Climate change alters stability and species ... - Wiley Online Library

Received: 6 March 2017

|

Revised: 25 August 2017

|

Accepted: 26 August 2017

DOI: 10.1111/gcb.13891

PRIMARY RESEARCH ARTICLE

Climate change alters stability and species potential interactions in a large marine ecosystem Gary P. Griffith1,2

| Peter G. Strutton1,3 | Jayson M. Semmens1

1

Institute for Marine and Antarctic Studies, University of Tasmania, Hobart, Australia 2

Norwegian Polar Institute, Tromsø, Norway 3

Australian Research Council Centre of Excellence for Climate System Science, University of Tasmania, Hobart, Tasmania, Australia Correspondence Gary P. Griffith, Institute for Marine and Antarctic Studies, University of Tasmania, Hobart, Australia. Email: [email protected]

Abstract We have little empirical evidence of how large-scale overlaps between large numbers of marine species may have altered in response to human impacts. Here, we synthesized all available distribution data (>1 million records) since 1992 for 61 species of the East Australian marine ecosystem, a global hot spot of ocean warming and continuing fisheries exploitation. Using a novel approach, we constructed networks of the annual changes in geographical overlaps between species. Using indices of changes in species overlap, we quantified changes in the ecosystem stability, species robustness, species sensitivity and structural keystone species. We then compared the species overlap indices with environmental and fisheries data to identify potential factors leading to the changes in distributional overlaps between species. We found that the structure of the ecosystem has changed with a decrease in asymmetrical geographical overlaps between species. This suggests that the ecosystem has become less stable and potentially more susceptible to environmental perturbations. Most species have shown a decrease in overlaps with other species. The greatest decrease in species overlap robustness and sensitivity to the loss of other species has occurred in the pelagic community. Some demersal species have become more robust and less sensitive. Pelagic structural keystone species, predominately the tunas and billfish, have been replaced by demersal fish species. The changes in species overlap were strongly correlated with regional oceanographic changes, in particular increasing ocean warming and the southward transport of warmer and saltier water with the East Australian Current, but less correlated with fisheries catch. Our study illustrates how large-scale multispecies distribution changes can help identify structural changes in marine ecosystems associated with climate change. KEYWORDS

biodiversity, climate change, fisheries, marine conservation, marine ecosystems, ocean warming, species interactions

1 | INTRODUCTION

species in the rate of distribution change may substantially re-organize regional marine ecosystem functioning, potentially triggering

Recent analyses of marine species’ response to climate change,

regime changes (Ling & Johnson, 2009).

based on the leading or trailing edge of species distributions, show

New indices, such as the velocity of temperature change, have

that distribution shifts of many species are consistent with ocean

shown the speed and direction of climate change (Burrows et al.,

temperature changes (Poloczanska et al., 2013). Differences between

2011). However, we have little empirical evidence or understanding

e90

|

© 2017 John Wiley & Sons Ltd

wileyonlinelibrary.com/journal/gcb

Glob Change Biol. 2018;24:e90–e100.

GRIFFITH

|

ET AL.

e91

of how complex large-scale distributional changes between species

The East Australian marine ecosystem is an ideal location to

may be impacted (Berlow et al., 2004). One way forward is to con-

investigate this approach. The Australian Commonwealth and Scien-

sider the distributional patterns between species, rather than the

tific Industrial Research Organisation (CSIRO) has collated extensive

leading or trailing edge of the distribution of one or several species.

distribution data (>1 million catch records) since 1992 on 180 dem-

This approach, by considering greater complexity, can simplify rather

ersal and pelagic species. Oceanographic analysis has shown the

than complicate our understanding of the consequences of environ-

region to be a global warming hot spot (Suthers et al., 2011). Ocean

mental changes on large-scale multispecies distribution (Berlow

warming in the region has been a consequence of the strengthening

et al., 2009; Griffith & Fulton, 2014).

and poleward extension of the East Australian Current (EAC), the

Using a network approach, the strength of potential interactions

major western boundary current of the South-Pacific sub-tropical

between species can be quantified as the proportion of their geo-

gyre (Ridgeway, 2007). The EAC is predicted to further strengthen


jo et al., 2011). In the congraphical distribution that overlaps (Arau

and warm significantly into the future (Cai, Shi, Cowan, Bi, & Ribbe,

text of climate change, while species may not have a direct biotic

2005).

interaction but geographically overlap, they may have a common

Here, we use changes in the distributional overlap between spe-

response to an environmental driver such as ocean warming due to

cies to show how the structure of the East Australian marine ecosys-

preferred temperature ranges that may alter the proportion of over-

tem has changed since 1992. First, we track changes in the overall


jo et al., 2011; Pollock et al., 2014; Wilson lap between species (Arau

robustness and stability of the ecosystem. Second, we quantify

et al., 2016).

changes in species contributions to robustness and changes in the

Studies show that the distributional patterns between co-occur-

sensitivity of species to the loss of interactions with other species.

ring species are generally asymmetrical (Bascompte, Jordano, & Ole-

Third, we determine changes in structural keystone species. The

sen, 2006). That is, species A may potentially depend strongly on

changes in species overlaps are then correlated with changes in fish-

species B, whereas species B may only potentially depend weakly

eries management and regional oceanography.

on species A. In the context of climate change, asymmetrical distributions have two important consequences. First, species A would need to track the distribution of species B with environmental changes, but species B would not need to track species A so
jo, Rozenfeld, Rahbek, & Marquet, 2011; Salomon, strongly (Arau

2 | MATERIALS AND METHODS 2.1 | Species data and study area

Connolly, & Bode, 2010). Second, analyses of complex networks

Species distribution data for the study area were extracted from the

show that asymmetrical distributions have a few key species with a

CSIRO data set that records the location of all species caught by

high number of interactions with others species, while the majority

commercial fisheries and scientific cruises from 1992 to 2011. This

of species have a limited number of interactions (Dunne, Williams,

data set represents the most extensive and complete information on

& Martinez, 2002). This type of network is considered robust to the

fish species distributions in the region. A data set of 180 species (>1

random loss of species from environmental change as most species

million records) was extracted to encompass the East Australian mar-

are poorly connected. However, they are very sensitive to the loss

ine system (20°S to 46°S, 145°W to 160°W; Fig. S1). This region

of key species. In contrast, symmetrical distributions imply that spe-

was chosen as the study area for three reasons. First, it represents

cies depend equally on one another and are more likely to respond

the most densely observed and most complete species occurrence

equally, and move as a cohort under environmental pressure. Con-

data in the southern hemisphere. Second, it is the subject of on-

sequently, a greater proportion of species are highly connected with

going extensive marine research and encompasses significant com-

other species and are more susceptible and less robust to the ran-

mercial fisheries (Fulton, Smith, & Smith, 2007). Third, it encom-

dom removal of species, but more robust to the loss of key species

passes the rapidly changing East Australian Current (EAC).


jo et al., 2011). (Arau The proportion of the geographical overlaps by all species with an individual species may be interpreted as a measure of each spe-

2.2 | Spatial co-occurrence

cies’ contribution to the robustness of the ecosystem. This approach

Our underlying concepts are (i) the potential interaction intensity

is a quantitative extension of species degree, defined as the number

between co-occurring pairs of species A and B can be quantified by

of species that a target species interacts with (Bascompte et al.,

calculating the proportion of the geographical area of species A that

2006). Conversely, the proportion of an individual species’ geograph-

overlaps with species B and vice versa; and (ii) that the degree of

ical range that overlaps with all species can indicate sensitivity to

overlapping indicates the degree of symmetry between the species

the loss of interactions with other species. Spatially, this type of net-


jo et al., 2011). If an equal area of species A lies within species (Arau

work analysis can also suggest the direction and extent of changes

B and vice versa, then the degree of overlap between the two spe-

in distributional overlap between many species. A quantitative analy-

cies will be symmetrical and the potential interaction intensity

sis of the changes in species overlaps could then be used to also

between species will be equal. If more of B lies within the area of A

investigate if these changes can be associated with environmental

than the area of A lies within B, the link will be asymmetrical

factors.

(Fig. S2). The degree of symmetry Sim(A,B) between the two species

|

e92

GRIFFITH

ET AL.

ecological index on the assumption that the greater the degree of a

can be calculated as: SimðA; BÞ ¼

species, the greater the species contribution to ecosystem robust-

j OðA; BÞ  OðB; AÞ j max½OðA; BÞ; OðB; AÞ

(1)

ness (Dunne et al., 2002). However, species degree assumes a simple

O(A,B) defines the degree of overlapping a ⋂ b/a as the propor-

realistic asymmetrical relationships between species and quantifies

tion of a (Species A’s area) that overlaps b (Species B’s area). O(B,A)


jo et al., 2011). any changes in the geographical asymmetry (Arau

defines the degree of overlapping b ⋂ a/a as the proportion of b

Here, we refer to Sin as species overlap robustness. That is, Sin is the

(Species B’s area) that overlaps a (Species A’s area). The denominator

contribution of a species to the robustness of the network of geo-

is the maximum overlap of either O(A,B) or O(B,A). Note that while

graphical overlaps between all species.

symmetrical relationship between species. Sin accounts for the more

this quantifies the potential interaction intensity between Species A and Species B and vice versa, it does not indicate the direction of the potential interaction as Sim(A,B) = Sim(B,A). When Sim(A,B) = 0,

2.4 | Species sensitivity

the link is completely symmetric. When Sim(A,B) = 1, the link is com-

We define an additional measure of the sensitivity of an individual

pletely asymmetrical (Fig. S2). Matrices of the degree of symmetry

species to the loss of potential interactions with other species, called

between pairs of species were constructed for each year and species

species overlap sensitivity Sout, calculated as the sum of the propor-

to form a matrix structure of 180 species 9 180 species 9 20 lay-

tion of the geographic distribution of species A that overlaps with

ers, where the layers are years.

B1. . .n as:

Tests for statistical significance with patterns of co-occurrence Sout ðAÞ ¼

are not universally agreed on and remain a huge area of statistical research

(Veech,

2013).

Considering

this,

we

applied

OðA; Bi Þ

(3)

i

three

approaches to test for overlaps that may be due to chance alone.

n X

A high Sout implies that the species is potentially highly dependent

First, borrowing from graph theory, we computed correlation statis-


jo on other species, whereas a low Sout implies that it is not (Arau

tics (Pearson correlation, Jaccard coefficient, Goodman Kruskal

et al., 2011).

Gamma and Hamming distance) between cells (rows and columns) of each matrix (Borgatti, Everett, & Johnson, 2013). Second, we computed correlations between each yearly matrix using QAP correlation (Borgatti et al., 2013). Third, we applied a probabilistic model that computes the probabilities that two species would co-occur more or

2.5 | Structural keystone index of species interaction robustness Structural keystone species can be determined by examining the fre-

less frequently (or greater) than they actually do (Griffith, Veech, &

quency distribution P(k) of the species degree (ki): that is, the fre-

Marsh, 2016; Veech, 2013). The model is strictly analytical, is distri-

quency distribution of the number of species that a given species

bution-free and requires no randomization. Previous work with this

interacts with (Dunne et al., 2002). Typically, P(k) follows an expo-

model has shown it to have very low Type 1 and low Type II error

nential or power-law (scale-free) distribution. In this type of distribu-

rates (Veech, 2013). In all, 61 species of the original 180 species

tion, most species have few connections with other species while a

showed a strong relationship not due to chance and were used in

few species have many connections to other species. The well-con-

the analysis (Table S1).

nected species can be considered structural keystone species (Dunne et al., 2002) as opposed to functional keystone species (Jord
an, Okey, Bauer, & Libralato, 2008). In scale-free networks, there are a

2.3 | Species robustness

few well-connected keystone species with high ki values, while most If we assume that the degree of overlap between species indicates

species are poorly connected. Scale-free frequency distributions

the degree to which the species may potentially interact, we can

imply that networks are very robust to the random loss of poorly

define a measure of the potential interaction strength between all

connected species (those at the low ki tail), but very sensitive to the


jo et al., 2011). We can other species and a target species (Arau

loss of high ki keystone species (Dunne et al., 2002). In contrast,

quantify the normalized sum of the proportion of the geographical

networks with a greater proportion of highly connected species are

distribution of all other species B1. . .n that overlap with target species

more sensitive to the random removal of species. Here, the concept

A as:

is extended to examine the frequency distribution of Sin (species Sin ðAÞ ¼

n X

overlap robustness), that is expected to generally have the same OðBi ; AÞ

(2)

i

properties as P(k) (Barret, Barthelemy, Pastor-Satorras, & Vespignani, 2004). Potential power-law distributions were tested using maximum

where Sin (A) is a measure of how strongly other species may poten-

likelihood fitting methods (Clauset, Shalizi, & Newman, 2009; http://

tially interact with A.

www.santafe.edu/~aaronc/powerlaws/). One-sample Kolmogorov–

Sin (A) is a quantitative extension of species degree (ki), which is

Smirnov goodness-of-fit test was applied to each distribution. Differ-

the number of links that connect a target species with all other spe-

ences between distributions were tested with the two-sample

cies (Bascompte et al., 2006). Species degree is considered a useful

Kolmogorov–Smirnov test (Matlab R2017a).

GRIFFITH

|

ET AL.

2.6 | Caveats

e93

examined using scatter plots. Data were then detrended by subtracting the mean from each time series. Time series were compared

There are a number of important caveats to our assumptions and

using cross-correlation analysis with lags (Matlab R2017a, Signal Pro-

approach:

cessing Toolbox). Correlations were computed using Pearson linear correlation coefficient (Buck, Daniel, & Singer, 2002).

1. Species may overlap and have a functional relationship but not interact,

3 | RESULTS

2. Species may overlap and not have any biotic interaction, 3. Sampling is unlikely to reflect the full geographical distribution

3.1 | Changes in asymmetry

and seasonal ranges of each species, 4. Sampling has not included all species in the ecosystem,

Between 1992 and 2000, the frequency of overlaps between co-

5. Overlaps that we eliminated as occurring by chance alone may

occurring species showed an asymmetric distribution, skewed

be meaningful, 6. Knowledge of the life history and ecology of many of these species are poorly known.

towards 1 (Figure 1a). The frequency distribution followed a powerlaw (scale-free) distribution / = 3.46, xmin = 17.4, p = .04) (Clauset et al., 2009). Between 2001 and 2011, the geographical overlap between species showed less asymmetrical overlaps and was more Gaussian (Figure 1b). The two distributions were significantly differ-

2.7 | Ocean temperatures and nutrients

ent when compared using the two-sample Kolmogorov–Smirnov test

Temperature, salinity, dissolved oxygen, phosphate, nitrate and sili-

(h = 1, k = 0.29, 5% significance).

cate were extracted from the CSIRO Atlas of Regional Seas (CARS). This is a high-resolution (0.5° 9 0.5°) seasonal atlas available in monthly and seasonal values (Ridgeway, Dunn, & Wilkin, 2002). Full details on CARS can be found at http://www.cmar.csiro.au/cars. Monthly satellite sea surface temperature (SST) data at a spatial resolution of 1° 9 1° were downloaded from the GES-DISC Interactive Online Visualization and Analysis Infrastructure (GIOVANNI) as part of NASA’S Goddard Earth Sciences (GES) Data and Information Services Center (http://www.giovanni.sci.gsfc.nasa.gov/giovanni). Ocean temperature data at the surface, 50 and 100 m depth were extracted from the Australian Ocean Data Network (http:// www.portal.aodn.org.au) of the Integrated Marine Observing System (IMOS) for the Port Hacking (34°50 S, 151°150 E) coastal station to compose yearly anomalies of temperature and salinity at 0, 50 and 100 m depth. This data set, sampled weekly since 1953, represents one of the most detailed and longest sampling of biogeochemical oceanographic data.

3.2 | Changes in species overlap robustness and sensitivity Of the 61 species in our study, 49 species showed a decrease in contribution to ecosystem robustness between 1992 and 2011 (negative values in Fig. S3a). In all, 32 species showed a decrease in their sensitivity to potential interactions with other species (positive values in Fig. S3b). In all, 29 species showed an increase in their sensitivity index (negative values in Fig. S3b). Overall species overlap robustness (Sin) of the ecosystem declined from 1994 with the greatest decrease occurring since 2005 (Figure 2a). When grouped by pelagic and demersal species, we found that the Sin of the pelagic community decreased (Figure 2b), but for the demersal community it increased, except for a decline between 2008 and 2010 (Figure 2c). Pelagic species overlap sensitivity (Sout) to the loss of potential interactions with other species increased until 2006, then declined (Figure 3a). During our study period, demersal species became less sensitive to the loss of potential interactions

2.8 | East Australian Current (EAC) & Fisheries data Monthly EAC net transport (southward) data were extracted from the

with other species (Figure 3b).

BLUElink Reanalysis (BRAN) ocean model. BLUElink is a partnership

3.3 | Structural keystone species

between CSIRO, Bureau of Meteorology and the Royal Australian

Using our index of species overlap robustness (Sin), which accounts for

Navy (Oke, Brassington, Griffin, & Schiller, 2008). The model is an

the full range of symmetrical to asymmetrical overlaps that all species

eddy resolving ocean reanalysis that combines model estimates with

may have with a particular species, we ranked the species that had

observations to produce more realistic hind cast and forecast fields of

the highest proportion of overlaps (Table 1) as structural keystone

circulation around Australia at ~10 km resolution (Oke et al., 2013).

species. In 1992, keystone species were mostly pelagic predator spe-

Monthly fisheries catch data were extracted for the study area from

cies such as the tunas and billfish. At year 2011, structural keystone

CSIRO holdings of Commonwealth and State fisheries catch records.

species had altered to mainly demersal fish species and cephalopods.

2.9 | Statistical methods

3.4 | Oceanographic and fisheries correlations

Time series data of each of the species overlap indices (Sin and Sout)

Overall ecosystem robustness anomaly was negatively correlated

and each of the oceanographic and fisheries data were first

with EAC net southerly transport anomaly at a time lag of 3 years

e94

|

GRIFFITH

(a)

18

ET AL.

1992−2000

16 14

freq. (x103)

12 10 8 6 4 2 0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

0.7

0.8

0.9

1.0

asymmetry

(b) 20

2001−2011

18 16

freq.(x103)

14 12 10 8 6 4 2 0

0.1

0.2

0.3

0.4

0.5

0.6

asymmetry

F I G U R E 1 Frequency distribution of the asymmetry Sim (A,B) between pairs of co-occurring species in the East Australian marine ecosystem: (a) between years 1992 and 2000 and (b) between years 2001 and 2011. When Sim = 0, the overlap is completely symmetric. When Sim = 1, the overlap is completely asymmetric. (Data points = 74,420, no of species = 61)

(Table 2). That is, as EAC net southerly transport increased, network

showed a correlation with observed fish catch. The strongest corre-

robustness decreased 3 years later. Pelagic species overlap robustness

lation was between the pelagic species changes in sensitivity and

(Sin) and species overlap sensitivity (Sout) were also negatively corre-

observed fish catch. However, the correlations were not as signifi-

lated with satellite sea surface temperature and also the 50 and

cant as for the oceanographic metrics such as EAC net transport

100 m temperature data from the Port Hacking station data at up to

south and seawater temperature changes.

a 4-year time lag. Pelagic species overlap sensitivity (Sout) to the potential loss of other species was also correlated with EAC net southerly transport. Demersal species overlap robustness (Sin) was

4 | DISCUSSION

strongly negatively correlated with changes in satellite sea surface temperature, the 50 m Port Hacking temperature data and EAC net

A key question in global change biology concerns not just the human

southerly transport at a 3- to 4-year lag.

impacts on individual species or small groups of species, but the con-

Pelagic species overlap sensitivity (Sout) was strongly correlated

sequences for the stability and persistence of entire ecosystems

with EAC net southerly transport, changes in sea surface tempera-

(Ings et al., 2009). This is the first attempt to create a spatially expli-

ture and salinity at a 0- to 3-year time lag. Demersal Sin was strongly

cit understanding of how the changes in the complex geographical

negatively correlated with changes in increasing sea level, increasing

overlaps between species have altered the stability of a large marine

salinity and increasing EAC net transport south at a 0- to 4-year

ecosystem.

time lag.

Using a network approach, we quantified the nature of the over-

Changes in the overall ecosystem robustness did not show any

laps between species accounting for the realistic continuum of distri-

significant correlation with the observed fish catch. The robustness

butional overlaps between species that ranges from symmetrical to

of the pelagic community and demersal community separately

asymmetrical. Empirical and theoretical models of aquatic and

|

ET AL.

(a)

0.4

Species contribution to robustness (stability)

GRIFFITH

0.3

e95

All species

0.2 0.1 0 −0.1 −0.2 −0.3 −0.4 −0.5

1993

1995

1997

1999

2001

2003

2005

2007

2009

2011

2003

2005

2007

2009

2011

2003

2005

2007

2009

2011

Year

Species contribution to robustness (stability)

(b) 0.5

Pelagic species

0.4 0.3 0.2 0.1 0 −0.1 −0.2 −0.3 −0.4 −0.5

1993

1995

1997

1999

2001

(c)

0.5

Species contribution to robustness (stability)

Year

0.4

Demersal species

0.3 0.2 0.1 0 −0.1 −0.2 −0.3

1993

1995

1997

1999

2001 Year

F I G U R E 2 Anomalies of the normalized annual species overlap robustness (Sin) used here as a measure of species contribution to robustness in the demersal and pelagic communities: (a) All species, (b) Pelagic community and (c) Demersal community. Data smoothed with a 3-year running mean. Positive anomalies indicate increased overlaps between species and negative anomalies indicate decreased overlaps between species (n = 61 species) terrestrial ecosystems show that asymmetry promotes stability and,

frequency of asymmetrical overlaps between pairs of species had

that a high degree of asymmetry may be required to prevent ecosys-

decreased. In marine systems, where many species can alter their

tem collapse (Holt, 2006). In contrast, increasing symmetry can

feeding and other functional interactions, making inferences between

reduce the capacity of a network to recover from environmental

structural and functional relationships is difficult. We have insuffi-

perturbations and increase the potential for transient dynamics. For

cient evidence to conclude that asymmetrical geographic distribu-

the East Australian marine ecosystem, since 1992, we found that the

tions of species are linked to functional asymmetry of food web

e96

|

GRIFFITH

0.3

(a)

ET AL.

Pelagic species

Sensitivity to loss of other species

0.2 0.1 0 −0.1 −0.2 −0.3 −0.4 −0.5 −0.6

1993

1995

1997

1999

2001

2003

2005

2007

2009

2011

2003

2005

2007

2009

2011

Year

(b)

0.5

Demersal species

Sensitivity to loss of other species

0.4 0.3 0.2 0.1 0 −0.1 −0.2 −0.3

1993

1995

1997

1999

2001 Year

F I G U R E 3 Anomalies of the species overlap sensitivity (Sout) used as a measure of species sensitivity to the loss of potential interactions with other species: (a) Pelagic community and (b) Demersal community. Data smoothed with a 3-year running mean. Positive anomalies indicate increased overlaps between species and negative anomalies indicate decreased overlaps between species. (n = 61 species) T A B L E 1 Top ranked structural keystone species at year 1992 and year 2011, ranked by the species overlap robustness index (Sin) Rank

Species at year 1992

Species at year 2011

1

Porbeagle shark

Lamna nasus

Gould’s flying squid

Nototodarus gouldi

2

Yellowfin tuna

Thunnus albacares

Frostfishes

Benthodesmus spp.

Thunnus alalunga

King fish

Seriola ialandi

3

Albacore tuna

4

Sharks other

5

Broad billed swordfish

Xiphias gladius

Blue-eye trevalla

Hyperoglyphe antarctica

Gulper shark

Centrophorus granulosus

6

Southern bluefin tuna

Thunnus maccoyii

Barracouta

Thyrsites atun

7

Smooth oreo dory

Pseudocyttus maculatus

Alfosino

Beryx decadactylus

8

Skates

Rajidae

Hapuka

Polyprion oxygeneios

9

Orange roughy

Hoplostethus atlanticus

Mahi mahi

Coryphaena hippurus

10

Blue warehou

Seriolella brama

Silver dory

Cyttus australis

interactions. Nevertheless, a number of terrestrial studies have

In ecological terms, highly connected species may be keystone

shown a relationship between spatial co-occurrence and functional

species that have an unexpectedly large effect on other species.

interactions

between

species


jo (Arau

et al.,

2011;

Azaele,

Many increasingly mathematically complex indices have been pro-

Muneepeerakul, Rinaldo, & Rodriquez-Iturbe, 2010). The high inci-

posed to identify keystone species for food web stability (Lai, Lie, &

dence of asymmetrical interactions between species is also consid-

Jord
an, 2012; Livi, Jord
an, Lecca, & Okey, 2011). The index used

ered an evolutionary consequence of species interaction correlated

here, based on graph theory, assumes that the positional importance

with species abundance (V
azquez et al., 2007).

of a species in food web interactions can be determined from the

GRIFFITH

|

ET AL.

T A B L E 2 Significant correlations (p < .01) between the changes in species overlap robustness (Sin) or species overlap sensitivity (Sout), oceanographic and fisheries data Species overlap indices

Lag in years

Pearson linear correlation

Species overlap robustness

EAC transport southward Satellite sea level

2016). We extended network degree (k), the simplest measure of topological species importance based on the number of connections for a given species, to a measure of species importance based on the distantly, this method accounts for the degree of asymmetry in the

0

0.67

overlaps. We identified potential structural keystone species that we

3

0.65

know also have a functional relationship with the other species we

0

0.75

measured and are known to be ecologically important in the ecosystem (Condie, Johnson, Fulton, & Bulman, 2014).

Pelagic species Primary productivity

reliable guide for management strategies (McDonald-Madden et al.,

tributional overlap each species has with all other species. Impor-

Ecosystem all species Sea level (FDS)

e97

2

0.58

Based on our network index, there has been a significant change

0

0.70

in structural keystone species in the East Australian marine system.

5

0.57

For instance, pelagic predator species such as albacore and yellow

0

0.82

Temperature (PHS-surface)

4

0.52

Satellite sea surface temperature

3

0.53

Salinity (PHS-surface)

4

0.57

Observed total fish catch

2

0.68

Temperature (PHS-50 m depth)

3

0.46

EAC net transport southward

4

0.83

Sea level (FDS) Observed total catch Satellite sea level

Demersal species

In contrast, demersal species such as trevalla have become the most important structural species increasing their potential interactions with other species by ~40%. The change in structural keystone species may have important implications for the on-going energy flow within the ecosystem. Structural asymmetry and the stability of the system are enhanced through keystone predator species that act as couplers of strong and weak interaction of energy flow (Rooney, McCann, Gellner, & Moore, 2006). Food web architecture based on multiple asymmetric energy channels provides ecosystems with a et al., 2006). Any reduction in the number of potential interactions

Pelagic species EAC net transport southward

3

0.53

Observed total fish catch

1

0.76

Salinity (PHS-50 m depth)

2

0.67

0

0.61

2

0.55

Temperature (PHS-50 m depth)

other species, as quantified by the species overlap robustness index.

potent mechanism for responding to large perturbations (Rooney

Species overlap sensitivity

Satellite sea surface temperature

fin tuna have seen a decrease in ~60% of potential interactions with

by predator species that act as couplers of fast and slow energy channels via both strong and weak geographical interaction, such as we have shown, has the potential to decrease the energy flux and stability of the system. Our present understanding of changes in species potential interactions has focused on individual species or small numbers of species. This is a limitation to understanding changes in species interactions under climate change (Tylianakis et al., 2008). To address these limitations, the idea of “community modules” originally

Demersal species

conceived by Holt (1997) has been proposed as a means to simplify

0

0.75

4

0.57

Satellite sea level

0

0.80

Salinity (PHS-50 m depth)

0

0.81

Satellite sea surface temperature

0

0.66

Temperature (PHS-50 m depth)

3

0.53

large pelagic and small pelagic species, the contribution of species to

Observed total catch

1

0.53

tive of decreasing stability and decreasing resilience of species to

Sea Level (FDS) EAC net transport southward

FDS, Fort Dennison Station; PHS, Port Hacking Station.

complex food web interactions into tractable communities, to help understand the effect of changes in environment on predator–prey interactions and changes in keystone species (Gillman, Urban, Tewksbury, Gilchrist, & Holt, 2010). We followed this idea, simplifying the large species spatial distribution data into pelagic and demersal communities. We showed that in the community networks of the robustness of each community has decreased over time, indicathe potential loss of other species. Demersal communities have shown an increase in robustness and a decrease in their sensitivity to the loss of geographical overlaps with other species in the ecosys-

number of potential direct and indirect links that a species has with

tem. Modelling of the potential changes in community composition

other species. Increasingly, the indices are being used to prioritize

and biomass, as a consequence of ocean warming, ocean acidifica-

and determine optimal species management strategies although

tion and continued fisheries exploitation show restructuring of the

recent studies show that no single keystone index provides a

demersal and pelagic communities by year 2040 (Fulton, 2010;

|

e98

GRIFFITH

ET AL.

Griffith, Fulton, & Richardson, 2011). In comparison to the model

Pacific (Tian et al., 2004), significant correlations between climate

results, our study using empirical data suggests that structural

and oceanographic indices are consistent with climate variability

changes have already occurred over the last two decades.

forcing changes in biological variables and processes. The changes

Changes in oceanographic conditions and the effect of commer-

we observed can be attributed to warming and changes in the EAC.

cial fisheries are plausible factors responsible for the changes in

However, we cannot attribute the changes to a long-term climate

species overlap robustness (Sin) and species overlap sensitivity (Sout).

signal as our time period of species distribution is too short for sug-

There have been a number of studies showing changes in oceano-

gesting a long-term correlation and the changes may be part of the

graphic conditions and individual species distribution for the east

decadal variability of the South Pacific gyre (Hill et al., 2011; Roem-

Australian marine ecosystem (Suthers et al., 2011). The dominant

mich et al., 2007).

oceanographic feature is the East Australian Current (EAC) that

Fisheries catch data were negatively correlated with changes in

transports warm, salty, low nutrient water southward along the East

our species overlap indices for both the demersal and pelagic groups

Australian coast. The South Pacific gyre and consequently the EAC

(Table 2). Analysis of a wide array of ecosystems showed that

have exhibited a long-term strengthening trend related to strength-

despite significant differences in ocean dynamics and exploitation

ening of the subtropical westerly winds in the South Pacific (Hill,

history, fishing was usually the dominant driver of reduced species

Rintoul, Ridgway, & Oke, 2011; Roemmich et al., 2007). Our study

abundance and loss of species (Link et al., 2011). Trawling intensity

area has experienced increasing ocean warming and increased net

in our study area has never reached the high levels of exploitation

transport of warmer water southward consistent with this prevailing

found in other productive fisheries parts of the world such as the

view.

North Sea (Novaglio, Ferratti, Smith, & Frusher, 2016). Results based

Recent studies on distributional shifts of ectothermic species

on species–area relationships (SAR) from historic trawl surveys of

with ocean warming in other regions indicate that the leading and

demersal species within the East Australian marine ecosystem show

trailing edges of the distribution of some species are responding to

that demersal fish communities have undergone some changes in

ocean warming although many species showed no response

species richness and evenness from the removal of fish and inverte-

(Poloczanska et al., 2013). In recent years, the strengthening of the

brates with bottom disturbance from trawl nets (Novaglio et al.,

EAC and associated ocean warming has been linked to a southward

2016). The effect of fishing exploitation on the slope of SARs in

range expansion of a number of near-shore marine species, changes

exploited demersal south-east Australian fish communities was con-

in the structure of near-shore zooplankton and population-level

sidered low.

changes in some important commercial invertebrate species (Johnson

Species overlap robustness (Sin) for the pelagic community was

et al., 2011). In our study area, ocean temperatures have already

highest during the 1990s. This was also the period of intensive and

reached a threshold that is causing increased metabolic costs and

increased fishing pressure on the highly connected pelagic tuna and

decreased growth for the Banded Morwong (Cheilodactylus spectabo-

billfish species (Fulton et al., 2007). While abundance of these spe-

lis; Neuheimer, Thresher, Lyle, & Semmens, 2011).

cies has remained high with improved fisheries management, these

We found significant correlations between changes in species

species show fewer network interactions and have become less

overlap robustness (Sin) and species overlap sensitivity (Sout) with ocean

robust to species loss. Our analysis shows a significant structural

warming in the region. This is not unexpected, as shifts in distribu-

change and reduction in the keystone importance of these species

tion of both demersal, pelagic species are largely driven by changes

that would not be evident from species abundance alone.

in water temperature, and redistribution of preferred temperature

There is a major on-going debate within the field of climate

ranges (Hobday & Evans, 2013). The correlation was most significant

change between the two opposing views that species will respond

for the demersal community with increasing species overlaps to the


jo et al., 2011). to change individualistically or as assemblages (Arau

south and south-east suggesting that ocean warming may be the

Our results show that pelagic species appear to be responding indi-

major driver in the increase in potential interactions for these spe-

vidualistically with no consistent pattern to the direction of

cies. Ocean warming from climate-related changes has been

changes in distribution. In comparison, demersal species are

reported for many coastal and continental shelf fish and benthic

responding as a cohort with increased predominantly poleward

assemblages (Nye, Link, Hare, & Overholtz, 2009; Shackell, Bundy,

interactions.

Nye, & Link, 2012). In the North Sea, shifts in species distribution

In conclusion, recent studies suggest that differences in the

seem to have been in response to a positive North Atlantic Oscilla-

response of individual species and populations to climate change

tion (NAO); the dominant mode of winter variability in the North

effects may substantially re-organize species interactions and marine

Atlantic region. The positive winter NAO index from stronger than

ecosystem functioning at a regional scale. This study based on net-

usual subtropical high pressure and a deeper than normal Icelandic

work theory and extensive observational data shows that substantial

low is characterized by elevated temperatures and stronger westerly

re-organization of the complex distributional overlaps between spe-

winds. In the 1980s, this resulted in a regime shift with a reduction

cies may have already occurred for the East Australian marine

in recruitment of North Sea cod, whereas Baltic sprat started to

ecosystem. Our approach provides a way forward to incorporate

thrive (Alheit et al., 2005). In the north-east Pacific (McFarlane, King,

complex spatial changes in multi-species distribution into an analysis

& Beamish, 2000; Tian, Akamine, & Suda, 2004) and the north-west

of species and ecosystem responses to climate change.

GRIFFITH

|

ET AL.

ACKNOWLEDGEMENTS We would like to thank Neil Klaer of the Commonwealth Scientific and Industrial Research Organisation (CSIRO) for his help with collating the fisheries catch data. Lara Wilcox of the University of Otago for his assistance with data analysis. The Department of Education, Employment and Workplace Relations Grant–ANNIMS Springboard Program funding supported the work. We thank Caleb Gardner, Colin Simpfendorfer, Michelle Heupel and Anya Waite for their roles in obtaining the funding and getting the project started.

DATA ACCESSIBILITY Data, programs and analysis are available at the IMAS Data Portal (http://data.imas.utas.edu.au), the NPI Data Portal (http://data.npola r.no) or from the corresponding author.

ORCID Gary P. Griffith

http://orcid.org/0000-0002-7136-4237

REFERENCES €llmann, C., Dutz, J., Kornilovs, G., Loewe, P., Mohrholz, V., & Alheit, J., Mo Wasmund, N. (2005). Synchronous ecological regime shifts in the central Baltic and the North Sea in the late 1980s. ICES Journal of Marine Science, 62, 1205–1215.
jo, M. B., & Peterson, T. (2012). Uses and misuses of bioclimatic Arau enveloped modeling. Ecology, 93, 1527–1539.
jo, M. B., Rozenfeld, A., Rahbek, C., & Marquet, P. A. (2011). Using Arau species co-occurrence networks to assess the impacts of climate change. Ecography, 34, 897–908. Azaele, S., Muneepeerakul, R., Rinaldo, A., & Rodriquez-Iturbe, I. (2010). Inferring plant ecosystem organization from occurrences. Journal of Theoretical Ecology, 262, 323–329. Barret, A., Barthelemy, M., Pastor-Satorras, R., & Vespignani, A. (2004). The architecture of complex weighted networks. Proceedings of the National Academic Science USA, 101, 3747–3752. Bascompte, J., Jordano, P., & Olesen, J. M. (2006). Asymmetric coevolutionary networks facilitate biodiversity maintenance. Science, 312, 431–433. Berlow, E. L., Dunne, J. A., Martinez, N. D., Stark, P. B., Williams, R. J., & Brose, U. (2009). Simple prediction of interaction strengths in complex food webs. Proceedings of the National Academy of Sciences, 106, 187–191. Berlow, E. L., Neutel, A.-M., Cohen, J. E., De Ruiter, P. C., Ebenman, B., Emmerson, M., . . . Petchey, O. (2004). Interaction strengths in food webs: Issues and opportunities. Journal of Animal Ecology, 73, 585– 598. Borgatti, S. P., Everett, M. G., & Johnson, J. C. (2013). Analyzing social networks. London: SAGE Publications Ltd. Buck, J. R., Daniel, M. N., & Singer, A. C. (2002). Computer explorations in signals and systems using MATLAB, 2nd ed. Upper Saddle River, NJ: Prentice Hall. Burrows, M. T., Schoeman, D. S., Buckley, L. B., Moore, P., Poloczanska, E. S., Brander, K. M., . . . Richardson, A. J. (2011). The pace of shifting climate in marine and terrestrial ecosystems. Science, 334, 652–655. Cai, W., Shi, G., Cowan, T., Bi, D., & Ribbe, J. (2005). The response of the southern annular mode, the east Australian current, and the southern

e99

mid-latitude ocean circulation to global warming. Geophysical Research Letters, 32, 1–4. Clauset, A., Shalizi, C. R., & Newman, M. E. J. (2009). Power-law distributions in empirical data. SIAM Review, 51, 661–703. Condie, S. A., Johnson, P., Fulton, E. A., & Bulman, C. M. (2014). Relating food web structure to resilience, keystone status and uncertainty in ecological responses. Ecosphere, 5, 1–6. Dunne, J. A., Williams, R. J., & Martinez, N. D. (2002). Food-web structure and network theory: The role of connectance and size. Proceedings of the National Academy of Sciences USA, 99, 12917– 12922. Fulton, E. A. (2010). Approaches to end-to-end ecosystem models. Journal of Marine Systems, 81, 171–183. Fulton, E. A., Smith, A. D. M., & Smith, D. C. (2007). Alternative management strategies for southeast Australian Commonwealth Fisheries: stage 2: quantitative management strategy evaluation. Australian Fisheries Management Authority Report. Gillman, S. E., Urban, M. C., Tewksbury, J., Gilchrist, G. W., & Holt, R. D. (2010). A framework for community interactions under climate change. Trends in Ecology and Evolution, 25, 325–331. Griffith, G. P., & Fulton, E. A. (2014). New approaches and challenges to simulate the complex interaction effects of multiple human impacts on the marine environment. ICES Journal of Marine Science, 71, 764–774. Griffith, G. P., Fulton, E. A., & Richardson, A. J. (2011). Effects of fishing and acidification-related benthic mortality on the south-east Australian marine ecosystem. Global Change Biology, 17, 3058–3074. Griffith, D. M., Veech, J. A., & Marsh, C. J. (2016). cooccur: Probabilistic species co-occurrence analysis in R. Journal of Statistical Software, 69, 1–6. Hill, K. L., Rintoul, S. R., Ridgway, K. R., & Oke, P. R. (2011). Decadal changes in the South Pacific western boundary current system revealed in observations and ocean state estimates. Journal of Geophysical Research Oceans, 116, C01009. https://doi.org/10.1029/ 2009JC005926 Hobday, A. J., & Evans, K. (2013). Detecting climate impacts with oceanic fish and fisheries data. Climatic Change, 119, 49–62. Holt, R. D. (1997). Predation, apparent competition, and the structure of prey communities. Theoretical Population Biology, 12, 197–229. Holt, R. D. (2006). Asymmetry and stability. Nature, 442, 252–253. €thgen, N., Brown, J., DorIngs, T. C., Montoya, J. M., Bascompte, J., Blu mann, C. F., . . . Woodward, G. (2009). Ecological networks-beyond food webs. Journal of Animal Ecology, 78, 253–269. Johnson, C. R., Banks, S. C., Barrett, N., Cazassus, F., Dunstan, P. K., Edgar, G. J., . . . Taw, N. (2011). Climate change cascades: Shifts in oceanography, species’ ranges and subtidal marine community dynamics in eastern Tasmania. Journal of Experimental Marine Biology and Ecology, 400, 17–32. Jord
an, F., Okey, T. A., Bauer, B., & Libralato, S. (2008). identifying important species: Linking structure and function in ecological networks. Ecological Modelling, 216, 75–80. Lai, S.-M., Lie, W. C., & Jord
an, F. (2012). On the centrality and uniqueness of species from the network perspective. Biology Letters, 8, 570–573. Ling, S. D., & Johnson, C. R. (2009). Population dynamics of an ecologically important range-extender: Kelp beds versus sea urchin barrens. Marine Ecology Progress Series, 374, 113–125. Link, J. S., Bundy, A., Overholtz, W., Shackell, N., Manderson, J., Duplisea, D., . . . Friedland, K. D. (2011). Ecosystem-based fisheries management in the Northwest Atlantic. Fish and Fisheries, 12, 152–170. Livi, C. M., Jord
an, F., Lecca, P., & Okey, T. A. (2011). Identifying key species in ecosystems with stochastic sensitivity analysis. Ecological Modelling, 222, 2542–2551. McDonald-Madden, E., Sabbadin, R., Game, E. T., et al. (2016). Using food-web theory to conserve ecosystems. Nature Communications, 7, 10245. https://doi.org/10.1038/ncomms10245

e100

|

McFarlane, G. A., King, J. R., & Beamish, R. J. (2000). Have there been recent changes in climate? Ask the fish. Progress in Oceanography, 47, 147–169. Neuheimer, A. B., Thresher, R. E., Lyle, J. M., & Semmens, J. M. (2011). Tolerance limit for fish growth exceeded by warming waters. Nature Climate Change, 1, 110–113. Novaglio, C., Ferratti, F., Smith, A. D. M., & Frusher, S. (2016). Speciesarea relationships as indicators of human impacts on demersal fish communities. Diversity and Distributions, 22, 1186–1198. Nye, J. A., Link, J. S., Hare, J. A., & Overholtz, W. J. (2009). Changing spatial distribution of fish stocks in relation to climate and population size on the Northeast United States continental shelf. Marine Ecology Progress Series, 393, 111–129. Oke, P. R., Brassington, G. B., Griffin, D. A., & Schiller, A. (2008). The Bluelink Ocean Data Assimilation System (BODAS). Ocean Modelling, 21, 46–70. Oke, P. R., Sakov, P., Cahill, M. L., Dunn, J. R., Fiedler, R., Griffin, D. A., . . . Schiller, A. (2013). Towards a dynamically balanced eddy-resolving reanalysis: BRAN3. Ocean Modelling, 21, 46–70. Pollock, L. J., Tingley, R., Morris, W. K., Golding, N., O’Hara, R. B., Parris, K. M., . . . McCarthy, M. A. (2014). Understanding cooccurrence by modelling species simultaneously with a Joint Distribution Model (JSDM). Methods in Ecology and Evolution, 5, 397–406. Poloczanska, E. S., Brown, C. J., Sydeman, W., Kiessling, W., Schoeman, D. S., Moore, P. J., . . . Richardson, A. J. (2013). Global imprint of climate change on marine Nature. Climate Change, 3, 919–925. Ridgeway, K. R. (2007). Long-term trend and decadal variability of the southward penetration of the East Australian current. Geophysical Research Letters, 34, L13613. https://doi.org/10.1029/2006JC003898 Ridgeway, K. R., Dunn, J. R., & Wilkin, J. L. (2002). Ocean interpolation by four-dimensional least squares-application to the waters around Australia. Journal of Atmospheric and Oceanic Technology, 19, 1357–1375. Roemmich, D., Gilson, J., Davis, R., Sutton, P., Wijfells, S., & Riser, S. (2007). Decadal spin-up of the South Pacific subtropical gyre. Journal of Physical Oceanography, 37, 172–173. Rooney, N., McCann, K., Gellner, G., & Moore, J. C. (2006). Structural asymmetry and the stability of diverse food webs. Nature, 442, 265–269. Salomon, Y., Connolly, S. R., & Bode, L. (2010). Effects of asymmetric dispersal on the coexistence of competing species. Ecology Letters, 13, 432–441.

GRIFFITH

ET AL.

Shackell, N. L., Bundy, A., Nye, J. A., & Link, J. S. (2012). Common large-scale responses to climate and fishing across ecosystems in the northwest Atlantic. ICES Journal of Marine Science, 69, 151– 162. Suthers, I. M., Young, J. W., Baird, M. E., Roughan, M., Everett, J. D., Brassington, G. B., . . . Ridgeway, K. (2011). The strengthening East Australian Current, its eddies and biological effects—An introduction and overview. Deep Sea Research, 58, 538–546. Tian, Y., Akamine, T., & Suda, M. (2004). Modeling the influence of oceanic-climatic changes on the dynamics of Pacific saury in the NorthWestern Pacific using a life-cycle model. Fisheries Oceanography, 13, 125–137. Tylianakis, J. M., Didham, R. K., Bascompte, J., & Wardle, D. A. (2008). Global change and species interactions in terrestrial ecosystems. Ecology Letters, 11, 1351–1363. €thgen, N., Krasnov, B. Vazquez, D. P., Melian, C. J., Williams, N. M., Blu R., & Poulin, R. (2007). Species abundance and asymmetric interaction strength in ecological networks. Oikos, 116, 1120–1127. Veech, J. A. (2013). A probabilistic model for analysing species co-occurrence. Global Ecology and Biogeography, 22, 252–260. Wilson, L. J., Fulton, C. J., Mc Hogg, A., Joyce, K. E., Radford, B. T. M., & Fraser, C. (2016). Climate-driven changes to ocean circulation and their inferred impacts on marine dispersal patterns. Global Ecology and Biogeography, 25, 923–939.

SUPPORTING INFORMATION Additional Supporting Information may be found online in the supporting information tab for this article.

How to cite this article: Griffith GP, Strutton PG, Semmens JM. Climate change alters stability and species potential interactions in a large marine ecosystem. Glob Change Biol. 2018;24:e90–e100. https://doi.org/10.1111/gcb.13891