See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/299594496

A neutral-metabolic theory of latitudinal biodiversity

Article · March 2016

DOI: 10.1111/geb.12451

CITATIONS READS

34 531

2 authors:

Derek P Tittensor Boris Worm

Dalhousie University Dalhousie University
158 PUBLICATIONS 12,727 CITATIONS 216 PUBLICATIONS 39,023 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Global Hotspots of Transshipment of Fish Catch at Sea View project

Nereus Program View project

All content following this page was uploaded by Boris Worm on 10 November 2017.

The user has requested enhancement of the downloaded file.

Global Ecology and Biogeography, (Global Ecol. Biogeogr.) (2016)

RESEARCH A neutral-metabolic theory of latitudinal

PA P E R
biodiversity
Derek P. Tittensor1,2* and Boris Worm1

1
Biology Department, Dalhousie University,
Halifax, NS, B3H 4R2, Canada, 2Science
Programme, United Nations Environment
Programme World Conservation Monitoring
Centre, Cambridge CB3 0DL, UK
*Correspondence: Derek P. Tittensor, Biology
Department, Dalhousie University, Halifax,
NS, B3H 4R2, Canada. E-mail:
derek.tittensor@unep-wcmc.org
ABSTRACT
Aim Latitudinal gradients of species richness represent Earth’s first-order
biodiversity pattern. Most species groups display a near-monotonic decline in
richness from the equator to the poles, yet there exists little mechanistic theory
to derive such patterns from first principles. Here we integrate two key
advances – neutral theory and the metabolic theory of ecology – to reconstruct
global species richness gradients and test underlying causes.
Location Simulated global meta-community.
Methods We constructed a spatially explicit global meta-community with
constant per capita rates of disturbance, speciation and dispersal. No gradient
emerged in this neutral base model. Focusing on the oceans as a model system,
we added a water temperature gradient of 0–308C that independently affected
rates of community turnover and speciation based on established metabolic
scaling laws. We also added a gradient in habitat area that roughly parallels the
observed decrease in ocean area from tropical to polar waters.
Results Thermal effects on the community turnover rate caused a transient
latitudinal gradient that ultimately disappeared. Thermal effects on speciation
produced a dynamically stable but relatively weak gradient. Increasing habitat
area towards the equator in combination with thermal effects on speciation
rate produced a more realistic gradient that emerged from the combined effects
of species–area and species–energy theory.
Main conclusions This reasonably simple model provides a platform to
explore support for processes underpinning large-scale biodiversity gradients,
and the ability of prominent ecological theories, independently or in
combination, to capture them.
Keywords
Global model, latitudinal diversity gradient, macroecology, metabolic theory,
neutral theory.

INTRODUCTION
Biodiversity, here expressed as the number of species in a
given area, is distributed non-randomly across the globe.
Striking gradients of declining species richness from tropical
to polar areas have fascinated naturalists since the time of
von Humboldt and Darwin and comprise a prominent ‘first-
order’ biogeographical pattern that holds both on land and
in the oceans (Pianka, 1966; Rohde, 1992; Gaston, 2000; Hill-
ebrand, 2004; Kreft & Jetz, 2007; Krug et al., 2009; Tittensor
et al., 2010; Brown, 2014). Although not all species groups
show a monotonic decrease in richness with increasing lati-
tude (see, e.g., Rutherford et al., 1999; Worm et al., 2005),
the vast majority of taxa display peak richness in tropical or
subtropical climates (Hillebrand, 2004).
There is broad agreement that a near-ubiquitous pattern
like the latitudinal gradient in species richness probably has a
consistent mechanistic explanation that should apply across
different life histories and habitats (Pianka, 1966; Rohde,
1992). It has further been theorized that variation in the total

C 2016 John Wiley & Sons Ltd

V DOI: 10.1111/geb.12451
http://wileyonlinelibrary.com/journal/geb 1
D. P. Tittensor and B. Worm
number of species among regions should fundamentally
result from variation in one or more of three basic ecological
and evolutionary processes that ultimately control the diver-
sity of life, namely speciation rates, extinction rates and
immigration/colonization rates; these processes have been
studied by theoreticians (Hubbell, 2001), palaeontologists
(Jablonski et al., 2006) and evolutionary biologists (Dowle
et al., 2013) alike. Hence the question: what is the relative
importance of these processes in controlling observed gra-
dients, and how might spatial variation in environmental
parameters such as temperature, habitat area and productiv-
ity affect them?
In a classic review of the topic, Rohde (1992) systemati-
cally surveyed the internal logic and empirical evidence for
28 published hypotheses explaining the general latitudinal
gradient. He concluded that an ‘evolutionary speed hypothe-
sis’ might best explain latitudinal gradients, probably as the
result of shorter generation times, faster mutation rates and
faster selection at higher temperatures. Comprehensive evi-
dence now supports these mechanisms; specifically, higher
rates of molecular evolution (Gillooly et al., 2005) and speci-
ation (Jablonski et al., 2013) in the tropics, particularly for
ectotherms. More recent empirical analyses of globally well-
represented species groups have identified spatial gradients in
temperature and habitat area, as well as productivity, as pri-
mary environmental correlates of richness (Kreft & Jetz,
2007; Tittensor et al., 2010; Jetz & Fine, 2012), but the mech-
anisms by which such correlates may affect observed patterns
are still debated (Currie et al., 2004; Dowle et al., 2013).
Notably, no unifying body of theory has yet been devel-
oped to explain the relative importance of these mechanisms
in generating observed latitudinal gradients (Whittaker et al.,
2001). Yet significant advances have been made to capture
the effects of dispersal, speciation and extinction on local to
regional biodiversity patterns in a unified neutral model
(Bell, 2001; Hubbell, 2001; Rosindell et al., 2011). Independ-
ent of these efforts, metabolic theory offers a very general
explanation of the effects of temperature and body size on
biodiversity (Brown et al., 2004; Allen & Gillooly, 2006;
Stegen et al., 2009). In line with a recent proposal (Marquet
et al., 2014), we combined these two bodies of theory,
because together they represent prominent ecological and
evolutionary processes that might help explain latitudinal
gradients from fundamental principles.
As a basis we begin with Hubbell’s (2001) spatial null
model of ecological equivalence. Neutral theory has the sim-
plifying assumption of identical ecological traits across spe-
cies, and constant per capita rates of dispersal and speciation.
We then add latitudinal thermal gradients that explicitly
drive variation in community turnover rate and speciation
rate as predicted by the metabolic theory of ecology (Brown
et al., 2004). This model remains fully neutral at the local
community level, but features a metabolically induced gradi-
ent of (independently or in combination) community turn-
over and speciation rates across the global meta-community.
We also explore the effects of varying area (or area-specific
2 Global Ecology and Biogeography, V
productivity) by varying the number of individuals residing
within each community (Currie et al., 2004). Thus our aim
is to take a simple neutral model and add additional ecologi-
cal processes to gauge how adequately it can approximate
observed patterns in latitudinal richness, and to explore the
relative importance of hypothesized mechanisms in generat-
ing these patterns. We apply these mechanisms (independ-
ently or synergistically) to a hypothetical global ocean that
features temperature and area gradients of similar magnitude
to those observed in the real world. Our focus is on the
oceans as a model system because of strong empirical evi-
dence that high temperature and large habitat area consis-
tently correlate with high regional species richness across
species groups (Tittensor et al., 2010). In addition, the glob-
ally interconnected nature of the oceans and greater facility
for dispersal are more easily approximated using the simple
cylindrical model described below. Two computational meth-
ods are used to run the models (see Methods), one to exam-
ine transient dynamics (‘forward’ mode) and one to run
large numbers of simulations to examine dynamically stable
results (‘coalescence’ mode).
METHODS
Herein, we attempt to explore global spatial patterns of bio-
diversity as a function of fundamental ecological and evolu-
tionary processes, beginning with a spatially explicit
metacommunity model derived from the unified neutral
theory of biodiversity and biogeography (Hubbell, 2001). As
per Hubbell’s original theory, two basic principles inform the
model design. The first is that all processes (birth, death, dis-
persal and speciation) are modelled as per capita rates, such
that interactions and events have an identical probability for
each individual in a local community. Secondly, we also
assume zero-sum dynamics, meaning that there are a fixed
number of individuals competing for a finite pool of resour-
ces supported by a given habitat area. We note that this type
of model only applies to ecologically similar species within a
trophic level (e.g. trees, corals, planktivorous fish) and not
necessarily to entire assemblages and food webs (Hubbell,
2001).
Basic model
We begin with the basic model of a local community that is
composed of n ‘cells’ or ‘sites’. Each of these sites is occupied
by a single individual of a particular species. Multiple local
communities are linked into a global, spatially explicit meta-
community grid of size i 3 j (Fig. 1). This grid consists of
local communities arranged around a cylinder; that is, cells
are connected over the longitudinal meridian but not the
poles. Apart from the rows of cells at the highest and lowest
latitudes, which have fewer connections, each community is
connected to its eight nearest neighbours, the so-called ‘Moore
neighbourhood’ of cellular automata theory (Hubbell, 2001),
via dispersal.
C 2016 John Wiley & Sons Ltd
 Neutral-metabolic theory of latitudinal diversity

Global metacommunity Local community

Cold

Thermal gradient

Hot Large area

Small area

Cold

Figure 1 Model schematic. A spatially explicit global metacommunity is depicted. It consists of 21 3 21 local communities, each
composed of n individuals, where n is a function of habitat area or productivity. Different colours in local communities symbolize
different species. Communities are connected through dispersal (m) to neighbouring cells (arrows). Although the grid is depicted as flat,
the left-hand and right-hand borders are connected, forming a cylindrical shape. The dashed line in the centre represents the equator,
and the top and bottom of the main grid represent the poles. Disturbance and replacement of individuals on the grid is a per capita
function that is constant (in the fully neutral model) or a function of temperature (in the combined neutral-metabolic model) acting
through turnover rates (generation times) and/or speciation rates. The effect of increasing area (or productivity) can be simulated
through varying the number of individuals in any specific local community.

When a site gets disturbed, with all sites disturbed with
equal probability, the individual at that site is replaced at
random, either with a newly formed species (with probability
v) or, more commonly, via dispersal (with probability 1 – v).
If a dispersal event occurs, the dispersal is from a neighbour-
ing community (the Moore neighbourhood) with probability
m, or alternatively from within the local community with
probability 1 – m. The probability of a specific species immi-
grating from the local community and the Moore neighbour-
hood is proportional to the relative abundance of species in
these areas. Thus a species can only move to immediately
neighbouring communities but can ultimately spread through
the global community via successive Moore neighbourhoods
(Bell, 2001), with faster dispersal at higher values of m.
To enable comparison with Hubbell’s spatial metacom-
munity model, we set the number of cells in each local com-
munity, n, to be 16, and the number of communities in
longitudinal and latitudinal directions, i and j, to be 21. We
explore a range of values for the speciation probability, v,
ranging from 0.0001 to 0.1, and for regional dispersal, m,
from 0.001 to 1. Furthermore, we run an analysis in which
the dispersal parameter is set to either 0.005 or 0.5 and the
speciation rate to 0.000709, giving a fundamental biodiversity
number (h) of 10 for direct comparison with Hubbell (2001).
However, note that unlike Hubbell we do not embed the
21 3 21 grid of communities in a large plane to avoid edge
effects. We further examine the sensitivity of results to local
community size by changing the number of cells within each
local community (n 5 8, 16, 64 and 256 individuals; see Fig.
S1 in the Supporting Information).
Temperature-dependent mechanisms
After exploring the base model, we introduce two separate
mechanisms that may affect latitudinal species richness via a
thermal gradient, and can be applied independently or in
concert. Temperature is the only environmental variable that
has been consistently identified as important in terms of
marine biodiversity patterns (Tittensor et al., 2010), which
are the focus of our study. Several mechanisms have been
hypothesized as underlying this effect, and here we focus on
two: the effect of temperature on generation time or commu-
nity turnover rate, and its effect on speciation rates (Allen
et al., 2006). According to the metabolic theory of ecology,
biochemical reaction rates, metabolic rates and other rates of
biological activity increase exponentially as a function of the
internal temperature of an organism and body size. In the
context of our neutral model we can ignore the effect of
body mass, as all organisms are assumed to be ecological
equivalents within the same trophic level. The temperature
dependence is described by the Boltzmann factor:
e 2E=kT (1)
where E is the empirically derived activation energy (c. 0.65 eV
for respiratory-based processes in ectotherms; Allen et al.,

Global Ecology and Biogeography, V

C 2016 John Wiley & Sons Ltd 3
D. P. Tittensor and B. Worm
2006), k is the Boltzmann constant (8.617 3 1025 eV K21)
and T is the absolute internal temperature in Kelvin (K)
(Boltzmann 1872; Arrhenius 1889). Here we model this effect
for ectotherms by assuming a 308 gradient in average (surface)
water temperature, from 08C at the poles to 308C at the equa-
tor, values that approximate those empirically observed in the
modern surface oceans (Tittensor et al., 2010). At this stage,
we do not consider endotherms, and only note that their lati-
tudinal patterns may be different (Davies et al., 2010; Tittensor
et al., 2010) because their metabolic rate will not scale in the
same manner with external temperature. This gradient gives an
exponential curve with a 15-fold difference in metabolic rate
between the equator and the poles.
Thermally mediated community turnover rate model
To simulate the mechanism of a metabolic effect on genera-
tion time and community turnover rate, we make the rate of
disturbance and recolonization (and hence community turn-
over rate and its inverse, generation time) dependent on tem-
perature. When selecting an individual to be disturbed, we
set a disturbance probability p for each individual that
remains identical within each local community, but for each
row j of communities at a specific point along the thermal
gradient individuals have a disturbance probability scaled to
the metabolic rate at that temperature, normalized by the
total number of individuals in the model. That is, for an
individual h in community i, j, the probability of acting
(being disturbed) in any individual time-step is:
e 2E=kTj
ph 5 P 2E=kTj
: (2)
j i:n:e
Purely as a result of a relatively faster disturbance and turn-
over rate, and hence shorter generation times, we will also
observe more speciation events per unit time at lower lati-
tudes (Allen et al., 2006). This ‘community turnover rate’
model is still strictly a neutral model in that all individuals
are ecologically equivalent, with identical per capita rates in
an identical environment and identical responses to tempera-
ture gradients. Hence, a coalescence algorithm can still be
applied using the Markov chain approach (see the descrip-
tion below). However, individual metabolism and species
interactions speed up at higher temperatures, such that local
communities operate at different rates.
Thermally mediated speciation rate model
Metabolic theory invokes a second effect of temperature on spe-
ciation rates independent of turnover times. For ectotherms,
there is empirical evidence that mutation rates in terms of nucle-
otide substitutions increase as a function of temperature (Allen
et al., 2006). This may be explained by the increased production
of free oxygen radicals at higher metabolic rates, which increases
nucleotide substitution and ultimately speciation rates (Allen
et al., 2006). We can simulate thermal effects on speciation in
4 Global Ecology and Biogeography, V
our model by setting the speciation rate, v, to be a normalized
function of temperature:
vbase e 2E=kTj
vj 5   (3)
min e 2E=kTj
where vbase is the ‘basic’ speciation rate at the lowest temper-
ature, and the metabolic effect of temperature, e 2E=kTj is nor-
malized by dividing through by its minimum value. Thus at
the poles the speciation probability is vbase and it increases
exponentially towards the equator. This effect can be applied
independently of or in combination with the effect of tem-
perature on turnover rates.
Area model
The final mechanism that we examine relates to community
size. Some communities will support larger numbers of indi-
viduals, possibly because of increased habitat availability or
alternatively higher productivity; this is the ‘species–energy
hypothesis’ sensu (Wright, 1983), or the ‘more individuals
hypothesis’ sensu (Srivastava & Lawton, 1998). We can allow
local communities to have different numbers of individuals,
but keep all per capita rates within local communities
constant.
The probability of any individual h in community i, j
being disturbed is then:
1
ph 5 P : (4)
i;j ni;j
Hence larger communities with more individuals will see
more frequent disturbances overall.
For habitat area, we use a strictly proportional relationship,
whereby twice the habitat area supports twice the number of
individuals. In our exploratory model herein, we assume a lin-
ear five-fold increase in area from the poles to the equator, a
value roughly equivalent to the area increase in the global
oceans, averaged across the Northern and Southern Hemi-
spheres (see Fig. 3 in Allen et al., 2006). We note that
increases in productivity also produce increases in the number
of individuals, but for the sake of this exploration we do not
attempt to model both processes at the same time.
Combined models
We explore all two-factor combinations of the three mecha-
nisms described above, as well as the model with all three
effects included. Adding the effect of area in concert with the
latitudinal thermal effect on community turnover rate pro-
duces a model in which the probability of any individual h
in community i, j being disturbed is then:
e 2E=kTj
ph 5 P 2E=kTj
: (5)
i;j ni;j :e
We can include the thermally mediated effect on speciation rate
in a two- or three-factor model simply by applying equation (3).
C 2016 John Wiley & Sons Ltd

Computational methods
The model is implemented using two different algorithms or
modes: forward mode (to examine transient dynamics) and
coalescence mode (to run large numbers of simulations in a
computationally tractable manner). The forward mode (Hub-
bell, 2001) proceeds by disturbing one individual per time
step until a finite number of time steps or community turn-
overs have been completed. With this approach, the model
state at every time step can be analysed, allowing us to exam-
ine the development of spatial gradients over time. The grid
is initially seeded with a single species; we note that the tran-
sient dynamics in terms of the direction of the gradient are
dependent upon the initial number of species, but ultimately
settle to the same state at equilibrium (as per Hubbell, 2001).
As an alternative simulation approach, the coalescence algo-
rithm uses techniques originally developed to track neutral
evolutionary markers in population genetics (Felsenstein,
2004). These powerful algorithms have a number of important
advantages over traditional simulation techniques. Coalescence
begins with a present-day global community where each indi-
vidual is of unknown species identity and lineage, and then
works backwards in time, applying the rules of the model in
reverse and determining speciation and dispersal events that
can uniquely identify individuals as being derived from the
same lineage (Rosindell et al., 2008). This efficiently uncovers
the ancestry and phylogeny of contemporary individuals in
the sample, but without modification can only be applied to
strictly neutral models. Crucially, it provides a way to simulate
individual-based theory for large and complex spatial struc-
tures; no known alternative algorithm would be tractable in
these situations (J. Rosindell, pers. comm.; Rosindell et al.,
2008, 2011).This gain in speed occurs through the tracking of
individuals whose lineage has not yet been determined but not
those whose identities have been established, whereas the for-
ward model continues to track every individual during all
time steps. Coalescence also provides an objective end-point,
unlike the forward mode for which it can be difficult to deter-
mine that the model has fully ‘settled’. The end-point is
reached when the identity of each species has been established,
giving a complete picture of the present state based on the
ecological and evolutionary events that have occurred. It can-
not, however, be used to examine transient dynamics as only
the end-state community is produced. The rapid computation
time and stochastic nature of the coalescence algorithm allows
us to run numerous Monte Carlo simulations and determine
mean values and ranges. In this study, we run 100 simulations
for each parameter combination and present the mean and
the 2.5% and 97.5% percentiles.
For both the forward and coalescence modes, in contrast
to Hubbell (2001), we do not disturb a single cell in every
local community in each time step. Instead a randomly
selected site from the entire global community (i.e. all indi-
viduals in all communities) is chosen for disturbance proba-
bilistically, where the probability of a cell being selected is
identical for each site in the basic model but non-identical
Global Ecology and Biogeography, V
C 2016 John Wiley & Sons Ltd
Neutral-metabolic theory of latitudinal diversity
when the community turnover rate mechanism is operating.
In the long run, this ‘Markov chain’ approach settles to a
mean state identical to that produced by having a single
disturbance within every community every time-step (see
Fig. S2 for comparison with Hubbell, 2001), but with the
advantage that differential rates of disturbance and commu-
nity turnover under environmental gradients can be modelled
without complex timing structures being needed to ensure
that synchronization and action rates for individuals are
equivalent to the specified probabilities.
All code for the model was written in Processing 3.10a
(https://www.processing.org/) with analysis of results in R,
version 3.2.0 (R Core Team, 2015). Code is freely available
from https://github.com/derekt/Latitudinal-biodiversity-model.
RESULTS
Running the basic neutral models to a dynamic equilibrium
(25,000 complete community turnovers in forward mode,
all species identified in coalescence mode), produced rank-
abundance patterns that were statistically indistinguishable
from those in Chapter 7 of Hubbell (2001) (Fig. S2). These
results suggest that the basic algorithm successfully
implements the neutral model in both forward and coales-
cence modes; from herein, due to its computational tract-
ability, we use the coalescence mode to identify equilibrium
patterns and the forward mode to examine transient
dynamics.
When we implemented the basic coalescence model with-
out any thermal or area mechanisms, no latitudinal gradient
emerged at equilibrium (Fig. 2a,b). This effect was consistent
for the parameter combinations that we tested regardless of
speciation rate, dispersal parameter value or local community
size. The global community was thus made up of similarly
species-rich local communities, though of course with some
level of stochasticity.
Adding a metabolic effect on community turnover rate
(1/generation time) to the model via a latitudinal thermal
gradient produced similar results (Fig. 2c,d), with no gradi-
ent at dynamic equilibrium. However, when a metabolic
effect on speciation rate was applied, a shallow equilibrium
gradient emerged (Fig. 2e,f). The mean richness at the equa-
tor, 7.3 species, was around 2.1 times that at the poles (3.6
species), which had a near-identical value to that of the base
model (3.5 species). The mean gradient was near linear,
within the variability induced by the stochastic processes
within our neutral model.
Running the basic model but with a five-fold area gradient
resulted in larger community sizes towards the equator
and produced a stronger equilibrium latitudinal gradient
(Fig. 2g,h), with mean values at the equator almost 4.4 times
bigger than those at the poles.
When modelling the area and temperature mechanisms in
combination, their effects appeared multiplicative (Fig. 3).
Applying both the community turnover rate and speciation
rate gradients produced a near-identical equilibrium gradient
5
D. P. Tittensor and B. Worm

A B multiplicative (Fig. 3e,f), producing a stronger equilibrium

20
gradient with a sharper peak, and a mean equatorial richness
value (35.4 species) that was 9.2 times higher than the polar

15
Local richness
value (3.9). Applying all three mechanisms (Fig. 3g,h) pro-

Latitude
duced a dynamic equilibrium gradient that was identical to

10
the speciation rate and area model
We explored the sensitivity of community properties to var-

5
iation in speciation rate and dispersal parameters for model
0 20 40 runs with thermal gradient effects on both speciation and
Local richness
community turnover rates (Fig. 4). Mean local community
C D

20
species richness increased with both dispersal and speciation
rate (Fig. 4a). Mean global meta-community richness showed
15
Local richness

an even more pronounced increase with increasing speciation

Latitude
rate (Fig. 4b). The effect of the dispersal parameter on global
10

richness, however, was minimal, acting only to slightly increase

the total richness at low dispersal values. When assessing the
5

effect of both parameters on the strength of the latitudinal

biodiversity gradient (Fig. 4c), we saw an interaction between
0 20 40
Local richness the effect of speciation rate and dispersal. While increasing
E F speciation rate increased the latitudinal richness gradient
20

across most dispersal parameter values tested, the effect was

more pronounced at low than at high dispersal rates (Fig. 4c).
15
Local richness

Dispersal had a slight positive effect on gradient strength at

Latitude

low speciation rates, but a negative effect at the highest specia-

10

tion rates
Varying base local community size from 4, 16, 64 and 256
5

individuals across all communities, again with both thermally

0 20 40 mediated mechanisms included, produced similar gradients,
Local richness although at progressively higher mean richness, as more species
G H can coexist in larger communities (Fig. S1). We also observed a
20

moderate and asymptotic increase in the steepness of the latitu-

15

dinal gradient with increasing community size (mean differ-

Local richness

Latitude

ence between polar and equatorial richness 1.48, 1.98, 2.21 and
10

2.28 respectively). We further note that for more ‘realistic’

parameter values of low speciation and dispersal rates, the gra-
5

dient became more pronounced in larger communities.

0 20
Local richness
Figure 2 Emergence of latitudinal biodiversity patterns. The left-
hand column depicts a global grid of species richness. The right
hand column depicts a latitudinal gradient, with the solid line
indicating the mean and shading representing the 95% range. (a, b)
Neutral base model. (c, d) Including a thermal gradient effect on
community turnover rate. (e, f) Including a thermal gradient effect
on speciation rate. (g, h) Including an area effect. Dispersal
parameter m 5 0.1, speciation rate v 5 0.01, global metacommunity
size Jm 5 7056 individuals. Thermal effects are parameterized as per
the metabolic theory of ecology. Area and temperature gradients
approximate those found in the global oceans.
to that of the speciation rate gradient on its own (Fig. 3a,b).
A similar result occured when the community turnover rate
and area mechanisms were applied together (Fig. 3c,d). The
effects of combining speciation rate and area effects were
Finally, when we examined transient dynamics in the for-
40
ward model with only a thermal metabolic effect on relative
community turnover rate, this produced a temporary latitu-
dinal gradient (Fig. 5), in contrast to the equilibrium state
(Fig. 2c,d). This gradient persisted for around 150 complete
community turnovers, but then weakened and disappeared
by around 1000 community turnovers. The strength of this
gradient, peaking at around a 2.1-fold difference between
equator and poles, was roughly comparable to that resulting
from thermal metabolic effects on speciation. Adding both
thermal effects together (i.e. community turnover rate and
speciation rate) in the forward model resulted in a multipli-
cative effect that produced a strongly peaked, but again tran-
sient, gradient (Fig. S3). At dynamic equilibrium, the stable
gradient emerged, due to the thermal effect on speciation
rate. The transient gradient also appeared in the community
turnover rate and area model, as well as the full three-
mechanism model (area, community turnover rate, specia-
tion rate).

6 Global Ecology and Biogeography, V

C 2016 John Wiley & Sons Ltd
 Neutral-metabolic theory of latitudinal diversity

A B transient or dynamically stable, depending on the mecha-

20
nisms invoked. Adding an area gradient produces a stronger

15
latitudinal effect, particularly in combination with a thermal

Local richness
effect on speciation rate. These results support empirical syn-

Latitude
10
theses which, at least in the oceans, invoke temperature and
area as general environmental correlates of diversity (Titten-

5
sor et al., 2010). A key strength of this combined model is
that it incorporates processes operating in both evolutionary
0 20 40
and ecological time, and that it can be easily expanded to
Local richness
include the effects of other potential factors structuring bio-
C D
20
diversity such as productivity or changes in disturbance
15 regime, among others.
Local richness

Our basic spatial neutral model without a temperature or

Latitude
area effect does not produce a latitudinal gradient (Fig. 2a,b).
10

This is perhaps unsurprising, since the per capita ecological

equivalence and constancy of rates and local community sizes
5

give no reason for a spatial pattern to emerge (Hubbell,

0 20 40
2001). In this basic model, stochasticity explains any differen-
Local richness ces among local communities, with their properties governed
E F by the basic parameters of the neutral model, namely the bal-
20

ance between immigration, speciation and extinction.

When we combine neutral with metabolic theory and
15
Local richness

introduce a temperature gradient that approximates observed

Latitude

surface temperature gradients in the oceans, latitudinal pat-

10

terns emerge, although their strength and stability depends

on the invoked mechanism. With temperature-driven
5

increases in community turnover rate only, a transient gradi-

0 20 40 ent rapidly appears in the forward model (Fig. 5), but then
Local richness
flattens and collapses. This gradient does not persist in the
G H
20

results from a coalescence model (Fig. 2c,d), which represent

dynamic equilibrium conditions. This transience is due to a
15

metabolically driven shorter generation time in the tropics

Local richness

Latitude

that results in a faster approach towards the equilibrium

10

state, and furthermore a relatively higher chance of speciation

per unit time. If starting from lower than equilibrium rich-
5

ness conditions, such as a single species everywhere, the

0 20
Local richness
Figure 3 Combined effects of area and temperature gradients
on latitudinal biodiversity patterns. (a, b) Combined effect of
temperature-driven gradients in community turnover rate and
speciation rate. (c, d). Combined effect of area and temperature
effect on community turnover rate. (e, f) Synergistic effect of
area/productivity and speciation gradient. (g, h). Synergistic
effects of area and gradients in community turnover rate and
speciation rate. Symbols and base parameter settings as in Fig. 2.
Thermal effects are parameterized as per the metabolic theory of
ecology. Area and temperature gradients approximate those
found in the global oceans.
DISCUSSION
Our results show that a combination of neutral and meta-
bolic theories can generate a latitudinal gradient of bio-
diversity. The latitudinal gradients produced can be either
more rapid speciation will produce greater short-term equa-
40
torial richness, even if most new species do not persist for
long. Eventually, however, the polar regions will ‘catch up’
and reach the same asymptotic richness value as equatorial
areas (Fig. 2c,d). The transient gradient thus partly results
from the lack of any ‘stabilizing force’ to maintain an asymp-
totic gradient of equilibrium richness (such as an effect of
temperature on speciation rates), and partly from the initial
richness being lower than equilibrium richness. This latter
point can be demonstrated by starting a forward-mode
model with local richness values set to be higher than the
mean equilibrium value, in which case the more rapid
approach of equatorial regions towards the equilibrium value
produces a transient inverse latitudinal gradient (results not
shown). We note that it actually remains an open question as
to whether the Earth has reached an equilibrium biodiversity
value, a fact that is deemed unlikely given the continuing
increase of fossil biodiversity up to the present, as well as the
large-scale disturbances events that ‘reset’ biodiversity at

Global Ecology and Biogeography, V

C 2016 John Wiley & Sons Ltd 7
D. P. Tittensor and B. Worm

Local species richness Figure 4 Exploration of parameter space for a model including
16 thermal effects on both community turnover rate and speciation
rate. Community properties are shown as a function of variation
14 in dispersal and speciation rate parameters. (a) Effects on mean
local community species richness. (b) Effects on global meta-
12
community species richness. (c) Effects on the strength of the
15
latitudinal gradient, represented as the difference between mean
equatorial and mean polar richness. Shown are mean values of
10
10 100 model runs. Thermal effects are parameterized as per the
Richness

metabolic theory of ecology. Temperature gradients approximate

8
5
those found in the global oceans.

1 0.1

0.1 0.01
Di s
4 regional and sometimes global scales (Raup & Sepkoski,
sp
er 0.01 0.001 ate
sa
l ra ion
R 1982; Benton & Twitchett, 2003).
tes iat
ec
0.001 1e−04 Sp 2 A dynamically stable equilibrium biodiversity gradient is
produced by a thermally mediated latitudinal gradient in spe-
ciation rates. When the strength of this gradient is parame-
Global species richness terized according to metabolic theory (Allen et al., 2006) and
observed differences in surface ocean temperature, local rich-
ness at the equator is around twice the polar value. This is
4000
weaker than most empirical observations (Hillebrand, 2004;
Tittensor et al., 2010), suggesting that other processes need
to be invoked. It does, however, along with the transient gra-
4000 dient discussed above, support empirical results that suggest
3000
a thermal effect on biodiversity patterns in the oceans (Tit-
3000
Richness

tensor et al., 2010), and indeed on land (Rohde, 1992; Allen

2000 & Gillooly, 2006).
1000
2000 Our examination of the effects on the results of increasing
local community size for every community suggests that lati-
1 0.1 tudinal gradients appear regardless of the community size,
0.1 0.01 1000
with increasing gradient strength as local community size
Di
sp a tes increases, though this effect appears to be asymptotic. This
er 0.01 0.001
sa io nR
l ra iat
tes ec
suggests that relatively small community sizes, at least for
0.001 1e−04 Sp
some combinations of speciation and dispersal, might be rea-
0
sonable proxies for estimating the strength of gradients in
larger communities.
Diff. in equatorial and polar richness
The effects of changing dispersal and speciation parameters
12
on diversity properties were also examined (Fig. 4). Both
local and global species richness strongly increase with
10
increases in speciation rate (Fig. 4a,b). In contrast, increases
in dispersal rate do not strongly affect global richness (Fig.
4b) but enhance local richness by spreading new species to
10 8
other communities before they can go extinct, and maintain-
Delta richness

8
ing diversity in the face of stochastic dominance by abundant
6
6 species (Hubbell, 2001). The effect of dispersal and speciation
4 rates on the strength of the latitudinal gradient (the differ-
2 ence between polar and equatorial richness) is more nuanced
4
(Fig. 4c). In general, increased speciation leads to a steeper
1 0.1
gradient, due to the exponential character of the metabolic
0.1 0.01
Di
sp ate
s 2 relationship, which disproportionately increases richness at
er 0.01 0.001
sa nR
l ra
tes iat
io low latitudes as the speciation rate increases. At very high
ec
0.001
1e−04 Sp
levels of dispersal, however, increasing the speciation rate
0
produces no effect on the gradient strength, and in fact the
gradient is less steep than at lower dispersal rates. This is
probably due to the more rapid spread of species through

8 Global Ecology and Biogeography, V

C 2016 John Wiley & Sons Ltd
 Neutral-metabolic theory of latitudinal diversity

A B I J

10

10
8

8
Local richness

Local richness
Latitude

Latitude
6

6
4

4
2

2
0 5 10 0 5 10
Local richness Local richness

C D K L
10

10
8

8
Local richness

Local richness
Latitude

Latitude
6

6
4

4
2

2
0 5 10 0 5 10
Local richness Local richness

E F M N
10

10
8

8
Local richness

Local richness
Latitude

Latitude
6

6
4

4
2

2
0 5 10 0 5 10
Local richness Local richness

G H O P
10

10
8

8
Local richness

Local richness
Latitude

Latitude
6

6
4

4
2

0 5 10 0 5 10
Local richness Local richness

Figure 5 Emergence of a transient biodiversity gradient in a model including a thermal effect on community turnover rate. Shown is
the appearance of a biodiversity gradient in a forward-mode neutral model after (a, b) 1 community turnover (7056 time steps), (c, d)
10 community turnovers, (e, f) 50 community turnovers, (g, h) 100 community turnovers, (i, j) 125 community turnovers, (k, l) 250
community turnovers, (m, n) and its disappearance within 1000 community turnovers. (o, p): Coalescence model results are shown for
comparison. Symbols and base parameter settings are as in Fig. 2. Thermal effects are parameterized as per the metabolic theory of
ecology. Temperature gradients approximate those found in the global oceans.

the global meta-community, homogenizing the biota and
minimizing local community endemics. At low speciation
rates, dispersal had a minor positive effect on gradient
strength, probably by allowing newly formed species to
spread before they go locally extinct.
For a model including thermal effects on both the com-
munity turnover rate and speciation rate, the results at equi-
librium are identical to those of a model that only includes
thermal effects on speciation rate. This is consistent with the
transient nature of the turnover-driven gradient (Fig. 5). The
transient gradient produced, however, is very rapid and steep,
before declining to a dynamic equilibrium that consists of a
relatively weaker gradient (Fig. S3). We conclude that the
effects of increasing community turnover rate alone may not
be sufficient to explain the existence of general and persistent
latitudinal gradients across many species groups.

Global Ecology and Biogeography, V

C 2016 John Wiley & Sons Ltd 9
D. P. Tittensor and B. Worm
Increasing the area of local communities, and hence the
number of individuals therein, towards the equator also
produced a latitudinal diversity gradient (Fig. 2g,h). This gra-
dient did not change when combined with a temperature
effect on community turnover rate, but increased both the
height and the steepness of the gradient when combined
with a thermal effect on speciation rate (Fig. 3e,f). The com-
bined effect of these processes was multiplicative, rather than
additive, which explains the steepness of the resulting gradi-
ent. A combination of increasing area and increasing specia-
tion rate towards the tropics thus would appear to be most
consistent with the empirical richness data and their environ-
mental correlates (Kreft & Jetz, 2007; Tittensor et al., 2010;
Jetz & Fine, 2012), as well as well-established theory regard-
ing the general effects of increasing habitat area (MacArthur
& Wilson, 1967; Hubbell, 2001) and temperature (Brown
et al., 2004; Allen & Gillooly, 2006). In reality, of course,
thermal effects on community turnover rate and speciation
rate are both likely to occur, but including the effect on com-
munity turnover rate does not change the dynamically stable
gradient. The relative strength of the gradient (a 9.2-fold
increase in species richness using realistic area and tempera-
ture changes from the poles to the equator) is within the
general range of observed latitudinal gradients in richness of
marine oceanic (about a five-fold increase) and coastal exo-
therms (about a 10-fold increase, on average; Tittensor et al.,
2010).
An obvious next step would be to model these processes
more closely in direct comparison with empirical data. The
relative importance of temperature, area and productivity
(which will also increase the number of individuals in a local
community) in driving empirically observed gradients could
in principle be assessed by fixing those to observed values
and determining which process, or combination of processes,
produces a gradient that most closely reflects reality. In order
to do this, however, a number of other problems must be
confronted. It is not clear, for example, how community size
actually scales with productivity, on average. Moreover, how
large should the individual local communities be before they
can be considered a reasonable approximation of real sys-
tems? And how would real dispersal constraints such as con-
tinental boundaries and current patterns be implemented?
How would multiple trophic levels be modelled? What
should the dispersal and speciation rate parameters be set to?
These questions represent rich opportunities for making
theory conform more closely to empirical realities. We fur-
ther note that not all empirical patterns follow the canonical
decrease in richness from the equator towards the poles.
Some species groups peak at higher latitudes, which could be
related, for example, to varying gradients of temperature
(Tittensor et al., 2010) and productivity (Davies et al., 2010),
or the effects of endothermy (Tittensor et al., 2010) or niche
overlap (Beaugrand et al., 2013) that are not captured by our
simple models.
Although we present here the first combined application of
neutral and metabolic theory to global biodiversity, other
10
approaches can produce similar patterns, though these have
tended to be applied non-spatially or at smaller scales. For
example, Stegen et al. (2009, 2012) present an alternative
approach, where they generalize basic metabolic theory by
considering a range of possible metabolic scaling relationships
for species characterized by different body sizes, and by
including consumer–resource dynamics. Their model is non-
spatial, as only the general shape of the temperature–richness
relationship is considered. It is based on niches and adaptation
rather than neutral theory, with the evolving size-structured
trophic networks being considerably more complex than our
neutral communities and requiring more parameters to be
specified. Unlike the monotonically increasing gradients pro-
duced by our model, Stegen et al. (2009, 2012) produce a
range of plausible diversity gradients (positive, negative, unim-
odal or asymptotic with increasing temperature), depending
on the specified thermal dependences of basal resource supply,
consumer vital rates and mutation rates. However, these gra-
dients tend to diminish over time, in contrast to the stable
gradients that can emerge from our models.
The strength and persistence of the latitudinal gradient in
our model depends on the relative balance of speciation,
extinction and dispersal, and thus the parameters from the
basic neutral model. We note that with this model we are
not aiming for ‘realism’ nor a direct comparison with data or
a comprehensive evaluation of the effects of the various
parameters on the results. Rather, we view this as an explora-
tory study that simply examines whether indeed a latitudinal
gradient can be produced by integrating two prominent eco-
logical theories. The ability of the model to produce a latitu-
dinal biodiversity pattern must be interpreted in the light of
its assumptions and inherent limitations. These include the
often-criticized, biologically unrealistic assumption of ecolog-
ical equivalence. Here, we subscribe to the view that the use-
fulness of a theory lies in its ability to produce observed
patterns, and in its relative simplicity and generality rather
than specific complexity (Marquet et al., 2014). This useful-
ness lies in presenting a very simple base theory of ecological
and evolutionary processes which can then be extended to
include more biological realism. We extended it first by
including metabolic theory that scales community turnover
and speciation rates to ambient temperature. The tempera-
ture gradient we assumed is very simplistic and does not
reflect patterns of variability (e.g. seasonal or longitudinal
variability) observed in the real world. It is also not reflective
of past environmental conditions which may have been influ-
ential in shaping the evolutionary trajectories of species we
see today. Likewise, we introduced a crude area effect into
the model that is based on the average increase in total area
observed across all oceans; clearly this would be different, for
example, for coastal species where area would scale with the
available coastline habitat and on land where different rela-
tionships between total area and latitude exist. Another unre-
alistic assumption concerns the absence of dispersal barriers,
such that all species are assumed to be able to disperse and
survive everywhere. Clearly some form of niche conservatism
Global Ecology and Biogeography, V
C 2016 John Wiley & Sons Ltd

(Wiens & Graham, 2005) would be more realistic, and would
likely enhance the observed gradient by restricting dispersal
of newly evolved species to specific latitudes (Beaugrand
et al., 2013), which feature different habitats and temperature
regimes. Perhaps some combination of the neutral and meta-
bolic theories shown here and aspects of niche theory might
prove promising for future unification of latitudinal
gradient-generating mechanisms.
Spatial neutral models have previously been used to
explore the relative distribution of abundance, the the rela-
tion between range and abundance, the species–area law and
the turnover of species composition (Bell, 2001; Hubbell,
2001; Rosindell et al., 2011). However, to the best of our
knowledge, they have not yet been applied to understand
and approximate biogeographical patterns at the largest scale.
Our results suggest that a stable latitudinal gradient from the
equator to the poles can be generated in a fully neutral
model by at least two mechanisms, namely incorporating
metabolic theory through the effect of thermal gradients on
biological rates or through accounting for increasing habitat
area (or productivity) towards the equator. While not
attempting to be ‘realistic’ in a strict sense, this combined
theory does mechanistically produce testable hypotheses
about the strength and direction of global diversity gradients,
an important component of any model or theory (Marquet
et al., 2014). We hope that this stimulates further work
towards combining these and other prominent ecological the-
ories via simple models. A next step could be to integrate
aspects of niche theory to model actual barriers to dispersal
that keep species from freely spreading to habitats they did
not evolve in, in an attempt to more precisely delineate the
relative importance of niche and neutral processes in generat-
ing observed latitudinal gradients and other prominent mac-
roecological patterns.
ACKNOWLEDGEMENTS
We gratefully acknowledge helpful comments and advice from
J. Rosindell, particularly with regard to the implementation of
the coalescence model. Financial support was provided by The
National Science and Engineering Research Council of Canada.
All code used to run the models is available at the Github reposi-
tory of Derek Tittensor (https://github.com/derekt).
REFERENCES
Allen, A.P. & Gillooly, J.F. (2006) Assessing latitudinal gra-
dients in speciation rates and biodiversity at the global
scale. Ecology Letters, 9, 947–954.
Allen, A.P., Gillooly, J.F., Savage, V.M. & Brown, J.H. (2006)
Kinetic effects of temperature on rates of genetic diver-
gence and speciation. Proceedings of the National Academy
of Sciences USA, 103, 9130–9135.
Beaugrand, G., Rombouts, I. & Kirby, R.R. (2013) Towards
an understanding of the pattern of biodiversity in the
oceans. Global Ecology and Biogeography, 22, 440–449.
Bell, G. (2001) Neutral macroecology. Science, 293, 2413–2418.
Global Ecology and Biogeography, V
C 2016 John Wiley & Sons Ltd
Neutral-metabolic theory of latitudinal diversity
Benton, M.J. & Twitchett, R.J. (2003) How to kill (almost)
all life: the end-Permian extinction event. Trends in Ecology
and Evolution, 18, 358–365.
Brown, J.H. (2014) Why are there so many species in the
tropics? Journal of Biogeography, 41, 8–22.
Brown, J.H., Gillooly, J.F., Allen, A.P., Savage, V.M. & West,
G.B. (2004) Toward a metabolic theory of ecology. Ecology,
85, 1771–1789.
Currie, D.J., Mittelbach, G.G., Cornell, H.V., Field, R., Guegan,
J.-F., Hawkins, B.A., Kaufman, D.M., Kerr, J.T., Oberdorff,
T., O’Brien, E. & Turner, J.R.G. (2004) Predictions and tests
of climate-based hypotheses of broad-scale variation in taxo-
nomic richness. Ecology Letters, 7, 1121–1134.
Davies, R.G., Irlich, U.M., Chown, S.L. & Gaston, K.J. (2010)
Ambient, productive and wind energy, and ocean extent
predict global species richness of procellariiform seabirds.
Global Ecology and Biogeography, 19, 98–110.
Dowle, E., Morgan-Richards, M. & Trewick, S. (2013) Molec-
ular evolution and the latitudinal biodiversity gradient.
Heredity, 110, 501–510.
Felsenstein, J. (2004) Inferring phylogenies. Sinauer Associates,
Sunderland, MA.
Gaston, K.J. (2000) Global patterns in biodiversity. Nature,
405, 220–227.
Gillooly, J.F., Allen, A.P., West, G.B. & Brown, J.H. (2005)
The rate of DNA evolution: effects of body size and tem-
perature on the molecular clock. Proceedings of the
National Academy of Sciences USA, 102, 140–145.
Hillebrand, H. (2004) On the generality of the latitudinal
diversity gradient. The American Naturalist, 163, 192–211.
Hubbell, S.P. (2001) The unified neutral theory of biodiversity
and biogeography. Princeton University Press, Princeton, NJ.
Jablonski, D., Roy, K. & Valentine, J.W. (2006) Out of the
tropics: evolutionary dynamics of the latitudinal diversity
gradient. Science, 314, 102–106.
Jablonski, D., Belanger, C.L., Berke, S.K., Huang, S., Krug,
A.Z., Roy, K., Tomasovych, A. & Valentine, J.W. (2013)
Out of the tropics, but how? Fossils, bridge species, and
thermal ranges in the dynamics of the marine latitudinal
diversity gradient. Proceedings of the National Academy of
Sciences USA, 110, 10487–10494.
Jetz, W. & Fine, P.V.A. (2012) Global gradients in vertebrate
diversity predicted by historical area–productivity dynamics
and contemporary environment. PLoS Biology, 10,
e1001292.
Kreft, H. & Jetz, W. (2007) Global patterns and determinants
of vascular plant diversity. Proceedings of the National
Academy of Sciences USA, 104, 5925–5930.
Krug, A.Z., Jablonski, D., Valentine, J.W. & Roy, K. (2009)
Generation of Earth’s first-order biodiversity pattern.
Astrobiology, 9, 113–124.
MacArthur, R.H. & Wilson, E.O. (1967) The theory of island
biography. Princeton University Press, Princeton, NJ.
Marquet, P.A., Allen, A.P., Brown, J.H., Dunne, J.A., Enquist,
B.J., Gillooly, J.F., Gowaty, P.A., Green, J.L., Harte, J.,
Hubbell, S.P., O’Dwyer, J., Okie, J.G., Ostling, A., Ritchie,
11
 D. P. Tittensor and B. Worm
M., Storch, D. & West, G.B. (2014) On theory in ecology.
BioScience, 64, 701–710.
Pianka, E.R. (1966) Latitudinal gradients in species diversity:
a review of concepts. The American Naturalist, 100, 33–46.
R Core Team (2015) R: a language and environment for statis-
tical computing. R Foundation for Statistical Computing,
Vienna, Austria. Available at: http://www.R-project.org/.
Raup, D.M. & Sepkoski, J.J.J. (1982) Mass extinctions in the
marine fossil record. Science, 215, 1501–1503.
Rohde, K. (1992) Latitudinal gradients in species-diversity –
the search for the primary cause. Oikos, 65, 514–527.
Rosindell J., Wong Y. & Etienne R.S. (2008) A coalescence
approach to spatial neutral ecology. Ecological Informatics,
3, 259–271.
Rosindell, J., Hubbell, S.P. & Etienne, R.S. (2011) The unified
neutral theory of biodiversity and biogeography at age ten.
Trends in Ecology and Evolution, 26, 340–348.
Rutherford, S., D’Hondt, S. & Prell, W. (1999) Environmen-
tal controls on the geographic distribution of zooplankton
diversity. Nature, 400, 749–753.
Srivastava, D.S. & Lawton, J.H. (1998) Why more productive
sites have more species: an experimental test of theory
using tree-hole communities. The American Naturalist, 152,
510–529.
Stegen, J.C., Enquist, B.J. & Ferriere, R. (2009) Advancing
the metabolic theory of biodiversity. Ecology Letters, 12,
1001–1015.
Stegen, J.C., Ferriere, R. & Enquist, B.J. (2012) Evolving eco-
logical networks and the emergence of biodiversity patterns
across temperature gradients. Proceedings of the Royal Soci-
ety B: Biological Sciences, 279, 1051–1060.
Tittensor, D.P., Mora, C., Jetz, W., Lotze, H.K., Ricard, D.,
Berghe, E.V. & Worm, B. (2010) Global patterns and pre-
dictors of marine biodiversity across taxa. Nature, 466,
1098–1101.
Whittaker, R.J., Willis, K.J. & Field, R. (2001) Scale and spe-
cies richness: towards a general, hierarchical theory of spe-
cies diversity. Journal of Biogeography, 28, 453–470.
12
View publication stats
Wiens, J.J. & Graham, C.H. (2005) Niche conservatism: inte-
grating evolution, ecology, and conservation biology. Annual
Review of Ecology, Evolution, and Systematics, 36, 519–539.
Worm, B., Sandow, M., Oschlies, A., Lotze, H.K. & Myers,
R.A. (2005) Global patterns of predator diversity in the
open oceans. Science, 309, 1365–1369.
Wright, D.H. (1983) Species-energy theory: an extension of
species–area theory. Oikos, 41, 496–506.
SUPPORTING INFORMATION
Additional Supporting Information may be found in the
online version of this article:
Figure S1. Rank–abundance plots for comparison to
Hubbell’s original meta-community model.
Figure S2. Effects of local community size on latitudinal
gradient strength.
Figure S3. Emergence of a partially transient gradient in
biodiversity for a model with thermal effects on both
turnover and speciation (contrast with Fig. 5).
BIOS KE TC H
Derek Tittensor is a marine biologist with interests
including macroecology, biodiversity and human
impacts upon the oceans. He is senior marine biodi-
versity scientist at the United Nations Environment
Programme World Conservation Monitoring Centre
and an adjunct professor in the Department of Biology
at Dalhousie University
Boris Worm is professor for marine conservation biol-
ogy at Dalhousie University, Halifax, Canada. His
interests focus on the causes and consequences of
global biodiversity patterns, both in space and time.
Editor: Brian Enquist
Global Ecology and Biogeography, V
C 2016 John Wiley & Sons Ltd