Phytoplankton diversity and community structure affected by
oceanic dispersal and mesoscale turbulence
Marina Lévy1*, Oliver Jahn2, Stephanie Dutkiewicz2, Michael J. Follows2
DISPERSAL IMPACT ON PLANKTON DIVERSITY
* Corresponding author, [email protected]
1
LOCEAN-IPSL, CNRS/UPMC/IRD/MNHN, Paris, France
2
Department of Earth, Atmospheric and Planetary Sciences, Massachusetts Institute of
Technology, Cambridge, USA
1
1
Abstract
2
We explore the role of oceanic dispersal in setting patterns of phytoplankton diversity,
3
with emphasis on the role of mesoscale turbulence, in the context of numerical simulations
4
that resolve mesoscale eddies and a diverse set of phytoplankton types. The model suggests
5
that dispersal of phytoplankton by oceanic transport processes increases phytoplankton
6
diversity at the local scale of O(10-100) km (α-diversity), extends the range of many
7
phytoplankton types, and decreases the ability of rare types to persist in isolated areas. As a
8
consequence, assemblages are modified and diversity is decreased at the regional scale of
9
O(1000) km (γ-diversity). By progressively accounting for different classes of motion, we
10
show that the increase of α-diversity ensues from vertical mixing of the organisms, dispersal
11
by mean lateral currents and, in slightly larger proportion, by dispersal due to eddies. With the
12
progressive inclusion of mechanisms of dispersal, the population becomes dominated by a
13
smaller number of types but with larger degree of co-existence, in larger home range areas. A
14
resource competition interpretation shows that physical transport can reduce the effective
15
subsistence concentration of a limiting resource. In addition, co-existence of types with
16
unequal fitness is enabled by disequilibrium. The simulations suggest that mesoscale
17
turbulence plays a particular role, concomitantly providing a means for different types to
18
achieve comparable fitness and extending the exclusion timescale for less competitive types.
19
20
21
22
23
Keywords: plankton, biodiversity, dispersion, eddies
24
25
2
25
26
27
Lay abstract
28
freely transported by the flow and make up the base of the oceanic food web. This work
29
explores how their diversity is affected by the dispersal induced by different scales of motions
30
in the flow. Previous theoretical studies, based mostly on terrestrial ecosystems, have
31
suggested that the number of types able to co-exist at the same location should increase with
32
increasing levels of dispersal. Here we test this hypothesis for a realistic range of oceanic
33
conditions using a set of computer simulations of phytoplankton ecology coupled to a
34
turbulent ocean circulation model. Our results illustrate that with the progressive increase of
35
means for dispersal, the global population becomes dominated by a smaller number of
36
phytoplankton types but with larger ability for co-existence and in larger home-range areas.
37
We show that dispersal due to oceanic eddies contribute significantly to these changes.
38
Moreover, we show that mixing of populations can only partly explain the increase in local
39
co-existence. We offer a complementary view in which dispersal allows more types to
40
become equally competitive.
Phytoplankton are an extremely diverse set of floating, microscopic organisms which are
41
42
3
42
Introduction
43
[1] Phytoplankton form the base of the marine food web and critically mediate the earth's
44
carbon cycle. The significant diversity of this set of organisms is important for the ecology
45
and biogeochemistry of the ocean, and almost certainly lends stability to the system (Ptacnik
46
et al. 2008; Tilman et al. 1997). The question of what maintains the magnitude and patterns of
47
diversity have long been discussed (e.g., Hutchinson 1961; Barton et al. 2010) and are not yet
48
resolved. Here we focus on the role of dispersal in the fluid environment of the open ocean.
49
[2] Dispersal in the terrestrial environment has been explored and understood as a
50
community-structuring mechanism (Pulliam 1988), expanding niches (Pulliam 2000) and
51
affecting diversity (Mouquet and Loreau 2003). Ecologists distinguish between diversity at
52
the local scale and at the regional scale: α-diversity describes the number of locally co-
53
existing species at a given location; in contrast, at broader scale, γ-diversity designates the
54
sum of existing local communities over an extended region; those do not necessarily co-exist
55
locally. Theoretical results suggest that moderate dispersal has the ability to increase α–
56
diversity at constant γ –diversity (Mouquet and Loreau 2003). Compilation of experimental
57
data (Cadotte 2006) confirmed this theoretical trend for α–diversity, but are more ambiguous
58
as to what happens to γ –diversity.
59
[3] Dispersal is an essential element of the oceanic habitat, yet only recently have
60
oceanographers started to relate dispersal and phytoplankton diversity at basin scale in the
61
open ocean (though much study has been undertaken with dispersal of planktonic larvae in the
62
coastal oceans, e.g. Cowen et al, 2000). Some recent open ocean studies suggest that
63
phytoplankton assemblages might be structured by dispersal including recent global-scale,
64
remote-sensing approaches (d’Ovidio et al. 2010; De Monte et al. 2013) and numerical
65
simulations (Barton et al. 2010; Clayton et al. 2013). These studies suggest that dispersal
66
could explain the existence of localized phytoplankton diversity hotspots. In parallel,
67
idealized numerical modeling studies with limited “two types competing for one resource”
68
setups have emphasized that the two types could only co-exist after inclusion of means for
69
dispersal (Bracco et al. 2000; Perruche et al. 2010, 2011). Importantly, these studies have
70
revealed the influence of dispersal arising from different scales of motion (Table 1): the large-
71
scale lateral motions associated with the mean currents (Barton et al.2010; Clayton et al.
72
2013), the 3D-motions associated with mesoscale eddies and frontal dynamics (hereafter
4
73
mesoscale turbulence; Bracco et al. 2000; Perruche et al. 2011) and the vertical motions
74
resulting from convective mixing (hereafter vertical diffusion, Perruche et al. 2010).
75
[4] Here we examine the relationship between α–diversity, γ –diversity and dispersal in
76
the ocean and identify which scales of motion might be the most significant in shaping
77
assemblages. We test the hypothesis that dispersal of phytoplankton by oceanic transport
78
processes (1) increases the ability of different types to co-exist, (2) expands the area of their
79
niches, and (3) that transport processes at the mesoscale and submesoscale contribute
80
significantly to these changes. We investigate the consequences on diversity at the regional
81
scale and whether the changes ensue from mixing of nearby communities or are associated
82
with modifications of the community structure.
83
[5] To unravel the role of vertical mixing, mean currents and mesoscale turbulence in
84
regulating phytoplankton local diversity, regional diversity and community structure, we
85
employ a recent ecosystem model framework (referred to as “DARWIN” in the following,
86
Follows et al 2007), which here resolves a hundred distinct phytoplankton types, resulting in
87
flexible self-assembly of communities. Previous studies have used this tool to examine the
88
role of mechanisms including competitive exclusion, near-neutral co-existence (Barton et al.
89
2010), top-down control (Prowe et al. 2012), and immigration (Clayton et al. 2013) in shaping
90
patterns of diversity. Here we embed this ecosystem model within a high-resolution physical
91
simulation representative of the Northwest Atlantic (Lévy et al. 2010) that captures basin
92
scale circulation and resolves mesoscale features.
93
[6] In the following section, we describe in more detail the modeling approach, model
94
diagnostics and measures of diversity. Then we introduce a theoretical framework to explore
95
the roles of diffusion and transport in setting diversity in a hierarchy of idealized simulations
96
for which we present and discuss the results.
97
Method
98
[7] We used a model that resolves features analogous to the Gulf Stream, stratified
99
subtropics, and strongly seasonal subpolar regimes (Fig. 1). A physical simulation (Levy et al
100
2010) provided the physical fields used to drive the DARWIN ecosystem model in a series of
101
"off-line" experiments. Next we present the model used, the series of experiments, the model
102
diagnostics and the measures of diversity that will be analyzed.
5
103
Physical simulation
104
[8] The physical model is an idealized representation of the North Atlantic seasonally
105
varying subpolar and subtropical gyres at high-resolution (Lévy et al. 2010). In this model, a
106
baroclinicaly unstable jet (the model’s Gulf Stream) separates a warm, oligotrophic
107
subtropical gyre from a colder, more productive subpolar gyre (Fig. 1A,B). A region of
108
highest eddy kinetic energy (> 0.15 m2.s-2) is located in the offshore extension of the jet,
109
with areas of moderate eddy kinetic energy (> 0.05 m2.s-2) on the southern and northern
110
flanks of the jet, consistent with the levels observed in the North Atlantic (Fig. 1C,D). The
111
emerging mesoscale turbulence is characterized by a large number of interacting mesoscale
112
eddies, which give rise to intense vertical circulation at submesoscale fronts (Fig. 2A,B). Five
113
years of high-resolution velocities, temperature and vertical mixing saved from the physical
114
run of Lévy et al (2010) were used to drive the DARWIN model (see Appendix A for more
115
details).
116
DARWIN ecosystem model
117
[9] The ecosystem model is modified from Barton et al (2010) (see Dutkiewicz et al. 2009
118
for details, equations and parameter values). It resolves the cycling of nitrogen, phosphorus
119
and silica through inorganic, living and dissolved and particulate organic forms. The
120
simulations resolve separate pools of ammonium, nitrite, and nitrate, but does not nitrogen
121
fixation. There are 100 phytoplankton types each with unique physiology. Light-,
122
temperature- and resource-dependent growth, along with sinking, grazing, and other mortality
123
shape their relative fitness. Transport and mixing by the fluid flow also significantly influence
124
the interactions and development of populations. The phytoplankton types are initialized with
125
randomly assigned broad range of physiological attributes for temperature, light and nutrients.
126
We resolve two grazers size classes.
127
Offline Experiments
128
[10] To examine the influence of dispersal on phytoplankton diversity, we conducted a
129
series of off-line experiments using the physical fields provided by the physical run as
130
external forcings. All offline experiments were run within the MIT-GCM (Massachusets
131
Institute of Technology General Circulation Model) framework (Marshall et al. 1997) at 1/9°
132
horizontal resolution and integrated forward for 20 years, by repeating 4 times the 5 years of
133
forcings provided by the physical model. After about 7 years a repeating annual cycle in
6
134
ecosystem structure emerged. Results shown in this paper are shown for the last 3 years of the
135
simulations.
136
[11] In the most comprehensive experiment (hereafter HR for High Resolution), the
137
DARWIN ecosystem model was forced by the high-resolution physical fields (velocity,
138
vertical mixing and temperature) from the physical run (Figs. 2A, 2B). For details on
139
DARWIN initial conditions, the reader is referred to Appendix A. We used the HR simulation
140
to evaluate the realism of the model and to provide the monthly nutrient field that forces the
141
following four experiments.
142
[12] To unravel the influence of the different physical transport mechanisms driving
143
phytoplankton dispersal (i.e., mean advection, eddy advection and vertical mixing), we then
144
conducted a coherent series of four off-line experiments where we progressively modified the
145
advecting velocity and vertical mixing fields (Table 2). To ensure that the only difference was
146
in the phytoplankton transport equation (Eq.1) and not other environmental conditions, we
147
imposed identical nutrient and temperature distributions in this set of “forced” simulations.
148
Practically, treating nutrient as external forcing was done by strongly restoring (with a one
149
day time scale) the nutrient field to the monthly mean of experiment HR, coarse-grained to 1°
150
(Fig. 2D). For temperature, we used the temperature from the physical run (Fig. 1A) monthly
151
averaged and coarse-grained to 1°. Hence, in the four forced runs the only sources of small-
152
scale fluctuation were plankton advection and plankton vertical mixing. The suite of
153
experiments was conceived as follows: starting from no transport (experiment 0D) we
154
progressively added vertical mixing (experiment 1D), mean advection (experiment 3D-m) and
155
eddy advection (experiment 3D-e) in the phytoplankton transport equation. Thus in the 0D
156
experiment, there was no transport of phytoplankton. In the 1D experiment, we added
157
constant vertical mixing throughout the water column Kc=10-4 m2s-1. In the 3D-m experiment,
158
phytoplankton was diffused vertically with Kc and advected with the annual mean velocity of
159
the physical run, coarse-grained at 1° (Fig. 2C). In the 3D-e experiment, phytoplankton were
160
diffused vertically with Kc and advected with the high-resolution velocities of the physical run
161
(Fig. 2A,B). Note that since they were started from the same base field, the mean currents in
162
3D-m and 3D-e were therefore identical. We made the choice of using a constant vertical
163
mixing coefficient in the four forced runs to make them more comparable. In the HR run, the
164
K field has strong seasonal variations, particularly in the north.
7
165
Emerging community structure and diversity
166
[13] After an initial adjustment, the biomass of some of the initial 100 types of
167
phytoplankton fell below the threshold of numerical noise, and these types were assumed to
168
have become extinct. The remaining types Pj, and their relative proportions p j =
Pj
∑P
,
j
j
169
constitute what we refer to as the “community structure”. In our experiments, the community
170
171
structure is not imposed, it “self assembles” according to the relative fitness of the
€
phytoplankton types. Our suite of experiments enables different community structure to
172
emerge, which we compare. We evaluate the “change in community structure” between 2
173
model experiments A and B as
∑
j
174
175
p Aj − p Bj
2
. This index varies between 0 and 1. It equals 0
when all types are in exact same proportion in the two runs, and 1 when all types differ.
[14] Different types
€ Pj in the community are not present everywhere, but occupy specific
176
habitat whose surface, the “home range area”, varies. We empirically defined it as the area
177
where the annual mean, depth integral of Pj was larger than a concentration threshold of 10-2
178
mmole P/m2. With this definition, the home range is essentially the "realized niche"
179
(Hutchinson, 1957). We defined “rare types” as the phytoplankton types whose home range
180
area occupied less than 5% of the domain.
181
[15] For diversity, we defined the metric ‘‘richness’’ to be the number of phytoplankton
182
types Pj that exceed a relative threshold biomass concentration of Pj > 10-5 max(Pj) (similar to
183
Barton et al. 2010; Prowe et al. 2012). We distinguished between local richness (at the scale
184
of the model grid, O(10)km), from regional richness (at the scale the model domain,
185
O(1000)km). Practically, local richness was computed at each grid cell and for each month,
186
using local, monthly mean concentrations for Pj. Local richness were sensibly the same when
187
using full resolution Pj or 1° coarse-grained Pj, revealing that local richness represented the
188
richness over the O(10-100)km scale range. Regional richness was computed using domain-
189
averaged concentrations for Pj. With these definitions in mind, local richness provides a
190
measure of the α-diversity, while regional richness corresponds to γ-diversity.
191
192
[16] We also used the local Shannon Index (H) which gives a complementary view: it
measures the joint influence of types richness and evenness (Stirling and Wilsey 2001):
8
193
j
194
195
€
( )
H = −∑ p j ln p j , where pj is the biomass of Pj divided by the total biomass ( p j = P j
∑ P ).
j
j
H has its maximum value of ln(n), with n the number of types, when all types are represented
by equal amounts.
€
196
[17] There is a certain degree of seasonality and vertical structure in the local richness,
197
because different phytoplankton types are adapted to different light and nutrient levels. Here,
198
we present the richness after averaging over the first 100 meters and over the year. We noted
199
that using annual mean Pj instead of monthly mean did not significantly change our results,
200
and that surface richness showed very similar patterns to the depth integrals. Regional
201
richness did not show any seasonal variations, because a type has to persist all year long to be
202
selected by the model. The modeled richness depends to a limited extent on the physiological
203
traits of phytoplankton types initialized (as indicated by an ensemble of related simulations
204
with different initial sets of parameters for the phytoplankton types; Barton et al. 2010) as
205
well as on the chosen threshold. However, the key patterns and results discussed here are
206
robust.
207
Theoretical Framework
208
[18] To illustrate how dispersal might affect the community structure, and the α and γ
209
diversities we first describe two theoretical concepts: “mimimum R*” and “contemporaneous
210
disequilibrium”.
211
[19] Minimum R*: This theoretical framework follows from resource competition theory
212
(Tilman 1977, 1982) with modifications here for transport terms. We can represent
213
phytoplankton changes in time be the following equation (simpler than the numerical model)
214
where several types j with biomass Pj competing for a resource R:
215
∂P j
R
= µj
P − m j P j + V j P j + M j P j +E j P j
∂t
R+k j j
(1)
216
€
217
Here, µj is a maximal growth rate of type j, a function of light and temperature. Nutrient
218
type j, and mj represents a simple parameterization of sinking, grazing, viral lysis and other
219
loss terms. Vj is the vertical mixing per unit biomass ( V j =
limitation is parameterized as a Monod function where kj is the half-saturation constant for
€
1
∂ K∂ P ), Mj is the mean
Pj z z j
9
1
∇⋅ uP j ), and Ej is eddy transport per unit biomass (
Pj
220
transport per unit biomass ( M j = −
221
1
∇⋅ u' P' j ), where u is the mean current, u’ is the eddy circulation and K the vertical
Pj
€
mixing coefficient. Thus the first two term on left hand side of the equation denote local
222
223
€
224
Ej =−
growth and loss, the third represents vertical dispersion, the fourth represents the large scale
advection, and the fifth the eddy motions. The convention we use here is that any addition of
225
biomass by any of these processes to a location has a positive value (i.e. Vj, Mj, Ej>0) and
226
removal has a negative value.
227
228
229
230
€
231
232
[20] To explore how competition for resources can set community structure, we solve for
the equilibrium solution at steady state Rj*. We get the expression:
R*j =
(
k j m j − M j − E j −Vj
(
)
µj − m j − M j − E j −Vj
)
(2)
In the absence of transport or mixing (as in Experiment 0D) Eq. (2) reduces to:
R*j =
k jm j
µj − mj
(3)
[21] The equilibrium resource concentration Rj* suggests that the ambient concentration
€
233
of the limiting resource is determined by characteristics of the organism including its
234
maximum growth rate (µj) nutrient half-saturation constant (kj), and mortality rate (mj). The
235
ambient resource concentration will be drawn down to the lowest positive R* amongst the
236
organisms present and other organisms will be excluded over time [Stewart and Levin, 1973].
237
If steady-state conditions are satisfied, types can co-exist if they have the same (lowest
238
positive) Rj* which can be accomplished by various combination of physiological parameters
239
kj, mj and µj.
240
[22] In the presence of vertical mixing (neglecting large scale transport and eddy
241
transport for present, as in Experiment 1D), there become more ways to have the same Rj*:
242
Eq. (4)
243
If a type has a transport source supplied to it (i.e. Vj>0) this will reduces its effective R*,
244
making it potentially more competitive with another local population.
10
245
[23] Consider for example two depths z1 and z2 and three types of plankton A, B and C.
246
Assume that the types are identical except for their growth rate (i.e. kA=kB=kC, mA=mB=mC).
247
Here we assume that at depth z1, µ1A>µ1B>µ1C and at depth z2 µ2B>µ2A>µ2C. Without vertical
248
mixing, R*1A<R*1B<R*1C suggesting that plankton type A will be more competitive at depth
249
z1and type B and C will be excluded. Similarly at depth z2, type B will be more competitive
250
and A and C will be excluded. Thus the community structure will consist of type A at depth z1
251
and type B at depth z2; there will be no co-existence of A and B and type C will be excluded
252
everywhere. Now assume mixing between depth z1 and z2: at depth z1, V1A is negative - type A
253
is removed, but V1B is positive - type B is brought into that depth (visa versa at depth z2 for
254
V2A and V2B). Thus, at depth z1 the R*1A for type A will now include the V1A (negative) term,
255
and this will increase the effective R*1A, but the R*1B for type B decreases since V1B is positive.
256
If V1A and V1B are large enough, R*1A=R*1B and the two types will co-exist at depth z1.
257
Moreover type C can become competitive if V1C is sufficiently large. On the other hand, if V1A
258
is strongly negative, type A will have a potential loss of competitivity (R*1A will become
259
greater than R*1B) and A will become extinct. Thus vertical mixing adds an extra dimension
260
for the competition between types that implies more potential for co-existence and potential
261
for change of the γ –diversity and community structure.
262
[24] Similarly, adding large-scale transport Mj (Experiment 3D-m) or eddy terms Ej
263
(Experiment 3D-e), adds yet another dimension that allows for different levels of competition
264
and potential for co-existence. Though, it should be stressed that eddy mixing is intermittent
265
by nature and thus potentially constantly disrupts the steady state hypothesis that supports the
266
definition of R*. In fact, to properly account for eddy terms in the definition of R*, we must
267
assume that the system is in statistical steady-state, i.e. in steady-state after averaging over
268
many possible realizations of the turbulent system.
269
[25] The R* scenario is thus relevant when the system is close to statistical steady-state
270
equilibrium, i.e. all year long in the subtropical ocean and during summer in the subpolar
271
ocean, as shown by Dutkiewicz et al. (2009; their Fig. 4). Indeed in the subpolar ocean,
272
summer equilibrium is halted by nutrient entrainment in winter leading to an intense bloom in
273
spring. This sudden disruption eventually leads to competitive exclusion of all but the single
274
phytoplankton type that grows fastest (Barton et al. 2010).
11
275
[26] Contemporaneous disequilibrium: The “contemporaneous disequilibrium”
276
(Richerson et al. 1970), in contrast, suggests that dispersal can maintain types with unequal
277
fitness out of equilibrium. Let’s consider that, in the absence of dispersion, the community
278
consists of two types A and B living in adjacent habitats SA and SB: A is at a competitive
279
advantage in SA, and B is more competitive in SB; A and B do not co-exist. At any location,
280
there is only one type, A or B. In this situation, α –diversity is one and γ –diversity is two.
281
Dispersal can allow type A to continuously invade SB. If this occurs frequently enough and at
282
a rate faster than the competitive exclusion time scale, than types A and B can now co-exist
283
over SB. α –diversity becomes two over SB, γ –diversity remains two. This theoretical
284
framework was illustrated by the model experiments of Perruche et al. (2010; 2011) with two
285
types of phytoplankton. In this scenario, dispersal increases α –diversity but without changing
286
the community structure and the γ –diversity.
287
[27] To sum up, in the “minimum R* scenario”, the phytoplankton types with the lowest
288
R* are expected to outcompete other phytoplankton types over time. Transport processes lead
289
to more ways to have the same minimum R*, but also can lead to some types having
290
increased R* and becoming less competitive. Thus community structure and γ –diversity can
291
be altered. This scenario is particularly relevant to the subtropical ocean. In contrast, in the
292
subpolar ocean, the steady-state hypothesis breaks and the “contemporaneous disequilibrium”
293
scenario is more plausible. In that case, α –diversity is expected to increase with increased
294
dispersal but without significant changes in the community structure and γ –diversity.
295
Results
296
[28] We now describe the changes in diversity, home range areas and community
297
structure in our series of forced offline experiments: 0D, 1D, 3D-m, 3D-e, where dispersion
298
was progressively increased by adding more and more ingredients to the transport equation.
299
Results from a second set of experiments (LR and HR, Table 2) which mimic 3D-m and 3D-e
300
but where environmental factors such as temperature and nutrient supply are allowed to
301
dynamically vary are discussed in Appendix B. They show similar results than the set of
302
experiments presented here. We start by evaluating our model for the HR experiment, which
303
is the most realistic in terms of circulation and forcing.
304
Model evaluation
12
305
[29] The total phytoplankton biomass distribution in our model was qualitatively and
306
quantitatively consistent with remote and in situ observations in the Northwest Atlantic (Fig.
307
3A,B), and similar to what was obtained with a model with a single phytoplankton type (Lévy
308
et al. 2012). It was characterized by a larger biomass in the subpolar gyre and lower
309
concentrations in the oligotrophic subtropical gyre.
310
[30] Richness was significantly larger in the subtropical gyre, where it averaged 14, and
311
dropped to half of that value in the subpolar gyre (Fig. 3C). There was also a marked diversity
312
hotspot over the jet area (Fig. 3C). The decline of diversity with increasing latitude and the
313
presence of diversity hotspots over boundary currents are patterns that had previously been
314
identified with a global coarse-resolution configuration of the DARWIN model (Barton et al.
315
2010). The hotspot over our model’s jet is corroborated by the observed interleaving of
316
several dominant groups of phytoplankton determined from remote-sensing observations
317
based on their optical anomalies (De Monte et al. 2013) over the Gulf Stream path (Fig. 3D).
318
Greater diversity over subtropical regions is not seen from remote-sensing observations (Fig.
319
3D). But this pattern is evident in in-situ observations of many taxa in both marine and
320
terrestrial ecosystems (Currie 1991; Hillebrand 2004).
321
[31] Thus, despite the very simplified geometry and forcing, the emerging community
322
structure in our model showed some consistency with observed phytoplankton distribution
323
and diversity which give us confidence in the ability of our model in representing the
324
processes responsible for this community structure. In the more dynamically consistent
325
simulations (3D-m, 3D-e), these patterns are also evident. However there are very strong
326
differences between the different experiments in terms of local and regional diversity, home
327
range and community structure that we describe and discuss further in the following.
328
Changes in local and regional phytoplankton diversity
329
[32] A key result is the increase of α-diversity, accompanied by a decrease of γ-diversity,
330
with increasing levels of dispersal (Fig. 4). In the absence of physical transport (experiment
331
0D), local diversity was low everywhere. No more than two types were able to co-exist at a
332
single location, while a total of more than 40 types (out of the initialized 100) persisted across
333
the domain. The progressive addition of transport processes (experiments 1D, 3D-m, 3D-e)
334
led to the progressive increase in the number of types that co-existed (α-diversity), from
335
averaged values of ~5 in 1D, to ~8 in 3D-m and ~15 in 3D-e. In parallel to this increase of
13
336
local diversity, we observed a significant decrease in the total number of types that persisted
337
in the domain (γ-diversity): from ~40 in 0D, to ~30 in 1D and ~20 in 3D-m.
338
Changes in home range area
339
[33] How can α-diversity be increasing and at the same time γ-diversity be decreasing?
340
This apparent paradox can be explained by examination of the change in home range area
341
with dispersal. In the absence of dispersal, types that persisted occupied limited geographic
342
areas in the horizontal and in the vertical and were extinct elsewhere (Fig. 5). These habitats
343
were characterized by a temperature, light and nutrient range, close to optimum values for the
344
types (not shown). The geographical distribution of the habitats was thus very zonal in close
345
association with the isolines of temperature and nutrients (Fig. 1 and 2), and followed
346
constant depth in the vertical, corresponding to homogeneous light levels. There was a clear
347
segregation between types that occupied the surface, those at subsurface, types in the subpolar
348
gyre, those in the subtropical gyre and some others in narrower regions between the two
349
gyres. About 50% of the types that persisted could be categorized as “rare”, meaning that they
350
had very small home range areas that covered less than 5% of the total area (Fig. 6). Thus in
351
the absence of dispersion, the physical landscape was organized in a large number of adjacent
352
areas of very small extension. In this extreme situation, the local diversity was low, i.e. very
353
few types coexisted at the same location but the regional diversity was large and there was a
354
large proportion of rare types. In presence of dispersal, the decrease of regional diversity
355
mostly reflected the disappearance of rare types (Fig. 6) that were excluded by invaders. In
356
parallel, the overall size of the area occupied by each phytoplankton type increased: in
357
experiment 3D-e, 12 types occupied more than 50% of the domain while in 0D and 1D, not a
358
single type occupied more than 30%. This implied that in 3D-e, there were large intersections
359
between areas, where many types co-existed. Our result thus illustrate that with the
360
progressive inclusion of means for dispersion, the global population became dominated by a
361
smaller number of phytoplankton types but with a higher degree of co-existence, and larger
362
ranges.
363
Changes in community structure
364
[34] Changes in diversity were accompanied by a change in community structure (Fig. 7).
365
The change in community structure was more complex than the simple extinction of some of
366
the types. Although the eight most dominant types in 3D-e were present in all runs, they
14
367
changed in proportion. It was not necessarily the less abundant types in 0D and 1D that
368
became extinct in 3D, but also some that were quite abundant (such as #s 95, 68 or 46). The
369
changes in community structure ranged between 20 to 50% (Table 3). They illustrated that, at
370
the type level, transport could either be beneficial or unfavorable, and the net effect would
371
depend on the type. But at the community level, more transport was favorable to more co-
372
existence.
373
[35] Figure 8 provides a schematic illustration of the biogeographic changes associated
374
with increasing dispersal that summarizes our results. Habitats of the different phytoplankton
375
types composing the community structure are represented by ellipses, whose size are
376
proportional to the home range area and whose color serve to characterize individual types.
377
The habitats are organized along the north-south temperature gradient. α-diversity is large
378
where ellipses intersect, γ-diversity corresponds to the number of ellipses, larger home range
379
areas correspond to larger ellipses. Note the change in community structure identified by the
380
partial change in colors between the low dispersion and high dispersion cases.
381
Discussion
382
Dispersal and α-diversity
383
[36] We have addressed the question of whether dispersal increases co-existence in
384
phytoplankton communities and, if so, why? To do so, we conducted a suite of model
385
experiments, with velocity and temperature fields from a high-resolution physical hydro-
386
dynamical model, coupled with a high-complexity (100 phytoplankton types) ecosystem and
387
biogeochemical model. The physical model simulated a mesoscale turbulence regime close to
388
oceanic conditions in the Gulf Stream region. The biogeochemical model enabled the
389
emergence of diverse phytoplankton communities. Our models suggest that dispersal and
390
mesoscale turbulence increased ability for co-existence at the local scale of O(10-100) km
391
(i.e. increased α-diversity), and changed the community structure. That local diversity should
392
increase with dispersal is consistent with hypotheses from theoretical ecological studies (e.g.
393
Mouquet and Loreau, 2003) and a meta-analysis of terrestrial ecosystem observations
394
(Cadotte, 2006).
395
396
[37] α-diversity, however, is far from homogeneous over our model’s domain (Fig. 3C).
Barton et al. (2010) obtained α-diversity patterns similar to ours in a global ocean model
15
397
coupled with the same DARWIN ecosystem model. They explained the larger diversity in the
398
subtropics by the relatively weak seasonality that enables coexistence of multiple
399
phytoplankton types with comparable fitness R*. In contrast, they argued that strong seasonal
400
variability of the environment in the subpolar ocean leads to competitive exclusion of
401
phytoplankton with slower growth rates and explains the lower diversity. In a subsequent
402
study at higher resolution, Clayton et al. (2013) suggested that the confluence of biomes along
403
with enhanced nutrient supplies partly explained the presence of diversity hotspots over
404
western boundary regions.
405
[38] Interestingly in this study, it was particularly in the subtropical gyre and jet area that
406
the progressive addition of transport processes (experiments 1D, 3D-m, 3D-e) led to the
407
progressive increase of α-diversity (Fig. 9). The increase was just as pronounced for the
408
Shannon index as for richness, revealing that the additional types were present at relatively
409
high abundances. In contrast, in the subpolar gyre, the increase in α-diversity was not as
410
marked. In fact, despite a moderate increase in the Shannon Index, local richness only
411
increased when mesoscale motions were resolved. This highlights that dispersal did not
412
significantly increase the number of co-existing types but made their relative proportions
413
more even.
414
[39] We can understand the large increase of α-diversity in the subtropical gyre with the
415
R* scenario. In the case of 0-D, competition reduces to Eq. 3: the phytoplankton with the
416
physiological characteristics that leads to the lowest R* will dominate. Since µj is a function
417
of light and temperature, different phytoplankton type will have the lowest R* in different
418
vertical and horizontal regions of the model domain. Co-existence within each grid-cell
419
occurs as there is a seasonal cycle of both temperature and light. Similar R* and long
420
exclusion times (Barton et al., 2010) allow there to be co-existence on the timescale of these
421
integrations. With the progressive inclusion of vertical mixing Vj, mean transport, Mj, and
422
eddy mixing Ej, competition within a grid cell becomes more complex. There are many more
423
ways to have similar R* (Eq. 2). Thus, though in a single grid cell, a phytoplankton type is
424
physiologically best suited (as in 0-D), transport or mixing out of the grid cell will increase
425
the R*. At the same time, transport or mixing into the grid cell of a different type from a large
426
population outside of the grid cell will lead to a decrease of that type’s R*. With the inclusion
427
of more ways to be transported or mixed in each of 1D, 3D-m, 3D-e more types in any
428
location have a similar R* and co-exist.
16
429
[40] In contrast in the subpolar region, the local diversity was low and previous studies
430
(Dutkiewicz et al. 2009; Barton et al. 2010) have shown that the types that have the fastest
431
growth rates dominate. Dispersion can potentially lead to the sustained mixing of nearby
432
areas, allowing for more types to co-exist in a state of “contemporaneous disequilibrium”.
433
[41] Moreover, the theory suggests changes in community structure should occur in the
434
case of the R* scenario, but not in the case of contemporaneous disequilibrium. Thus the
435
more drastic changes in community structure in the subtropical gyre compared to the subpolar
436
gyre (Fig. 10) further support the expectation that “R* minimum” is more likely to explain the
437
observed changes in the subtropical gyre and “contemporaneous disequilibrium” in the
438
subpolar gyre.
439
[42] We should note that we obtain a similar (although slightly larger) increase in α-
440
diversity in our set of forced experiments at controlled nutrient levels than in our set of free
441
runs (Fig. 4). This enables us to conclude that the increased ability for co-existence largely
442
ensues from phytoplankton dispersal by the eddying flow. However, differential nutrient
443
supplies over frontal regions allows there to be a larger increase in α-diversity in our free runs
444
(Fig. 11) suggesting that hotspots in jet regions develop due to mingling of different seed
445
populations provided with increased nutrient supply, consistent with Clayton et al (2013). The
446
complex controls on diversity in frontal regions still remain to be fully understood.
447
What are the consequences of dispersal on γ-diversity?
448
[43] Theoretical ecological studies hypothesized that global diversity would remain
449
constant with more dispersion (e.g. Mouquet and Loreau, 2003) but this hypothesis could not
450
be clearly tested against observations (Cadotte, 2006). The decrease in γ-diversity with
451
dispersal in our simulations was not anticipated, but we find the decrease is due to the strong
452
reduction in the number of rare types. In the absence of dispersion, these rare types are
453
characterized by very small home range area where they are best adapted and outcompete the
454
rest of the types. In presence of dispersion, emerging types have larger home range or
455
“realized niche” (Fig. 6) although in all experiments the "fundamental niche" (defined by
456
Hutchinson 1957 as the niche a organism would fill in the absence of any competition) are
457
identical. This highlights that types must be adapted to a wider range of environmental
458
conditions in order to survive and that the quality of competition is very different depending
459
on the amounts of dispersal.
17
460
[44] In a fully mixed system with sufficiently long exclusion times, α and γ-diversity
461
would be the same. However, though they are tending to towards similar values in our model
462
studies with increasing dispersal (Fig. 4), they do not reach the same number. This is because
463
the jet between the two gyres acts as a partial barrier and because the timescales for water
464
masses to mix throughout the domain are on the scale of, or longer than, the exclusion
465
timescales. Thus we always expect a γ-diversity larger than the α-diversity.
466
Which scales of motion are the most crucial to maintaining diversity?
467
[45] The forced experiments progressively added vertical mixing, mean transport and
468
eddy transport (0D, 1D, 3D-m, 3D-e) to explore how each altered diversity. All play a
469
significant role (Fig. 4). In principle, vertical diffusion and mean advection are very different,
470
because one adds dispersal on the vertical and the other on the horizontal direction; on the
471
other hand, eddy advection acts both on the horizontal and vertical directions because of the
472
strong sub-mesoscale vertical velocities (Fig. 2D).
473
[46] An intriguing result is that mesoscale turbulence, although the most effective at
474
maintaining high levels of diversity at the local scale, was not responsible for any significant
475
decline of diversity at the regional scale (unlike other means for dispersion) (Fig. 4).
476
However, the changes detected in the community structure between 3D-m (large scale
477
transport alone) and 3D-e (including eddies) were substantial (Fig. 7). Some types, such as #s
478
63, 17 and 44, were significantly more abundant in 3D-m, some others (#s 57, 4, 94) were
479
less; some were present in 3D-m but extinct in 3D-e (#s 61, 21, 80); others were present in
480
3D-e but extinct in 3D-m (#s 13, 23, 89).
481
[47] A possible explanation for this paradox lies in the particular transport properties of
482
mesoscale turbulence. Mesoscale eddies, when they form, decay or interact, are an effective
483
mean for mixing and dispersal. On the other hand, they also have the ability to isolate water
484
masses in their core for very long periods of time (e. g. Lehahn et al. 2011; d’Ovidio et al.
485
2013). The temporary refuge provided by eddies thus adds a level of complexity which could
486
explain the maintenance of less competitive types (Bracco et al., 2000; d’Ovido et al., 2010).
487
Along this line, we can hypothesis that mesoscale turbulence allows the minimum R* and
488
contemporary disequilibrium scenario to work in concert with one another: by improving
489
more ways to achieve the same fitness (thus increasing α-diversity) and by avoiding the
18
490
extinction of the less competitive types in the new, fluctuating environment (thus maintaining
491
γ-diversity).
492
Significance to aquatic environments
493
[48] This study suggests that dispersal leads to a complex pattern of diversity of
494
phytoplankton types in the ocean. The small ranges of types in the absence of dispersal
495
become significantly altered when the full spectrum of transport processes is accounted for. In
496
general dispersal leads to an increased home range, but not always. Best locally adapted types
497
are not necessarily the same in a stable environment and in an environment which puts them
498
constantly in movement. Hence dispersal does more than allowing co-existence of nearby
499
types by permanently mixing them: it allows more ways to co-exist.
500
[49] More precisely, we considered two ways through which dispersal can allow co-
501
existence: either by allowing more ways to achieve the same fitness (minimum R* scenario)
502
or by maintaining types with unequal fitness out of equilibrium (contemporary disequilibrium
503
scenario). A resource competition perspective suggests that the relative fitness of any given
504
phytoplankton community is regulated by a variety of factors, including physical conditions,
505
predation, competition for resources, variability of the environment and dispersal. In this
506
context, the types that are more likely to coexist are those who achieve the maximum fitness
507
(minimum R*) and this fitness depends on physiological parameters as well as physical
508
transport. Advection/ diffusion potentially increases the capacity for coexistence because it
509
provides more ways to achieve the same minimum R*. On the other hand, in the
510
contemporary disequilibrium case, advection/diffusion continuously puts in contact
511
communities that are best fit in nearby habitats. Because this occurs at a rate faster than
512
competitive exclusion, the communities can co-exist. The minimum R* appears to explain the
513
diversity and community structure changes in the stable subtropics. On the other hand in the
514
more dynamic subpolar regions changes are better explained by contemporary disequilibrium.
515
[50] An important message is that local competition between types in aquatic
516
environments cannot be understood in terms of local resource availability alone. We must also
517
account for the transport of populations. This suggests that in the aquatic environments,
518
diversity will be highly dependent not only on local nutrient and temperature conditions but
519
also on velocities, mixing, and on those conditions in surrounding linked regions.
520
19
520
521
522
Appendix A: Model details
Construction of physical forcing fields
523
[51] The physical forcing fields used to drive the DARWIN model in off-line mode come
524
from the high-resolution physical model run described in Levy et al (2010). This physical run
525
was performed on a high-resolution horizontal grid (1/54°) with 30 vertical levels with and
526
primitive equation ocean model NEMO (Madec 2008).
527
[52] The physical model domain is an idealized representation of the North Atlantic, in
528
the form of a rectangle rotated by 45° on the β-plane (Fig. 1). This choice was made to
529
encompass model equivalents of the Gulf Stream, subtropical and subpolar regimes in a
530
domain small enough to allow long integrations at high resolution. The model was forced at
531
the surface with seasonal buoyancy fluxes and wind.
532
[53] The model was spun-up for 850 years at low resolution (1°) and for another fifty
533
years at high-resolution (1/54°). We used 5 years of physical fields (velocities, temperature
534
and vertical mixing) saved from the original model run at a time-resolution of 2 days and a
535
grid resolution of 1/9°. This degradation does not affect the resolution of the finest resolved
536
structures: no difference can be seen between the coarse-grained 1/9° fields (Fig. 2A,B) and
537
the original fields at 1/54° resolution. The reason for this stems from the level of
538
dissipation/diffusion required during the online integration of the original model (Lévy et al.
539
2012b). We refer to these fields as the “high-resolution” physical fields.
540
Construction of initial conditions
541
[54] In Lévy et al. (2012a), this physical model was coupled online to the simple
542
biogeochemical model LOBSTER that comprises a single phytoplankton and run for another
543
50 years. We used the spun-up nitrate field from the Lévy et al. (2012a) run to initialize
544
DARWIN in the HR simulation. The LOBSTER model did not include phosphate or silica.
545
Thus phosphate was initialized as 1/16th of the nitrate field and silica with a uniform depth
546
profile consistent with the North Atlantic (Garcia et al. 2006). To initialize phytoplankton and
547
zooplankton, we used identical distributions of biomass for all.
548
20
548
549
550
Appendix B: Free runs
551
which mimic 3D-m and 3D-e but where environmental factors such as nutrient supply and
552
temperature vary dynamically. More precisely, in the set of experiments presented in this
553
paper (0D, 1D, 3D-m, 3D-e), we treated nutrients and temperature as a set of coarse-
554
resolution external forcings. This enabled us to impose the same external environment
555
conditions for all runs and to conduct a coherent suite where the only difference was in the
556
phytoplankton transport equation. In this second set of experiments (LR, HR), we run the full
557
ecosystem model freely at two different grid resolution: at high-resolution (HR), hence
558
capturing the full strength of eddy turbulence, and at low resolution (LR) thus only capturing
559
large-scale currents (Table 2). In the LR experiment, the full ecosystem model was run freely
560
with the velocity, vertical mixing and temperature fields from an original low resolution (1°)
561
online run, and initialized with the nutrient field of that run. In the HR experiment, the full
562
ecosystem was freely run with the high-resolution fields from the original high resolution run.
563
The environment conditions in the HR and LR experiments differ in many respects because
564
the dynamical fields were integrated at different resolutions. For instance in the HR run, the
565
Gulf Stream position is shifted south by 5° compared to the LR run, due to non-linear terms in
566
the momentum equation (Lévy et al., 2010). The change in resolution also induces different
567
supplies of nutrients, because the nitracline depth and the vertical velocities are different in
568
the two runs (Lévy et al., 2012a).
569
[55] Here we present and discuss a second set of experiments (LR and HR, Table 2)
[56] The change of local diversity between LR and HR was comparable in magnitude to
570
the change obtained between 3D-m and 3D-e (Fig. 4). Moreover, no significant change in
571
regional diversity was detected between CR and HR, as for 3D-m and 3D-e (Fig. 4). The
572
decrease of local richness in the LR run compared to the HR run was particularly strong in the
573
jet area and subtropical gyre (Fig. 11A), as for 3D-m and 3D-e (Fig. 9). This increase in local
574
diversity was also associated with an increase of the area occupied by each type (Figs. 11B
575
and 6). Finally, the change in community structure between LR and HR (0.31) was of the
576
same order than the change between 3D-m and 3D-e (0.24, Table 3). Thus the differences
577
between LR and HR are very similar to those between 3D-m and 3D-e.
578
21
578
579
580
581
582
583
584
585
586
587
Acknowledgements: ML is grateful to the Earth and Planetary Science department of MIT
for inviting her as a Houghton lecturer, which enabled to initiate this work. This manuscript
was improved thanks to the insightful comments of an anonymous reviewer. We acknowledge
useful discussions with Sophie Clayton, Alice Soccodato and Oliver Buhler. Fig. 3D was
kindly processed by Francesco d’Ovidio. We are grateful for support from CNRS/INSU
(TANGGO grant from LEFE) and National Science Foundation (grant OCE-1048926
MOBY).
References
588
Alvain, S., C. Moulin, Y. Dandonneau, and H. Loisel. 2008. Seasonal distribution and
589
succession of dominant phytoplankton groups in the global ocean: A satellite view. Global
590
Biogeochem. Cycles. 22(3): GB3001, doi:10.1029/2007GB003154.
591
Barton, A. D., S. Dutkiewicz, G. Flierl, J Bragg, and M. J. Follows. 2010. Patterns of
592
Diversity in Marine Phytoplankton. Science. 327(5972):1509–1511,
593
doi:10.1126/science.1184961.
594
Bracco, A., Provenzale, A., & Scheuring, I. 2000. Mesoscale vortices and the paradox of
595
the plankton. Proceedings of the Royal Society B: Biological Sciences. 267(1454):1795–
596
1800, doi:10.1098/rspb.2000.1212.
597
598
599
Cadotte, M.W. 2006. Dispersal and species diversity: A meta-analysis. Am. Nat. 167: 913924
Clayton, S., S. Dutkiewicz, O. Jahn, and M.J. Follows (submitted). Ocean eddies and
600
dispersal maintain phytoplankton diversity. Submitted to Limnology and Oceanography:
601
Fluids and Environment.
602
Cowen, R. K., K M. M. Lwiza, S Sponaugle, C. B. Paris, D B. Olson. 2000. Connectivity
603
of Marine Populations: Open or Closed? Science 287, 857, doi:10.1126/science.287.5454.857
604
Currie, D. J. 1991. Energy and large-scale patterns of animal-and plant-species richness.
605
606
607
Am. Nat. 137(1): 27–49.
De Monte, S., A. Soccodato, S. Alvain and F. d’Ovidio. 2013. Can we detect oceanic
biodiversity hotspots from space? ISME journal. doi:10.1038/ismej.2013.72.
22
608
Dutkiewicz, S., M. J. Follows and J. G. Bragg. 2009. Modeling the coupling of ocean
609
ecology and biogeochemistry. Glob. Biogeochem. Cycl. 23(4): GB4017,
610
doi:10.1029/2008GB003405.
611
612
Follows, M. J., S. Dutkiewicz, S. Grant and S. Chisholm. 2007. Emergent biogeography of
microbial communities in a model ocean. Science. 315(5820): 1843-1846.
613
Garcia, H. E., R. Locarnini, T. B. Boyer and J. L. Antonov. 2006. Word Ocean Atlas 2005,
614
Volume 4 : Nutrients (phospahte, nitrate, silicate). NOAA Atlas NESDIS 64. US Gov printing
615
office, Washington, D.C., 396 pp.
616
617
618
619
Hillebrand, H. 2004. On the generality of the latitudinal diversity gradient. Am. Nat.
163(2): 192–211.
Hutchinson, G.E. 1957. "Concluding remarks" . Cold Spring Harbor Symposia on
Quantitative Biology 22 (2): 415–427.
620
Hutchinson, G.E. 1961. The paradox of the plankton. Am. Nat. 95: 137–145.
621
Jeffrey, S. and G. Hallegraeff. 1980. Studies of phytoplankton species and photosynthetic
622
pigments in a warm core eddy of the East Australian Current. I. Summer populations. Mar.
623
Ecol. Prog. Ser. 3: 285–294.
624
Lehahn, Y., F. d'Ovidio, M. Lévy, Y. Amitai and E. Heifetz. 2011. Long range transport of
625
a quasi isolated chlorophyll patch by an Agulhas ring. Geoph. Res. Lett. 38:L16610,
626
doi:10.1029/2011GL048588.
627
Lévy, M., P. Klein, A.-M. Tréguier, D. Iovino, G. Madec, S. Masson, and K. Takahashi.
628
2010. Modifications of gyre circulation by sub-mesoscale physics. Ocean Modell. 34:1–15,
629
doi:10.1016/j.ocemod.2010.04.001.
630
Lévy, M., D. Iovino, L. Resplandy, P. Klein, A.-M. Tréguier, G. Madec, S. Masson, and K.
631
Takahashi. 2012a. Large-scale impacts of sub-mesoscale dynamics on phytoplankton: Local
632
and remote effects. Ocean Modell.43-44: 77–93, doi:10.1016/j.ocemod.2011.12.003.
23
633
Lévy, M. L. Resplandy, P. Klein, X. Capet, D. Iovino and C. Ethé. 2012b. Grid
634
degradation of submesoscale resolving ocean models: Benefits for offline passive tracer
635
transport. Ocean Modell. 48:1-9, doi :10.1016/j.ocemod.2012.02.004.
636
637
638
639
640
641
642
643
644
Madec, G. 2008. NEMO Ocean Engine, Note du Pole de modélisation de l’Institut Pierre
Simon Laplace, 27:1-217.
Marshall, J. C., C. Hill, L. Perelman and A. Adcroft. 1997. Hydrostatic, quasi-hydrostatic
and non-hydrostatic ocean modeling. J. Geophys. Res.102: 5733–5752.
Mouquet, N. and M. Loreau. 2003. Community patterns in source-sink metacommunities.
Am. Nat. 162:544–557.
d'Ovidio, F., S. De Monte, S. Alvain, Y. Dandonneau and M. Lévy. 2010. Fluid dynamical
niches of phytoplankton types. Proc. Nat. Acc. Sc.. doi: 10.1073/pnas.1004620107.
d’Ovido, F. S. De Monte, A. Della Penna, C. Cotté and C. Guinet. 2013. Ecological
645
implications of eddy retention in the open ocean: a Lagrangian approach, J. Phys. A. 254023.
646
Perruche C., P.Rivière, G. Lapeyre, P. Pondaven and X. Carton. 2010. Phytoplankton
647
competition and coexistence: Intrinsic ecosystem dynamics and impact of vertical
648
mixing. J. Mar. Syst. 81: 99-111.
649
650
Perruche C., P.Rivière, G. Lapeyre, X. Carton and P. Pondaven. 2011. Effects of SQG
turbulence on Phytoplankton competition and coexistence. J. Mar. Res. 69:105-135.
651
Ptacnik, R., Solimini, A.G., Andersen, T., Tamminen, T., Brettum, P., Lepisto, L., Willén,
652
and E. Rekolainen, S. 2008. Diversity predicts stability and resource use efficiency in natural
653
phytoplankton communities. Proc. Nat. Acc. Sc. 105: 5134–5138.
654
Prowe, A. E. F., M. Pahlow, S. Dutkiewicz, M. J. Follows and A. Oschlies. 2012. Top-
655
down control of marine phytoplankton diversity in a global ecosystem model. Progress
656
Ocean. 101: 1-13, doi:10.1016/j.pocean.2011.11.016.
657
Pulliam, H.R. 1988. Sources, sinks, and populations regulation. Am. Nat.. 132: 652-661.
24
658
659
660
Pulliam, H.R. 2000. On the relationships between niche and disturbance. Ecol. Lett. 3:
349-361.
Resplandy, L., A.P. Martin, F. Le Moigne, P. Martin, A. Aquilina, L. Mémery, M. Lévy
661
and R. Sanders. 2012. How does dynamical spatial variability impact 234Th-derived
662
estimates of organic export. Deep-Sea Res. I. 68: 24-45.
663
Richerson, P., R. Armstrong and C. R. Goldman. 1970. Contemporaneous disequilibrium,
664
a new hypothesis to explain the "paradox of the plankton", US PNAS 67(4) 1710-1714.
665
Stirling, G. and B. Wilsey. 2001. Empirical relationships between species richness,
666
667
668
669
670
671
672
evenness, and proportional diversity. Am. Nat. 158, 286–299.
Tilman, D. 1977. Resource competition between planktonic algae: An experimental and
theoretical approach. Ecology. 58: 338–348.
Tilman, D. 1982, Resource Competition and Community Structure, Monogr. Popul. Biol.
17: 296 pp. Princeton Univ. Press. Princeton, N. J.
Tilman, D., C. Lehman and K. Thomson. 1997. Plant diversity and ecosystem productivity:
theoretical considerations. Proc. Nat. Acc. Sc. 94: 1857–1861.
673
25
673
Table 1: Scales of ocean motion responsible for dispersal examined in this work.
Spatial scale
Temporal scale
Direction
Large scale circulation
100-1000 km Month to year
XY
Mesoscale turbulence
1-100 km
Week to month
XYZ
Vertical mixing
10-1000 m
Day to week
Z
674
675
Table 2: Characteristics of the offline model runs. Uhr, Khr and Thr are high-resolution
676
velocity, vertical mixing and temperature from the original model run. Ucg is Uhr coarse-
677
grained at 1° resolution and averaged annually. Tcg is Thr coarse-grained at 1° resolution and
678
averaged monthly. NO3cg is nitrate from the HR run, coarse-grained at 1° resolution and
679
averaged monthly. Ulr, Klr and Tlr are coarse-resolution velocity, vertical mixing and
680
temperature obtained from the same original model but run at coarse-resolution (1°). “Free
681
run” means that nitrate is let free to evolve after initialization and “forced run” means that
682
nitrate is restored to a prescribed value.
U
K
NO3
T
Free run
HR
Uhr
Khr
Free
Thr
Free run
LR
Ulr
Klr
Free
Tlr
Forced run
3D-e
Uhr
1.e-4
NO3cg
Tcg
Forced run
3D-m
Ucg
1.e-4
NO3cg
Tcg
Forced run
1D
None
1.e-4
NO3cg
Tcg
Forced run
0D
None
None
NO3cg
Tcg
683
684
26
684
Table 3: Change in community structure between pairs of experiments, ranging between 0
685
(identical community structure) and 1 (complete change of community structure) (see text for
686
more details).
687
0D
1D
3D-m
3D-e
0D
0.00
0.29
0.46
0.36
1D
0.29
0.00
0.28
0.17
3D-m
0.46
0.28
0.00
0.24
3D-e
0.36
0.17
0.24
0.00
27
687
Figure Legends
688
Figure 1. Model configuration. Annual mean sea surface temperature (SST) and eddy kinetic
689
energy (EKE) in the original high-resolution physical model compared with satellite
690
observations from the North Atlantic. The arrow on panel A shows the location of the mean
691
jet (Gulf Stream equivalent) that separates the two gyres. Satellite SST is from AVHRR
692
Oceans Pathfinder and EKE is derived from multi-satellite altimetric product Aviso
693
(www.aviso.oceanobs.com). Adapted from Resplandy et al. (2012).
694
695
Figure 2. Physical and chemical fields used to force the DARWIN simulations. A) High-
696
resolution surface current used to force experiments HR and 3D-e; Dec, 1st snapshot. B) High-
697
resolution vertical velocity at 50m used to force experiments HR and 3D-e; Dec, 1st snapshot.
698
C) Coarsed-grained surface current used to force experiment 3D-m. D) Coarsed-grained
699
surface nitrate distribution used to force experiments 3D-e, 3D-m, 1D and 0D; December
700
distribution. See text for more details on how these fields are constructed.
701
702
Figure 3. Model evaluation. A) Model annual mean surface chlorophyll (in HR run). B)
703
Satellite annual mean surface chlorophyll (SeaWifs). C) Model annual mean local richness (in
704
HR run). D) Proxy of phytoplankton diversity based on area-based mixing (De Monte et al.,
705
2013) of satellite optical anomalies (Alvain et al., 2008).
706
707
Figure 4. Change of α-diversity (black) and γ-diversity (grey) with dispersion. The α and γ-
708
diversities are measured as the domain integral of the annual local and regional richness,
709
respectively. Experiments 0D, 1D, 3D-m and 3D-e are ranked along the x-axis by increasing
710
level of dispersion in the flow field. Starting with no dispersion (0D), means for dispersion
711
are progressively accounted for: vertical mixing, advection by mean currents, eddy advection.
712
Results from the LR and HR experiments are marked with a dot and a star, respectively. Our
713
results show an increase in α-diversity with increasing means for dispersion, and a decrease in
714
γ-diversity.
715
716
Figure 5. Habitats of a sub-sample of six phytoplankton types (# 79, 65, 17, 62, 59,95) in our
717
set of experiments with increasing level of dispersion. For each type, the upper rectangle
28
718
shows the annual mean surface biomass over the x/y domain , the lower rectangle shows the
719
annual mean vertical distribution over the x/z domain (from 0 to 100m depth).
720
721
Figure 6. Area of each phytoplankton type, sorted in decreasing order along the x-axis. The
722
area is expressed in % of the total area and represents the fraction of the domain where the
723
type is present. The area associated with rank n is the area of the nth more spatially spread
724
type. The largest area (rank 0) occupies ~70% of the domain in 3D-e, but only ~35% in 0D.
725
Inversely, the smallest area in 3D-e is ~10% (rank 20), and 0.1% in 1D (rank 31) and 0D
726
(rank 43). Types occupying less than 5% of the domain (shaded in grey) are defined as “rare”.
727
728
Figure 7. Changes in community structure. Abundance of individual phytoplankton types, in
729
experiments 0D, 1D, 3D-m and 3D-e and sorted by their rank in experiment 3D-e.
730
731
Figure 8. Schematic representation of the changes in community structure and bio-geography
732
at high and low levels of dispersion. At low dispersion, there are a large number of types, over
733
small home range areas that are not overlapping: there is no co-existence. The different
734
habitats are organized along the temperature gradient. At high dispersion, the number of types
735
decreases, but the size of the home ranges increases and they intersect: different types co-
736
exist. Note that the types are not all the same between the high dispersion and low dispersion
737
cases. The different colors illustrate the change in community structure.
738
739
Figure 9. Changes of α-diversity with dispersion in the subtropical gyre, subpolar gyre and
740
jet area. Annual-mean, depth-integrated, zonally-averaged A) local richness and B) local
741
Shannon index, against meridional direction (in km) in experiments 0D, D, 3D-m and 3D-e.
742
743
Figure 10. Changes in community structure between pairs of experiments in the subtropical
744
gyre, subpolar gyre and jet area.
745
746
Figure 11. Results for the free runs LR (dashed lines) and HR (plain lines). A) Annual-mean,
747
depth-integrated, zonally-averaged local richness against meridional direction (in km). B)
748
Ranked area.
749
29
749
750
Figure 1
30
Figure 2
31
Figure 3
32
Figure 4
33
Figure 5
34
Figure 6
35
Figure 7
36
Figure 8
37
Figure 9
38
Figure 10
39
Figure 11
40