Variance estimate and taxonomic resolution: an analysis of macrobenthic spatial patterns at different scales in a Western Mediterranean coastal lagoon.

Abstract

The effects of taxonomic resolution on the variance estimates of macrobenthic assemblages were studied at four spatial scales in a Mediterranean coastal lagoon. The assemblages exhibited significant differences at all the investigated scales; however, spatial variability was mainly associated with the smallest and the largest scales. The decrease of taxonomic resolution (from species to family) was not related to a decrease of the overall variability and similar estimates of variance components were obtained using species and family resolution levels. The ordination models derived from species and family abundances were very similar both in terms of location and dispersion effect, while further aggregation to the class level began to alter the observed spatial patterns. In future studies aimed at assessing changes in the lagoon, resources derived from the cost reductions achieved using family level could be employed to plan more frequent surveys and/or to adopt complex spatial sampling designs with a high number of replicates.

Full text

Page 1

Accepted Manuscript
Variance estimate and taxonomic resolution: an analysis of macrobenthic spatial
patterns at different scales in a Western Mediterranean coastal lagoon
M. Tataranni, F. Maltagliati, A. Floris, A. Castelli, C. Lardicci
PII:S0141-1136(09)00026-9
DOI:10.1016/j.marenvres.2009.02.003
Reference:MERE 3319
To appear in:
Marine Environmental Research
Received Date:4 March 2008
Revised Date:23 February 2009
Accepted Date:26 February 2009
Please cite this article as: Tataranni, M., Maltagliati, F., Floris, A., Castelli, A., Lardicci, C., Variance estimate and
taxonomic resolution: an analysis of macrobenthic spatial patterns at different scales in a Western Mediterranean
coastal lagoon, Marine Environmental Research (2009), doi: 10.1016/j.marenvres.2009.02.003
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Page 2

ACCEPTED MANUSCRIPT

1
Variance estimate and taxonomic resolution: an analysis of macrobenthic
1
spatial patterns at different scales in a Western Mediterranean coastal
2
lagoon
3

4
5
6
7
8
9
M. Tataranni a,*, F. Maltagliati a, A. Floris b, A. Castelli a, C. Lardicci a
a Dipartimento di Biologia, Università di Pisa, via Derna 1, 56126 Pisa, Italy
b Dipartimento di Zoologia e Genetica Evoluzionistica, Università di Sassari, via Muroni 25,
07100 Sassari, Italy
*Corresponding author. Tel.: +39 050 2211403/5, fax: +39 050 2211410
E-mail address: mtataranni@biologia.unipi.it (M. Tataranni)
10
11

Abstract
12
The effects of taxonomic resolution on the variance estimates of macrobenthic assemblages
13
were studied at four spatial scales in a Mediterranean coastal lagoon. The assemblages
14
exhibited significant differences at all the investigated scales; however, spatial variability was
15
mainly associated with the smallest and the largest scales. The decrease of taxonomic
16
resolution (from species to family) was not related to a decrease of the overall variability and
17
similar estimates of variance components were obtained using species and family resolution
18
levels. The ordination models derived from species and family abundances were very similar
19
both in terms of location and dispersion effect, while further aggregation to the class level
20
began to alter the observed spatial patterns. In future studies aimed at assessing changes in the
21
lagoon, resources derived from the cost reductions achieved using family level could be
22
employed to plan more frequent surveys and/or to adopt complex spatial sampling designs
23
with a high number of replicates.
24

25
KEY WORDS: spatial scales - taxonomic resolution - multivariate analysis - macrobenthos -
26
coastal lagoon - monitoring - Western Mediterranean
27

Page 3

ACCEPTED MANUSCRIPT

2

28
1. Introduction
29
Macrobenthic invertebrates are an essential component in soft-sediment environments playing
30
important roles in ecosystem processes, such as dispersion, burial, nutrient cycling and energy
31
flow (Snelgrove, 1998). A deep knowledge of spatial variability patterns in macrobenthic
32
assemblages is relevant to properly characterise one of the major sources of biotic diversity in
33
natural environments; moreover, such information represents a requirement to develop
34
strategies of management and conservation (e.g. Lubchenco et al., 1991), as well as to advise
35
suitable guidelines for periodical monitoring programs (e.g. Underwood, 1997). Soft bottom
36
environments are usually considered homogeneous habitats; however structural analyses in
37
marine and brackish systems have repeatedly demonstrated that patchiness of macrofaunal
38
assemblages is a common feature at both small-medium scales, (Thrush et al., 1989; Hewitt et
39
al., 2002; Noren and Lindegarth, 2005) and large scales (Morrisey et al., 1992; Edgar and
40
Barret, 2002; Ysebaert and Herman, 2002). In particular, those previous studies emphasized
41
the lacking of a single correct scale at which assemblages can be described (Levin, 1992),
42
since different patterns of distribution can be obtained depending on the spatial scale of
43
observation. The description of the distributional patterns at multiple spatial scales and
44
identification of the most relevant ones are needed to formulate possible explanations about
45
ecological processes, or unnatural impacts structuring ecosystems (Underwood and Chapman,
46
1996; Underwood et al., 2000; Ysebaert et al., 2003). Furthermore, such information can be
47
useful for avoiding erroneous interpretations of spatial patterns observed at a particular scale
48
and also to advise useful guidelines for routine environmental monitoring programs.
49
In order to estimate the proportion of variability associated with each examined scale and to
50
identify the most relevant spatial scale, the hierarchical sampling approach is considered the
51
most appropriate method (Underwood, 1997; Hewitt et al., 1998). In hierarchical designs
52

Page 4

ACCEPTED MANUSCRIPT

3
small-scaled sampling units are nested within larger-scaled ones, allowing unconfounded
53
statistical comparisons among each spatial scale (Underwood, 1981; Kotliar and Wiens,
54
1990). Nested designs have been successfully used to investigate populations and
55
assemblages across a wide range of marine habitats and organisms. Most studies focused on
56
intertidal and subtidal rocky shores (i.a. Archambault and Bourget, 1996; Underwood and
57
Chapman, 1996; Benedetti-Cecchi, 2001a; Fraschetti et al., 2001; 2005; Chapman and
58
Underwood, 2008), while soft-bottoms have been less explored (Morrisey et al., 1992; Stark
59
et al., 2003; Noren and Lindegarth, 2005; Terlizzi et al., 2008b).
60
The increase of soft bottom macrobenthic studies is often hidden because identifying and
61
enumerating all organisms are time-consuming and labour-intensive processes (Warwick,
62
1993; Olsgard et al., 1998), that requires taxonomic expertise (Terlizzi et al., 2003). In the
63
last two decades, many studies have analysed data at several taxonomic resolutions, showing
64
that results obtained using species or family level are very similar (i.a. Warwick, 1988;
65
Vanderklift et al., 1996; Olsgard et al., 1998; Karakassis and Hatziyanni, 2000; Lampadariou
66
et al., 2005; Wlodarska-Kowalczuk and Kedra, 2007). Identifying a taxonomic level higher
67
than species that is sufficient for detecting differences in assemblage composition without
68
losing important information is a concept termed “taxonomic sufficiency” (TS, Ellis, 1985).
69
The TS method might have some practical implications, in particular routine monitoring
70
programs could become less expensive and faster than those conducted at the species level
71
resolution and therefore macrobenthic assemblages could be analysed more frequently.
72
However, at present, most studies have usually compared different levels of TS at a single
73
spatial scale (Vanderklift et al., 1996; Olsgard et al., 1998; Karakassis and Hatziyanni, 2000;
74
De Biasi et al., 2003; Wlodarska-Kowalczuk and Kedra, 2007), while few researchers have
75
investigated the effects of TS on the spatial distribution patterns observed at multiple scales
76
(Chapman, 1998; Anderson et al., 2005; Dethier and Schoch, 2006). Moreover in these
77

Page 5

ACCEPTED MANUSCRIPT

4
previous works, spatial variability was usually not distinguished in relation to differences in
78
location or dispersion among groups of samples, while the ecological heterogeneity is
79
considered a valuable feature of any habitat which can provide important information on
80
biological assemblages (Anderson, 2006; Terlizzi et al., 2008a).
81
In the present study, abundance and composition of soft bottom benthic macrofaunal
82
assemblages in a Western Mediterranean coastal lagoon were described with particular
83
attention to their variability across different spatial scales. The Santa Giusta lagoon can be
84
considered as representative of small microtidal brackish environments characterizing the
85
Mediterranean region (Basset et al., 2006). Coastal lagoons are areas of considerable
86
naturalistic interest but often are located close to urban or industrial centres, therefore they are
87
possibly affected by direct (e.g. sewage discharge, aquaculture) or indirect (e.g.
88
eutrophication) human activities (i.a. Barnes, 1991; Lardicci et al., 2001). Given the
89
naturalistic and economic importance of these biotopes, research that may provide appropriate
90
quantitative data is relevant for their conservation and management.
91
In this study, a hierarchical sampling design including four spatial scales (ranging from
92
meters up to thousands of metres) was used i) to estimate the relative importance and test
93
statistical significance of macrofauna variability at different spatial scales, in order to identify
94
the spatial scale associated with the highest variability; ii) to examine if spatial patterns are
95
influenced mainly by changes in species composition or relative abundances, comparing
96
results obtained from several transformations of species abundance; and iii) to analyse if
97
lower levels of taxonomic resolution (family and class) show similar spatial patterns with
98
respect to those obtained at species level, both in terms of location and dispersion effects.
99
Results will allow to increase the knowledge of macrobenthic spatial distribution in Santa
100
Giusta lagoon and to assess the applicability of TS method for decreasing time and cost in
101
subsequent routine surveys. The methodological approach employed in the present
102

Page 6

ACCEPTED MANUSCRIPT

5
investigation could provide interesting practical implications for future studies, not only in
103
this lagoon, but also in other similar brackish environments.
104

105
2. Materials and Methods
106
2.1 Study area and sampling
107
The Santa Giusta lagoon (Western Mediterranean, Italy) is one of the largest coastal brackish
108
environments of Sardinia island, it is a polyhaline basin located along the central-western
109
coast of Sardinia. The lagoon is included in the Ramsar convention (1971) and belongs to a
110
complex system of transitional waters of high natural and economic value. The Santa Giusta
111
lagoon is approximately circular in shape with an area of 7.9 km2 and a mean depth of 1 m; it
112
is located near the town of Oristano and Santa Giusta, in the plain of Pesaria, an agricultural
113
area that is intensively cultivated with rice. The lagoon has no natural attributes and is
114
separated from the sea by a longshore bar, it is also connected with two inner small basins
115
called Pauli Maiori and Pauli Figu (Figure 1). Central and peripheral canals have been
116
dredged about 2 m deep in order to facilitate seawater flow into the lagoon. As a consequence,
117
waters of Santa Giusta lagoon are now well mixed as regards circulation and stratification
118
(Sechi et al., 2001). Salinity ranges from 25‰ to 42‰, with a mean annual value of 30‰
119
(Sechi et al., 2001; Luglié et al., 2002). There is a prevalent sandy-muddy bottom, with small
120
patches of both macroalgae and angiosperms (e.g. Enteromorpha sp., Gracilaria sp., Ruppia
121
cirrhosa, Zostera sp.), which are distributed all over the lagoon.
122
In this context, three sampling zones were randomly selected among the three habitats in the
123
lagoon with different sediment and hydrodynamic properties (Luglié et al., 2002) (Figure 1).
124
Specifically, the alpha zone was in the central-northern part of the lagoon and it was mostly
125
influenced by the urban and industrial wastewaters from Oristano and Santa Giusta areas. It
126
has been considered a low-intermediate hydrodynamic energy environment with a prevalence
127

Page 7

ACCEPTED MANUSCRIPT

6
of clay-silty sediments (Luglié et al., 2002). The beta zone was in the central-southern part of
128
the lagoon and possibly influenced by the drainage from the surrounding farmlands. This zone
129
was characterised by an intermediate-high hydrodynamic energy and sand-silty sediments
130
(Luglié et al., 2002). Then the gamma zone was an area near the Pesaria canal, connecting the
131
lagoon with the sea; it was characterized mainly by sandy sediments and high hydrodynamic
132
energy (Luglié et al., 2002), being closer to the sea (Figure 1). Samples were collected in
133
November 2002, according to a hierarchical sampling design. Within each of the three zones,
134
four random sites were selected and within each site, four areas were randomly chosen. Two
135
replicate samples were taken within each area for a total of 96 samples. Spatial variability was
136
estimated at four hierarchical scales: among zones (103 m apart), among sites within zones
137
(102 m apart), among areas within sites (10s m apart) and among replicates (1 m apart). Since
138
small vegetal patches are widely distributed throughout the lagoon, replicate samples were
139
carefully placed away from vegetal patches (at least 5 m from the closest patch), in order to
140
minimise the possible effects of background heterogeneity on macrofaunal composition. Soft-
141
sediment samples were collected on bare bottom with a box-corer (10 x 17 cm2), sieved
142
through a 0.5 mm mesh and preserved in 4 % formaldehyde. All collected macrozoobenthic
143
organisms were sorted and identified to the species level and abundances (number of
144
specimens per taxon) were calculated. Time needed for classification and counting of diverse
145
taxa was recorded. All analyses were performed by researchers, with low experience in
146
taxonomic identification but supported by skilled taxonomists. The time spent to identify all
147
organisms at the species, family and class levels was 255, 95 and 5 hours, respectively.
148

149
2.2 Statistical analyses
150
Permutational multivariate analysis of variance (PERMANOVA; Anderson, 2001) was used
151
to test the null hypothesis of no differences among assemblages at different spatial scales,
152

Page 8

ACCEPTED MANUSCRIPT

7
according to a three factors (zone, site, area) nested design through 4999 permutations of
153
residuals under a reduced model (Anderson and Ter Braak, 2003). At species level, data were
154
analysed using the Bray-Curtis dissimilarity measure on untransformed and transformed data
155
(square-root, fourth-root, presence/absence), in order to evaluate if assemblages are mainly
156
driven by compositional or relative abundance changes. Using stronger and stronger
157
transformations, the emphasis of results can be shift from the most abundant species to the
158
rarest ones; in particular, variability measured by presence/absence data reflects only
159
compositional changes, while variation in relative abundances is more important in analyses
160
based on other transformations or untransformed data (Clarke and Gorley, 2001; Anderson et
161
al., 2005b). Furthermore, mean squares calculated by PERMANOVA were used to estimate
162
multivariate variance components associated at each spatial scale, in a way analogous to
163
univariate partitioning using ANOVA (Searle et al., 1992; Benedetti-Cecchi, 2001b). For a
164
better comparison, in y-axis the variability at each spatial scale was expressed as square-root
165
of variance components; therefore, the values could be interpreted as percentages of Bray-
166
Curtis dissimilarity (Anderson et al., 2005a). Separate analyses were performed using the
167
square-root transformed data at species, family and class levels of taxonomic resolution. The
168
family level was chosen because it has been often indicated as the most effective in
169
minimizing the cost-benefit ratio (Lardicci and Rossi, 1998; Karakassis and Hatziyanni, 2000;
170
De Biasi et al., 2003; Lampadariou et al., 2005), while the class level was chosen to assess the
171
effectiveness of a further higher resolution. Since a significant result for a given factor from
172
PERMANOVA could indicate that the groups differ in their location and/or dispersion,
173
PERMDISP analyses were also performed to focus only on dispersion effects, testing the
174
factors “zone” and “site” (Anderson, 2006). Analogous analyses were performed using
175
separate data sets for the three main taxonomic groups.
176

Page 9

ACCEPTED MANUSCRIPT

8
The local species richness was visualised as a function of number of replicate samples in
177
species-sample accumulation curves based on 999 permutations. To visualize multivariate
178
patterns in assemblages across the three zones, non-metric multidimensional scaling (nMDS)
179
ordination plots were produced. A separate plot was done for the overall species community
180
and also for data aggregated at family and class level of resolution. All plots were done on the
181
basis of Bray-Curtis dissimilarity matrix of square-root transformed data. To evaluate the
182
degree of similarity among matrices obtained using different taxonomic aggregations, the
183
RELATE routine was used to test the null hypothesis of independence of the two similarity
184
matrices. On the other hand, in order to detect which species contributed most to dissimilarity
185
among the three different zones, a similarity percentage (SIMPER) routine was performed
186
(cut off 80%) (Clarke, 1993). All accumulation curves, nMDS plots, RELATE tests and
187
SIMPER results were obtained using the PRIMER v.6 software (Clarke and Gorley, 2001).
188

189
3. Results
190
3.1 Faunal composition
191
A total of 23 878 individuals belonging to 83 species, 43 families, and 4 classes were
192
collected. The time spent to identify all organisms at the species, family and class levels was
193
255, 95 and 5 hours, respectively.
194
Considering the number of individuals, crustaceans accounted for 43.5% of total abundance
195
followed by polychaetes and molluscs representing respectively 30.0% and 26.5% of total
196
abundance. Instead, considering the number of species, polychaetes were the most
197
representative group (54 species) followed by crustaceans (19) and molluscs (10). The
198
number of species per family varied widely (Figure 2), with most of families (29) represented
199
by only one species. The most species-rich families were within the polychaete class (e.g. 9
200
Syllidae, 6 Spionidae, 4 Capitellidae, 4 Phyllodocidae and 4 Serpulidae) and to a lesser extent
201

Page 10

ACCEPTED MANUSCRIPT

9
in the crustaceans (e.g. 3 Corophiidae and 3 Gammaridae). The majority of crustaceans (12
202
species) and molluscs (7 species) spanned a large part of the lagoon, being recorded in more
203
than 12 of the 48 sampling areas; whereas most polychaetes (24 species) were restricted in
204
less than 12 sampling areas (Figure 3). Seven species (five polychaetes and two crustaceans)
205
were limited to a single area and they were also represented by only one individual. Five
206
species (four polychaetes and one mollusc) were restricted to only two sampling areas and
207
were represented by very few individuals (Figure 3). The mean abundance values for
208
polychaetes and crustaceans were quite similar in all three sampling zones of the study site,
209
while molluscs showed a higher variability being the most abundant taxa in gamma, but
210
almost absent in alpha zone (Figure 4).
211
The local species richness was higher in the alpha zone than in beta or gamma ones and
212
cumulative samples from each zone were representative reaching an asymptote rather quickly
213
(Figure 5). In particular, the number of species collected would be just slightly reduced, even
214
analysing only 2 sampling sites (namely 8 areas or 16 replicates) in each zone of the lagoon
215
(Table 1). A highly significant difference in the faunal composition of the three zones was
216
detected by SIMPER analysis, with the greatest dissimilarity recorded for the alpha - gamma
217
zone pair (75%), followed by the beta - gamma (69%) and the alpha - beta (68%) ones.
218
In alpha zone, despite the smallest number of individuals (5 550), the highest number of
219
species (75) was recorded, with five species that accounted for 60.4% of total abundance.
220
This group included species typical of brackish habitats, such as Monocorophium sextonae,
221
Cymodoce truncata, Hydroides elegans and Corophium acherusicum (Ruffo, 1998) and a
222
species typical of sandy-muddy bottom such as Pseudopolydora antennata (Lardicci et al.,
223
2001). Moreover, SIMPER analyses highlighted the value of other species in determining the
224
dissimilarity among the three zones: a polychaete, Cirriformia tentaculata, and a crustacean
225
occurring in areas with abundant algal coverage, Pseudolirius kroyeri (Table 2). In the beta
226

Page 11

ACCEPTED MANUSCRIPT

10
zone 60 species were found and the five most abundant ones (51.3% of total 7 601
227
individuals) were typical brackish species like Cymodoce truncata, Abra ovata,
228
Monocorophium sextonae, as well as Minuspio multibranchiata and Microdeutopus
229
anomalus, occurring where macroalgae are present (Ruffo, 1998). Opportunistic species
230
indicating organic enrichment (Pearson and Rosenberg, 1978; Cognetti, 1982), such as Phylo
231
foetida and Neanthes caudata, were also characteristic of the beta zone (Table 2), as well as
232
suspension feeders (Loripes lacteus and Cerastoderma glaucum) and grazers (Cumella
233
limicola and Iphinoe serrata). In gamma zone 10 730 organisms from 59 species were
234
collected and 80% of total abundance was reached with only five species: Mytilaster minimus
235
(alone accounting for 33%), Cymodoce truncata and Tanais dulongii, all typical of brackish
236
environments, besides Minuspio multibranchiata and Naineris laevigata.
237

238
3.2 Scales of multivariate spatial variability and taxonomic resolution
239
At the taxonomic level of species, PERMANOVA showed that there was a highly significant
240
variability at all spatial scales considered (Table 3). The greatest variability occurred at the
241
largest spatial scale, among zones, for which the average Bray-Curtis dissimilarity was around
242
37%; then, the successive variation component was that among replicate samples (35% of
243
dissimilarity), followed by less variability among areas (22%) and sites (20%) (Figure 6).
244
Furthermore, the relative importance of different spatial scales in the hierarchy did not vary
245
with different data transformations. Similar spatial patterns were obtained for analyses based
246
on untransformed and transformed data (square root, fourth root, presence/absence) (Figure
247
6).
248
Highly significant variability at all spatial scales was also detected, at the family and class
249
levels (Table 3). In addition, similar variance components, as well as the relative importance
250
of different spatial scales, were maintained proceeding from species to family analysis (Figure
251

Page 12

ACCEPTED MANUSCRIPT

11
7). Instead, using the class level of resolution, the variance components decreased showing
252
less dissimilarity among assemblages at all spatial scales analysed; moreover, proportional
253
amount of variation changed showing the highest variability at the smallest scale, among
254
replicate samples (Figure 7). At the species level, differences among zones were mainly due
255
to differences in their location, since a significant dispersion effect was revealed only between
256
alpha and gamma zones (Table 4). At the scale of site, the source of variability changed
257
depending on the sampling zone as emerged by pairwise tests, within the alpha zone some
258
sites were not significantly different from each other (i.e. P-values of both PERMANOVA
259
and PERMDISP tests were not significant), in some cases sites differed in their location (i.e.
260
P-values of PERMANOVA significant, P-values of PERMDISP not significant), in other
261
cases sites differed both in their location and dispersion (i.e. P-values of both PERMANOVA
262
and PERMDISP tests were significant). Similar results were obtained within the beta zone,
263
where sites were different also because of their dispersion (i.e. P-values of PERMANOVA
264
not significant, P-values of PERMDISP significant). Dispersion effect never contributed to
265
differences among sites in the gamma zone (Table 4). All but two pairwise results were
266
likewise detected using the family level of taxonomic resolution; while at the class level, the
267
majority of results were not significant, therefore indicating different relationships compared
268
to those obtained at finer taxonomic levels, both at zone and site spatial scales (Table 4).
269
Separate analyses for the three collected taxonomic groups showed a highly significant
270
variability at all spatial scale; only the variance component for molluscs at the site scale was
271
found not different from zero because of the greater variability at the smaller spatial scale of
272
area (Figure 8). Such results matched the pairwise tests which showed that in very few cases
273
sites differed, mainly in their dispersion (Table 5). For crustaceans and molluscs the greatest
274
variability occurred at the largest spatial scale (Figure 8) and for both groups it was mainly
275
due to differences in location among zones (Table 5). For polychaetes, the sources of the high
276

Page 13

ACCEPTED MANUSCRIPT

12
variability at the zone scale were differences in location and dispersion (Table 5); however
277
polychaetes showed the greatest variability among replicate samples and such spatial scale
278
was also important for the other taxonomic groups (Figure 8). Except for molluscs, a small
279
variability was associated with the two intermediate spatial scales (Figure 8); in particular at
280
the site scale, there was a prevalent location effect both for polychaetes and crustaceans
281
although differences in dispersion were also detected especially for crustaceans within the
282
beta zone (Table 5).
283
The nMDS plot based on the species abundance data showed some differences among the
284
three sampled zones. In particular, gamma samples were clearly clustered and separated from
285
the other zones, while alpha and beta samples were partially overlapped (Figure 9a). At the
286
family level, nMDS ordination was very similar to that obtained at the species level (Figure
287
9b). Further aggregation to the class level produced a different ordination pattern, with
288
substantially higher levels of overlapping of the three zones (Figure 9c). Relationships
289
between similarity matrices calculated for the three taxonomic levels were confirmed by
290
RELATE results, which showed ρ = 0.960, p<0.001 between species and family levels and ρ
291
= 0.565, p<0.001 between species and class levels.
292

293
4. Discussion
294
The first result that stands out from our work is that the benthic assemblages of Santa Giusta
295
lagoon were extremely variable, with significant differences at all considered spatial scales,
296
from metres up to thousands of metres. This outcome was highly consistent with results
297
reported in studies analysing spatial variability by nested design, whatever the habitat
298
investigated (see Fraschetti et al., 2005 for a review). In the Santa Giusta lagoon most of the
299
variation was associated with the smallest and the largest spatial scale, thus indicating that
300
both small-scale and large-scale processes play a major role in shaping benthic community
301

Page 14

ACCEPTED MANUSCRIPT

13
spatial patterns. Variability among replicates at small spatial scale is usually considered a
302
widespread feature of many different assemblages, being mainly determined by biological
303
interactions and/or local physical factors (i.a. Ekman, 1979; Underwood and Chapman, 1986;
304
Wilson, 1991; Morrisey et al., 1992; Benedetti Cecchi et al., 2001a; Coleman et al., 2002;
305
Rossi and Lardicci, 2002; Fraschetti et al., 2005; Chapman and Underwood, 2008). On the
306
other hand, differences in assemblages at large spatial scales have been mostly related to
307
abiotic processes (Thrush et al., 1989; Thrush, 1991). Factors such as hydrodynamic energy,
308
trophic status, seawater and freshwater influence, nutrients supply and confinement could
309
differently characterise the three sampling zones of the Santa Giusta lagoon, according to the
310
models of zonation proposed for other Mediterranean coastal lagoons (Guelorget and
311
Perthuisot, 1982; Lardicci et al., 1993; 1997; Pérez-Ruzafa et al., 2007). In addition, benthic
312
communities could be unevenly subjected to the two main sources of anthropogenic
313
disturbance affecting this lagoon at all the study scales. Results of this study showed that the
314
three sampling zones were clearly distinct and characterised by typical features but analysing
315
and explaining the effects of abiotic or biotic factors responsible of such differences were not
316
among the explicit aims of this study.
317
A number of papers reported that results of statistical analyses can be greatly influenced by
318
the choice of data transformation; in fact, the ability to detect differences along strong
319
environmental gradients was affected more by changing the data transformation rather than
320
the level of taxonomic identification (i.a. Olsgard et al., 1998; Karakassis and Hatziyanni,
321
2000). This is also consistent with results by Chapman (1998) and Lasiak (2003), who
322
observed that the type of transformation altered patterns of variability within sites, which may
323
be important for some research programmes. The choice of transformation determines the
324
relative contribution of quantitative and qualitative intersample differences in the final
325
outcome of all multivariate analyses. Strong transformations (fourth root, presence/absence)
326

Page 15

ACCEPTED MANUSCRIPT

14
give little weight to differences in abundance, whereas weak (square root) or null
327
transformations provide patterns mainly reflecting the differences of the most abundant
328
species (Olsgard et al., 1998; Karakassis and Hatziyanni, 2000; Clarke and Gorley, 2001). In
329
particular, the variability estimated by analysing presence/absence data explicitly reflects the
330
compositional changes of assemblages at different spatial or temporal scales and this can be
331
compared with analyses based on other transformations (or untransformed data), mostly
332
describing relative abundance differences (Anderson et al., 2005b). On this basis, some
333
studies demonstrated that variability of benthic assemblages at larger scales is mainly
334
“compositional”, as indicated by the presence/absence analyses, while variability at medium
335
or smaller scales is driven by changes in relative abundance, particularly by numerically
336
dominant taxa (Archambault and Bourget, 1996; Chapman, 1998; Anderson et al., 2005b).
337
Conversely, our results showed that the relative importance of examined spatial scales (from
338
meters up to thousands of metres) was always maintained, despite different transformations
339
used. These findings underpinned that in Santa Giusta lagoon, large scale processes shaped
340
three distinct zones characterized by different species. At the other investigated scales,
341
differences in species composition were possibly caused by the presence of diverse
342
microhabitats, which created high spatial heterogeneity. For example, small vegetal patches
343
might possibly influence the faunal composition of the surrounding bare bottoms; however,
344
such patches are distributed all over the lagoon, thus producing a high spatial heterogeneity at
345
the smallest spatial scale in all the three sampling zones of Santa Giusta. This outcome
346
indicated that spatial heterogeneity is not necessarily related to the extent of the study area.
347
Therefore the unambiguous interpretation of results can be promoted using nested designs
348
also in small environments, like the Santa Giusta lagoon. The multi-scale approach is
349
recommended as a basic tool for spatial distribution analyses, especially when such
350
information is still scarcely known in the investigated environment. In particular for future
351