Britton-Simmons 2004
MARINE ECOLOGY PROGRESS SERIES
Vol. 277: 61–78, 2004 Published August 16
Mar Ecol Prog Ser
Direct and indirect effects of the introduced alga
Sargassum muticum on benthic, subtidal
communities of Washington State, USA
Kevin H. Britton-Simmons1,*
Department of Ecology and Evolution, The University of Chicago, 1101 East 57th Street, Chicago, Illinois 60637, USA
1
Present address: University of Washington, Friday Harbor Laboratories, 620 University Road, Friday Harbor,
Washington 98250, USA
ABSTRACT: Introduced algae have become a prominent component of the marine flora in many
regions worldwide. In the NE Pacific, the introduced Japanese alga Sargassum muticum (Yendo)
Fensholt is common and abundant in shallow, subtidal, rocky habitats, but its effects on subtidal, ben-
thic communities in this region have not previously been studied. I measured the response of native
species to experimental manipulation of S. muticum in field experiments in the San Juan Islands of
Washington State. Native canopy (brown) and understory (red) algae were more abundant in plots
from which S. muticum had been removed, and the native kelp Laminaria bongardiana (the most
abundant species of brown alga in the absence of S. muticum) grew more than twice as fast in plots
where S. muticum was absent. The negative effects of S. muticum on native algae appear to be a re-
sult of shading, rather than changes in water flow, sedimentation, or nutrient availability. S. muticum
also had a strongly negative indirect effect on the native sea urchin Stronglyocentrotus droebachi-
ensis by reducing abundances of the native kelp species on which it prefers to feed. My results
indicate that S. muticum has a substantial impact on native communities in this region, including
effects at multiple trophic levels. Because of their worldwide distribution and capacity to alter native
communities, non-indigenous algae are potentially important agents of global ecological change.
KEY WORDS: Sargassum muticum · Introduced algae · Competition · Indirect effects · Stronglylo-
centrotus droebachiensis · Shading
Resale or republication not permitted without written consent of the publisher
INTRODUCTION (Yendo) Fensholt, on native, subtidal kelp communities
in Washington State, USA.
Introduced algae have become a prominent compo- Sargassum muticum is native to SE Asia (Yendo
nent of the marine flora in many regions worldwide 1907), but its present distribution as an invasive spe-
(Rueness 1989, Verlaque 1994a, DeWreede 1996, cies is widespread, including Europe, the Mediter-
Cohen et al. 2001). Despite their widespread distribu- ranean Sea and the west coast of North America. In the
tion, few studies have rigorously documented the United States, S. muticum was introduced to Washing-
effects of exotic algae and as a consequence many of ton State in the early 20th century, probably with ship-
their possible impacts remain speculative. Neverthe- ments of Japanese oysters that were imported for
less, available evidence suggests that introduced algae aquaculture beginning in 1902 (Scagel 1956). Follow-
do have the potential to substantially alter native com- ing its establishment in this region, it subsequently
munities (Verlaque 1994b, Villele & Verlaque 1995, invaded over 3000 km of coastline on the west coast of
Walker & Kendrick 1998, Levin et al. 2002). In the North America (Scagel 1956, Setzer & Link 1971). S.
present study, I experimentally evaluate the impact muticum has now become established in low intertidal
of the introduced Japanese alga Sargassum muticum and shallow subtidal habitats throughout Puget Sound
*Email: aquaman@kevinbs.net © Inter-Research 2004 · www.int-res.com
62 Mar Ecol Prog Ser 277: 61–78, 2004
and the San Juan Islands (own pers. obs.), where it using SCUBA. Removal experiments and associated
occurs in densities as high as 126 plants m–2 (own kelp growth experiments were carried out at 2 sites
unpubl. data). within the San Juan Islands Marine Preserve network
In areas where Sargassum muticum is abundant in adjacent to San Juan and Shaw Islands, known locally
the San Juan Islands, it forms a dense covering that as Colin’s Cove (48° 33.16’ N, 122° 58.79’ W) and Point
towers up to 2 m above all but 1 native algal species, George (48° 32.97’ N, 123° 00.33’ W), respectively. These
Nereocystis luetkeana. These dense stands of S. reserves were established in 1990 and are closed to
muticum may reduce light, dampen flow, increase harvesting with very limited exceptions (e.g. salmon).
sedimentation and reduce ambient nutrient concentra- Laboratory experiments were completed at FHL.
tions available for native kelp species (order Laminari- The 2 field sites vary in their physical characteristics.
ales). Because kelps are an important source of carbon The habitat at Colin’s Cove is composed primarily of
in coastal food webs (Duggins et al. 1989) and the algal relatively flat, rocky reefs with some small boulders.
communities they are associated with provide habitat In contrast, the substratum at Point George is more het-
and food for a wide variety of marine animals (Jones erogeneous, with rocky reef overlaid by a layer of small
1971, Bernstein & Jung 1979, Ebeling et al. 1985), any to medium-sized boulders. Point George is also sub-
negative effects of S. muticum on these communities jected to more intense tidal currents than Colin’s Cove.
may have broader consequences for this ecosystem. Despite these differences in abiotic habitat, both
Previous studies have varied substantially in their field sites are similar in the biological community they
conclusions about how strong an effect Sargassum support. The native kelp communities that dominate
muticum has on native communities. While intertidal shallow, subtidal habitats in this region are species-
studies in British Columbia (DeWreede 1983) and Spain rich and structurally complex. The upper layer of the
(Viejo 1997) suggest that it competes with native algae, algal community is composed of a canopy of large,
a study in California tidepools found no effect (Wilson brown algae in the orders Laminariales and Desmares-
2001). Subtidal studies have found evidence for inhibi- tiales, referred to herein as canopy algae. The middle
tion of giant kelp recruitment in California (Ambrose & layer consists of an assemblage of fleshy red algae
Nelson 1982) and for competition with native algae in from a variety of orders, referred to herein as under-
Denmark (Staehr et al. 2000). Interestingly, in the San story algae. The bottom layer is formed by encrusting
Juan Islands, S. muticum supports a more abundant coralline algae and filamentous, turf-forming algae.
and species-rich epibiont community than the native Herbivores in this shallow subtidal system include the
kelp Laminaria saccharina (Giver 1999), which it is green sea urchin Strongylocentrotus droebachiensis
thought to displace. Previous studies have focused and a variety of grazing molluscs including chitons,
almost exclusively on competitive interactions of S. limpets and snails.
muticum with other algae and have not investigated Background information on Sargassum muticum.
its potential influence on other trophic levels (but see S. muticum is a brown alga in the order Fucales. Al-
Wilson 2001). though the plant is a perennial, its lateral branches are
The goal of this study was to investigate the effect of only present for a portion of the year. The phenology of
Sargassum muticum on native kelp communities in the S. muticum’s life cycle varies regionally. In the San
San Juan Islands of Washington State by measuring Juan Islands, its numerous lateral branches begin
the response of native algae and invertebrates to ex- growing in early March, reach their maximum height
perimental manipulation of the presence of S. muticum. in June, and begin to senesce in mid- to late-August.
I was also interested in the mechanism(s) underlying Only the short (5 to 10 cm) basal holdfast portion over-
the effects of S. muticum. By measuring light, nutri- winters in a quiescent state. In the San Juan Islands,
ents, water flow and sedimentation in experimental each holdfast produces as many as 18 laterals in the
plots in the field, I tested several hypotheses about early spring, each of which can grow as tall as 3 m.
how S. muticum changed the abiotic environment. Small pneumatocysts along the primary axis of each
I predicted that S. muticum should decrease light, lateral make it positively buoyant and cause the later-
water flow and nutrients, but that it should increase als to extend vertically into the water column. S.
sedimentation. muticum has a simple life cycle. Reproductive struc-
tures called receptacles are borne along secondary
branches and contain both oogonia and antheridia.
MATERIALS AND METHODS After eggs are released from the oogonia, they adhere
to the external surface of the receptacle, where they
Study sites. This research was based at Friday Har- are fertilized. Fertilized embryos remain attached to
bor Laboratories (hereafter FHL) on San Juan Island, the receptacle until they develop tiny, adhesive rhi-
Washington State, USA. Field work was accomplished zoids, at which point they detach and recruit to the
Britton-Simmons: Effects of an introduced alga 63
substratum in close proximity to the parent plant (Dey- were taken simultaneously at surface and at depth
sher & Norton 1982). using a Licor LI-1000 data logger coupled to 2 quantum
Sargassum muticum removal experiments. I de- sensors. Surface measurements were taken from a
signed experiments to test the hypothesis that the small boat, and measurements at depth were taken
presence of S. muticum was influencing the structure 30 cm above the substratum in the center of each con-
of the native community. During June 1999, 2 removal trol and removal treatment plot. Because Nereocystis
experiments were simultaneously initiated at Point luetkeana, the only native algae species that is taller
George and Colin’s Cove. At each site, ten 50 × 50 cm than S. muticum, was not present in any of my plots at
plots, at a depth of –2 m mean lower low water (MLLW) the time and because the remaining kelp species have
and containing high densities of S. muticum, were per- demersal blades that typically extend less than 30 cm
manently marked using metal-stamped stainless steel above the substratum, measuring light at 30 cm al-
washers and marine epoxy (Z-Spar splash zone com- lowed me to isolate the effect of S. muticum. Light was
pound™). At each site, the experimental plots were measured at 1 s intervals for 1 min in each plot. Plots
arranged in a line parallel to shore; the average dis- were sampled systematically, beginning at one end of
tance between plots was 2.5 m. Plots were then the site and proceeding to the other end in quick suc-
randomly assigned to control or removal treatments. cession, in order to minimize the time elapsed between
Control plots were not altered. Removal plots had all S. samples. The same sampling protocol was used on all
muticum plants eliminated, both from within the plots sampling dates. Light measurements were always
and from a 50 cm buffer surrounding the plots. Each S. taken at both sites on the same day and the sampling
muticum plant was removed by carefully prying its period straddled noon. Whenever possible, samples
small (1 to 2 cm in diameter), discoid holdfast off the were taken on a cloudless day. The average of the 60
rock substrate with a small dive knife. The buffer zone instantaneous measurements at the surface and at
was created in order to reduce shading or other effects depth was used to calculate percent transmittance
of adjacent S. muticum plants. The initial S. muticum (light at depth/light at surface), which is the quantity
removal began on June 30, 1999 and was completed at reported and used in statistical analyses.
both sites by July 5, 1999. The removal treatment was Water flow measurements. Because Sargassum muti-
maintained over the course of the experiment by cum’s fronds form dense stands extending several
removing any new S. muticum recruits from removal meters into the water column, they have the potential
plots and adjacent buffer zones several times per year. to dampen water flow. I tested the hypothesis that S.
All experimental plots were censused once prior to muticum reduced water flow, using alabaster dissolu-
imposing the removal treatment and then 2 to 6 times tion blocks following the methods of Eckman et al.
per year thereafter. At the time of each census, I (1989). Small blocks of alabaster (di-hydrated calcium
recorded the identity and number of all macroalgae in sulfate) were cut from a larger block and sanded to
each experimental plot. In addition, I estimated the uniform dimensions (average of 46 × 42 × 10 mm).
percentage of primary space on the rock substrate that Blocks (1 per plot) were anchored to threaded steel
was covered by crustose algae, filamentous algae, rods at a height of 30 cm above the substratum, which
fleshy red algae, fleshy brown algae, Sargassum muti- is above all kelp species present (Nereocystis luet-
cum and bare rock. The percent cover estimates were keana was not in any experimental plot at this time)
carried out on a 25 × 25 cm subplot. Benthic inverte- but below S. muticum, thereby isolating the effect of S.
brates were counted in the baseline and first post- muticum on flow. The threaded steel rods were
removal census and then twice per year thereafter. screwed into tubes of hexagonal nuts (8 nuts stacked
Time and weather constraints prevented complete on top of one another so that the threaded holes lined
censuses of all taxa of interest on some sampling dates. up) that had been attached to the rock substratum with
Light measurements. I measured light levels in all marine epoxy (Z-Spar splash zone compound™). In
removal experiment plots on 4 dates in 2000 (March order to prevent abrasion by seaweed, blocks were
21, May 12, June 9 and August 22). Sargassum muti- enclosed in hardware cloth cages (mesh squares =
cum’s fronds are present for only part of the year (typi- 12 mm on a side), which were also attached to the
cally March to August), and therefore any effect they threaded steel rods by stainless steel hardware. Before
have on light should be limited to this time period. and after deployment, the blocks were dried to a con-
Because my sampling dates spanned the months of stant weight at 60°C and weighed to the nearest 1 mg.
March to August, and included dates when the fronds The blocks were deployed on August 27, 1999 and
were absent and present, I was able to use these data collected on August 29, 1999 for a total deployment
to test the hypothesis that S. muticum reduced ambient time of 42.5 h.
light levels in plots where it was present. Measurements Sedimentation measurements. Any effect that Sar-
of photosynthetically active radiation (400 to 700 nm) gassum muticum has on water flow is likely to influ-
64 Mar Ecol Prog Ser 277: 61–78, 2004
ence the flux of particulate matter from the water col- length of time necessary to complete the collection and
umn to the substrate. I measured sedimentation rates processing of samples for both sites exceeded the slack
in each removal experiment plot during July 2001 in current period, the 2 sites were sampled on successive
order to test the hypothesis that S. muticum increases days. Point George water samples were taken on
sedimentation rates. Many native algae (including August 5, 1999 and Colin’s Cove water samples were
kelps) are recruiting during this time and increased taken on August 6, 1999.
sedimentation rates could have important effects on Nalgene bottles (1 l) were acid-washed in a 10% HCl
community development by altering patterns of algal solution and dried. Each plot was sampled by placing a
settlement or survivorship. Although nearly all sedi- bottle in the center of the plot, removing the lid and
ment traps have drawbacks and no trap design has allowing it to fill. Capped bottles were brought to the
been developed specifically for shallow, subtidal habi- surface and immediately placed in an ice-filled cooler
tats, I chose a design that was recommended for use in until being processed in the boat. A subsample of the
strong currents like those characteristic of my system water in each bottle was extracted with a syringe and
(Gardner 1980, Jeurg 1996). Specifically, a cylinder filtered through a glass microfiber filter into a smaller
with a height-to-diameter ratio of at least 4.5 and a bottle which was placed in an ice-filled cooler. The
minimum diameter of 3.9 cm is recommended for remaining water in the 1 l bottles was then discarded.
strong currents to prevent resuspension of sediments Larger bottles were used for collecting the water sam-
that have already been trapped (Gardner 1980). I con- ples underwater in order to facilitate the extraction of
structed cylindrical traps from PVC that were 32 cm in the water to be analyzed once they were brought to the
height and 6 cm in diameter. The traps were attached surface. Samples were immediately transported to the
to the rock substrate just adjacent to each plot by laboratory where they were frozen (and stored) at
screwing a bolt that protruded from the bottom of each –70°C until being shipped to the University of Wash-
trap into a tube of hexagonal nuts that had been ington’s School of Oceanography, where they were
attached to the rock with marine epoxy (Z-Spar splash processed using standard protocols (UNESCO 1994)
zone compound™). Traps were deployed for a single for determination of ammonium, nitrate, nitrite, phos-
10 d sampling period, which spanned portions of both phorus, and silica concentrations.
a spring (when sedimentation should be lowest) and a Kelp growth experiment. I used a kelp growth ex-
neap (when sedimentation should be the highest) tide periment at 2 sites in July 2000 to test the hypothesis
series, and therefore should give an indication of the that Sargassum muticum negatively affects kelp
average effect of S. muticum on sedimentation. Prior growth. I used the native kelp Laminaria bongardiana
to removal, each trap was capped with a PVC lid to (formerly L. groenlandica), which was the most com-
prevent loss of contents during transport to FHL for mon kelp at the 2 removal experiment sites, as the
processing. All samples were immediately frozen at focal species. L. bongardiana plants that were similar
–70°C until they could be processed. in size (8.1 ± 2.6 g = mean ± 1 SD) were collected intact
After removing large animals (e.g. hermit crabs) and from the field and brought to the laboratory where
pieces of algae with a coarse (4 mm) mesh, I used a vac- they were kept in flow-through seawater tanks until
uum pump to filter the sediment onto glass microfiber their return to the field less than 24 h later. Transplant-
filters that had been baked in a muffle furnace at 500°C ing was accomplished as follows. A cinder block
(to remove any organic content) for 6 h. The sediment- (39.5 × 14.5 × 19 cm) with eye bolts inserted at each
laden filters were then dried to a constant weight at end was placed in the area just adjacent to each exper-
60°C , weighed (to the nearest mg), baked for 6 h in a imental plot from which S. muticum had been removed
300°C muffle furnace (to burn off organic content) and (removal) or left intact (control). A single length of
weighed again. This protocol allowed me to separate braided nylon rope, containing 3 L. bongardiana whose
out organic and inorganic components of the sediment. holdfasts had been woven into it, was then attached to
Nutrient measurements. Dense stands of Sargassum each cinder block by tying the ends of the rope to the
muticum have the potential to reduce ambient levels of eye bolts. The cinder block served to elevate trans-
critical nutrients to levels that might negatively affect planted kelps above the substratum slightly, reducing
the growth of native algae. In August 1999, I took a shading by adjacent kelps and thereby better isolating
single water sample from the center of each experi- the effect of S. muticum. Each plant was individually
mental plot to test the hypothesis that S. muticum marked with a piece of numbered flagging tape tied
reduced ambient nutrient concentrations in plots gently around the stipe. Plants were spaced at 10 cm
where it was present. All samples were taken at slack intervals on the ropes and were randomly assigned to
current when water movement is at a minimum, and sites and treatments.
any dilution effect of S. muticum on the local nutrient I used 3 metrics to quantify growth during the 28 d
pool would be most likely to be detected. Because the experiment: (1) change in mass was determined by
Britton-Simmons: Effects of an introduced alga 65
weighing kelps to the nearest 0.1 g before and after tained a single specimen of each of the 3 algal species
transplanting them to the field; prior to weighing, each that had been collected from the field on the day the
plant was held aloft and shaken vigorously and then experiment was set up. Plants were attached to the
patted dry with paper towels to remove excess water. bottom in random order by inserting holdfasts into
(2) Change in blade area was measured by tracing the small incisions that were positioned 10 cm apart in a
blade area of each plant onto butcher paper before and RubbermaidTM bathmat adhering tightly to the bottom.
after transplanting them to the field; blade traces were In addition to being attached at the holdfast, each Sar-
converted to area by cutting them out, weighing them gassum muticum plant was bent over and cable-tied to
and converting those weights to area using a conver- the bathmat at a point approximately 30 cm away from
sion factor determined by weighing pieces of butcher the holdfast. This ensured that all species of algae
paper of known area. (3) Linear blade growth was were equally accessible and experimental results
measured by using a cork borer to remove a circular reflected actual food preferences, unconfounded by
piece of blade tissue 2 cm in diameter at the midpoint the urchin’s inability to access and feed on S. muticum.
of the blade, 4 cm from the meristem region and mea- Urchins were collected from the field on the morning
suring the distance the hole moved distally during the the experiment was set up (May 1, 2001) and kept in a
experiment. flow-through seawater tank until being placed in
Urchin food-preference experiment. Data collected experimental tanks. The 2 sides of each tank were ran-
in the removal experiment showed that green urchins domly assigned to either urchin or control treatments
avoided foraging in control plots, where Sargassum and 6 urchins (6 to 10 cm in diameter) were placed in
muticum was present. I compared the feeding of Stron- each half-tank assigned to the urchin treatment.
gylocentrotus droebachiensis on S. muticum relative to Urchins were assigned to tanks such that each tank
the native kelps Agarum fimbriatum and Laminaria had a range of sizes and total urchin mass per tank was
bongardiana, to test the hypothesis that the urchin’s approximately equal. The density of urchins used in
distaste for S. muticum caused it to avoid S. muticum- this experiment (3.5 urchins m–2) is within the range of
dominated plots. These 2 native kelps were the numer- natural densities I have recorded in the field. Con-
ically dominant canopy species in removal plots at both sumption of each algae species over the 3 d experi-
experimental sites, but they differ considerably in ment was calculated by pairing control and urchin
terms of palatability to urchins. Whereas L. bongar- treatments from each tank and applying the formula
diana is a preferred food of green urchins in this [Ti (C f /Ci )] – Tf , where Ti and Tf are the initial and final
region, A. fimbriatum is consistently avoided in food algae masses in treatment tanks and C i and C f are the
preference experiments (Vadas 1977). This difference initial and final algae masses in the randomly paired
in preference probably reflects the 5 to 10 times higher control tank (Peterson & Renaud 1989). Prior to weigh-
levels of polyphenolics (chemicals known to deter ing algae at the beginning and end of the experiment,
herbivory) in A. fimbriatum than in L. bongardiana each individual was held aloft and shaken gently to
(Hammerstrom et al. 1998). By scaling the green remove excess water from its surface, and then spun in
urchin’s feeding rate on S. muticum against 2 native a salad spinner for 20 s.
kelp species whose relative palatability was already Urchin predation experiment. Increased predation
known, I hoped to better characterize its response to by sea stars in Sargassum muticum-dominated areas
S. muticum as a potential food source. is an alternative hypothesis that could also explain
I had 2 hypotheses about why urchins may not feed the near-absence of green urchins in the control plots
on Sargassum muticum in the field: (1) they find its in the removal experiment at Point George. Algal
tissue unpalatable, or (2) its positively buoyant fronds canopies dominated by S. muticum might enable the
make it inaccessible to urchins. I designed this feeding sea star Pycnopodia helianthoides (a locally common
experiment to test Hypothesis 1 and controlled the predator on green urchins) to prey more effectively on
accessibility of the 3 algae species in this experiment green urchins. Green urchins typically cover their
to ensure that all 3 species were equally available for aboral surface with kelp or other algae while feeding,
urchins to feed on. but S. muticum’s positively buoyant fronds and wiry
The experiment was carried out in 7 circular outdoor morphology make it unsuitable for this purpose.
aquaria (105 cm in diameter, 45 cm tall). All aquaria Therefore, green urchins might be more vulnerable to
had a constant supply of unfiltered seawater from the predation by P. helianthoides in areas where S.
FHL seawater system throughout the experiment. muticum is the dominant alga because they may be
Each aquarium was divided in half with plastic Vexar unable to effectively hide themselves.
mesh (oval holes measuring 2.6 × 1.9 cm); this allowed I used a laboratory experiment in seawater tanks at
water to flow freely throughout the tank while isolat- FHL in May 2003 to test whether predation on urchins
ing urchins on 1 side. Each side of each aquarium con- by sea stars differed in a Sargassum muticum mono-
66 Mar Ecol Prog Ser 277: 61–78, 2004
culture versus a mixed kelp canopy. I divided 7 out- the number of urchins that were visible from above in
door aquaria (105 cm in diameter, 45 cm tall) in half each tank. These individuals should be more vulnera-
using plastic Vexar mesh (oval holes measuring 2.6 × ble to predation by visual predators (e.g. crabs) and
1.9 cm). This setup allowed water to flow freely tactile predators (e.g. sea stars). At the end of the
throughout the tank while preventing urchins from experiment all sea stars were removed from the tanks
moving between the 2 sides of the tanks. All aquaria and the remaining urchins were counted.
had a constant supply of unfiltered seawater from the Statistical analyses. Most statistical analyses were
FHL seawater system throughout the experiment. The carried out using SYSTAT version 9.0 (SPSS). Because
halves of each tank were randomly assigned to 1 of 2 many native species were not present in all removal
experimental treatments: (1) S. muticum monoculture, experiment plots, and because I expected functionally
or (2) mixed canopy of Laminaria bongardiana and similar species to respond similarly to the presence of
Agarum fimbriatum; so that each tank contained Sargassum muticum, I grouped native species into
1 replicate of each treatment. The densities of S. functional groups in addition to carrying out single-
muticum (29 m–2) and kelps (25 m–2) used in the exper- species analyses for the most common native species.
iment were determined from the average densities of Macroalgae were separated into 2 groups: (1) under-
these species in the field. The algae used in the exper- story — small, red species (e.g. Rhodoptilum plumo-
iment were collected on the day preceding and the day sum, Odonthalia spp., Plocamium cartilagineum,
of beginning the experiment, and were attached to the Gigartina papillata, Laurencia spectabilis, Opuntiella
bottom of the tanks by inserting their holdfasts into californica and Callophyllis spp.), and (2) canopy —
small incisions in Rubbermaid™ bathmats adhered large, brown species (Laminaria bongardiana, L. com-
tightly to the bottom of the tanks. I used a mixture of plenata, Agarum cribrosum, A. fimbriatum, Costaria
sizes for each species in order to mimic the size costata, Nereocystis luetkeana, Alaria marginata, Des-
structure typical of algae populations in the field (S. marestia ligulata and D. viridis). Native invertebrates
muticum: 30 to 60 cm tall; A. fimbriatum and L. bon- were separated into 2 main groups: (1) herbivorous
gardiana: 25 to 60 cm tall). The order of species (in the molluscs (Mopalia spp., Cryptochiton stelleri, Tonicella
kelp treatment) and their spatial arrangement on the lineata, Acmea mitra, Diodora aspera, Margarites
mats was haphazard. pupillus and Lacuna vincta), and (2) detritivores (Bit-
Once the algal treatments were set up, I placed 8 tium eschrichtii, Pandalus spp., and Pagurus spp.). The
Strongylocentrotus droebachiensis on each side of data for Strongylocentrotus droebachiensis were ana-
each tank. All urchins were dropped into the tanks lyzed independently from those for other herbivores.
around the perimeter and were allowed to acclimate Because the single-species analyses did not yield any
for 18 h before the sea stars were added. The urchins new insights (with the exception of S. droebachiensis),
were collected over a 1 mo long period preceding the the data presented in this paper are generally for the
experiment and were kept in flow-through seawater functional groupings.
tanks and fed a mixed algal diet prior to their use in Biological data from the removal experiments were
the experiment. I divided the urchins into 2 size classes analyzed using repeated-measures ANOVA, blocking
(50 to 65 mm, and 65 to 80 mm) and randomly selected by site. Site (2 levels, Point George and Colin’s Cove) and
4 urchins from each for each experimental unit. The Treatment (2 levels, control and removal) were both
sea stars Pycnopodia helianthoides (40 to 60 cm in treated as fixed factors. For each response variable, I first
diameter) used in the experiment were collected on the performed a 2-way ANOVA on the pre-removal census
day preceding and the day of beginning the experi- data to test the hypothesis that control and removal
ment, and were kept in flow-through seawater tanks plots differed prior to imposing the removal treatment. A
without food until being placed in the experimental pre-removal difference was detected only for understory
tanks. In order to ensure that the variation in sea star density, and therefore I included the pre-removal data as
size was evenly distributed across the treatments, I a covariate in a repeated-measures ANCOVA analyses
paired sea stars of equal size and randomly assigned for that variable. The assumption of normality was tested
them to opposite sides of the same tank until each tank using a Kolmogorov-Smirnov test (α = 0.05). The as-
had 1 sea star per side. The tanks were lightly shaded sumption of equality of variances was tested using
using a double layer of thick, cotton fishing net (4 cm2 an F max test (α = 0.05; Sokal & Rohlf 1995). Data were
mesh size) in order to mimic the shallow subtidal transformed as necessary using square-root, arcsine and
light conditions where these species normally interact. natural log functions to conform to the assumptions of
Additional light reduction was achieved by wrapping ANOVA. I also performed power analyses using GPower
the south-facing half of each tank in a layer of black (Faul & Erdfelder 1992) on each response variable that
plastic. This experiment was allowed to run for a total showed no response to the Sargassum muticum removal
of 8 d. On Days 3 and 8 of the experiment I recorded treatment.
Britton-Simmons: Effects of an introduced alga 67
I performed an additional analysis on the ratio of the I used the SEM module in Statistica for Windows
number of Laminaria bongardiana to the number of (Release 6.0) to estimate unstandardized structural
Agarum fimbriatum in removal experiment plots. For coefficients using a maximum likelihood algorithm.
this analysis I only used data from the last 5 censuses in The statistical significance of structural equation
order to exclude the transitory dynamics which ap- coefficients was determined by the software, using
peared to occur in removal plots following the removal multiple regression for each set of dependent and
of Sargassum muticum. Unfortunately, the large num- independent variables. Alternative models were con-
ber of zeros in the data set (primarily in the control structed using information from my experiments, the
plots, where there were few kelps) precluded both the published literature, and my own knowledge of the
calculation of ratios and the use of ANOVA. Instead, I natural history of the system. For the purposes of SEM
took the average abundance of each of the kelp spe- analyses, I excluded canopy algae that are known to
cies across all replicates for each treatment on each be unpalatable to Strongylocentrotus droebachiensis
sampling date and used those averages to calculate the (Vadas 1968) because a preliminary analysis of the
ratio of the 2 species for each treatment on each sam- data from Point George indicated that urchins
pling date. Because these data points are not indepen- responded strongly to those species that are known to
dent (they are repeated measures), I used the non- be preferred food items based on laboratory feeding
parametric Scheirer-Ray-Hare test (extension of the trials (Vadas 1968) and weakly to the remaining,
Kruskal-Wallis; Sokal & Rohlf 1995) to determine unpalatable species. Palatable canopy species typi-
whether the ratio of L. bongardiana to A. fimbriatum cally contain low levels of polyphenolics, and this
differed between treatments. category included Alaria marginata, Laminaria bon-
In addition to the ANOVA analyses, I analyzed the gardiana, L. saccharina, L. complenata and Nereocys-
abundance data for a subset of the taxa from my tis luetkeana. In contrast, unpalatable canopy species
removal experiment at Point George using structural typically have high levels of polyphenolic compounds
equation modeling (SEM), a form of multiple regres- or other chemical defenses against herbivory, and
sion related to path analysis (Hayduk 1987, Shipley these species included Agarum fimbriatum, A. cribro-
2000), to test whether the effect of Sargassum muticum sum, Desmarestia viridis and D. ligulata.
on Strongylocentrotus droebachiensis was a direct or I tested 2 alternative models to determine whether
indirect effect. SEM is a type of statistical analysis in urchins were responding to the presence of Sargassum
which causal relationships between variables (species muticum or the absence of native algae. The presence
in this case) are hypothesized in the form of an inter- of unpalatable S. muticum might directly affect urchins
action web and tested using a system of linear equa- if their foraging behavior differs in areas with and
tions (Hayduk 1987). Structural equation coefficients without S. muticum. The presence of unpalatable
are calculated using a maximum likelihood algorithm algae species has been shown to alter the feeding
and the significance of each coefficient is tested using behavior of herbivores on palatable algae when both
multiple regression. Structural equation coefficients types of algae are growing together in a phenomenon
indicate how a change in the predictor variable would called an ‘associational plant refuge’ (Pfister & Hay
change the target variable, holding all other variables 1988). For example, an unpalatable congener of S.
constant. The net effect of an indirect pathway be- muticum altered the feeding behavior of the urchin
tween 2 species that involves multiple links can be cal- Arbacia punctulata on palatable red algae when both
culated by multiplying the coefficients for the relevant types of algae were present (Pfister & Hay 1988). Alter-
links (Wootton 1994). A predicted correlation matrix natively, S. muticum could indirectly affect green
between the variables is calculated based on the spec- urchins by competing with their preferred prey spe-
ified model and compared to the actual correlation cies. The first model contained a direct pathway
matrix, using a χ2 distribution, to ask whether the two between S. muticum and urchins, in addition to an
differ significantly. A non-significant χ2 statistic indi- indirect pathway (via palatable canopy). The second
cates that the causal relationships specified in the model lacked the direct pathway in order to test
model cannot be rejected as a good caricature of the whether removing that link altered the fit of the model
species interactions in nature. The relative fit of alter- to the data. Because I experimentally manipulated the
native models can be compared by comparing their χ2 presence of S. muticum, I included treatment as a vari-
statistics (when models are nested) as well as their able in the models. I used a combination of 2 statistics
Akaike information criterion (AIC) values. AIC is an to assess the fit of the models to the data: the χ2 test
information-theoretic criterion for model selection that statistic and the AIC.
takes into account both the goodness of fit of the model Strongylocentrotus droebachiensis abundance data
as well as the complexity of the model required to from Point George (green urchins were not present at
achieve that fit (Burnham & Anderson 1998). Colin’s Cove) were analyzed separately because this
68 Mar Ecol Prog Ser 277: 61–78, 2004
species showed variability in abundance independent height of the Sargassum muticum fronds in control
of other taxa. On many sampling dates there were no plots varied over the course of my light measurements
urchins in any control plot (i.e. there was no variance), and I expected their effect on light to increase as they
precluding the use of repeated-measures ANOVA to grew, and subsequently decrease as they senesced at
analyze these data. Instead, I took the average across the end of the summer. I used the sequential Bonfer-
all sample dates for each plot and performed a t-test on roni method (Sokal & Rohlf 1995) to correct my critical
the square-root transformation of these data to look for p-value for multiple comparisons within each site.
an overall effect of the treatment on average urchin I analyzed urchin algal preference data using non-
abundance. parametric statistics because transformations failed to
Sediment, nutrient, water flow and kelp growth data make these data homoscedastic. I tested the null hypo-
were each analyzed using a 2-way ANOVA with site thesis that Strongylocentrotus droebachiensis con-
and treatment as fixed factors. sumed all 3 algal species presented to it equally using
Light data were analyzed using a 2-way repeated- a Kruskal-Wallis non-parametric test. I then used non-
measures ANOVA with site and treatment as fixed fac- parametric unplanned comparisons (Zar 1999) to de-
tors. Percent transmission light data were arcsin-trans- termine how prey differed from one another.
formed prior to analyses. I followed up the ANOVA on I used a t-test to compare the number of urchins
light data with a 1-tailed t-test (because I had an a pri- eaten in tanks with Sargassum muticum versus tanks
ori expectation that light would be lower in control containing native kelp species in order to determine
plots than removal plots) on each sampling date to whether predation risk differed between those 2 treat-
determine on which dates light differed significantly. ments. The data were log-transformed prior to analysis
This follow-up analysis was important because the to achieve homoscedasticity.
Table 1. Effect of Sargassum muticum on native algae. Results of
ANOVA and ANCOVA (for understory abundance) testing effect
RESULTS
of S. muticum removal
Removal experiments
Source of variation SS df MS F p
The structure of the native algae community
Canopy abundance was substantially altered by the removal of Sar-
Site 4.74 1 4.74 2.30 0.149 gassum muticum. Canopy algae were less abun-
Treatment 47.06 1 47.06 22.86 0.000
Site × Treatment 3.03 1 3.03 1.47 0.242 dant in control plots (p < 0.001; Table 1, Fig. 1)
Error 32.93 16 2.06 and this effect did not differ between the 2 sites
Time 13.50 13 1.04 3.90 0.000 (Table 1). The significant time × site interaction
Time × Site 8.14 13 0.63 2.35 0.006 (p = 0.006; Table 1, Fig. 1) indicates that the
Time × Treatment 6.65 13 0.51 1.92 0.029
Time × Site × Treatment 5.59 13 0.43 1.61 0.083 temporal dynamics of canopy abundance differed
Error 55.40 208 0.27 between the sites and was probably caused by
Understory abundance the somewhat delayed response of canopy algae
Site 6.77 1 6.77 4.95 0.042 to removal at Colin’s Cove (Fig. 1). Finally, a sig-
Treatment 12.05 1 12.05 8.81 0.009 nificant time × treatment interaction (p = 0.029;
Site × Treatment 0.03 1 0.03 0.02 0.889
Covariate (June 1999) 8.16 1 8.16 5.97 0.027
Table 1, Fig. 1) showed that the difference in
Error 20.52 15 1.37 canopy abundance between the 2 treatments
Time 2.37 9 0.26 1.87 0.061 fluctuated over time, which could be explained
Time × Site 2.93 9 0.32 2.31 0.019 by the slower response of canopy algae at
Time × Treatment 0.51 9 0.06 0.40 0.931
Colin’s Cove as well as the increase in canopy
Time × Site × Treatment 1.38 9 0.15 1.09 0.372
Time × Covariate 1.38 9 0.15 1.09 0.375 abundance in Colin’s Cove controls in early 2000
Error 18.99 135 0.14 relative to Point George. Although there was
Canopy richness not a significant time × site × treatment interac-
Site 2.75 1 2.75 3.77 0.070 tion (p = 0.083, Table 1), it did appear that the
Treatment 10.87 1 10.87 14.92 0.001 recovery of canopy algae in removal plots was
Site × Treatment 0.48 1 0.48 0.66 0.427
Error 11.65 16 0.73 delayed at Colin’s Cove relative to Point George
Time 7.96 13 0.61 4.48 0.000 (Fig. 1). Despite this possible difference in the
Time × Site 3.25 13 0.25 1.83 0.041 timing of recovery, native canopy algae were 4
Time × Treatment 2.15 13 0.16 1.21 0.275 to 5 times as abundant in removal plots com-
Time × Site × Treatment 4.98 13 0.38 2.80 0.001
Error 28.43 208 0.14 pared to control plots at both sites by the last
census date.
Britton-Simmons: Effects of an introduced alga 69
Understory algae were also less abundant in control
plots (p = 0.009; Table 1, Fig. 2) and this effect did not
differ between sites (Table 1). Including pre-removal
understory abundance as a covariate explained a sig-
nificant amount of variation in the ANCOVA model
(p = 0.027; Table 1). As with the data for canopy algae,
there was a significant time × site interaction (p = 0.019;
Table 1), which was probably caused by the somewhat
delayed recovery of understory algae in removal plots
at Colin’s Cove (Fig. 2). Although the recovery of
understory algae at Colin’s Cove lagged behind that at
Point George slightly, by the end of the experiment
understory algae were twice as abundant in removal
plots as control plots at both sites compared to a similar
100% initial difference at Point George and a 62%
initial difference at Colin’s Cove.
Native canopy richness was lower in control plots
than removal plots (p = 0.001; Table 1, Fig. 3). On aver-
age, Sargassum muticum displaced 1 native species at
both sites. As was the case for canopy and under-
story abundance, a significant time × site interaction
(p = 0.041; Table 1) probably was caused by the some-
what delayed recovery of canopy algae in removal
Fig. 2. Abundance (mean ± SE) of native understory algae in
Sargassum muticum removal experiments at Point George and
Colin’s Cove (n = 5). First data point in each series is for
pre-removal census
plots at Colin’s Cove (Fig. 3). Finally, the significant
time × site × treatment interaction (p = 0.001; Table 1)
was caused by a lack of concordance in the temporal
dynamics of the treatment effect at the 2 sites. During
1999, the control and removal means diverged at Point
George but remained roughly equal at Colin’s Cove
(Fig. 3). The following year, in 2000, the treatment
means at Point George began to converge again while
those at Colin’s Cove had just begun to diverge (Fig. 3).
The relative abundance of the 2 most common native
kelp species, Laminaria bongardiana and Agarum fim-
briatum, differed between control and removal plots.
The ratio of L. bongardiana to A. fimbriatum was lower
in control plots, where Sargassum muticum was
present (Scheirer-Ray-Hare extension of the Kruskal-
Wallis test, H = 7.22, p < 0.01; Fig. 4). This effect did
not differ between sites (H = 0.63, p < 0.5), and there
was no indication of a site × treatment interaction
(H = 0.37, p < 0.9).
The abundance of green urchins was negatively
Fig. 1. Abundance (mean ± SE) of native canopy algae in
Sargassum muticum removal experiments at Point George
affected by the presence of Sargassum muticum. Al-
and Colin’s Cove (n = 5). First data point in each series is for though there were no urchins in any plot during the
pre-removal census first summer of the experiment, a year later they had
70 Mar Ecol Prog Ser 277: 61–78, 2004
begun to forage regularly in removal plots at Point all sample dates for each plot showed that urchins
George (Fig. 5). Their absence during the first summer were significantly more abundant in removal plots
of the experiment probably reflects the near-absence than control plots (t 8 = –6.34, p < 0.001; Fig. 5) at Point
of kelps, their preferred food, in experimental plots. George. Green urchins were never observed at Colin’s
The appearance of urchins in removal plots was pre- Cove, but red urchins (S. franciscanus) were present at
ceded by an increase in canopy algae in removal plots deeper depths at that site throughout the experiment.
at Point George (Fig. 1). A t-test on the average Several variables showed no response to Sargassum
Strongylocentrotus droebachiensis abundance across muticum manipulation. I found no evidence that
S. muticum altered the percent cover of crustose
coralline algae, filamentous turf-forming algae, or bare
rock (Table 2). Similarly, there was no evidence for an
effect of S. muticum on the abundance of detritivores,
the abundance of herbivorous molluscs, or the species
richness of the invertebrate community (Table 2).
Power analysis revealed that I had low statistical power
to detect effects on some of these variables (Table 2).
However, in several cases (bare rock, crustose algae
and invertebrate richness) the extremely small effect
sizes involved would have made it difficult to detect a
significant effect, even if a more powerful experimen-
tal design had been employed. In the case of turf-form-
ing algae, it appears that a large amount of variation
may have obscured the effect of S. muticum (Table 2).
The results of the structural equation modeling
(SEM) are presented in Fig. 6. Structural equation
coefficients (values next to arrows in Fig. 6) indicate
the sign and strength of the effect of one variable on
another. More specifically, each coefficient indicates
the change in abundance of the dependent variable
that would result from a 1-unit change in the predictor
variable. Thick arrows indicate statistically significant
pathways (p < 0.05) and thin arrows indicate non-
significant paths.
The SEM results suggest that the effect of Sargas-
sum muticum on green urchins was an indirect effect,
Fig. 3. Species richness (mean ± SE) of native canopy algae in
Sargassum muticum removal experiments at Point George
and Colin’s Cove (n = 5). First data point in each series is for
pre-removal census
Fig. 5. Strongylocentrotus droebachiensis. Abundance (mean
Fig. 4. Ratio of the number of Laminaria bongardiana (L.b.) to ± SE) in Sargassum muticum removal experiment at Point
the number of Agarum fimbriatum (A.f.) (n = 5). Both species George (n = 5). First data point in each series is for pre-
are native kelps removal census
Britton-Simmons: Effects of an introduced alga 71
Table 2. Response variables that showed no statistically significant (α = 0.05) response to removal of Sargassum muticum (n = 5 for
each mean). Rock, turf-forming algae and coralline algae data are % cover. All abundance data are means ± 1 SE for each
experimental plot averaged across all sampling dates. Power = 1 – β, or probability of correctly rejecting H o if it were false
Variable Point George Colin’s Cove Power
Control Removal Control Removal
Bare rock 04.9 ± 2.3 12.1 ± 3.4 07.9 ± 4.3 01.1 ± 0.6 0.05
Crustose coralline algae 21.5 ± 7.9 34.1 ± 9.0 18.7 ± 6.1 20.1 ± 5.9 0.20
Turf-forming algae 49.1 ± 8.8 34.5 ± 8.3 53.5 ± 7.7 42.4 ± 5.1 0.40
Invertebrate richness 07.1 ± 0.6 07.8 ± 0.4 04.7 ± 0.5 04.7 ± 0.6 0.07
Detritivores 13.5 ± 3.0 05.8 ± 1.4 06.1 ± 1.4 06.8 ± 2.0 0.54
Herbivorous molluscs 07.0 ± 1.0 11.0 ± 0.9 04.0 ± 1.1 04.0 ± 1.2 0.62
Light
The overall ANOVA on light data from both sites
indicated a highly significant effect of treatment
(p < 0.001; Table 3, Fig. 7) on light transmittance, and
this effect did not differ between the 2 sites (p = 0.570;
Table 3). These results reflect the fact that light inten-
sity was (on average) more intense in removal plots
than control plots at both sites (Fig. 7), even though the
absolute difference between treatments varied over
time as Sargassum muticum’s fronds grew, matured,
Fig. 6. Structural equation model for Point George data in
and senesced. There was also a significant site effect
Sargassum muticum removal experiment (urchins = Strongy-
locentrotus droebachiensis; palatable canopy = Laminaria (p = 0.001; Table 3), with percent transmission of light
bongardiana, L. complenata, Alaria marginata, Nereocystis at Point George higher than at Colin’s Cove (Fig. 7).
luetkeana). Arrows indicate direction of causality. Thick This systematic difference is probably due to the fact
arrows indicate statistically significant (p < 0.05) paths from that the shoreline at Point George faces W–SW and
multiple-regression analysis; structural equation coefficients
are shown next to each arrow. Removing direct link between
receives more intense sunlight than the E-facing shore
S. muticum and urchins did not change model fit or values of at Colin’s Cove. Finally, a significant time × treatment
the remaining coefficients. Observed correlation matrix interaction (p < 0.001; Table 3) indicated that the treat-
did not differ significantly from that predicted by model ment effect changed over time. To determine when
(χ22 = 5.08, p = 0.08; Akaike information criterion, AIC = 0.267)
significant differences in light level occurred, I fol-
lowed up this analysis with t-tests on each sampling
date within the 2 sites. A 1-tailed t-test on each date
not a direct effect (Fig. 6). The direct pathway from S. revealed that light transmittance was lower in control
muticum to urchins was very weak and was not statis- plots at both sites in May (p < 0.05 and p < 0.01 at Point
tically significant (coefficient = 0.03, p = 0.90; Fig. 6). In George and Colin’s Cove, respectively; Fig. 7) and
contrast, the net effect of the indirect pathway from S. June (p < 0.001 at Point George and Colin’s Cove;
muticum to urchins, obtained by multiplying the coef- Fig. 7).
ficients for each of the links involved (Wootton 1994),
was very strong (–1.07 × 0.94 = –1.01) and both of the
Table 3. Results of ANOVA testing effect of Sargassum muticum
component pathways (S. muticum to palatable canopy on light transmittance in removal experiments
and palatable canopy to urchins) were statistically sig-
nificant. Removing the direct link between S. muticum Source of variation SS df MS F p
and urchins did not change the model fit or the values
of the remaining coefficients. In fact, the 2 models Site 0.128 1 0.128 16.07 0.001
were statistically indistinguishable (χ22 = 5.08, p = 0.08, Treatment 0.251 1 0.251 31.54 0.000
Site × Treatment 0.003 1 0.003 00.34 0.570
AIC = 0.267 for both models), providing additional evi-
Error 0.127 160 0.008
dence that the direct effect of S. muticum on urchins Time 0.696 3 0.232 70.75 0.000
was negligible. In summary, the SEM analysis sug- Time × Site 0.009 3 0.003 00.95 0.423
gests that the negative effect of S. muticum on urchins Time × Treatment 0.239 3 0.080 24.31 0.000
was an indirect effect that also involved palatable Time × Site × Treatment 0.021 3 0.007 02.18 0.103
Error 0.157 480 0.003
canopy algae (Fig. 6).
72 Mar Ecol Prog Ser 277: 61–78, 2004
Fig. 8. Dissolution of gypsum blocks (mean ± SE) in experi-
mental plots where Sargassum muticum was present (Con-
trol) and in plots from which it had been removed (Removal)
more numerous in removal plots (Figs. 1 & 2). The
deposition of organic sediment was not different in
control and removal plots (ANOVA, F1,14 = 3.73, p =
0.074; Fig. 9). The deposition of inorganic sediment did
not differ between the treatments either (ANOVA,
F1,14 = 0.43, p = 0.524), but there was a significant
site × treatment interaction for this response variable
(ANOVA, F1,14 = 5.42, p = 0.035). This interaction prob-
ably occurred because inorganic sedimentation tended
Fig. 7. Percent transmission (mean ± SE) of photosynthetically to be slightly higher in removal plots than control plots
active radiation (400 to 700 nm, PAR) in Sargassum muticum at Point George, but higher in control than removal
removal experimental plots at Point George and Colin’s Cove plots at Colin’s Cove (Fig. 9). Finally, total sediment
in 2000. Secondary y-axis shows height of S. muticum (± SE) deposition did not differ between the 2 treatments
in control plots at time each set of light measurements was
taken. Asterisks indicate sample dates where t-test indicated (ANOVA, F1,14 = 0.96, p = 0.344), but it also showed
significant difference between treatments (*, **, ***: p < 0.05, some suggestion of a site × treatment interaction
0.01, and 0.001, respectively) (ANOVA, F1,14 = 4.37, p = 0.055), which likely resulted
for the same reason stated above for inorganic sedi-
mentation. Overall, there was no indication that S.
Water flow muticum removal had an effect on sedimentation by
the time that native algae had recovered to replace the
I found no evidence that the Sargassum muticum S. muticum that had been removed 2 yr earlier.
removal treatment had an effect on water flow at a
distance of 30 cm above the substratum (ANOVA,
F1,15 = 0.04, p = 0.85; Fig. 8). However, water flow at Nutrients
Point George was considerably higher than at Colin’s
Cove (ANOVA, F1,15 = 65.29, p < 10– 5; Fig. 8). There None of the 5 nutrients assayed differed between
was no evidence of a site × treatment interaction. control and removal plots (Table 4) (ANOVA, F1,16,
Sedimentation Table 4. Nutrient concentrations (µM means ± 1 SE) of water samples taken from
Sargassum muticum removal experiment plots (n = 5 for each mean)
Analysis of sedimentation data col-
Nutrient Point George Colin’s Cove
lected in July 2001 yielded no
Control Removal Control Removal
evidence that Sargassum muticum
altered sedimentation rates. At the PO4 01.96 ± 0.012 01.97 ± 0.007 1.91 ± 0.007 01.89 ± 0.013
time these data were collected, canopy Si(OH)4 49.05 ± 0.256 48.56 ± 0.195 46.97 ± 0.2360 47.04 ± 0.113
and understory algae had already NO3 22.47 ± 0.088 22.46 ± 0.129 21.88 ± 0.0960 21.67 ± 0.374
NO2 00.31 ± 0.003 00.30 ± 0.002 0.30 ± 0.002 00.30 ± 0.002
responded to the S. muticum removal NH4 00.52 ± 0.015 00.53 ± 0.021 0.57 ± 0.010 00.54 ± 0.028
treatment, with both types of algae
Britton-Simmons: Effects of an introduced alga 73
of replicates. At Colin’s Cove, 1 of the control plots lost
all its plants, and consequently there were only 4 con-
trol replicates from that site. There was no statistically
significant relationship between site (ANOVA, p =
0.826) or treatment (ANOVA, p = 0.826) and the num-
ber of plants lost. Likewise, there was no relationship
between the number of plants lost and growth of the
remaining kelps (R21,17 = 0.13, p = 0.133).
Linear blade growth of Laminaria bongardiana was
2 to 3 times faster in plots from which Sargassum muti-
cum was absent than in those where it was present
(ANOVA, F1,15 = 25.95, p < 0.001; Fig. 10). Growth did
Fig. 9. Sediment accumulation (mean ± SE) in traps placed in not differ between the 2 sites (ANOVA, F1,15 = 0.20, p >
experimental plots at Point George and Colin’s Cove where 0.30) and there was no site × treatment interaction
Sargassum muticum was present (Control) and in plots from
(ANOVA, F1,15 = 0.11, p > 0.73). Analysis of blade area
which it had been removed (Removal). Total sediment is sum
of organic and inorganic components of sediment and kelp mass data yielded comparable results (data
not shown).
Urchin food preferences
Strongylocentrotus droebachiensis distinguished
among the 3 species of algae that were presented to
it in the preference experiment (Kruskal-Wallis, p =
0.001; Fig. 11). Non-parametric unplanned compar-
isons (Zar 1999) showed that Laminaria bongardiana
was preferred over both Agarum fimbriatum (p < 0.01)
and Sargassum muticum (p < 0.005), but that A. fim-
briatum and S. muticum were equally ignored. This
experiment supported the hypothesis that green
urchins do not feed on S. muticum in the field because
Fig. 10. Laminaria bongardiana. Growth (mean ± SE) of
native kelp in plots where Sargassum muticum was present they found its tissue unpalatable. Nevertheless, even if
(Control) and plots from which it had been removed they find S. muticum palatable its morphology could
(Removal) at Point George and Colin’s Cove (n = 5) prevent them from effectively exploiting it as a food
resource in the field.
p > 0.30 in every case). However, phosphate (ANOVA,
F1,16 = 40.24, p < 0.001), silicate (ANOVA, F1,16 = 75.18,
p < 0.001) and nitrate (ANOVA, F1,16 = 10.92, p < 0.01)
concentrations were significantly higher at Point
George (Table 4). Since water samples were taken on
different (but consecutive) days at each site, these site
differences may reflect day-to-day fluctuations as
water masses move through this region.
Kelp growth
Each plot began the kelp growth experiment with 3
Laminaria bongardiana, but several plots lost 1 or more
plants during the course of the experiment. I averaged Fig. 11. Algal mass (mean + SE) consumed by native sea
the results for all surviving plants within each plot and urchin Strongylocentrotus droebachiensis in food choice
experiment where 2 native kelps (Laminaria bongardiana and
used those means in the statistical analysis. Thus, the Agarum fimbriatum) and Sargassum muticum were offered
extra plants in each plot increased the accuracy of the simultaneously. Letters indicate which means differ signifi-
growth measurements but did not increase the number cantly (non-parametric unplanned comparisons)
74 Mar Ecol Prog Ser 277: 61–78, 2004
interaction (F1,12 = 4.32, p = 0.08), the difference
between the 2 treatments did appear to decline from
Days 3 to 8 (Fig. 13). Despite the absence of an over-
all difference between treatments, further analysis
showed that significantly more urchins were visible in
the S. muticum treatment on Day 3 (t = –2.843, p =
0.01), but that this difference had dissipated by Day 8
(t = –0.830, p = 0.42). Even though there was little
difference between the treatments at the end of the
experiment (Day 8), it does appear that a higher pro-
portion of urchins in the kelp treatment were hidden
for the majority of the experiment (Fig. 13). Contrary to
expectation, the urchins in the kelp treatment were not
Fig. 12. Strongylocentrotus droebachiensis. Number of green
urchins (mean + SE) eaten by sea star Pycnopodia heliantho- less vulnerable to sea star predation (Fig. 12). This
ides in tanks containing Sargassum muticum monoculture experiment does provide some evidence that green
or mixture of native kelp species (n = 7) urchins can more easily hide themselves in algal
canopies dominated by native kelps than in those dom-
inated by S. muticum. Furthermore, these data are a
Urchin predation conservative indication of the differential use of S.
muticum and kelps by urchins, because in this labora-
The sea star predation hypothesis was not supported tory experiment urchins were able to make use of S.
by experimental data. There was no difference in pre- muticum in a way they cannot commonly do in the
dation by Pycnopodia helianthoides on Strongylocen- field. Most of the urchins that managed to hide them-
trotus droebachiensis between the Sargassum muti- selves in the S. muticum treatment did so by climbing
cum and mixed-kelp treatments (t = –0.632, p = 0.545; the walls of the tank, grasping onto the fronds of
Fig. 12). While the average number of urchins visible nearby S. muticum, and using those fronds to cover
(≥ 25% of test) was greater on Days 3 and 8 (Fig. 13), their tests (own pers. obs.). Although S. muticum does
repeated-measures ANOVA analysis indicated no grow adjacent to vertical rock surfaces in the field at
statistically significant difference between treatments some sites (e.g. at the base of walls or next to large
(F1,12 = 4.30, p = 0.06). There was a significant effect of boulders), these places are relatively rare and I have
time in the ANOVA model (F1,12 = 15.87, p = 0.002), never observed green urchins using S. muticum to
which resulted because both treatments means de- hide themselves in this manner in the field. Overall,
creased between the 2 sampling dates (Fig. 13). the results of this experiment suggest that increased
Although there was no significant time × treatment predation by sea stars in S. muticum-dominated areas
cannot explain why green urchins were less abundant
in control plots in the S. muticum removal experiment.
DISCUSSION
Removal experiments showed that Sargassum muti-
cum has substantial direct effects on the native algae
community characteristic of the San Juan Islands.
Competition with S. muticum reduced the abundance
of native canopy algae by approximately 75% and
native understory algae by about 50% (Figs. 1, 2 & 6).
S. muticum also displaced (on average) 1 native spe-
cies of canopy algae (Fig. 3), thereby reducing the spe-
cies richness of native canopy species, but leaving total
algal richness unchanged. However, the negative
effect on species richness that occurred at the small
Fig. 13. Strongylocentrotus droebachiensis. Number of visible scale of my experimental plots (0.25 m2) may not be
green urchins (mean ± SE) in tanks containing mixture of
native kelps versus tanks containing Sargassum muticum
important at larger scales.
(n = 7). Treatment means differed on Day 3 (t-test, p = 0.01) In addition to affecting the abundance and richness
but not on Day 8 (t-test, p = 0.42) of native canopy algae, Sargassum muticum changed
Britton-Simmons: Effects of an introduced alga 75
the relative abundance of the 2 most common native this scenario, palatable algae gain protection from her-
kelp species, Laminaria bongardiana and Agarum fim- bivores when they are spatially associated with un-
briatum. L. bongardiana was relatively less abundant palatable algae because herbivores avoid foraging in
in plots where S. muticum was present (Fig. 4). This areas where the unpalatable species are present. This
result was probably caused by interspecific differences hypothesis was not supported by SEM analyses of my
in light requirements. Laminaria bongardiana is a spe- removal experiment data (Fig. 6). (2) Algal canopies
cies that has few chemical defenses (Hammerstrom et dominated by S. muticum could enable the sea star
al. 1998), grows relatively fast, is found only in the Pycnopodia helianthoides (a locally common green
shallow subtidal zone (own pers. obs.), and grows urchin predator) to prey more effectively on green
more slowly when transplanted beneath S. muticum urchins. Green urchins typically cover their aboral sur-
(Fig. 10). A. fimbriatum, on the other hand, has high face with kelp or other algae while feeding, but S.
concentrations of polyphenolics to deter herbivory muticum’s positively buoyant fronds and wiry mor-
(Hammerstrom et al. 1998), grows relatively slowly, phology could make it unsuitable for this purpose.
and is the deepest-occurring kelp species in this region Therefore, green urchins might be more vulnerable to
(commonly found at depths of 15 m or more; own pers. predation by foraging sea stars in areas where S.
obs.). Thus, L. bongardiana is probably more sensitive muticum is the dominant alga. A laboratory experi-
than A. fimbriatum to the shading caused by S. muti- ment designed to test this hypothesis showed that
cum (Fig. 7), and was relatively less abundant in although green urchins seemed better able to conceal
control plots as a result. themselves when feeding on native kelps compared to
Field measurements suggest that the effects of Sar- S. muticum (Fig. 13), there was no difference in sea
gassum muticum on macroalgae are probably a result star (P. helianthoides) predation between the 2 treat-
of shading (Fig. 7). There was no evidence that S. ments (Fig. 12). My conclusion that S. muticum nega-
muticum had an effect on nutrients (Table 4), and be- tively affects urchins is in accordance with a study of
cause algae in this region do not appear to be nutrient- S. muticum in California tidepools that found a similar
limited (Wootton 1991, Pfister & Van Alstyne 2003) it result for a congener, Strongylocentrotus purpuratus
is unlikely that competition for nutrients plays an (Wilson 2001). However, the indirect effect that was
important role in these interactions. However, my the cause of the negative effect on urchins in my sys-
limited sampling of water flow (Fig. 8) and sedimenta- tem has not been demonstrated in any previous study.
tion (Fig. 9) makes it difficult to rule out these factors Although the removal experiments at Point George
completely, and other resources I did not take into and Colin’s Cove yielded largely similar results, there
account (e.g. space) could also be important. were some notable differences between the 2 sites. For
Structural equation modeling (SEM) showed that example, the recovery dynamics of canopy algae fol-
Sargassum muticum-induced changes in palatable lowing the removal of Sargassum muticum differed
canopy algae had an important indirect effect on a between them (Fig. 1). Whereas the recovery at Point
native herbivore, the green sea urchin Strongylocen- George was rapid, it was considerably delayed at
trotus droebachiensis (Fig. 6). Green urchins were not Colin’s Cove. Source-populations of native canopy and
recorded in experimental plots at either site during the understory species were only 2 to 3 m deeper than
first year of the experiments (Fig. 5), although they my experimental plots at both sites. Under most con-
were present at Point George at depths deeper than ditions, kelps are unlikely to be strongly dispersal-
my experiments (own pers. obs.). However, by the limited over this distance (Reed et al. 1988, Fredriksen
summer of 2000, green urchins had begun foraging et al. 1995, Forrest et al. 2000). Nevertheless, stronger
regularly in the removal plots at Point George, where a tidal currents at Point George (Fig. 8) may have facili-
robust community of kelps and red algae had devel- tated the dispersal of algae into plots from which S.
oped following the removal of S. muticum (Fig. 5). It muticum had been removed.
appears that green urchins avoided control plots The presence of green urchins at Point George but
because the kelp genera Laminaria, Alaria and Nereo- not Colin’s Cove was another source of variation
cystis, which are their preferred food (Fig. 11 and between sites. These urchins can be locally abundant,
Vadas 1968, 1977, Larson et al. 1980), were less abun- but have a discontinuous distribution across sites. Thus
dant due to competition with S. muticum (Fig. 1). it is not surprising that they were present at only 1 of
I also tested 2 additional hypotheses that might my sites. Canopy algae abundance in removal plots
explain why green urchins avoided plots where Sar- at Point George decreased following the arrival of
gassum muticum was abundant: (1) The presence of urchins at that site in the spring of 2000 (Fig. 1). Native
unpalatable S. muticum could reduce grazing by sea canopy algae richness was higher in the absence of
urchins on nearby palatable native species (so-called Sargassum muticum at both sites, but urchin grazing in
associational plant refuge; e.g. Pfister & Hay 1988). In the removal plots at Point George caused a decline in
76 Mar Ecol Prog Ser 277: 61–78, 2004
canopy richness during 2000 and briefly led to the 2000, own pers. obs.), where it is susceptible to desic-
convergence of control and treatment dynamics at cation and frost (Norton 1977), and it generally reaches
that site (Fig. 3). lower densities in the intertidal compared with the
Because kelps are the numerically and physically subtidal, even in tidepools (Wilson 2001, own pers.
dominant plants in these algal communities, it would obs.). Thus, one might expect it to have less of an
be useful to know whether only 1 or both phases of impact on native species in the intertidal simply
their life cycles are affected by Sargassum muticum. because it is less abundant there.
The lowest light levels I recorded in plots containing Although different sites in the San Juan Islands vary
S. muticum (29 and 36 µE m–2 s–1 at Colin’s Cove and considerably in the density of Sargassum muticum
Point George, respectively) are below the threshold at they contain, it is extremely difficult to find sites which
which light limitation in kelp sporophytes is expected it has not yet invaded (own pers. obs.). Given the
(150 to 200 µE m–2 s–1, Lüning 1981), and growth results of these experiments, 2 impacts may be of most
experiments clearly showed that sporophytes grew concern: First, although my experiments were con-
more slowly under the S. muticum canopy (Fig. 10). ducted on a small scale, it seems likely that the total
However, these light levels exceed the threshold abundance of native kelp in the San Juan Islands has
where kelp gametophytes are saturated for vegetative been reduced by S. muticum. Since a wide variety of
growth (20 µE m–2 s–1, Lobban & Harrison 1994). Thus, taxa, including even marine mammals and birds, uti-
the effect of S. muticum on the kelp component of the lize these shallow subtidal kelp communities, the con-
algal community is probably due to its impact on the sequences of this invasion may extend well beyond the
sporophyte phase, not the gametophyte phase of the benthic organisms that were the focus of this study.
kelp life cycle. One important caveat to this conclusion Second, the avoidance by green urchins of areas with
is that my light measurements reflect maximum irradi- dense S. muticum has important potential indirect
ance values because the data were taken at midday; effects for kelp communities. Urchins are an important
average light intensity was undoubtedly much lower. disturbance agent because they clear patches of rock,
In general, many species of native algae are likely to thereby resetting the successional sequence of the
be affected due to shading by Sargassum muticum community (Vadas 1968, Duggins 1980). As a conse-
because of the timing of its life cycle in this region. The quence, algal diversity is enhanced by the creation of
fronds of S. muticum usually begin growing in early a mosaic of patches that differ in their successional
March each year, and in 2000 it was already having stages (Vadas 1968, Duggins 1980). The absence of
a significant effect on light by mid-May (Fig. 7). green urchins could ultimately cause a decline in
Although some perennial kelp species are capable of macroalgal diversity in shallow, subtidal kelp commu-
reproducing in the winter (e.g. Agarum fimbriatum nities. Furthermore, by concentrating their grazing in
and Laminaria bongardiana; Vadas 1968, and own areas outside S. muticum populations, green urchins
pers. obs.) many annual kelps (e.g. Costaria costata) may facilitate the spread of this invader, which has
must reproduce in the spring and summer. Further- higher recruitment in areas that have been experi-
more, most native red algae (i.e. understory algae) are mentally denuded of algae (Britton-Simmons 2003).
either only present during the spring and summer, or The response of green urchins to the presence of
experience most of their growth during that time (own Sargassum muticum suggests that this generalist her-
pers. obs.). The months during which S. muticum is bivore is presently unlikely to slow the rate of S.
having its strongest effect on light (Fig. 7) is also a crit- muticum’s spread (Figs. 6 & 11). The food preferences
ical period of time for the growth and reproduction of of green urchins could change over time so that S.
many species of native algae. muticum becomes a more preferred food resource.
Previous studies of Sargassum muticum have varied Moreover, natural selection could eventually favor
widely in their conclusions about its effect on native such a shift, especially if S. muticum continues to in-
communities. However, my study is in accordance with crease in abundance and displace palatable native
1 generalization that emerges from a review: studies in kelp species. However, S. muticum’s wiry morpho-
the intertidal zone have found little or no impact of logy and positive buoyancy may make it largely
S. muticum (DeWreede 1983, DeWreede & Vander- inaccessible to green urchins and thereby preclude an
meulen 1988, Viejo 1997, Wilson 2001), but studies in evolutionary shift in feeding preferences.
the subtidal zone indicate relatively strong effects My removal experiments were conducted at sites
(Ambrose & Nelson 1982, Staehr et al. 2000, present where the abundance of Sargassum muticum is at the
study). Considering the vertical distribution of S. upper end of its distribution of densities. If the effect of
muticum, this general trend is not surprising. The S. muticum on light is proportional to its abundance,
lower intertidal is at the upper edge of S. muticum’s then I would expect its effect on native species at any
vertical distribution (DeWreede 1983, Staehr et al. particular site to also be proportional to abundance.
Britton-Simmons: Effects of an introduced alga 77
Nevertheless, a more complex relationship is possible. Cohen BF, McArthur MA, Parry GD (2001) Exotic marine
For example, other studies have shown that the sign of pests in the Port of Melbourne, Victoria. MAFRI Rep 25:
1–96
an interaction between exotic and native species can
DeWreede RE (1983) Sargassum muticum (Fucales, Phaeo-
change as their relative densities change (Reush & phyta): regrowth and interaction with Rhodomela larix
Williams 1998). Thus, it may be necessary to evaluate (Ceramiales, Rhodophyta). Phycologia 22:153–160
the effects of S. muticum at various densities in order DeWreede RD (1996) The impact of seaweed introductions on
to understand more completely how it interacts with biodiversity. Global Biodiv 6(3):2–9
DeWreede RE, Vandermeulen R (1988) Lithothrix aspergillum
native species. At low densities it may have a relatively (Rhodophyta): regrowth and interaction with Sargassum
small effect on light, although even this small effect muticum (Phaeophyta) and Neorhodomela larix (Rhodo-
may be sufficient to alter the competitive outcome of phyta). Phycologia 27:469–476
interspecific interactions among native kelps without Deysher L, Norton R (1982) Dispersal and colonization in Sar-
gassum muticum (Yendo) Fensholt. J Exp Mar Biol Ecol
reducing overall kelp abundance.
56:179–185
This study illustrates that an introduced alga can Duggins DO (1980) Kelp beds and sea otters: an experimental
have important effects on native communities. These approach. Ecology 61(3):447–453
results, in combination with other studies that have Duggins DO, Simenstad CA, Estes JA (1989) Magnification of
also demonstrated strong effects of non-indigenous secondary production by kelp detritus in coastal marine
ecosystems. Science 245:170–173
algae species (Verlaque 1994b, Villele & Verlaque Ebeling AW, Laur DR, Rowley RJ (1985) Severe storm distur-
1995, Walker & Kendrick 1998, Levin et al. 2002) high- bances and reversal of community structure in a southern
light the importance of future research into the spread California kelp forest. Mar Biol 84:287–294
and impact of introduced algae. Because of their Eckman JE, Duggins DO, Sewell AT (1989) Ecology of under-
story kelp environments. I. Effects of kelps on flow and
worldwide distribution and capacity to alter native
particle transport on the bottom. J Exp Mar Biol Ecol 129:
communities, non-indigenous algae are potentially 173–187
important agents of global ecological change. Faul F, Erdfelder E (1992) GPOWER: a priori, post-hoc, and
compromise power analyses for MS-DOS (Computer pro-
gram). Bonn University, Department of Psychology, Bonn
Acknowledgements. None of this work would have been pos- Forrest BM, Brown SN, Taylor MD, Hurd CL, Hay CH (2000)
sible without the many individuals who volunteered their The role of natural dispersal mechanisms in the spread
time to dive with me on this project. I am especially grateful to of Undaria pinnatifida (Laminariales, Phaeophyceae).
J. Kido, E. Iyengar, B. Pister, L. Self, and C. Catton for being Phycologia 39:547–553
my ‘buddies’. For thoughtful discussions which improved the Fredriksen S, Sjotun F, Lein T, Rueness J (1995) Spore disper-
quality of this work I thank T. Wootton, C. Pfister, M. Leibold, sal in Laminaria hyperborea (Laminariales, Phaeophy-
J. Bergelson, G. Dwyer, S. Williams, D. Duggins, T. Klinger, J. ceae). Sarsia 80:47–53
Kido, S. Hall, K. Polivka, P. Geddes, S. Harrell and B. Pister. Gardner WD (1980) Sediment trap dynamics and calibration:
J. Bergelson, C. Pfister, J. Byers, M. Wilson, K. Abbott, T. a laboratory evaluation. J Mar Res 38:41–52
Wootton, C. Peterson, and 2 anonymous reviewers helped me Giver KJ (1999) Effects of the invasive seaweed Sargassum
transform early versions of this manuscript into its present muticum on native marine communities in northern Puget
form. T. Wootton gave guidance and encouragement through- Sound, Washington. MS thesis, Western Washington
out this project. The staff of Friday Harbor Laboratories pro- University, Bellingham
vided logistical support including boats, dive facilities and Hammerstrom K, Dethier MN, Duggins DO (1998) Rapid
laboratory space, and for that I am thankful. This work was phlorotannin induction and relaxation in five Washington
funded by grants to K.B.S. from The Packard Foundation and kelps. Mar Ecol Prog Ser 165:293–305
The University of Chicago’s Hinds Fund, and grants to T. Hayduk LA (1987) Structural equation modeling with LISREL.
Wootton from The Marine Ecosystem Health Program and Johns Hopkins University Press, Baltimore
The Mellon Foundation. K.B.S. was supported by a Depart- Jones LC (1971) Studies on selected small herbivorous inver-
ment of Education GAANN fellowship during the course of tebrates inhabiting Macrocystis canopies and holdfasts
this study. in southern California kelp beds. Nova Hedwigia 32:
343–367
Juerg B (1996) Towards a new generation of sediment traps
LITERATURE CITED and a better measurement/understanding of settling parti-
cle flux in lakes and oceans: a hydrodynamical protocol.
Ambrose RF, Nelson BV (1982) Inhibition of giant kelp re- Aquat Sci 58:283–296
cruitment by an introduced brown alga. Bot Mar 25:265–267 Larson BR, Vadas RL, Keser M (1980) Feeding and nutritional
Bernstein BB, Jung N (1979) Selective pressures and coevolu- ecology of the sea urchin Stronglyocentrotus droebachi-
tion in a kelp canopy community in southern California. ensis in Maine, USA. Mar Biol 59:49–62
Ecol Monogr 49:335–355 Levin PS, Coyer JA, Petrik R, Good TP (2002) Community-
Britton-Simmons KH (2003) Establishment, spread, and im- wide effects of non-indigenous species on temperate
pact of the introduced Japanese seaweed, Sargassum rocky reefs. Ecology 83:3182–3193
muticum, in the San Juan Islands, WA. PhD thesis, The Lobban CS, Harrison PJ (1994) Seaweed ecology and physio-
University of Chicago, Chicago logy. Cambridge University Press, New York
Burnham KP, Anderson DR (1998) Model selection and infer- Lüning K (1981) Light. In: Lobban CS, Wynne MJ (eds) The
ence: a practical information–theoretic approach. Springer biology of seaweeds. University of California Press, Berke-
Verlag, New York ley, p 340–342
78 Mar Ecol Prog Ser 277: 61–78, 2004
Norton TA (1977) The growth and development of Sargassum Study (JGOFS) core measurements. Intergovernmental
muticum (Yendo) Fensholt. J Exp Mar Biol Ecol 26:41–53 Oceanographic Commission (IOC), manual and guides No
Peterson CH, Renaud PE (1989) Analysis of feeding prefer- 29. Paris, France
ence experiments. Oecologia 80:82–86 Vadas RL (1968) The ecology of Agarum and the kelp bed
Pfister CA, Hay ME (1988) Associational plant refuges: con- community. PhD thesis, University of Washington, Seattle
vergent patterns in marine and terrestrial communities Vadas RL (1977) Preferential feeding: an optimization
result from differing mechanisms. Oecologia 77:118–129 strategy in sea urchins. Ecol Monogr 47:337–371
Pfister CA, Van Alstyne KL (2003) An experimental as- Verlaque M (1994a) Checklist of introduced plants in the
sessment of the effects of nutrient enhancement on the Mediterranean: origins and impact on the environment
intertidal kelp Hedophyllum sessile (Laminariales, Phaeo- and human activities. Oceanol Acta 17:1–23
phyceae). J Phycol 39:285–290 Verlaque M (1994b) Mediterranean algal communities are
Reed DC, Laur DR, Ebeling AW (1988) Variation in algal dis- changing in the face of the invasive alga Caulerpa taxi-
persal and recruitment: the importance of episodic events. folia (Vahl) C. Agardh. Oceanol Acta 17:659–672
Ecol Monogr 58:321–335 Viejo RM (1997) The effects of colonization by Sargassum
Reush TBH, Williams SL (1998) Variable responses of native muticum on tidepool macroalgal assemblages. J Mar Biol
eelgrass Zostera marina to a non-indigenous bivalve Assoc UK 77:325–340
Musculista senhousia. Oecologia 113:428–441 Villele X de, Verlaque M (1995) Changes and degradation
Rueness J (1989) Sargassum muticum and other introduced in a Posidonia oceanica bed invaded by the introduced
Japanese macroalgae: biological pollution of European tropical alga Caulerpa taxifolia in the north western
coasts. Mar Pollut Bull 20(4):173–176 Mediterranean. Bot Mar 38:1–9
Scagel RF (1956) Introduction of a Japanese alga, Sargassum Walker DI, Kendrick GA (1998) Threats to macroalgal
muticum, into the Northeast Pacific. Fish Res Pap St Wash diversity: marine habitat destruction and fragmentation,
1:1–10 pollution and introduced species. Bot Mar 41:105–112
Setzer B, Link C (1971) The wanderings of Sargassum muti- Wilson M (2001) Is Sargassum muticum a benign invader of
cum and other relations. Stomatopod 2:5–6 tidepools on the Pacific coast of North America? MS thesis,
Shipley B (2000) Cause and correlation in biology (a user’s California State University Monterey Bay, Seaside
guide to path analysis, structural equations and causal Wootton JT (1991) Direct and indirect effects of nutrient on
inference). Cambridge University Press, Cambridge intertidal community structure: variable consequences of
Sokal RR, Rohlf FJ (1995) Biometry: the principle and practice seabird guano. J Exp Mar Biol Ecol 151:139–153
of statistics in biological research, 3rd edn. WH Freeman, Wootton JT (1994) Predicting direct and indirect effects: an
New York integrated approach using experiments and path analysis.
Staehr PA, Pedersen MF, Thomsen MS, Wernberg T, Krause- Ecology 75:151–165
Jensen D (2000) The invasion of Sargassum muticum in Lim- Yendo K (1907) The Fucaceae of Japan. J Coll Sci Imp Univ
fjorden (Denmark) and its possible impact on the indigenous Tokyo 21(12):1–174
macroalgal community. Mar Ecol Prog Ser 207:79–88 Zar JH (1999) Biostatistical analysis, 4th edn. Prentice Hall,
UNESCO (1994) Protocols for the Joint Global Ocean Flux Upper Saddle River, NJ
Editorial responsibility: Charles Peterson (Contributing Editor), Submitted: February 3, 2003; Accepted: February 3, 2004
Morehead City, North Carolina, USA Proofs received from author(s): August 9, 2004
Vol. 277: 61–78, 2004 Published August 16
Mar Ecol Prog Ser
Direct and indirect effects of the introduced alga
Sargassum muticum on benthic, subtidal
communities of Washington State, USA
Kevin H. Britton-Simmons1,*
Department of Ecology and Evolution, The University of Chicago, 1101 East 57th Street, Chicago, Illinois 60637, USA
1
Present address: University of Washington, Friday Harbor Laboratories, 620 University Road, Friday Harbor,
Washington 98250, USA
ABSTRACT: Introduced algae have become a prominent component of the marine flora in many
regions worldwide. In the NE Pacific, the introduced Japanese alga Sargassum muticum (Yendo)
Fensholt is common and abundant in shallow, subtidal, rocky habitats, but its effects on subtidal, ben-
thic communities in this region have not previously been studied. I measured the response of native
species to experimental manipulation of S. muticum in field experiments in the San Juan Islands of
Washington State. Native canopy (brown) and understory (red) algae were more abundant in plots
from which S. muticum had been removed, and the native kelp Laminaria bongardiana (the most
abundant species of brown alga in the absence of S. muticum) grew more than twice as fast in plots
where S. muticum was absent. The negative effects of S. muticum on native algae appear to be a re-
sult of shading, rather than changes in water flow, sedimentation, or nutrient availability. S. muticum
also had a strongly negative indirect effect on the native sea urchin Stronglyocentrotus droebachi-
ensis by reducing abundances of the native kelp species on which it prefers to feed. My results
indicate that S. muticum has a substantial impact on native communities in this region, including
effects at multiple trophic levels. Because of their worldwide distribution and capacity to alter native
communities, non-indigenous algae are potentially important agents of global ecological change.
KEY WORDS: Sargassum muticum · Introduced algae · Competition · Indirect effects · Stronglylo-
centrotus droebachiensis · Shading
Resale or republication not permitted without written consent of the publisher
INTRODUCTION (Yendo) Fensholt, on native, subtidal kelp communities
in Washington State, USA.
Introduced algae have become a prominent compo- Sargassum muticum is native to SE Asia (Yendo
nent of the marine flora in many regions worldwide 1907), but its present distribution as an invasive spe-
(Rueness 1989, Verlaque 1994a, DeWreede 1996, cies is widespread, including Europe, the Mediter-
Cohen et al. 2001). Despite their widespread distribu- ranean Sea and the west coast of North America. In the
tion, few studies have rigorously documented the United States, S. muticum was introduced to Washing-
effects of exotic algae and as a consequence many of ton State in the early 20th century, probably with ship-
their possible impacts remain speculative. Neverthe- ments of Japanese oysters that were imported for
less, available evidence suggests that introduced algae aquaculture beginning in 1902 (Scagel 1956). Follow-
do have the potential to substantially alter native com- ing its establishment in this region, it subsequently
munities (Verlaque 1994b, Villele & Verlaque 1995, invaded over 3000 km of coastline on the west coast of
Walker & Kendrick 1998, Levin et al. 2002). In the North America (Scagel 1956, Setzer & Link 1971). S.
present study, I experimentally evaluate the impact muticum has now become established in low intertidal
of the introduced Japanese alga Sargassum muticum and shallow subtidal habitats throughout Puget Sound
*Email: aquaman@kevinbs.net © Inter-Research 2004 · www.int-res.com
62 Mar Ecol Prog Ser 277: 61–78, 2004
and the San Juan Islands (own pers. obs.), where it using SCUBA. Removal experiments and associated
occurs in densities as high as 126 plants m–2 (own kelp growth experiments were carried out at 2 sites
unpubl. data). within the San Juan Islands Marine Preserve network
In areas where Sargassum muticum is abundant in adjacent to San Juan and Shaw Islands, known locally
the San Juan Islands, it forms a dense covering that as Colin’s Cove (48° 33.16’ N, 122° 58.79’ W) and Point
towers up to 2 m above all but 1 native algal species, George (48° 32.97’ N, 123° 00.33’ W), respectively. These
Nereocystis luetkeana. These dense stands of S. reserves were established in 1990 and are closed to
muticum may reduce light, dampen flow, increase harvesting with very limited exceptions (e.g. salmon).
sedimentation and reduce ambient nutrient concentra- Laboratory experiments were completed at FHL.
tions available for native kelp species (order Laminari- The 2 field sites vary in their physical characteristics.
ales). Because kelps are an important source of carbon The habitat at Colin’s Cove is composed primarily of
in coastal food webs (Duggins et al. 1989) and the algal relatively flat, rocky reefs with some small boulders.
communities they are associated with provide habitat In contrast, the substratum at Point George is more het-
and food for a wide variety of marine animals (Jones erogeneous, with rocky reef overlaid by a layer of small
1971, Bernstein & Jung 1979, Ebeling et al. 1985), any to medium-sized boulders. Point George is also sub-
negative effects of S. muticum on these communities jected to more intense tidal currents than Colin’s Cove.
may have broader consequences for this ecosystem. Despite these differences in abiotic habitat, both
Previous studies have varied substantially in their field sites are similar in the biological community they
conclusions about how strong an effect Sargassum support. The native kelp communities that dominate
muticum has on native communities. While intertidal shallow, subtidal habitats in this region are species-
studies in British Columbia (DeWreede 1983) and Spain rich and structurally complex. The upper layer of the
(Viejo 1997) suggest that it competes with native algae, algal community is composed of a canopy of large,
a study in California tidepools found no effect (Wilson brown algae in the orders Laminariales and Desmares-
2001). Subtidal studies have found evidence for inhibi- tiales, referred to herein as canopy algae. The middle
tion of giant kelp recruitment in California (Ambrose & layer consists of an assemblage of fleshy red algae
Nelson 1982) and for competition with native algae in from a variety of orders, referred to herein as under-
Denmark (Staehr et al. 2000). Interestingly, in the San story algae. The bottom layer is formed by encrusting
Juan Islands, S. muticum supports a more abundant coralline algae and filamentous, turf-forming algae.
and species-rich epibiont community than the native Herbivores in this shallow subtidal system include the
kelp Laminaria saccharina (Giver 1999), which it is green sea urchin Strongylocentrotus droebachiensis
thought to displace. Previous studies have focused and a variety of grazing molluscs including chitons,
almost exclusively on competitive interactions of S. limpets and snails.
muticum with other algae and have not investigated Background information on Sargassum muticum.
its potential influence on other trophic levels (but see S. muticum is a brown alga in the order Fucales. Al-
Wilson 2001). though the plant is a perennial, its lateral branches are
The goal of this study was to investigate the effect of only present for a portion of the year. The phenology of
Sargassum muticum on native kelp communities in the S. muticum’s life cycle varies regionally. In the San
San Juan Islands of Washington State by measuring Juan Islands, its numerous lateral branches begin
the response of native algae and invertebrates to ex- growing in early March, reach their maximum height
perimental manipulation of the presence of S. muticum. in June, and begin to senesce in mid- to late-August.
I was also interested in the mechanism(s) underlying Only the short (5 to 10 cm) basal holdfast portion over-
the effects of S. muticum. By measuring light, nutri- winters in a quiescent state. In the San Juan Islands,
ents, water flow and sedimentation in experimental each holdfast produces as many as 18 laterals in the
plots in the field, I tested several hypotheses about early spring, each of which can grow as tall as 3 m.
how S. muticum changed the abiotic environment. Small pneumatocysts along the primary axis of each
I predicted that S. muticum should decrease light, lateral make it positively buoyant and cause the later-
water flow and nutrients, but that it should increase als to extend vertically into the water column. S.
sedimentation. muticum has a simple life cycle. Reproductive struc-
tures called receptacles are borne along secondary
branches and contain both oogonia and antheridia.
MATERIALS AND METHODS After eggs are released from the oogonia, they adhere
to the external surface of the receptacle, where they
Study sites. This research was based at Friday Har- are fertilized. Fertilized embryos remain attached to
bor Laboratories (hereafter FHL) on San Juan Island, the receptacle until they develop tiny, adhesive rhi-
Washington State, USA. Field work was accomplished zoids, at which point they detach and recruit to the
Britton-Simmons: Effects of an introduced alga 63
substratum in close proximity to the parent plant (Dey- were taken simultaneously at surface and at depth
sher & Norton 1982). using a Licor LI-1000 data logger coupled to 2 quantum
Sargassum muticum removal experiments. I de- sensors. Surface measurements were taken from a
signed experiments to test the hypothesis that the small boat, and measurements at depth were taken
presence of S. muticum was influencing the structure 30 cm above the substratum in the center of each con-
of the native community. During June 1999, 2 removal trol and removal treatment plot. Because Nereocystis
experiments were simultaneously initiated at Point luetkeana, the only native algae species that is taller
George and Colin’s Cove. At each site, ten 50 × 50 cm than S. muticum, was not present in any of my plots at
plots, at a depth of –2 m mean lower low water (MLLW) the time and because the remaining kelp species have
and containing high densities of S. muticum, were per- demersal blades that typically extend less than 30 cm
manently marked using metal-stamped stainless steel above the substratum, measuring light at 30 cm al-
washers and marine epoxy (Z-Spar splash zone com- lowed me to isolate the effect of S. muticum. Light was
pound™). At each site, the experimental plots were measured at 1 s intervals for 1 min in each plot. Plots
arranged in a line parallel to shore; the average dis- were sampled systematically, beginning at one end of
tance between plots was 2.5 m. Plots were then the site and proceeding to the other end in quick suc-
randomly assigned to control or removal treatments. cession, in order to minimize the time elapsed between
Control plots were not altered. Removal plots had all S. samples. The same sampling protocol was used on all
muticum plants eliminated, both from within the plots sampling dates. Light measurements were always
and from a 50 cm buffer surrounding the plots. Each S. taken at both sites on the same day and the sampling
muticum plant was removed by carefully prying its period straddled noon. Whenever possible, samples
small (1 to 2 cm in diameter), discoid holdfast off the were taken on a cloudless day. The average of the 60
rock substrate with a small dive knife. The buffer zone instantaneous measurements at the surface and at
was created in order to reduce shading or other effects depth was used to calculate percent transmittance
of adjacent S. muticum plants. The initial S. muticum (light at depth/light at surface), which is the quantity
removal began on June 30, 1999 and was completed at reported and used in statistical analyses.
both sites by July 5, 1999. The removal treatment was Water flow measurements. Because Sargassum muti-
maintained over the course of the experiment by cum’s fronds form dense stands extending several
removing any new S. muticum recruits from removal meters into the water column, they have the potential
plots and adjacent buffer zones several times per year. to dampen water flow. I tested the hypothesis that S.
All experimental plots were censused once prior to muticum reduced water flow, using alabaster dissolu-
imposing the removal treatment and then 2 to 6 times tion blocks following the methods of Eckman et al.
per year thereafter. At the time of each census, I (1989). Small blocks of alabaster (di-hydrated calcium
recorded the identity and number of all macroalgae in sulfate) were cut from a larger block and sanded to
each experimental plot. In addition, I estimated the uniform dimensions (average of 46 × 42 × 10 mm).
percentage of primary space on the rock substrate that Blocks (1 per plot) were anchored to threaded steel
was covered by crustose algae, filamentous algae, rods at a height of 30 cm above the substratum, which
fleshy red algae, fleshy brown algae, Sargassum muti- is above all kelp species present (Nereocystis luet-
cum and bare rock. The percent cover estimates were keana was not in any experimental plot at this time)
carried out on a 25 × 25 cm subplot. Benthic inverte- but below S. muticum, thereby isolating the effect of S.
brates were counted in the baseline and first post- muticum on flow. The threaded steel rods were
removal census and then twice per year thereafter. screwed into tubes of hexagonal nuts (8 nuts stacked
Time and weather constraints prevented complete on top of one another so that the threaded holes lined
censuses of all taxa of interest on some sampling dates. up) that had been attached to the rock substratum with
Light measurements. I measured light levels in all marine epoxy (Z-Spar splash zone compound™). In
removal experiment plots on 4 dates in 2000 (March order to prevent abrasion by seaweed, blocks were
21, May 12, June 9 and August 22). Sargassum muti- enclosed in hardware cloth cages (mesh squares =
cum’s fronds are present for only part of the year (typi- 12 mm on a side), which were also attached to the
cally March to August), and therefore any effect they threaded steel rods by stainless steel hardware. Before
have on light should be limited to this time period. and after deployment, the blocks were dried to a con-
Because my sampling dates spanned the months of stant weight at 60°C and weighed to the nearest 1 mg.
March to August, and included dates when the fronds The blocks were deployed on August 27, 1999 and
were absent and present, I was able to use these data collected on August 29, 1999 for a total deployment
to test the hypothesis that S. muticum reduced ambient time of 42.5 h.
light levels in plots where it was present. Measurements Sedimentation measurements. Any effect that Sar-
of photosynthetically active radiation (400 to 700 nm) gassum muticum has on water flow is likely to influ-
64 Mar Ecol Prog Ser 277: 61–78, 2004
ence the flux of particulate matter from the water col- length of time necessary to complete the collection and
umn to the substrate. I measured sedimentation rates processing of samples for both sites exceeded the slack
in each removal experiment plot during July 2001 in current period, the 2 sites were sampled on successive
order to test the hypothesis that S. muticum increases days. Point George water samples were taken on
sedimentation rates. Many native algae (including August 5, 1999 and Colin’s Cove water samples were
kelps) are recruiting during this time and increased taken on August 6, 1999.
sedimentation rates could have important effects on Nalgene bottles (1 l) were acid-washed in a 10% HCl
community development by altering patterns of algal solution and dried. Each plot was sampled by placing a
settlement or survivorship. Although nearly all sedi- bottle in the center of the plot, removing the lid and
ment traps have drawbacks and no trap design has allowing it to fill. Capped bottles were brought to the
been developed specifically for shallow, subtidal habi- surface and immediately placed in an ice-filled cooler
tats, I chose a design that was recommended for use in until being processed in the boat. A subsample of the
strong currents like those characteristic of my system water in each bottle was extracted with a syringe and
(Gardner 1980, Jeurg 1996). Specifically, a cylinder filtered through a glass microfiber filter into a smaller
with a height-to-diameter ratio of at least 4.5 and a bottle which was placed in an ice-filled cooler. The
minimum diameter of 3.9 cm is recommended for remaining water in the 1 l bottles was then discarded.
strong currents to prevent resuspension of sediments Larger bottles were used for collecting the water sam-
that have already been trapped (Gardner 1980). I con- ples underwater in order to facilitate the extraction of
structed cylindrical traps from PVC that were 32 cm in the water to be analyzed once they were brought to the
height and 6 cm in diameter. The traps were attached surface. Samples were immediately transported to the
to the rock substrate just adjacent to each plot by laboratory where they were frozen (and stored) at
screwing a bolt that protruded from the bottom of each –70°C until being shipped to the University of Wash-
trap into a tube of hexagonal nuts that had been ington’s School of Oceanography, where they were
attached to the rock with marine epoxy (Z-Spar splash processed using standard protocols (UNESCO 1994)
zone compound™). Traps were deployed for a single for determination of ammonium, nitrate, nitrite, phos-
10 d sampling period, which spanned portions of both phorus, and silica concentrations.
a spring (when sedimentation should be lowest) and a Kelp growth experiment. I used a kelp growth ex-
neap (when sedimentation should be the highest) tide periment at 2 sites in July 2000 to test the hypothesis
series, and therefore should give an indication of the that Sargassum muticum negatively affects kelp
average effect of S. muticum on sedimentation. Prior growth. I used the native kelp Laminaria bongardiana
to removal, each trap was capped with a PVC lid to (formerly L. groenlandica), which was the most com-
prevent loss of contents during transport to FHL for mon kelp at the 2 removal experiment sites, as the
processing. All samples were immediately frozen at focal species. L. bongardiana plants that were similar
–70°C until they could be processed. in size (8.1 ± 2.6 g = mean ± 1 SD) were collected intact
After removing large animals (e.g. hermit crabs) and from the field and brought to the laboratory where
pieces of algae with a coarse (4 mm) mesh, I used a vac- they were kept in flow-through seawater tanks until
uum pump to filter the sediment onto glass microfiber their return to the field less than 24 h later. Transplant-
filters that had been baked in a muffle furnace at 500°C ing was accomplished as follows. A cinder block
(to remove any organic content) for 6 h. The sediment- (39.5 × 14.5 × 19 cm) with eye bolts inserted at each
laden filters were then dried to a constant weight at end was placed in the area just adjacent to each exper-
60°C , weighed (to the nearest mg), baked for 6 h in a imental plot from which S. muticum had been removed
300°C muffle furnace (to burn off organic content) and (removal) or left intact (control). A single length of
weighed again. This protocol allowed me to separate braided nylon rope, containing 3 L. bongardiana whose
out organic and inorganic components of the sediment. holdfasts had been woven into it, was then attached to
Nutrient measurements. Dense stands of Sargassum each cinder block by tying the ends of the rope to the
muticum have the potential to reduce ambient levels of eye bolts. The cinder block served to elevate trans-
critical nutrients to levels that might negatively affect planted kelps above the substratum slightly, reducing
the growth of native algae. In August 1999, I took a shading by adjacent kelps and thereby better isolating
single water sample from the center of each experi- the effect of S. muticum. Each plant was individually
mental plot to test the hypothesis that S. muticum marked with a piece of numbered flagging tape tied
reduced ambient nutrient concentrations in plots gently around the stipe. Plants were spaced at 10 cm
where it was present. All samples were taken at slack intervals on the ropes and were randomly assigned to
current when water movement is at a minimum, and sites and treatments.
any dilution effect of S. muticum on the local nutrient I used 3 metrics to quantify growth during the 28 d
pool would be most likely to be detected. Because the experiment: (1) change in mass was determined by
Britton-Simmons: Effects of an introduced alga 65
weighing kelps to the nearest 0.1 g before and after tained a single specimen of each of the 3 algal species
transplanting them to the field; prior to weighing, each that had been collected from the field on the day the
plant was held aloft and shaken vigorously and then experiment was set up. Plants were attached to the
patted dry with paper towels to remove excess water. bottom in random order by inserting holdfasts into
(2) Change in blade area was measured by tracing the small incisions that were positioned 10 cm apart in a
blade area of each plant onto butcher paper before and RubbermaidTM bathmat adhering tightly to the bottom.
after transplanting them to the field; blade traces were In addition to being attached at the holdfast, each Sar-
converted to area by cutting them out, weighing them gassum muticum plant was bent over and cable-tied to
and converting those weights to area using a conver- the bathmat at a point approximately 30 cm away from
sion factor determined by weighing pieces of butcher the holdfast. This ensured that all species of algae
paper of known area. (3) Linear blade growth was were equally accessible and experimental results
measured by using a cork borer to remove a circular reflected actual food preferences, unconfounded by
piece of blade tissue 2 cm in diameter at the midpoint the urchin’s inability to access and feed on S. muticum.
of the blade, 4 cm from the meristem region and mea- Urchins were collected from the field on the morning
suring the distance the hole moved distally during the the experiment was set up (May 1, 2001) and kept in a
experiment. flow-through seawater tank until being placed in
Urchin food-preference experiment. Data collected experimental tanks. The 2 sides of each tank were ran-
in the removal experiment showed that green urchins domly assigned to either urchin or control treatments
avoided foraging in control plots, where Sargassum and 6 urchins (6 to 10 cm in diameter) were placed in
muticum was present. I compared the feeding of Stron- each half-tank assigned to the urchin treatment.
gylocentrotus droebachiensis on S. muticum relative to Urchins were assigned to tanks such that each tank
the native kelps Agarum fimbriatum and Laminaria had a range of sizes and total urchin mass per tank was
bongardiana, to test the hypothesis that the urchin’s approximately equal. The density of urchins used in
distaste for S. muticum caused it to avoid S. muticum- this experiment (3.5 urchins m–2) is within the range of
dominated plots. These 2 native kelps were the numer- natural densities I have recorded in the field. Con-
ically dominant canopy species in removal plots at both sumption of each algae species over the 3 d experi-
experimental sites, but they differ considerably in ment was calculated by pairing control and urchin
terms of palatability to urchins. Whereas L. bongar- treatments from each tank and applying the formula
diana is a preferred food of green urchins in this [Ti (C f /Ci )] – Tf , where Ti and Tf are the initial and final
region, A. fimbriatum is consistently avoided in food algae masses in treatment tanks and C i and C f are the
preference experiments (Vadas 1977). This difference initial and final algae masses in the randomly paired
in preference probably reflects the 5 to 10 times higher control tank (Peterson & Renaud 1989). Prior to weigh-
levels of polyphenolics (chemicals known to deter ing algae at the beginning and end of the experiment,
herbivory) in A. fimbriatum than in L. bongardiana each individual was held aloft and shaken gently to
(Hammerstrom et al. 1998). By scaling the green remove excess water from its surface, and then spun in
urchin’s feeding rate on S. muticum against 2 native a salad spinner for 20 s.
kelp species whose relative palatability was already Urchin predation experiment. Increased predation
known, I hoped to better characterize its response to by sea stars in Sargassum muticum-dominated areas
S. muticum as a potential food source. is an alternative hypothesis that could also explain
I had 2 hypotheses about why urchins may not feed the near-absence of green urchins in the control plots
on Sargassum muticum in the field: (1) they find its in the removal experiment at Point George. Algal
tissue unpalatable, or (2) its positively buoyant fronds canopies dominated by S. muticum might enable the
make it inaccessible to urchins. I designed this feeding sea star Pycnopodia helianthoides (a locally common
experiment to test Hypothesis 1 and controlled the predator on green urchins) to prey more effectively on
accessibility of the 3 algae species in this experiment green urchins. Green urchins typically cover their
to ensure that all 3 species were equally available for aboral surface with kelp or other algae while feeding,
urchins to feed on. but S. muticum’s positively buoyant fronds and wiry
The experiment was carried out in 7 circular outdoor morphology make it unsuitable for this purpose.
aquaria (105 cm in diameter, 45 cm tall). All aquaria Therefore, green urchins might be more vulnerable to
had a constant supply of unfiltered seawater from the predation by P. helianthoides in areas where S.
FHL seawater system throughout the experiment. muticum is the dominant alga because they may be
Each aquarium was divided in half with plastic Vexar unable to effectively hide themselves.
mesh (oval holes measuring 2.6 × 1.9 cm); this allowed I used a laboratory experiment in seawater tanks at
water to flow freely throughout the tank while isolat- FHL in May 2003 to test whether predation on urchins
ing urchins on 1 side. Each side of each aquarium con- by sea stars differed in a Sargassum muticum mono-
66 Mar Ecol Prog Ser 277: 61–78, 2004
culture versus a mixed kelp canopy. I divided 7 out- the number of urchins that were visible from above in
door aquaria (105 cm in diameter, 45 cm tall) in half each tank. These individuals should be more vulnera-
using plastic Vexar mesh (oval holes measuring 2.6 × ble to predation by visual predators (e.g. crabs) and
1.9 cm). This setup allowed water to flow freely tactile predators (e.g. sea stars). At the end of the
throughout the tank while preventing urchins from experiment all sea stars were removed from the tanks
moving between the 2 sides of the tanks. All aquaria and the remaining urchins were counted.
had a constant supply of unfiltered seawater from the Statistical analyses. Most statistical analyses were
FHL seawater system throughout the experiment. The carried out using SYSTAT version 9.0 (SPSS). Because
halves of each tank were randomly assigned to 1 of 2 many native species were not present in all removal
experimental treatments: (1) S. muticum monoculture, experiment plots, and because I expected functionally
or (2) mixed canopy of Laminaria bongardiana and similar species to respond similarly to the presence of
Agarum fimbriatum; so that each tank contained Sargassum muticum, I grouped native species into
1 replicate of each treatment. The densities of S. functional groups in addition to carrying out single-
muticum (29 m–2) and kelps (25 m–2) used in the exper- species analyses for the most common native species.
iment were determined from the average densities of Macroalgae were separated into 2 groups: (1) under-
these species in the field. The algae used in the exper- story — small, red species (e.g. Rhodoptilum plumo-
iment were collected on the day preceding and the day sum, Odonthalia spp., Plocamium cartilagineum,
of beginning the experiment, and were attached to the Gigartina papillata, Laurencia spectabilis, Opuntiella
bottom of the tanks by inserting their holdfasts into californica and Callophyllis spp.), and (2) canopy —
small incisions in Rubbermaid™ bathmats adhered large, brown species (Laminaria bongardiana, L. com-
tightly to the bottom of the tanks. I used a mixture of plenata, Agarum cribrosum, A. fimbriatum, Costaria
sizes for each species in order to mimic the size costata, Nereocystis luetkeana, Alaria marginata, Des-
structure typical of algae populations in the field (S. marestia ligulata and D. viridis). Native invertebrates
muticum: 30 to 60 cm tall; A. fimbriatum and L. bon- were separated into 2 main groups: (1) herbivorous
gardiana: 25 to 60 cm tall). The order of species (in the molluscs (Mopalia spp., Cryptochiton stelleri, Tonicella
kelp treatment) and their spatial arrangement on the lineata, Acmea mitra, Diodora aspera, Margarites
mats was haphazard. pupillus and Lacuna vincta), and (2) detritivores (Bit-
Once the algal treatments were set up, I placed 8 tium eschrichtii, Pandalus spp., and Pagurus spp.). The
Strongylocentrotus droebachiensis on each side of data for Strongylocentrotus droebachiensis were ana-
each tank. All urchins were dropped into the tanks lyzed independently from those for other herbivores.
around the perimeter and were allowed to acclimate Because the single-species analyses did not yield any
for 18 h before the sea stars were added. The urchins new insights (with the exception of S. droebachiensis),
were collected over a 1 mo long period preceding the the data presented in this paper are generally for the
experiment and were kept in flow-through seawater functional groupings.
tanks and fed a mixed algal diet prior to their use in Biological data from the removal experiments were
the experiment. I divided the urchins into 2 size classes analyzed using repeated-measures ANOVA, blocking
(50 to 65 mm, and 65 to 80 mm) and randomly selected by site. Site (2 levels, Point George and Colin’s Cove) and
4 urchins from each for each experimental unit. The Treatment (2 levels, control and removal) were both
sea stars Pycnopodia helianthoides (40 to 60 cm in treated as fixed factors. For each response variable, I first
diameter) used in the experiment were collected on the performed a 2-way ANOVA on the pre-removal census
day preceding and the day of beginning the experi- data to test the hypothesis that control and removal
ment, and were kept in flow-through seawater tanks plots differed prior to imposing the removal treatment. A
without food until being placed in the experimental pre-removal difference was detected only for understory
tanks. In order to ensure that the variation in sea star density, and therefore I included the pre-removal data as
size was evenly distributed across the treatments, I a covariate in a repeated-measures ANCOVA analyses
paired sea stars of equal size and randomly assigned for that variable. The assumption of normality was tested
them to opposite sides of the same tank until each tank using a Kolmogorov-Smirnov test (α = 0.05). The as-
had 1 sea star per side. The tanks were lightly shaded sumption of equality of variances was tested using
using a double layer of thick, cotton fishing net (4 cm2 an F max test (α = 0.05; Sokal & Rohlf 1995). Data were
mesh size) in order to mimic the shallow subtidal transformed as necessary using square-root, arcsine and
light conditions where these species normally interact. natural log functions to conform to the assumptions of
Additional light reduction was achieved by wrapping ANOVA. I also performed power analyses using GPower
the south-facing half of each tank in a layer of black (Faul & Erdfelder 1992) on each response variable that
plastic. This experiment was allowed to run for a total showed no response to the Sargassum muticum removal
of 8 d. On Days 3 and 8 of the experiment I recorded treatment.
Britton-Simmons: Effects of an introduced alga 67
I performed an additional analysis on the ratio of the I used the SEM module in Statistica for Windows
number of Laminaria bongardiana to the number of (Release 6.0) to estimate unstandardized structural
Agarum fimbriatum in removal experiment plots. For coefficients using a maximum likelihood algorithm.
this analysis I only used data from the last 5 censuses in The statistical significance of structural equation
order to exclude the transitory dynamics which ap- coefficients was determined by the software, using
peared to occur in removal plots following the removal multiple regression for each set of dependent and
of Sargassum muticum. Unfortunately, the large num- independent variables. Alternative models were con-
ber of zeros in the data set (primarily in the control structed using information from my experiments, the
plots, where there were few kelps) precluded both the published literature, and my own knowledge of the
calculation of ratios and the use of ANOVA. Instead, I natural history of the system. For the purposes of SEM
took the average abundance of each of the kelp spe- analyses, I excluded canopy algae that are known to
cies across all replicates for each treatment on each be unpalatable to Strongylocentrotus droebachiensis
sampling date and used those averages to calculate the (Vadas 1968) because a preliminary analysis of the
ratio of the 2 species for each treatment on each sam- data from Point George indicated that urchins
pling date. Because these data points are not indepen- responded strongly to those species that are known to
dent (they are repeated measures), I used the non- be preferred food items based on laboratory feeding
parametric Scheirer-Ray-Hare test (extension of the trials (Vadas 1968) and weakly to the remaining,
Kruskal-Wallis; Sokal & Rohlf 1995) to determine unpalatable species. Palatable canopy species typi-
whether the ratio of L. bongardiana to A. fimbriatum cally contain low levels of polyphenolics, and this
differed between treatments. category included Alaria marginata, Laminaria bon-
In addition to the ANOVA analyses, I analyzed the gardiana, L. saccharina, L. complenata and Nereocys-
abundance data for a subset of the taxa from my tis luetkeana. In contrast, unpalatable canopy species
removal experiment at Point George using structural typically have high levels of polyphenolic compounds
equation modeling (SEM), a form of multiple regres- or other chemical defenses against herbivory, and
sion related to path analysis (Hayduk 1987, Shipley these species included Agarum fimbriatum, A. cribro-
2000), to test whether the effect of Sargassum muticum sum, Desmarestia viridis and D. ligulata.
on Strongylocentrotus droebachiensis was a direct or I tested 2 alternative models to determine whether
indirect effect. SEM is a type of statistical analysis in urchins were responding to the presence of Sargassum
which causal relationships between variables (species muticum or the absence of native algae. The presence
in this case) are hypothesized in the form of an inter- of unpalatable S. muticum might directly affect urchins
action web and tested using a system of linear equa- if their foraging behavior differs in areas with and
tions (Hayduk 1987). Structural equation coefficients without S. muticum. The presence of unpalatable
are calculated using a maximum likelihood algorithm algae species has been shown to alter the feeding
and the significance of each coefficient is tested using behavior of herbivores on palatable algae when both
multiple regression. Structural equation coefficients types of algae are growing together in a phenomenon
indicate how a change in the predictor variable would called an ‘associational plant refuge’ (Pfister & Hay
change the target variable, holding all other variables 1988). For example, an unpalatable congener of S.
constant. The net effect of an indirect pathway be- muticum altered the feeding behavior of the urchin
tween 2 species that involves multiple links can be cal- Arbacia punctulata on palatable red algae when both
culated by multiplying the coefficients for the relevant types of algae were present (Pfister & Hay 1988). Alter-
links (Wootton 1994). A predicted correlation matrix natively, S. muticum could indirectly affect green
between the variables is calculated based on the spec- urchins by competing with their preferred prey spe-
ified model and compared to the actual correlation cies. The first model contained a direct pathway
matrix, using a χ2 distribution, to ask whether the two between S. muticum and urchins, in addition to an
differ significantly. A non-significant χ2 statistic indi- indirect pathway (via palatable canopy). The second
cates that the causal relationships specified in the model lacked the direct pathway in order to test
model cannot be rejected as a good caricature of the whether removing that link altered the fit of the model
species interactions in nature. The relative fit of alter- to the data. Because I experimentally manipulated the
native models can be compared by comparing their χ2 presence of S. muticum, I included treatment as a vari-
statistics (when models are nested) as well as their able in the models. I used a combination of 2 statistics
Akaike information criterion (AIC) values. AIC is an to assess the fit of the models to the data: the χ2 test
information-theoretic criterion for model selection that statistic and the AIC.
takes into account both the goodness of fit of the model Strongylocentrotus droebachiensis abundance data
as well as the complexity of the model required to from Point George (green urchins were not present at
achieve that fit (Burnham & Anderson 1998). Colin’s Cove) were analyzed separately because this
68 Mar Ecol Prog Ser 277: 61–78, 2004
species showed variability in abundance independent height of the Sargassum muticum fronds in control
of other taxa. On many sampling dates there were no plots varied over the course of my light measurements
urchins in any control plot (i.e. there was no variance), and I expected their effect on light to increase as they
precluding the use of repeated-measures ANOVA to grew, and subsequently decrease as they senesced at
analyze these data. Instead, I took the average across the end of the summer. I used the sequential Bonfer-
all sample dates for each plot and performed a t-test on roni method (Sokal & Rohlf 1995) to correct my critical
the square-root transformation of these data to look for p-value for multiple comparisons within each site.
an overall effect of the treatment on average urchin I analyzed urchin algal preference data using non-
abundance. parametric statistics because transformations failed to
Sediment, nutrient, water flow and kelp growth data make these data homoscedastic. I tested the null hypo-
were each analyzed using a 2-way ANOVA with site thesis that Strongylocentrotus droebachiensis con-
and treatment as fixed factors. sumed all 3 algal species presented to it equally using
Light data were analyzed using a 2-way repeated- a Kruskal-Wallis non-parametric test. I then used non-
measures ANOVA with site and treatment as fixed fac- parametric unplanned comparisons (Zar 1999) to de-
tors. Percent transmission light data were arcsin-trans- termine how prey differed from one another.
formed prior to analyses. I followed up the ANOVA on I used a t-test to compare the number of urchins
light data with a 1-tailed t-test (because I had an a pri- eaten in tanks with Sargassum muticum versus tanks
ori expectation that light would be lower in control containing native kelp species in order to determine
plots than removal plots) on each sampling date to whether predation risk differed between those 2 treat-
determine on which dates light differed significantly. ments. The data were log-transformed prior to analysis
This follow-up analysis was important because the to achieve homoscedasticity.
Table 1. Effect of Sargassum muticum on native algae. Results of
ANOVA and ANCOVA (for understory abundance) testing effect
RESULTS
of S. muticum removal
Removal experiments
Source of variation SS df MS F p
The structure of the native algae community
Canopy abundance was substantially altered by the removal of Sar-
Site 4.74 1 4.74 2.30 0.149 gassum muticum. Canopy algae were less abun-
Treatment 47.06 1 47.06 22.86 0.000
Site × Treatment 3.03 1 3.03 1.47 0.242 dant in control plots (p < 0.001; Table 1, Fig. 1)
Error 32.93 16 2.06 and this effect did not differ between the 2 sites
Time 13.50 13 1.04 3.90 0.000 (Table 1). The significant time × site interaction
Time × Site 8.14 13 0.63 2.35 0.006 (p = 0.006; Table 1, Fig. 1) indicates that the
Time × Treatment 6.65 13 0.51 1.92 0.029
Time × Site × Treatment 5.59 13 0.43 1.61 0.083 temporal dynamics of canopy abundance differed
Error 55.40 208 0.27 between the sites and was probably caused by
Understory abundance the somewhat delayed response of canopy algae
Site 6.77 1 6.77 4.95 0.042 to removal at Colin’s Cove (Fig. 1). Finally, a sig-
Treatment 12.05 1 12.05 8.81 0.009 nificant time × treatment interaction (p = 0.029;
Site × Treatment 0.03 1 0.03 0.02 0.889
Covariate (June 1999) 8.16 1 8.16 5.97 0.027
Table 1, Fig. 1) showed that the difference in
Error 20.52 15 1.37 canopy abundance between the 2 treatments
Time 2.37 9 0.26 1.87 0.061 fluctuated over time, which could be explained
Time × Site 2.93 9 0.32 2.31 0.019 by the slower response of canopy algae at
Time × Treatment 0.51 9 0.06 0.40 0.931
Colin’s Cove as well as the increase in canopy
Time × Site × Treatment 1.38 9 0.15 1.09 0.372
Time × Covariate 1.38 9 0.15 1.09 0.375 abundance in Colin’s Cove controls in early 2000
Error 18.99 135 0.14 relative to Point George. Although there was
Canopy richness not a significant time × site × treatment interac-
Site 2.75 1 2.75 3.77 0.070 tion (p = 0.083, Table 1), it did appear that the
Treatment 10.87 1 10.87 14.92 0.001 recovery of canopy algae in removal plots was
Site × Treatment 0.48 1 0.48 0.66 0.427
Error 11.65 16 0.73 delayed at Colin’s Cove relative to Point George
Time 7.96 13 0.61 4.48 0.000 (Fig. 1). Despite this possible difference in the
Time × Site 3.25 13 0.25 1.83 0.041 timing of recovery, native canopy algae were 4
Time × Treatment 2.15 13 0.16 1.21 0.275 to 5 times as abundant in removal plots com-
Time × Site × Treatment 4.98 13 0.38 2.80 0.001
Error 28.43 208 0.14 pared to control plots at both sites by the last
census date.
Britton-Simmons: Effects of an introduced alga 69
Understory algae were also less abundant in control
plots (p = 0.009; Table 1, Fig. 2) and this effect did not
differ between sites (Table 1). Including pre-removal
understory abundance as a covariate explained a sig-
nificant amount of variation in the ANCOVA model
(p = 0.027; Table 1). As with the data for canopy algae,
there was a significant time × site interaction (p = 0.019;
Table 1), which was probably caused by the somewhat
delayed recovery of understory algae in removal plots
at Colin’s Cove (Fig. 2). Although the recovery of
understory algae at Colin’s Cove lagged behind that at
Point George slightly, by the end of the experiment
understory algae were twice as abundant in removal
plots as control plots at both sites compared to a similar
100% initial difference at Point George and a 62%
initial difference at Colin’s Cove.
Native canopy richness was lower in control plots
than removal plots (p = 0.001; Table 1, Fig. 3). On aver-
age, Sargassum muticum displaced 1 native species at
both sites. As was the case for canopy and under-
story abundance, a significant time × site interaction
(p = 0.041; Table 1) probably was caused by the some-
what delayed recovery of canopy algae in removal
Fig. 2. Abundance (mean ± SE) of native understory algae in
Sargassum muticum removal experiments at Point George and
Colin’s Cove (n = 5). First data point in each series is for
pre-removal census
plots at Colin’s Cove (Fig. 3). Finally, the significant
time × site × treatment interaction (p = 0.001; Table 1)
was caused by a lack of concordance in the temporal
dynamics of the treatment effect at the 2 sites. During
1999, the control and removal means diverged at Point
George but remained roughly equal at Colin’s Cove
(Fig. 3). The following year, in 2000, the treatment
means at Point George began to converge again while
those at Colin’s Cove had just begun to diverge (Fig. 3).
The relative abundance of the 2 most common native
kelp species, Laminaria bongardiana and Agarum fim-
briatum, differed between control and removal plots.
The ratio of L. bongardiana to A. fimbriatum was lower
in control plots, where Sargassum muticum was
present (Scheirer-Ray-Hare extension of the Kruskal-
Wallis test, H = 7.22, p < 0.01; Fig. 4). This effect did
not differ between sites (H = 0.63, p < 0.5), and there
was no indication of a site × treatment interaction
(H = 0.37, p < 0.9).
The abundance of green urchins was negatively
Fig. 1. Abundance (mean ± SE) of native canopy algae in
Sargassum muticum removal experiments at Point George
affected by the presence of Sargassum muticum. Al-
and Colin’s Cove (n = 5). First data point in each series is for though there were no urchins in any plot during the
pre-removal census first summer of the experiment, a year later they had
70 Mar Ecol Prog Ser 277: 61–78, 2004
begun to forage regularly in removal plots at Point all sample dates for each plot showed that urchins
George (Fig. 5). Their absence during the first summer were significantly more abundant in removal plots
of the experiment probably reflects the near-absence than control plots (t 8 = –6.34, p < 0.001; Fig. 5) at Point
of kelps, their preferred food, in experimental plots. George. Green urchins were never observed at Colin’s
The appearance of urchins in removal plots was pre- Cove, but red urchins (S. franciscanus) were present at
ceded by an increase in canopy algae in removal plots deeper depths at that site throughout the experiment.
at Point George (Fig. 1). A t-test on the average Several variables showed no response to Sargassum
Strongylocentrotus droebachiensis abundance across muticum manipulation. I found no evidence that
S. muticum altered the percent cover of crustose
coralline algae, filamentous turf-forming algae, or bare
rock (Table 2). Similarly, there was no evidence for an
effect of S. muticum on the abundance of detritivores,
the abundance of herbivorous molluscs, or the species
richness of the invertebrate community (Table 2).
Power analysis revealed that I had low statistical power
to detect effects on some of these variables (Table 2).
However, in several cases (bare rock, crustose algae
and invertebrate richness) the extremely small effect
sizes involved would have made it difficult to detect a
significant effect, even if a more powerful experimen-
tal design had been employed. In the case of turf-form-
ing algae, it appears that a large amount of variation
may have obscured the effect of S. muticum (Table 2).
The results of the structural equation modeling
(SEM) are presented in Fig. 6. Structural equation
coefficients (values next to arrows in Fig. 6) indicate
the sign and strength of the effect of one variable on
another. More specifically, each coefficient indicates
the change in abundance of the dependent variable
that would result from a 1-unit change in the predictor
variable. Thick arrows indicate statistically significant
pathways (p < 0.05) and thin arrows indicate non-
significant paths.
The SEM results suggest that the effect of Sargas-
sum muticum on green urchins was an indirect effect,
Fig. 3. Species richness (mean ± SE) of native canopy algae in
Sargassum muticum removal experiments at Point George
and Colin’s Cove (n = 5). First data point in each series is for
pre-removal census
Fig. 5. Strongylocentrotus droebachiensis. Abundance (mean
Fig. 4. Ratio of the number of Laminaria bongardiana (L.b.) to ± SE) in Sargassum muticum removal experiment at Point
the number of Agarum fimbriatum (A.f.) (n = 5). Both species George (n = 5). First data point in each series is for pre-
are native kelps removal census
Britton-Simmons: Effects of an introduced alga 71
Table 2. Response variables that showed no statistically significant (α = 0.05) response to removal of Sargassum muticum (n = 5 for
each mean). Rock, turf-forming algae and coralline algae data are % cover. All abundance data are means ± 1 SE for each
experimental plot averaged across all sampling dates. Power = 1 – β, or probability of correctly rejecting H o if it were false
Variable Point George Colin’s Cove Power
Control Removal Control Removal
Bare rock 04.9 ± 2.3 12.1 ± 3.4 07.9 ± 4.3 01.1 ± 0.6 0.05
Crustose coralline algae 21.5 ± 7.9 34.1 ± 9.0 18.7 ± 6.1 20.1 ± 5.9 0.20
Turf-forming algae 49.1 ± 8.8 34.5 ± 8.3 53.5 ± 7.7 42.4 ± 5.1 0.40
Invertebrate richness 07.1 ± 0.6 07.8 ± 0.4 04.7 ± 0.5 04.7 ± 0.6 0.07
Detritivores 13.5 ± 3.0 05.8 ± 1.4 06.1 ± 1.4 06.8 ± 2.0 0.54
Herbivorous molluscs 07.0 ± 1.0 11.0 ± 0.9 04.0 ± 1.1 04.0 ± 1.2 0.62
Light
The overall ANOVA on light data from both sites
indicated a highly significant effect of treatment
(p < 0.001; Table 3, Fig. 7) on light transmittance, and
this effect did not differ between the 2 sites (p = 0.570;
Table 3). These results reflect the fact that light inten-
sity was (on average) more intense in removal plots
than control plots at both sites (Fig. 7), even though the
absolute difference between treatments varied over
time as Sargassum muticum’s fronds grew, matured,
Fig. 6. Structural equation model for Point George data in
and senesced. There was also a significant site effect
Sargassum muticum removal experiment (urchins = Strongy-
locentrotus droebachiensis; palatable canopy = Laminaria (p = 0.001; Table 3), with percent transmission of light
bongardiana, L. complenata, Alaria marginata, Nereocystis at Point George higher than at Colin’s Cove (Fig. 7).
luetkeana). Arrows indicate direction of causality. Thick This systematic difference is probably due to the fact
arrows indicate statistically significant (p < 0.05) paths from that the shoreline at Point George faces W–SW and
multiple-regression analysis; structural equation coefficients
are shown next to each arrow. Removing direct link between
receives more intense sunlight than the E-facing shore
S. muticum and urchins did not change model fit or values of at Colin’s Cove. Finally, a significant time × treatment
the remaining coefficients. Observed correlation matrix interaction (p < 0.001; Table 3) indicated that the treat-
did not differ significantly from that predicted by model ment effect changed over time. To determine when
(χ22 = 5.08, p = 0.08; Akaike information criterion, AIC = 0.267)
significant differences in light level occurred, I fol-
lowed up this analysis with t-tests on each sampling
date within the 2 sites. A 1-tailed t-test on each date
not a direct effect (Fig. 6). The direct pathway from S. revealed that light transmittance was lower in control
muticum to urchins was very weak and was not statis- plots at both sites in May (p < 0.05 and p < 0.01 at Point
tically significant (coefficient = 0.03, p = 0.90; Fig. 6). In George and Colin’s Cove, respectively; Fig. 7) and
contrast, the net effect of the indirect pathway from S. June (p < 0.001 at Point George and Colin’s Cove;
muticum to urchins, obtained by multiplying the coef- Fig. 7).
ficients for each of the links involved (Wootton 1994),
was very strong (–1.07 × 0.94 = –1.01) and both of the
Table 3. Results of ANOVA testing effect of Sargassum muticum
component pathways (S. muticum to palatable canopy on light transmittance in removal experiments
and palatable canopy to urchins) were statistically sig-
nificant. Removing the direct link between S. muticum Source of variation SS df MS F p
and urchins did not change the model fit or the values
of the remaining coefficients. In fact, the 2 models Site 0.128 1 0.128 16.07 0.001
were statistically indistinguishable (χ22 = 5.08, p = 0.08, Treatment 0.251 1 0.251 31.54 0.000
Site × Treatment 0.003 1 0.003 00.34 0.570
AIC = 0.267 for both models), providing additional evi-
Error 0.127 160 0.008
dence that the direct effect of S. muticum on urchins Time 0.696 3 0.232 70.75 0.000
was negligible. In summary, the SEM analysis sug- Time × Site 0.009 3 0.003 00.95 0.423
gests that the negative effect of S. muticum on urchins Time × Treatment 0.239 3 0.080 24.31 0.000
was an indirect effect that also involved palatable Time × Site × Treatment 0.021 3 0.007 02.18 0.103
Error 0.157 480 0.003
canopy algae (Fig. 6).
72 Mar Ecol Prog Ser 277: 61–78, 2004
Fig. 8. Dissolution of gypsum blocks (mean ± SE) in experi-
mental plots where Sargassum muticum was present (Con-
trol) and in plots from which it had been removed (Removal)
more numerous in removal plots (Figs. 1 & 2). The
deposition of organic sediment was not different in
control and removal plots (ANOVA, F1,14 = 3.73, p =
0.074; Fig. 9). The deposition of inorganic sediment did
not differ between the treatments either (ANOVA,
F1,14 = 0.43, p = 0.524), but there was a significant
site × treatment interaction for this response variable
(ANOVA, F1,14 = 5.42, p = 0.035). This interaction prob-
ably occurred because inorganic sedimentation tended
Fig. 7. Percent transmission (mean ± SE) of photosynthetically to be slightly higher in removal plots than control plots
active radiation (400 to 700 nm, PAR) in Sargassum muticum at Point George, but higher in control than removal
removal experimental plots at Point George and Colin’s Cove plots at Colin’s Cove (Fig. 9). Finally, total sediment
in 2000. Secondary y-axis shows height of S. muticum (± SE) deposition did not differ between the 2 treatments
in control plots at time each set of light measurements was
taken. Asterisks indicate sample dates where t-test indicated (ANOVA, F1,14 = 0.96, p = 0.344), but it also showed
significant difference between treatments (*, **, ***: p < 0.05, some suggestion of a site × treatment interaction
0.01, and 0.001, respectively) (ANOVA, F1,14 = 4.37, p = 0.055), which likely resulted
for the same reason stated above for inorganic sedi-
mentation. Overall, there was no indication that S.
Water flow muticum removal had an effect on sedimentation by
the time that native algae had recovered to replace the
I found no evidence that the Sargassum muticum S. muticum that had been removed 2 yr earlier.
removal treatment had an effect on water flow at a
distance of 30 cm above the substratum (ANOVA,
F1,15 = 0.04, p = 0.85; Fig. 8). However, water flow at Nutrients
Point George was considerably higher than at Colin’s
Cove (ANOVA, F1,15 = 65.29, p < 10– 5; Fig. 8). There None of the 5 nutrients assayed differed between
was no evidence of a site × treatment interaction. control and removal plots (Table 4) (ANOVA, F1,16,
Sedimentation Table 4. Nutrient concentrations (µM means ± 1 SE) of water samples taken from
Sargassum muticum removal experiment plots (n = 5 for each mean)
Analysis of sedimentation data col-
Nutrient Point George Colin’s Cove
lected in July 2001 yielded no
Control Removal Control Removal
evidence that Sargassum muticum
altered sedimentation rates. At the PO4 01.96 ± 0.012 01.97 ± 0.007 1.91 ± 0.007 01.89 ± 0.013
time these data were collected, canopy Si(OH)4 49.05 ± 0.256 48.56 ± 0.195 46.97 ± 0.2360 47.04 ± 0.113
and understory algae had already NO3 22.47 ± 0.088 22.46 ± 0.129 21.88 ± 0.0960 21.67 ± 0.374
NO2 00.31 ± 0.003 00.30 ± 0.002 0.30 ± 0.002 00.30 ± 0.002
responded to the S. muticum removal NH4 00.52 ± 0.015 00.53 ± 0.021 0.57 ± 0.010 00.54 ± 0.028
treatment, with both types of algae
Britton-Simmons: Effects of an introduced alga 73
of replicates. At Colin’s Cove, 1 of the control plots lost
all its plants, and consequently there were only 4 con-
trol replicates from that site. There was no statistically
significant relationship between site (ANOVA, p =
0.826) or treatment (ANOVA, p = 0.826) and the num-
ber of plants lost. Likewise, there was no relationship
between the number of plants lost and growth of the
remaining kelps (R21,17 = 0.13, p = 0.133).
Linear blade growth of Laminaria bongardiana was
2 to 3 times faster in plots from which Sargassum muti-
cum was absent than in those where it was present
(ANOVA, F1,15 = 25.95, p < 0.001; Fig. 10). Growth did
Fig. 9. Sediment accumulation (mean ± SE) in traps placed in not differ between the 2 sites (ANOVA, F1,15 = 0.20, p >
experimental plots at Point George and Colin’s Cove where 0.30) and there was no site × treatment interaction
Sargassum muticum was present (Control) and in plots from
(ANOVA, F1,15 = 0.11, p > 0.73). Analysis of blade area
which it had been removed (Removal). Total sediment is sum
of organic and inorganic components of sediment and kelp mass data yielded comparable results (data
not shown).
Urchin food preferences
Strongylocentrotus droebachiensis distinguished
among the 3 species of algae that were presented to
it in the preference experiment (Kruskal-Wallis, p =
0.001; Fig. 11). Non-parametric unplanned compar-
isons (Zar 1999) showed that Laminaria bongardiana
was preferred over both Agarum fimbriatum (p < 0.01)
and Sargassum muticum (p < 0.005), but that A. fim-
briatum and S. muticum were equally ignored. This
experiment supported the hypothesis that green
urchins do not feed on S. muticum in the field because
Fig. 10. Laminaria bongardiana. Growth (mean ± SE) of
native kelp in plots where Sargassum muticum was present they found its tissue unpalatable. Nevertheless, even if
(Control) and plots from which it had been removed they find S. muticum palatable its morphology could
(Removal) at Point George and Colin’s Cove (n = 5) prevent them from effectively exploiting it as a food
resource in the field.
p > 0.30 in every case). However, phosphate (ANOVA,
F1,16 = 40.24, p < 0.001), silicate (ANOVA, F1,16 = 75.18,
p < 0.001) and nitrate (ANOVA, F1,16 = 10.92, p < 0.01)
concentrations were significantly higher at Point
George (Table 4). Since water samples were taken on
different (but consecutive) days at each site, these site
differences may reflect day-to-day fluctuations as
water masses move through this region.
Kelp growth
Each plot began the kelp growth experiment with 3
Laminaria bongardiana, but several plots lost 1 or more
plants during the course of the experiment. I averaged Fig. 11. Algal mass (mean + SE) consumed by native sea
the results for all surviving plants within each plot and urchin Strongylocentrotus droebachiensis in food choice
experiment where 2 native kelps (Laminaria bongardiana and
used those means in the statistical analysis. Thus, the Agarum fimbriatum) and Sargassum muticum were offered
extra plants in each plot increased the accuracy of the simultaneously. Letters indicate which means differ signifi-
growth measurements but did not increase the number cantly (non-parametric unplanned comparisons)
74 Mar Ecol Prog Ser 277: 61–78, 2004
interaction (F1,12 = 4.32, p = 0.08), the difference
between the 2 treatments did appear to decline from
Days 3 to 8 (Fig. 13). Despite the absence of an over-
all difference between treatments, further analysis
showed that significantly more urchins were visible in
the S. muticum treatment on Day 3 (t = –2.843, p =
0.01), but that this difference had dissipated by Day 8
(t = –0.830, p = 0.42). Even though there was little
difference between the treatments at the end of the
experiment (Day 8), it does appear that a higher pro-
portion of urchins in the kelp treatment were hidden
for the majority of the experiment (Fig. 13). Contrary to
expectation, the urchins in the kelp treatment were not
Fig. 12. Strongylocentrotus droebachiensis. Number of green
urchins (mean + SE) eaten by sea star Pycnopodia heliantho- less vulnerable to sea star predation (Fig. 12). This
ides in tanks containing Sargassum muticum monoculture experiment does provide some evidence that green
or mixture of native kelp species (n = 7) urchins can more easily hide themselves in algal
canopies dominated by native kelps than in those dom-
inated by S. muticum. Furthermore, these data are a
Urchin predation conservative indication of the differential use of S.
muticum and kelps by urchins, because in this labora-
The sea star predation hypothesis was not supported tory experiment urchins were able to make use of S.
by experimental data. There was no difference in pre- muticum in a way they cannot commonly do in the
dation by Pycnopodia helianthoides on Strongylocen- field. Most of the urchins that managed to hide them-
trotus droebachiensis between the Sargassum muti- selves in the S. muticum treatment did so by climbing
cum and mixed-kelp treatments (t = –0.632, p = 0.545; the walls of the tank, grasping onto the fronds of
Fig. 12). While the average number of urchins visible nearby S. muticum, and using those fronds to cover
(≥ 25% of test) was greater on Days 3 and 8 (Fig. 13), their tests (own pers. obs.). Although S. muticum does
repeated-measures ANOVA analysis indicated no grow adjacent to vertical rock surfaces in the field at
statistically significant difference between treatments some sites (e.g. at the base of walls or next to large
(F1,12 = 4.30, p = 0.06). There was a significant effect of boulders), these places are relatively rare and I have
time in the ANOVA model (F1,12 = 15.87, p = 0.002), never observed green urchins using S. muticum to
which resulted because both treatments means de- hide themselves in this manner in the field. Overall,
creased between the 2 sampling dates (Fig. 13). the results of this experiment suggest that increased
Although there was no significant time × treatment predation by sea stars in S. muticum-dominated areas
cannot explain why green urchins were less abundant
in control plots in the S. muticum removal experiment.
DISCUSSION
Removal experiments showed that Sargassum muti-
cum has substantial direct effects on the native algae
community characteristic of the San Juan Islands.
Competition with S. muticum reduced the abundance
of native canopy algae by approximately 75% and
native understory algae by about 50% (Figs. 1, 2 & 6).
S. muticum also displaced (on average) 1 native spe-
cies of canopy algae (Fig. 3), thereby reducing the spe-
cies richness of native canopy species, but leaving total
algal richness unchanged. However, the negative
effect on species richness that occurred at the small
Fig. 13. Strongylocentrotus droebachiensis. Number of visible scale of my experimental plots (0.25 m2) may not be
green urchins (mean ± SE) in tanks containing mixture of
native kelps versus tanks containing Sargassum muticum
important at larger scales.
(n = 7). Treatment means differed on Day 3 (t-test, p = 0.01) In addition to affecting the abundance and richness
but not on Day 8 (t-test, p = 0.42) of native canopy algae, Sargassum muticum changed
Britton-Simmons: Effects of an introduced alga 75
the relative abundance of the 2 most common native this scenario, palatable algae gain protection from her-
kelp species, Laminaria bongardiana and Agarum fim- bivores when they are spatially associated with un-
briatum. L. bongardiana was relatively less abundant palatable algae because herbivores avoid foraging in
in plots where S. muticum was present (Fig. 4). This areas where the unpalatable species are present. This
result was probably caused by interspecific differences hypothesis was not supported by SEM analyses of my
in light requirements. Laminaria bongardiana is a spe- removal experiment data (Fig. 6). (2) Algal canopies
cies that has few chemical defenses (Hammerstrom et dominated by S. muticum could enable the sea star
al. 1998), grows relatively fast, is found only in the Pycnopodia helianthoides (a locally common green
shallow subtidal zone (own pers. obs.), and grows urchin predator) to prey more effectively on green
more slowly when transplanted beneath S. muticum urchins. Green urchins typically cover their aboral sur-
(Fig. 10). A. fimbriatum, on the other hand, has high face with kelp or other algae while feeding, but S.
concentrations of polyphenolics to deter herbivory muticum’s positively buoyant fronds and wiry mor-
(Hammerstrom et al. 1998), grows relatively slowly, phology could make it unsuitable for this purpose.
and is the deepest-occurring kelp species in this region Therefore, green urchins might be more vulnerable to
(commonly found at depths of 15 m or more; own pers. predation by foraging sea stars in areas where S.
obs.). Thus, L. bongardiana is probably more sensitive muticum is the dominant alga. A laboratory experi-
than A. fimbriatum to the shading caused by S. muti- ment designed to test this hypothesis showed that
cum (Fig. 7), and was relatively less abundant in although green urchins seemed better able to conceal
control plots as a result. themselves when feeding on native kelps compared to
Field measurements suggest that the effects of Sar- S. muticum (Fig. 13), there was no difference in sea
gassum muticum on macroalgae are probably a result star (P. helianthoides) predation between the 2 treat-
of shading (Fig. 7). There was no evidence that S. ments (Fig. 12). My conclusion that S. muticum nega-
muticum had an effect on nutrients (Table 4), and be- tively affects urchins is in accordance with a study of
cause algae in this region do not appear to be nutrient- S. muticum in California tidepools that found a similar
limited (Wootton 1991, Pfister & Van Alstyne 2003) it result for a congener, Strongylocentrotus purpuratus
is unlikely that competition for nutrients plays an (Wilson 2001). However, the indirect effect that was
important role in these interactions. However, my the cause of the negative effect on urchins in my sys-
limited sampling of water flow (Fig. 8) and sedimenta- tem has not been demonstrated in any previous study.
tion (Fig. 9) makes it difficult to rule out these factors Although the removal experiments at Point George
completely, and other resources I did not take into and Colin’s Cove yielded largely similar results, there
account (e.g. space) could also be important. were some notable differences between the 2 sites. For
Structural equation modeling (SEM) showed that example, the recovery dynamics of canopy algae fol-
Sargassum muticum-induced changes in palatable lowing the removal of Sargassum muticum differed
canopy algae had an important indirect effect on a between them (Fig. 1). Whereas the recovery at Point
native herbivore, the green sea urchin Strongylocen- George was rapid, it was considerably delayed at
trotus droebachiensis (Fig. 6). Green urchins were not Colin’s Cove. Source-populations of native canopy and
recorded in experimental plots at either site during the understory species were only 2 to 3 m deeper than
first year of the experiments (Fig. 5), although they my experimental plots at both sites. Under most con-
were present at Point George at depths deeper than ditions, kelps are unlikely to be strongly dispersal-
my experiments (own pers. obs.). However, by the limited over this distance (Reed et al. 1988, Fredriksen
summer of 2000, green urchins had begun foraging et al. 1995, Forrest et al. 2000). Nevertheless, stronger
regularly in the removal plots at Point George, where a tidal currents at Point George (Fig. 8) may have facili-
robust community of kelps and red algae had devel- tated the dispersal of algae into plots from which S.
oped following the removal of S. muticum (Fig. 5). It muticum had been removed.
appears that green urchins avoided control plots The presence of green urchins at Point George but
because the kelp genera Laminaria, Alaria and Nereo- not Colin’s Cove was another source of variation
cystis, which are their preferred food (Fig. 11 and between sites. These urchins can be locally abundant,
Vadas 1968, 1977, Larson et al. 1980), were less abun- but have a discontinuous distribution across sites. Thus
dant due to competition with S. muticum (Fig. 1). it is not surprising that they were present at only 1 of
I also tested 2 additional hypotheses that might my sites. Canopy algae abundance in removal plots
explain why green urchins avoided plots where Sar- at Point George decreased following the arrival of
gassum muticum was abundant: (1) The presence of urchins at that site in the spring of 2000 (Fig. 1). Native
unpalatable S. muticum could reduce grazing by sea canopy algae richness was higher in the absence of
urchins on nearby palatable native species (so-called Sargassum muticum at both sites, but urchin grazing in
associational plant refuge; e.g. Pfister & Hay 1988). In the removal plots at Point George caused a decline in
76 Mar Ecol Prog Ser 277: 61–78, 2004
canopy richness during 2000 and briefly led to the 2000, own pers. obs.), where it is susceptible to desic-
convergence of control and treatment dynamics at cation and frost (Norton 1977), and it generally reaches
that site (Fig. 3). lower densities in the intertidal compared with the
Because kelps are the numerically and physically subtidal, even in tidepools (Wilson 2001, own pers.
dominant plants in these algal communities, it would obs.). Thus, one might expect it to have less of an
be useful to know whether only 1 or both phases of impact on native species in the intertidal simply
their life cycles are affected by Sargassum muticum. because it is less abundant there.
The lowest light levels I recorded in plots containing Although different sites in the San Juan Islands vary
S. muticum (29 and 36 µE m–2 s–1 at Colin’s Cove and considerably in the density of Sargassum muticum
Point George, respectively) are below the threshold at they contain, it is extremely difficult to find sites which
which light limitation in kelp sporophytes is expected it has not yet invaded (own pers. obs.). Given the
(150 to 200 µE m–2 s–1, Lüning 1981), and growth results of these experiments, 2 impacts may be of most
experiments clearly showed that sporophytes grew concern: First, although my experiments were con-
more slowly under the S. muticum canopy (Fig. 10). ducted on a small scale, it seems likely that the total
However, these light levels exceed the threshold abundance of native kelp in the San Juan Islands has
where kelp gametophytes are saturated for vegetative been reduced by S. muticum. Since a wide variety of
growth (20 µE m–2 s–1, Lobban & Harrison 1994). Thus, taxa, including even marine mammals and birds, uti-
the effect of S. muticum on the kelp component of the lize these shallow subtidal kelp communities, the con-
algal community is probably due to its impact on the sequences of this invasion may extend well beyond the
sporophyte phase, not the gametophyte phase of the benthic organisms that were the focus of this study.
kelp life cycle. One important caveat to this conclusion Second, the avoidance by green urchins of areas with
is that my light measurements reflect maximum irradi- dense S. muticum has important potential indirect
ance values because the data were taken at midday; effects for kelp communities. Urchins are an important
average light intensity was undoubtedly much lower. disturbance agent because they clear patches of rock,
In general, many species of native algae are likely to thereby resetting the successional sequence of the
be affected due to shading by Sargassum muticum community (Vadas 1968, Duggins 1980). As a conse-
because of the timing of its life cycle in this region. The quence, algal diversity is enhanced by the creation of
fronds of S. muticum usually begin growing in early a mosaic of patches that differ in their successional
March each year, and in 2000 it was already having stages (Vadas 1968, Duggins 1980). The absence of
a significant effect on light by mid-May (Fig. 7). green urchins could ultimately cause a decline in
Although some perennial kelp species are capable of macroalgal diversity in shallow, subtidal kelp commu-
reproducing in the winter (e.g. Agarum fimbriatum nities. Furthermore, by concentrating their grazing in
and Laminaria bongardiana; Vadas 1968, and own areas outside S. muticum populations, green urchins
pers. obs.) many annual kelps (e.g. Costaria costata) may facilitate the spread of this invader, which has
must reproduce in the spring and summer. Further- higher recruitment in areas that have been experi-
more, most native red algae (i.e. understory algae) are mentally denuded of algae (Britton-Simmons 2003).
either only present during the spring and summer, or The response of green urchins to the presence of
experience most of their growth during that time (own Sargassum muticum suggests that this generalist her-
pers. obs.). The months during which S. muticum is bivore is presently unlikely to slow the rate of S.
having its strongest effect on light (Fig. 7) is also a crit- muticum’s spread (Figs. 6 & 11). The food preferences
ical period of time for the growth and reproduction of of green urchins could change over time so that S.
many species of native algae. muticum becomes a more preferred food resource.
Previous studies of Sargassum muticum have varied Moreover, natural selection could eventually favor
widely in their conclusions about its effect on native such a shift, especially if S. muticum continues to in-
communities. However, my study is in accordance with crease in abundance and displace palatable native
1 generalization that emerges from a review: studies in kelp species. However, S. muticum’s wiry morpho-
the intertidal zone have found little or no impact of logy and positive buoyancy may make it largely
S. muticum (DeWreede 1983, DeWreede & Vander- inaccessible to green urchins and thereby preclude an
meulen 1988, Viejo 1997, Wilson 2001), but studies in evolutionary shift in feeding preferences.
the subtidal zone indicate relatively strong effects My removal experiments were conducted at sites
(Ambrose & Nelson 1982, Staehr et al. 2000, present where the abundance of Sargassum muticum is at the
study). Considering the vertical distribution of S. upper end of its distribution of densities. If the effect of
muticum, this general trend is not surprising. The S. muticum on light is proportional to its abundance,
lower intertidal is at the upper edge of S. muticum’s then I would expect its effect on native species at any
vertical distribution (DeWreede 1983, Staehr et al. particular site to also be proportional to abundance.
Britton-Simmons: Effects of an introduced alga 77
Nevertheless, a more complex relationship is possible. Cohen BF, McArthur MA, Parry GD (2001) Exotic marine
For example, other studies have shown that the sign of pests in the Port of Melbourne, Victoria. MAFRI Rep 25:
1–96
an interaction between exotic and native species can
DeWreede RE (1983) Sargassum muticum (Fucales, Phaeo-
change as their relative densities change (Reush & phyta): regrowth and interaction with Rhodomela larix
Williams 1998). Thus, it may be necessary to evaluate (Ceramiales, Rhodophyta). Phycologia 22:153–160
the effects of S. muticum at various densities in order DeWreede RD (1996) The impact of seaweed introductions on
to understand more completely how it interacts with biodiversity. Global Biodiv 6(3):2–9
DeWreede RE, Vandermeulen R (1988) Lithothrix aspergillum
native species. At low densities it may have a relatively (Rhodophyta): regrowth and interaction with Sargassum
small effect on light, although even this small effect muticum (Phaeophyta) and Neorhodomela larix (Rhodo-
may be sufficient to alter the competitive outcome of phyta). Phycologia 27:469–476
interspecific interactions among native kelps without Deysher L, Norton R (1982) Dispersal and colonization in Sar-
gassum muticum (Yendo) Fensholt. J Exp Mar Biol Ecol
reducing overall kelp abundance.
56:179–185
This study illustrates that an introduced alga can Duggins DO (1980) Kelp beds and sea otters: an experimental
have important effects on native communities. These approach. Ecology 61(3):447–453
results, in combination with other studies that have Duggins DO, Simenstad CA, Estes JA (1989) Magnification of
also demonstrated strong effects of non-indigenous secondary production by kelp detritus in coastal marine
ecosystems. Science 245:170–173
algae species (Verlaque 1994b, Villele & Verlaque Ebeling AW, Laur DR, Rowley RJ (1985) Severe storm distur-
1995, Walker & Kendrick 1998, Levin et al. 2002) high- bances and reversal of community structure in a southern
light the importance of future research into the spread California kelp forest. Mar Biol 84:287–294
and impact of introduced algae. Because of their Eckman JE, Duggins DO, Sewell AT (1989) Ecology of under-
story kelp environments. I. Effects of kelps on flow and
worldwide distribution and capacity to alter native
particle transport on the bottom. J Exp Mar Biol Ecol 129:
communities, non-indigenous algae are potentially 173–187
important agents of global ecological change. Faul F, Erdfelder E (1992) GPOWER: a priori, post-hoc, and
compromise power analyses for MS-DOS (Computer pro-
gram). Bonn University, Department of Psychology, Bonn
Acknowledgements. None of this work would have been pos- Forrest BM, Brown SN, Taylor MD, Hurd CL, Hay CH (2000)
sible without the many individuals who volunteered their The role of natural dispersal mechanisms in the spread
time to dive with me on this project. I am especially grateful to of Undaria pinnatifida (Laminariales, Phaeophyceae).
J. Kido, E. Iyengar, B. Pister, L. Self, and C. Catton for being Phycologia 39:547–553
my ‘buddies’. For thoughtful discussions which improved the Fredriksen S, Sjotun F, Lein T, Rueness J (1995) Spore disper-
quality of this work I thank T. Wootton, C. Pfister, M. Leibold, sal in Laminaria hyperborea (Laminariales, Phaeophy-
J. Bergelson, G. Dwyer, S. Williams, D. Duggins, T. Klinger, J. ceae). Sarsia 80:47–53
Kido, S. Hall, K. Polivka, P. Geddes, S. Harrell and B. Pister. Gardner WD (1980) Sediment trap dynamics and calibration:
J. Bergelson, C. Pfister, J. Byers, M. Wilson, K. Abbott, T. a laboratory evaluation. J Mar Res 38:41–52
Wootton, C. Peterson, and 2 anonymous reviewers helped me Giver KJ (1999) Effects of the invasive seaweed Sargassum
transform early versions of this manuscript into its present muticum on native marine communities in northern Puget
form. T. Wootton gave guidance and encouragement through- Sound, Washington. MS thesis, Western Washington
out this project. The staff of Friday Harbor Laboratories pro- University, Bellingham
vided logistical support including boats, dive facilities and Hammerstrom K, Dethier MN, Duggins DO (1998) Rapid
laboratory space, and for that I am thankful. This work was phlorotannin induction and relaxation in five Washington
funded by grants to K.B.S. from The Packard Foundation and kelps. Mar Ecol Prog Ser 165:293–305
The University of Chicago’s Hinds Fund, and grants to T. Hayduk LA (1987) Structural equation modeling with LISREL.
Wootton from The Marine Ecosystem Health Program and Johns Hopkins University Press, Baltimore
The Mellon Foundation. K.B.S. was supported by a Depart- Jones LC (1971) Studies on selected small herbivorous inver-
ment of Education GAANN fellowship during the course of tebrates inhabiting Macrocystis canopies and holdfasts
this study. in southern California kelp beds. Nova Hedwigia 32:
343–367
Juerg B (1996) Towards a new generation of sediment traps
LITERATURE CITED and a better measurement/understanding of settling parti-
cle flux in lakes and oceans: a hydrodynamical protocol.
Ambrose RF, Nelson BV (1982) Inhibition of giant kelp re- Aquat Sci 58:283–296
cruitment by an introduced brown alga. Bot Mar 25:265–267 Larson BR, Vadas RL, Keser M (1980) Feeding and nutritional
Bernstein BB, Jung N (1979) Selective pressures and coevolu- ecology of the sea urchin Stronglyocentrotus droebachi-
tion in a kelp canopy community in southern California. ensis in Maine, USA. Mar Biol 59:49–62
Ecol Monogr 49:335–355 Levin PS, Coyer JA, Petrik R, Good TP (2002) Community-
Britton-Simmons KH (2003) Establishment, spread, and im- wide effects of non-indigenous species on temperate
pact of the introduced Japanese seaweed, Sargassum rocky reefs. Ecology 83:3182–3193
muticum, in the San Juan Islands, WA. PhD thesis, The Lobban CS, Harrison PJ (1994) Seaweed ecology and physio-
University of Chicago, Chicago logy. Cambridge University Press, New York
Burnham KP, Anderson DR (1998) Model selection and infer- Lüning K (1981) Light. In: Lobban CS, Wynne MJ (eds) The
ence: a practical information–theoretic approach. Springer biology of seaweeds. University of California Press, Berke-
Verlag, New York ley, p 340–342
78 Mar Ecol Prog Ser 277: 61–78, 2004
Norton TA (1977) The growth and development of Sargassum Study (JGOFS) core measurements. Intergovernmental
muticum (Yendo) Fensholt. J Exp Mar Biol Ecol 26:41–53 Oceanographic Commission (IOC), manual and guides No
Peterson CH, Renaud PE (1989) Analysis of feeding prefer- 29. Paris, France
ence experiments. Oecologia 80:82–86 Vadas RL (1968) The ecology of Agarum and the kelp bed
Pfister CA, Hay ME (1988) Associational plant refuges: con- community. PhD thesis, University of Washington, Seattle
vergent patterns in marine and terrestrial communities Vadas RL (1977) Preferential feeding: an optimization
result from differing mechanisms. Oecologia 77:118–129 strategy in sea urchins. Ecol Monogr 47:337–371
Pfister CA, Van Alstyne KL (2003) An experimental as- Verlaque M (1994a) Checklist of introduced plants in the
sessment of the effects of nutrient enhancement on the Mediterranean: origins and impact on the environment
intertidal kelp Hedophyllum sessile (Laminariales, Phaeo- and human activities. Oceanol Acta 17:1–23
phyceae). J Phycol 39:285–290 Verlaque M (1994b) Mediterranean algal communities are
Reed DC, Laur DR, Ebeling AW (1988) Variation in algal dis- changing in the face of the invasive alga Caulerpa taxi-
persal and recruitment: the importance of episodic events. folia (Vahl) C. Agardh. Oceanol Acta 17:659–672
Ecol Monogr 58:321–335 Viejo RM (1997) The effects of colonization by Sargassum
Reush TBH, Williams SL (1998) Variable responses of native muticum on tidepool macroalgal assemblages. J Mar Biol
eelgrass Zostera marina to a non-indigenous bivalve Assoc UK 77:325–340
Musculista senhousia. Oecologia 113:428–441 Villele X de, Verlaque M (1995) Changes and degradation
Rueness J (1989) Sargassum muticum and other introduced in a Posidonia oceanica bed invaded by the introduced
Japanese macroalgae: biological pollution of European tropical alga Caulerpa taxifolia in the north western
coasts. Mar Pollut Bull 20(4):173–176 Mediterranean. Bot Mar 38:1–9
Scagel RF (1956) Introduction of a Japanese alga, Sargassum Walker DI, Kendrick GA (1998) Threats to macroalgal
muticum, into the Northeast Pacific. Fish Res Pap St Wash diversity: marine habitat destruction and fragmentation,
1:1–10 pollution and introduced species. Bot Mar 41:105–112
Setzer B, Link C (1971) The wanderings of Sargassum muti- Wilson M (2001) Is Sargassum muticum a benign invader of
cum and other relations. Stomatopod 2:5–6 tidepools on the Pacific coast of North America? MS thesis,
Shipley B (2000) Cause and correlation in biology (a user’s California State University Monterey Bay, Seaside
guide to path analysis, structural equations and causal Wootton JT (1991) Direct and indirect effects of nutrient on
inference). Cambridge University Press, Cambridge intertidal community structure: variable consequences of
Sokal RR, Rohlf FJ (1995) Biometry: the principle and practice seabird guano. J Exp Mar Biol Ecol 151:139–153
of statistics in biological research, 3rd edn. WH Freeman, Wootton JT (1994) Predicting direct and indirect effects: an
New York integrated approach using experiments and path analysis.
Staehr PA, Pedersen MF, Thomsen MS, Wernberg T, Krause- Ecology 75:151–165
Jensen D (2000) The invasion of Sargassum muticum in Lim- Yendo K (1907) The Fucaceae of Japan. J Coll Sci Imp Univ
fjorden (Denmark) and its possible impact on the indigenous Tokyo 21(12):1–174
macroalgal community. Mar Ecol Prog Ser 207:79–88 Zar JH (1999) Biostatistical analysis, 4th edn. Prentice Hall,
UNESCO (1994) Protocols for the Joint Global Ocean Flux Upper Saddle River, NJ
Editorial responsibility: Charles Peterson (Contributing Editor), Submitted: February 3, 2003; Accepted: February 3, 2004
Morehead City, North Carolina, USA Proofs received from author(s): August 9, 2004