Introduction

Ecklonia radiata (C. Ag.) J. Agardh is a small kelp reaching a maximum length of about 2 m. It has its main distribution in the southern hemisphere where it is common, often dominant, in the warm-temperate parts of South Africa (Field et al. 1980), Australia (Steinberg and Kendrick 1999) and New Zealand (Choat and Schiel 1982). E. radiata is also found in the northern hemisphere, but so far it has been described only from Oman on the Arabian Peninsula (Bolton and Anderson 1994). It is likely, however, that a revision of the genus Ecklonia would include one North African and several Japanese Ecklonia species under E. radiata (Bolton and Anderson 1994), increasing its northern hemisphere distribution considerably.

E. radiata is remarkably successful in terms of its ecological dominance and it is probably the most abundant macroalga in Australia (Steinberg and Kendrick 1999). It has a profound influence on associated algal assemblages (Kendrick et al. 1999; Melville and Connell 2001) because canopies of large macroalgae modify the physical environment (Eckman et al. 1989; Kennelly 1989; Critchley et al. 1990). The contribution of individual kelps to modification of the environment is a function of their morphology (Velimirov and Griffiths 1979; Kennelly 1989). Consequently, if E. radiata exhibits systematic spatial variation in morphology, it will likewise systematically exert spatially variable effects on algal assemblages. This may be a factor contributing to the large variability commonly encountered in comparisons of E. radiata assemblages across various spatial scales (Kennelly and Underwood 1992; Kendrick et al. 1999; Fowler-Walker and Connell 2002).

The objective of the present study was to quantify the morphological variation in E. radiata along its geographic range in Australasia. Geographic separation may cause morphological differences among populations of macroalgae as genetic exchange among populations becomes restricted (Kusumo and Druehl 2000) and they are exposed to grossly different environmental conditions. These differences may be continuously changing following major environmental clines in, for example, insolation, temperature or salinity (Russell 1986; Kalvas and Kautsky 1998). Consequently, we tested the hypotheses that E. radiata has different morphology at different locations and that the degree of morphological difference depends on spatial distances among locations.

Materials and methods

Sampling

Sampling was carried out during summer from November 2000 to March 2001. Eleven locations were selected; eight along the south-west Australian coastline and three from other Australasian regions (Fig. 1; Table 1). Locations within Western Australia were selected on the basis of accessibility and proximity. Locations outside Western Australia were found by requesting collaboration via the ALGAL-L email list. These 11 locations span 18° north–south and 37° east–west, and they cover >6,000 km of coastline including the geographic range of Ecklonia radiata from its northern limit in Western Australia (Kalbarri: Womersley 1987) and south around Australia to Sydney, and New Zealand. Three random sites were sampled within each location. Large, presumably fully developed, individuals were targeted during the sampling (Rice et al. 1985; Jackelman and Bolton 1990). Thus, in situ a SCUBA diver haphazardly collected large stage 3 sporophytes (sensu Kirkman 1981) by carefully dislodging the holdfast from the reef with a dive knife. Only solitary individuals where the holdfasts did not overlap were collected, and collections were generally made within an area of a few square metres. The collected specimens were kept moist and refrigerated until processing.

Fig. 1.
figure 1

The geographical position of sampling locations. Global positioning system co-ordinates and location codes are given in Table 1

Full size image
Table 1. Sampling locations. Three sites were sampled within each location. Ten individuals were sampled at each site, except ADE3 (n=6) and DSN1–3 (n=5)
Full size table

Within each site 11 morphological characters were sampled from n=5–10 individuals (Table 1). These morphological characters were chosen to cover most parts of the thallus (see illustrations in Bolton and Anderson 1994). The morphological characters and procedures for measurement are given in Table 2.

Table 2. Morphological characters measured
Full size table

Data analyses

Firstly, we tested whether the morphology of Ecklonia radiata is different among locations along its geographic range in Australasia (H 0: there are no differences in E. radiata morphology among locations separated by 10s–1,000s of kilometres). Each morphological character was tested with a one-way ANOVA (Zar 1996) where sites were nested within locations. Before testing, the data set was balanced by randomly removing individuals from each of the sites to meet the lowest common denominator (n=5; cf. Table 1) and checked for homogeneity of variances (Cochran's C-test: Winer et al. 1991). Because of the large number of tests (11 ANOVAs) Bonferroni corrected α-values (α/11) were used to determine significance levels. Further, the total variation was decomposed and partitioned into the relative contribution from each level of nesting, and the relative magnitude of effect (ω 2) was calculated (Graham and Edwards 2001).

Multivariate analyses were performed in PRIMER v5 (Clarke and Gorley 2001). To test the independence of each character, product moment correlation was performed among all pairs of variables using the "Draftmans plot" routine. A normalised euclidean distance dissimilarity matrix was calculated from untransformed data using the full unbalanced data set on all morphological characters. Difference in morphology among locations was tested with a two-way nested ANOSIM (analysis of similarities) with sites nested in locations, followed by pairwise comparisons. The data were represented graphically in non-metric multidimensional scaling (nMDS) plots. SIMPER (similarity of percentages) was performed to identify which morphological characters contributed most to the observed patterns.

Secondly, we tested whether the magnitude of morphological difference was a function of distance between locations (H 0: there is no correlation between distance and the degree of morphological similarity among locations). The matrix of pairwise R-values from the ANOSIM was correlated (Spearman rank correlation) to an among-location distance matrix generated from GPS (global positioning system) co-ordinates (cf. Table 1).

Results

Spatial variability of individual morphological characters

Most morphological characters of Ecklonia radiata showed considerable variation among locations (Tables 3, 4). For example, total thallus length differed by >65 cm between the individuals in Kalbarri and the individuals in Sydney. Similarly, there was >130 g dry weight difference between the thallus biomass of individuals from Kalbarri and individuals from Mandurah. Lateral spinousity was the most variable character (CV=48%), whereas stipe diameter was the least variable (CV=3%).

Table 3. Ecklonia radiata. Summary statistics of morphological characters. Each location: data given are means (standard errors) of n=3 sites. Total: data given are means (standard errors) of n=11 locations. Location codes are given in Table 1
Full size table
Table 4. Ecklonia radiata. Results from ANOVA on individual morphological characters [df=10, 22 and 132 for location, site(location) and individuals within site]. No transformations were necessary to homogenise variances (Cochran's C-test: P>0.05). [Var. Comp. and ω 2 are the variance components and magnitude of effect (Graham and Edwards 2001), respectively] ***, P<0.001/11; **, P<0.01/11; *, P<0.05/11; n.s., P>0.05/11
Full size table

Five of the 11 morphological characters were significantly different (P<0.05/11) among locations (Table 4). All morphological characters, except lateral spinousity, were also significantly different (P<0.05/11) among sites within locations (Table 4). Although sites within locations in most cases were highly significantly different (P<0.001/11), the variation at this level generally contributed <30% to the total variation, as indicated by the magnitudes of effect in Table 4. Location was the dominant contributor to variance in lamina width, numbers of laterals and in lateral spinousity. Variation among individuals within a site (i.e. the error term) generally had a high, often >40%, contribution towards the total variation, particularly in the number of lamina twists per plant.

Most morphological characters were poorly to moderately correlated to each other (−0.5<R<0.5; Table 5). Exceptions were thallus length versus stipe length (R=0.53), lamina length (R=0.85) and number of laterals (R=0.74), and number of laterals versus lamina length (R=0.70). These correlations could however be expected as some of these characters are defined from each other. For example, total length is the sum of stipe length, lamina length and the length of laterals that extend beyond the tip of the central lamina. Contrary to what would be expected from simple allometric scaling (Larkum 1986; Wernberg et al. 2001), there was no correlation between thallus dry weight and thallus length (R=0.11) or the number of laterals (R=0.11).

Table 5. Ecklonia radiata. Correlation coefficients (Pearson product moment correlation) among the morphological characters on all the sampled individuals (n=311). Morphological codes are given in Table 2
Full size table

Spatial variability of multivariate morphology

The ordination based on all morphological characters (Fig. 2) showed large separation (high dissimilarity) among some locations (e.g. DSN and KAL) and very small separations (high similarity) among other locations (e.g. JUR and ALB). Sites within locations also exhibited large separation within some locations (e.g. FRE and DSN). The ANOSIM revealed a moderate, but highly significant, separation of locations (Clarke's R=0.512, P<0.001). There was also a marked, although smaller, separation of sites nested within locations (Clarke's R=0.326, P<0.001).

Fig. 2.
figure 2

Ecklonia radiata. Non-metric multidimensional scaling plot of samples within sites among locations along the geographical distribution of E. radiata in Australasia. Location codes as in Table 1

Full size image

The multivariate pairwise comparisons among locations (Table 6) yielded very small or even negative Clarke's R values among ADE–JUR–HAM–ALB, suggesting that E. radiata from sites within these locations were particularly similar. ADE had relatively small Clarke's R values in comparisons with all locations. Conversely, MAN and DSN were different to most other locations.

Table 6. Ecklonia radiata. Clarke's R-values from post hoc multiple comparisons of dissimilarity between location pairs. R-values in bold indicate well-separated locations (R>0.75: Clarke and Gorley 2001). P≥0.1 in all cases because of the limited maximum number of permutations. Location codes are given in Table 1
Full size table

Thallus dry weight was by far the most important morphological character, on average accounting for >30% of the dissimilarity in E. radiata morphology between location pairs (Table 7). Thallus dry weight also had the most consistent contribution to location dissimilarity.

Table 7. Ecklonia radiata. SIMPER results. Mean (±standard deviation) values for the 55 possible location combinations
Full size table

There was no correlation between distance and location dissimilarity (Spearman's R=0.08, P=0.57, n= 55; Fig. 3). Some E. radiata populations separated by <200 km were just as different from each other as they were from populations >3,000 km away. There was a trend of populations separated by >3,000 km to be relatively different (Clarke's R≥0.37), although this must be interpreted conservatively because of the lack of closely located regions outside Western Australia and the potentially confounding effects of sampling depths (cf. Table 1).

Fig. 3.
figure 3

Ecklonia radiata. Location dissimilarity (pairwise Clarke's R) versus linear distance between locations

Full size image

Discussion

The morphological variation within and among 11 populations of Ecklonia radiata in Australasia was determined to be highly variable. There were no consistent patterns of spatial variation in individual morphological characters: variation among locations, sites within locations and individuals within sites were all main contributors to total variation in one character or another. When all morphological characters were considered concomitantly, distinctly different E. radiata morphologies were found among some locations. Yet, there was no relationship between the degree of morphological difference and the linear distance between locations. We thus reject our first null hypothesis of no differences in morphology among locations, but retain our second null hypothesis that there is no relationship between distance and degree of morphological difference. Consequently, our results do not support a model of geographic clines in E. radiata morphology. Instead we propose that the "total-morphology" at any one site reflects multiple forcing factors operating on different morphological characters at different spatial scales.

Spatial variation in morphology

We found that variations in very few characters were correlated, and that different spatial scales were important to variation in individual morphological characters. This suggests that most morphological characters develop independently of each other, possibly stimulated by processes operating at different spatial scales. For example, the majority of variation in lamina length was associated with spatial scales of kilometres (among sites within location), possibly reflecting gross differences in wave exposure among inshore and offshore sites (e.g. Phillips et al. 1997). In contrast, most of the variation in lateral spinousity appears to be associated with scales of 100s–1,000s of kilometres (among locations) and perhaps reflects regional differences in grazing pressure (Fowler-Walker and Connell 2002; Vanderklift 2002). Regional differences in grazing pressure have been suggested to play a role in morphological differentiation between populations of Fucus vesiculosus L. (Kalvas and Kautsky 1993). However, morphological characters are not necessarily affected only by processes operating at one scale: lateral spinousity could, at the same time, reflect a protective adaptation to large-scale differences in grazing pressure as well as a physiological adaptation to reduce the thickness of diffusion boundary layers (Hurd 2000) under small-scale differences in hydrodynamic regimes (Bell and Denny 1994). Indeed, the large importance of the smallest scale, among individuals within site, to variation in all morphological characters suggests wide genotypic plasticity. The differences in scales of variation also emphasise the potential bias in only looking at a few arbitrarily selected morphological characters.

Thallus dry weight was identified as the morphological character contributing most to differences in kelp morphology among locations. The poor correlation between dry weight and other size-related measures (e.g. total length, number of laterals) suggests that the influence of differences in dry weight among locations is more than simple allometric scaling, whereby, for example, dry weight increases because the length increases (Larkum 1986; Wernberg et al. 2001). Dry weights integrate aspects from many morphological characters, and the independent variation in several characters concomitantly may result in unpredictable variation in dry weight depending on whether variations in one character compensate or exacerbate the effect of variation in other characters. Biomass, including dry weight, may be affected directly by tattering (Blanchette 1997) or grazing (Steinberg 1995) or indirectly through the effect of external factors on the development of individual morphological characters such as thallus thickness (depth and wave exposure: Molloy and Bolton 1996) and stipe elongation (light and wave exposure: Sjøtun et al. 1993; nutrient concentrations and wave exposure: Gerard and Mann 1979).

Studies on intertidal algae have demonstrated considerable morphological variation among locations separated by a few kilometres to 1,000s of kilometres (Rice et al. 1985; Kalvas and Kautsky 1993; Ralph et al. 1998), but there is little evidence to suggest that these differences are continuously distributed in space or follow major clines (Rice et al. 1985; Ralph et al. 1998; Scott et al. 2001; but see Rice and Kenchington 1990). Also, variation among sites within locations is often of similar magnitude (or larger) to variation among locations, and a large proportion of the total variation is generally present at the lowest level of analysis: among individuals within sites (Rice et al. 1985; Kalvas and Kautsky 1998; Ralph et al. 1998; Scott et al. 2001). Morphological variation in macroalgae is often explained in terms of adaptation to hydrodynamic environments (see review by Hurd 2000). As pointed out by Rice et al. (1985) these environments are not continuously distributed in space over large geographic scales. Rather, microtopography (Carpenter and Williams 1993; Bell and Denny 1994) along with reef and coastline topography (Phillips et al. 1997; Gaylord 1999) causes hydrodynamic environments to vary discretely on scales from centimetres to several kilometres. Thus, variation in the hydrodynamic environment is a plausible cause of mosaics of morphs on several spatial scales, if E. radiata morphology varies consistently with the hydrodynamic regime.

Morphological differences among locations are not necessarily caused by either pheno- or genotypical variation. Unless the age distributions of the investigated populations are known to be similar, ontogenetic differences cannot be ruled out (e.g. Kalvas and Kautsky 1993) as many morphological characters are known to correlate with age within several laminarian algae (Novaczek 1981; Hymanson et al. 1990; Sjøtun and Fredriksen 1995). Differences in age structure, for example caused by difference in wave exposure (Sjøtun et al. 1993) or grazing (Andrew and Jones 1994), could contribute towards regional differences in morphology because E. radiata from northern New Zealand has been reported to attain ages of >15 years (Novaczek 1981) whereas E. radiata from Fairlight Bay in New South Wales (Larkum 1986) and Marmion Lagoon in Western Australia (T. Wernberg, unpublished data) are considerably younger, with maximum ages of 2–4 years. Thus, in this study the large, fully developed individuals sampled from Doubtful Sound, New Zealand, might have been considerably older than those from other locations, leading to among-location differences that are unrelated to factors associated with geographic separation. Similarly, if kelp beds are composed of individuals from several cohorts (Dayton et al. 1984; Sjøtun et al. 1993), age differences among individuals within sites could possibly explain the large morphological variation among individuals within sites.

Taxonomic issues

The morphology of Ecklonia radiata is highly variable and this has led to the formation of an intractable suite of names for the various morphotypes, covering several levels in the taxonomic hierarchy (Bolton and Anderson 1994). An example of this is E. radiata v. biruncinata or E. biruncinata (Bory) Papenf., a morph in which the surface of the thallus is covered with small spines. No studies have systematically tested whether the differences in morphological characters justify taxonomic distinction, and it is unknown whether these morphological variants are geno- or phenotypes. Because of the confusion and inconsistency in the systematics of the various geographical and morphological varieties of "E. radiata", Bolton and Anderson (1994) found it more appropriate to group them under "E. radiata complex" rather than classifying them as distinct species or subgroups under E. radiata.

Substantial morphological variation is a common phenomenon in marine macroalgae (Rice et al. 1985; Jackelman and Bolton 1990; Molloy and Bolton 1996; Ralph et al. 1998) that has led to taxonomic confusion and debate about the nature of such variation and the level at which it should be recognised (Bolton and Anderson 1994; Ralph et al. 1998). We have found no evidence of distinct morphotypes among the 11 populations of E. radiata sampled. This study, however, did not include the large stipitate forms, with stipe lengths in excess of 1 m known from northern New Zealand and Queensland (Novaczek 1984; Cole and Syms 1999), nor did it include the peculiar form of E. radiata from Hamelin Bay, with haptera forming at the lateral apices, tentatively identified as E. brevipes J. Agardh by Huisman (2000). Our findings support the very wide species definition suggested by others (Bolton and Anderson 1994, and references herein). However, as there are relatively large variations within location that appear to be phenotypic, morphological distinctness ("ecotypes") may warrant the recognition of types on some level below species. This, however, needs further experimental research directed at the influence of specific environmental factors such as light, temperature, grazing pressure and hydrodynamic regime. Also, genetic analyses (e.g. Kusumo and Druehl 2000) would be a very strong tool to elucidate the nature of differences among E. radiata populations and their morphology.