AQUATIC CONSERVATION: MARINE AND FRESHWATER ECOSYSTEMS

Aquatic Conserv: Mar. Freshw. Ecosyst. 20: 452–463 (2010)

Published online 30 March 2010 in Wiley InterScience
(www.interscience.wiley.com). DOI: 10.1002/aqc.1111

A rapid biodiversity assessment methodology tested on intertidal

rocky shores

TIMOTHY D. O’HARAa,, PRUE F. E. ADDISONa, RUTH GAZZARDa, TRUDY L. COSTAb

and JACQUELINE B. POCKLINGTONb
a
Museum Victoria, GPO Box 666, Melbourne 3001, Australia
b
Zoology Department, University of Melbourne, Melbourne 3010, Australia

ABSTRACT
1. Conservation managers require biodiversity assessment tools to estimate the impact of human activities on
biodiversity and to prioritize resources for habitat protection or restoration. Large-scale programs have been
developed for freshwater ecosystems which grade sites by comparing measured versus expected species richness.
These models have been applied successfully to habitats that suffer from systemic pressures, such as poor water
quality. However, pressures in other habitats, such as rocky intertidal shores, are known to induce more subtle
changes in community composition.
2. This paper tests a biodiversity assessment methodology that uses the ANOSIM R statistic to quantify the
biological dissimilarity between a site being assessed and a series of reference sites selected on the basis of their
similar environmental profile. Sites with high R values for assemblage composition have an anomalous
assemblage for their environmental profile and are potentially disturbed.
3. This methodology successfully identified moderate to heavily perturbed sites in a pilot study on 65 rocky
intertidal sites in south-eastern Australia. In general, measures based on percentage cover (flora and sessile
invertebrates) were more sensitive than abundance (fauna). Copyright r 2010 John Wiley & Sons, Ltd.

Received 19 August 2009; Revised 23 December 2009; Accepted 27 January 2010

KEY WORDS: biodiversity assessment; ecosystem health; intertidal; invertebrates; algae; dissimilarity; multivariate; ANOSIM

INTRODUCTION test a biodiversity assessment methodology to rank the

condition of at least 60 rocky intertidal sites across 1100 km
Coastal marine habitats are affected by various human of coastline in the State of Victoria in temperate south-eastern
activities including habitat destruction and changed Australia. This posed a number of methodological challenges
hydrology (e.g. for sea walls, boat ramps, marinas, etc.); that would be typical of meso-scale projects in coastal
water pollution (e.g. nutrients, toxins, salinity, temperature); environments. First, there were no consistent quantitative
trampling or displacement by visitors; direct harvesting of measurements of environmental stressors across the study area.
organisms for food, bait, aquaria or curiosity; introduced Monitoring of pollutants or nutrient enrichment was spatially
species; and climate change (Suchanek, 1994; O’Hara, 2002). concentrated around a few sewerage outfalls and popular
In response, marine managers frequently request standardized beaches (O’Hara, unpubl. data). Measurement of recreational
biodiversity assessment indices of site condition for ‘state of use was sporadic at best (Addison et al., 2008). Second, the
environment’ reporting and to guide the allocation of scarce biological assemblages were potentially influenced by the range
management resources (Bailey et al., 2004). This information is of strong environmental (e.g. temperature, primary production,
often required at large spatial scales on a limited budget. wave exposure, geology) and biogeographical gradients that
Typically managers want to rank site condition across the were known to occur across the study area (O’Hara, 2000;
entire jurisdiction of their agency. O’Hara and Poore, 2000; Underwood and Chapman, 2007).
For example, the Natural Heritage Trust of the Australian Third, the intertidal assemblages were likely to be characterized
Government commissioned Museum Victoria to develop and by high spatial and temporal variability in species abundance,

*Correspondence to: Timothy D. O’Hara, Museum Victoria, GPO Box 666, Melbourne 3001, Australia. E-mail: tohara@museum.vic.gov.au

Copyright r 2010 John Wiley & Sons, Ltd.

 TESTING A RAPID BIODIVERSITY ASSESSMENT METHODOLOGY 453

particularly at small (o1 m) and regional (1–10 km) scales sites, which were quantitatively selected on the basis of their
(Archambault and Bourget, 1996; Underwood and Chapman, similar environmental profile. ANOSIM R is based on
1998; Fraschetti et al., 2005). similarity coefficients which can utilize abundance or
Consequently, without direct measurement of stressors, site presence/absence data and make no assumption about the
condition had to be assessed solely from the composition and/ direction of the assemblage change. Although ANOSIM R was
or abundance of the fauna and flora, assuming that a change developed as a test statistic to calculate the significance level of
will occur in the ecosystem in response to exposure to stressors multivariate differences between groups (Clarke, 1993), it is a
(Bailey et al., 2004; Pinedo et al., 2007). However, the variation useful comparative measure in its own right because it is
induced by human stressors had to be distinguished from the derived from ranked rather than absolute similarity measures
high environmental variability. The lack of stressor and then scaled to lie between –1 and 1 (Clarke and Warwick,
measurements precluded partitioning environmental and 2001; Anderson et al., 2008). For example, comparative
impact factors across a multivariate regression framework ANOSIM R measures are used to optimally split groups of
such as distance-based linear modelling (McArdle and samples in the linkage-tree procedure (Clarke et al., 2008).
Anderson, 2001). Moreover, pilot data suggested that a Here we systematically calculate ANOSIM R values between
single comprehensive multivariate analysis of all sites would each site and its group of selected reference sites with a similar
be unlikely to derive identifiable groups of sites with an environmental profile, using these values to rank sites with
affected assemblage because of the presence of many strong anomalous assemblages for further investigation.
environmental gradients across the large study area. This biodiversity assessment methodology, hereafter called
An alternative potential methodology was the reference MAVRIC (monitoring and assessment of Victoria’s rocky
condition approach (RCA) developed for freshwater systems intertidal coastline), was tested on both macro-invertebrate
that uses ‘sites’ as the basic sampling unit, factors out faunal and floral assemblages from 65 rocky intertidal sites in
environmental variation through the careful selection of south-eastern Australia, one of the largest intertidal surveys
reference (control) sites, and measures the difference in conducted in the southern hemisphere. Sites ranged from being
biological assemblage composition between test and reference putatively affected by sewage pollution (adjacent to outfalls) or
sites as the basis for a preliminary assessment of site condition trampling (popular recreational sites) to relatively inaccessible
(Bailey et al., 2004). The RCA forms the basis for large-scale sites with no identified site-specific human impacts.
programs to monitor freshwater systems in various countries,
including RIVPACS in the United Kingdom (Wright, 1995;
Clarke et al., 2003), AUSRIVAS in Australia (Simpson and
METHODS
Norris, 2000) and BEAST in Canada (Reynoldson et al., 1995).
However, there were potential problems in applying these
The MAVRIC biodiversity assessment methodology
freshwater methodologies to coastal systems. Existing RCA
implementations develop an expected taxon list for a test site The MAVRIC biodiversity assessment procedure has two basic
based on taxa found at the matching reference sites and steps: (1) selection of a set of reference sites for comparison
compute an observed over expected taxa (O/E) ratio for the with a test site on the basis of their shared environmental
test site, which is used as an index of site condition. Not only characteristics (termed ‘nearest-neighbour’ sites after Linke
does this probabilistic presence/absence approach require et al., 2004); and (2) calculation of ANOSIM R statistics for
species-rich systems (Marchant et al., 1997), it also assumes both biological and environmental data, using the test site as
that environmental degradation will cause a loss of taxa at a one group and the selected nearest-neighbour sites as the
site (or at least the loss of the ability to detect them). However, second group. These R values are a comparative measure of the
on rocky shores, impacts such as trampling or harvesting are multivariate assemblage difference between the test site and the
known experimentally to induce subtle species-specific nearest-neighbour reference set. High biological R values
responses in abundance rather than result in local extinction indicate that the assemblage is clearly distinct from those at
(Keough and Quinn, 1998, 2000). Models based on abundance the nearest-neighbour sites. The calculation of environmental R
are likely to be more sensitive than models based on values validates the reference site selection. High environmental
presence–absence (Hewitt et al., 2005). Moreover, at R indicates that the site was not well matched with the available
moderately polluted sites, nutrient enrichment can cause an reference sites (see below). This can be visualized as a series of
increase in species richness (Pearson and Rosenberg, 1978; ordinations (Figure 1). A set of reference sites nearest to the
Bishop et al., 2002). These difficulties can be avoided by using test site are selected from multi-dimensional environmental
a multivariate measure of the difference in assemblages space (Figure 1(a)). The test site is then compared with this set
between a test site and a selected group of reference sites. of nearest-neighbour sites both environmentally (Figure 1(b))
One solution is to use the residuals from a PCA analysis or and biologically (Figure 1(c)).
multivariate regression as the basis of an index of disturbance ANOSIM R values are calculated by subtracting the mean
(e.g. the CDI method, see Flåten et al., 2007). However, ranked similarity between pairs of the nearest-neighbour sites
residuals reflect model error in addition to natural variability from the mean ranked similarity between the test site and each
and human disturbance. PCA in particular is rarely nearest-neighbour site, and the result then scaled to lie between
appropriate for the non-linearities and zero-inflated data 1 and 1 (Clarke and Green, 1988). Negative values of R
common in ecological studies (Clarke and Warwick, 2001). indicate that the test site cannot be distinguished from the
Consequently it was decided to test a novel methodology nearest-neighbour sites. Positive values of R indicate that
that used the non-parametric ANOSIM R statistic (Clarke, the test site lies outside the range of variation encountered in
1993) to quantify the overall difference in assemblage the set of nearest-neighbour sites. It is important to note that,
composition between a test site and a series of reference although the magnitude of the positive R values gives an

Copyright r 2010 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 20: 452–463 (2010)
454 T. D. O’HARA ET AL.

Environmental data
Stress: 0.08 (Clarke and Ainsworth, 1993) or BEST (Clarke et al., 2008)
procedures can be used to find the combination of
Reference site environmental variables that best match the assemblage
Test site pattern (using the same non-parametric multivariate
framework as ANOSIM) by correlating ranked similarity
matrices generated from both sets of data for all reference sites
using the Spearman coefficient. The use of ranked similarity
coefficients here avoids the requirement in some other
methodologies (RIVPACS and AUSRIVAS) to divide
reference sites into discrete ecological groups, when in reality
they form a continuum along various environmental gradients
(a) (Bailey et al., 2004; Dauvin, 2007). It also avoids the need to
Environmental data Stress: 0.01 Faunal data Stress: 0.1 use ordinations (e.g. the ANNA model, Linke et al., 2004) or
cluster diagrams to match environmental predictors with the
faunal pattern which impose an artificial dimensionality on the
data (Clarke and Warwick, 2001).
Determining the number of nearest-neighbour sites to use
in the comparison with a test site is more problematic. Small
numbers of nearest-neighbour sites are likely to generate
(b) (c)
variable results reflecting the lack of site replication. On the
Figure 1. The MAVRIC concept. Reference sites that are other hand, increasing the number of sites will increase the
environmentally most similar to a test site are identified using environmental heterogeneity within the nearest-neighbour
dissimilarity coefficients. The test site is then compared
group which in turn will alter the likelihood that a test site
environmentally and biologically with the selected reference sites
using the ANOSIM R statistic, which for this purpose subtracts the will be considered distinct. This can be investigated empirically
mean ranked similarity between pairs of reference sites from that by treating each reference site in turn as the test site and
between the test site and each reference site and then is scaled to lie calculating the Environmental R for increasing numbers of
between 1 and 1. This can be visualized as a series of ordinations. In nearest-neighbour sites. The optimal solution has minimum
this example (a) eight reference sites nearest environmentally to the test
site (Pickering Point) are identified; the test site (b) falling within the standard deviation (variability) and mean (distinctiveness) of
range of variation of these reference sites environmentally (ANOSIM environmental R across all these relatively undisturbed sites.
R 5 0.22) but (c) having a distinct faunal assemblage (R 5 0.74). A similar procedure was used to test the effect of varying
data transformations of the biological abundance data
indication of how distinct the test site is from the nearest- (binary, fourth-root, log(x11), square root, and none) and
neighbour group, negative values cannot be interpreted in the combinations of environmental variables. Finally, tests were
same way, as these latter results are influenced by made to determine whether models based on selecting nearest-
compositional heterogeneity (‘clumpiness’) within the neighbour sites using the ‘best’ match of environmental
nearest-neighbour group itself rather than the relationship variables performed better (i.e. lower mean and standard
between the test and nearest-neighbour sites. deviation of environmental R) than selecting sites based (1)
It is more informative to consider the test/nearest- solely on latitude/longitude, or (2) chosen at random without
neighbour site relationship as falling into four groups based replacement.
on whether the R statistic is positive or negative for
environmental and assemblage data. Negative R values for Species analyses
both indicate the test site has a suite of environmental factors
and biological composition within the variation found in the The percentage contribution of each species to the assemblage
group of nearest-neighbour sites. Positive R values for Bray–Curtis dissimilarity between test and nearest-neighbour
environmental data indicate that although the best-matched sites can be determined by using the SIMPER (similarity of
nearest-neighbour sites were chosen for comparison, the test percentages) procedure (Clarke, 1993). The percentage
site still has a distinct environmental signature. The test site contributions of individual species were aggregated into
may be unique, or inadequate reference sites surveyed. These broad taxonomic groups for interpretability, by (1) assigning
test sites can either have a positive or negative assemblage R a positive value to each species contribution if the average
depending on whether the unusual environmental conditions abundance was greater at test sites and a negative value if
are reflected in the biological assemblage. Sites with positive greater at the reference sites, and (2) summing the overall
assemblage R and negative environmental R have an unusual contributions for each taxonomic group. The groups were (a)
assemblage despite their typical environmental conditions. for sessile taxa: blue-green, green, brown and red algae;
These sites, particularly those with relatively high R (e.g. lichens; bivalves; serpulid and spionid polychaetes; ascidians,
40.5), are the most interesting from a management and (b) for motile taxa: anemones; chitons, bivalves,
perspective and potentially disturbed by human activities. gastropods, pulmonates; barnacles, crabs; seastars; ascidians.

Selection of reference sites Data sets

The accurate matching of test to nearest-neighbour sites is A survey of intertidal rock platforms was conducted along the
dependent on the selection of environmental predictors and the 1100 km Victorian coast in south-eastern Australia (Figure 2).
number of nearest-neighbour sites chosen. The BIOENV Fifty-eight reference sites were surveyed between February and

Copyright r 2010 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 20: 452–463 (2010)
 TESTING A RAPID BIODIVERSITY ASSESSMENT METHODOLOGY 455

Figure 2. Map of the survey sites and bioregions defined by the Interim Marine and Coastal Regionalisation for Australia (IMCRA) (Thackway and
Cresswell, 1998). Named sites are discussed in the text.

May (autumn) 2005. The sites were selected from prior field (e.g. Xenostrobus pulex and Brachidontes rostratus), gastropod
experience on the basis of geographic spread, size and limpets (e.g. Cellana tramoserica and Siphonaria diemenensis)
accessibility. Inaccessible (relatively unimpacted) sites were and gastropod snails (e.g. Chlorodiloma odontis and Bembicium
surveyed where possible, including some on offshore islands nanum); low shore — dominated by algae (e.g. Hormosira
and others without ready vehicular and/or pedestrian access. banksii). The kelp-dominated fringes at the reef edge and rock
Several sites were within marine protected areas established in pools (40.1 m deep) were considered subtidal and not surveyed.
November 2002. On the more accessible sections of the The abundances of each macro-invertebrate species
coastline, sites were chosen that were furthest from (44 mm) were counted for each quadrat (the ‘faunal’
pedestrian access points. Eleven of these sites were re- dataset). Dense aggregations of barnacles, mussels and
sampled between February and March 2006 in order to littorinid molluscs were occasionally estimated from four
measure temporal variability within reference sites. subquadrats (62.5  62.5 mm). The percentage cover of
An additional eight test sites were selected for their macroalgae and sessile invertebrates (particularly mussels,
vulnerability to (a) pollution from sewerage outfalls, or barnacles, tubicolous polychaetes, ascidians) were also
(b) disturbance by anthropogenic recreational use. One site estimated for each quadrat using a 50 point grid (the ‘cover’
(Boags Rocks) was surveyed immediately adjacent to a major dataset). Some species that were difficult to identify in the field
sewerage outfall and another two were 0.5 (Boags East) and (particularly filamentous and encrusting algae) were aggregated
3.2 km (Fingals Beach) to the east. This outfall discharges into higher taxa. The mean quadrat abundance and/or cover
370 ML day 1of effluent from the Eastern Treatment Plant, for each resulting taxon were then calculated for each site.
which treats 42% of the sewage from Melbourne, the capital
city of Victoria (Newell et al., 1999). A fourth putatively
polluted site was surveyed adjacent to a smaller outfall at
Environmental predictors
Pyramid Rock, which discharges 0.6 ML day 1 of effluent
from Phillip Island. Consistent data for coastal visitation do In order to calculate environmental data for each site, a
not exist across Victoria, however, four test sites were chosen geographic information system (GIS) was constructed using
on the basis of their likely (Point Grey, Barwon Heads, Point ArcGIS 9.0 software (ESRI, 2004) from (a) environmental
Lonsdale) or known (Sorrento, see Addison et al., 2008) high data derived from oceanographic and climate models, (b) geo-
rates of recreational use. Test sites were surveyed between referenced, ortho-rectified and mosaicked aerial (1.2 m pixel)
February 2005 and April 2007. and satellite (1 m pixel) imagery, and (c) survey data.
While it would have been preferable to use quantitative
in situ rather than modelled measurements of many variables
Biological data
(e.g. for wave energy see Helmuth and Denny, 2003), such data
Sites were surveyed during daylight hours, on days where the sets were not available for the Victorian coasts and were
predicted low tide was between 0 and 0.4 m above datum. At judged to be prohibitively expensive to collect. Consequently,
each site, one 50 m wide location was chosen, and two 10 m wide the best available modelled or qualitative data sets were used.
transects were surveyed within each location. Along each However, the use of environmental factors in the methodology
transect, a randomly placed group of five quadrats was not to build a comprehensive predictive model of how
(50  50 cm) were surveyed within three 5  5 m areas, one in these factors influence assemblage composition, but only to
each of the high, mid and low shore levels. In the absence of improve the selection of nearest-neighbour sites. The non-
accurate height-above-shore data (see below), shore levels were parametric environmental selection methodology used here
determined visually by the following criteria (Bennett and Pope, (BIOENV, see above) selects the set of variables that ‘best
1953): high shore — dominated by littorinid snails (mainly match’ the overall assemblage pattern regardless of their
Austrolittorina unifasciata); mid shore — dominated by mussels resolution or collinearity.

Copyright r 2010 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 20: 452–463 (2010)
456 T. D. O’HARA ET AL.

Oceanographic data were derived from the CSIRO Atlas of environmental data. This latter metric is calculated by
Regional Seas (CARS2000) data sets (Ridgway et al., 2002; linearly scaling the data to lie between 0 and 1 (using the
Commonwealth of Australia, 2005) and included mean sea- minimum and maximum data values), and then scaling the
surface temperature, primary production, salinity, phosphate, resultant Euclidean distance to also lie between 0 and 1 by
nitrate, silica, and oxygen. Primary production was derived from dividing by the square root of the number of variables.
satellite observations of ocean colour data, a surrogate of
phytoplankton concentrations in surface waters (Commonwealth
of Australia, 2005). Data for each survey site were obtained by
RESULTS
spline interpolation from the original raster datasets (which
included coastal areas) with resolutions of 0.02–0.0431. Root Assemblage patterns
mean square wave height (Hrms) and wave period (T) were
derived from three years (30/07/2002 to 30/07/2005) of diurnal In total 96 macro-invertebrate species were identified and
predictions made by a wind-driven offshore model with a counted from the 58 reference sites, including six anemones,
resolution of 0.1251 latitude/longitude (Australian Bureau of eight chitons, seven bivalves, one opisthobranch, 42
Meteorology), combined into a wave power variable using the prosobranch gastropods, four pulmonates, nine barnacles,
formula P 5 (rG2/4p)(H2rms/8)T, where r is the density of water nine crabs, one brachiopod, five seastars, two sea cucumbers,
and G the gravitational constant (Leigh et al., 1987), and one brittlestar and one ascidian. An MDS ordination of site/
extrapolated to the coastline using ordinary Kriging with the mean species abundance (Figure 3(a)) showed embayment sites
semi-variogram fitted to a spherical model. dispersed to the right, largely differentiated from those on the
To determine whether the site faunal assemblage was open coast to the left. The open coast sites formed a broad
influenced by a species–area effect (Schoener, 1976), the area geographical gradient, with those from the east (Twofold and
of intertidal reef was calculated in three ways, the size of the Flinders bioregions; Thackway and Cresswell, 1998) forming
immediate reef (bounded by significant areas of sand or mud), distinct groups at the bottom of the ordination. Sites from the
the amount of available reef within a 1 km radius, and the two western bioregions were not as clearly differentiated.
amount of reef within a 10 km radius. These values were SIMPER analyses revealed that the distinction between
calculated by manually digitizing intertidal reef from aerial embayments and open coast was primarily driven by a shift in
and satellite imagery (1 m pixel size) and ground-truthing areas barnacle and gastropod species. Along the open coast
of uncertainty using methodologies based on Congleton et al. Chamaesipho, Chthamalus and Catomerus were the most
(1999) and Zharikov et al. (2005). The amount of reef within a common barnacle genera (contributing 16.4% of the total
radius was calculated using the ‘Extract by Circle’ analysis in dissimilarity), whereas Tetraclitella and Elminius were the most
ArcGIS 9.0 (ESRI, 2004). common in embayments (6.6%). The littorinids Austrolittorina
The rock type for each site was categorized from geological and Afrolittorina, generally abundant in the wave splash zone
site descriptions (Bird, 1993) and then transformed into an along the open coast, were much less common in embayments
ordinal variable using the qualitative Moh’s scale of hardness (7.8%). Embayments also had fewer limpets (e.g. Cellana,
(in general, granite 4 basalt 4 calcarenite 4 limestone 4 Notoacmea, Siphonaria, in total 16.8%) but more trochids (e.g.
sandstone). Sites were classified in the field into two additional Austrocochlea, 6.1%). The difference in east (Twofold,
binary variables, reef type (boulder reefs or flat rock platforms)
and coastal type (bay or open coast). The aspect of each reef
Stress: 0.16
was derived from the perpendicular angle to the shoreline (e.g. Otway
0 5 north, 180 5 south) determined from aerial imagery. Central Victoria
At the smallest scale, reef rugosity, latitude and longitude Flinders
were measured for each quadrat. Rugosity was measured by Twofold Shelf
trailing a 3 mm wide chain over the surface of the reef under Victorian Embayments
two perpendicular sides of each quadrat and measuring the
total extended length once removed. Latitude and longitude
were taken from portable Garmin 72 GPS units. Ordinary
GPS-derived altitude was found to be inaccurate at the scales
required (0–3 m above sea level) and not used. For these (a)
factors, quadrat values were averaged (mean) to calculate a site Stress: 0.21
centroid value. Otway
Central Victoria
Flinders
Data transformations and similarity coefficients
Twofold Shelf
The abundance data were transformed to ensure that analyses Victorian Embayments

were not dominated by a few common data. The spread of

mean abundances (0–3,504) and mean percentage cover (0–17)
suggested a severe (fourth-root) and a moderate (square-root)
transformation, respectively. However, ANOSIM R values
derived from binary, fourth-root, log(x11), square root, and (b)
untransformed data were generated for comparison (see Figure 3. MDS ordinations of (a) faunal abundance and (b)
above). All analyses used the Bray–Curtis similarity for percentage cover from the 58 reference sites, grouped by bioregion
biological data and double-scaled Euclidean distance for (see Figure 2).

Copyright r 2010 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 20: 452–463 (2010)
 TESTING A RAPID BIODIVERSITY ASSESSMENT METHODOLOGY 457

Flinders) and west (Otway, Central Victoria) along the open 0.6

coast was driven principally by differences in relative 0.4

Environmental ANOSIM R
abundance of a range of species rather than biogeographical 0.2
turnover. Barnacles (25.0%) and ascidians (Pyura, 3.3%) were 0.0
generally more common in the east and littorinids
-0.2
(Austrolittorina, Afrolittorina, 6.1%) in the west. Species
-0.4
contributing more than 1% to the dissimilarity between Null Model (100 iterations)
Cape Paterson
eastern and western bioregions that were known to have a -0.6 Cat Bay

distributional limit within Victoria included the western -0.8

5 10 15 20 25 30 35 40
gastropods Nerita atramentosa (2.2%), Notoacmea mayi (a) Number of nearest-neighbour sites
(1%), Chlorodiloma odontis (1%), C. adelaidae (1%) and the 0.7
eastern Nerita melanotragus (2.9%) and the chiton

Environmental ANOSIM R
0.5
Sypharochiton pelliserpentis (1%).
In total 59 species were counted as percentage cover, 0.3
including two mytilid bivalves, one ascidian (Pyura), two
aggregating polychaetes (the spionid Boccardia and the 0.1

serpulid Galaeolaria), 10 green algae, 26 brown algae, 15 red -0.1

Null Model (100 iterations)
Standard deviation
algae, two blue-green algae (Rivularia) and one lichen Mean

(Lichenia). An MDS ordination of these data (Figure 3(b)) -0.3

5 10 15 20 25 30 35 40
showed less bioregional differentiation. In particular, (b) Number of nearest-neighbour sites
embayment sites were not differentiated, even in the 0.7
equivalent 3D ordination (not shown).
0.5
Faunal ANOSIM R

Reference site selection 0.3

MAVRIC model
Standard deviation
The BIOENV analysis selected six environmental predictors as 0.1
Mean

best matching the overall biotic pattern for all reference sites: -0.1
Null Model (100 iterations)
Standard deviation
rock hardness, wave power, primary production, salinity, Mean

phosphate and the binary bay/open coast variable (Spearman -0.3

5 10 15 20 25 30 35 40
r 5 0.63 for macro-invertebrates, r 5 0.32 for cover). The (c) Number of nearest-neigbour sites
results were slightly improved for cover by adding oxygen
(r 5 0.34) rather than salinity and phosphate, however, to Figure 4. Trends in ANOSIM R values with increasing number of
ensure comparability, the faunal set of predictors were used for reference sites. (a) Two example sites showing how they can become
increasingly (higher R) or decreasingly environmentally differentiated as
both data sets. Predictors based on latitude, longitude, reef more nearest-neighbour sites are included in the model. (b) The mean
area, sea surface temperature and rugosity were not and standard deviation of environment R values for all 58 reference
emphasized in any of the top BIOENV groupings. sites. Optimal models have both low standard deviation and mean. (c)
The number of nearest-neighbour sites to be used in the For faunal abundance, the MAVRIC model (selecting reference sites
based on their similar environmental characteristics) performed better
model was determined empirically by treating each (relatively
than a null model (selecting sites at random) for less than 20 sites.
undisturbed) reference site in turn as a test site, and
investigating the behaviour of the MAVRIC model. Sites
responded differently to increased numbers of nearest- R had lower standard deviation for 4–9 sites and lower mean
neighbour sites (Figure 4(a)). Some sites (e.g. Cat Bay) were for 4–40 sites (not shown). A spatial model, based solely on
well matched environmentally to a few others, but then latitude and longitude as environmental predictors (not
become increasingly distinct as more environmentally remote shown), also had higher standard deviation and mean of R
sites were added to the comparison, increasing the than the MAVRIC model for both faunal abundance and
environmental ANOSIM R. Other sites (e.g. Cape Paterson) percentage cover. The spatial model performed more poorly
were distinct initially but gradually became indistinguishable than the null model for large numbers of nearest-neighbour
as the size of the comparative reference group was increased. sites (420).
Others (not shown) were highly volatile initially but trended to A similar procedure was used to assess the affect on
a stable value of R with increased numbers of sites or were ANOSIM R of varying the severity of the faunal and cover
relatively stable throughout. The optimal number of nearest- data transformation from none, square root, log(x11), fourth
neighbour sites to be used was determined from the lowest root to binary. Only the binary transformation had a notable
standard deviation of environmental R across all 58 reference effect on the standard deviation and mean of the R values (not
sites, i.e. the number where most reference sites were not shown), generating higher values (i.e. a poorer model) than
differentiated environmentally from their nearest-neighbour abundance transformations.
group. This was eight sites for this data set (Figure 4(b)).
Faunal R values for all reference sites were compared with
Test sites
a null model (Figure 4(c)), where nearest-neighbour group for
each site was selected at random and the resulting R averaged The environmental and biotic R were generated from eight
over 100 iterations. The MAVRIC model had lower standard nearest-neighbour sites for all 58 reference and eight putative
deviation than the null model for fewer than 20 nearest- polluted or trampled test sites (Figure 5). Sixty of the sites had a
neighbour sites and lower mean for 4–14 sites. MAVRIC cover negative environmental R indicating a good environmental match

Copyright r 2010 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 20: 452–463 (2010)
458 T. D. O’HARA ET AL.

1.0 sites (see Q3 on Figure 5(a)). Relatively few of these sites had
Q1 Pickering Settlement Q2 large positive values of faunal or cover R (Q1). For example,
0.8 Point Point only four had a faunal R over 0.5 (Figure 5(a)) and six had a
Boags Rocks Doughboy Is cover R over 0.5 (Figure 5(b)).
0.6 Urquharts Bluff The eight putative polluted or trampled test sites all had
negative environmental R but varying faunal and cover R
0.4 (Figure 5). Two polluted (Boags Rocks R 5 0.7, Boags East
R 5 0.1) and one trampled site (Barwon Heads R 5 0.1), had
Faunal ANOSIM R

Point Lonsdale
0.2 Boags East positive faunal R; and three polluted (Boags Rocks R 5 0.98,
Boags East R 5 0.77, Fingals R 5 0.64) and one trampled site
0.0
Barwon Heads (Sorrento R 5 0.38) had positive cover R.
-0.2 The taxa that were driving the large (40.5) faunal and
Fingals
cover R values were investigated using SIMPER analyses
-0.4 Sorrento (Tables 1 and 2). For the putatively-heavily polluted site at
Pyramid Boags Rocks, there was a clear decline in most faunal and
-0.6 Rocks Point Grey cover groups compared with the nearest-neighbour sites, most
notably gastropods, barnacles, brown (Hormosira) and red
-0.8 algae, and an increase in spionid polychaetes (Boccardia),
Q3 Q4
green algae (Ulva), anemones (particularly Aulactinia), and
-1.0
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 bivalves (Brachidontes, Lasaea). The nearby site of Boags East
(a) Environmental ANOSIM R also had reduced brown algal cover, but lacked the spionids
and had above average green (Ulva) and red (foliose coralline
1.0 Q1 Gabo Is Q2 and fleshy turfing) algae. A few kilometres to the east, at
Boags Rocks Fingals Beach, the dense algal turf included the brown alga
0.8 Boags East
Capreolia implexa, foliose coralline and fleshy red algae (e.g.
Childers Cove Fingals
Laurencia). The dominant macro-brown alga Hormosira was
0.6
Secret Beach Cape Paterson present, although with a low percentage cover.
Sorrento The three reference sites with high cover ANOSIM R
0.4
Point varied. Childers Cove and Secret Beach lacked the typical
Cover ANOSIM R

Lonsdale
0.2 Hormosira cover, the former site having above average green
Pyramid Rock alga Caulerpa fragilis and the latter having above average
0.0 Laurencia, foliose corallines, encrusting corallines and Ulva.
Point Grey
A repeat visit to Secret Beach indicated that Hormosira was
-0.2
present but only in pools scattered among boulders, none of
-0.4 which were in the paths of the surveyed transects. Cape
Barwon Heads Paterson on the other hand differed in having fewer mussels
-0.6 (Xenostrobus) and a greater diversity of brown algae (including
Cystophora, Scytosiphon and Halopteris) that were generally
-0.8 restricted to deeper rock pools at other sites.
Q3 Q4
The three reference sites with high faunal ANOSIM R also
-1.0
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 varied. Settlement Point, an embayment site surrounded by
(b) Environmental ANOSIM R muddy sediments, had an odd assemblage with relatively few
gastropods, but higher numbers of crabs, chitons and
Figure 5. Plot of ANOSIM R values for 58 reference (solid circles) and echinoderms (and a few specimens of other unusual groups
eight test sites (open squares) generated by the MAVRIC procedure from such as brachiopods). Pickering Point had relatively few
environmental predictors and (a) faunal abundance, and (b) percentage molluscs or barnacles, but more anemones and seastars.
cover. Quadrant 1 (Q1) 5 Sites that are well matched environmentally Urquhart Bluff had reduced numbers of anemones,
with their nearest-neighbour sites but have an anomalous assemblage.
Q2 5 Sites with an anomalous environmental profile and assemblage. pulmonates and barnacles but a typical gastropod assemblage.
Q3 5 Sites with an environmental profile and assemblage that cannot be
distinguished from their nearest-neighbour set. Q4 5 Sites with an
anomalous environmental profile but with an assemblage that cannot Temporal differences
be distinguished from the nearest-neighbour sites.
A comparison of the test sites with a subset of 11 reference sites
from central Victoria surveyed for two consecutive years
between the test site and their nearest-neighbour group. The showed some stability in the pattern for both percentage cover
exceptions were reference sites in embayments (e.g. Doughboy and faunal abundance between surveys (Figure 6). A few sites
Island in Corner Inlet) and the far-east of Victoria (e.g. Gabo had a substantially changed R (40.3) between year 1 and year
Island). These were uncommon habitats within the study area and 2 including (a) for fauna, the test site Boags East ( 0.4) and
consequently were not well matched environmentally to the the reference sites Point Addis (10.6), Point Roadknight
selected group of eight nearest-neighbour sites. ( 0.4), Flat Rocks (10.6); and (b) for cover, the reference sites
Of the sites with negative environmental R, 36 also had Cape Paterson ( 0.8) and Flat Rocks (10.5). However, with
negative faunal R indicating a good correspondence between the exception of percentage cover at Cape Paterson, the sites
environmental predictors and the faunal assemblage for these with negative environmental R and assemblage R40.5

Copyright r 2010 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 20: 452–463 (2010)
 Table 1. SIMPER percentage contribution to the assemblage dissimilarity of cover between selected test and eight nearest-neighbour sites accumulated into major biotic
components.
Site Site type ANOSIM R Accumulated SIMPER percentage contribution

Environment Cover Blue-green Lichens Green Brown Red Bivalves Serpulid Spionid Ascidians
algae algae algae algae polychaetes polychaetes

Boags Rocks Test site (polluted) 0.38 0.98 1.4 4.2 9.2 26.4 12.4 3.7 6.0 19.3 4.2
Boags East Test site (polluted) 0.38 0.77 1.3 3.8 20.0 9.8 21.7 6.7 4.5 0.0 5.3
Fingals Test site (polluted) 0.34 0.64 0.7 2.9 10.0 4.7 26.4 5.2 2.6 0.0 0.0

Copyright r 2010 John Wiley & Sons, Ltd.

Childers Cove Reference site 0.53 0.63 3.3 3.4 7.5 26.9 13.9 1.0 7.8 0.0 0.0
Cape Paterson Reference site 0.05 0.53 4.3 4.3 2.7 15.0 3.5 15.1 3.1 0.0 0.5
Secret Beach Reference site 0.13 0.58 1.3 8.0 7.8 25.3 18.5 11.7 3.3 0.0 3.0
Positive percentage contributions indicate that the average percentage cover (square-root transformed) is greater at the test site, negative contributions are the reverse. Only sites with a negative
Environmental R and Cover R40.5 are listed.

Table 2. SIMPER percentage contribution to the assemblage dissimilarity of faunal abundance between selected test and eight nearest-neighbour sites accumulated into major
biotic components.
Site Site type ANOSIM R Accumulated SIMPER percentage contribution
TESTING A RAPID BIODIVERSITY ASSESSMENT METHODOLOGY

Environment Fauna Anemones Chitons Bivalves Gastropods Pulmonates Barnacles Crabs Seastars Ascidians

Settlement Point Reference site 0.16 0.78 2.8 4.1 5.3 20.8 3.9 3.1 10.1 1.6 0.5
Pickering Point Reference site 0.22 0.74 2.8 4.1 0.0 39.2 8.5 10.7 3.1 3.9 0.0
Boags Rocks Test site (polluted) 0.38 0.72 8.5 3.6 4.7 32.4 8.5 9.8 3.7 3.2 4.0
Urquhart Bluff Reference site 0.14 0.66 8.0 0.8 0.7 0.4 4.1 5.4 0.7 7.0 2.2
Positive percentage contributions indicate that the average abundance (double square-root transformed) is greater at the test site, negative contributions are the reverse. Only sites with a negative
Environmental R and Faunal R40.5 are listed.
459

Aquatic Conserv: Mar. Freshw. Ecosyst. 20: 452–463 (2010)

460 T. D. O’HARA ET AL.

Year 1 Year 2
1.0 1.0

0.8 Boags Rocks 0.8 Boags Rocks

0.6 0.6
Boags East
Point Addis
0.4 0.4
Barwon Heads
Faunal ANOSIM R

Faunal ANOSIM R
Point Roadknight
Flat Rocks
0.2 Barwon Heads 0.2 Boags
East
0.0 Fingals 0.0 Sorrento
Sorrento Fingals
Point Roadknight
-0.2 Point Grey -0.2
Point Addis
Pyramid
-0.4 -0.4 Rock Point Grey
Flat Rocks
Pyramid Rock
-0.6 -0.6

-0.8 -0.8

-1.0 -1.0
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
(a) Environmental ANOSIM R (b) Environmental ANOSIM R
Year 1 Year 2

1.0 Boags Rocks 1.0 Boags Rocks

Boags East
Boags East
0.8 0.8
Fingals
Pyramid Rock
Pyramid Rock
0.6 0.6
Fingals Cape Paterson
0.4 0.4
Cover ANOSIM R

Cover ANOSIM R

Sorrento Sorrento Flat Rocks

0.2 0.2

0.0 0.0
Point Grey
-0.2 Flat Rocks -0.2
Barwon Cape Paterson
-0.4 Barwon
Point Grey -0.4 Heads
Heads
-0.6 -0.6

-0.8 -0.8

-1.0 -1.0
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
(c) Environmental ANOSIM R (d) Environmental ANOSIM R

Figure 6. Comparison of (a, b) faunal and (c, d) cover ANOSIM R between consecutive survey years for 11 reference and seven test sites in central
Victoria. Open squares represent test sites and solid circles reference sites.

remained the same for the two years. A SIMPER analysis of sites that have an assemblage that was clearly distinct from the
the change in percentage cover at Cape Paterson (not shown) variation shown by the selected nearest-neighbour sites. This
indicated that a more typical assemblage was surveyed in Year system has advantages over existing meso- to large-scale RCA
2 than the unusual assemblage in Year 1 (see above), with programs in not relying on the artificial categorization of sites
increased cover of the brown alga Hormosira and reduced into habitat groups, being able to detect changes in abundance
cover of rock-pool brown algae being recorded. rather than testing for species gain or loss, and not making
assumptions about the direction of assemblage change.
This study successfully identified anomalous assemblages at
sites that were putatively identified as moderately to heavily
DISCUSSION polluted by sewage. One site in particular (Boags Rocks) had a
very distinct assemblage measured using either faunal
The RCA as an exploratory tool
abundance or floral/faunal percentage cover. Other
This study tested a rapid RCA-type biodiversity assessment comparative studies have described reduced brown algal
methodology that used the ANOSIM R statistic to quantify (Hormosira) cover and mats of the spionid worm Boccardia
the biological and environmental dissimilarity between a site (Brown et al., 1990; Bellgrove et al., 1997) at this site. Two
being assessed and a series of reference (nearest-neighbour) adjacent sites (Boags East and Fingals) also had very distinct
sites objectively selected on the basis of their similar assemblages based on floral/faunal percentage cover but not
environmental profile. These dissimilarity results were then faunal abundance. Conversely, assemblages at sites known to
compared with a threshold (for example R40.5) to identify be subject to medium to high levels of visitor pressure were not

Copyright r 2010 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 20: 452–463 (2010)
 TESTING A RAPID BIODIVERSITY ASSESSMENT METHODOLOGY 461

distinguished, although that may be an artefact of the spatial composition, only that sites with similar values for these
scale of the sampling design (see below). variables also share a relatively consistent flora and fauna.
Apart from the known polluted sites, sites with ‘distinct’
assemblages fell into two other groups. The first group was
characterized by large environmental as well as assemblage Applications
dissimilarity. They represented unique or marginal habitats
The methodology tested here is suitable for meso- to large-
which were not well matched environmentally to other sites.
scale assessments of ecosystem condition, particularly for
These environmental ‘outliers’ (e.g. Doughboy Is, Gabo Is)
systems with strong ecological and biogeographic gradients,
should be removed from operational implementations if
and/or when there is an absence of quantitative impact data.
additional reference sites cannot be found.
The methodology will identify sites that have anomalous
The second group had distinct assemblages even though
assemblages compared with a selected group of sites with
a priori they were not understood to be impacted by
similar environmental profiles. This is a rapid biodiversity
anthropogenic activities and were environmentally well
assessment methodology that will identify moderately to
matched with their nearest-neighbour sites. These sites
heavily perturbed sites at relatively low cost, particularly if
clearly warrant further study. For example, the depauperate
careful attention is given to the scale of sampling compared
faunal assemblage at Pickering Point warrants further
with the scale of expected impact. Anomalous sites can then be
management attention, as, although it is in a marine park,
investigated further to determine the underlying causes of the
the nearby Merri River estuary has been extensively modified
measured dissimilarity. Given the inadequate funding of many
by human activities, the Warrnambool city sewage outfall is
environmental and conservation agencies, the focus on
only 500 m to the west, and during summer the site is subject to
moderately to severely affected sites is unlikely to be an
heavy visitation (Parks Victoria, 2006).
impediment to management action (Bottrill et al., 2008).
It is important to emphasize that this approach is
A database (Microsoft Access 2007r) with the biological/
exploratory, identifying assemblage outliers, those sites that
environmental data, and software to run the calculations, is
vary more than expected from background variation. It should
available from the corresponding author on request.
be used as an exploratory management tool. It does not
identify causal relationships between assemblage composition
and anthropogenic threats.
ACKNOWLEDGEMENTS

Model refinement The project was funded by grants from the Australian
In this study, assemblages at sites known to be subject to National Heritage Trust (NHT Project no: 202244), Parks
medium to high levels of trampling or low levels of sewage Victoria, and the Victorian Environmental Protection
were not distinguishable from the nearest-neighbour references Authority. Special thanks to Dr Richard Marchant (Museum
sites. Other studies have reported similar or variable findings. Victoria) for many statistical discussions and explaining the
The lack of sensitivity in this study may be related to the underlying methods of RIVPACS and AUSRIVAS models;
definition of a site (50 m of coastline) that masks the localized the late Clarrie Handreck (Marine Research Group of
scale or spatially heterogeneous effect of these impacts Victoria) for his invaluable help is selecting field sites and
(Keough and Quinn, 2000; Bishop et al., 2002). For example, identifying problematic animals; Liz Greaves, Rebecca Koss
even though repeatedly trampled ‘paths’ could be clearly seen and Anna McCallum (Museum Victoria) for assisting with
at Sorrento they were not of sufficient scale to alter average field work; Dr Anthony Boxshall and numerous rangers (Parks
assemblage composition across the whole site. If small-scale Victoria) for facilitating access to remote sites; and two
impacts are of management interest, smaller sampling units anonymous reviewers for their insightful comments.
may be required.
This survey measured both percentage cover and faunal
abundance. Both have advantages and disadvantages. Faunal REFERENCES
abundance was better matched environmentally to the
available predictors. However, it appeared relatively
Addison PFE, Koss RS, O’Hara TD. 2008. Recreational use
insensitive to anthropogenic disturbance, with only the of a rocky intertidal reef in Victoria: implications for
highly polluted test site at Boags Rocks being identified as ecological research and management. Australasian Journal
distinct. Cover showed less obvious environmental pattern, of Environmental Management 15: 169–179.
possibly because fewer taxa could be consistently identified to Anderson ML, Gorley RN, Clarke KR. 2008. PERMANOVA1
species in the field (particularly encrusting coralline and for PRIMER: Guide to Software and Statistical Methods.
filamentous algae). Conversely, cover appeared more PRIMER-E: Plymouth, UK.
sensitive to anthropogenic disturbance, particularly the loss Archambault P, Bourget E. 1996. Scales of coastal heterogeneity
of brown macro-algae such as Hormosira. Abundance models and benthic intertidal species richness, diversity and
performed better than those based on presence–absence. abundance. Marine Ecology Progress Series 136: 111–121.
Bailey RC, Norris RH, Reynoldson TB. 2004. Bioassessment
The selection of nearest-neighbour sites using a suite of
of Freshwater Ecosystems: using the Reference Condition
environmental variables, chosen with the BIOENV procedure, Approach. Kluwer Academic Publishers: Boston, USA.
resulted in superior models (lower overall assemblage Bellgrove A, Clayton MN, Quinn GP. 1997. Effects of
variation) than models based on spatial coordinates alone or secondarily treated sewerage effluent on intertidal macro-
random selection of sites. This does not imply that these algal recruitment processes. Marine and Freshwater Research
environmental variables alone are driving assemblage 48: 137–146.

Copyright r 2010 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 20: 452–463 (2010)
462 T. D. O’HARA ET AL.

Bennett I, Pope EC. 1953. Intertidal zonation of the exposed Keough MJ, Quinn GP. 2000. Legislative vs. practical
rocky shore of Victoria, together with a rearrangement of protection of an intertidal shoreline in southeastern
the biogeographic provinces of temperate Australian shores. Australia. Ecological Applications 10: 871–881.
Australian Journal of Marine and Freshwater Research 4: Leigh Jr EG, Paine RT, Quinn JF, Suchanek TH. 1987. Wave
105–159. energy and intertidal productivity. Proceedings of the
Bird CF. 1993. The Coast of Victoria: The Shaping of Scenery. National Academy of Sciences of the United States of
Melbourne University Press: Victoria. America 84: 1314–1318.
Bishop MJ, Underwood AJ, Archambault P. 2002. Sewage Linke S, Norris RH, Faith DP, Stockwell D. 2004. ANNA: a
and environmental impacts on rocky shores: necessity of new predictive method for bioassessment programs.
identifying relevant spatial scales. Marine Ecology Progress Freshwater Biology 50: 147–158.
Series 236: 121–128. Marchant R, Hirst A, Norris RH, Butcher R, Metzeling L,
Bottrill MC, Joseph LN, Carwardine J, Bode M, Cook C, Tiller D. 1997. Classification and prediction of
Game ET, Grantham H, Kark S, Linke S, McDonald- macroinvertebrate assemblages from running waters in
Madden E et al. 2008. Finite conservation funds mean triage Victoria, Australia. Journal of the North American
is unavoidable. Trends in Ecology and Evolution 24: 183–184. Benthological Society 16: 664–681.
Brown B, Davies SA, Synnot RN. 1990. Long-term McArdle BH, Anderson MJ. 2001. Fitting multivariate
monitoring of the effects of treated sewage effluent on the models to community data: a comment on distance-based
intertidal macroalgal community near Cape Schanck, redundancy analysis. Ecology 82: 290–297.
Victoria, Australia. Botanica Marina 33: 85–98. Newell B, Molloy R, Fox D. 1999. Environmental Impact
Clarke KR. 1993. Non-parametric multivariate analyses of Assessment and Review of Effluent Disposal Options for
changes in community structure. Australian Journal of Eastern Treatment Plant. Final Report June 1999. CSIRO
Ecology 18: 117–143. Environmental Projects Office, Wembley, Western Australia.
Clarke KR, Ainsworth M. 1993. A method of linking O’Hara TD. 2000. Victorian Province, Australia. In Seas at the
multivariate community structure to environmental Millennium: An Environmental Evaluation, Sheppard CRC
variables. Marine Ecology Progress Series 92: 205–219. (ed.). Pergamon: Oxford; 661–671.
Clarke KR, Green RH. 1988. Statistical design and analysis O’Hara TD. 2002. Endemism, rarity and vulnerability of
for a ‘biological effects’ study. Marine Ecology Progress marine species along a temperate coastline. Invertebrate
Series 46: 213–226. Systematics 16: 671–684.
Clarke KR, Warwick RM. 2001. Change in Marine O’Hara TD, Poore GCB. 2000. Distribution and origin of
Communities: An Approach to Statistical Analysis and Southern Australian echinoderms and decapods. Journal of
Interpretation, 2nd edn. Natural Environment Research Biogeography 27: 1321–1335.
Council and Plymouth Marine Laboratory: UK. Parks Victoria. 2006. Merri Marine Sanctuary. Draft
Clarke KR, Somerfield PJ, Gorley RN. 2008. Testing of null Management Plan June 2006. Parks Victoria: Melbourne.
hypotheses in exploratory community analyses: similarity Pearson TH, Rosenberg R. 1978. Macrobenthic succession in
profiles and biota-environment linkage. Journal of relation to organic enrichment and pollution of the marine
Experimental Marine Biology and Ecology 366: 56–69. environment. Oceanography and Marine Biology Annual
Clarke RT, Wright JF, Furse MT. 2003. RIVPACS models for Review 16: 229–311.
predicting the expected macroinvertebrate fauna and Pinedo S, Garcı́a M, Satta MP, de Torres M, Ballesteros E.
assessing the ecological quality of rivers. Ecological 2007. Rocky-shore communities as indicators of water
Modelling 160: 219–233. quality: a case study in the northwestern Mediterranean.
Commonwealth of Australia. 2005. National Marine Marine Pollution Bulletin 55: 126–135.
Bioregionalisation of Australia. Commonwealth of Reynoldson TB, Day KE, Bailey RC, Norris RH. 1995.
Australia: Hobart. Biological guidelines for freshwater sediment based on
Congleton W, Pearce B, Parker M, Beal B. 1999 Mariculture benthic assessment of sediment (the BEAST) using a
siting: a GIS description of intertidal areas. Ecological multivariate approach for predicting biological state.
Modelling 116: 63–75. Australian Journal of Ecology 20: 198–219.
Dauvin J-C. 2007. Paradox of estuarine quality: benthic Ridgway KR, Dunn JR, Wilkin JL. 2002. Ocean interpolation
indicators and indices, consensus or debate for the future. by four-dimensional least-squares—application to waters
Marine Pollution Bulletin 55: 271–281. around Australia. Journal of Atmospheric and Oceanographic
ESRI. 2004. ArcGIS 9.0. ESRI Software: Redlands, USA. Technology 19: 1357–1375.
Flåten GR, Botnen H, Grung B, Kvalheim OM. 2007. Schoener TW. 1976. The species–area relation within
Quantifying disturbances in benthic communities— archipelagoes: models and evidence from island land birds.
comparison of the community disturbance index (CDI) to In Proceedings of the XVI International Ornithological
other multivariate methods. Ecological Indicators 7: Congress, Firth HJ, Calaby JH (eds). Australian Academy
254–276. of Science: Canberra, ACT; 629–642.
Fraschetti S, Terlizzi A, Benedetti-Cecchi L. 2005. Patterns of Simpson J, Norris RH. 2000. Biological assessment of water
distribution of marine assemblages from rocky shores: quality: development of AUSRIVAS models and outputs. In
evidence of relevant scales of variation. Marine Ecology RIVPACS and Similar Techniques for Assessing the
Progress Series 296: 13–29. Biological Quality of Freshwaters, Wright JF, Sutcliffe
Helmuth B, Denny MW. 2003. Predicting wave exposure in the DW, Furse MT (eds). Freshwater Biological Association
rocky intertidal zone: do bigger waves always lead to larger and Environment Agency: Ableside, Cumbria, UK;
forces? Limnology and Oceanography 48: 1338–1345. 125–142.
Hewitt JE, Anderson MJ, Thrush SF. 2005. Assessing and Suchanek TH. 1994. Temperate coastal marine communities:
monitoring community health in marine systems. Ecological biodiversity and threats. American Zoologist 34: 100–114.
Applications 15: 942–953. Thackway R, Cresswell ID. 1998. Interim Marine and
Keough MJ, Quinn GP. 1998. Effects of periodic disturbances Coastal Regionalisation for Australia: An Ecosystem-
from trampling on rocky intertidal algal beds. Ecological Based Classification for Marine and Coastal
Applications 8: 141–161. Environments. Version 3.3. Environment Australia,

Copyright r 2010 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 20: 452–463 (2010)
 TESTING A RAPID BIODIVERSITY ASSESSMENT METHODOLOGY 463

Commonwealth Department of the Environment: Canberra, Wright JF. 1995. Development and use of a system for
Australia. predicting macroinvertebrate fauna in flowing waters.
Underwood AJ, Chapman MG. 1998. Spatial analysis of Australian Journal of Ecology 10: 181–197.
intertidal assemblages on sheltered rocky shores. Australian Zharikov Y, Skilleter G, Loneranga N, Taranto T, Cameron B.
Journal of Ecology 23: 138–157. 2005. Mapping and characterising subtropical estuarine
Underwood AJ, Chapman MG. 2007. Intertidal temperate landscapes using aerial photography and GIS for potential
rocky shores. In Marine Ecology, Connell SD, Gillanders application in wildlife conservation and management.
BM (eds). Oxford University Press: Oxford; 402–427. Biological Conservation 125: 87–100.

Copyright r 2010 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 20: 452–463 (2010)