The Stability of Boundary Regions between Kelp Beds and Deforested Areas
Author(s): Brenda Konar and James A. Estes
Source: Ecology, Vol. 84, No. 1, (Jan., 2003), pp. 174-185
Published by: Ecological Society of America
Stable URL: http://www.jstor.org/stable/3108007
Accessed: 09/07/2008 22:25


Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.


Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=esa.


Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.




JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org.




                                               http://www.jstor.org
Ecology, 84(1), 2003, pp. 174-185
C 2003 by the Ecological Society of America




    THE STABILITY OF BOUNDARY REGIONS BETWEEN KELP BEDS
            AND DEFORESTED AREAS
                       BRENDA  KONAR"12 AND JAMES A. ESTES'

 U.S. Geological Survey and Department of Biology, A-316 Earth and Marine Sciences Building, University of California,
                     Santa Cruz, California 95064 USA

         Abstract. Two distinct organizational states of kelp forest communities, foliose algal
       assemblages and deforested barren areas, typically display sharp discontinuities. Mecha-
       nisms responsible for maintaining these state differences were studied by manipulating
       various features of their boundary regions. Urchins in the barren areas had significantly
       smaller gonads than those in adjacent kelp stands, implying that food was a limiting resource
       for urchins in the barrens. The abundance of drift algae and living foliose algae varied
       abruptly across the boundary between kelp beds and barren areas. These observations raise
       the question of why urchins from barrens do not invade kelp stands to improve their fitness.
       By manipulating kelp and urchin densities at boundary regions and within kelp beds, we
       tested the hypothesis that kelp stands inhibit invasion of urchins. Urchins that were ex-
       perimentally added to kelp beds persisted and reduced kelp abundance until winter storms
       either swept the urchins away or caused them to seek refuge within crevices. Urchins invaded
       kelp bed margins when foliose algae were removed but were prevented from doing so when
       kelps were replaced with physical models. The sweeping motion of kelps over the seafloor
       apparently inhibits urchins from crossing the boundary between kelp stands and barren
       areas, thus maintaining these alternate stable states. Our findings suggest that kelp stands
       are able to defend themselves from their most important herbivores by combining their
       flexible morphology with the energy of wave-generated surge. The inhibitory influence of
       this interaction may be an important mechanism maintaining the patchwork mosaics of
       barren areas and kelp beds that characterize many kelp forest ecosystems.
         Key words:    Aleutian Islands; boundary stability; grazing; kelp; sea urchins; Strongylocentrotus
       polyacanthus.


             INTRODUCTION                of alternate stable-state communities should develop
                                   where forces capable of driving a community beyond
  Thirty years ago, Lewontin (1969) raised the ques-
                                   its domain of attraction are localized and spatially asyn-
tion of whether biological communities are globally
                                   chronous. Such events, including avalanches, fires,
stable or take multiple stable states. Five years later,
                                   treefall gaps, seed masts, and physical disturbance from
Sutherland (1974) presented empirical evidence for
                                   wave-borne rocks and logs, are probably ubiquitous in
multiple stable states in the marine fouling community
                                   nature. Alternate domains of attraction in systems with
at Beaufort, North Carolina, and argued that other sys-
                                   multiple stable states are separated by unstable equi-
tems behave in a similar manner. Although ecologists
                                   libria (May 1977, Scheffer et al. 2001) that should cre-
have since debated the evidence needed to demonstrate
                                   ate sharp transitions in space and time. Thus, one ap-
multiple stable-state communities (Connell and Sousa
                                   proach to understanding multiple stable communities
1983, Peterson 1984), there are now many examples
                                   is to focus on their boundaries.
showing that different organizational states of the same
                                     Numerous studies, done mostly at temperate lati-
species do exist in nature (Holling 1973, Noy-Meir
                                   tudes in the northern hemisphere, indicate that kelp
1975, May 1977, Dayton and Tegner 1984, Ash 1988,
Knowlton 1992, Law and Morton 1993, Maron and             forests exist in two organizational states: one domi-
Jefferies 1999, Petraitis and Dudgeon 1999, Scheffer         nated by lush kelp assemblages (hereafter termed "kelp
et al. 2001).                             beds") and the other by intense sea urchin grazing and
                                   a dearth of foliose algae (hereafter termed "barren
  How and why multiple stable communities occur is
less clear. To answer these questions, one must under-        areas") (see Dayton 1984, Harrold and Pearse 1987,
                                   Witman and Dayton 2000 for reviews). Both organi-
stand both the generation and maintenance of the al-
                                   zational states contain the same species, although the
ternative states (Petraitis and Latham 1999). Mosaics
                                   rate of plant tissue loss to herbivory by sea urchins is
                                   low in kelp beds and high in barren areas (Steinberg
  Manuscript received 20 November 2000; revised 30 May       et al. 1995, Estes et al. 1998). Variation in sea urchin
2002; accepted 5 June 2002. CorrespondingEditor:J. D. Witman.
  2
   Present address: School of Fisheries and Ocean Sciences,     behavior (and not abundance per se) is often the driving
University of Alaska Fairbanks, P.O. Box 757220, Fairbanks,      force in determining which of these states occurs at
Alaska 99775-7220 USA.                        any place or time. Sea urchins in kelp beds are typically
                                  174
   2003
January                STABILITYOF BOUNDARY REGIONS                          175

weakly motile or sessile, employing a sit-and-wait for-   TABLE 1. Changesin kelp abundance Adak and Shemya
                                                 at
                               Island (in the Semichis) between 1987 (when sea otters
aging strategy in which they consume detrital fallout     were absent from the Semichis and abundantat Adak) to
(hereafter termed "drift") from the kelp canopy (Pearse    1997 (when sea otterpopulationdensitywas similarat both
1980, Duggins 1981, Cowen et al. 1982, Keats et al.      islands).
1984, Ebeling et al. 1985, Harrold and Reed 1985,
Rogers-Bennett et al. 1995, Konar 2000a). In barren                      Date
areas, these herbivores move extensively in their quest              1987           1997
for food, destroying kelps and other foliose algae in                Percentage       Percentage
the process (Russo 1979, Cowen et al. 1982, Harris et    Island      n   w/o kelps     n   w/o kelps
al. 1984, Harrold and Reed 1985, Scheibling 1986,      Adak      786    20.9     780    71.9
Rogers-Bennett et al. 1995). While sea urchin predators   Semichi     845    87.5     840    64
can drive shallow reef environments from one state to     Notes: The analysis is based on counts of kelp plants in
the other, especially in the northern hemisphere (Estes   randomlyplaced 0.25 m2-quadrats 35 randomlyselected
                                                at
and Palmisano 1974, Dayton 1984, Harrold and Pearse     sites in the Semichi Islandsand 45 similarlyselected sites at
                              Adak Island (see Estes and Duggins [1995] for details of
1987, Foster and Schiel 1988, Steinberg et al. 1995),    sampling procedures).Number of quadratssampled = n;
smaller scale mosaics (from meters to hundreds of me-    "percentagew/o kelps" refers to the percentage of these
ters across) of kelp beds and barren areas characterize   quadrats which no kelps were counted.
                                   in
many kelp forests (Duggins 1983, Watson 1993). These
mosaics often exist within what appear to be uniform
habitats, and switches between states in specific patch-  1974). The state change is abrupt for both increasing
es of habitat occur through time (Ebeling et al. 1985,   (Estes and Duggins 1995) and declining (Estes et al.
Harrold and Reed 1985, Konar 2000a). Ebeling et al.     1998) otter numbers. Furthermore, after making thou-
(1985) found that episodic storms drove this system     sands of dives at dozens of islands in the Aleutian
from one state to another, contingent on the starting    archipelago, it was apparent that when kelp forests and
point (i.e., storms drove kelp forests to urchin barrens  barren areas co-occur in these transitional communi-
and visa versa) although they did not attempt to explain  ties, they always do so as a distinct patchwork mosaic
the abrupt spatial transitions between kelp forests and   with one state or the other predominating over the larg-
barren areas. In some cases these boundary regions are   er scale of entire islands. Thus, our efforts to under-
transitory as actively grazing sea urchins form "fronts"  stand the mechanisms of change have focused on these
that invade the kelp beds (Leighton 1971, Camp et al.    mosaics, and especially the boundary regions between
1973, Witman 1985), whereas in other cases they per-    them.
sist for long periods (Mattison et al. 1977).         The work reported here was done at Shemya (52?43'
  Despite this extensive evidence for multiple stable   N, 174?07' E) and Adak (51?84' N, 176?64' W) islands
states in kelp forest communities, the transitional dy-   in the Aleutian archipelago. Shemya Island was recol-
namics between kelp forests and barren areas are poorly   onized by sea otters in the late 1980s or early 1990s
understood. Here we examined why sea urchins in bar-    and supported a small population during the time of
ren areas at the margins of kelp patches did not invade   our study (Konar 2000b). The reef habitat around She-
the adjacent kelp beds. First, we investigated the pos-   mya was mostly barren areas, interspersed with small
sibility that urchins at the border regions are not food  foliose algal beds formed by the perennial kelps, Aga-
limited and do not move into kelp areas for this reason.  rum cribrosum, Thalassiophyllum clathrus, Laminaria
Our findings did not support this idea. We next hy-     dentigera, and L. yezoensis, and during summer the
pothesized that the kelps somehow prevented urchin     annual species Desmarestia viridis, D. ligulata, and the
intrusions, and tested that hypothesis by manipulating   surface canopy%-forming   kelp, Alaria fistulosa. The
kelp and urchin populations at the border regions and    boundaries between kelp stands and barren areas are
within established kelp beds. A strong inhibitory effect  sharp and persist over time intervals of weeks to
by the kelps in the border regions was demonstrated.    months. Adak Island supported an abundant sea otter
We also conducted a series of related experiments and    population from the early 1960s through the 1980s
measurements to clarify the inhibitory mechanism.      (Kenyon 1969), during which time the reef habitats
                              surrounding the island were kelp dominated (Estes and
           THE SYSTEM
                              Duggins 1995). Increased predation by killer whales
  Our work was done in the western Aleutian archi-     drove this population sharply downward during the
pelago of the North Pacific Ocean. Whether rocky reef    1990s. By 1999, when the work reported herein was
systems in this region exist as barren areas or kelp beds  done, both otter densities (Doroff et al. 2003) and the
depends largely on whether or not sea otters (Enhydra    kelp forest community at Adak and Shemya (Estes et
lutris) are present. Sea otters consume sea urchins     al. 1998, Table 1) were similar.
(Strongylocentrotus polyacanthus), thereby limiting      Our focus here is on how the mosaics of kelp stands
their size and population density, and preventing the    and urchin barrens are maintained. While prior exper-
development of barren areas (Estes and Palmisano      iments have demonstrated that urchin grazing prevents
176                 BRENDA KONAR AND JAMES A. ESTES             Ecology,Vol. 84, No. 1

kelps from invading the barren areas (Paine and Vadas            Urchin gonad indices
1969, Duggins 1981, Harrold and Pearse 1987), the
                               Echinoderm gonads, besides producing gametes,
factor or factors that prevent sea urchins from invading
                              serve the added function of energy storage (Lares and
the adjacent kelp beds were unclear. In most cases,
                              Pomoroy 1998, Russell 1998). Thus, gonad size varies
there was no evident variation in the physical habitat
                              both seasonally and with food availability (Hagen
(i.e., depth, substrate type, or water movement) be-
                              1998, Konar 1998, Meidel and Scheibling 1999, Vadas
tween these community states.
                              et al. 2000). Gonadal indices of sea urchins ([gonad
           METHODS               mass/total wet mass] X 100; Gonor 1972) were mea-
      The community and study sites         sured from individuals distributed across the kelp bed-
                              barrens border as a relative measure of fitness variation
  Except for descriptions of the transitions between   between the two community states and at their inter-
kelp beds and barren areas (described in the following   face. Although fitness is defined by both growth and
paragraph), all of our research was done in a mosaic    reproductive output, we selected this simple measure
of kelp beds and barren areas on the southwest shore    because we were interested in the degree to which food
of Shemya Island. The percent cover of major algal     might be limiting to sea urchins across the barrens-
taxa in three distinct beds was visually estimated (De-  kelp forest interface at the time of our study.
thier et al. 1993) in randomly placed 0.25-M2 quadrats.    Twelve urchins of varying test diameters were hap-
Brown algae were identified to species and grouped as   hazardly collected 1 m into the kelp beds, 1 m into the
annuals or perennials, while foliose red algae and en-   barren grounds, and at the points of transition from
crusting coralline algae were grouped, but not identi-   kelp to barren areas. Only sea urchins >40 mm test
fied to species. Our characterization of these kelp beds  diameter were used in this analysis, because studies
was based on a sample of 54 quadrats, taken from be-    elsewhere show that smaller individuals of a similar
tween 8 and 13 m water depth during June-August      species, S. purpuratus, have proportionally smaller go-
1996.                           nads (Gonor 1972). The urchins were weighed and dis-
  Transitions between kelp beds and barren areas     sected to determine gonad mass. These measurements
                              were taken in June 1997.
  To quantify the nature of transitions between kelp
beds and barren areas, we located three small kelp beds       Urchin additions and algal removals
in the Kuluk Bay region of Adak Island. The species
composition and mosaic nature of these beds were sim-     We conducted an experiment to determine if sea ur-
ilar to those that occurred at Shemya Island (B. Konar,  chins could generate and maintain barren areas within
personal observation). Border regions were located     established kelp beds. Four replicate 100-M2 blocks
haphazardly, and contiguous 0.25-M2 quadrats were     were established within kelp beds. Each block was di-
sampled perpendicularly across each border. All foliose  vided into four 5 X 5 m plots, which served as our
brown algae were identified to species and counted, sea  experimental units. Urchin density and algal cover were
urchins were counted, and percent covers estimated for   monitored monthly in each block using six randomly
foliose red algae. For each of these variables, we used  placed 0.25-M2 quadrats. Both the manipulations and
a randomized block design (in vs. out of kelp stands    measurements were located sufficiently far from the
and distance from border as treatments, beds as blocks)  block borders to avoid interactions with treatments in
to assess statistically significant spatial patterns be-  the adjacent blocks. Measurements were taken from
tween and within the alternate community state.      July through November of 1996 (no sampling was done
                              in October because of persistent inclement weather),
       Algal drift measurements           and in August 1997. The experimental units were sub-
  Drift algae are known to influence the foraging be-   jected to the following treatments, assigned at random
havior of sea urchins, and thus their tendency to over-  in each block.
graze kelp beds (Ebeling et al. 1985, Harrold and Reed    Foliose algae removed/urchins added.-The pur-
1985, Konar 2000a). The abundance and species com-     poses of this treatment were to determine if high urchin
position of drift algae were measured along transects   densities could persist within a kelp bed when the fo-
run perpendicular to the kelp bed-urchin barren inter-   liose algae were removed, and if so, whether or not
face to determine how this potential food source for    these cleared patches were maintained as barren areas.
urchins varied across the border between these two     To make this determination, all foliose algae were re-
community states. Drift algae were quantified 3 m into   moved and urchins were added (from the nearby bar-
the kelp bed and 3 m into the urchin barren by collec-   rens) at a density of 68.0 individuals2 (average den-
tions from six haphazardly placed 10-M2 circular plots   sity in the nearby barrens).
in each habitat. Only horizontal, rocky substrates were    Foliose algae removed/no urchins added.-The pur-
sampled. The drift algae were weighed to the nearest    poses of this treatment were to determine if urchins
gram (wet mass). Brown algae were identified to spe-    moved from within or through the kelp bed into a
cies and grouped as perennials or annuals.         cleared area, and if not, to measure recovery of the
   2003
January                STABILITYOF BOUNDARY REGIONS                        177

foliose algae relative to Treatment 1. To make this de-     Unmanipulated controls.-These were done to mon-
termination, all foliose algae were removed from the     itor natural fluctuations in urchin density at the kelp
experimental units, but urchins were not added.       bed margins and on the periphery of the barren grounds
  Foliose algae not removed/urchins added.-The       during the course of the experiment.
purposes of this treatment were to determine if urchins     Nested ANOVAs (quadrats nested within experi-
could remove the foliose algae from within an estab-     mental units nested within sites) were conducted on
lished kelp bed, and if so, to chronicle the development   urchin densities in the kelp beds and barren areas im-
and persistence of the resulting urchin barren. To make   mediately before the manipulations to determine if the
this determination, urchins were added to the experi-    availability of urchins varied among treatment areas.
mental units at a density of 68.0 individuals/,   but   The same analyses were conducted on urchin densities
the foliose algae were not removed.             10 d after the manipulations to determine if there were
  Unmanipulated control.-These    were included to   significant treatment effects.
determine the natural variation of urchin and foliose
algal abundance over the course of the experiment.               Algal abrasion
  We analyzed these data as a three-way model (ur-
chins present/absent X foliose algae present/absent X      The previously described algal removal experiments
areas). The ANOVAs were run separately for each time     showed that both fleshy macroalgae and the surveyor's
period in order to make explicit changes through time    tape-models inhibited sea urchins from invading the
evident.                           kelp beds. We suspected that wave-induced abrasion
                               was the mechanism causing this inhibition. Because
     Invasion of urchins into kelp beds         the kelp beds were comprised of several common mac-
  Foliose algae were removed from kelp beds at their    roalgal species, we wished to determine whether these
interface with barren areas to determine if the algae    varied in abrasive function, and if so, which ones
inhibited sea urchin invasion. Here again, the experi-    caused the most abrasion. Dissolution rates of clod
mental units were 5 X 5 m quadrats, extending into      cards (Denny 1985) were used for this purpose.
the kelp bed from the barren area interface. Twelve       Our clod cards, similar to those described by Doty
experimental units were established, and the treatments   (1971), were a mixture of plaster of paris and latex
were assigned randomly among them. Urchin densities     paint that were molded and hardened in ice cube trays.
were measured from six randomly placed 0.25-M2        Once hard, the clod cards were glued to a small sheet
quadrats in each experimental unit and the adjacent     of PVC and placed in a seawater aquarium for one week
urchin barren. Measurements were taken immediately      to cure. They were then removed from the seawater,
prior to the manipulations and 10 d later. These ex-     dried, and weighed at successive times until a constant
periments were initiated in June 1997 and consisted of    value was attained. Each clod card then was attached
the following four treatments, each replicated three     by a cable tie to a cement brick (40 X 20 X 6 cm) and
times.                            placed in the field. The clod cards used for each ex-
  Total algal removals.-This treatment was under-      periment were made from a single batch in order to
taken to determine if urchins would invade the kelp     eliminate variation between the relative amounts of
bed when all of the foliose algae were removed from     plaster of paris and latex paint used. Treatment/han-
its border. A strong initial response was obtained, and   dling controls were taken into the field and then re-
hence the following treatments were added to refine     turned to the seawater aquarium.
our understanding of the mechanisms.              Field measurements were made during July 1997, a
  Removal of annuals only.-Only the annual algae      period when sea conditions were relatively calm. Clod
(Alaria fistulosa and Desmarestia spp.) were removed.    cards were haphazardly placed in the field under nat-
This was done to determine the role of annual vs. pe-    urally occurring patches of Alaria fistulosa, Desma-
rennial algae in inhibiting sea urchin invasions.      restia viridis, Agarum cribrosum, or Laminaria den-
  Structural replacements of the foliose algae.-All     tigera that were intermixed within the same kelp stand.
foliose algae were removed, but they were replaced      One clod card was placed under each patch. Additional
with clumps of surveyor's flagging that were nailed to    cards were placed in cracks in the substratum. Each
the substrate in similar densities to the algae. These    treatment was replicated four times. After four days,
"artificial plants" were added to recreate the physical   all the cards were collected and returned to the labo-
presence of foliose algae, but without any of their bi-   ratory, and dried until a constant mass was attained.
ological properties, to determine if algal structure alone  Mass loss by the control cards was subtracted from
could prevent sea urchins from invading the kelp beds.    mass loss by the various treatment cards to obtain an
While these models are dissimilar in form to many of     index of abrasion.
the foliose algae (the morphology of which also varies     Unless otherwise stated, standard parametric statis-
substantially both within and among species), they are    tical analyses were used. Enumeration data were In +
similar to most of these algae in their flexibility and   1 transformed, and percent cover data were arcsine
movement with wave-induced surge.              square-root transformed prior to analysis.
178                           BRENDA KONAR AND JAMES A. ESTES                  Ecology, Vol. 84, No. 1



   NE 60 -
   LC)
N
   (N

<a4                                   urchins
o m'                     tE             brown algal stipes
W      40    12                         foliose red algae       FIG. 1. Number of urchins and brown algal
.n    O           I    T       T                      stripesand percent cover of foliose red algae in
0-(0Cn             *   l      *  l0.25-M2                        quadrats placed along a contiguous
  *?
 =15             *   |      *  *                      transect from 1.25 m within the kelp bed to 1.25
en $)*                            *                   m within the barren area at the study site in the
                                                    **
  s,      20                                         Aleutian archipelago. Values are means and I
       0                                           ~~~~~~~~~~~~~~~~~~~~SE.

       0

            1.25   1.0 0.75 0.5 0.25      0.25   0.5 0.75 1.0    1.25
                within barren ground          within kelp bed
                            Distance


                     RESULTS                     Transitions between kelp beds and barren areas
                  The community                   The borders between kelp beds and barren areas at
                                          Adak Island were sharply defined (Fig. 1). Sea urchins
  The percent algal cover in kelp beds at Shemya Is-
                                          were abundant from the edge of the border outward
land varied substantially among species. Perennial
                                          into the urchin barren, but urchin density declined
brown algae (Agarum cribrosum, Thalassiophyllum
                                          abruptly from the border into the kelp beds. Similarly,
clathrus, Laminaria dentigera, and L. yezoensis) con-
                                          kelp density and red algal cover varied sharply over
tributed most of the canopy cover (52.4 + 3.4% [mean
                                          the same gradient, but in the opposite direction All of
  1 SE]), followed by annual brown algae (Alaria, 10.9
                                          these variables differed significantly between the kelp
+ 1.9% and Desmarestia, 3.0 + 0.9%), foliose red
                                          stands and adjacent barren areas (urchins, F,,44 = 329.7;
algae (20.9 + 2.5%), and encrusting coralline algae
                                          kelp stipes, F144 = 221.6; foliose red algae, F144 =
(11.7 + 2.5%). Foliose algae of all kinds were absent
                                          30.5; P < 0.01 for each). None of these variables dif-
from the adjacent barrens where the substratum was
                                          fered significantly among the five distances from the
covered almost entirely with encrusting coralline algae
                                          borders across the kelp beds (urchins, F420 = 0.692, P
and occasional small (<0.25 M2) patches of the en-
                                          = 0.606; kelp stipes, F4,20 = 0.836, P = 0.518; foliose
crusting green alga, Codium setchelli.
                                          red algae, F4,20 = 0.421, P = 0.792) or the barren areas
                                          (urchins, F420 = 1.049, P = 0.407; kelp stipes, F420 =
           100                              0.478, P = 0.751; foliose red algae, F420 = 0.387, P
               reds                          =  0.815).
E      80      n perennialbrowns.
o             E2annual browns                                  Drift algae
                                       a)   The abundance of drift algae also varied sharply
   v
   60)      eo'-:                        ~~~~~~~~~~~4 kelp bed-urchin barren interface (7.1
                                        across the                     ? 3.5
Cl)
                                          g wet mass/10 m2 in barren area vs. 99.5 ? 41.9 g/10
M 40-                                       m2 in kelp bed, Fig. 2). This difference was highly
E
                                      2~~~~~~0
                                        significant    (t = 5.02, df = 10, P < 0.001).   Most drift
a)20-                                       in the barrens was red algae (81.7% of the total). In
                               -
                                          contrast, most drift in the kelp bed was perennial
         0                            0    (64.1% of the total) and annual (34.9% of the total)
            in barren       on edge    in kelp        brown algae.
                        Site
                                                    Urchin gonad indices
  FIG.2. Drift algal abundance and sea urchin gonadal in-
dices (mean ? 1 SE) across the barren-area-kelp-bed interface             The relative size of sea urchin gonads also varied
in 15 m water depth on the Pacific Ocean side of Shemya              sharply across the barren-area-kelp-bed interface (Fig.
Island, Alaska. Drift algae were quantified 1.0 m into the
                                          2). These were significantly less at the kelp-barren in-
barren area and 1.0 m into the kelp bed from the barren-kelp-
bed interface using six randomly placed 10-M2 circular plots            terface and 1 m into the barren ground (1.8 ? 0.3 and
in each habitat. Gonads were quantified by sampling 12 ur-             1.6 ? 0.1, respectively) than they were only 1 m into
chins in each habitat and along the interface.                   the kelp bed (5.6 + 0.6; ANCOVA: Area, F2,32 = 24.97,
January 2003                           STABILITY OF BOUNDARY REGIONS                                179

                                 - U - foliose cover--*-urchin           density

                A) Treatment1: Algae                                    2:
                                                        B) Treatment Algae
               removed, urchinsadded                                  removed

              100                            24         100                 24

         'a    80           L       X        2     i
                                               XE      80             S
                                                  0
            00                                      0~~~~~~
               0                            6002
                                                        40~~~~~~~~~~~~~~(
                                                   U4
            i   40~
               40          ~     ~    ~~6.
                                               L~~~~~~~~~~                    6
            .220                              0   6       20                 6
                                                  2
         LL 5    0U--60~1                            cz               +  *      0
                 C) Treatment 3: Urchins                            D) Treatment 4: Control
            IL             added               8 Lo                             8  Lo


       m     >  100 Th         C1n    E            24      0                     24

  s80 u                                                                     18  m
            0
                                                  ~~~~~~~~06
            o60                                    -0060
                                   A        i2 oS                          12
                 40      ~~~~~~~~~.0                         400

         .2    20--0                I          6~~~6 .0       2
                                                      2                  6    6
                                                  U
         U-                                                  A
               0        1     I    I    I      0           0 AT.         A a
                   7_           o        c                  7   (0  0)  0    CU

                      -
                            <
                               E    E    >~                  ~     E  E    >
                                                                 0    0
                                   0    0


  FIG. 3. Urchin density (mean ?I   SE) and mean percent cover foliose algae at the various algal removal and urchin
addition manipulations. The design of this manipulation consisted of four replicate 100-rn2 manipulation blocks that were
set up in three separate kelp beds on the Pacific Ocean side of Shemya Island, Alaska, between 8 and 13 m water depth.



P < 0.001; Urchin mass a covariate, F132 = 7.291, P                        d). Site effects on urchin density were statistically sig-
-  0.011).                                            nificant during two of the sample periods (Table 2).
                                                  Interaction effects were similarly small and inconsis-
   Urchin additions and algal removals within the                        tent through time (Table 2).
            kelp beds                                   Foliose algal cover in the control plots remained
  Urchins added to both cleared and uncleared areas                        largely unchanged through the study, ranging between
within the kelp beds remained at high densities (65.6                       71.7% and 96.3% (Fig. 3d). When all foliose algae were
to 84.0 urchins/M2, respectively) from July 1996 when                       removed and urchins were not added, foliose cover
the translocations were initiated, until the first winter                     increased by -20% after one month (by August) and
storms in October (Fig. 3a, c). Highly significant urchin                     increased steadily thereafter to -90% of the initial val-
treatment effects on urchin densities for the period of                      ue one year later (Fig. 3b). When sea urchins were
July through September reflect this pattern (Table 2).                       added to otherwise unmanipulated kelp beds, they ac-
By November, urchin densities had declined in both                         tively grazed the plots, causing the foliose algal cover
urchin addition treatments (0.8 to 10.4 urchins/M2, re-                      to decline from 75.4% to 24.5% between July and Sep-
spectively), and all of those that remained occurred in                      tember (Fig. 3c). Defoliation continued until the first
cracks and crevices within the substrate. At this point,                      winter storms, at which time most of the urchins dis-
urchin densities did not vary significantly among treat-                      appeared and foliose algal cover increased to 49.5%
ments (Table 2). This pattern remained unchanged                          by November 1996 and 91.3% by August 1997. Ur-
through August 1997, although by that time significant                       chins added to cleared areas prevented foliose algal
treatment effects again had developed because of in-                        recovery from July through September (Fig. 3a). Most
creases in both urchin addition treatments (Fig. 3a, c).                      of these urchins disappeared or retreated to cryptic hab-
Urchin densities in the other treatment plots (Foliose                       itats with the onset of winter storms in October, where-
algae removed/no urchins added, and Unmanipulated                         upon foliose algal cover increased (to 26.6% by No-
controls) remained low throughout the study (Fig. 3b,                       vember and to 91.3% by August 1997). After one year,
180                   BRENDA KONAR AND JAMES A. ESTES               Ecology, Vol. 84, No. 1

TABLE 2.  Summary of statistical results, reported by time periods, from the experimental addition of sea urchins into kelp
 forests at Shemya Island, Alaska.

                                         Month
                                July                  August
          Effect      df        F         P          F         P
Sea urchin density
 Kelp               1,  3       1.38       0.325         4.46        0.125
 Urchin              1,  3     1032.56      <0.001        517.29       <0.001
 Site               3,  16      4.08       0.025         2.42        0.104
 Kelp X Urchin           1,  3       0.05       0.831         0.003       0.961
 Kelp X Site            3,  16      1.13       0.367         1.06        0.394
 Urchin X Site           3,  16      0.78       0.524         1.16        0.357
 Kelp X Urchin X Site       3,  16      0.87       0.475         1.85        0.179
Percent cover foliose algae
 Kelp               1,  3      38.22       0.009        87.07       0.003
 Urchin              1,  3       0.54       0.517        11.67       0.042
 Site               3,  16      1.17       0.354         7.87       0.002
 Kelp X Urchin           1,  3       0.14       0.732         0.05       0.841
 Kelp X Site            3,  16      16.51      <0.001         4.05       0.026
 Urchin X Site           3,  16      0.64       0.598        17.81       <0.001
 Kelp X Urchin X Site       3,  16      1.79       0.189         2.6       0.088
 Notes: Data were analyzed as a 2 X 2 X 4 factorial experiment, with Kelp (+ or - at beginning of experiment) and Urchin
(+ or - at beginning of experiment) as fixed effects and Site as a random effect. The ANOVAs were conducted on sea
urchin densities (In + 1 transformed) and percent cover of foliose algae (arcsine square-root transformed).



the percent covers of foliose algae in all the manipu-         Inhibition of urchin invasions across the
lated treatments had returned to their initial values and              kelp bed border
were not significantly different from the controls (Fig.
                                 Sea urchin densities did not differ significantly
3). These patterns are reflected by highly significant
                                among treatment plots prior to the manipulations, thus
algal and urchin treatment effects on foliose algal cover
                                demonstrating the preexperimental homogeneity of our
for each sample period from July-November 1996 and
                                experimental units (Table 3). Likewise, the absence of
statistically insignificant effects by summer 1997 (Ta-
                                significant treatment effects in the adjacent barren areas
ble 2). Site effects were mostly insignificant through
                                demonstrates that all of the plots were similar in terms
the summer months of 1996. Site effects were highly
                                of available sea urchins (Table 3). However, the various
significant in November 1996 and summer 1997. In-
                                treatment manipulations strongly influenced subse-
teraction effects were small and inconsistent through
                                quent sea urchin density. Sea urchins quickly occupied
time (Table 2).
                                areas that were cleared of all foliose algae at the kelp-
                                barren interface (Fig. 4). Within ten days, urchin den-
                                sities in these plots increased from low values to those
                          * kelp   occurring in the adjacent barrens (42.4 ? 2.8 urchins/
    12 -E0                    barren
                                m2vs. 44.8 ? 2.8 urchins/M2, respectively). During the
    10    T
                                same time period, urchin densities in the kelp bed mar-
LO)
                                gins did not change notably when only the annual algae
                                were removed, when all algae were removed and sur-
                                veyor's flagging was added, or in the unmanipulated
                                treatments (Fig. 4).
 01    -         -        -
                                           Algal abrasion
                                The rate of mass loss varied significantly across the
                              various algal species (Table 4; ANOVA: F = 19.62, df
      all cleared annual structure  added, control  = 4, 15, P < 0.001). The loss rate was greatest for
            cleared   all cleared
                              clod cards placed under the two annual species, Alaria
               Treatment           fistulosa (19.7 ? 0.6 g) and Desmarestia viridis (17.8
  FIG. 4. Sea urchindensitiesfollowing kelp manipulations ? 1.8 g), and the perennial kelp Agarum cribrosum
at the kelp-bed-barren-area interfaceon ShemyaIsland,Alas- (15.0 ? 1.1 g); less for cards placed under Laminaria
ka. After 10 d, three replicatetreatmentswere surveyedfor
urchindensities using six randomlyplaced 0.25-in2quadrats dentigera (6.9 ? 1.0 g); lower yet for cards placed in
in each treatment adjacentbarrenarea.Valuesare means substrate cracks and crevices (3.1 ? 1.0 g); and lowest
          and
and 1 SE.                         for the laboratory controls (1.4 + 0.05 g; n = 5, Scheffr
January 2003                 STABILITY OF BOUNDARY REGIONS                              181


TABLE  2.    Extended.


                      Month
           September                   November                   One year
       F          P            F            P           F         P


     10.93          0.046          0.001         0.977          0.002         0.986
    771.93          <0.001          7.62          0.07         235.33         0.001
     1.77          0.193         12.51         <0.001          0.672         0.582
     0.002          0.963          0.36         0.59          1.1          0.372
     0.089          0.965          1.507         0.251          0.91         0.458
     1.36          0.29          22.756        <0.001          0.257         0.855
     0.28          0.836          1.884         0.173          0.253         0.858

     17.77          0.024         28.29         0.013          3.254         0.169
    258.97          <0.001         21.48         0.019          2.349         0.223
     1.66          0.217         20.65         <0.001         40.57         <0.001
     10.09          0.05          0.55         0.512          0.353         0.594
     7.5           0.002          1.61         0.226          4.07         0.025
     1.9           0.17          11.73         <0.001          4.61         0.016
     3.21          0.051          3.63         0.036          2.16         0.133



multiple range test, P < 0.05). An expanded analysis            in barren areas to move toward food, which is abundant
including data from the crevice treatments and labo-            in the adjacent kelp beds (Mattison et al. 1977, Russo
ratory controls showed that loss rates under L. dentig-           1979, Konar 2000a), a behavior that seems inconsistent
era were significantly elevated (ANOVA: F = 64.09,             with stable boundary areas. The common explanation
df = 5, 19, P < 0.001, Scheff6 multiple range test, P            for boundary area stability has been that the kelp bed
< 0.05).                                  supplies enough drift algae to urchins at the edge of
                                      the barren area that these individuals need not move
               DISCUSSION
                                      in search of food (Russo 1979, Harrold and Reed 1985).
 Alternate stable states in kelp forest communities            This was not the case in our study, as the abundance
  Many complex systems may be characterized by               of drift algae declined to the typically low value of
multiple equilibria because of the nonlinear nature of           barren areas within a meter of the kelp bed border (Fig.
their forcing functions (Scheffer et al. 2001). Kelp beds          2). Food limitation for urchins living at the margin of
and barren areas in northern hemisphere kelp forests            the barren ground is further indicated by the abrupt
seem to provide a particularly good example of alter-            decline in their gonadal indices at the kelp forest-bar-
nate stable states for biological communities. Spatio-           ren interface (Fig. 2).
temporal discontinuities in these states have been dem-            Our experiments show that macroalgae living at the
onstrated in numerous studies, but the essential role of          border between kelp beds and barren areas deter in-
history in understanding these systems (Lewontin              vasions by sea urchins. When all foliose algae at the
1969, Sutherland 1974) only became clear following             kelp bed margin were removed, urchin densities rapidly
Ebeling et al.'s (1985) long-term study of Naples Reef.           increased within the clearings (Fig. 4). Urchin inva-
Other studies have shown a strong tendency for urchins           sions also were inhibited when the cleared algae were

           TABLE 3. Summary statistics for nested ANOVAs conducted on sea urchin numbers measured
            immediately before and 10 days following the border manipulations.

                                 Initial             After 10 days
               Source         F       df     P      F      df      P
           Barren area
            Treatment          0.958.     3, 45   0.421    0.251    3, 45    0.86
            Replicate(Treatment)    0.739      8, 45   0.657    1.41    8, 45    0.219
            Quadrat(Replicate)     0.561     15, 45   0.889    0.995   15, 45    0.476
           Kelp area
            Treatment          0.334      3, 45   0.799    90.03    3, 45    <0.001
            Replicate(Treatment)    0.482      8, 45   0.862    0.537    8, 45    0.822
            Quadrat(Repricate)     0.94      15, 45   0.53     1.115   15, 45    0.371
            Notes: Results are reported separately for measurements taken from within the kelp forests
           and the adjacent barren areas. Data were In + 1 transformed prior to analysis.
182                  BRENDA KONAR AND JAMES A. ESTES                Ecology,Vol. 84, No. 1

TABLE 4. Mass loss (g; mean + l   SE)  of clod cards placed   Different algal species varied in their abrasive action
 under different algal species.
                                 (Table 4). We did not measure flow rate, and thus it is
  Algae               Mean + 1     SE
                                 possible that some of the differences among species
                                 was a consequence of differential dissolution due to
Alaria                19.7  +  0.6  A
Desmarestia             17.8  +  1.8  A      variable water flow rather than abrasion per se. How-
Agarum                15.0  +  1.1  A      ever, we believe that the differences in the clod card
Laminaria               6.9  +  1.7  B      mass loss were due to algal abrasion rather than dif-
In crack               3.1  +  1.0  B
                                 ferences in water flow around dissimilar algae because
  Notes: Four cards were placed under each experimental     in another study done on Shemya Island, clod cards
species and collected after four- days. Similar letters to the  were found with significantly more abrasion under al-
right of the standard errors denote nonsignificant differences
using a post hoc ScheffM F test, P < 0.05.            gae in kelp beds than in algal cleared kelp beds (20.5
                                 g + 1.6 vs. 5.8 g + 0.5; Konar 2000a). In fact, mass
                                 loss of clod cards due to water motion alone was min-
replaced with model kelp plants, thus demonstrating        imal. In general, annuals caused more abrasion than
that some feature of the kelp's physical presence (as       did perennials. Nonetheless, the perennials by them-
opposed to induced defenses) was likely the key mech-       selves prevented urchins from invading the kelp beds.
anism. These observations led us to wonder why the        These findings explain why the borders we studied per-
same food species are beneficial in some circumstances      sisted through the winter when the annual algae were
and inhibitory in others.                     absent. The only alga that did not cause significant
  Although sea urchins did not move naturally from        abrasion was Laminaria dentigera. This probably is
barrens into kelp beds, those that were translocated       because L. dentigera possess a long, thick stipe that
across the boundary areas into the centers of the beds      holds its single blade above the substrate. Velimirov
both persisted and defoliated the surrounding areas        and Griffiths (1979) showed that abrasion patterns be-
(Fig. 3), thus indicating that kelps living in the interior    neath L. pallida (a species with similar morphology)
portions of the bed are vulnerable to urchin grazing,       varied with plant size. Smaller plants (1-5-cm stipe
whereas those at the margin are not. Induced chemical       length) in Velimirov and Griffiths' study abraded the
defense by kelps in the border region is one possibility.     substrate evenly with an area defined by the arc of their
While we did not test for this directly, we think it is      blade length whereas the abrasion beneath larger plants
unlikely for several reasons. One is that we have found      (20-150-cm stipe length) was less near their holdfasts.
very little variation in secondary metabolite levels       Most individual L. dentigera in the kelp beds we stud-
(both phlorotannins and nonpolar compounds) in Alar-       ied were large (50-75-cm stipe length).
ia or Laminaria species collected from the western          The tendency for sea urchins to reside in cryptic
Aleutian archipelago (Estes and Steinberg 1988; P D.       habitats has been viewed largely as a means of predator
Steinberg and J. A. Estes, unpublished data). Another       avoidance (Himmelman and Steele 1971, Lowry and
is that sea urchins in barren areas readily consume even     Pearse 1973, Nelson and Vance 1979, Bernstein et al.
the most well defended kelp species (Agarum and Thal-       1981, Witman 1985, McClanahan 1998), facilitated by
assiophyllum, Steinberg et al. 1995; Desmarestia, Kon-      urchins' ability to employ a sit-and-wait foraging strat-
ar 2000a). Finally, if individuals in the border areas      egy where drift algae is abundant. Algal abrasion may
were protected by chemical defenses, we would expect       also contribute to this behavior, as our clod card ex-
to find the most well defended species in these areas,      periments demonstrate that abrasion rate is lowest in
which was not the case. We think it more likely that       cracks and crevices (Table 4). This interpretation is
the pattern relates to differences in flow and the direc-     supported by the fact that all of the urchins we trans-
tionality of urchin attack between borders and interiors.     located into the kelp beds that survived the winter
Kelps living at the borders are vulnerable to attack from     storms did so by residing in cryptic habitats. Barren
only one direction, whereas those living in the midst       areas persist through the winter, even in shallow water.
of translocated urchins are vulnerable to attack from       However, when kelp beds are present, sea urchin den-
all directions. Himmelman (1984) noted that sea ur-        sity may be limited by the amount of cryptic habitat.
chins usually attacked individual kelps simultaneously      Other factors (such as drift abundance) set higher den-
from all sides and that the resulting large number of       sity limits in these same habitats when kelps are absent.
attacking urchins often anchored the plant to the sea-
floor, thus preventing its normal wave-induced motion          Interaction variation as a mechanism for
from pummeling and thereby warding off attackers.            maintaining alternate stable communities
Because kelp beds impart significant drag on their sur-       Our findings indicate that the same species of plants
rounding water mass (Duggins 1987, Eckman and Dug-        and herbivores may interact in qualitatively different
gins 1993, Friedland and Denny 1995), the strength of       ways depending on relationships between their behav-
surge-induced motion by kelps is probably less in the       ior and initial abundance. This is of more general in-
center of a kelp bed than it is at the border (Seymour      terest because consumer-prey interactions are essential
et al. 1989).                           organizational features of all natural communities.
   2003
January               STABILITYOF BOUNDARY REGIONS                           183

Thus, historically mediated contingencies in the nature   tion of whether such density-dependent   shifts in the
of consumer-prey interactions might lead to the broad    qualitative nature of consumer-prey  interactions is a
occurrence of multiple stable communities. At the level   common mechanism for the generation and mainte-
of population regulation, these interactions take four   nance of multiple stable-state communities. The answer
possible forms, depending upon whether the prey and     most likely will come from studies of systems under
consumer populations are enhanced or reduced by their    strong top-down control and for which multiple or-
respective interactors. For plant-herbivore interactions  ganizational states are evident in space or time.
specifically, both negative and positive effects of her-
                                        ACKNOWLEDGMENTS
bivores on plants are known. Negative effects occur in
                                This research was supported by the U.S. Air Force's Legacy
the numerous cases where plant tissue loss from her-    Program and the USGS-Biological Resources Division. We
bivory reduces a plant's future fitness. Positive effects  thank Eugene Augustine, Daniel Boone, and Joseph Meehan
through overcompensation, though less well docu-      for support and assistance. We thank Dan Doak, Pete Rai-
mented, are also possible. Examples include increased    mondi, Mike Foster, Ingrid Parker, Peter Petraitis, Jon Wit-
flowering and seed production in montaine shrubs      man, and two anonymous referees for helpful criticisms and
                              suggestions. We also thank our many field assistants: Chris-
(Paige and Whitman 1987), increased growth rates in     tian McDonald, Matt Edwards, Nicolas Ladizinsky, Jeanine
encrusting coralline algae (Steneck et al. 1991), en-    Sidran, Cassandra Roberts, Bill Maloney, Bernard Friedman,
hanced plant growth and primary production by un-      Jos Selig, Clare Dominic, Yale Passamaneck, Cynthia Clock,
gulate grazing in African grasslands (Augustine and     Jeanne Brown, Chad King, and Jeff Roller. Many thanks also
                              go to Mike Kenner (U.C. Santa Cruz) and Jim Bodkin, Dan
McNaughton 1998), and enhanced intertidal diatom      Monson, and George Esslinger (USGS-Biological Resources
populations by grazing limpets (Connor and Quinn      Division, Anchorage) for informal support. Thanks also need
1984). Positive effects of plants on herbivores are ubiq-  to go to the U.S. Fish and Wildlife Service-Alaska Maritime
uitous in nature, occurring in all ecosystems fueled by   Refuge and the U.S. Coast Guard for logistical support.
photosynthesis. Negative effects of plant defenses on              LITERATURE  CITED
herbivores also are well known (Lindquist and Hay
                              Ash, J. 1988. Rainforest boundaries. Journal of Biogeogra-
1996, Bolser et al. 1998, Schnitzler et al. 1998, Witman   phy 15:619-630.
and Dayton 2000). However, while herbivore fitness is    Augustine, D. J., and S. J. McNaughton. 1998. Ungulate
often reduced relative to that of an herbivore feeding    effects on the functional species composition of plant com-
on undefended plants, the herbivores are generally      munities: herbivore selectivity and plant tolerance. Journal
                               of Wildlife Management 62:1165-1183.
thought to respond either through the evolution of re-   Barkai, A., and C. McQuaid. 1988. Predator-prey role re-
sistance or by feeding on more palatable species.       versal in a marine benthic ecosystem. Science 242:62-64.
Chemically defended plants may create variation in     Bernstein, B. B., B. E. Williams, and K. H. Mann. 1981. The
herbivore abundance, with the herbivores aggregating     role of behavioral responses to predators in modifying ur-
                               chins' (Strongylocentrotus droebachiensis) destructive
around more palatable prey species. Our findings dem-
                               grazing and seasonal foraging patterns. Marine Biology 63:
onstrate the importance of physically mediated inhi-     39-49.
bition of herbivores by plants, the result being a very   Bolser, R. C., M. E. Hay, N. Lindquist, W. Fenical, and D.
strong negative effect on herbivore population density    Wilson. 1998. Chemical defenses of freshwater macro-
by their preferred prey.                   phytes against crayfish herbivory. Journal of Chemical
                               Ecology 24:1639-1658.
  Density-dependent shifts in bottom-up forcing by     Camp, D. K., S. P. Cobb, and J. F van Breedveld. 1973.
plants on herbivores (from positive in barren areas to    Overgrazing of seagrasses by a regular urchin Lytechinus
negative at the kelp bed's border) in turn changes the    variegatus. Bioscience 23:37-38.
strength of top-down forcing by herbivores on plants    Connell, J. H., and W. P Sousa. 1983. On the evidence needed
                               to judge ecological stability or persistence. American Nat-
(from weak in the kelp beds to strong in the barren      uralist 121:789-824.
areas), thus maintaining the two alternate community    Connor, V. M., and J. F Quinn. 1984. Stimulation of food
states. We know of only one other example of a con-      species growth by limpet mucus. Science 225:843-844.
sumer-prey role reversal causing such a shift in com-    Cowen, R. K., C. R. Agegian, and M. S. Foster. 1982. The
munity state: Barkai and McQuaid's (1988) report of      maintenance of community structure in a central California
                               giant kelp forest. Journal of Experimental Marine Biology
predator-prey switching between rock lobsters and       and Ecology 64:189-201.
whelks in South Africa. At normal densities, rock lob-   Dayton, P. K. 1984. Processes structuring some marine com-
sters consume whelks and limit whelk populations,       munities: are they general? Pages 179-197 in D. R. Strong,
whelks consume mussels and other bivalves, and thus      D. Simberloff, L. G. Abele, and A. B. Thistle, editors.
                               Ecological communities: conceptual issues and the evi-
lobsters help maintain a mussel-dominated reef com-
                               dence. Princeton University Press, Princeton, New Jersey,
munity. When lobster populations were reduced from      USA.
overfishing or physical disturbance, whelk populations   Dayton, P. K., and M. J. Tegner. 1984. Catastrophic storms,
increased, thus causing mussel beds to decline. The      El Nino, and patch stability in a southern California kelp
superabundant whelks became lobster predators, so that    community. Science 224:283-285.
                              Denny, M. W. 1985. Water motion. Pages 405-424 in M. M.
when lobsters were reintroduced into the system, high     Littler and D. S. Littler, editors. Handbook of phycological
mortality from whelk predation prevented them from      methods, ecological field methods: macroalgae. Cambridge
becoming reestablished. These findings raise the ques-    University Press, Cambridge, UK.
184                  BRENDA KONAR AND JAMES A. ESTES                 Ecology, Vol. 84, No. 1


Dethier, M. N., E. S. Graham, S. Cohen, and L. M. Tear.     Himmelman, J. H., and D. H. Steele. 1971. Food and pred-
 1993. Visual versus random-point percent cover estima-      ators of the green sea urchin Strongylocentrotus droeba-
 tions: 'objective' is not always better. Marine Ecology Pro-   chiensis in Newfoundland waters. Marine Biology 9:315-
 gress Series 96:93-100.                     322.
Doroff, A. M., J. A. Estes, M. T. Tinker, D. M. Burn, and T.   Holling, C. S. 1973. Resilience and stability of ecological
 J. Evans. 2003. Sea otter population declines in the Aleu-    systems. Annual Review of Ecology and Systematics 4: 1-
 tian archipelago. Journal of Mammalogy, in press.        23.
Doty, M. S. 1971. Measurement of water movement in ref-     Keats, D. W., D. H. Steele, and G. R. South. 1984. Depth-
 erence to benthic algal growth. Botanica Marina 14:32-35.    dependent reproductive output of the green sea urchin,
Duggins, D. 0. 1981. Sea urchins and kelp: the effects of     Strongylocentrotus droebachiensis (O. F Muller), in rela-
 short-term changes in urchin diet. Limnology Oceanogra-     tion to the nature and availability of food. Journal of Ex-
 phy 26:391-394.                         perimental Marine Biology and Ecology 80:77-91.
Duggins, D. 0. 1983. Starfish predation and the creation of   Kenyon, K. W. 1969. The sea otter in the eastern Pacific
 mosaic patterns in a kelp-dominated community. Ecology      Ocean. North American Fauna 68:1-352.
 63:1610-1619.                         Knowlton, N. 1992. Thresholds and multiple stable states in
Duggins, D. 0. 1987. The effects of kelp forests on nearshore   coral reef community dynamics. American Zoologist 32:
 environments: biomass, detritus, and altered flow. Pages     674-682.
 192-201 in G. R. VamBlaricom and J. A. Estes, editors.     Konar, B. 1998. Mechanisms that structure marine benthic
 The community ecology of sea otters. Springer-Verlag, Ber-    communities at the Semichi Islands, Alaska. Dissertation.
 lin, Germany.                          University of California, Santa Cruz, California, USA.
Ebeling, A. W., D. R. Laur, and R. J. Rowley. 1985. Severe    Konar, B. 2000a. Seasonal inhibitory effects of marine plants
 storm disturbances and reversal of community structure in    on sea urchins: structuring communities the algal way. Oec-
 a southern California kelp forest. Marine Biology 84:287-    ologia 125:208-217.
 294.                              Konar, B. 2000b. Limited effects of a keystone species: trends
Eckman, J. E., and D. 0. Duggins. 1993. Effects of flow      of sea otters and kelp forests at the Semichi Islands, Alaska.
 speed on growth of benthic suspension feeders. Biological    Marine Ecology Progress Series 199:271-280.
 Bulletin 185:28-41.                      Lares, M. T., and C. M. Pomoroy. 1998. Use of body com-
Estes, J. A., and D. 0. Duggins. 1995. Sea otters and kelp     ponents during starvation in Lytechinus variegates (La-
 forests in Alaska: generality and variation in a community    mark) (Echinodermata: Echinoidea). Journal of Experi-
 ecological paradigm. Ecological Monographs 65:75-100.      mental Marine Biology and Ecology 225:99-106.
Estes, J. A., and J. F Palmisano. 1974. Sea otters: their role  Law, R., and R. D. Morton. 1993. Alternate permanent states
 in structuring nearshore communities. Science 185:1058-     of ecological communities. Ecology 74:1347-1361.
 1060.                             Leighton, D. L. 1971. Grazing activities on benthic inver-
Estes, J. A., and P. D. Steinberg. 1988. Predation, herbivory,   tebrates in kelp beds. Nova Hedwigia 32:421-453.
 and kelp evolution. Paleobiology 14:19-36.           Lewontin, R. C. 1969. The meaning of stability. Brookhaven
Estes, J. A., M. T. Tinker, T. M. Williams, and D. F Doak.     Symposium of Biology 22:13-24.
 1998. Killer whale predation on sea otters linking oceanic   Lindquist, N., and M. E. Hay. 1996. Palatability and chemical
                                  defense of marine invertebrate larvae. Ecological Mono-
 and nearshore ecosystems. Science 282:473-476.
                                  graphs 66:431-450.
Foster, M. S., and D. R. Schiel. 1988. Kelp communities and
                                 Lowry, L. F., and J. S. Pearse. 1973. Abalones and sea urchins
 sea otters: keystone species or just another brick in the
                                  in an area inhabited by sea otters. Marine Biology 23:213-
 wall? Pages 92-115 in G. R. Van Blaricom and J. A. Estes,
                                  219.
 editors. The community ecology of sea otters. Springer-
                                 Maron, J. L., and R. L. Jefferies. 1999. Bush lupine mortality,
 Verlag, Berlin, Germany.                     altered resource availability, and alternative vegetation
Friedland, M. T., and M. W. Denny. 1995. Surviving hydro-     states. Ecology 80:443-454.
 dynamic forces in a wave-swept environment: consequenc-    Mattison, J. E., J. D. Trent, A. L. Shanks, T. B. Akin, and J.
 es of morphology in the feather boa kelp, Egregia menziesii   S. Pearse. 1977. Movement and feeding activity of red sea
 (Turner). Journal of Experimental Marine Biology and       urchins (Strongylocentrotus franciscanus) adjacent to a
 Ecology 190:109-133.                       kelp forest. Marine Biology 39:25-30.
Gonor, J. J. 1972. Gonad growth in the sea urchin, Stron-    May, R. M. 1977. Thresholds and breakpoints in ecosystems
 gylocentrotus purpuratus (Stimpson) (Echinodermata: Ech-     with a multiciplicity of stable states. Nature 269:471-477.
 inoidea) and the assumptions of gonad index methods. Jour-   McClanahan, T. R. 1998. Predation and the distribution and
 nal of Experimental Marine Biology and Ecology 10:89-      abundance of tropical sea urchin populations. Journal of
 103.                               Experimental Marine Biology and Ecology 221:231-255.
Hagen, N. T. 1998. Effect of food availability and body size   Meidel, S. K., and R. E. Scheibling. 1999. Effects of food
 on out-of-season gonad yield in the green sea urchin, Stron-   type on reproductive maturation and growth of the sea ur-
 gylocentrotus droebachiensis. Journal of Shellfish Research   chin Strongylocentrotus droebachiensis. Marine Biology
 17:1533-1539.                          134:155-166.
Harris, L. G., A. W. Ebeling, D. R. Laur, and R. J. Rowley.   Nelson, B. V., and R. R. Vance. 1979. Diel foraging patterns
 1984. Community recovery after storm damage: a case of      of the sea urchin Centrostephanus coronatus as a predator
 facilitation in primary succession. Science 224:1336-1338.    avoidance strategy. Marine Biology 51:251-258.
Harrold, C., and J. S. Pearse. 1987. The ecological role of   Noy-Meir, I. 1975. Stability in grazing ecosystems: an ap-
 echinoderms in kelp forests. Echinoderm Studies 2:137-      plication of predator-prey graphs. Journal of Ecology 63:
 234.                               459-481.
Harrold, C., and D. C. Reed. 1985. Food availability, sea    Paige, K. N., and T. G. Whitman. 1987. Overcompensation
 urchin grazing, and kelp forest community structure. Ecol-    in response to mammalian herbivory: the advantage to be-
 ogy 66:1160-1169.                        ing eaten. American Naturalist 129:407-416.
Himmelman, J. H. 1984. Urchin feeding and macroalgal dis-    Paine, R. T., and R. L. Vadas. 1969. The effects of grazing
 tribution in Newfoundland, eastern Canada. Le Naturaliste    by sea urchins, Strongylocentrotus spp. on benthic algal
 Canadien 111:337-348.                      populations. Limnology and Oceanography 14:710-719.
January
   2003                STABILITYOF BOUNDARY REGIONS                            185

Pearse, J. S. 1980. Synchronization of gametogenesis in the     bivorous amphipod (Ampithoe longimana) but not by an
 sea urchins Strongylocentrotus purpuratus and S. francis-     herbivorous sea urchin (Arbacia punctulata). Journal of
 canus. Pages 53-68 in W. H. Clark, Jr. and T. S. Adams,      Chemical Ecology 24: 1715-1732.
 editors. Advances in invertebrate reproduction. Elsevier     Seymour, R. J., M. J. Tegner, P. K. Dayton, and P. E. Parnell.
 North Holland, New York, New York, USA.              1989. Storm wave induced mortality of giant kelp, Ma-
Peterson, C. H. 1984. Does a rigorous criterion for environ-     crocystis pyrifera, in southern California. Estuarine, Coast-
 mental identity preclude the existence of multiple stable     al and Shelf Science 28:277-292.
 points? American Naturalist 124:127-133.             Steinberg, P. D., J. A. Estes, and F C. Winter. 1995. Evo-
Petraitis, P. S., and S. R. Dudgeon. 1999. Experimental ev-     lutionary consequences of food chain length in kelp forest
 idence for the origin of alternative communities on rocky     communities. Proceedings of the National Academy of Sci-
 intertidal shores. Oikos 84:239-245.                ences (USA) 92:8145-8148.
Petraitis, P. S., and R. E. Latham. 1999. The importance of    Steneck, R. S., S. D. Hacker, and M. N. Dethier. 1991. Mech-
 scale in testing the origins of alternative stable states with   anisms of competitive dominance between crustose cor-
 examples from marine and terrestrial ecosystems. Ecology      alline algae: an herbivore-mediated competitive reversal.
 80:429-442.                            Ecology 72:938-950.
Rogers-Bennett, L., W. A. Bennett, H. C. Fastenau, and C.     Sutherland, J. P. 1974. Multiple stable points in natural com-
 M. Dewees. 1995. Spatial variation in red sea urchin re-      munities. American Naturalist 108:859-873.
 production and morphology: implications for harvest re-     Vadas, R. L., B. Beal, T. Dowling, and J. C. Fegley. 2000.
 fugia. Ecological Applications 5:1171-1180.            Experimental field tests of natural algal diets on gonad
Russell, M. P. 1998. Resource allocation plasticity in sea      index and quality in the green sea urchin, Strongylocen-
 urchins: rapid, diet induced, phenotypic changes in the      trotus droebachiensis: a case for rapid summer production
 green sea urchin Strongylocentrotus droebachiensis (Muell-     in post-spawned animals. Aquaculture 182:115-135.
 er). Journal of Experimental Marine Biology and Ecology     Velimirov, B., and C. L. Griffiths. 1979. Wave-induced kelp
 220:1-14.                             movement and its importance for community structure. Bo-
Russo, A. R. 1979. Dispersion and food differences between      tanica Marina 22:169-172.
 two populations of the sea urchin Strongylocentrotusfran-    Watson, J. 1993. The effects of sea otter (Enhydra lutris)
 ciscanus. Journal of Biogeography 6:407-414.            foraging on shallow rocky communities off northwestern
Scheffer, M., S. Carpenter, J. A. Foley, C. Folkes, and B.      Vancouver Island, British Columbia. Dissertation. Univer-
 Walker. 2001. Catastrophic shifts in ecosystems. Nature      sity of California, Santa Cruz, California, USA.
 413:591-596.                           Witman, J. D. 1985. Refuges, biological disturbance, and
Scheibling, R. 1986. Increased macroalgal abundance fol-       rocky subtidal community structure in New England. Eco-
 lowing mass mortalities of sea urchins (Strongylocentrotus     logical Monographs 55:421-445.
 droebachiensis) along the Atlantic coast of Nova Scotia.     Witman, J. D., and P. K. Dayton. 2000. Rocky subtidal com-
 Oecologia 68:186-198.                       munities. Pages 339-366 in M. D. Bertness, S. D. Gaines,
Schnitzler, I., W. Boland, and M. E. Hay. 1998. Organic sulfur    and M. E. Hay, editors. Marine community ecology. Sin-
 compounds from Dictyopteris spp. deter feeding by an her-     auer Press, Sunderland, Massachusetts, USA.