Collodaria and Dinophyceae were not categorized and are therefore
not represented). Metazoan groups score high (except Porifera sponges,
probably at the larval stage, which we excluded from statistical tests), phototrophs
score low, and phagotrophs occupy an intermediate position (ANOVA,Ftest).
Parasites score just below metazoans, which may reflect shared biogeography
because of the importance of metazoans as hosts.RESEARCH | REPORTS