Oecologia (2008) 156:883–894
DOI 10.1007/s00442-008-1043-9

GLOBAL CHANGE ECOLOGY - ORIGINAL PAPER



Range expansion of a habitat-modifying species leads to loss
of taxonomic diversity: a new and impoverished reef state
S. D. Ling




Received: 27 December 2007 / Accepted: 11 April 2008 / Published online: 15 May 2008
Ó Springer-Verlag 2008


Abstract Global climate change is predicted to have         intact macroalgal beds in terms of total numbers of taxa,
major negative impacts on biodiversity, particularly if       total individuals and Shannon diversity. In contrast, the
important habitat-modifying species undergo range shifts.      faunal community of the barrens habitat is overwhelmingly
The sea urchin Centrostephanus rodgersii (Diadematidae)       impoverished. Of 296 individual floral/faunal taxa recor-
has recently undergone poleward range expansion to rela-       ded, only 72 were present within incipient barrens, 253
tively cool, macroalgal dominated rocky reefs of eastern       were present in the recovered patches, and 221 were
Tasmania (southeast Australia). As in its historic environ-     present within intact macroalgal beds. Grazing activity of
ment, C. rodgersii in the extended range is now found in       C. rodgersii results in an estimated minimum net loss of
association with a simplified ‘barrens’ habitat grazed free      approximately 150 taxa typically associated with Tasma-
of macroalgae. The new and important role of this habitat-      nian macroalgal beds in this region. Such a
modifier on reef structure and associated biodiversity was      disproportionate effect by a single range-expanding species
clearly demonstrated by completely removing C. rodgersii       demonstrates that climate change may lead to unexpectedly
from incipient barrens patches at an eastern Tasmanian site     large impacts on marine biodiversity as key habitat-modi-
and monitoring the macroalgal response relative to unma-       fying species undergo range modification.
nipulated barrens patches. In barrens patches from which
C. rodgersii was removed, there was a rapid proliferation      Keywords Biodiversity Á Centrostephanus rodgersii Á
of canopy-forming macroalgae (Ecklonia radiata and          Climate change Á Kelp beds Á Sea urchin barrens
Phyllospora comosa), and within 24 months the algal
community structure had converged with that of adjacent
macroalgal beds where C. rodgersii grazing was absent. A       Introduction
notable scarcity of limpets on C. rodgersii barrens in
eastern Tasmania (relative to the historic range) likely       Global climate change is predicted to have major negative
promotes rapid macroalgal recovery upon removal of the        consequences for marine biodiversity (reviewed by Ro-
sea urchin. In the recovered macroalgal habitat, faunal       senzweig et al. 2007). While impacts on species are widely
composition redeveloped similar to that from adjacent        anticipated to occur directly as a result of shifts in bio-
                                   climate envelopes (e.g. Hijmans and Graham 2006), eco-
                                   system effects mediated by a range shift of key habitat-
Communicated by Pete Peterson.                    modifying species may result in disproportionately large
                                   impacts on marine biodiversity (e.g. Hughes 2000; Harley
Electronic supplementary material The online version of this
article (doi:10.1007/s00442-008-1043-9) contains supplementary    et al. 2006). If habitat-modifying species undergo range
material, which is available to authorized users.          shift, the occurrence of ‘catastrophic shifts’ (Scheffer et al.
                                   2001) in marine ecosystems are likely to become more
S. D. Ling (&)
                                   common, with altered ecosystem states having major
School of Zoology, University of Tasmania, Private Bag 5,
Hobart 7001, Australia                        impacts on biodiversity (e.g. Elmqvist et al. 2003; Folke
e-mail: sdling@utas.edu.au                      et al. 2004; Hughes et al. 2005).


                                                             123
884                                                   Oecologia (2008) 156:883–894


  Driven by increased poleward penetration of the warm           Since it was first detected off the Tasmanian mainland at
East Australian Current (EAC, see Ridgway 2007), the sea         St Helens in 1978, the abundance of C. rodgersii has
urchin Centrostephanus rodgersii (Diadematidae) has           increased, the range has moved further south and wide-
recently undergone a southern range extension in temperate        spread barrens habitat (continuous across hundreds of
southeastern Australia (Edgar et al. 2004, 2005; Johnson         metres) now occur in some locations (Johnson et al. 2005).
et al. 2005; Ling et al. 2008). Extending its range from         While widespread barrens currently occur in relatively few
New South Wales (NSW) south to the Tasmanian coastline          places in eastern Tasmania, a major feature of C. rodgersii
(Fig. 1), C. rodgersii is just one of many species observed       grazing on this coast is the occurrence of small incipient
to have undergone recent range extension in this region         barrens patches (1–10 m in diameter) within dense and
(Edgar 1997; reviewed by Poloczanska et al. 2007).            diverse macroalgal beds (Johnson et al. 2005). Given the
However, it is the range extension of C. rodgersii that         anticipated positive effects of climate change on C. rod-
appears to be particularly important to the benthic com-         gersii dispersal and larval development within Tasmania
munity given the sea urchins’ ability to eliminate            (Ling et al. 2008; Ling, submitted), barrens habitat could
macroalgal habitat and effect a catastrophic shift to an         potentially expand on this coastline to reflect patterns
alternative sea urchin ‘barrens’ state (e.g. Fletcher 1987;       already observed in NSW (Johnson et al. 2005). Thus,
Andrew 1991, 1993; Andrew and Underwood 1993; Hill            C. rodgersii grazing in eastern Tasmania is considered to
et al. 2003). Such is the importance of this herbivore that       pose a major threat to the structure and functioning of the
within its historic NSW range approximately 50% of all          biologically diverse macroalgal-dominated rocky reefs
near-shore rocky reef is urchin barrens as a result of          (e.g. Edgar et al. 2004, Ling et al. 2008) and the important
grazing by this single sea urchin species (Andrew and          resources that they support (Johnson et al. 2005). The aim
O’Neill 2000).                              of this study was to explicitly examine the impact of this
                                     range-extending species on reef habitat structure and
                                     associated biodiversity within the extended range by using
                                     controlled sea urchin removals.
       N

       -30
              NSW
                                     Materials and methods

               148                     Experimental manipulation

                                     Manipulations testing the effect of C. rodgersii grazing on
                                     the structure and biodiversity of rocky reefs within the
                                     extended range were undertaken at Bicheno on the east
-41.5                                  coast of Tasmania (Fig. 1). Six discrete incipient barrens
                                     patches, ranging in size from approx. 3 to 6 m in diameter,
 Latitude




              Tasmania                   each supporting 8–116 resident C. rodgersii (density 1.3–
                                     3.6 m-2), were randomly assigned as complete C. rod-
                                     gersii ‘removal’ or ‘unmanipulated’ control patches; no
                                     attempt was made to standardise urchin numbers across the
                                     naturally occurring patches. As described for NSW,
          100 km                        C. rodgersii in Tasmania is highly nocturnal and displays a
                                     homing behaviour so that grazing is largely manifest as
                                     halos radiating from crevices used for daytime shelter
                                     (reviewed by Andrew and Byrne 2001). Typical of the
                  Longitude  148           Tasmanian east coast, the incipient C. rodgersii barrens
                                     investigated occurred deeper than 8 m where a combina-
Fig. 1 Map of Tasmania showing the experimental site at Bicheno
                                     tion of wave action and mechanical abrasion by
(asterisk). Inset indicates the distribution of Centrostephanus rod-
gersii in southeastern Australia, solid line indicates New South Wales  macroalgae appears to determine the shallow limit of the
range (after Andrew and Byrne 2001), broken line indicates range     barrens (Johnson et al. 2005) and also the passage by
extension to Tasmania. On the main map of Tasmania, solid grey line   urchins between neighbouring patches.
indicates range over which barrens patches can be found, broken grey
                                      The temporal response of the algal community follow-
line indicates range where individuals, but not barrens patches, have
been observed (after Johnson et al. 2005; J. Valentine, personal     ing C. rodgersii removal was assessed using a
communication)                              non-destructive spatially nested sampling design consisting


123
Oecologia (2008) 156:883–894                                              885


of three replicate incipient barrens patches within each    cover of various taxa (and bare rock) to be estimated from
treatment (removal vs. unmanipulated control) and four     the photographs by enumerating the taxa present at the 100
replicate quadrats (0.25 m2) haphazardly sampled within    equidistant points defined by the intersecting lines. Taxa
each patch on each sampling occasion. Manipulations      were identified to species level where possible; otherwise,
commenced in spring 2003 (19 November) with a single      functional groups were used, such as foliose red, filamen-
pair of barren patches (depth 9–10 m) randomly assigned    tous red, filamentous brown and filamentous green algae. It
as either urchin removal or unmanipulated control. The     was difficult to assess the cover of non-geniculate coralline
four additional barrens patches (approx. 150 m from ori-    algae due to the loss of pigmentation in some plants
ginal site in slightly deeper water, depth 14–15 m) were    because it was generally unknown whether bleached areas
discovered 4 months later. To reduce possible bias of     were living or dead. Thus, for the purpose of this study,
seasonal variability on algal recruitment on algal recovery,  encrusting coralline algae included both bleached and
the additional barrens patches were manipulated during the   pigmented components.
following spring (24 November 2004). These patches were
monitored over the preceding 8 months prior to random     Habitat and faunal structure of ‘recovered’ macroalgal
assignment of the urchin removal treatment to two of the    beds—destructive sampling
patches. Thus, there was a total of three replicate urchin
removal and three replicate control patches. Treatments    To assess the impact of C. rodgersii barrens on reef
were maintained and patches sampled approximately every    structure and associated fauna, all experimental patches
2 months for a period of about 36 months. The response of   were sampled destructively at the end of the experiment
canopy-forming algal species was of a priori interest, and   after the macroalgal canopy had re-established. On termi-
comparison among treatments was planned at 6, 12 and      nation of the experiment in November 2006, the original
24 months post-removal of C. rodgersii.            urchin removal patch had experienced 36 months of
  Reflecting the spatially circumscribed nature of patches,  recovery, whereas the additional manipulated patches had
the limited movement of adult C. rodgersii and the       experienced 24 months post-removal of C. rodgersii. Thus,
apparent low recruitment of juveniles over the duration of   the destructive sampling design at the conclusion of the
the study, there was minimal re-invasion of patches from    experiment captured both spatially and temporally variable
which urchins had been removed (i.e. fewer than ten      components across patches nested within the urchin
individuals were required to be removed during routine     removal treatment. To enable a comparison of habitat and
maintenance of the urchin removal treatment as compared    faunal structure between the urchin removal treatment and
to a total of 169 urchins removed during the initial appli-  adjacent intact macroalgal beds of similar topography
cation of the treatment). Other large benthic herbivores    (boulder reef) and depth (9–15 m) but unaffected by
present on the study reef included the sea urchin Helio-    C. rodgersii (i.e. where grazing had not been observed for
cidaris erythrogramma, the lucrative blacklip abalone     at least 7 years), the nested experimental design was
(Haliotis rubra), albeit rarely, and the herbivorous fish    extended to include three adjacent ‘intact macroalgal bed’
Odax cyanomelas. In particular, H. erythrogramma        patches. Selected by randomised fin kicks and compass
occurred commonly within barrens patches occupied by      directions, these patches were destructively sampled upon
C. rodgersii; however, this endemic species is not known    termination of the urchin removal experiment. Hence, the
to form barrens at exposed sites in eastern Tasmania      extended design included three levels of ‘treatment’
(Johnson et al. 2005). In notable contrast to sea urchin    (unmanipulated barrens, sea urchin removal and unma-
barrens within the historic NSW range, large limpets      nipulated intact macroalgal beds), three levels of ‘patch’
([20 mm) were found to be absent on C. rodgersii barrens    nested within ‘treatment’ and four quadrats within each
in eastern Tasmania. Benthic herbivores other than       ‘patch’, providing an estimate of error.
C. rodgersii were not manipulated as part of this study.     Because the routine photoquadrat monitoring provided
                                only a 2D representation of the substratum, patterns in the
Monitoring algal response                   total habitat structure (i.e. inclusive of macroalgal canopy,
                                understorey and basal substratum components) were
On each sampling occasion, the four replicate 0.25-m2     examined in detail at the final assessment by sequentially
quadrats within each experimental patch were photo-      photographing and then destructively sampling each stra-
graphed to obtain a planar two dimensional (2D) image of    tum from top to bottom. The abundance and total length of
the benthos. Reference to subsurface buoys ensured that    individual canopy-forming macroalga within each quadrat
sampling occurred only within the original boundaries of    was also measured, and the total algal biomass of macro-
the barrens patches. Each quadrat was dissected by a grid   algal canopy and understorey strata was calculated from dry
of 10 9 10 equidistant lines, which enabled the percentage   weights obtained by drying algal samples at 70°C for 48 h.


                                                         123
886                                                 Oecologia (2008) 156:883–894


The cover of encrusting and structural invertebrates was      pi is the proportion of the community that belongs to the Ith
estimated from photographs once the overlying algae had      taxa. The significance of differences in assemblage struc-
been removed. Associated benthic fauna were sampled        ture was estimated using permutational multivariate
from each quadrat by sealing underwater all excised mac-      analysis of variance (PERMANOVA; Anderson 2001,
roalgal habitat and structural invertebrates in plastic bags,   2005; McArdle and Anderson 2001). All PERMANOVA
while the remaining benthic fauna were extracted from the     tests were based on 9999 permutations of Bray–Curtis
substratum using a venturi suction sampler connected to 1-     dissimilarity matrices generated from non-standardised
mm mesh bags (each quadrat was systematically suctioned      fourth-root transformed data. Significant terms were
for 3 min at a flow rate of 180 L min-1). Fauna contained      investigated with a posteriori pairwise comparisons using
within algal habitat and/or benthic suction samples were      the PERMANOVA t statistic based on distances of the
extracted by thoroughly agitating samples in seawater       correct permutable units. Corrections for type-I error rate
before passing them through a 1-mm sieve. Faunal samples      were made using the Dunn–Sidak method.
were then sorted and enumerated at the species level where
possible. Where species could not be identified, individual
specimens were assigned to taxonomic groups based on the      Results
finest taxonomic resolution possible.
                                  Recovery of macroalgal habitat
Analysis
                                  In all incipient barrens patches from which C. rodgersii was
Univariate analyses The effects of C. rodgersii removal      removed, a structurally complex assemblage of foliose
were analysed with a temporal series of one-way Model III     algae developed that was ultimately dominated by the
nested analysis of variance (ANOVA) of factors ‘treatment’     canopy-forming species Ecklonia radiata and Phyllospora
(fixed effect) and ‘patch nested within treatment’ (random     comosa (Fig. 2). Filamentous algae and macroalgal sporo-
effect) at the pre-planned (a priori) times of interest of 0, 6,  phytes (height \20 mm) recruited to available space and
12 and 24 months post-sea urchin removal. Data collected      began to overgrow the substratum within 1 month of the sea
by destructive sampling at the conclusion of the experiment    urchins being removed. Effects of C. rodgersii removal on
were analysed with the same nested ANOVA structure         the areal cover of canopy-forming macroalgae were statis-
except for the addition of a third level of treatment, the     tically detectable at all pre-planned times (approx. 6, 12 and
‘intact macroalgal bed’. All univariate statistical analyses    24 months) after removal of the sea urchin (Table 1). The
were undertaken using SASÒ (v. 6.12; SAS Institute, Cary      pattern of re-colonisation for E. radiata (by cover) occurred
NC), and data were checked for conformity to assumptions      consistently across C. rodgersii removal patches, while
of homoscedasticity and normality. Where data were het-      significant between-patch variability was detected for
eroscedastic, the transformation to stabilise variances was    P. comosa and total canopy cover (Table 1).
determined by the relationship between group standard         Assessment of benthic habitat structure by destructive
deviations and means (Draper and Smith 1981). The         sampling revealed that removal of C. rodgersii resulted in
appropriate transformation for each variable is expressed in    the replacement of the open barrens substratum with a
terms of the untransformed variable Y. Where lower levels     structurally heterogeneous benthic habitat composed of a
of nesting revealed non-significant results (P [ 0.25, Winer    macroalgal canopy and accompanying understorey (Fig. 3).
et al. 1991), data at higher levels were pooled a posteriori to  Nested ANOVA revealed significant differences in both
provide a more powerful test of lower order terms. Multiple    canopy and understorey components between treatment
range tests were conducting using the Ryan–Einot–Gabriel–     groups, with pairwise comparisons revealing differences
Welsch (REGW) procedure. Size–frequency distributions of      between urchin removal and unmanipulated barrens, but not
canopy-forming macroalgae, as assessed at the termination     between urchin removal and intact macroalgal beds
of the experiment, were compared between ‘urchin removal’     (Fig. 3). Patterns in the cover of the basal substratum layer
and ‘intact macroalgal beds’ using the Kolmogorov–Smir-      varied between treatment groups (Fig. 3), showing consis-
nov test.                             tency across treatments in the cover of encrusting corallines
                                  but significantly higher cover of bare rock and filamentous
Multivariate analyses Comparisons of communities were       algae on the unmanipulated barrens in comparison to the
visualised using nMDS ordination, and the species con-       urchin removal and intact macroalgal treatments.
tributing most to dissimilarity were revealed using the        Canopy-forming macroalgae occurred at a higher abun-
SIMPER software routine (PRIMER 5, ver. 5.2.9). Taxo-       dance within the urchin removal treatment than on the
nomic diversity of each sample was calculated using the
             À           Á          barrens habitat or intact macroalgal beds; whereas in barrens
                 P
Shannon Diversity Index H 0 ¼ À s pi loge pi ; where
                  i¼1              patches there were low numbers of minute (length\50 mm)


123
Oecologia (2008) 156:883–894                                                               887

Fig. 2 Response of incipient                a  Eck lonia radiata                b  Encrusting corallines
                           100                     Urchin
barren patches to the removal of
Centrostephanus rodgersii                                     removal
                            80
showing percentage cover                                     Control
(mean ± SE) by canopy-                 60
forming macroalgae (a, c, e) and            40
by the basal substratum layer
(b, d, f). Urchin removal               20
treatment (filled circles, n = 3)
                            0
and unmanipulated ‘control’
patches (open circles, n = 3)
are shown. Dashed vertical line            100  c Phyllospora comosa                d  Filamentous algae
in each column shows the
                   Percent cover  80
timing of urchin removal;
negative values on the X-axis             60
represent months prior to
removals, and vertical arrows             40
indicate pre-planned times of             20
interest for analysis of the
urchin removal effect. Note:              0
n = 2 for months prior to
removals                       100  e  Total canopy                  f  Bare rock

                            80

                            60

                            40

                            20

                            0
                              -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20 22 24  -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20 22 24

                                           Months post C. rodgersii removal
Table 1 Results of nested Model III ANOVA testing of the effect of Centrostephanus rodgersii removal on macroalgal cover at pre-planned
months post-removal
Response variable                        Source         Orthogonal pre-planned comparisons
(transformation 6, 12, 24 months)
                                            T = approx. 6 monthsa      T = 12 months    T = 24 months

Ecklonia radiata                        Treatment        F1,22 = 53.15          F1,4 = 33.97    F1,22 = 80.22
 [log(Y + 0.0001), Y0.28, Y0.24]                            P \ 0.0001*           P = 0.0043*     P \ 0.0001*
                                Patch (treatment)    F4,18 = 0.54          F4,18 = 2.28    F4,18 = 0.51
                                            P = 0.7070           P = 0.1006     P = 0.7317
Phyllospora comosa                       Treatment        F1,4 = 585.85          F = 7.87      F1,22 = 33.34
 [log(Y + 0.0001), Y0.45, log(Y + 0.0001)]                       P \ 0.0001*           P = 0.0485*     P \ 0.0001*
                                Patch (treatment)    F4,18 = 5.19          F4,18 = 10.07    F4,18 = 0.68
                                            P = 0.0059*           P = 0.0002*     P = 0.6133
Total canopy macroalgae [, Y0.67,]               Treatment        F1,4 = 14.08          F1,4 = 10.07    F1,4 = 17.21
                                            P = 0.0199*           P = 0.0338*     P = 0.0143*
                                Patch (treatment)    F4,18 = 5.89          F4,18 = 9.45    F4,18 = 15.35
                                            P = 0.0033*           P = 0.0003*     P \ 0.0001*
* Significant P values
Note: prior to applying the C. rodgersii removal treatment, no differences in the cover of barrens substratum components were detected between
incipient patches (encrusting corallines, F1,4 = 0.1200, P = 0.7418; bare rock [trans. = Y0.22], F1,4 = 0.1247, P = 0.8030; filamentous algae,
F1,4 = 0.01 P = 0.9460)
a
  The appox. 6-month sample was only attainable at 7 months post manipulation

individuals, macroalgae in the urchin removal patches were              relatively more larger individuals (Fig. 4). A comparison of
dominated by small individuals tailing to large-sized classes,            macroalgal size–frequency distribution between urchin
and in the intact macroalgal habitat, there were fewer but              removal and intact macroalgal beds revealed significantly


                                                                        123
888                                                                Oecologia (2008) 156:883–894

Fig. 3 Percentage cover                   Canopy
(mean ± SE) of unmanipulated             100  a  Ecklonia radiata         b Phyllospora comosa        c  Total canopy
barrens (incipient barrens),                                                             a    a
barrens with urchin removal and            80
intact macroalgal beds for                        a      a
canopy (a, b, c), understorey             60
(d, e, f) and substratum (g, h, i)
components of benthic habitat             40                                 a
                                                        a
structure. Bars with identical            20
letters indicate Ryan–Einot–
Gabriel–Welsch (REGW)                     b                   b                  b
                            0
groupings of treatments for each
                              Understorey
response variable. a = 0.05
                   Percent cover  100  d  Foliose reds           e B rowns              f  Total understorey

                           80

                           60
                                                                           a
                           40                                                   a
                                           a             a
                                     a
                           20                                 a
                                b                   b                  b
                            0

                              Substratum
                           100  g  Encrust. corallines        h  Filamentous algae        i  Bare rock

                           80
                                a     a
                           60
                                           a
                           40
                                                                     a
                           20
                                                   a
                            0
                                                        b     b             b    b
                              Incipient  Urchin  Intact     Incipient  Urchin   Intact   Incipient   Urchin   Intact
                              barrens  removal  macro-     barrens  removal   macro-   barrens   removal   macro-
                                         algae                 algae                algae




different size distributions (Kolmogorov–Smirnoff,                    re-colonisation of this habitat by a benthic faunal assemblage
P \ 0.0001). For urchin removal patches, total algal bio-                vastly different to that of the barrens, but not different to that
mass m-2 (canopy plus understorey species, excluding                   observed in intact macroalgal beds (Fig. 6b; see Table 2 for
encrusting corallines) was not statistically different to that              PERMANOVA summaries). The removal of C. rodgersii
of intact macroalgal beds, but it was much greater than                 clearly increased taxonomic richness, total abundance and
for the barrens habitat [total algal biomass for urchin                 Shannon diversity of benthic fauna (independent of struc-
removal patches = 845.68 ± 451.81 (SE) g m-2, for intact                 ture-forming invertebrates); however, there was little
macroalgal beds = 844.22 ± 127.45 g m-2, for unmanip-                  difference in the composition of benthic faunal communities
ulated barrens = 0.20 ± 0.04 g m-2, nested Model III                   between urchin removal and intact macroalgal bed treat-
ANOVA; trans. = (log(Y + 0.0001), treatment, F(2,6) =                  ments (Fig. 7). The taxa contributing most to dissimilarity in
84.07, P \ 0.0001; patch (treatment), F(6,27) = 1.98, P =                faunal abundance between macroalgal bed and barrens
0.1035]. Other benthic structural components, namely                   habitats were Amphipoda (38.2%); Polychaeta (8.76%);
sessile encrusting and erect invertebrates, contributed to the              Isopoda (6.91%); Gastropoda (6.16%); Tanaidacea (3.95%);
physical structure of recovered macroalgal and intact mac-                Hirudinea (3.59%); Bivalvia (3.47%); Echinodermata
roalgal habitats, but they contributed little to the barrens               (3.06%); Mysidaceae (2.66%); Serpulidae (2.65%);
habitat (Fig. 5).                                    Decapoda (2.05%); Brachiopoda (1.80%); Terebellidae
                                             (1.68%); Oligochaeta (1.51%). Graphical examination of
Effect of barrens on taxonomic diversity                         whole benthic communities (flora and fauna), based on the
                                             presence/absence of all described taxa (including structure-
Recovery of canopy-forming macroalgae within C. rodgersii                forming invertebrates) revealed overwhelmingly different
removal patches (Fig. 6a) resulted in an associated                   benthic communities in the presence of C. rodgersii grazing


123
Oecologia (2008) 156:883–894                                                                      889


           a Incipient barrens                          a  Bryzoans -start         b Encrust. sponges -start
       6
                                            20

       4                                    15




                                                          Not sampled




                                                                             Not sampled
                                            10
       2




                                    Percent cover
                                            5
                                                                   a
       0                                                                a
                                            0
           b  Urchin removal
                                              c  Bryzoans -end          d  Encrust. sponges -end
       60                                                                     a
Frequency




                                            20
       40
                                            15

       20                                    10
                                                     a                   a
       0                                     5
                                                          a
                                                a                  b
           c  Intact macroalgae                       0
                                              Incipient  Urchin  Intact     Incipient  Urchin  Intact
       6                                       barrens  removal  macro-     barrens  removal  macro-
                                                         algae                 algae
       4
                                    Fig. 5 Effect of Centrostephanus rodgersii on the cover of habitat-
                                    forming bryzoans (a, c) and encrusting sponges (b, d) for incipient
       2
                                    barrens, urchin removal and intact macroalgal patches. Start of the
                                    experiment prior to sea urchin removal, bryzoans absent; sponge
       0                            cover (nested Model III ANOVA; trans. = Y0.69, ‘treatment’,
         0       500      1000  1500  2000
                                    F(1,5) = 1.09, P = 0.3548; ‘patch (treatment)’, F(4,18) = 24.09, P \
                    Algal length (mm)        0.0001). End of experiment, bryzoans (trans. = log(Y + 0.0001),
                                    ‘treatment’, F(2,6) = 3.23, P = 0.1116; ‘patch (treatment)’,
Fig. 4 Size–frequency distributions of canopy-forming macroalgal    F(6,27) = 6.02, P = 0.0004); sponges (nested Model III ANOVA;
species (Ecklonia radiata and Phyllospora comosa) at termination of  trans. = log(Y + 0.0001), ‘treatment’, F(2,33) = 11.05, P = 0.0002;
the experiment in incipient barrens patches (a; n = 5), Centrosteph-  ‘patch (treatment)’, F(6,27) = 0.85, P = 0.5423). Bars with identical
anus rodgersii removal patches (b; n = 287), intact macroalgal beds  letters indicate REGW groupings of treatments within each sampling
(c; n = 70). Note different scales for the Y-axis           period, a = 0.05


(Fig. 6c; see Table 2 for PERMANOVA summary). Of the          Andrew and Byrne 2001) and broadly typical of sea urchin
296 individual floral and faunal taxa recorded, only 72 were      ‘coralline’ barrens throughout the world (reviewed by
present within incipient barrens, 253 were present in the       Pinnegar et al. 2000). The removal of C. rodgersii from
urchin removal patches and 221 were recorded within intact       barrens patches in eastern Tasmania resulted in a rapid
macroalgal beds (see Appendix 1 of the Electronic Supple-       replacement of the flat structurally homogeneous substra-
mentary Material). Thus, the formation of barrens by          tum of the initial urchin barrens with a structurally
C. rodgersii is estimated to result in a minimum localised       heterogeneous 3D benthic habitat complete with macroal-
loss of approximately 150 taxa from with eastern Tasmanian       gal canopy, diverse algal understorey and structural basal
macroalgal beds.                            invertebrates. Indeed, the dramatic and consistent pattern
                                    of algal recovery across all urchin removal patches indi-
                                    cated that the timing of urchin removals from barrens
Discussion                               patches (September 2003 as opposed to September 2004)
                                    was unimportant. While patterns in canopy cover and algal
Effect of sea urchin range expansion on reef habitat          biomass clearly converged on that observed for intact
                                    macroalgal beds, recovering patches were still biased
Climate change is leading to a re-distribution of marine        towards smaller and yet more abundant plants, indicating
species and altering ecosystem dynamics (e.g. Harley et al.      that effects of previous grazing on community succession
2006; Rosenzweig et al. 2007). Within the newly extended        were still detectable [24 months after removal of the sea
eastern Tasmanian range of Centrostephanus rodgersii,         urchin. Most importantly, however, return to the macroal-
this sea urchin now deconstructs the macroalgal habitat and      gal-dominated ecosystem state (macroalgal canopy
maintains a simplistic and homogeneous benthic habitat         cover [50%) was achieved rapidly (within approx.
typical of barrens described from its endemic range (e.g.       15 months) after urchin removal (for comparison of algal


                                                                            123
890                                                   Oecologia (2008) 156:883–894


                                    foliose algae and often slower or less complete recovery of
      aA.
                                    canopy-forming species, a result consistently attributed to
                                    patterns in propagule supply (Fletcher 1987; Andrew 1991,
                                    1993, 1998; Hill et al. 2003). A notable difference in the
                                    barrens assemblage across eastern Tasmania is the general
                                    lack of limpet ‘mesograzers’ that occur in high abundances
                                    on C. rodgersii barrens in NSW and which are capable of
                                    delaying macroalgal recovery following C. rodgersii
                                    removal (Fletcher 1987). Thus, the functional redundancy
                      Stress= 0.08
                      Stress=0.08         of the grazer group on barrens throughout eastern Tasma-
                                    nia would likely be enhanced if limpets were to establish at
                                    high densities. While regional differences in macroalgal
      b.
      B.
      A.
      bB.                             growth rates and grazer interactions are likely, experiments
                                    on NSW reefs have in general been undertaken on, or near,
                                    widespread barrens habitat. Conversely, I manipulated
                                    small incipient barrens patches (scale of metres) sur-
                                    rounded by reef dominated by dense macroalgal habitat,
                                    which likely provided a saturating supply of algal propa-
                                    gules at this scale. Therefore, direct scaling-up of these
                                    results is likely to lead to over-expectations of macroalgal
                       Stress= 0.10
                       Stress=0.10        recovery rates for larger scale barrens (102–103 m) where
                                    algal propagule supply may become limiting (reviewed by
                                    Dayton 1985). Unlike the dynamic recovery of macroalgal
      cC.                             habitat following C. rodgersii removal, un-manipulated
                                    barrens patches displayed a high stability over the 3-year
                                    duration of the study. In combination with in situ obser-
                                    vations at several sites over [8 years (author, personal
                                    observations), C. rodgersii barrens in eastern Tasmania
                                    appear to constitute a truly alternative and persistent state,
                                    as also reported for conspecific barrens in NSW (reviewed
                                    by Andrew and Byrne 2001).
                      Stress=0.10
                      Stress= 0.10
                                    Effect of sea urchin grazing on taxonomic diversity
                                    within the expanded range
Fig. 6 Ordinations (nMDS) showing the effect of Centrostephanus
rodgersii on benthic algal assemblages (a), benthic faunal assem-
blages (b) and entire benthic assemblages (flora plus fauna) (c) at   Examination of the benthic fauna in barrens patches con-
termination of experiment. Symbols represent individual quadrats    firmed the major effects of C. rodgersii grazing that extend
nested within replicate barrens patches (crosses), urchin removal    to the entire benthic community. While C. rodgersii is
patches (triangles) and intact macroalgal beds (circles). Ordinations
are based on Bray–Curtis similarity matrices obtained from fourth-
                                    known to be omnivorous, consuming encrusting and
root transformed percentage cover data for algae, from abundance    structure-forming invertebrates as well as algae (A. Pile,
data for faunal assemblages and from the presence/absence data for   personal communication; author, personal observation), the
whole benthic assemblages. Faunal and whole assemblage ordinations   greatest faunal impacts by C. rodgersii appear to be those
are overlaid with a bubbleplot (grey) representing macroalgal canopy
cover (largest bubbles represent 100% macroalgal canopy cover);
                                    caused by the loss of macroalgal habitat due to intense
dashed ellipses encompass the space occupied by the alternative     herbivory. Indeed, the barrens state is characterised by an
assemblages of barrens and macroalgal ecosystem states         impoverished benthic community, with approximately 150
                                    taxa fewer than adjacent macroalgal beds (also see
recovery in other systems, see Duggins 1980; Himmelman         Himmelman et al. 1983; Bodkin 1988; Graham 2004).
et al. 1983; Keats et al. 1990; Johnson and Mann 1993;         When the potential number of species that are either directly
Leinass and Christie 1996).                       consumed by sea urchins or simply associated with the
  In contrast to the rapid and consistent pattern of mac-       macroalgal habitat (e.g. Graham 2004) are considered, the
roalgal recovery observed in the current study,             total number of taxa potentially impacted by C. rodgersii
experimental removals of C. rodgersii in NSW have            grazing in eastern Tasmania may increase dramatically. As
resulted in a less predictable transition to assemblages of       an example, intensive grazing by C. rodgersii eliminates


123
Oecologia (2008) 156:883–894                                                        891


Table 2 PERMANOVA testing the effect of Centrostephanus rodgersii on algal, faunal and entire benthic assemblages at the conclusion of the
experiment

PERMANOVA                 Algal assemblage         Faunal assemblage          Whole benthic assemblage
Source             df     F       P (perm)      F          P (perm)    F        P (perm)

Treatment            2     15.87     0.0129*      6.41        0.0096*     7.33       0.0076*
Patch (treatment)        6      2.76     0.0071*      1.69        0.0130*     1.70       0.0186*
Residual            27
Total             35
Tests among ‘treatment’
Groups               Unique perm.     t        P (MC)       t       P (MC)     t      P (MC)

Barrens vs. removal         10          4.46      0.0015*      2.80      0.0097*    3.02     0.0055*
Barrens vs. intact         10          5.34      0.0009*      2.71      0.0098*    2.93     0.0081*
Removal vs. intact         10          1.02      0.3996       1.12      0.3170     1.13     0.3207

Average Bray–Curtis percentage dissimilarities within and between treatments: macroalgal, faunal and whole assemblages
           Barrens            Removal            Intact

Barrens       21.88; 46.98; 43.90
Removal       64.85; 69.48; 67.95      25.38; 25.00; 23.72
Intact        65.31; 68.92; 68.44      21.06; 27.05; 26.32      18.21; 27.21; 26.69

*Significant values. Pair-wise a posteriori comparisons were made after adjusting the type I error rate, a = 0.017
Results are given for one-way mixed model nested PERMANOVA, tests among treatments and dissimilarities within and between treatments.
For the pair-wise tests, Monte Carlo (MC) asymptotic P values were used given the small number of unique permutations (after Anderson 2005)


almost all algal species, of which there are an estimated 373        result in negative impacts for nektonic species that asso-
species in Tasmanian coastal waters alone (reviewed by           ciate with macroalgal habitat either as a result of direct
Sanderson 1997). Thus, one may expect that the rate of           habitat loss or the loss of an abundance of prey items
species accumulation with increasing sampling area (the           associated with vegetated habitats (e.g. Edgar and Shaw
species–area curve) is likely to be much greater for heter-         1995). While the spatial grain of the current study can be
ogeneous macroalgal habitat relative to homogenous             considered to be too small for an adequate examination of
barrens where a consistent community containing relatively         the effects of C. rodgersii barrens on fish assemblages
few species is observed.                          (reef fish in Tasmania typically possess home ranges
  In a similar study by Vance (1979) in California, over-         [2000 m2; Barrett 1995), of the few small cryptic fishes
grazing of the macroalgal habitat by the congeneric             (length \100 mm) sampled from the benthos (a total of
Centrostephanus coronatus also dramatically decreased            15 individuals in seven taxonomic groups), none were
local taxonomic diversity. Interestingly, the author con-          recorded from the barrens.
sidered that a patchwork of grazed patches among
macroalgal habitat may have the net effect of increasing           Effects of barrens on ecosystem functioning
the diversity of the community as a whole because local-
ised barrens patches may provide a habitat for grazer            As evidenced by the dramatic recovery of standing stocks
resistant taxa that were otherwise rarely observed. While          in algal biomass and associated benthic fauna, vast changes
there were few taxa (less than six) that were unique to the         in the physical and community structure of rocky reefs
barrens patches studied in eastern Tasmania (other than           occur with the transition from macroalgal beds to C. rod-
C. rodgersii itself; see Appendix I of the Electronic Sup-          gersii barrens. What remains less clear is how such shifts
plementary Material), it is clearly the catastrophic shift to        impact ecosystem functioning. However, given that epi-
widespread barrens (102–103 m), via the coalescence of            fauna are known to be major contributors to the flux of
incipient barrens patches, that will lead to the loss of           materials in macroalgal dominated reef habitats (e.g.
taxonomic diversity across increasingly large and eco-            Taylor 1998), the loss of fauna on barrens implies major
logically important spatial scales. Furthermore, the             functional differences between alternative macroalgal and
formation of C. rodgersii barrens may also be expected to          barrens states. Ultimately, the conversion of macroalgal


                                                                  123
892                                                                          Oecologia (2008) 156:883–894

     Taxonomic richness         200                                 independently on individual species within a community
                            a                             (e.g. Parmesan and Yohe 2003). Thus, the loss of local
                                          a
                       150                           a      habitat as a result of range extension by habitat-modifying
                                                         organisms coupled with large-scale shifts in the suitable
                       100                                 ‘climate envelope’ (e.g. Hijmans and Graham 2006) may
                                b                         be particularly devastating for some populations, particu-
                          50
                                                         larly those with contracted ranges to begin with. These
                                                         kinds of interactions are acutely relevant in places such as
                          0
                                                         Tasmania where poleward range retreat is prevented by a
                                                         lack of contiguous poleward land mass. Indeed, the large-
Total no. individuals




                  3000         b             a
                                                         scale decline of the giant kelp Macrocystis pyrifera in
                                                   a      eastern Tasmania over the past 50 years appears to be the
                  2000
                                                         result of the new regime of warm, nutrient-poor water on
                                                         this coast (e.g. Edgar et al. 2005; see also Ridgway 2007).
                  1000             b
                                                         While C. rodgersii grazing does not appear to be respon-
                                                         sible for the decline of this macroalga over large scales,
                          0                               localised barrens formation may prevent the recovery of
                                                         this alga at some sites even if poor nutrient conditions for
             Shannon Diversity Index




                            c                             plant growth were temporally reversed. Moreover, because
                          4                a         a      further strengthening of the EAC and greater thermal
                                                         stratification are predicted for southeastern Australia under
                          3
                                b                         global climate change (Cai et al. 2005), coastal waters off
                          2                               eastern Tasmania appear to be committed to a warm and
                          1
                                                         oligotrophic trajectory (reviewed by Poloczanska et al.
                                                         2007). This trend will have a positive effect on the repro-
                          0                               ductive success of C. rodgersii (Ling et al. 2008) but will
                            Incipient barrens  Urchin removals  Intact macroalgae
                                                         negatively influence macroalgal growth and likely result in
Fig. 7 Effect of Centrostephanus rodgersii on benthic faunal diver-                        more frequent dieback events (e.g. Valentine and Johnson
sity assessed at the end of the experiment on incipient barren patches,                      2004). Thus, the warming climate of this coast appears
urchin removal patches and intact macroalgal beds. Data shown are
means per square metre ± SE and do not include habitat-forming
                                                         poised to tilt macroalgal–urchin dynamics in favour of
invertebrates. a Taxonomic richness, i.e. total number of taxa (Model                       further sea urchin grazing and disproportionately large
III one-way nested ANOVA; trans. = Y0.5, Treatment, F(2,6) =                           effects on reef biodiversity.
125.47, P \ 0.0001; patch (treatment), F(6,27) = 1.71, P = 0.1575).
b Total number of individuals (trans. = Y0.22, treatment, F(2,27) =                        Acknowledgments I thank many dive volunteers who assisted with
57.45, P \ 0.0001; patch (treatment), F(6,27) = 1.34, P = 0.2728).                        the fieldwork, particularly Anthony Reid, Dave Stevenson, Adam
c Shannon diversity index (treatment, F(2,33) = 123.42, P \ 0.0001;                        Stephens and Ryan Downie. I am grateful for the assistance in faunal
patch (treatment), F(6,27) = 0.56, P = 0.7589). Note that the index                        identification received from Graham Edgar. This work was supported
was calculated using loge. Bars with identical letters indicate REGW                       by the School of Zoology and the Tasmanian Aquaculture and
groupings, a = 0.05                                                Fisheries Institute—University of Tasmania, plus the CSIRO-UTAS
                                                         joint programme in Quantitative Marine Science. This manuscript
                                                         was improved by comments received from Craig Johnson and Joseph
beds to widespread C. rodgersii barrens within the exten-
                                                         Valentine.
ded Tasmanian range is anticipated to reduce benthic
primary (after Chapman 1981; Babcock et al. 1999) and
secondary productivity (e.g. Duggins et al. 1989), with                              References
flow-on effects to many species, including species of
commercial importance (Andrew and Underwood 1992;                                 Anderson MJ (2001) A new method for non-parametric multivariate
Andrew et al. 1998; Johnson et al. 2005).                                       analysis of variance. Austral Ecol 26:32–46
                                                         Anderson MJ (2005) PERMANOVA: a FORTRAN computer
                                                           program for permutational multivariate analysis of variance.
Habitat loss coupled with changing climate                                      Department of Statistics, University of Auckland, New Zealand
                                                         Andrew NL (1991) Changes in subtidal habitat following mass
The results presented here demonstrate that dispropor-                                mortality of sea urchins in Botany Bay, New South Wales. Aust
                                                           J Ecol 16:353–362
tionately large impacts on marine biodiversity may occur
                                                         Andrew NL (1993) Spatial heterogeneity, sea urchin grazing, and
as a result of range shifts of key habitat-modifying                                 habitat structure on reefs in temperate Australia. Ecology
organisms. Furthermore, climate change also acts                                   74:292–302



123
Oecologia (2008) 156:883–894                                                        893

Andrew NL, Byrne M (2001) The ecology of Centrostephanus         Harley CDG, Hughes RA, Hultgren KM, Hultgren KM, Miner BG,
   rodgersii. In: Lawrence JM (ed) Edible sea urchins: biology and     Sorte CJB, Thornber CS, Rodriguez LF, Tomanek L, Williams
   ecology. Elsevier, Amsterdam, pp 149–160                 SL (2006) The impact of climate change in coastal marine
Andrew NL, O’Neill AL (2000) Large-scale patterns in habitat         systems. Ecol Lett 9:228–241
   structure on subtidal rocky reefs in New South Wales. Mar      Hijmans RJ, Graham CH (2006) The ability of climate envelope
   Freshw Res 51:255–263                          models to predict the effect of climate change on species
Andrew NL, Underwood AJ (1992) Associations and abundance of         distributions. Glob Chang Biol 12:2272–2281
   sea urchins and abalone on shallow subtidal reefs in southern    Hill NA, Blount C, Poore AGB, Worthington D, Steinberg P (2003)
   New South Wales. Aust J Mar Freshw Res 43:1547–1559           Grazing effects of the sea urchin Centrostephanus rodgersii in
Andrew NL, Underwood AJ (1993) Density-dependent foraging in the       two contrasting rocky reef habitats: effects of urchin density and
   sea urchin Centrostephanus rodgersii on shallow subtidal reefs in    its implications for the fishery. Mar Freshw Res 54:691–700
   New South Wales, Australia. Mar Ecol Prog Ser 99:89–98       Himmelman JH, Cardinal A, Bourget E (1983) Community develop-
Andrew NL, Worthington DG, Brett PA, Bentley N, Chick RC,           ment following removal of urchins, Strongylocentrotus
   Blount C (1998) Interactions between the abalone fishery and sea     droebachiensis, from the rocky subtidal zone of the St. Lawrence
   urchins in New South Wales. FRDC Final Report, Project No.        Estuary, Eastern Canada. Oecologia 59:27–39
   93/102. Fisheries Research and Development Corporation,       Hughes L (2000) Biological consequences of global warming: is the
   Deakin West, ACT, Australia                       signal already apparent across natural systems? TREE 15:56–61
Babcock RC, Kelly S, Shears NT, Walker JW, Willis TJ (1999)       Hughes TP, Bellwood DR, Folke C, Steneck RS, Wilson J (2005)
   Changes in community structure in temperate marine reserves.       New paradigms for supporting the resilience of marine ecosys-
   Mar Ecol Prog Ser 189:125–134                      tems. TREE 20:380–386
Barrett NS (1995) Short and long-term movement patterns of six      Johnson CR, Mann KH (1993) Rapid succession in subtidal
   temperate reef fishes (Families: Labridae and Monacanthidae).       understorey seaweeds during recovery from overgrazing by sea
   Mar Freshw Res 46:853–860                        urchins in eastern Canada. Bot Mar 36:63–77
Bodkin JL (1988) Effects of kelp forest removal on associated fish    Johnson CR, Ling SD, Ross J, Shepherd S, Miller K (2005)
   assemblages in central California. J Exp Mar Biol Ecol 117:227–     Establishment of the long-spined sea urchin (Centrostephanus
   238                                   rodgersii) in Tasmania: first assessment of potential threats to
Cai WJ, Shi G, Cowan T, Bi D, Ribbe J (2005) The response of         fisheries. FRDC Final Report, Project No. 2001/044. Fisheries
   southern annular mode, the East Australian Current, and the       Research and Development Corporation, Deakin West, ACT,
   southern midlatitude ocean circulation to global warming.        Australia
   Geophys Res Lett 32:L23706                     Keats DW, South GR, Steele DH (1990) Effects of an experimental
Chapman ARO (1981) Stability of sea urchin dominated barren          reduction in grazing green sea urchins on a benthic macroalgal
   grounds following destructive grazing of kelp in St. Margaret’s     community in eastern Newfoundland. Mar Ecol Prog Ser
   Bay, Eastern Canada. Mar Biol 62:307–311                 68:181–193
Dayton PK (1985) Ecology of kelp communities. Annu Rev Ecol       Leinass HP, Christie H (1996) Effects of removing sea urchins
   Evol Syst 16:215–245                           (Strongylocentrotus droebachiensis): stability of the barren state
Draper N, Smith H (1981) Applied regression analysis. Wiley, New       and succession of kelp forest recovery in the east Atlantic.
   York                                   Oecologia 105:524–536
Duggins DO (1980) Kelp beds and sea otters: an experimental       Ling SD, Johnson CR, Frusher S, King CK (2008) Reproductive
   approach. Ecology 61:447–453                       potential of a marine ecosystem engineer at the edge of a newly
Duggins DO, Simenstad CA, Estes JA (1989) Magnification of           expanded range. Glob Chang Biol 14:907–915
   secondary production by kelp detritus in coastal marine ecosys-   McArdle BH, Anderson MJ (2001) Fitting multivariate models to
   tems. Science 245:170–173                        community data: a comment on distance based redundancy
Edgar GJ (1997) Australian marine life. Reed, Kew               analysis. Ecology 82:290–297
Edgar GJ, Shaw C (1995) The production and trophic ecology of      Parmesan C, Yohe G (2003) A globally coherent fingerprint of
   shallow-water fish assemblages in southern Australia. II. Diets of    climate change. Nature 421:37–42
   fishes and trophic relationships between fishes and benthos at    Pinnegar JK, Polunin NVC, Francour P, Badalamenti F, Chemello R,
   Western Port. Victoria. J Exp Mar Biol Ecol 194:83–106          Harmelin-Vivien ML, Hereu B, Milazzo M, Zabala M, D’Anna
Edgar GJ, Barrett NS, Morton AJ, Samson CR (2004) Effects of algal      G, Pipitone C (2000) Trophic cascades in benthic marine
   canopy clearance on plant, fish and macroinvertebrate commu-       ecosystems: lessons for fisheries and protected-area manage-
   nities on eastern Tasmanian reefs. J Exp Mar Biol Ecol 312:67–      ment. Environ Conserv 27:179–200
   87                                 Poloczanska ES, Babcock RC, Butler A, Hobday AJ, Hoegh-
Edgar GJ, Samson CR, Barrett NS (2005) Species extinction in the       Guldberg O, Kunz TJ, Matear R, Milton DA, Okey TA,
   marine environment: Tasmania as a regional example of          Richardson AJ (2007) Climate change and Australian marine
   overlooked losses in biodiversity. Conserv Biol 19:1294–1300       life. Oceanogr Mar Biol Annu Rev 45:409–480
Elmqvist T, Folke C, Nystrom M, Peterson G, Bengtsson J, Walker B,    Ridgway KR (2007) Long-term trend and decadal variability of the
   Norberg J (2003) Response diversity and ecosystem resilience.      southward penetration of the East Australian Current. Geophys
   Front Ecol Environ 1:488–94                       Res Lett 34:L13613. doi:10.1029/2007GL030393
Fletcher WJ (1987) Interactions among subtidal Australian sea      Rosenzweig C, Casassa G, Karoly DJ, Imeson A, Liu C, Menzel A,
   urchins, gastropods and algae: effects of experimental removals.     Rawlins S, Root TL, Seguin B, Tryjanowski P (2007) Assess-
   Ecol Monogr 57:89–109                          ment of observed changes and responses in natural and managed
Folke C, Carpenter S, Walker B, Scheffer M, Elmqvist T, Gunderson       systems. In: Parry ML, Canziani OF, Palutikof JP, van der
   L, Holling CS (2004) Regime shifts, resilience, and biodiversity     Linden PJ, Hanson CE (eds) Climate change 2007: impacts,
   in ecosystem management. Annu Rev Ecol Evol Syst 35:557–81        adaptation and vulnerability. Contribution of Working Group II
Graham MH (2004) Effects of local deforestation on the diversity and     to the Fourth Assessment Report of the Intergovernmental Panel
   structure of southern California giant kelp forest food webs.      on Climate Change. Cambridge University Press, Cambridge, pp
   Ecosystems 7:341–357                           79–131


                                                                  123
894                                                     Oecologia (2008) 156:883–894

Sanderson JC (1997) Subtidal macroalgal assemblages in temperate    Valentine JP, Johnson CR (2004) Establishment of the introduced
  Australian coastal waters. State of the Environment Technical      kelp Undaria pinnatifida following dieback of the native
  Paper Series (Estuaries and the Sea). Department of the         macroalga Phyllospora comosa in Tasmania, Australia. Mar
  Environment, Canberra                          Freshw Res 55:223–230
Scheffer M, Carpenter S, Foley JA, Folke C, Walker B (2001)       Vance RR (1979) Effects of grazing by the sea urchin, Centrosteph-
  Catastrophic shifts in ecosystems. Nature 413:591–596          anus coronatus, on prey community composition. Ecology
Taylor RB (1998) Density, biomass and productivity of animals in      60:537–546
  four subtidal rocky reef habitats: the importance of small mobile  Winer BJ, Brown DR, Michels KM (1991) Statistical principles in
  invertebrates. Mar Ecol Prog Ser 172:37–51               experimental design, 3rd edn edn. McGraw Hill, New York




123