Keough and Quinn 98
Effects of Periodic Disturbances from Trampling on Rocky Intertidal Algal Beds
Author(s): Michael J. Keough and G. P. Quinn
Source: Ecological Applications, Vol. 8, No. 1 (Feb., 1998), pp. 141-161
Published by: Ecological Society of America
Stable URL: http://www.jstor.org/stable/2641317
Accessed: 25/01/2010 14:06
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=esa.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Ecological Society of America is collaborating with JSTOR to digitize, preserve and extend access to
Ecological Applications.
http://www.jstor.org
Ecological Applicationis, 8(1), 1998, pp. 141-161
C) 1998 by the Ecological Society of America
EFFECTS OF PERIODIC DISTURBANCES FROM TRAMPLING ON
ROCKY INTERTIDAL ALGAL BEDS
MICHAEL J. KEOUGH1 AND G. P. QUINN2
Department of Zoology, University of Melbourne, Parkville, Victoria 3052, Australia
2Department of Ecology and Evolutionary Biology, Monash University,
Clayton, Victoria 3168, Australia
Abstract. We investigated the ability of an assemblage of animals and plants on rocky
shores in southeastern Australia to resist and/or recover from repeated pulse disturbances
in the form of trampling. Disturbances of four different intensities were applied experi-
mentally over six summers, with no human access at other times of the year. The dominant
intertidal plant, the brown alga Hormosira banksii, was affected by trampling, but the
effects were heterogeneous between sites. At two sites, a series of pulse disturbances
produced a series of pulse responses, although the effect of a given pulse varied among
years, possibly related to the severity of summer desiccating conditions each year. At the
third site, pulse disturbances produced a press response; at high levels of trampling, Hor-
mosira was almost eliminated within 2 yr, and at two intermediate levels of trampling,
cover was reduced from >90 to 60-70%, where it remained for 4 yr. Effects of trampling
showed little small-scale spatial variation. Untrampled areas did fluctuate through time,
often as a result of summer burnoff of algae. Natural disturbances occurred irregularly
through the study, and their effects varied on very small spatial scales (among plots <30
m apart).
Trampling enhanced the densities of a range of herbivorous mollusks, especially limpets,
and reduced the abundance of articulated coralline algae, which were abundant in the
understory of Hormosira mats. These effects varied among sites but showed much less
variation on smaller spatial scales. The reductions in coralline algae may be a direct effect
of trampling, but increases in mollusk abundance occurred some time after changes to
Hormosira cover, and those changes may be an indirect effect of trampling.
We compared the effects of trampling on areas of the shore that had been trampled for
two and four summers, to test whether a past history of disturbance influenced the effect
of a new disturbance. No significant effects were found on algae or mobile animals, although
a mild summer may have made our test of history relatively weak.
Hormosira banksii fits the definition of a keystone species or engineer and, as such, is
an appropriate focus for management and as an indicator. Spatially heterogeneous effects
of a constant physical perturbation, however, mean that management of these rocky shores
requires more complex models and indicate that caution should be used in adopting this
species as a uniform indicator of environmental change.
Key words: disturbance, pulse and press; Hormosira;human impact; intertidal algae; press dis-
turbance; pulse disturbance; rocky shores, management of; trampling.
INTRODUCTION and Osenberg 1996). In the terminology of disturbance
(Bender et al. 1984), these events are press distur-
Human activities often constitute a disturbance to
bances. Other human activities are more variable in
natural environments, and in attempting to assess the
impacts of those activities or to develop models that time, and have been termed pulse disturbances. Spills
lead to their management, it is important to determine of toxicants often are short pulses through a system;
how they compare to natural disturbances that may be other inputs of nutrients and toxicants often vary great-
present. ly in association with high rainfall events and associ-
In nearshore coastal environments, anthropogenic ated storm water runoff. Many commercial fisheries are
disturbances vary in nature. Some, such as industrial seasonal, and any impacts of human recreational ac-
discharges, are more or less continuous stresses, and tivities are likely also to show annual cycles, reflecting
there is an extensive literature on their effects and the seasonal changes in visitation to coastal areas. These
methodology for detecting these impacts (e.g., Schmitt latter activities may be a series of pulses of distur-
bances, followed by potential recovery periods.
Many natural disturbances also have spatially vari-
Manuscript received 27 August 1996; revised 20 May
1997; accepted 16 June 1997; final version received 24 July able effects and/or patterns of recovery (e.g., Connell
1997. 1979, Paine and Levin 1981, Dayton et al. 1984, 1992,
141
142 MICHAELJ. KEOUGHAND G. P. QUINN Ecological Applications
Vol. 8, No. I
and see reviews by Sousa 1984, 1985, Connell and METHODS
Keough 1985, Lake 1990), although there are relatively
Our main study areas were extensive rocky limestone
few natural disturbances showing periodicity (but see,
platforms within Mornington Peninsula National Park,
e.g., Bertness and Ellison 1987).
in southeastern Australia. The national park extends
Periodic activities can be viewed as a series of pulse
over -30 km of moderately exposed ocean coastline,
disturbances, and a population or assemblage could re- and is accessible to the general public. A section of 8
spond in a number of ways. If the disturbance is not km was formerly under the control of the Department
too intense or the system has high resilience, it may of Defence, and was incorporated into the national park
be able to recover in the intervals between disturbances in 1989. This section of the park has remained closed
(Petraitis et al. 1989). Recovery may also occur if the to public access, so shores in that area have been pro-
interval between disturbances is long. In the termi- tected from direct human influence for >75 yr. The
nology of recent disturbance theory, repeated pulse dis- dominant intertidal habitat is provided by the fucoid
turbances may produce a series of pulse-like recoveries, alga Hormosira banksii, which forms large monotypic
or act as a press disturbance. Whether a particular dis- beds (Fig. 1). The individual plants have a basal hold-
turbance regime produces pulse or press responses is fast, from which are produced fronds. Fronds are com-
also likely to depend on the intensity of the disturbance: posed of chains of vesicles (Fig. 1). The general de-
low intensity disturbances, i.e., those that cause little scription of the major habitat types in this area is pro-
damage during a given pulse, may allow rapid recov- vided by Povey and Keough (1991). Hormosira beds
ery, but there may be a critical intensity beyond which provide habitat for a range of smaller gastropod mol-
persistent changes occur. lusks and crustaceans. Other mollusks feed primarily
On rocky shorelines of southern Australia, one of in open (i.e., lacking macroalgal cover) areas, and their
the most prominent species is the perennial fucoid alga abundances are negatively correlated with the presence
Hormosira banksii, which forms extensive monotypic of large algae (G. P. Quinn and M. J. Keough, personal
stands at midtidal levels of rock platforms (Fig. 1). observations).
Hormosira beds are habitat for a range of mobile an-
imals, and their presence is negatively associated with Experimental designs
other species. H. banksii is sensitive to short-term tra-
The main trampling experiment ran for 6 yr, and
pling (Povey and Keough 1991) and is deleteriously
consisted of areas of the shore being trampled over
affected by other anthropogenic activities, including
summer, followed by a recovery period from midau-
discharge of sewage (Brown et al. 1990, Fairweather
tumn to early summer. On these shores, there are rel-
1990). We used this alga to test the effects of recurrent
atively low levels of visitation by humans until late
disturbances at a range of intensities, with the distur-
December, when levels become high through January
bances recurring annually over six summers. during a major holiday period, taper off in February,
In particular, do seasonally recurrent disturbances and remain at that level until approximately Easter
caused by humans produce pulse or press responses? (March) when another brief holiday period occurs
Are the responses spatially consistent? Does the kind (King 1992). Our experiment followed those broad pat-
of response to disturbance vary with intensity of dis- terns of use.
turbance? We also compared the (controlled) human We established three trampling sites at hapazardly
disturbances to changes produced by natural events. chosen areas within the protected area of the national
We also considered the possibility that organisms' park. Two sites were on different parts of the intertidal
sensitivity to new disturbances might be related to their platforms at Cheviot Beach; Harry's Pool was adjacent
history. For example, a history of competition may to a large rock pool and Cheviot Mid was -200 m
make an organism more sensitive to physical distur- along the shore, but set back from the edge of the
bance (Peterson and Black 1988) or change some life platform. The third site, Grenade Range, was at a plat-
history parameters later in its life (Scott 1994). Tanner form -1 km away, separated from Cheviot Beach by
et al. (1996) provide an overview of historical effects, two headlands. At each site, we initially established
although their own data did not show a strong effect two plots, --10 X 3 m, separated by 30-50 m, where
of history on community dynamics and structure. A the percentage cover of Hormosira exceeded 90. We
natural or anthropogenic disturbance may stress an or- marked eight trampling strips within each plot, each
ganism, inhibiting its ability to respond to new chal- strip being 50 cm wide, and 2-3 m long, with the exact
lenges, repeated occurrences of the same disturbance length varying among plots, depending on the size of
or novel stresses. Intertidal algae could become the Hormosira patch in which the plot was placed.
stressed by desiccation and become more vulnerable to Strips were separated by at least 1 m and parallel to
trampling, or vice versa. We took advantage of our each other. Each plot included two replicates of each
long-term experiment to contrast the responses of trampling treatment.
plants with a long history of disturbance to those of The main trampling experiment used four intensities
plants with little or no history of trampling. of trampling. Our focus here is on changes occurring
February 1998 EFFECTS OF PERIODIC DISTURBANCES 143
FIG. 1. Hormosirabanksii plants. The top panel shows a false-color infrared image of a small rock platform at Cheviot
Beach. The red area in the center of the platform is dense Hormo.sira,with small patches elsewhere on the platform. The
photograph covers an area -50 m wide. The lower panel shows individual plants that have been subject to moderate trampling,
with some reduction in cover. Note the morphology of the plants. with fronds of spherical vesicles in long chains, and a
combination of intact and damaged chains. The picture covers an area 10 cm wide.
144 MICHAELJ. KEOUGHAND G. P. QUINN Ecological Applications
Vol. 8, No. 1
each summer, so the intensity of disturbance is the num- plots were identical to the existing two at each site,
ber of passages per summer (which is equivalent to the and were within the same large algal mat. We repeated
number/day), as the number of trampling days was con- this procedure at the beginning of the summer of 1994-
sistent across treatments. One passage consisted of a 1995, so at Harry's Pool and Cheviot Mid, we had two
person of average size walking at normal pace along plots that had been trampled for 4 yr, one that had been
a strip, and strips received either 0, 5, 10, or 25 passages trampled for 2 yr, and one with no history of trampling.
on a given low tide. For repeated passages, tramplers All of the new strips were trampled and sampled in the
moved beyond the end of the strip before turning, to manner described above, including a census of the com-
prevent more severe forces associated with pivoting of plete fauna in autumn of 1995.
the feet. All trampling was done by average-sized
adults, wearing rubber-soled athletic shoes or sandals, Analyses
footwear of a similar type to that used by recreational All data were analyzed by analysis of variance. Our
visitors (Povey 1989). Each summer, we used 6-8 d of design for the main experiment involved six factors,
such trampling on every strip, spread haphazardly over and corresponded to a split-plot or repeated-measures
the suitably low tides during summer. We began the design. We use the latter terminology, for clarity. The
trampling in the summer of 1990-1991 and continued spatial component of the design was partly nested, with
through the summer of 1995-1996. The experiment has Sites, and Plots within Sites. Both factors were crossed
continued, and in this paper we present analyses of five with Trampling, with two strips within each Plot-Tram-
years of data, plus descriptions of events occurring dur- pling combination. Each strip was then sampled 10
ing the sixth summer (1995-1996). times, with those 10 samples falling into 5 yr, and
We sampled the experiment twice annually, at the before/after each summer. In repeated-measures ter-
beginning of summer, and again after Easter, after the minology, the "subjects" were strips, the between-sub-
last trampling period. Easter holidays represent the last jects factors were Site, Plot, and Trampling, and the
major recreational influx to coastal areas before winter. within-subjects factors were Years and Before-After
The measurement after Easter was intended to assess Summer. Trampling, Before-After Summer, and Years
the effects of trampling, while the measurement in early were fixed factors, the latter because we had sampled
summer provided a measure of recovery over the pre- for all five years in the period. Strips, Plots, and Sites
ceding seasons. Each strip was sampled using a 70 x were random factors, the latter to allow us to generalize
35 cm quadrat placed in the center of each strip, with about spatial variation in responses to disturbance.
its long axis running parallel to the strip. We photo- The detailed censuses at the end of 3 and 5 yr were
graphed the quadrat with color slide film (first 3 yr) or each analyzed by partly nested analysis of variance,
Hi-8 video (second 3 yr), and back in the laboratory, i.e., the above design with no within-subjects factors
we projected the images and calculated the cover of (Plots within Sites, crossed with Trampling, and two
Hormosira using 100 randomly placed dots superim- replicates). We analyzed the abundance of all common
posed on the image. taxa. Coralline algae can not be identified to species
On two occasions, after three and five summers (i.e., in the field, although they are numerically dominated
in autumn of 1993 and 1995) we counted understory by species of Corallina, and we separated them into
algae and mobile animals. We did this using two quad- articulate and encrusting forms because of the different
rats, placed end-to-end in the center of each strip. Algae ecological properties of those growth forms (Steneck
were estimated using a grid of 100 points, and we and Dethier 1994). Uncalcified turfing or encrusting
counted all algae beneath each point. Animals were algae were also abundant enough to analyze, even
identified and counted, after a thorough search of each though individual species, such as Cladophora and
quadrat. We used two quadrats because some animals Ralfsia, could not be analyzed. We pooled all algae
were uncommon, and we required a larger sample, but other than Hormosira to create a further plant variable.
numbers were averaged to provide a single value for The animals were dominated by mollusks, and herbi-
each strip. We did not do complete censuses often, vores in particular. The predatory whelk Thais orbita
because we considered the sampling procedure to be and the scavenging Cominella lineolata were present,
potentially disruptive to understory algae or associated but not sufficiently abundant for analysis. We analyzed
invertebrates. abundance of the two true limpets Cellana tramoserica
Historical effects.-After 2 yr, we tested whether and Patelloida alticostata, the two pulmonate limpets
plants with a history of disturbance and recovery might Siphonaria diemenensis and S. zelandica, plus the lit-
be more resilient or more sensitive to a new distur- torinid Bembicium nanum. We created two additional
bance, by adding a new plot to the Harry's Pool and pooled herbivore groups, (true) limpets (Cellana + Pa-
Cheviot Mid sites. At both sites, Hormosira had re- telloida alticostata + P. latistrigata) and nonlimpet
covered by the beginning of summer. At Grenade grazers (Siphonaria spp., Bembicium, plus Austro-
Range, Hormosira had declined in treatment plots (see cochlea constricta, Turbo undulata). We also assessed
Results), so we could not compare the response of these the performance of various "community" statistics; we
plants to that of previously untrampled areas. The new calculated the taxonomic richness (number of recog-
February 1998 EFFECTS OF PERIODIC DISTURBANCES 145
TABLE 1. Analysis of changes in cover of Hormosira banksii through time, as a function of level of trampling.
Denom.
Source of variation df no.t MS F P P (Chev)
Between-strips (i.e., pooled across time) effects
1. Sites (S) 2 2 2255.4 4.26 0.133 0.815
2. Plots within Sites (P{S}) 3 6 529.2 0.44 0.725
3. Trampling (T) 3 6 15778.5 13.19 0.000
4. Trampling X Site 6 5 4186.9 15.67 0.000 0.087
5. T X P{S} 9 6 267.1 0.22 0.988
6. Strips within Plots (Residual) 24 1196.5
Within-strips (i.e., temporal) effects
7. Years (Y) 4 13 17053.7 84.67 0.000
8. Year X Site 8 9 1804.9 3.91 0.017 0.302
9. Years X P{S} 12 13 462.1 2.29 0.013
10. Year X Trampling 12 13 284.4 1.41 0.201
11. Y X T X S 24 11 450.3 2.25 0.013 0.458
12. Y X T X P{S} 36 13 200.0 0.99 0.493
13. Strips X Years (Residual) 96 201.4
14. Before-After Summer (BA) 1 20 32472.3 394.82 0.000
15. BA X Sites 2 16 124.8 0.13 0.880 0.897
16. BA X P{S} 3 20 936.6 11.39 0.000
17. BA X Trampling 3 20 826.2 10.05 0.000
18. BA X S X T 6 19 107.8 2.35 0.120 0.279
19. BA X T X P{S} 9 20 45.9 0.56 0.817
20. Strips X BA Residual 24 82.2
21. Years X Before-After Summer 4 27 2919.5 38.17 0.000
22. Y x BA x S 8 23 534.5 0.90 0.546 0.383
23. T X BA X P{S} 12 27 594.4 7.77 0.000
24. Y X BA X Trampling 12 27 516.4 6.75 0.000
25. Y x BA X T X S 24 26 68.3 0.79 0.728 0.336
26. Y X BA x T X P{S} 36 27 86.8 1.14 0.308
27. Y X BA X Strips Residual 96 76.5
Notes: Significant effects (at (- = 0.05) are shown in boldface. The right-most column shows the P values associated with
tests for heterogeneous effects of trampling at the two sites on the Cheviot Beach Platform.
t Terms are numbered, and the denominators used to test each effect are indicated using those numbers.
nizable taxa), species richness of gastropods, total Illinois). For all analyses, we examined primarily the
number of individuals, and the Shannon-Wiener di- assumption of normality, by examining residuals by
versity index (H'). probability plots. There were generally too few repli-
The experiment to examine history of disturbance cates at a given level of the design for a meaningful
was analyzed using data collected at the end of the comparison of variances. When we used repeated-mea-
1994-1995 summer. In the analysis, we found no sig- sures or partly hierarchical analyses, we also examined
nificant variation between the two Cheviot sites (at ax the more conservative Greenhouse-Geiser and Huynh-
= 0.25), so we omitted them from the analysis, to give Feldt corrected F tests, which provide some protection
four levels of trampling, and three levels of history (0, against violations of assumptions of compound sym-
2, or 4 yr) with either four, two, or two plots within metry. Those tests did not produce results markedly
each level of history, respectively. The data were an- different from the uncorrected ones, and we saw no
alyzed as a partly nested analysis, with History, evidence of strong violation of these assumptions, so
Plots{History}, Trampling, H X T, and T X Plots as only the standard F tests are presented here.
the terms in the analysis, and History and Trampling Note, in all analysis tables, probabilities are rounded
as fixed factors and Plots as a random factor. Our con- to three decimal places for brevity; values <0.0005 are
clusions would not be altered by the more conservative shown as 0.000.
step of retaining sites within the analysis. We analyzed
Hormosira cover and abundances of all common taxa. RESULTS
To be more confident of detecting an effect on Hor-
Effects on Hormosira banksii
mosira, we analyzed the percentage covers from the
postsummer 1994-1995 survey, and data from pre- and We found striking temporal and spatial variation in
postsummer of 1995-1996, so we could examine the the percentage cover of H. banksii, and find it helpful
profile through time of plots with different histories. to separate effects involving trampling from those that
The three values were treated as repeated measures, presumably represent natural variation.
using the statistical model described above. Effects of trampling.-Trampling affected Hormo-
All data analysis was done using SYSTAT for Win- sira beds dramatically, with the effects varying through
dows, version 5.03 (SYSTAT Incorporated, Evanston, time and through space (Table 1). Individual plants
146 MICHAELJ. KEOUGHAND G. P. QUINN EcologicalApplications
Vol. 8, No. 1
Cheviot Platform Sites
100
0 40
U
20 erk| |None 5 10 25|
eror
Grenade Range Site
100
N0 -V
60 -~ P-
0
40t40
'-
- errors |.
20 .
o~~~~~~~~~~~
1991 1992 1993 1994 1995 1996
FIG. 2. Changes in cover of Hormosira banksii on two platforms, under different levels of trampling. Data were pooled
from the two sites at Cheviot Beach, as they showed no significant heterogeneity. The bar at the top of the figure indicates
periods during which trampling occurred (as dark blocks) and times when there was no disturbance (clear blocks). The error
bars at the base of each figure indicate three standard errors, calculated using the variance components for error terms used
to test Trampling X time effects, using data from Table 1. The left error bar is the geometric mean of the time X Strips
residuals from Table 1, the middle error represents the Trampling X time X Plots term (based on three components: Year X
T X P, BA X T X P and Y X BA X T X P). The right error bar indicates variance for assessing variation that is independent
of trampling, i.e., time X Plots terms.
were initially damaged by the loss of chains of vesicles, cover through to late 1994. Plants in control strips cov-
and ultimately by whole fronds, as described by Povey ered 80-100% of space until late 1994, while the major
and Keough (1991). Under severe damage, they were changes occurred in the most heavily disturbed areas,
reduced to holdfasts (Fig. 1). The effects of trampling where cover declined after each period of trampling,
were quite different at the three sites (see Trampling with little or no recovery in the intervening seasons,
x Site, Year x Site X Trampling interactions on Table so that by late 1994, cover had fallen to <10%. In the
1). At Grenade Range, there was a decline in percentage austral summer of 1994-1995, there was a major de-
cover after the first summer's trampling, and the rate cline across all treatments, with cover falling by -30%
of decline was proportional to the intensity of trampling in controls and the two intermediate levels of distur-
(Fig. 2). The plants recovered by the beginning of the bance, and falling to a few percent in the most heavily
following summer, but then declined even more under disturbed areas (Fig. 2). After this decline, there was
the second year's trampling. There was little subse- some recovery in the control and intermediate treat-
quent recovery, and in the third and fourth years plants ments, although cover did not return to its levels of the
in the two intermediate treatments remained at 60-70% spring of 1995, and there was little recovery in the
February 1998 EFFECTS OF PERIODIC DISTURBANCES 147
heavily disturbed areas. Trampling in the summer of cant Site x Trampling interactions for articulate cor-
1995-1996 had little apparent effect (Fig. 2). alline algae, total algae, Cellana tramoserica, Siphon-
The situation was very different at the two Cheviot aria diemenensis, and the pooled categories of limpets
Beach sites. The effects of trampling did not differ and nonlimpet grazers. When the data for the two Chev-
significantly between the two sites (Table 1). We there- iot platforms were analyzed, the only significant effect
fore discuss the two sites together. For the first two of trampling was a Plots x Trampling interaction for
years, the Cheviot sites followed a trajectory similar Patelloida alticostata (Table 2). For Cellana and S.
to that shown by Grenade Range, with an initial de- diemenensis, there was no consistent relationship be-
cline, a complete recovery, then a more severe decline tween intensity of trampling and abundance at the
during the disturbances of the second year (Fig. 2). Cheviot sites, but at Grenade Range, a 2-4-fold in-
From that stage, however, all plots recovered com- crease in abundance at the highest intensity of tram-
pletely, and for the next 3 yr, we saw little effect of pling (Fig. 3). There were similar patterns for Patel-
trampling. There was a marked decline in cover in the loida and S. zelandica, although they were more vari-
summer of 1994-1995, as at Grenade Range, but this able, and the analyses did not show significant effects
decline was consistent across all trampling treatments. after 5 yr. When the data were pooled, the nonlimpet
There had been complete recovery by midspring of grazers and limpets showed strong relationships with
1995, and trampling had little effect in that summer, trampling at Grenade Range, but again, no apparent
with an increase in cover in three treatments, and a pattern at the two Cheviot Beach sites.
decline only under heavy trampling. Articulate corallines varied dramatically in abun-
The variation that we observed in the effects of tram- dance between sites, and responded variably to tram-
pling was almost all at larger spatial scales; we were pling. They formed a major part of the understory at
able to test whether replicate plots at each site showed Cheviot Mid, with a mean cover of -50%, yet covered
the same effect of trampling, and all four tests incor- no more than 10% at the other two sites (Fig. 3). At
porating a Trampling x Plot interaction were nonsig- the two Cheviot sites, there was no relationship with
nificant (Table 1). Similarly, the variation among rep- trampling, but at Grenade Range, their cover declined
licate strips within plots was quite small; the four terms with trampling from -8% in control areas to 0 in the
incorporating Strips on Table 1 together accounted for most heavily trampled treatments (Fig. 3). There was
only 18% of the total sum of squares in the analysis no pattern for encrusting corallines or for turfing algae,
(compared to -35% for trampling effects). and the patterns for total algae reflected the hetero-
Temporal variation independent of trampling.-The geneous results of individual taxa.
pattern of variation was different when we examined At the time of the first census, after three summers
the effects that were unrelated to trampling. There was of trampling, the only species to show consistent effects
strong variation in percentage cover of Hormosira of trampling was the limpet Cellana tramoserica (Table
among Years and Before-After Summer (Table 1), as 3). There were isolated small-scale effects of trampling
well as small-scale variation in cover. Changes over on another limpet, Patelloida alticostata, and the cover
summer reflect burning-off of the algae, with a con- of articulated coralline algae, both of which showed
sequent loss of biomass and/or cover, and these effects significant Trampling x Plot variation (Table 3). There
varied among years (Years x BA interaction; Table 1). was also a significant effect of trampling on all limpets
This pattern was not significantly heterogeneous among pooled. Almost all common taxa varied significantly
the three sites. Cover declined over the summers of among plots, but not among sites (Table 3). When the
1990-1991 and 1994-1995, and showed either weak two Cheviot sites only were compared, the results were
changes (1991-1992) or no discernible change (1992- similar except that there was no effect of trampling on
1993, 1993-1994, 1995-1996) in the other summers the abundance of Patelloida, and significant Plot X
(Fig. 2). The variation among years was not consistent Trampling effects on the abundance of the pulmonate
among the three sites, but this effect was small, and limpet Siphonaria diemenensis, Cellana tramoserica,
not clear from the graphs (Fig. 2). and the pooled categories of nonlimpet grazers and
In contrast to the trampling effects, there was small- limpets (Table 3).
scale variation in the effects of Year and Before-After The derived variables showed relatively weak ef-
Summer, with significant interactions between Plots fects. There was no significant effect of trampling on
and these two factors. taxonomic richness, gastropod species richness, or H',
regardless of whether all sites or just the Cheviot plat-
Effects on other organisms forms were compared (Table 3). The number of indi-
Trampling affected other organisms, but they were viduals was affected by trampling, with the effect vary-
generally heterogeneous among the three sites. After 5 ing between sites (Table 3). For this variable, residual
yr, the effects of trampling varied significantly among plots did not completely support the assumptions of
sites, rather than among plots (Table 2), with the sites the analysis for raw or log-transformed data, but were
effect generated by the difference between Grenade intermediate. However, the effect of trampling was con-
Range and the two Cheviot Sites. There were signifi- sistent for raw and log-transformed data, and disap-
148 MICHAEL J. KEOUGH AND G. P. QUINN EcologicalApplications
Vol. 8, No. 1
TABLE2. Analyses of the abundance of major animals and plants after 5 yr of trampling, highlighting effects of trampling
at all three sites and at only the two sites on Cheviot Beach platform.
df df Articulate Encrust.
Trampling (num.) (denom.) corallines corallines All corallines Turf All algae
Overall (all sites)
Sites (S) 2 3 0.343 0.786 0.359 0.584 0.382
Trampling (T) 3 6 0.388 0.499 0.624 0.130 0.615
S X T 6 9 0.018 0.916 0.283 0.565 0.013
Plots within Sites 3 24 0.000 0.244 0.000 0.000 0.000
T X Plots 9 24 0.997 0.105 0.397 0.272 0.870
MS Residual 24 38.8 0.02 66.9 0.01 72.8
R2 0.98 0.53 0.96 0.82 0.96
Cheviot only
Sites 1 2 0.362 0.821 0.354 0.745 0.359
Trampling 3 2 0.343 0.807 0.436 0.647 0.532
S X T 3 6 0.143 0.190 0.169 0.138 0.113
Plots within Sites 2 16 0.000 0.266 0.000 0.000 0.000
T X Plots 6 16 0.983 0.765 0.912 0.749 0.893
MS Residual 16 52.0 5.2 49.8 45.1 43.4
R2 0.98 0.41 0.98 0.75 0.98
Historical effects
History 2 5 0.665 0.388 0.464 0.689 0.645
Trampling 3 15 0.690 0.882 0.195 0.914 0.852
History X Trampling 6 15 0.185 0.031 0.379 0.312 0.468
Plots within Histories 5 32 0.000 0.289 0.015 0.000 0.000
Trampling x Plots 15 32 0.781 0.930 0.976 0.752 0.535
MS Residual 32 56.4 13.5 144.9 70.6 58.7
RI2 0.97 0.45 0.51 0.96 0.97
Notes: The table also shows the results of analyses to assess the effects of a history of trampling for Cheviot platform.
For each taxon, the table shows the probabilities from the ANOVA associated with tests of hypotheses, plus the residual MS
and the variance explained by the model and the degrees of freedom associated with numerator and denominator for each F
ratio. Combining the degrees of freedom, P values, and MS Residual allows reconstruction of the complete analysis table.
All tests of significance were done at a = 0.05. Significant effects are shown in bold.
peared when only the Cheviot sites were compared, ables also showed strong variation among plots that
indicating that the primary difference was, again, be- was unrelated to the levels of trampling.
tween Grenade Range and the two Cheviot sites, with Because a particular level of trampling produced a
the number of individuals rising strongly at Grenade different cover of Hormosira at different sites, it is
Range as the intensity of trampling increased. All vari- probably not surprising that most of the effects on other
TABLE 3. Analyses of the abundance of major animals and plants after three summers of trampling, highlighting effects of
trampling at all three sites and at only the two sites on Cheviot Beach platform.
df df Articulate Encrust.
Trampling (num.) (denom.) corallines corallines All corallines Turf All algae
Overall (all sites)
Sites (S) 2 3 0.415 0.092 0.427 0.184 0.464
Trampling (T) 3 6 0.980 0.243 0.990 0.236 0.668
S XT 6 9 0.510 0.456 0.402 0.539 0.464
Plots within Sites 3 24 0.000 0.516 0.000 0.012 0.000
T X Plots 9 24 0.030 0.931 0.061 0.694 0.355
MS Residual 24 15.4 0.005 17.1 0.007 37.6
R2 0.99 0.45 0.99 0.70 0.98
Cheviot only
Sites 1 2 0.412 0.292 0.415 0.869 0.432
Trampling 3 2 0.975 0.720 0.981 0.470 0.899
S XT 3 6 0.387 0.700 0.329 0.298 0.387
Plots within Sites 2 16 0.000 0.068 0.000 0.017 0.000
T x Plots 6 16 0.057 0.399 0.065 0.716 0.423
MS Residual 16 22.8 0.0 22.1 0.0 37.1
R2 0.99 0.58 0.99 0.56 0.99
Notes: For each taxon, the table shows the probabilities from the ANOVA associated with tests of hypotheses, plus the
residual MS and the variance explained by the model and the degrees of freedom associated with numerator and denominator
for each F ratio. Combining the degrees of freedom, P values, and MS Residual allows reconstruction of the complete analysis
table. All tests of significance were done at a = 0.05. Significant effects are shown in bold.
February 1998 EFFECTS OF PERIODIC DISTURBANCES 149
TABLE 2. Extended.
Cellana Patelloida Siphonaria Siphonaria Bembicium Nonlimpet
tramoserica alticostata diemenensis zelandica nanum grazers Limpets
0.705 0.839 0.692 0.153 0.604 0.494 0.747
0.532 0.519 0.606 0.469 0.543 0.562 0.508
0.005 0.191 0.001 0.086 0.092 0.001 0.006
0.000 0.041 0.000 0.098 0.000 0.000 0.000
0.579 0.471 0.913 0.889 0.343 0.950 0.645
3.07 2.15 1355 489 0.30 6.95
0.82 0.60 0.80 0.62 0.88 0.76 0.80
0.766 0.942 0.721 0.542 0.929 0.693 0.814
0.987 0.265 0.966 0.274 0.528 0.955 0.788
0.166 0.871 0.198 0.736 0.128 0.151 0.350
0.000 0.001 0.000 0.000 0.000 0.000 0.000
0.538 0.048 0.465 0.957 0.150 0.497 0.332
2.5 0.85 718 29.0 0.24 742 3.6
0.80 0.75 0.87 0.69 0.90 0.89 0.84
0.581 0.304 0.972 0.930 0.572 0.974 0.507
0.323 0.606 0.922 0.499 0.855 0.932 0.247
0.743 0.865 0.797 0.157 0.463 0.702 0.732
0.000 0.000 0.000 0.000 0.000 0.000 0.000
0.384 0.037 0.574 0.931 0.345 0.546 0.359
2.23 0.57 1535 29.0 0.45 1674 3.29
0.79 0.76 0.71 0.67 0.84 0.74 0.82
organisms were manifest as a Site x Trampling or Plot and Plots as random effects). We found only one sig-
X Trampling interaction. Our conceptual model for nificant interaction between Hormosira cover and Site,
these shores is that Hormosira is the habitat-forming for Cellana tramoserica, and no significant Plot X Hor-
species, and that the abundances of other organisms mosira interaction, suggesting that the trampling ef-
may track changes in Hormosira. To take account of fects may have been indirect responses to changes in
variation in cover of Hormosira, we reanalyzed the data Hormosira.
in the upper part of Table 2 (i.e., all three sites) as a There were significant effects of Hormosira for ar-
nested analysis of covariance (Sites, Plots within Sites, ticulated corallines, all coralline and all algae, Patel-
and cover of Hormosira as the covariate, with Sites loida, Siphonaria diemenensis, and the nonlimpet graz-
TABLE 3. Extended.
Cellana Patelloida Siphonaria Siphonaria Bembicium Nonlimpet
tramoserica alticostata diemenensis zelandica nanum grazers Limpets
0.735 0.527 0.670 0.638 0.102 0.775 0.639
0.027 0.313 0.181 0.182 0.253 0.152 0.044
0.894 0.430 0.355 0.684 0.396 0.388 0.805
0.062 0.004 0.000 0.000 0.119 0.000 0.006
0.570 0.017 0.094 0.913 0.099 0.124 0.187
0.932 0.276 64.0 0.781 0.100 70.7 1.604
0.52 0.78 0.78 0.64 0.75 0.77 0.68
0.720 0.669 0.750 0.530 0.147 0.820 0.696
0.029 0.562 0.377 0.547 0.509 0.197 0.047
0.917 0.249 0.236 0.467 0.272 0.520 0.862
0.003 0.008 0.000 0.002 0.095 0.000 0.000
0.035 0.744 0.007 0.921 0.116 0.027 0.029
0.188 0.172 5.37 1.00 0.119 11.0 0.266
0.79 0.60 0.93 0.63 0.77 0.90 0.85
150 J. AND G. P. QUINN
MICHAEL KEOUGH Ecological Applications
Vol. 8, No. 1
Grenade Harry'sPool Cheviot Mid
l.l i I
i-. 0................_-
60 Artic. coral. 0.3 Enc. coral.
0 ~~~~~~0
U 0 U 0.
L
40 ~ Cellana 20 Patelloida
0-**- ..*..
0 C
800 Siph. diem. 240 Siph. z.
4 Bembicium 0.5 - Turf algae
.0 ~ ~ ~ ~ .
-0 ~~ ~
0< 0 5 X,
0 5 10 25 0 5 10 25
Level of Trampling
FIG. 3. Abundance of algae and herbivorous snails after 5 yr of trampling. The organisms are articulate coralline algae,
encrusting coralline algae, the limpets Cellana tramoserica and Patelloida alticostata, the pulmonate limpets Siphonaria
diemenensis and S. zelandica, the littorinid snail Bembicium nanum, and a group of foliose algae (Turf). Each panel on the
graph shows the mean abundance, as percentage cover or number per square meter (N). Three standard error bars are shown
at the lower left corner of each panel. The left bar indicates the variation among replicate strips within plots, the middle bar
indicates the Plots{Site} x Trampling variation, while the third indicates the Plots{Sites} variation. The middle bar is the
most important for interpreting the figures, as it represents the error term used to assess the Sites x Trampling interaction.
Note that encrusting coralline algae and turf algae are plotted as arcsine transformed values, and Bembicium data are shown
as square-root transformed values. These scales represent the data used in the analyses.
February 1998 EFFECTS OF PERIODIC DISTURBANCES 151
TABLE 4. Regressions of the abundance of individual taxa on percentage cover of Hormosira for all sites, Grenade Range
only, and the Cheviot sites. Data were individual trampling strips, and sample sizes were 48 (all sites), 16, and 32. The
direction of the relationship, i.e., the sign of the regression slope, is also shown.
Direction All sites Grenade Range Cheviot sites
Taxon +/- r2 p r2 p r2 p
Articulated coralline algae + 16 0.000 73 0.000 18 0.001
Encrusting corallines + 4 0.068 4 0.461 6 0.052
All corallines + 20 0.000 25 0.047 21 0.000
Turf algae - 6 0.031 11 0.202 13 0.003
All algae + 22 0.000 41 0.007 22 0.000
Cellana tramoserica - 35 0.000 56 0.001 23 0.000
Patelloida alticostata - 21 0.000 31 0.024 12 0.006
Siphonaria diemenensis - 50 0.000 61 0.000 43 0.000
Siphonaria zelandica - 18 0.000 30 0.029 10 0.010
Bembicium nanum - 14 0.001 9 0.269 27 0.000
Nonlimpet grazers - 50 0.000 60 0.000 43 0.000
Limpets - 35 0.000 52 0.002 24 0.000
Austrocochlea constricta + 2 0.266 14 0.150 1 0.584
Derived variables
Taxonomic richness -0+t 2 0.343 <1 0.865 13 0.040
Gastropod richness - 3 0.237 12 0.189 2 0.418
Total individuals - 57 0.000 60 0.000 52 0.000
Log (individuals) - 64 0.000 73 0.000 65 0.000
H' 2 0.381 20 0.082 15 0.026
Note: Significant effects are shown in bold.
t Overall negative effect, weak positive at Grenade Range, negative at Cheviot sites.
t Overall positive, negative at GR, positive at Cheviot.
ers and limpets. Simple regressions of the abundance summer of 1995-1996, and postsummer 1994-1995),
of each of these taxa against Hormosira cover at the and considered the temporal profiles using a repeated-
time of the census were used to indicate the direction measures analysis. Over that time period, we did not
of the effects. These regressions were significant for find even a significant effect of trampling (see Table
most taxa, although the strong relationships were with 5), with only a marginally nonsignificant main effect
cover of articulate coralline algae, Cellana, Patelloida, of trampling after summer of 1995-1996 even hinting
Siphonaria diemenensis, and the composite grazing cat- at a change. There was significant variation through
egories (Table 4). We also analyzed Grenade Range time and among plots, but all effects involving history
and the two Cheviot sites separately, and relationships (i.e., the main effect, plus the interactions with time
were generally stronger at Grenade Range than Cheviot and trampling) were far from significant (Table 5). The
(Table 4; paired t test using r2 values for each taxon, lack of a significant effect did not appear to be a result
t = 3.12, df = 12, P = 0.009). At Grenade Range, of low power; rather, plots with the three different his-
there were particularly strong effects of Hormosira tories of trampling showed very similar patterns (Fig.
cover for articulate coralline algae, Cellana and Si- 5).
phonaria diemenensis. Cover of both kinds of coralline The censuses of other plants and animals after 5 yr
algae (and the pooled variables total coralline and total also showed little effect of history. There were no sig-
algae) varied positively with Hormosira cover, as did nificant simple effects of history, and only a single
the abundance of the herbivorous snail Austrocochlea interaction with trampling, for encrusting coralline al-
constricta, whereas the slopes of the regressions were gae (Table 2). Main effects of history would be difficult
negative for turfing algae and all other herbivorous to identify, as we found substantial variation among
snails (Table 4). replicate plots, the level of variation used for assessing
Trends in the derived variables also became clearer this main effect. The test of the History x Trampling
with the analysis of covariance. The Site x Trampling interaction, however, had 6 and 15 degrees of freedom,
effects on the number of individuals disappeared, and and there was little variation among plots in the effect
there was a very strong overall effect of Hormosira of trampling. The Plots X Trampling interaction was
cover (Fig. 4). There was no effect of algal cover on used to test the History x Trampling term. The mean
either taxonomic or gastropod richness, and, interest- abundances of the major taxa showed no hint of an
ingly, a positive effect of Hormosira on H' at the Chev- effect, with the plots with no history having the highest
iot sites and a (nonsignificant) negative trend at Gre- means for some taxa, while plots with the longest his-
nade Range (Table 4). tory of trampling were greatest for other taxa (Fig. 6).
Even the one taxon showing a significant effect, the
Historical effects of trampling encrusting coralline algae, did not have a pattern of
We found few strong effects of history. We analyzed means that could be easily interpreted with respect to
the cover of Hormosira at three times (pre- and post- the effects of history or trampling (Fig. 6).
152 MICHAEL J. KEOUGH AND G. P. QUINN Ecological
Applications
Vol. 8, No. 1
15 15
ww A AU
*AAA AAL AA
10 AA A A 10
U A H A HA
LI A n o A A
L \AI *
AA A
I A
5 5 * AA AA nA n
n no A A L n A
A LI ~ UA** n
0 III0 II
0 20 40 60 80 100 0 20 40 60 80 100
4.0 3.0
AU
3.0 L A A A A LI
AA A 2.0 ~ ALI
A ~~~~~~~~~~~~AAU
2.0 A , A on
A AA ALIn
IiEALI
AA1.0A
1.0 LI*A A OLA A
LI A n A
A nL
AA
0.0 -----J0.0
0 20 40 60 80 100 0 20 40 60 80 100
Hormosira cover (%)
FIG. 4. Variation in derived variables listed on Table 6, plotted as a function of Hormosira cover. On each panel, the
three symbols indicate main sites (squares are Grenade Range, and triangles and stars represent the two Cheviot Platform
sites). Filled and unfilled symbols identify the two plots at each site, and each point is a replicate strip.
TABLE 5. Effects of history of trampling on Hormosira banksii. The table shows the results of analysis of variance of the
percentage cover of H. banksii at each of three times (upper panel), together with a repeated measures analysis examining
all time simultaneously (lower panel).
All data
df df Postsummer Presummer Postsummer (repeated
Source of variation (num.) (denom.) 1994-1995 1995-1996 1995-1996 measures)
History 2 5 0.791 0.848 0.871 0.849
Trampling 3 15 0.583 0.659 0.062 0.273
History x Trampling 6 15 0.795 0.589 0.758 0.565
Plots within Histories 5 32 0.000 0.325 0.094 0.015
Trampling x Plots 15 32 0.771 0.930 0.763 0.912
MS Residual 32 405.2 337.2 243.3 774.3
r2 0.61 0.34 0.48
Repeated measures analysis: within-strips effects
Time 2 10 0.016
Time x History 4 10 0.917
Time x Trampling 6 30 0.536
Time x History x Trampling 12 30 1.000
Time x Plots within Histories 10 0.000
Time X Trampling X Plots 30 0.459
Time x MS Residual 64 105.6
r2 ~~~~~~~~~~~~~~~~~~
Notes: Probabilities are shown, with significant effects (at 0.05) shown in bold. The multiple r2 is shown for each
a
analysis, and for the repeated measures, the r2was calculated by treating the two residual (MS Residual and Strips MS Residual)
terms as unexplained variation.
February 1998 EFFECTS OF PERIODIC DISTURBANCES 153
* - * o -o *................. sulted in a rapid decline in the dominant alga and little
4 yr 2 yr No history recovery, a press response. At all sites, there was a
relationship between change in algal cover and inten-
100 sity of disturbance, but nowhere did increasing inten-
sity of disturbance cause a shift from a pulse to a press
response. There was no hint of any "intermediate"
effect of disturbance (Connell 1978, Lubchenco 1978);
trampling at intermediate levels produced results that
were intermediate, and there was no hint of a peak in
EndSummer 1995-1996 measures of diversity at intermediate levels.
0 1S I Why were the effects different at the three sites? We
found strong variation among sites in the effect of tram-
pling, but little variation in the effects of trampling at
100 the level of plots, suggesting that variation in responses
to disturbance is determined by habitat variation at
larger spatial scales. All three sites and all plots were
on horizontal sections of rock platform, and were at
approximately the same height on the shore. Our im-
pression was that the Grenade Range site was slightly
V Pre-Summer1995-1996 more exposed to wave action, but we expected any
0 I--SE
difference to reduce the levels of desiccation, and make
the algae more, rather than less resilient. It is difficult
to identify a cause of this variation; at the largest spatial
100
scale, we had only three replicate sites, and to correlate
EndSummer1994-1995 the sensitivity of a site with environmental variables
(e.g., coastal geomorphology, height, orientation, and
other factors) would require data from many more sites.
Our study was designed to assess variation among sites
in general, rather than focusing on these three precise
0 I I I locations (hence sites as a random factor in the anal-
yses).
Another primary aim of our study was to identify
0 5 10 25 "thresholds" of disturbance, levels of trampling that
Trampling Intensity could be sustained, but beyond which recovery was
incomplete. Again, we did not reach a simple conclu-
FIG. 5. Effects of history of trampling on the percentage sion, but instead have shown increased complexity in
cover of Hormosira banksii. Each panel shows the mean cover the relationship with intensity of disturbance (level of
of H. banksii, pooled over plots at the two Cheviot platform
sites, against the level of trampling, and data are shown for trampling). At the two Cheviot sites, there was a re-
three censuses. Error bars are shown in the corner of each lationship between the intensity of disturbance during
panel, calculated as described in Fig. 3. those summers when trampling had a strong effect. In
other years, we found little effect of trampling, and
hence no relationship between intensity and change in
The community variables also showed no effect of algal abundances. In contrast, there was clear evidence
history, whether we examined taxonomic or gastropod of a nonlinear relationship between intensity of dis-
species richness, total numbers of animals, or diversity turbance and algal cover at Grenade Range: intense
of the animals (Table 6). disturbances produced a very large change, but the two
DISCUSSION intermediate levels (5 and 10 passages/day) produced
very similar patterns of algal abundance, and resulted
Anthropogenic disturbances and Hormosira in Hormosira cover being intermediate between un-
For this intertidal system, a series of repeated pulse disturbed and heavily disturbed areas.
disturbances did not consistently produce either a pulse
or press response. At our two Cheviot sites, each dis- Were the effects of trampling on other species
turbance pulse produced a response, but recovery gen- direct or indirect?
erally occurred over the following 8-9 mo, so we saw The changes in the abundances of other organisms
essentially a series of pulse responses. After six sum- seemed not to happen at the same time as changes in
mers of trampling, we found no strong effect of re- Hormosira. After three summers, few species showed
peated disturbances, and certainly no cumulative effect. strong effects, even though there had been considerable
In contrast, the series of pulses at Grenade Range re- changes in Hormosira cover by that time. These species
154 MICHAEL J. KEOUGH AND G. P. QUINN Applications
Ecological
Vol. 8, No. 1
No history 2 4
40 Artic. coral. 0.3 Enc. coral
0 0
0 0.0
6 6
*1 | | Patelloida
Cellana
0 ~~~~~~~~~~0
400 200 -
'I Siph. diem. Siph. z.
0 0
4 I Bembicium || Turfalgae
0 1,l 110 ESnlf
0 5 10 25 0 5 10 25
Level of Trampling
FIG. 6. Effects of history of trampling on the abundance of algae and herbivorous snails. The species are described in
the caption to Fig. 3. Each panel shows the mean abundance of the taxon for four levels of trampling and three histories.
The error bars are the first two bars described in Fig. 3.
may have been directly affected by trampling, but just peated their experiments in areas lacking macroalgae,
responded more slowly (and positively in some cases!), and found little direct effect on gastropod abundances,
but our results for herbivorous snails are consistent although their test may not have been sensitive because
with short-term trampling experiments on the platforms of the likelihood of movement of individuals between
at Cheviot Beach. Povey and Keough (1991) found that experimental and surrounding areas. Supporting evi-
increases in the densities of herbivorous snails occurred dence is provided by our analyses of the 1995 census
some time after the cessation of trampling in Hormo- in which we replaced the level of trampling with the
sira mats, suggesting a response to the removal of mac- cover of Hormosira, and found that the interactions
roalgae, rather than a direct effect of feet. They re- between trampling and space (sites, plots) disappeared.
February 1998 EFFECTS OF PERIODIC DISTURBANCES 155
TABLE 6. Analyses of derived variables after five years of trampling, highlighting effects of trampling at all three sites and
at only the two sites on Cheviot Beach platform. The variables were taxonomic richness (S), species richness of gastropods
(Sgast), numbers of animals (raw and log-transformed), and Shannon-Wiener diversity (H').
df df
Trampling (num.) (denom.) S Sgast N Log N H'
Overall (all sites)
Sites (S) 2 3 0.719 0.492 0.508 0.678 0.883
Trampling (T) 3 6 0.730 0.872 0.555 0.785 0.814
S x T 6 9 0.071 0.331 0.001 0.019 0.507
Plots within Sites 3 24 0.003 0.027 0.000 0.000 0.000
T x Plots 9 24 0.873 0.826 0.946 0.913 0.688
MS Residual 24 2.063 1.438 2956 0.112 0.262
R2 0.6 0.5 0.8 0.7 0.7
Cheviot only
Sites 1 2 0.698 0.450 0.703 0.839 0.721
Trampling 3 2 0.387 0.175 0.948 0.545 0.990
S x T 3 6 0.078 0.746 0.162 0.367 0.498
Plots within Sites 2 16 0.000 0.008 0.000 0.000 0.000
T X Plots 6 16 0.934 0.555 0.486 0.719 0.460
MS Residual 16 1.406 1.063 800.1 0.104 0.246
R2 0.71 0.64 0.89 0.76 0.75
Historical effects
History 2 5 0.564 0.318 0.978 0.964 0.309
Trampling 3 15 0.190 0.039 0.939 0.977 0.574
History X Trampling 6 15 0.089 0.116 0.688 0.518 0.682
Plots within Histories 5 32 0.000 0.002 0.000 0.000 0.000
Trampling X Plots 15 32 0.913 0.909 0.545 0.827 0.572
MS Residual 32 1.797 1.5 1750 0.114 0.214
R2 0.69 0.65 0.75 0.68 0.78
Notes: The table also shows the results of analyses to assess the effects of a history of trampling for Cheviot platform.
For each variable, the table shows the probabilities from the ANOVA associated with tests of hypotheses, plus the residual
MS and the variance explained by the model and the degrees of freedom associated with numerator and denominator for each
F ratio. Combining the degrees of freedom, P values, and MS Residual allows reconstruction of the complete analysis table.
All tests of significance were done at ot = 0.05, and significant effects are shown in bold.
Our interpretation is that, for the other animals, we events, in which coralline algae lose pigment and frag-
were not seeing spatially variable direct effects of tram- ment, usually after midday low tides on very hot sum-
pling, but spatially consistent indirect responses to the mer days. It is possible that Hormosira canopies ame-
loss of Hormosira canopies. The abundances of a range liorate these effects. Algal turfs were weakly, but neg-
of common herbivores were correlated negatively with atively, correlated with the presence of Hormosira, but
the cover of Hormosira. The weakest relationship was again, we can not determine whether this result reflects
with the pulmonate limpet Siphonaria zelandica, which direct or indirect effects. Hormosira may compete with
occurs in small, wet depressions on the rock surface. some species in this category, either by direct shading
These depressions occur independently of the presence or by preempting recruitment, and its removal could
of Hormosira. In general, the animals in trampled areas allow some of these algae to establish before the im-
were established individuals, so increased abundances migration of herbivores. We have no direct experi-
were the result of migration, rather than settlement. mental evidence to address these possibilities.
Algal responses varied; we found positive correla-
Our results from the two Cheviot sites emphasize the
tions between Hormosira and the two groups of cor-
value of long-term studies. After two years, which is
alline algae, and negative correlations with the abun-
a long time for ecological experiments, the change in
dance of fleshy turfing and encrusting algae. We can
Hormosira cover was consistent with a long-term de-
not determine whether the effect on articulated coral-
line algae was direct or indirect. It was seen only at cline, with incomplete recovery of the trampled plants
Grenade Range, where losses of Hormosira were most and increasing levels of damage in the second year.
severe. These algae form dense mats on the seaward, Had we terminated the experiments at that stage, we
more exposed edges of the platforms, where Hormosira would have concluded that repeated pulse disturbances
is absent (Povey and Keough 1991), so they are not produce a press response at higher intensities of dis-
associated obligately with Hormosira. Povey and turbance. Continuing the experiments over the next few
Keough ran short-term trampling experiments in areas years saw any such trend vanish. The relatively few
at Cheviot Beach dominated by coralline algae, and a other long-term studies on marine hard substrata have
decline in coralline algae only under very intense tram- also found that results from one or two years are not
pling. However, we have frequently observed bleaching necessarily representative of patterns over longer time
156 MICHAELJ. KEOUGHAND G. P. QUINN Ecological Applications
Vol. 8, No. I
20
periods (Dayton et al. 1992, Connell et al. 1997), a
finding consistent with other habitats.
0
We found no significant effect of a prior history of L
trampling on the sensitivity of Hormosira to a new
season's trampling, nor was there any effect on asso- -20 -
ciated organisms. We do not regard the two attempts
at this experiment, in the third and fifth summers, as -40
conclusive, however. In the first attempt, the summer
was very mild, and plants may not have been stressed
(see Discussion: Natural disturbances), and trampling
had only a weak effect on any plot at Cheviot Beach.
In our second attempt, with three levels of history, a -60
major natural disturbance overrode any effects of tram- 20
pling. When there is only a weak effect of trampling -20-
in any plot, our ability to detect differences in responses
among plots with different histories will be restricted -40 -
severely. It is possible that some strong historical ef- Pool
Harry's
fects would have been apparent, had they occurred, so -60 I
we regard our nonsignificant result with caution. We
20
view these results with some additional caution be-
cause, despite the relatively long-term nature of this
0
study compared to most experiments, it is possible that
historical effects develop over very long time scales
(lOs to lOOsof years), especially if they involve genetic
adaptation in a long-lived perennial plant such as Hor-
mosira. In this case, historical effects will be very dif-
CheviotMid
ficult to detect. Perhaps more interestingly, a history
-60 1
of trampling did not affect the plants' responses to the
natural disturbance in 1994-1995, suggesting that the 1991 1992 1993 1994 1995 1996
plants' resistance had not been altered by trampling. FIG. 7. Changes in Hormosira cover in untrampled areas
The history treatments do show that the increased ef- over each of six summers. The graph shows three panels, one
fects of trampling shown in the first 2 yr at Cheviot for each site, and on each panel, bars show the mean change
Beach represent year-to-year variation in effects of in cover for two replicate control areas on each plot, with
differently shaded bars representing replicate plots. Note that
trampling, and we can refute the hypothesis that plants the two rightmost bars in each cluster are plots from the
are initially affected strongly, but then become more history of disturbance treatment. They were recorded for a
resistant. limited number of summers.
Natural disturbances shore winds, high temperatures, and locally high at-
The cover of Hormosira also changed naturally, as mospheric pressure results in lower than predicted
seen in the trajectories of control plots. The most ob- tides, and consequently greater duration of exposure of
vious source of natural disturbances that we have ob- animals and plants at low tide. When these weather
served is burnoff, in which large sections of plants turn events coincide with new or full moons and midday
brown, shrivel, and break off. In most summers, there low tides, exposure increases further.Disturbances may
is some reduction in cover from this cause, but occa- also come from wave action, which occasionally tears
sionally there are severe events that result in large up plants, or from deposition of sand, but our subjective
changes. King (1992), for example, reported a burnoff impression is that these two sources are much less com-
in the Bunurong Marine Reserve, -80 km east of our mon than burnoff.
sites, in which some plots fell by -70% as the result To examine these natural changes, we replotted all
of a single hot day. In southeastern Australia, there are data from untrampled areas to show the change in cover
mixed semidiurnal tides, and the lower of the pair of that occurred each summer. Over the six summers at
low tides occurs in the daytime in summer and at night Pt Nepean, the change in algal cover varied greatly,
in winter, in contrast to many areas of North America. and the pattern of changes over that period varied on
As a result, summer low tides are likely to be physi- large (among sites) and small (among plots) spatial
ologically stressful. The dominant weather systems in scales (Fig. 7). We have so far observed one major
southern Australia are fronts that travel rapidly east- burnoff (1994-1995), two years in which little change
wards, and can have significant effects on tidal expo- occurred over summer (1992-1993 and 1995-1996),
sure. As a high pressure system moves across, the and three years with overall declines over the summer
winds become northerly, and the combination of off- (1990-1991, 1991-1992, and 1993-1994).
February 1998 EFFECTS OF PERIODIC DISTURBANCES 157
What causes variation among years? The biggest The small-scale patchiness in the effects of natural
changes occurred in 1994-1995, and were associated disturbances was striking; a given event, such as the
with a burnoff that occurred in early October. They 1994-1995 burnoff, did not have effects that were uni-
resulted from a single day of unusually hot spring form over whole platforms. Rather, a group of appar-
weather, and plants immediately began to turn brown. ently similar plots might show very different responses
Our estimates of that change were underestimates be- (e.g., Cheviot Mid, 1995, Fig. 7). Variation among plots
cause some loss of tissue had occurred before our first was common to all years; even when conditions were
sample that year. To understand some of the variation favorable or slightly unfavorable (e.g., 1996 and 1991,
among years, we obtained climate records over the time Fig. 7), individual plots varied widely. We have no clear
period of our experiment, including daily maxima from explanation for this variation; plots were established
a weather station on Phillip Island, 50 km to the east, on apparently uniform sections of Hormosira, with lit-
and hourly tidal records from Pt. Lonsdale, 5 km west tle vertical relief and consistent plant cover.
of Pt. Nepean. We designated 28?C as a hot day, cor-
responding approximately to the 90th percentile of Natural vs. anthropogenic disturbances
summer maxima. For each hot day, we identified the A central question about anthropogenic activities,
time of daytime low tide. We could also calculate the whether they act as physical disturbances or predation,
number of hourly recordings for which our sites were is the extent to which they represent selective pressures
exposed, using 0.5 m above MLLW as the tidal height at novel spatial and/or temporal scales, rather than just
at which the platform was first exposed. The duration changes in the frequency of existing comparable nat-
of exposure ranged from 0 to 7 h, with the most frequent ural events. On these rocky shores, natural disturbances
durations being 3, 4, 5, and 6 h (19, 16, 34, and 17% happen in some years, but not others. Over the 5-6 yr
of days, respectively). The number of such hot days of our study, they did not happen at a predictable time
varied dramatically between summers, as did the num- of year, although they were concentrated between late
ber of days on which high temperatures and midday spring and early autumn. A particular combination of
low tides coincided (Fig. 8). The hottest summer was weather conditions, i.e., a disturbance of a particular
1994-1995, with 11 such days. Interestingly, there apparent intensity, has effects that are variable on at
were only three spring days over the whole period when least two spatial scales.
the maximum temperature exceeded 28?C; on one day, In contrast, disturbances from trampling occur reg-
the platform was barely exposed, but the other occa- ularly; on these shores, there are low levels of visitation
sions were consecutive days in October 1994 (see Fig. during autumn, winter, and parts of spring, when weath-
8). The predicted low tide was not low enough to ex- er conditions are less pleasant, and suitable low tides
pose the platform, but the actual low tide was 20 cm occur more often at night. There is a predictable rise
lower than predicted on those days. In contrast, in in visitation in summer, and different rock platforms
1995-1996 most plots showed an overall increase over may have consistent rank orders of their levels of vis-
the summer (Fig. 7), and the summer of 1995-1996 itation (King 1992; M. J. Keough and G. P. Quinn,
was unusually mild, with relatively humid weather and personal observations). The effects of an experimen-
high rainfall, a mean daily maximum temperature close tally controlled level of trampling were very patchy at
to the long-term average, but no hot days with midday the level of whole rock platforms. A constant level of
low tides. The other years varied; there were only three disturbance applied to plots separated by 50-100 m
hot midday tides in 1991-1992, and on those days, the produced similar results, and there was little variation
temperature did not exceed 30?C. However, the plat- between replicate strips up to 8 m apart.
form was exposed for six or more hours on each event. Anthropogenic disturbances, then, show quite dif-
In 1992-1993 there was only one occasion when very ferent patterns of temporal and spatial predictability
hot days were accompanied by midday low tides. The from natural ones. Perhaps more importantly, the ef-
other hot period did not coincide with a low tide. In fects of these disturbances on the dominant algae show
1993-1994, there were four exposure periods, but only quite different scales of spatial variation.
one coincided with temperatures >30'C. These obser- Our conceptual model of these platforms, from our
vations of weather events are broadly consistent with work and that on eastern Australian shores by Under-
changes occurring in control plots in those years. wood and colleagues (Underwood 1980, Underwood
The three sites showed different patterns through and Jernakoff 1981, 1984, Underwood and Kennelly
time, but not in any consistent manner (Fig. 7). We 1990), is that established Hormosira mats are true hab-
have already dealt with our inability to explain the itat formers; they offer damp, shaded refuges to some
different responses of the sites to trampling, and the plants and animals (G. P. Quinn, M. J. Keough, and N.
same limitation applies here. A research priority is, Gust, personal observations), at the same time reducing
therefore, to describe mesoscale changes in algal beds, microalgal abundances on rock surfaces (Underwood
most likely through remote sensing, and to correlate and Jernakoff 1981, 1984). Reductions in microalgal
the changes with properties (e.g., aspect, slope of abundance result in mobile herbivores emigrating. As
shore) of individual platforms. the cover of Hormosira is reduced, there are higher
158 MICHAEL J. KEOUGH AND G. P. QUINN Ecological Applications
Vol. 8, No. 1
40
1
995-1996
10
i19941995 I 1 4I I2
43 I0
40[ 1993-1994 666 I . MA
10
40 1993-1994
10W2
S 0 N D J F M
Month
FIG. 8. Physical conditions during the period of the experiment. The figure shows, for each spring-summer period, daily
maximum temperatures from 1 September through 31 March. The horizontal line on each panel indicates 28?C, and symbols
above a particular temperature value indicate that a low tide of a height sufficient to expose the rock platform occurred
between 1100 and 1500 on that day. Numbers adjacent to the symbols indicate the number of daylight hours that the sites
were exposed.
standing crops of microalgae, the primary molluscan and probably can only be removed by disturbance or
food source (Underwood 1979, 1980, 1984, Creese and senescence. Their ability to regenerate from the hold-
Underwood 1982), leading to increased densities of fasts suggests that they may be very long-lived, and
herbivores. Underwood and Jernakoff (1984) suggest- we have marked plants that have survived for 5 yr and
ed that newly recruited macroalgae are vulnerable to were mature when tagged. Underwood and Jernakoff's
herbivores, and at Cheviot Beach, a range of snails can hypothesis was that areas free of macroalgae and dom-
reduce microalgal standing crops (Keough et al. 1997; inated by molluscs and areas dominated by macroalgae
B. Burton, personal communication). Established Hor- are alternative community states. Hormosira can take
mosira plants appear to have few or no predators, are over bare areas, but herbivores can not cause the re-
not colonized extensively by other sessile organisms, verse change. Any process that removes Hormosira
February 1998 EFFECTS OF PERIODIC DISTURBANCES 159
mats or, in the case of anthropogenic effects, that in- ommended as a suitable indicator for routine coastal
creases the rate of disturbance, can have persistent ef- monitoring (Quinn and Keough 1993). Our present re-
fects. Hormosira may be viewed as a keystone (sensu sults suggest some caution, however. With variation in
Paine 1995) in this intertidal system, or even an au- sensitivity of beds on different rock platforms, a mon-
togenic engineer (Jones et al. 1994). itoring program could be influenced by the chance des-
ignation of an unusually sensitive or resistant site as a
Management implications control or impact location. Such a problem would be
Our results have implications for marine conserva- reduced by the use of multiple control or impact areas,
tion, but also for environmental impact monitoring. The but at the very least, the variation documented here
most important implication for monitoring is the dif- would increase the background variation in the system,
ference in sensitivity of these algal mats. These plants and most likely would result in a given "impact" hav-
are sensitive to trampling, but there is not a critical ing a variable effect. Glasby and Underwood (1996)
level of trampling that is sustainable. Rather, individual have suggested that detection of pulse and press dis-
shores separated by only a few kilometers may be very turbances requires more sophisticated sampling designs
different. Natural disturbances show similar mesoscale than are often used. Our results suggest that both may
variation in their effects. King (1992) examined the be plausible responses for a single species, so moni-
ability of Hormosira plants in the Bunurong Marine toring designs may require additional complexity.
Reserve to recover from an intense disturbance (a re- On shores of southeastern Australia, other recrea-
duction in percentage cover to 5%). She found that the tional activities affect intertidal biota. Harvesting of
rate of recovery also varied between sites separated by molluscs in the intertidal and shallow subtidal has
a few kilometers, but, as with our study, was unable strong direct (Keough et al. 1993) and some indirect
with only three sites to correlate resilience with en- (Marshall and Keough 1994, Sharpe and Keough 1997)
vironmental variables. effects near Melbourne, the major urban center. On
These combined results suggest a few options for those shores, there are no extensive algal canopies, so
management of marine reserves that receive many vis- trampling is of little concern. In contrast, we have
itors (Keough 1996). At the moment, there is the po- found only very weak effects of harvesting along the
tential for severe damage, but no way of predicting the coastline used in the present study (G. P. Quinn, M. J.
sensitivity of an individual rock platform. One might Keough, and N. Gust, personal observations) or in ar-
acknowledge this limitation, and zone marine reserves eas to the east (King 1992). Trampling seems poten-
into high and low access areas, and be prepared to tially more important than harvesting along these coast-
tolerate some damage to high access areas, while main- lines.
taining low access areas in a more natural state. If they Elsewhere in the world, harvesting has severe effects
could be identified, it would be desirable to designate on the biota of rocky shores of South Africa and South
resistant areas, such as Cheviot Beach, as high access. America (reviewed extensively in Siegfried 1994).
Such an option would require little ongoing manage- Trampling has received much less attention, with only
ment. Alternatively, one could rotate areas, opening a handful of studies, which vary in their design (Beau-
some to visitors and closing others. The "open" and champ and Gowing 1982, Brosnan and Crumrine
"resting" times for an area would need to be designated 1994). In particular, it has received little attention in
conservatively, choosing the times for sensitive areas those areas where harvesting has such striking effects,
that had low resistance and low resilience. Such a man- such as Chile and the Transkei region of South Africa
agement scheme would require ongoing management (Siegfried 1994), and it is difficult to imagine that any
action, to open and close access tracks, but would not effects of trampling could rival these impacts.
require ongoing monitoring. The third alternative For rocky shores of southeastern Australia, we can-
would be to open most areas to access, but close them not classify disturbances from pedestrian traffic simply
if any signs of damage became apparent. This method as pulse or press. Some physical processes take the
would recognize the variation among shores, but would form of repeated pulses, not only on rocky shores, but
require continuous monitoring to detect declines in al- in other habitats (Lake 1990, Riffell et al. 1996). Such
gal beds and respond to them. Such an option also pulses do not necessarily produce simply a press or a
requires monitoring of low-access areas as controls, to pulse response, but the response may vary between
distinguish changes caused by visitor access from those apparently similar habitat patches, and a press-pulse
that reflect natural disturbances. It may be possible to split is too simple for many situations. Further com-
monitor large areas of coastline using remote sensing, plications could occur if pulses occur at random, rather
and to detect changes at a range of areas. than regular intervals, or if there is a lag between a
Hormosira is sensitive to other human activities, pulse of disturbance and the response of the organisms.
such as sewage discharge (Brown et al. 1990, Fair- An appropriate theory of disturbance for this system,
weather 1990), and has been developed as a bioassay then, must deal with the spatially heterogeneous effects
species for toxicants from pulp mills and the petroleum of natural pulse disturbances, which do not occur pre-
industry (Gunthorpe et al. 1995). It has also been rec- dictably in time, but also must describe the highly vari-
160 MICHAEL J. KEOUGH AND G. P. QUINN Ecological Applications
Vol. 8, No. 1
able responses to a constant level of periodic anthro- Jones, C. G., J. H. Lawton, and M. Shachak. 1994. Organ-
pogenic disturbances. isms as ecosystem engineers. Oikos 69:373-386.
Keough, M. J. 1996. Management of intertidal shores: case
ACKNOWLEDGMENTS studies from the Victorian coast. Pages 95-99 in R. Thack-
way, editor. Criteria and guidelines for the identification
This work was supported by grants from the Australian and selection of a national representative system of marine
Research Council. The National Parks Service gave us per- protected areas. Australian Nature Conservation Agency,
mission to work at Pt Nepean, without realizing how much Canberra, Australia.
damage we'd do. Various people helped with the field work, Keough, M. J., G. P. Quinn, and R. Bathgate. 1997. Geo-
including Nick Gust, Michael Holloway, Laura Stuart, Mi- graphic variation in interactions between size classes of the
chael Shirley, Jim Radford. We greatly appreciate all their
limpet Cellana tramoserica. Journal of Experimental Ma-
efforts, particularly those of Nick Gust. Air temperature data rine Biology and Ecology 215:19-34.
were provided by the Australian Bureau of Meteorology and
Keough, M. J., G. P. Quinn, and A. King. 1993. Correlations
tide information was provided by the National Tidal Facility.
between human collecting and intertidal mollusc popula-
The manuscript benefited from comments by Sam Lake and
tions on rocky shores. Conservation Biology 7:378-391.
Barbara Downes, and we appreciate discussions with them
King, A. 1992. Monitoring and management of human ac-
and Joe Connell. Peter Petraitis and an anonymous reviewer
tivity on rocky shores. Thesis. University of Melbourne,
provided helpful suggestions.
Melbourne, Australia.
LITERATURE CITED Lake, P. S. 1990. Disturbing hard and soft bottom commu-
nities: a comparison of marine and freshwater environ-
Beauchamp, K. A., and M. M. Gowing. 1982. A quantitative ments. Australian Journal of Ecology 15:477-489.
assessment of human trampling effects on a rocky intertidal Lubchenco, J. 1978. Plant species diversity in a marine in-
community. Marine Environmental Research 7:279-293. tertidal community: importance of herbivore food prefer-
Bender, E. A., T. J. Case, and M. E. Gilpin. 1984. Pertur- ence and algal competitive abilities. American Naturalist
bation experiments in community ecology: theory and prac- 112:23-39.
tice. Ecology 65: 1-33. Marshall, P. G., and M. J. Keough. 1994. Asymmetry in
Bertness, M. D., and A. M. Ellison. 1987. Determinants of intraspecific competition in the limpet Cellana tramnoserica
pattern in a New England salt marsh plant community. Sowerby. Journal of Experimental Marine Biology and
Ecological Monographs 57:129-147. Ecology 177:121-138.
Brosnan, D. M., and L. L. Crumrine. 1994. Effects of human Paine, R. T. 1995. A conversation on refining the concept of
trampling on marine rocky shore communities. Journal of keystone species. Conservation Biology 9:962-964.
Experimental Marine Biology and Ecology. 177:79-97.
Paine, R. T., and S. A. Levin. 1981. Intertidal landscapes:
Brown, V. B., S. A. Davies, and R. N. Synnot. 1990. Long-
disturbance and the dynamics of pattern. Ecological Mono-
term monitoring of the effects of treated sewage on the
graphs 51:145-178.
intertidal macroalgal community near Cape Schanck, Vic-
Peterson, C. H., and R. Black. 1988. Density-dependent mor-
toria, Australia. Botanica Marina 33:85-98.
tality caused by physical stress interacting with biotic his-
Connell, J. H. 1978. Diversity in tropical rain forests and
tory. American Naturalist 131:257-270.
coral reefs. Science 199:1302-1310.
Petraitis, P. S., R. E. Latham, and R. A. Niesenbaum. 1989.
. 1979. Tropical rain forests and coral reefs as open
The maintenance of species diversity by disturbance. Quar-
non-equilibrium systems. Pages 141-163 in R. Anderson,
B. Turner, and L. Taylor, editors. Population dynamics. terly Review of Biology 64:393-418.
Blackwell Scientific, Oxford, UK. Povey, A. J. 1989. People on a rocky intertidal shore, effects
of trampling on animals and algae. BSc. Honours thesis.
Connell, J. H., T. P. Hughes, and C. C. Wallace. 1997. Long-
term community dynamics of corals on Heron Island, Aus- University of Melbourne, Melbourne Australia.
tralia. Ecological Monographs, in press. Povey, A., and M. J. Keough. 1991. Effects of trampling on
Connell, J. H., and M. J. Keough. 1985. Disturbance and plant and animal populations on rocky shores. Oikos 61:
patch dynamics of subtidal marine animals on hard sub- 355-368.
strata. Pages 125-152 in S. T. A. Pickett and P. S. White, Quinn, G. P., and M. J. Keough. 1993. Biological monitoring
editors. The ecology of natural disturbance and patch dy- in Port Phillip Bay and its relationship to inputs of nutrients
namics. Academic Press, New York, New York, USA. and toxicants. Technical Report 365, Environment Protec-
Creese, R. G., and A. J. Underwood. 1982. Analysis of inter- tion Authority, Melbourne, Victoria, Australia.
and intra-specific competition amongst limpets with dif- Riffell, S. K., K. J. Gutzwiller, and S. H. Anderson. 1996.
ferent methods of feeding. Oecologia (Berlin) 53:337-346. Does repeated human intrusion cause cumulative declines
Dayton, P. K., V. Currie, T. Gerrodette, B. D. Keller, R. Ro- in avian richness and abundance? Ecological Applications
senthal, and D. Van Tresca. 1984. Patch dynamics and 6:492-505.
stability of some California kelp forest communities. Eco- Schmitt, R. J., and C. W. Osenberg, editors. 1996. Detecting
logical Monographs 54:253-289. ecological impacts: concepts and applications in coastal
Dayton, P. K., M. J. Tegner, P. E. Parnell, and P. B. Edwards. habitats. Academic Press, Orlando, Florida, USA.
1992. Temporal and spatial patterns of disturbance and Scott, D. E. 1994. The effect of larval density on adult de-
recovery in a kelp forest community. Ecological Mono- mographic traits in Ambystoma opacum. Ecology 75:1383-
graphs 62:421-445. 1396.
Fairweather, P. G. 1990. Sewage and the biota on seashores: Sharpe, A., and M. J. Keough. 1997. An investigation of the
assessment of impact in relation to natural variability. En- indirect effects of intertidal shellfish collection. Journal of
vironmental Monitoring and Assessment 14:197-210. Experimental Marine Biology and Ecology, in press.
Glasby, T. M., and A. J. Underwood. 1996. Sampling to Siegfried, W. R., editor. 1994. Rocky shores: exploitation in
differentiate between pulse and press perturbations. Envi- Chile and South Africa. Springer Verlag, Berlin, Germany.
ronmental Monitoring and Assessment 42:241-252. Sousa, W. P. 1984. The role of disturbance in natural com-
Gunthorpe, L., M. Nottage, D. Palmer, and R. Wu. 1995. The munities. Annual Review of Ecology and Systematics 15:
development of a fertilisation inhibition assay using ga- 353-391.
metes of the brown alga Hormosira banksii. Australasian . 1985. Disturbance and patch dynamics on rocky
Journal of Ecotoxicology 1:25-32. intertidal shores. Pages 101-124 in S. T. A. Pickett and P.
February 1998 EFFECTS OF PERIODIC DISTURBANCES 161
S. White, editors. The ecology of natural disturbance and . 1984. Vertical and seasonal patterns in competition
patch dynamics. Academic Press, New York, New York, for microalgae between intertidal gastropods. Oecologia
USA. (Berlin) 64:211-222.
Steneck, R. S., and M. N. Dethier. 1994. A functional group Underwood, A. J., and P. Jernakoff. 1981. Effects of inter-
approach to the structure of algal-dominated communities. actions between algae and grazing gastropods on the struc-
Oikos 69:476-498. ture of a low-shore intertidal algal community. Oecologia
Tanner, J. E., T. P. Hughes, and J. H. Connell. 1996. The (Berlin) 48:221-233.
Underwood, A. J. and P. Jernakoff. 1984. The effects of tidal
role of history in community dynamics: a modelling ap-
height, wave-exposure, seasonality and rockpools on graz-
proach. Ecology 77:108-117. ing and the distribution of intertidal macroalgae in New
Underwood, A. J. 1979. The ecology of intertidal gastropods. South Wales. Journal of Experimental Marine Biology and
Advances in Marine Biology 16:110-210. Ecology 75:71-96.
1980. The effects of grazing by gastropods and phys- Underwood, A. J., and S. J. Kennelly. 1990. Ecology of
ical factors on the upper limits of distribution of intertidal marine algae on rocky shores and subtidal reefs in tem-
macroalgae. Oecologia (Berlin) 46:201-213. perate Australia. Hydrobiologia 192:3-20.
Author(s): Michael J. Keough and G. P. Quinn
Source: Ecological Applications, Vol. 8, No. 1 (Feb., 1998), pp. 141-161
Published by: Ecological Society of America
Stable URL: http://www.jstor.org/stable/2641317
Accessed: 25/01/2010 14:06
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=esa.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Ecological Society of America is collaborating with JSTOR to digitize, preserve and extend access to
Ecological Applications.
http://www.jstor.org
Ecological Applicationis, 8(1), 1998, pp. 141-161
C) 1998 by the Ecological Society of America
EFFECTS OF PERIODIC DISTURBANCES FROM TRAMPLING ON
ROCKY INTERTIDAL ALGAL BEDS
MICHAEL J. KEOUGH1 AND G. P. QUINN2
Department of Zoology, University of Melbourne, Parkville, Victoria 3052, Australia
2Department of Ecology and Evolutionary Biology, Monash University,
Clayton, Victoria 3168, Australia
Abstract. We investigated the ability of an assemblage of animals and plants on rocky
shores in southeastern Australia to resist and/or recover from repeated pulse disturbances
in the form of trampling. Disturbances of four different intensities were applied experi-
mentally over six summers, with no human access at other times of the year. The dominant
intertidal plant, the brown alga Hormosira banksii, was affected by trampling, but the
effects were heterogeneous between sites. At two sites, a series of pulse disturbances
produced a series of pulse responses, although the effect of a given pulse varied among
years, possibly related to the severity of summer desiccating conditions each year. At the
third site, pulse disturbances produced a press response; at high levels of trampling, Hor-
mosira was almost eliminated within 2 yr, and at two intermediate levels of trampling,
cover was reduced from >90 to 60-70%, where it remained for 4 yr. Effects of trampling
showed little small-scale spatial variation. Untrampled areas did fluctuate through time,
often as a result of summer burnoff of algae. Natural disturbances occurred irregularly
through the study, and their effects varied on very small spatial scales (among plots <30
m apart).
Trampling enhanced the densities of a range of herbivorous mollusks, especially limpets,
and reduced the abundance of articulated coralline algae, which were abundant in the
understory of Hormosira mats. These effects varied among sites but showed much less
variation on smaller spatial scales. The reductions in coralline algae may be a direct effect
of trampling, but increases in mollusk abundance occurred some time after changes to
Hormosira cover, and those changes may be an indirect effect of trampling.
We compared the effects of trampling on areas of the shore that had been trampled for
two and four summers, to test whether a past history of disturbance influenced the effect
of a new disturbance. No significant effects were found on algae or mobile animals, although
a mild summer may have made our test of history relatively weak.
Hormosira banksii fits the definition of a keystone species or engineer and, as such, is
an appropriate focus for management and as an indicator. Spatially heterogeneous effects
of a constant physical perturbation, however, mean that management of these rocky shores
requires more complex models and indicate that caution should be used in adopting this
species as a uniform indicator of environmental change.
Key words: disturbance, pulse and press; Hormosira;human impact; intertidal algae; press dis-
turbance; pulse disturbance; rocky shores, management of; trampling.
INTRODUCTION and Osenberg 1996). In the terminology of disturbance
(Bender et al. 1984), these events are press distur-
Human activities often constitute a disturbance to
bances. Other human activities are more variable in
natural environments, and in attempting to assess the
impacts of those activities or to develop models that time, and have been termed pulse disturbances. Spills
lead to their management, it is important to determine of toxicants often are short pulses through a system;
how they compare to natural disturbances that may be other inputs of nutrients and toxicants often vary great-
present. ly in association with high rainfall events and associ-
In nearshore coastal environments, anthropogenic ated storm water runoff. Many commercial fisheries are
disturbances vary in nature. Some, such as industrial seasonal, and any impacts of human recreational ac-
discharges, are more or less continuous stresses, and tivities are likely also to show annual cycles, reflecting
there is an extensive literature on their effects and the seasonal changes in visitation to coastal areas. These
methodology for detecting these impacts (e.g., Schmitt latter activities may be a series of pulses of distur-
bances, followed by potential recovery periods.
Many natural disturbances also have spatially vari-
Manuscript received 27 August 1996; revised 20 May
1997; accepted 16 June 1997; final version received 24 July able effects and/or patterns of recovery (e.g., Connell
1997. 1979, Paine and Levin 1981, Dayton et al. 1984, 1992,
141
142 MICHAELJ. KEOUGHAND G. P. QUINN Ecological Applications
Vol. 8, No. I
and see reviews by Sousa 1984, 1985, Connell and METHODS
Keough 1985, Lake 1990), although there are relatively
Our main study areas were extensive rocky limestone
few natural disturbances showing periodicity (but see,
platforms within Mornington Peninsula National Park,
e.g., Bertness and Ellison 1987).
in southeastern Australia. The national park extends
Periodic activities can be viewed as a series of pulse
over -30 km of moderately exposed ocean coastline,
disturbances, and a population or assemblage could re- and is accessible to the general public. A section of 8
spond in a number of ways. If the disturbance is not km was formerly under the control of the Department
too intense or the system has high resilience, it may of Defence, and was incorporated into the national park
be able to recover in the intervals between disturbances in 1989. This section of the park has remained closed
(Petraitis et al. 1989). Recovery may also occur if the to public access, so shores in that area have been pro-
interval between disturbances is long. In the termi- tected from direct human influence for >75 yr. The
nology of recent disturbance theory, repeated pulse dis- dominant intertidal habitat is provided by the fucoid
turbances may produce a series of pulse-like recoveries, alga Hormosira banksii, which forms large monotypic
or act as a press disturbance. Whether a particular dis- beds (Fig. 1). The individual plants have a basal hold-
turbance regime produces pulse or press responses is fast, from which are produced fronds. Fronds are com-
also likely to depend on the intensity of the disturbance: posed of chains of vesicles (Fig. 1). The general de-
low intensity disturbances, i.e., those that cause little scription of the major habitat types in this area is pro-
damage during a given pulse, may allow rapid recov- vided by Povey and Keough (1991). Hormosira beds
ery, but there may be a critical intensity beyond which provide habitat for a range of smaller gastropod mol-
persistent changes occur. lusks and crustaceans. Other mollusks feed primarily
On rocky shorelines of southern Australia, one of in open (i.e., lacking macroalgal cover) areas, and their
the most prominent species is the perennial fucoid alga abundances are negatively correlated with the presence
Hormosira banksii, which forms extensive monotypic of large algae (G. P. Quinn and M. J. Keough, personal
stands at midtidal levels of rock platforms (Fig. 1). observations).
Hormosira beds are habitat for a range of mobile an-
imals, and their presence is negatively associated with Experimental designs
other species. H. banksii is sensitive to short-term tra-
The main trampling experiment ran for 6 yr, and
pling (Povey and Keough 1991) and is deleteriously
consisted of areas of the shore being trampled over
affected by other anthropogenic activities, including
summer, followed by a recovery period from midau-
discharge of sewage (Brown et al. 1990, Fairweather
tumn to early summer. On these shores, there are rel-
1990). We used this alga to test the effects of recurrent
atively low levels of visitation by humans until late
disturbances at a range of intensities, with the distur-
December, when levels become high through January
bances recurring annually over six summers. during a major holiday period, taper off in February,
In particular, do seasonally recurrent disturbances and remain at that level until approximately Easter
caused by humans produce pulse or press responses? (March) when another brief holiday period occurs
Are the responses spatially consistent? Does the kind (King 1992). Our experiment followed those broad pat-
of response to disturbance vary with intensity of dis- terns of use.
turbance? We also compared the (controlled) human We established three trampling sites at hapazardly
disturbances to changes produced by natural events. chosen areas within the protected area of the national
We also considered the possibility that organisms' park. Two sites were on different parts of the intertidal
sensitivity to new disturbances might be related to their platforms at Cheviot Beach; Harry's Pool was adjacent
history. For example, a history of competition may to a large rock pool and Cheviot Mid was -200 m
make an organism more sensitive to physical distur- along the shore, but set back from the edge of the
bance (Peterson and Black 1988) or change some life platform. The third site, Grenade Range, was at a plat-
history parameters later in its life (Scott 1994). Tanner form -1 km away, separated from Cheviot Beach by
et al. (1996) provide an overview of historical effects, two headlands. At each site, we initially established
although their own data did not show a strong effect two plots, --10 X 3 m, separated by 30-50 m, where
of history on community dynamics and structure. A the percentage cover of Hormosira exceeded 90. We
natural or anthropogenic disturbance may stress an or- marked eight trampling strips within each plot, each
ganism, inhibiting its ability to respond to new chal- strip being 50 cm wide, and 2-3 m long, with the exact
lenges, repeated occurrences of the same disturbance length varying among plots, depending on the size of
or novel stresses. Intertidal algae could become the Hormosira patch in which the plot was placed.
stressed by desiccation and become more vulnerable to Strips were separated by at least 1 m and parallel to
trampling, or vice versa. We took advantage of our each other. Each plot included two replicates of each
long-term experiment to contrast the responses of trampling treatment.
plants with a long history of disturbance to those of The main trampling experiment used four intensities
plants with little or no history of trampling. of trampling. Our focus here is on changes occurring
February 1998 EFFECTS OF PERIODIC DISTURBANCES 143
FIG. 1. Hormosirabanksii plants. The top panel shows a false-color infrared image of a small rock platform at Cheviot
Beach. The red area in the center of the platform is dense Hormo.sira,with small patches elsewhere on the platform. The
photograph covers an area -50 m wide. The lower panel shows individual plants that have been subject to moderate trampling,
with some reduction in cover. Note the morphology of the plants. with fronds of spherical vesicles in long chains, and a
combination of intact and damaged chains. The picture covers an area 10 cm wide.
144 MICHAELJ. KEOUGHAND G. P. QUINN Ecological Applications
Vol. 8, No. 1
each summer, so the intensity of disturbance is the num- plots were identical to the existing two at each site,
ber of passages per summer (which is equivalent to the and were within the same large algal mat. We repeated
number/day), as the number of trampling days was con- this procedure at the beginning of the summer of 1994-
sistent across treatments. One passage consisted of a 1995, so at Harry's Pool and Cheviot Mid, we had two
person of average size walking at normal pace along plots that had been trampled for 4 yr, one that had been
a strip, and strips received either 0, 5, 10, or 25 passages trampled for 2 yr, and one with no history of trampling.
on a given low tide. For repeated passages, tramplers All of the new strips were trampled and sampled in the
moved beyond the end of the strip before turning, to manner described above, including a census of the com-
prevent more severe forces associated with pivoting of plete fauna in autumn of 1995.
the feet. All trampling was done by average-sized
adults, wearing rubber-soled athletic shoes or sandals, Analyses
footwear of a similar type to that used by recreational All data were analyzed by analysis of variance. Our
visitors (Povey 1989). Each summer, we used 6-8 d of design for the main experiment involved six factors,
such trampling on every strip, spread haphazardly over and corresponded to a split-plot or repeated-measures
the suitably low tides during summer. We began the design. We use the latter terminology, for clarity. The
trampling in the summer of 1990-1991 and continued spatial component of the design was partly nested, with
through the summer of 1995-1996. The experiment has Sites, and Plots within Sites. Both factors were crossed
continued, and in this paper we present analyses of five with Trampling, with two strips within each Plot-Tram-
years of data, plus descriptions of events occurring dur- pling combination. Each strip was then sampled 10
ing the sixth summer (1995-1996). times, with those 10 samples falling into 5 yr, and
We sampled the experiment twice annually, at the before/after each summer. In repeated-measures ter-
beginning of summer, and again after Easter, after the minology, the "subjects" were strips, the between-sub-
last trampling period. Easter holidays represent the last jects factors were Site, Plot, and Trampling, and the
major recreational influx to coastal areas before winter. within-subjects factors were Years and Before-After
The measurement after Easter was intended to assess Summer. Trampling, Before-After Summer, and Years
the effects of trampling, while the measurement in early were fixed factors, the latter because we had sampled
summer provided a measure of recovery over the pre- for all five years in the period. Strips, Plots, and Sites
ceding seasons. Each strip was sampled using a 70 x were random factors, the latter to allow us to generalize
35 cm quadrat placed in the center of each strip, with about spatial variation in responses to disturbance.
its long axis running parallel to the strip. We photo- The detailed censuses at the end of 3 and 5 yr were
graphed the quadrat with color slide film (first 3 yr) or each analyzed by partly nested analysis of variance,
Hi-8 video (second 3 yr), and back in the laboratory, i.e., the above design with no within-subjects factors
we projected the images and calculated the cover of (Plots within Sites, crossed with Trampling, and two
Hormosira using 100 randomly placed dots superim- replicates). We analyzed the abundance of all common
posed on the image. taxa. Coralline algae can not be identified to species
On two occasions, after three and five summers (i.e., in the field, although they are numerically dominated
in autumn of 1993 and 1995) we counted understory by species of Corallina, and we separated them into
algae and mobile animals. We did this using two quad- articulate and encrusting forms because of the different
rats, placed end-to-end in the center of each strip. Algae ecological properties of those growth forms (Steneck
were estimated using a grid of 100 points, and we and Dethier 1994). Uncalcified turfing or encrusting
counted all algae beneath each point. Animals were algae were also abundant enough to analyze, even
identified and counted, after a thorough search of each though individual species, such as Cladophora and
quadrat. We used two quadrats because some animals Ralfsia, could not be analyzed. We pooled all algae
were uncommon, and we required a larger sample, but other than Hormosira to create a further plant variable.
numbers were averaged to provide a single value for The animals were dominated by mollusks, and herbi-
each strip. We did not do complete censuses often, vores in particular. The predatory whelk Thais orbita
because we considered the sampling procedure to be and the scavenging Cominella lineolata were present,
potentially disruptive to understory algae or associated but not sufficiently abundant for analysis. We analyzed
invertebrates. abundance of the two true limpets Cellana tramoserica
Historical effects.-After 2 yr, we tested whether and Patelloida alticostata, the two pulmonate limpets
plants with a history of disturbance and recovery might Siphonaria diemenensis and S. zelandica, plus the lit-
be more resilient or more sensitive to a new distur- torinid Bembicium nanum. We created two additional
bance, by adding a new plot to the Harry's Pool and pooled herbivore groups, (true) limpets (Cellana + Pa-
Cheviot Mid sites. At both sites, Hormosira had re- telloida alticostata + P. latistrigata) and nonlimpet
covered by the beginning of summer. At Grenade grazers (Siphonaria spp., Bembicium, plus Austro-
Range, Hormosira had declined in treatment plots (see cochlea constricta, Turbo undulata). We also assessed
Results), so we could not compare the response of these the performance of various "community" statistics; we
plants to that of previously untrampled areas. The new calculated the taxonomic richness (number of recog-
February 1998 EFFECTS OF PERIODIC DISTURBANCES 145
TABLE 1. Analysis of changes in cover of Hormosira banksii through time, as a function of level of trampling.
Denom.
Source of variation df no.t MS F P P (Chev)
Between-strips (i.e., pooled across time) effects
1. Sites (S) 2 2 2255.4 4.26 0.133 0.815
2. Plots within Sites (P{S}) 3 6 529.2 0.44 0.725
3. Trampling (T) 3 6 15778.5 13.19 0.000
4. Trampling X Site 6 5 4186.9 15.67 0.000 0.087
5. T X P{S} 9 6 267.1 0.22 0.988
6. Strips within Plots (Residual) 24 1196.5
Within-strips (i.e., temporal) effects
7. Years (Y) 4 13 17053.7 84.67 0.000
8. Year X Site 8 9 1804.9 3.91 0.017 0.302
9. Years X P{S} 12 13 462.1 2.29 0.013
10. Year X Trampling 12 13 284.4 1.41 0.201
11. Y X T X S 24 11 450.3 2.25 0.013 0.458
12. Y X T X P{S} 36 13 200.0 0.99 0.493
13. Strips X Years (Residual) 96 201.4
14. Before-After Summer (BA) 1 20 32472.3 394.82 0.000
15. BA X Sites 2 16 124.8 0.13 0.880 0.897
16. BA X P{S} 3 20 936.6 11.39 0.000
17. BA X Trampling 3 20 826.2 10.05 0.000
18. BA X S X T 6 19 107.8 2.35 0.120 0.279
19. BA X T X P{S} 9 20 45.9 0.56 0.817
20. Strips X BA Residual 24 82.2
21. Years X Before-After Summer 4 27 2919.5 38.17 0.000
22. Y x BA x S 8 23 534.5 0.90 0.546 0.383
23. T X BA X P{S} 12 27 594.4 7.77 0.000
24. Y X BA X Trampling 12 27 516.4 6.75 0.000
25. Y x BA X T X S 24 26 68.3 0.79 0.728 0.336
26. Y X BA x T X P{S} 36 27 86.8 1.14 0.308
27. Y X BA X Strips Residual 96 76.5
Notes: Significant effects (at (- = 0.05) are shown in boldface. The right-most column shows the P values associated with
tests for heterogeneous effects of trampling at the two sites on the Cheviot Beach Platform.
t Terms are numbered, and the denominators used to test each effect are indicated using those numbers.
nizable taxa), species richness of gastropods, total Illinois). For all analyses, we examined primarily the
number of individuals, and the Shannon-Wiener di- assumption of normality, by examining residuals by
versity index (H'). probability plots. There were generally too few repli-
The experiment to examine history of disturbance cates at a given level of the design for a meaningful
was analyzed using data collected at the end of the comparison of variances. When we used repeated-mea-
1994-1995 summer. In the analysis, we found no sig- sures or partly hierarchical analyses, we also examined
nificant variation between the two Cheviot sites (at ax the more conservative Greenhouse-Geiser and Huynh-
= 0.25), so we omitted them from the analysis, to give Feldt corrected F tests, which provide some protection
four levels of trampling, and three levels of history (0, against violations of assumptions of compound sym-
2, or 4 yr) with either four, two, or two plots within metry. Those tests did not produce results markedly
each level of history, respectively. The data were an- different from the uncorrected ones, and we saw no
alyzed as a partly nested analysis, with History, evidence of strong violation of these assumptions, so
Plots{History}, Trampling, H X T, and T X Plots as only the standard F tests are presented here.
the terms in the analysis, and History and Trampling Note, in all analysis tables, probabilities are rounded
as fixed factors and Plots as a random factor. Our con- to three decimal places for brevity; values <0.0005 are
clusions would not be altered by the more conservative shown as 0.000.
step of retaining sites within the analysis. We analyzed
Hormosira cover and abundances of all common taxa. RESULTS
To be more confident of detecting an effect on Hor-
Effects on Hormosira banksii
mosira, we analyzed the percentage covers from the
postsummer 1994-1995 survey, and data from pre- and We found striking temporal and spatial variation in
postsummer of 1995-1996, so we could examine the the percentage cover of H. banksii, and find it helpful
profile through time of plots with different histories. to separate effects involving trampling from those that
The three values were treated as repeated measures, presumably represent natural variation.
using the statistical model described above. Effects of trampling.-Trampling affected Hormo-
All data analysis was done using SYSTAT for Win- sira beds dramatically, with the effects varying through
dows, version 5.03 (SYSTAT Incorporated, Evanston, time and through space (Table 1). Individual plants
146 MICHAELJ. KEOUGHAND G. P. QUINN EcologicalApplications
Vol. 8, No. 1
Cheviot Platform Sites
100
0 40
U
20 erk| |None 5 10 25|
eror
Grenade Range Site
100
N0 -V
60 -~ P-
0
40t40
'-
- errors |.
20 .
o~~~~~~~~~~~
1991 1992 1993 1994 1995 1996
FIG. 2. Changes in cover of Hormosira banksii on two platforms, under different levels of trampling. Data were pooled
from the two sites at Cheviot Beach, as they showed no significant heterogeneity. The bar at the top of the figure indicates
periods during which trampling occurred (as dark blocks) and times when there was no disturbance (clear blocks). The error
bars at the base of each figure indicate three standard errors, calculated using the variance components for error terms used
to test Trampling X time effects, using data from Table 1. The left error bar is the geometric mean of the time X Strips
residuals from Table 1, the middle error represents the Trampling X time X Plots term (based on three components: Year X
T X P, BA X T X P and Y X BA X T X P). The right error bar indicates variance for assessing variation that is independent
of trampling, i.e., time X Plots terms.
were initially damaged by the loss of chains of vesicles, cover through to late 1994. Plants in control strips cov-
and ultimately by whole fronds, as described by Povey ered 80-100% of space until late 1994, while the major
and Keough (1991). Under severe damage, they were changes occurred in the most heavily disturbed areas,
reduced to holdfasts (Fig. 1). The effects of trampling where cover declined after each period of trampling,
were quite different at the three sites (see Trampling with little or no recovery in the intervening seasons,
x Site, Year x Site X Trampling interactions on Table so that by late 1994, cover had fallen to <10%. In the
1). At Grenade Range, there was a decline in percentage austral summer of 1994-1995, there was a major de-
cover after the first summer's trampling, and the rate cline across all treatments, with cover falling by -30%
of decline was proportional to the intensity of trampling in controls and the two intermediate levels of distur-
(Fig. 2). The plants recovered by the beginning of the bance, and falling to a few percent in the most heavily
following summer, but then declined even more under disturbed areas (Fig. 2). After this decline, there was
the second year's trampling. There was little subse- some recovery in the control and intermediate treat-
quent recovery, and in the third and fourth years plants ments, although cover did not return to its levels of the
in the two intermediate treatments remained at 60-70% spring of 1995, and there was little recovery in the
February 1998 EFFECTS OF PERIODIC DISTURBANCES 147
heavily disturbed areas. Trampling in the summer of cant Site x Trampling interactions for articulate cor-
1995-1996 had little apparent effect (Fig. 2). alline algae, total algae, Cellana tramoserica, Siphon-
The situation was very different at the two Cheviot aria diemenensis, and the pooled categories of limpets
Beach sites. The effects of trampling did not differ and nonlimpet grazers. When the data for the two Chev-
significantly between the two sites (Table 1). We there- iot platforms were analyzed, the only significant effect
fore discuss the two sites together. For the first two of trampling was a Plots x Trampling interaction for
years, the Cheviot sites followed a trajectory similar Patelloida alticostata (Table 2). For Cellana and S.
to that shown by Grenade Range, with an initial de- diemenensis, there was no consistent relationship be-
cline, a complete recovery, then a more severe decline tween intensity of trampling and abundance at the
during the disturbances of the second year (Fig. 2). Cheviot sites, but at Grenade Range, a 2-4-fold in-
From that stage, however, all plots recovered com- crease in abundance at the highest intensity of tram-
pletely, and for the next 3 yr, we saw little effect of pling (Fig. 3). There were similar patterns for Patel-
trampling. There was a marked decline in cover in the loida and S. zelandica, although they were more vari-
summer of 1994-1995, as at Grenade Range, but this able, and the analyses did not show significant effects
decline was consistent across all trampling treatments. after 5 yr. When the data were pooled, the nonlimpet
There had been complete recovery by midspring of grazers and limpets showed strong relationships with
1995, and trampling had little effect in that summer, trampling at Grenade Range, but again, no apparent
with an increase in cover in three treatments, and a pattern at the two Cheviot Beach sites.
decline only under heavy trampling. Articulate corallines varied dramatically in abun-
The variation that we observed in the effects of tram- dance between sites, and responded variably to tram-
pling was almost all at larger spatial scales; we were pling. They formed a major part of the understory at
able to test whether replicate plots at each site showed Cheviot Mid, with a mean cover of -50%, yet covered
the same effect of trampling, and all four tests incor- no more than 10% at the other two sites (Fig. 3). At
porating a Trampling x Plot interaction were nonsig- the two Cheviot sites, there was no relationship with
nificant (Table 1). Similarly, the variation among rep- trampling, but at Grenade Range, their cover declined
licate strips within plots was quite small; the four terms with trampling from -8% in control areas to 0 in the
incorporating Strips on Table 1 together accounted for most heavily trampled treatments (Fig. 3). There was
only 18% of the total sum of squares in the analysis no pattern for encrusting corallines or for turfing algae,
(compared to -35% for trampling effects). and the patterns for total algae reflected the hetero-
Temporal variation independent of trampling.-The geneous results of individual taxa.
pattern of variation was different when we examined At the time of the first census, after three summers
the effects that were unrelated to trampling. There was of trampling, the only species to show consistent effects
strong variation in percentage cover of Hormosira of trampling was the limpet Cellana tramoserica (Table
among Years and Before-After Summer (Table 1), as 3). There were isolated small-scale effects of trampling
well as small-scale variation in cover. Changes over on another limpet, Patelloida alticostata, and the cover
summer reflect burning-off of the algae, with a con- of articulated coralline algae, both of which showed
sequent loss of biomass and/or cover, and these effects significant Trampling x Plot variation (Table 3). There
varied among years (Years x BA interaction; Table 1). was also a significant effect of trampling on all limpets
This pattern was not significantly heterogeneous among pooled. Almost all common taxa varied significantly
the three sites. Cover declined over the summers of among plots, but not among sites (Table 3). When the
1990-1991 and 1994-1995, and showed either weak two Cheviot sites only were compared, the results were
changes (1991-1992) or no discernible change (1992- similar except that there was no effect of trampling on
1993, 1993-1994, 1995-1996) in the other summers the abundance of Patelloida, and significant Plot X
(Fig. 2). The variation among years was not consistent Trampling effects on the abundance of the pulmonate
among the three sites, but this effect was small, and limpet Siphonaria diemenensis, Cellana tramoserica,
not clear from the graphs (Fig. 2). and the pooled categories of nonlimpet grazers and
In contrast to the trampling effects, there was small- limpets (Table 3).
scale variation in the effects of Year and Before-After The derived variables showed relatively weak ef-
Summer, with significant interactions between Plots fects. There was no significant effect of trampling on
and these two factors. taxonomic richness, gastropod species richness, or H',
regardless of whether all sites or just the Cheviot plat-
Effects on other organisms forms were compared (Table 3). The number of indi-
Trampling affected other organisms, but they were viduals was affected by trampling, with the effect vary-
generally heterogeneous among the three sites. After 5 ing between sites (Table 3). For this variable, residual
yr, the effects of trampling varied significantly among plots did not completely support the assumptions of
sites, rather than among plots (Table 2), with the sites the analysis for raw or log-transformed data, but were
effect generated by the difference between Grenade intermediate. However, the effect of trampling was con-
Range and the two Cheviot Sites. There were signifi- sistent for raw and log-transformed data, and disap-
148 MICHAEL J. KEOUGH AND G. P. QUINN EcologicalApplications
Vol. 8, No. 1
TABLE2. Analyses of the abundance of major animals and plants after 5 yr of trampling, highlighting effects of trampling
at all three sites and at only the two sites on Cheviot Beach platform.
df df Articulate Encrust.
Trampling (num.) (denom.) corallines corallines All corallines Turf All algae
Overall (all sites)
Sites (S) 2 3 0.343 0.786 0.359 0.584 0.382
Trampling (T) 3 6 0.388 0.499 0.624 0.130 0.615
S X T 6 9 0.018 0.916 0.283 0.565 0.013
Plots within Sites 3 24 0.000 0.244 0.000 0.000 0.000
T X Plots 9 24 0.997 0.105 0.397 0.272 0.870
MS Residual 24 38.8 0.02 66.9 0.01 72.8
R2 0.98 0.53 0.96 0.82 0.96
Cheviot only
Sites 1 2 0.362 0.821 0.354 0.745 0.359
Trampling 3 2 0.343 0.807 0.436 0.647 0.532
S X T 3 6 0.143 0.190 0.169 0.138 0.113
Plots within Sites 2 16 0.000 0.266 0.000 0.000 0.000
T X Plots 6 16 0.983 0.765 0.912 0.749 0.893
MS Residual 16 52.0 5.2 49.8 45.1 43.4
R2 0.98 0.41 0.98 0.75 0.98
Historical effects
History 2 5 0.665 0.388 0.464 0.689 0.645
Trampling 3 15 0.690 0.882 0.195 0.914 0.852
History X Trampling 6 15 0.185 0.031 0.379 0.312 0.468
Plots within Histories 5 32 0.000 0.289 0.015 0.000 0.000
Trampling x Plots 15 32 0.781 0.930 0.976 0.752 0.535
MS Residual 32 56.4 13.5 144.9 70.6 58.7
RI2 0.97 0.45 0.51 0.96 0.97
Notes: The table also shows the results of analyses to assess the effects of a history of trampling for Cheviot platform.
For each taxon, the table shows the probabilities from the ANOVA associated with tests of hypotheses, plus the residual MS
and the variance explained by the model and the degrees of freedom associated with numerator and denominator for each F
ratio. Combining the degrees of freedom, P values, and MS Residual allows reconstruction of the complete analysis table.
All tests of significance were done at a = 0.05. Significant effects are shown in bold.
peared when only the Cheviot sites were compared, ables also showed strong variation among plots that
indicating that the primary difference was, again, be- was unrelated to the levels of trampling.
tween Grenade Range and the two Cheviot sites, with Because a particular level of trampling produced a
the number of individuals rising strongly at Grenade different cover of Hormosira at different sites, it is
Range as the intensity of trampling increased. All vari- probably not surprising that most of the effects on other
TABLE 3. Analyses of the abundance of major animals and plants after three summers of trampling, highlighting effects of
trampling at all three sites and at only the two sites on Cheviot Beach platform.
df df Articulate Encrust.
Trampling (num.) (denom.) corallines corallines All corallines Turf All algae
Overall (all sites)
Sites (S) 2 3 0.415 0.092 0.427 0.184 0.464
Trampling (T) 3 6 0.980 0.243 0.990 0.236 0.668
S XT 6 9 0.510 0.456 0.402 0.539 0.464
Plots within Sites 3 24 0.000 0.516 0.000 0.012 0.000
T X Plots 9 24 0.030 0.931 0.061 0.694 0.355
MS Residual 24 15.4 0.005 17.1 0.007 37.6
R2 0.99 0.45 0.99 0.70 0.98
Cheviot only
Sites 1 2 0.412 0.292 0.415 0.869 0.432
Trampling 3 2 0.975 0.720 0.981 0.470 0.899
S XT 3 6 0.387 0.700 0.329 0.298 0.387
Plots within Sites 2 16 0.000 0.068 0.000 0.017 0.000
T x Plots 6 16 0.057 0.399 0.065 0.716 0.423
MS Residual 16 22.8 0.0 22.1 0.0 37.1
R2 0.99 0.58 0.99 0.56 0.99
Notes: For each taxon, the table shows the probabilities from the ANOVA associated with tests of hypotheses, plus the
residual MS and the variance explained by the model and the degrees of freedom associated with numerator and denominator
for each F ratio. Combining the degrees of freedom, P values, and MS Residual allows reconstruction of the complete analysis
table. All tests of significance were done at a = 0.05. Significant effects are shown in bold.
February 1998 EFFECTS OF PERIODIC DISTURBANCES 149
TABLE 2. Extended.
Cellana Patelloida Siphonaria Siphonaria Bembicium Nonlimpet
tramoserica alticostata diemenensis zelandica nanum grazers Limpets
0.705 0.839 0.692 0.153 0.604 0.494 0.747
0.532 0.519 0.606 0.469 0.543 0.562 0.508
0.005 0.191 0.001 0.086 0.092 0.001 0.006
0.000 0.041 0.000 0.098 0.000 0.000 0.000
0.579 0.471 0.913 0.889 0.343 0.950 0.645
3.07 2.15 1355 489 0.30 6.95
0.82 0.60 0.80 0.62 0.88 0.76 0.80
0.766 0.942 0.721 0.542 0.929 0.693 0.814
0.987 0.265 0.966 0.274 0.528 0.955 0.788
0.166 0.871 0.198 0.736 0.128 0.151 0.350
0.000 0.001 0.000 0.000 0.000 0.000 0.000
0.538 0.048 0.465 0.957 0.150 0.497 0.332
2.5 0.85 718 29.0 0.24 742 3.6
0.80 0.75 0.87 0.69 0.90 0.89 0.84
0.581 0.304 0.972 0.930 0.572 0.974 0.507
0.323 0.606 0.922 0.499 0.855 0.932 0.247
0.743 0.865 0.797 0.157 0.463 0.702 0.732
0.000 0.000 0.000 0.000 0.000 0.000 0.000
0.384 0.037 0.574 0.931 0.345 0.546 0.359
2.23 0.57 1535 29.0 0.45 1674 3.29
0.79 0.76 0.71 0.67 0.84 0.74 0.82
organisms were manifest as a Site x Trampling or Plot and Plots as random effects). We found only one sig-
X Trampling interaction. Our conceptual model for nificant interaction between Hormosira cover and Site,
these shores is that Hormosira is the habitat-forming for Cellana tramoserica, and no significant Plot X Hor-
species, and that the abundances of other organisms mosira interaction, suggesting that the trampling ef-
may track changes in Hormosira. To take account of fects may have been indirect responses to changes in
variation in cover of Hormosira, we reanalyzed the data Hormosira.
in the upper part of Table 2 (i.e., all three sites) as a There were significant effects of Hormosira for ar-
nested analysis of covariance (Sites, Plots within Sites, ticulated corallines, all coralline and all algae, Patel-
and cover of Hormosira as the covariate, with Sites loida, Siphonaria diemenensis, and the nonlimpet graz-
TABLE 3. Extended.
Cellana Patelloida Siphonaria Siphonaria Bembicium Nonlimpet
tramoserica alticostata diemenensis zelandica nanum grazers Limpets
0.735 0.527 0.670 0.638 0.102 0.775 0.639
0.027 0.313 0.181 0.182 0.253 0.152 0.044
0.894 0.430 0.355 0.684 0.396 0.388 0.805
0.062 0.004 0.000 0.000 0.119 0.000 0.006
0.570 0.017 0.094 0.913 0.099 0.124 0.187
0.932 0.276 64.0 0.781 0.100 70.7 1.604
0.52 0.78 0.78 0.64 0.75 0.77 0.68
0.720 0.669 0.750 0.530 0.147 0.820 0.696
0.029 0.562 0.377 0.547 0.509 0.197 0.047
0.917 0.249 0.236 0.467 0.272 0.520 0.862
0.003 0.008 0.000 0.002 0.095 0.000 0.000
0.035 0.744 0.007 0.921 0.116 0.027 0.029
0.188 0.172 5.37 1.00 0.119 11.0 0.266
0.79 0.60 0.93 0.63 0.77 0.90 0.85
150 J. AND G. P. QUINN
MICHAEL KEOUGH Ecological Applications
Vol. 8, No. 1
Grenade Harry'sPool Cheviot Mid
l.l i I
i-. 0................_-
60 Artic. coral. 0.3 Enc. coral.
0 ~~~~~~0
U 0 U 0.
L
40 ~ Cellana 20 Patelloida
0-**- ..*..
0 C
800 Siph. diem. 240 Siph. z.
4 Bembicium 0.5 - Turf algae
.0 ~ ~ ~ ~ .
-0 ~~ ~
0< 0 5 X,
0 5 10 25 0 5 10 25
Level of Trampling
FIG. 3. Abundance of algae and herbivorous snails after 5 yr of trampling. The organisms are articulate coralline algae,
encrusting coralline algae, the limpets Cellana tramoserica and Patelloida alticostata, the pulmonate limpets Siphonaria
diemenensis and S. zelandica, the littorinid snail Bembicium nanum, and a group of foliose algae (Turf). Each panel on the
graph shows the mean abundance, as percentage cover or number per square meter (N). Three standard error bars are shown
at the lower left corner of each panel. The left bar indicates the variation among replicate strips within plots, the middle bar
indicates the Plots{Site} x Trampling variation, while the third indicates the Plots{Sites} variation. The middle bar is the
most important for interpreting the figures, as it represents the error term used to assess the Sites x Trampling interaction.
Note that encrusting coralline algae and turf algae are plotted as arcsine transformed values, and Bembicium data are shown
as square-root transformed values. These scales represent the data used in the analyses.
February 1998 EFFECTS OF PERIODIC DISTURBANCES 151
TABLE 4. Regressions of the abundance of individual taxa on percentage cover of Hormosira for all sites, Grenade Range
only, and the Cheviot sites. Data were individual trampling strips, and sample sizes were 48 (all sites), 16, and 32. The
direction of the relationship, i.e., the sign of the regression slope, is also shown.
Direction All sites Grenade Range Cheviot sites
Taxon +/- r2 p r2 p r2 p
Articulated coralline algae + 16 0.000 73 0.000 18 0.001
Encrusting corallines + 4 0.068 4 0.461 6 0.052
All corallines + 20 0.000 25 0.047 21 0.000
Turf algae - 6 0.031 11 0.202 13 0.003
All algae + 22 0.000 41 0.007 22 0.000
Cellana tramoserica - 35 0.000 56 0.001 23 0.000
Patelloida alticostata - 21 0.000 31 0.024 12 0.006
Siphonaria diemenensis - 50 0.000 61 0.000 43 0.000
Siphonaria zelandica - 18 0.000 30 0.029 10 0.010
Bembicium nanum - 14 0.001 9 0.269 27 0.000
Nonlimpet grazers - 50 0.000 60 0.000 43 0.000
Limpets - 35 0.000 52 0.002 24 0.000
Austrocochlea constricta + 2 0.266 14 0.150 1 0.584
Derived variables
Taxonomic richness -0+t 2 0.343 <1 0.865 13 0.040
Gastropod richness - 3 0.237 12 0.189 2 0.418
Total individuals - 57 0.000 60 0.000 52 0.000
Log (individuals) - 64 0.000 73 0.000 65 0.000
H' 2 0.381 20 0.082 15 0.026
Note: Significant effects are shown in bold.
t Overall negative effect, weak positive at Grenade Range, negative at Cheviot sites.
t Overall positive, negative at GR, positive at Cheviot.
ers and limpets. Simple regressions of the abundance summer of 1995-1996, and postsummer 1994-1995),
of each of these taxa against Hormosira cover at the and considered the temporal profiles using a repeated-
time of the census were used to indicate the direction measures analysis. Over that time period, we did not
of the effects. These regressions were significant for find even a significant effect of trampling (see Table
most taxa, although the strong relationships were with 5), with only a marginally nonsignificant main effect
cover of articulate coralline algae, Cellana, Patelloida, of trampling after summer of 1995-1996 even hinting
Siphonaria diemenensis, and the composite grazing cat- at a change. There was significant variation through
egories (Table 4). We also analyzed Grenade Range time and among plots, but all effects involving history
and the two Cheviot sites separately, and relationships (i.e., the main effect, plus the interactions with time
were generally stronger at Grenade Range than Cheviot and trampling) were far from significant (Table 5). The
(Table 4; paired t test using r2 values for each taxon, lack of a significant effect did not appear to be a result
t = 3.12, df = 12, P = 0.009). At Grenade Range, of low power; rather, plots with the three different his-
there were particularly strong effects of Hormosira tories of trampling showed very similar patterns (Fig.
cover for articulate coralline algae, Cellana and Si- 5).
phonaria diemenensis. Cover of both kinds of coralline The censuses of other plants and animals after 5 yr
algae (and the pooled variables total coralline and total also showed little effect of history. There were no sig-
algae) varied positively with Hormosira cover, as did nificant simple effects of history, and only a single
the abundance of the herbivorous snail Austrocochlea interaction with trampling, for encrusting coralline al-
constricta, whereas the slopes of the regressions were gae (Table 2). Main effects of history would be difficult
negative for turfing algae and all other herbivorous to identify, as we found substantial variation among
snails (Table 4). replicate plots, the level of variation used for assessing
Trends in the derived variables also became clearer this main effect. The test of the History x Trampling
with the analysis of covariance. The Site x Trampling interaction, however, had 6 and 15 degrees of freedom,
effects on the number of individuals disappeared, and and there was little variation among plots in the effect
there was a very strong overall effect of Hormosira of trampling. The Plots X Trampling interaction was
cover (Fig. 4). There was no effect of algal cover on used to test the History x Trampling term. The mean
either taxonomic or gastropod richness, and, interest- abundances of the major taxa showed no hint of an
ingly, a positive effect of Hormosira on H' at the Chev- effect, with the plots with no history having the highest
iot sites and a (nonsignificant) negative trend at Gre- means for some taxa, while plots with the longest his-
nade Range (Table 4). tory of trampling were greatest for other taxa (Fig. 6).
Even the one taxon showing a significant effect, the
Historical effects of trampling encrusting coralline algae, did not have a pattern of
We found few strong effects of history. We analyzed means that could be easily interpreted with respect to
the cover of Hormosira at three times (pre- and post- the effects of history or trampling (Fig. 6).
152 MICHAEL J. KEOUGH AND G. P. QUINN Ecological
Applications
Vol. 8, No. 1
15 15
ww A AU
*AAA AAL AA
10 AA A A 10
U A H A HA
LI A n o A A
L \AI *
AA A
I A
5 5 * AA AA nA n
n no A A L n A
A LI ~ UA** n
0 III0 II
0 20 40 60 80 100 0 20 40 60 80 100
4.0 3.0
AU
3.0 L A A A A LI
AA A 2.0 ~ ALI
A ~~~~~~~~~~~~AAU
2.0 A , A on
A AA ALIn
IiEALI
AA1.0A
1.0 LI*A A OLA A
LI A n A
A nL
AA
0.0 -----J0.0
0 20 40 60 80 100 0 20 40 60 80 100
Hormosira cover (%)
FIG. 4. Variation in derived variables listed on Table 6, plotted as a function of Hormosira cover. On each panel, the
three symbols indicate main sites (squares are Grenade Range, and triangles and stars represent the two Cheviot Platform
sites). Filled and unfilled symbols identify the two plots at each site, and each point is a replicate strip.
TABLE 5. Effects of history of trampling on Hormosira banksii. The table shows the results of analysis of variance of the
percentage cover of H. banksii at each of three times (upper panel), together with a repeated measures analysis examining
all time simultaneously (lower panel).
All data
df df Postsummer Presummer Postsummer (repeated
Source of variation (num.) (denom.) 1994-1995 1995-1996 1995-1996 measures)
History 2 5 0.791 0.848 0.871 0.849
Trampling 3 15 0.583 0.659 0.062 0.273
History x Trampling 6 15 0.795 0.589 0.758 0.565
Plots within Histories 5 32 0.000 0.325 0.094 0.015
Trampling x Plots 15 32 0.771 0.930 0.763 0.912
MS Residual 32 405.2 337.2 243.3 774.3
r2 0.61 0.34 0.48
Repeated measures analysis: within-strips effects
Time 2 10 0.016
Time x History 4 10 0.917
Time x Trampling 6 30 0.536
Time x History x Trampling 12 30 1.000
Time x Plots within Histories 10 0.000
Time X Trampling X Plots 30 0.459
Time x MS Residual 64 105.6
r2 ~~~~~~~~~~~~~~~~~~
Notes: Probabilities are shown, with significant effects (at 0.05) shown in bold. The multiple r2 is shown for each
a
analysis, and for the repeated measures, the r2was calculated by treating the two residual (MS Residual and Strips MS Residual)
terms as unexplained variation.
February 1998 EFFECTS OF PERIODIC DISTURBANCES 153
* - * o -o *................. sulted in a rapid decline in the dominant alga and little
4 yr 2 yr No history recovery, a press response. At all sites, there was a
relationship between change in algal cover and inten-
100 sity of disturbance, but nowhere did increasing inten-
sity of disturbance cause a shift from a pulse to a press
response. There was no hint of any "intermediate"
effect of disturbance (Connell 1978, Lubchenco 1978);
trampling at intermediate levels produced results that
were intermediate, and there was no hint of a peak in
EndSummer 1995-1996 measures of diversity at intermediate levels.
0 1S I Why were the effects different at the three sites? We
found strong variation among sites in the effect of tram-
pling, but little variation in the effects of trampling at
100 the level of plots, suggesting that variation in responses
to disturbance is determined by habitat variation at
larger spatial scales. All three sites and all plots were
on horizontal sections of rock platform, and were at
approximately the same height on the shore. Our im-
pression was that the Grenade Range site was slightly
V Pre-Summer1995-1996 more exposed to wave action, but we expected any
0 I--SE
difference to reduce the levels of desiccation, and make
the algae more, rather than less resilient. It is difficult
to identify a cause of this variation; at the largest spatial
100
scale, we had only three replicate sites, and to correlate
EndSummer1994-1995 the sensitivity of a site with environmental variables
(e.g., coastal geomorphology, height, orientation, and
other factors) would require data from many more sites.
Our study was designed to assess variation among sites
in general, rather than focusing on these three precise
0 I I I locations (hence sites as a random factor in the anal-
yses).
Another primary aim of our study was to identify
0 5 10 25 "thresholds" of disturbance, levels of trampling that
Trampling Intensity could be sustained, but beyond which recovery was
incomplete. Again, we did not reach a simple conclu-
FIG. 5. Effects of history of trampling on the percentage sion, but instead have shown increased complexity in
cover of Hormosira banksii. Each panel shows the mean cover the relationship with intensity of disturbance (level of
of H. banksii, pooled over plots at the two Cheviot platform
sites, against the level of trampling, and data are shown for trampling). At the two Cheviot sites, there was a re-
three censuses. Error bars are shown in the corner of each lationship between the intensity of disturbance during
panel, calculated as described in Fig. 3. those summers when trampling had a strong effect. In
other years, we found little effect of trampling, and
hence no relationship between intensity and change in
The community variables also showed no effect of algal abundances. In contrast, there was clear evidence
history, whether we examined taxonomic or gastropod of a nonlinear relationship between intensity of dis-
species richness, total numbers of animals, or diversity turbance and algal cover at Grenade Range: intense
of the animals (Table 6). disturbances produced a very large change, but the two
DISCUSSION intermediate levels (5 and 10 passages/day) produced
very similar patterns of algal abundance, and resulted
Anthropogenic disturbances and Hormosira in Hormosira cover being intermediate between un-
For this intertidal system, a series of repeated pulse disturbed and heavily disturbed areas.
disturbances did not consistently produce either a pulse
or press response. At our two Cheviot sites, each dis- Were the effects of trampling on other species
turbance pulse produced a response, but recovery gen- direct or indirect?
erally occurred over the following 8-9 mo, so we saw The changes in the abundances of other organisms
essentially a series of pulse responses. After six sum- seemed not to happen at the same time as changes in
mers of trampling, we found no strong effect of re- Hormosira. After three summers, few species showed
peated disturbances, and certainly no cumulative effect. strong effects, even though there had been considerable
In contrast, the series of pulses at Grenade Range re- changes in Hormosira cover by that time. These species
154 MICHAEL J. KEOUGH AND G. P. QUINN Applications
Ecological
Vol. 8, No. 1
No history 2 4
40 Artic. coral. 0.3 Enc. coral
0 0
0 0.0
6 6
*1 | | Patelloida
Cellana
0 ~~~~~~~~~~0
400 200 -
'I Siph. diem. Siph. z.
0 0
4 I Bembicium || Turfalgae
0 1,l 110 ESnlf
0 5 10 25 0 5 10 25
Level of Trampling
FIG. 6. Effects of history of trampling on the abundance of algae and herbivorous snails. The species are described in
the caption to Fig. 3. Each panel shows the mean abundance of the taxon for four levels of trampling and three histories.
The error bars are the first two bars described in Fig. 3.
may have been directly affected by trampling, but just peated their experiments in areas lacking macroalgae,
responded more slowly (and positively in some cases!), and found little direct effect on gastropod abundances,
but our results for herbivorous snails are consistent although their test may not have been sensitive because
with short-term trampling experiments on the platforms of the likelihood of movement of individuals between
at Cheviot Beach. Povey and Keough (1991) found that experimental and surrounding areas. Supporting evi-
increases in the densities of herbivorous snails occurred dence is provided by our analyses of the 1995 census
some time after the cessation of trampling in Hormo- in which we replaced the level of trampling with the
sira mats, suggesting a response to the removal of mac- cover of Hormosira, and found that the interactions
roalgae, rather than a direct effect of feet. They re- between trampling and space (sites, plots) disappeared.
February 1998 EFFECTS OF PERIODIC DISTURBANCES 155
TABLE 6. Analyses of derived variables after five years of trampling, highlighting effects of trampling at all three sites and
at only the two sites on Cheviot Beach platform. The variables were taxonomic richness (S), species richness of gastropods
(Sgast), numbers of animals (raw and log-transformed), and Shannon-Wiener diversity (H').
df df
Trampling (num.) (denom.) S Sgast N Log N H'
Overall (all sites)
Sites (S) 2 3 0.719 0.492 0.508 0.678 0.883
Trampling (T) 3 6 0.730 0.872 0.555 0.785 0.814
S x T 6 9 0.071 0.331 0.001 0.019 0.507
Plots within Sites 3 24 0.003 0.027 0.000 0.000 0.000
T x Plots 9 24 0.873 0.826 0.946 0.913 0.688
MS Residual 24 2.063 1.438 2956 0.112 0.262
R2 0.6 0.5 0.8 0.7 0.7
Cheviot only
Sites 1 2 0.698 0.450 0.703 0.839 0.721
Trampling 3 2 0.387 0.175 0.948 0.545 0.990
S x T 3 6 0.078 0.746 0.162 0.367 0.498
Plots within Sites 2 16 0.000 0.008 0.000 0.000 0.000
T X Plots 6 16 0.934 0.555 0.486 0.719 0.460
MS Residual 16 1.406 1.063 800.1 0.104 0.246
R2 0.71 0.64 0.89 0.76 0.75
Historical effects
History 2 5 0.564 0.318 0.978 0.964 0.309
Trampling 3 15 0.190 0.039 0.939 0.977 0.574
History X Trampling 6 15 0.089 0.116 0.688 0.518 0.682
Plots within Histories 5 32 0.000 0.002 0.000 0.000 0.000
Trampling X Plots 15 32 0.913 0.909 0.545 0.827 0.572
MS Residual 32 1.797 1.5 1750 0.114 0.214
R2 0.69 0.65 0.75 0.68 0.78
Notes: The table also shows the results of analyses to assess the effects of a history of trampling for Cheviot platform.
For each variable, the table shows the probabilities from the ANOVA associated with tests of hypotheses, plus the residual
MS and the variance explained by the model and the degrees of freedom associated with numerator and denominator for each
F ratio. Combining the degrees of freedom, P values, and MS Residual allows reconstruction of the complete analysis table.
All tests of significance were done at ot = 0.05, and significant effects are shown in bold.
Our interpretation is that, for the other animals, we events, in which coralline algae lose pigment and frag-
were not seeing spatially variable direct effects of tram- ment, usually after midday low tides on very hot sum-
pling, but spatially consistent indirect responses to the mer days. It is possible that Hormosira canopies ame-
loss of Hormosira canopies. The abundances of a range liorate these effects. Algal turfs were weakly, but neg-
of common herbivores were correlated negatively with atively, correlated with the presence of Hormosira, but
the cover of Hormosira. The weakest relationship was again, we can not determine whether this result reflects
with the pulmonate limpet Siphonaria zelandica, which direct or indirect effects. Hormosira may compete with
occurs in small, wet depressions on the rock surface. some species in this category, either by direct shading
These depressions occur independently of the presence or by preempting recruitment, and its removal could
of Hormosira. In general, the animals in trampled areas allow some of these algae to establish before the im-
were established individuals, so increased abundances migration of herbivores. We have no direct experi-
were the result of migration, rather than settlement. mental evidence to address these possibilities.
Algal responses varied; we found positive correla-
Our results from the two Cheviot sites emphasize the
tions between Hormosira and the two groups of cor-
value of long-term studies. After two years, which is
alline algae, and negative correlations with the abun-
a long time for ecological experiments, the change in
dance of fleshy turfing and encrusting algae. We can
Hormosira cover was consistent with a long-term de-
not determine whether the effect on articulated coral-
line algae was direct or indirect. It was seen only at cline, with incomplete recovery of the trampled plants
Grenade Range, where losses of Hormosira were most and increasing levels of damage in the second year.
severe. These algae form dense mats on the seaward, Had we terminated the experiments at that stage, we
more exposed edges of the platforms, where Hormosira would have concluded that repeated pulse disturbances
is absent (Povey and Keough 1991), so they are not produce a press response at higher intensities of dis-
associated obligately with Hormosira. Povey and turbance. Continuing the experiments over the next few
Keough ran short-term trampling experiments in areas years saw any such trend vanish. The relatively few
at Cheviot Beach dominated by coralline algae, and a other long-term studies on marine hard substrata have
decline in coralline algae only under very intense tram- also found that results from one or two years are not
pling. However, we have frequently observed bleaching necessarily representative of patterns over longer time
156 MICHAELJ. KEOUGHAND G. P. QUINN Ecological Applications
Vol. 8, No. I
20
periods (Dayton et al. 1992, Connell et al. 1997), a
finding consistent with other habitats.
0
We found no significant effect of a prior history of L
trampling on the sensitivity of Hormosira to a new
season's trampling, nor was there any effect on asso- -20 -
ciated organisms. We do not regard the two attempts
at this experiment, in the third and fifth summers, as -40
conclusive, however. In the first attempt, the summer
was very mild, and plants may not have been stressed
(see Discussion: Natural disturbances), and trampling
had only a weak effect on any plot at Cheviot Beach.
In our second attempt, with three levels of history, a -60
major natural disturbance overrode any effects of tram- 20
pling. When there is only a weak effect of trampling -20-
in any plot, our ability to detect differences in responses
among plots with different histories will be restricted -40 -
severely. It is possible that some strong historical ef- Pool
Harry's
fects would have been apparent, had they occurred, so -60 I
we regard our nonsignificant result with caution. We
20
view these results with some additional caution be-
cause, despite the relatively long-term nature of this
0
study compared to most experiments, it is possible that
historical effects develop over very long time scales
(lOs to lOOsof years), especially if they involve genetic
adaptation in a long-lived perennial plant such as Hor-
mosira. In this case, historical effects will be very dif-
CheviotMid
ficult to detect. Perhaps more interestingly, a history
-60 1
of trampling did not affect the plants' responses to the
natural disturbance in 1994-1995, suggesting that the 1991 1992 1993 1994 1995 1996
plants' resistance had not been altered by trampling. FIG. 7. Changes in Hormosira cover in untrampled areas
The history treatments do show that the increased ef- over each of six summers. The graph shows three panels, one
fects of trampling shown in the first 2 yr at Cheviot for each site, and on each panel, bars show the mean change
Beach represent year-to-year variation in effects of in cover for two replicate control areas on each plot, with
differently shaded bars representing replicate plots. Note that
trampling, and we can refute the hypothesis that plants the two rightmost bars in each cluster are plots from the
are initially affected strongly, but then become more history of disturbance treatment. They were recorded for a
resistant. limited number of summers.
Natural disturbances shore winds, high temperatures, and locally high at-
The cover of Hormosira also changed naturally, as mospheric pressure results in lower than predicted
seen in the trajectories of control plots. The most ob- tides, and consequently greater duration of exposure of
vious source of natural disturbances that we have ob- animals and plants at low tide. When these weather
served is burnoff, in which large sections of plants turn events coincide with new or full moons and midday
brown, shrivel, and break off. In most summers, there low tides, exposure increases further.Disturbances may
is some reduction in cover from this cause, but occa- also come from wave action, which occasionally tears
sionally there are severe events that result in large up plants, or from deposition of sand, but our subjective
changes. King (1992), for example, reported a burnoff impression is that these two sources are much less com-
in the Bunurong Marine Reserve, -80 km east of our mon than burnoff.
sites, in which some plots fell by -70% as the result To examine these natural changes, we replotted all
of a single hot day. In southeastern Australia, there are data from untrampled areas to show the change in cover
mixed semidiurnal tides, and the lower of the pair of that occurred each summer. Over the six summers at
low tides occurs in the daytime in summer and at night Pt Nepean, the change in algal cover varied greatly,
in winter, in contrast to many areas of North America. and the pattern of changes over that period varied on
As a result, summer low tides are likely to be physi- large (among sites) and small (among plots) spatial
ologically stressful. The dominant weather systems in scales (Fig. 7). We have so far observed one major
southern Australia are fronts that travel rapidly east- burnoff (1994-1995), two years in which little change
wards, and can have significant effects on tidal expo- occurred over summer (1992-1993 and 1995-1996),
sure. As a high pressure system moves across, the and three years with overall declines over the summer
winds become northerly, and the combination of off- (1990-1991, 1991-1992, and 1993-1994).
February 1998 EFFECTS OF PERIODIC DISTURBANCES 157
What causes variation among years? The biggest The small-scale patchiness in the effects of natural
changes occurred in 1994-1995, and were associated disturbances was striking; a given event, such as the
with a burnoff that occurred in early October. They 1994-1995 burnoff, did not have effects that were uni-
resulted from a single day of unusually hot spring form over whole platforms. Rather, a group of appar-
weather, and plants immediately began to turn brown. ently similar plots might show very different responses
Our estimates of that change were underestimates be- (e.g., Cheviot Mid, 1995, Fig. 7). Variation among plots
cause some loss of tissue had occurred before our first was common to all years; even when conditions were
sample that year. To understand some of the variation favorable or slightly unfavorable (e.g., 1996 and 1991,
among years, we obtained climate records over the time Fig. 7), individual plots varied widely. We have no clear
period of our experiment, including daily maxima from explanation for this variation; plots were established
a weather station on Phillip Island, 50 km to the east, on apparently uniform sections of Hormosira, with lit-
and hourly tidal records from Pt. Lonsdale, 5 km west tle vertical relief and consistent plant cover.
of Pt. Nepean. We designated 28?C as a hot day, cor-
responding approximately to the 90th percentile of Natural vs. anthropogenic disturbances
summer maxima. For each hot day, we identified the A central question about anthropogenic activities,
time of daytime low tide. We could also calculate the whether they act as physical disturbances or predation,
number of hourly recordings for which our sites were is the extent to which they represent selective pressures
exposed, using 0.5 m above MLLW as the tidal height at novel spatial and/or temporal scales, rather than just
at which the platform was first exposed. The duration changes in the frequency of existing comparable nat-
of exposure ranged from 0 to 7 h, with the most frequent ural events. On these rocky shores, natural disturbances
durations being 3, 4, 5, and 6 h (19, 16, 34, and 17% happen in some years, but not others. Over the 5-6 yr
of days, respectively). The number of such hot days of our study, they did not happen at a predictable time
varied dramatically between summers, as did the num- of year, although they were concentrated between late
ber of days on which high temperatures and midday spring and early autumn. A particular combination of
low tides coincided (Fig. 8). The hottest summer was weather conditions, i.e., a disturbance of a particular
1994-1995, with 11 such days. Interestingly, there apparent intensity, has effects that are variable on at
were only three spring days over the whole period when least two spatial scales.
the maximum temperature exceeded 28?C; on one day, In contrast, disturbances from trampling occur reg-
the platform was barely exposed, but the other occa- ularly; on these shores, there are low levels of visitation
sions were consecutive days in October 1994 (see Fig. during autumn, winter, and parts of spring, when weath-
8). The predicted low tide was not low enough to ex- er conditions are less pleasant, and suitable low tides
pose the platform, but the actual low tide was 20 cm occur more often at night. There is a predictable rise
lower than predicted on those days. In contrast, in in visitation in summer, and different rock platforms
1995-1996 most plots showed an overall increase over may have consistent rank orders of their levels of vis-
the summer (Fig. 7), and the summer of 1995-1996 itation (King 1992; M. J. Keough and G. P. Quinn,
was unusually mild, with relatively humid weather and personal observations). The effects of an experimen-
high rainfall, a mean daily maximum temperature close tally controlled level of trampling were very patchy at
to the long-term average, but no hot days with midday the level of whole rock platforms. A constant level of
low tides. The other years varied; there were only three disturbance applied to plots separated by 50-100 m
hot midday tides in 1991-1992, and on those days, the produced similar results, and there was little variation
temperature did not exceed 30?C. However, the plat- between replicate strips up to 8 m apart.
form was exposed for six or more hours on each event. Anthropogenic disturbances, then, show quite dif-
In 1992-1993 there was only one occasion when very ferent patterns of temporal and spatial predictability
hot days were accompanied by midday low tides. The from natural ones. Perhaps more importantly, the ef-
other hot period did not coincide with a low tide. In fects of these disturbances on the dominant algae show
1993-1994, there were four exposure periods, but only quite different scales of spatial variation.
one coincided with temperatures >30'C. These obser- Our conceptual model of these platforms, from our
vations of weather events are broadly consistent with work and that on eastern Australian shores by Under-
changes occurring in control plots in those years. wood and colleagues (Underwood 1980, Underwood
The three sites showed different patterns through and Jernakoff 1981, 1984, Underwood and Kennelly
time, but not in any consistent manner (Fig. 7). We 1990), is that established Hormosira mats are true hab-
have already dealt with our inability to explain the itat formers; they offer damp, shaded refuges to some
different responses of the sites to trampling, and the plants and animals (G. P. Quinn, M. J. Keough, and N.
same limitation applies here. A research priority is, Gust, personal observations), at the same time reducing
therefore, to describe mesoscale changes in algal beds, microalgal abundances on rock surfaces (Underwood
most likely through remote sensing, and to correlate and Jernakoff 1981, 1984). Reductions in microalgal
the changes with properties (e.g., aspect, slope of abundance result in mobile herbivores emigrating. As
shore) of individual platforms. the cover of Hormosira is reduced, there are higher
158 MICHAEL J. KEOUGH AND G. P. QUINN Ecological Applications
Vol. 8, No. 1
40
1
995-1996
10
i19941995 I 1 4I I2
43 I0
40[ 1993-1994 666 I . MA
10
40 1993-1994
10W2
S 0 N D J F M
Month
FIG. 8. Physical conditions during the period of the experiment. The figure shows, for each spring-summer period, daily
maximum temperatures from 1 September through 31 March. The horizontal line on each panel indicates 28?C, and symbols
above a particular temperature value indicate that a low tide of a height sufficient to expose the rock platform occurred
between 1100 and 1500 on that day. Numbers adjacent to the symbols indicate the number of daylight hours that the sites
were exposed.
standing crops of microalgae, the primary molluscan and probably can only be removed by disturbance or
food source (Underwood 1979, 1980, 1984, Creese and senescence. Their ability to regenerate from the hold-
Underwood 1982), leading to increased densities of fasts suggests that they may be very long-lived, and
herbivores. Underwood and Jernakoff (1984) suggest- we have marked plants that have survived for 5 yr and
ed that newly recruited macroalgae are vulnerable to were mature when tagged. Underwood and Jernakoff's
herbivores, and at Cheviot Beach, a range of snails can hypothesis was that areas free of macroalgae and dom-
reduce microalgal standing crops (Keough et al. 1997; inated by molluscs and areas dominated by macroalgae
B. Burton, personal communication). Established Hor- are alternative community states. Hormosira can take
mosira plants appear to have few or no predators, are over bare areas, but herbivores can not cause the re-
not colonized extensively by other sessile organisms, verse change. Any process that removes Hormosira
February 1998 EFFECTS OF PERIODIC DISTURBANCES 159
mats or, in the case of anthropogenic effects, that in- ommended as a suitable indicator for routine coastal
creases the rate of disturbance, can have persistent ef- monitoring (Quinn and Keough 1993). Our present re-
fects. Hormosira may be viewed as a keystone (sensu sults suggest some caution, however. With variation in
Paine 1995) in this intertidal system, or even an au- sensitivity of beds on different rock platforms, a mon-
togenic engineer (Jones et al. 1994). itoring program could be influenced by the chance des-
ignation of an unusually sensitive or resistant site as a
Management implications control or impact location. Such a problem would be
Our results have implications for marine conserva- reduced by the use of multiple control or impact areas,
tion, but also for environmental impact monitoring. The but at the very least, the variation documented here
most important implication for monitoring is the dif- would increase the background variation in the system,
ference in sensitivity of these algal mats. These plants and most likely would result in a given "impact" hav-
are sensitive to trampling, but there is not a critical ing a variable effect. Glasby and Underwood (1996)
level of trampling that is sustainable. Rather, individual have suggested that detection of pulse and press dis-
shores separated by only a few kilometers may be very turbances requires more sophisticated sampling designs
different. Natural disturbances show similar mesoscale than are often used. Our results suggest that both may
variation in their effects. King (1992) examined the be plausible responses for a single species, so moni-
ability of Hormosira plants in the Bunurong Marine toring designs may require additional complexity.
Reserve to recover from an intense disturbance (a re- On shores of southeastern Australia, other recrea-
duction in percentage cover to 5%). She found that the tional activities affect intertidal biota. Harvesting of
rate of recovery also varied between sites separated by molluscs in the intertidal and shallow subtidal has
a few kilometers, but, as with our study, was unable strong direct (Keough et al. 1993) and some indirect
with only three sites to correlate resilience with en- (Marshall and Keough 1994, Sharpe and Keough 1997)
vironmental variables. effects near Melbourne, the major urban center. On
These combined results suggest a few options for those shores, there are no extensive algal canopies, so
management of marine reserves that receive many vis- trampling is of little concern. In contrast, we have
itors (Keough 1996). At the moment, there is the po- found only very weak effects of harvesting along the
tential for severe damage, but no way of predicting the coastline used in the present study (G. P. Quinn, M. J.
sensitivity of an individual rock platform. One might Keough, and N. Gust, personal observations) or in ar-
acknowledge this limitation, and zone marine reserves eas to the east (King 1992). Trampling seems poten-
into high and low access areas, and be prepared to tially more important than harvesting along these coast-
tolerate some damage to high access areas, while main- lines.
taining low access areas in a more natural state. If they Elsewhere in the world, harvesting has severe effects
could be identified, it would be desirable to designate on the biota of rocky shores of South Africa and South
resistant areas, such as Cheviot Beach, as high access. America (reviewed extensively in Siegfried 1994).
Such an option would require little ongoing manage- Trampling has received much less attention, with only
ment. Alternatively, one could rotate areas, opening a handful of studies, which vary in their design (Beau-
some to visitors and closing others. The "open" and champ and Gowing 1982, Brosnan and Crumrine
"resting" times for an area would need to be designated 1994). In particular, it has received little attention in
conservatively, choosing the times for sensitive areas those areas where harvesting has such striking effects,
that had low resistance and low resilience. Such a man- such as Chile and the Transkei region of South Africa
agement scheme would require ongoing management (Siegfried 1994), and it is difficult to imagine that any
action, to open and close access tracks, but would not effects of trampling could rival these impacts.
require ongoing monitoring. The third alternative For rocky shores of southeastern Australia, we can-
would be to open most areas to access, but close them not classify disturbances from pedestrian traffic simply
if any signs of damage became apparent. This method as pulse or press. Some physical processes take the
would recognize the variation among shores, but would form of repeated pulses, not only on rocky shores, but
require continuous monitoring to detect declines in al- in other habitats (Lake 1990, Riffell et al. 1996). Such
gal beds and respond to them. Such an option also pulses do not necessarily produce simply a press or a
requires monitoring of low-access areas as controls, to pulse response, but the response may vary between
distinguish changes caused by visitor access from those apparently similar habitat patches, and a press-pulse
that reflect natural disturbances. It may be possible to split is too simple for many situations. Further com-
monitor large areas of coastline using remote sensing, plications could occur if pulses occur at random, rather
and to detect changes at a range of areas. than regular intervals, or if there is a lag between a
Hormosira is sensitive to other human activities, pulse of disturbance and the response of the organisms.
such as sewage discharge (Brown et al. 1990, Fair- An appropriate theory of disturbance for this system,
weather 1990), and has been developed as a bioassay then, must deal with the spatially heterogeneous effects
species for toxicants from pulp mills and the petroleum of natural pulse disturbances, which do not occur pre-
industry (Gunthorpe et al. 1995). It has also been rec- dictably in time, but also must describe the highly vari-
160 MICHAEL J. KEOUGH AND G. P. QUINN Ecological Applications
Vol. 8, No. 1
able responses to a constant level of periodic anthro- Jones, C. G., J. H. Lawton, and M. Shachak. 1994. Organ-
pogenic disturbances. isms as ecosystem engineers. Oikos 69:373-386.
Keough, M. J. 1996. Management of intertidal shores: case
ACKNOWLEDGMENTS studies from the Victorian coast. Pages 95-99 in R. Thack-
way, editor. Criteria and guidelines for the identification
This work was supported by grants from the Australian and selection of a national representative system of marine
Research Council. The National Parks Service gave us per- protected areas. Australian Nature Conservation Agency,
mission to work at Pt Nepean, without realizing how much Canberra, Australia.
damage we'd do. Various people helped with the field work, Keough, M. J., G. P. Quinn, and R. Bathgate. 1997. Geo-
including Nick Gust, Michael Holloway, Laura Stuart, Mi- graphic variation in interactions between size classes of the
chael Shirley, Jim Radford. We greatly appreciate all their
limpet Cellana tramoserica. Journal of Experimental Ma-
efforts, particularly those of Nick Gust. Air temperature data rine Biology and Ecology 215:19-34.
were provided by the Australian Bureau of Meteorology and
Keough, M. J., G. P. Quinn, and A. King. 1993. Correlations
tide information was provided by the National Tidal Facility.
between human collecting and intertidal mollusc popula-
The manuscript benefited from comments by Sam Lake and
tions on rocky shores. Conservation Biology 7:378-391.
Barbara Downes, and we appreciate discussions with them
King, A. 1992. Monitoring and management of human ac-
and Joe Connell. Peter Petraitis and an anonymous reviewer
tivity on rocky shores. Thesis. University of Melbourne,
provided helpful suggestions.
Melbourne, Australia.
LITERATURE CITED Lake, P. S. 1990. Disturbing hard and soft bottom commu-
nities: a comparison of marine and freshwater environ-
Beauchamp, K. A., and M. M. Gowing. 1982. A quantitative ments. Australian Journal of Ecology 15:477-489.
assessment of human trampling effects on a rocky intertidal Lubchenco, J. 1978. Plant species diversity in a marine in-
community. Marine Environmental Research 7:279-293. tertidal community: importance of herbivore food prefer-
Bender, E. A., T. J. Case, and M. E. Gilpin. 1984. Pertur- ence and algal competitive abilities. American Naturalist
bation experiments in community ecology: theory and prac- 112:23-39.
tice. Ecology 65: 1-33. Marshall, P. G., and M. J. Keough. 1994. Asymmetry in
Bertness, M. D., and A. M. Ellison. 1987. Determinants of intraspecific competition in the limpet Cellana tramnoserica
pattern in a New England salt marsh plant community. Sowerby. Journal of Experimental Marine Biology and
Ecological Monographs 57:129-147. Ecology 177:121-138.
Brosnan, D. M., and L. L. Crumrine. 1994. Effects of human Paine, R. T. 1995. A conversation on refining the concept of
trampling on marine rocky shore communities. Journal of keystone species. Conservation Biology 9:962-964.
Experimental Marine Biology and Ecology. 177:79-97.
Paine, R. T., and S. A. Levin. 1981. Intertidal landscapes:
Brown, V. B., S. A. Davies, and R. N. Synnot. 1990. Long-
disturbance and the dynamics of pattern. Ecological Mono-
term monitoring of the effects of treated sewage on the
graphs 51:145-178.
intertidal macroalgal community near Cape Schanck, Vic-
Peterson, C. H., and R. Black. 1988. Density-dependent mor-
toria, Australia. Botanica Marina 33:85-98.
tality caused by physical stress interacting with biotic his-
Connell, J. H. 1978. Diversity in tropical rain forests and
tory. American Naturalist 131:257-270.
coral reefs. Science 199:1302-1310.
Petraitis, P. S., R. E. Latham, and R. A. Niesenbaum. 1989.
. 1979. Tropical rain forests and coral reefs as open
The maintenance of species diversity by disturbance. Quar-
non-equilibrium systems. Pages 141-163 in R. Anderson,
B. Turner, and L. Taylor, editors. Population dynamics. terly Review of Biology 64:393-418.
Blackwell Scientific, Oxford, UK. Povey, A. J. 1989. People on a rocky intertidal shore, effects
of trampling on animals and algae. BSc. Honours thesis.
Connell, J. H., T. P. Hughes, and C. C. Wallace. 1997. Long-
term community dynamics of corals on Heron Island, Aus- University of Melbourne, Melbourne Australia.
tralia. Ecological Monographs, in press. Povey, A., and M. J. Keough. 1991. Effects of trampling on
Connell, J. H., and M. J. Keough. 1985. Disturbance and plant and animal populations on rocky shores. Oikos 61:
patch dynamics of subtidal marine animals on hard sub- 355-368.
strata. Pages 125-152 in S. T. A. Pickett and P. S. White, Quinn, G. P., and M. J. Keough. 1993. Biological monitoring
editors. The ecology of natural disturbance and patch dy- in Port Phillip Bay and its relationship to inputs of nutrients
namics. Academic Press, New York, New York, USA. and toxicants. Technical Report 365, Environment Protec-
Creese, R. G., and A. J. Underwood. 1982. Analysis of inter- tion Authority, Melbourne, Victoria, Australia.
and intra-specific competition amongst limpets with dif- Riffell, S. K., K. J. Gutzwiller, and S. H. Anderson. 1996.
ferent methods of feeding. Oecologia (Berlin) 53:337-346. Does repeated human intrusion cause cumulative declines
Dayton, P. K., V. Currie, T. Gerrodette, B. D. Keller, R. Ro- in avian richness and abundance? Ecological Applications
senthal, and D. Van Tresca. 1984. Patch dynamics and 6:492-505.
stability of some California kelp forest communities. Eco- Schmitt, R. J., and C. W. Osenberg, editors. 1996. Detecting
logical Monographs 54:253-289. ecological impacts: concepts and applications in coastal
Dayton, P. K., M. J. Tegner, P. E. Parnell, and P. B. Edwards. habitats. Academic Press, Orlando, Florida, USA.
1992. Temporal and spatial patterns of disturbance and Scott, D. E. 1994. The effect of larval density on adult de-
recovery in a kelp forest community. Ecological Mono- mographic traits in Ambystoma opacum. Ecology 75:1383-
graphs 62:421-445. 1396.
Fairweather, P. G. 1990. Sewage and the biota on seashores: Sharpe, A., and M. J. Keough. 1997. An investigation of the
assessment of impact in relation to natural variability. En- indirect effects of intertidal shellfish collection. Journal of
vironmental Monitoring and Assessment 14:197-210. Experimental Marine Biology and Ecology, in press.
Glasby, T. M., and A. J. Underwood. 1996. Sampling to Siegfried, W. R., editor. 1994. Rocky shores: exploitation in
differentiate between pulse and press perturbations. Envi- Chile and South Africa. Springer Verlag, Berlin, Germany.
ronmental Monitoring and Assessment 42:241-252. Sousa, W. P. 1984. The role of disturbance in natural com-
Gunthorpe, L., M. Nottage, D. Palmer, and R. Wu. 1995. The munities. Annual Review of Ecology and Systematics 15:
development of a fertilisation inhibition assay using ga- 353-391.
metes of the brown alga Hormosira banksii. Australasian . 1985. Disturbance and patch dynamics on rocky
Journal of Ecotoxicology 1:25-32. intertidal shores. Pages 101-124 in S. T. A. Pickett and P.
February 1998 EFFECTS OF PERIODIC DISTURBANCES 161
S. White, editors. The ecology of natural disturbance and . 1984. Vertical and seasonal patterns in competition
patch dynamics. Academic Press, New York, New York, for microalgae between intertidal gastropods. Oecologia
USA. (Berlin) 64:211-222.
Steneck, R. S., and M. N. Dethier. 1994. A functional group Underwood, A. J., and P. Jernakoff. 1981. Effects of inter-
approach to the structure of algal-dominated communities. actions between algae and grazing gastropods on the struc-
Oikos 69:476-498. ture of a low-shore intertidal algal community. Oecologia
Tanner, J. E., T. P. Hughes, and J. H. Connell. 1996. The (Berlin) 48:221-233.
Underwood, A. J. and P. Jernakoff. 1984. The effects of tidal
role of history in community dynamics: a modelling ap-
height, wave-exposure, seasonality and rockpools on graz-
proach. Ecology 77:108-117. ing and the distribution of intertidal macroalgae in New
Underwood, A. J. 1979. The ecology of intertidal gastropods. South Wales. Journal of Experimental Marine Biology and
Advances in Marine Biology 16:110-210. Ecology 75:71-96.
1980. The effects of grazing by gastropods and phys- Underwood, A. J., and S. J. Kennelly. 1990. Ecology of
ical factors on the upper limits of distribution of intertidal marine algae on rocky shores and subtidal reefs in tem-
macroalgae. Oecologia (Berlin) 46:201-213. perate Australia. Hydrobiologia 192:3-20.