Introduction

Species represent the fundamental units on which many analyses in fields such as ecology, biogeography, evolutionary biology, and conservation are based (Guerra-García et al. 2008). Biological research thus often depends on the accurate identification and discrimination of species. However, species identification for octocorals can present a challenge. Identifications based on macroscopic features observable in the field or in preserved specimens can be made for certain taxa or regions (e.g., Sánchez and Wirshing 2005), but in most cases, species discrimination requires the extraction and examination of microscopic sclerites (e.g. Williams 2013; Moore et al. 2016, 2017; van Ofwegen 2016, 2020), which requires a high level of taxonomic expertise, or genetic comparisons (e.g. McFadden et al. 2014a, 2014b, 2017; Lau and Reimer 2019; Xu et al. 2020), which can be time-consuming and expensive. The conflict between the need to identify and discriminate between species and the resources this requires is problematic, and as a result, species-level identification in octocorals may be beyond the scope of some studies, such as those relying on field observations, which may then be restricted to higher taxonomic level identifications (e.g., Fabricius and De’Ath 2008; Chanmethakul et al. 2010).

During taxonomic revision of the nominal soft coral species “Alcyonium aurantiacum” and its associated species complex in New Zealand (Kessel et al. 2022), it became apparent that two of the species involved can be extremely difficult to distinguish based on macroscopic morphological characters, despite being separated by consistent sclerite differences (for details on taxonomic characters see Kessel et al. 2022). Specimens of Kotatea aurantiaca Quoy & Gaimard, 1833 and the newly described K. lobata Kessel et al., 2022 (Alcyoniidae Lamouroux, 1812) can be similar in colony growth form and dimensions, as well as in the number and shapes of their colonies’ lobes. The growth form of K. aurantiaca is generally characterised by slender, digitate lobes, while K. lobata tends to produce thicker and more robust lobes (Kessel et al. 2022), but both species are highly variable and these morphologies grade into one another. As a result, intraspecific variation in macroscopic colony morphology overlaps between these species and can blur interspecific differences (Fig. 1), reflecting the broad patterns of indistinct morphological species boundaries previously noted among octocorals (Prada et al. 2008; Dueñas and Sánchez 2009; McFadden et al. 2017). This is exacerbated by the common behaviour of soft corals to hydrostatically expand and contract the coelenteron and thereby change the size and shape of a colony, which may also be affected by collection and preservation method (Fabricius and Alderslade 2001; Hellström and Benzie 2011; Davis et al. 2015). This means that a given collection of specimens will invariably display a spectrum of expansion/contraction states that may complicate, for example, the differentiation between expanded specimens of K. aurantiaca and contracted specimens of K. lobata.

Fig. 1
figure 1

A selection of Kotatea lobata (a–c) and K. aurantiaca (d–f) colonies (separated by lines), illustrating the typical range of expanded (a, d), partly expanded (e) and contracted (b, c, f) states exhibited by sets of preserved specimens. Note that depending on the state of expansion/contraction, the two species can resemble one another closely in colony growth form (e.g. differences are not as apparent between b and e or between c and f as they are between a and d)

Full size image

While K. aurantiaca tends to occur at more southerly latitudes and deeper depths than K. lobata, the two species overlap in their geographic and bathymetric distributions (Kessel et al. 2022). Geographic overlap is particularly pronounced around Northland, in the northern North Island of New Zealand. Regarding depth, overlap occurs at ~ 30 m, which corresponds approximately to the minimum depth so far observed for K. aurantiaca and maximum depth for K. lobata. The true extent of spatial overlap, however, is likely greater than can be inferred from the currently available material. Therefore, collection data may be of limited use in informing the species identity of a specimen. Moreover, phenotypic plasticity—defined as morphological responses to the environment that result in intraspecific variation (West-Eberhard 2003)—is common in octocorals, particularly along depth gradients (e.g., West et al. 1993; Rodríguez-Lanetty et al. 2003; Kim et al. 2004; Gori et al. 2012; Costantini et al. 2016; Calixto-Botía and Sánchez 2017), and may thus also contribute to the morphological variation observed in K. aurantiaca and K. lobata.

Kotatea aurantiaca and K. lobata are two of the most commonly encountered inshore octocorals in New Zealand, where they are endemic, but their colony morphologies have never been systematically measured or compared, and virtually nothing is known regarding any aspect of either species’ ecology or life history characteristics, which could inform why they occasionally adopt such similar forms. The aim of the present study was to ascertain if, and in what ways, the colony growth forms of K. aurantiaca and K. lobata differ, and to test the taxonomic value of macroscopic measurements using statistical species discrimination analyses. The reliability of statistical species discrimination methods that are based directly on morphological measurements was compared to methods using ratios formed from these measurements. Ratios were used to produce variables that are independent of the actual size of a specimen and can thus account for the expansion/contraction state of colonies. This marks the first time that ratios of colony-scale morphological measurements have been used for this purpose in octocorals. Considering the lack of ecological information currently available for K. aurantiaca and K. lobata, phenotypic plasticity with depth was also investigated for both species. It is hoped that this novel discrimination method will be used on other sets of morphologically similar soft coral species globally to elucidate taxonomically informative characters that may have previously been overlooked and to facilitate species identification in this group.

Material and methods

Sample selection and measurements of colony morphology

Forty preserved colonies were selected from each of the K. aurantiaca and K. lobata specimens listed by Kessel et al. (2022). Nearly all K. lobata colonies available were included, barring the smallest and most degraded individuals. Since K. aurantiaca colonies were more numerous, the 40 most-intact individuals were selected to achieve balanced sample sizes and to correspond to the colony size range observed among the K. lobata samples. Maximum colony height, width, and thickness (taken at 90˚ relative to the width measurement), as well as the length, diameter (at midpoint), and height (from colony base) of all primary lobes, were measured (primary lobes were here treated as lobes arising directly from the base of the colony, not from other lobes—they can be “mother” lobes and give rise to “daughter” lobes sensu Sánchez (2004) in regard to branches, but may also lack daughter lobes), and the total number of terminal lobe tips was also recorded for each individual. This allowed five direct measurements of colony morphology to be obtained: (1) estimated volume (= colony height × width × thickness); (2) mean lobe length (mean length of primary lobes); (3) mean lobe diameter (mean diameter of primary lobes at midpoint); (4) mean lobe height (mean height of primary lobes from colony base); and (5) number of tips (total number of terminal lobe tips).

Three ratios were calculated from the direct measurements to produce dimensionless values that are independent of a given colony’s state of expansion/contraction. The following ratios were selected because they characterise key colony growth form attributes: (1) the ratio of mean lobe length to mean lobe diameter (LL:LD): an indicator of lobe thickness, where a high ratio represents a long and thin lobe and a low ratio represents a thick lobe; (2) the ratio of mean lobe height to colony height (LH:CH): an indicator of colony stalk development, where a high ratio represents a well-developed stalk and a low ratio represents a poorly developed stalk; and (3) the ratio of number of tips to mean lobe diameter (Tips:LD): an indicator of tip development or lobe branching/splitting, where a high ratio represents much branching into daughter lobes and a low ratio represents little branching. Although the absolute number of tips present on a colony does not depend on its state of expansion/contraction, their conspicuousness does increase with colony expansion, with tips being less discernible and more prone to miscounts on highly contracted specimens.

Interspecific differences

All statistical analyses were carried out in IBM SPSS 27 and PAST4 (Hammer et al. 2001). First, two separate Hotelling’s T2 tests (a variation of one-way MANOVA where the independent variable has only two groups) were used to determine whether K. aurantiaca and K. lobata differ in either their direct measurements or their ratios of colony morphology. Note that alternative tests with fewer assumptions, such as PERMANOVA, may also be appropriate for exploring interspecific differences in soft corals, but MANOVA was here deemed preferable due to its higher statistical power as a parametric test.

All five direct measurements were transformed to better meet the assumptions of the test, while the three ratios were untransformed and calculated from untransformed data. Estimated volume and number of tips were both inverse transformed (inv), mean lobe length and mean lobe diameter were both square root transformed (sqrt), and mean lobe height was log10 transformed. These transformations were carried out because they performed best overall at improving the fit of the data to statistical assumptions (data not shown). The transformed data were used for all analyses described below.

Several assumptions of Hotelling’s T2 were violated and required further correction or consideration. Mean lobe height, estimated volume, and number of tips, as well as LH:CH and Tips:LD, were non-normally distributed (Shapiro–Wilk test p < 0.05), but Hotelling’s T2 (and MANOVA in general) is considered robust to departures from normality (Weinfurt 1995). The number of tips and all three ratios violated the assumption of homogeneity of variance (Levene’s test p < 0.05), which was corrected for by using Welch’s t-test for pairwise comparisons. The test on the three ratios did not exhibit homogeneity of variance–covariance matrices (Box's M test p < 0.001), but Hotelling’s T2 is considered robust to this violation when sample sizes are balanced (Tabachnick and Fidell 2014)—as is the case here—and Pillai’s trace was used instead of Wilks’ Λ (Olsen 1976). In both the Hotelling’s T2 tests (on direct measurements and on ratios), one multivariate outlier was detected for K. aurantiaca (assessed by Mahalanobis distance) as well as several univariate outliers for most variables in both species (assessed by boxplot), and although MANOVAs are regarded as sensitive to outliers (Tabachnick and Fidell 2014), these were retained in the analysis due to small sample sizes and to incorporate the true morphological variation observed (removal of outliers did not affect test results). In all other instances, all remaining assumptions were met, including no multicollinearity among dependent variables. Finally, pairwise comparisons between species were made using post-hoc independent-samples t-tests (or Welch’s t-test) with a Bonferroni-adjusted α level of 0.01 for direct measurements (5 tests) and 0.017 for ratios (3 tests).

To further explore interspecific differences among the ratios, Spearman’s rank-order correlations were used to assess the relationships between mean lobe length and mean lobe diameter, between mean lobe height and colony height, and between number of tips and mean lobe diameter for both species at a Bonferroni-adjusted α level of 0.008 (6 tests).

Species discrimination

Two separate binomial logistic regressions were used to classify individual colonies according to species based on either the five direct measurements of colony morphology or the three ratios described above. For both analyses, all independent variables (measurements and ratios) were linearly related to the logit of the dependent variable (species), as assessed using the Box-Tidwell procedure (Box and Tidwell 1962; Fox 2016). For the regression incorporating the direct measurements as independent variables, one standardised residual with a value of − 2.7 standard deviations was identified but retained in the analysis. For the regression using the ratios as independent variables, two standardised residuals with values of − 4.555 and − 3.066 standard deviations were identified but also retained in the analysis. These outliers were retained due to small sample sizes and to incorporate the full range of morphological variation observed among the specimens (removal of outliers did not affect test results). Note that step-wise model selection was also tested on the data, but the resulting models routinely performed worse at species discrimination and are thus not presented.

For both analyses, ROC (receiver operating characteristic) curves were calculated. In binomial logistic regression, the area under the ROC curve is equivalent to the concordance statistic, which is the most common measure of a generalized linear model’s ability to discriminate among the dependent variable categories (Gönen 2007; Steyerberg 2009). ROC curve coordinates were used to inform cut-off values for species discrimination, which were deemed acceptable when sensitivity (% of K. lobata specimens correctly assigned) was ≥ 80% and 1 − specificity (false positives or the percentage of K. aurantiaca specimens incorrectly assigned as K. lobata) was ≤ 20%.

Intraspecific responses to depth

Spearman’s rank-order correlations were employed to test for an association between depth and morphological variables (direct measurements as well as ratios) within K. aurantiaca and K. lobata.

Results

Interspecific differences

There was a statistically significant difference between K. aurantiaca and K. lobata in combined direct measurements of colony morphology (F[5, 74] = 15.143, Wilks’ Λ = 0.494, partial η2 = 0.506, p < 0.001). Mean lobe length was significantly longer in K. lobata (p < 0.001), mean lobe diameter was significantly thicker in K. lobata (p < 0.001), and K. aurantiaca had a significantly greater number of tips (Welch’s t-test p < 0.001) (Fig. 2b–d), while estimated volume and mean lobe height did not differ significantly (Fig. 2e, f). A statistically significant difference was also observed between K. aurantiaca and K. lobata in combined ratios of colony morphology (F[3, 76] = 24.018, Pillai’s trace = 0.487, partial η2 = 0.487, p < 0.001), with LL:LD, LH:CH, and Tips:LD all being significantly greater in K. aurantiaca (Welch’s t-test p < 0.001 in all cases) (Fig. 2g–i).

Fig. 2
figure 2

Mean differences between K. aurantiaca and K. lobata in depth distributions (a) and all measured variables of colony morphology (b–i) ± 95% CI and individual data points

Full size image

A statistically significant positive correlation was observed between mean lobe length and mean lobe diameter for both K. aurantiaca (rs[38] = 0.645, p < 0.001) and K. lobata (rs[38] = 0.602, p < 0.001) (Fig. 3a). The correlation between mean lobe height and colony height was significant and positive for K. aurantiaca (rs[38] = 0.473, p = 0.002), but not significant for K. lobata (rs[38] = 0.321, p = 0.044) (Fig. 3b). The correlation between number of tips and mean lobe diameter was significant and positive for K. aurantiaca (rs[38] = 0.588, p < 0.002), while that for K. lobata was not significant (rs[38] = 0.222, p < 0.169) (Fig. 3c).

Fig. 3
figure 3

Correlations between the variables comprising each of the morphological ratios for K. aurantiaca and K. lobata: a LL:LD; b LH:CH; c Tips:LD

Full size image

Species discrimination

For direct measurements of colony morphology, the logistic regression model was statistically significant (χ2[5] = 60.299, p < 0.001) and explained 70.6% of the variance in species classification (Nagelkerke R2). The model correctly classified 87.5% of cases, misclassifying five of 40 individuals for each species. Of the five predictor variables, only mean lobe diameter and number of tips explained significant variation in species classification (Table 1), with increasing mean lobe diameter and decreasing number of tips (since the variable was inverse transformed) associated with an increased likelihood of classification as K. lobata.

Table 1 Logistic regression results
Full size table

For ratios of colony morphology, the logistic regression model was also statistically significant (χ2[3] = 70.313, p < 0.001) and explained 78% of the variance in species classification (Nagelkerke R2). This model correctly classified 90% of cases, misclassifying only five K. aurantiaca individuals and three K. lobata individuals. All three ratios contributed significantly to the model (Table 1), with increasing LL:LD, LH:CH, and Tips:LD ratios all associated with a decreased likelihood of classification as K. lobata. The area under the ROC curve for the analysis of direct measurements was 0.933 (95% CI, 0.878–0.988) and for the analysis of ratios was 0.952 (95% CI, 0.905–0.998), both indicating an “outstanding” level of discrimination (sensu Hosmer et al. 2013). Potentially acceptable cut-off values could only be produced for mean number of tips and mean lobe diameter (Table 2).

Table 2 Variables with (potentially) acceptable and near-acceptable cut-off values for assignment of a given specimen as K. lobata based on ROC curve coordinates
Full size table

The logistic regression equation for direct measurements was:

$$\mathrm{log}\left(\frac{p}{1-p}\right)=-14.153-\left(1.615\times \mathrm{Estimated volume }\left[\mathrm{inv}\right]\right)+\left(1.317\times \mathrm{Mean lobe length }\left[\mathrm{sqrt}\right]\right)+\left(9.723\times \mathrm{Mean lobe diameter }\left[\mathrm{sqrt}\right]\right)-\left(1.374\times \mathrm{Mean lobe height }\left[\mathrm{log}10\right]\right)+\left(7.801\times \mathrm{Number of tips }\left[\mathrm{inv}\right]\right)$$

The logistic regression equation for ratios was:

$$\mathrm{log}\left(\frac{p}{1-p}\right)=7.329-\left(1.920\times \mathrm{LL}:\mathrm{LD}\right)-\left(8.431\times \mathrm{LH}:\mathrm{CH}\right)-\left(0.264\times \mathrm{Tips}:\mathrm{LD}\right)$$

Intraspecific responses to depth variation

Statistically significant correlations were only found between depth and three of the morphological variables examined for K. lobata. These were all weak to moderate and include a positive correlation between depth and LL:LD (rs[38] = 0.329, p = 0.038) (Fig. 4a) and negative correlations between depth and mean lobe height (rs[38] =  − 0.459, p = 0.004) as well as LH:CH (rs[38] =  − 0.433, p = 0.005) (Fig. 4b, c). For K. aurantiaca, no statistically significant correlations were found between any of the examined variables and depth.

Fig. 4
figure 4

Significant correlations between morphological variables and collection depth (m) for K. lobata: a LL:LD; b Mean lobe height; c LH:CH. Trendlines presented to help visualisation

Full size image

Discussion

In the present paper, we show how a novel species discrimination method elucidates colony growth form variation between the New Zealand soft corals Kotatea aurantiaca and K. lobata, which are morphologically similar. This has shown that macroscopic features of colony morphology can be taxonomically informative in soft corals and, although further testing is required, the general approach has the potential to differentiate between other soft coral species where molecular markers or sclerite comparisons cannot be used or are uninformative for species discrimination.

Interspecific differences between Kotatea aurantiaca and Kotatea lobata

Compared to K. lobata, K. aurantiaca has more numerous terminal lobe tips and greater tip development, forms a more prominent stalk, and possesses primary lobes that are shorter, thinner, and longer relative to their diameter. While primary lobe length increases with diameter in a similar way for both species, only in K. aurantiaca does lobe height increase with colony height, and the number of terminal tips increase with average primary lobe diameter. Thus, while primary lobes become thicker as they grow longer in both species, K. aurantiaca develops a stalk as the colony grows and continues to add terminal lobe tips, whereas K. lobata does not develop a stalk and grows mainly by progressively thickening its few lobes rather than by adding more. However, despite differences in mean values between the two species, K. aurantiaca and K. lobata overlap in every single measured morphological variable. These analyses thus confirm the initial observations that general morphological differences do exist between these species, but that these are difficult to discern for individual colonies due to a considerable degree of interspecific overlap.

Species discrimination

Determining how species can be discriminated is central to an understanding of biodiversity and the patterns and processes that drive it, because discrimination allows us to recognise whether a given set of specimens constitute members of the same species or not (Dayrat 2005). Morphometric analyses similar to those employed here are ubiquitous throughout comparative zoology (reviewed by Adams et al. 2004). In hexacorals and gorgonian octocorals, many more informative characters are available for interspecific comparison than in soft corals because these taxa do not exhibit colony expansion/contraction behaviour to the same degree. These include surface area, surface-to-volume ratio, average inter-calyx distance or inter-polyp, polyp density, branching pattern, and branching angles (e.g., Weinbauer and Velimirov 1998; Prada et al. 2008; Einbinder et al. 2009; Soto et al. 2018). Doszpot et al. (2019) used 3D modelling to analyse “area of occupied space” in scleractinian corals, for example. None of these characters is reliable for interspecific comparisons between Kotatea species, and we suggest that they are likely to be unreliable for other soft corals as well. Soft corals thus present an inherently difficult subject for examinations and taxonomic assessments based only on colony morphology.

Despite the observed morphological overlap, both logistic regression analyses were able to assign specimens correctly to species with a high degree of confidence. For the model based on direct measurements only, two of five variables—mean lobe diameter and number of tips—were significant, and both had much higher odds of classification as K. lobata for each 1-unit increase in these variables. This is because a 1-unit increase in the odds ratios covers almost the entire range of the transformed data, which is ~ 0.5–1.7 for the square root of the mean lobe diameter and ~ 0.01–1.00 for the inverse of the number of tips. Thus, greater lobe diameter and fewer tips translate to greater odds of classification as K. lobata.

For the regression model based on ratios, all three variables were significant, and increases in all were associated with a decreased likelihood of classification as K. lobata. Overall, the fact that 90% of specimens were correctly classified using ratios compared to 87.5% using direct measurements, and considering the Nagelkerke R2 and area under ROC curve values for both analyses, the regression model based on ratios performed slightly better. Misclassified specimens in both analyses were primarily composed of the smallest individuals. This may be because the smallest specimens tend not to exhibit much branching or very thick lobes and thus lack the features most useful to species discrimination. Regardless, these results indicate that the ratios are successfully controlled for a given specimen’s state of expansion/contraction and that species discrimination is less reliable when this is not taken into account. The three ratios improved the fit of the model compared to the two direct measurements suggesting that more useful morphological information can be extracted from ratios than from direct measurements. The use of morphometric ratios for species discrimination is a routine in the taxonomic study of many groups (Baur and Leuenberger 2011). While ratios have been used in gorgonian octocorals (e.g., Brazeau and Lasker 1988), they have not been used in soft corals. These new observations show that, while colony-scale features are often regarded as uninformative in soft corals, morphometric ratios can be taxonomically informative and may be useful for ongoing research into species boundaries, particularly for taxa between which sclerite differences are small.

Cut-off values are used as dividing points, which can separate continuous data into categories. Here, however, the application of cut-off values to separate specimens into species based on their morphological traits proved somewhat impractical. Chiefly, this is due to the extent of morphological overlap between K. aurantiaca and K. lobata. Cut-off values for mean lobe diameter and number of tips, if implemented, result in ~ 80% correct species assignment. Such information could be incorporated into future identification guides as it shows potential utility for field measurements and may, for example, aid in preliminary identifications of newly collected material. The next best cut-off value was for mean lobe length, with misclassification rates of ~ 30%. Depending on the study, such rates could possibly be acceptable and taken into account when collecting data or designing experiments. While these cut-off values can all be refined in the future through the use of larger sample sizes, K. aurantiaca and K. lobata may simply be too variable to be discriminated based on any single measure of colony morphology at much higher rates than these.

Phenotypic plasticity

Currently, one of the main challenges facing octocoral taxonomy is the uncertainty surrounding the extent and relevance of intraspecific morphological variation in conjunction with insufficient interspecific variation among many commonly used molecular markers (Pérez et al. 2016), which makes it difficult to establish species boundaries (McFadden et al. 2010). In octocorals, phenotypic plasticity is often examined with the aim of reviewing the taxonomic status of two or more morphotypes, which may then be split into several species (e.g., Soler-Hurtado et al. 2017) or retained as one (e.g., Bilewitch et al. 2010). Here, depth-related phenotypic plasticity was examined for K. aurantiaca and K. lobata to address the lack of ecological information available for both of these commonly encountered New Zealand endemics. Evidence for phenotypic plasticity being associated with depth was only found for K. lobata, which is surprising because K. aurantiaca occupies a much greater depth range. Interestingly, a near identical situation has been reported for two species of Mediterranean Sea gorgonians, whereby plasticity in a usually shallower-living species has resulted in it being misidentified as a more morphologically stable and deeper-ranging species (Pica et al. 2018).

It may be that K. lobata exhibits phenotypic plasticity because this species tends to inhabit much shallower depths, extending even into the intertidal zone, where it is subjected to greater environmental variability than the deeper-dwelling K. aurantiaca. This may include tidal water movement, wave stress, storm damage, and fluctuating light, salinity, temperature, and food availability—many of which are known to influence gorgonian colony form (e.g., Wainwright and Dillon 1969; Leversee 1976; Velimirov 1976; West 1997). Kotatea lobata has a higher LL:LD ratio at greater depths, equating to lobes that are longer and thinner than at shallower depths. It may be that the thick and robust lobes of K. lobata are necessary for its colonies to withstand wave stress or to optimise water retention when exposed intertidally, while at deeper depths, it can optimise surface area and food capture by forming colonies with thinner and more sprawling lobes, as has been reported for related soft corals (Sebens 1984). However, why mean lobe height and LH:CH should decrease with depth in K. lobata is unclear. This may be a plastic response related to sedimentation (Prada et al. 2008) or predation (West et al. 1993), but data incorporating a more complete range of depths and environmental conditions are needed to draw conclusions regarding the relationships between morphological variables and depth in both species.

Octocorals are highly species-specific in their responses to abiotic factors (Rodríguez-Lanetty et al. 2003), and thus, it is difficult to compare the results and ecological interpretations presented by different studies. While depth-related phenotypic plasticity has been recorded many times for gorgonians (e.g., West et al. 1993; Rodríguez-Lanetty et al. 2003; Kim et al. 2004; Gori et al. 2012; Costantini et al. 2016; Calixto-Botía and Sánchez 2017), this has not been reported for soft corals. Because of their contrasting colony architecture, the phenotypic plasticity observed in gorgonians (which do not exhibit colony expansion/contraction behaviour) cannot be directly compared to that observed in K. aurantiaca or K. lobata in most cases. Branch development is a notable exception, but comparisons of plasticity in this trait are problematic because the vast majority of taxa in which this has been examined are zooxanthellate, whereas K. aurantiaca and K. lobata are azooxanthellate. For example, zooxanthellate gorgonians commonly exhibit decreased branch development at greater depth (e.g., Lasker et al. 2003; Calixto-Botía and Sánchez 2017), which may minimise self-shading at reduced light levels (Brazeau and Lasker 1988). Kotatea lobata shows an opposite pattern, tending to be more finely divided at greater depths, while K. aurantiaca is variable across its entire depth range. Clearly, zooxanthellate and azooxanthellate species experience different selection-by-depth pressures and therefore express different phenotypic plastic responses.

The taxonomic status of K. aurantiaca and K. lobata

In light of the overlap in colony morphology and the lack of species-level monophyly in K. aurantiaca and K. lobata (Kessel et al. 2022), it may be posited that these are not two but one species, with K. lobata and K. aurantiaca representing shallow and deep ecotypes, respectively. Indeed, under this interpretation, observations made for K. aurantiaca and K. lobata closely match the results obtained by Prada et al. (2008), who demonstrated thicker branches and a tendency to grow in a single plane in a gorgonian species at shallower depths. Accordingly, K. lobata lives at shallow depths, can grow in a single-plane fashion (Kessel et al. 2022), and has thicker lobes than K. aurantiaca. However, this is unlikely to be the result of phenotypic plasticity and in this case is better explained by interspecific variation based on fixed morphological differences. This is strongly supported by the pronounced sclerite differences (Kessel et al. 2022) and the high accuracy of discrimination based on colony morphology presented here. Were both species part of a broader spectrum of phenotypic plasticity within a single species, one would expect to see less consistency in marked sclerite differences and lower accuracy in discriminating between (what would then be) ecotypes.

Studies comparing morphological and genetic variation in octocorals have variously concluded that distinct ecotypes may represent different species (e.g., Soler-Hurtado et al. 2017) or intraspecific variation (e.g., Gutiérrez-Rodríguez et al. 2009; Bilewitch et al. 2010), depending on whether genetic and morphological differences were concordant with one another. Occasionally, authors have also expressed uncertainty regarding the taxonomic status of apparent ecotypes (e.g., Gori et al. 2012; Costantini et al. 2016). Here, the interpretation that the lack of genetic variation between K. aurantiaca and K. lobata in mitochondrial mtMutS and nuclear 28S indicates a lack of species-level resolution in the selected markers is supported (Kessel et al. 2022). However, both K. aurantiaca and K. lobata meet the proposed threshold for accurate discrimination of species in Alcyonium identified by McFadden et al. (2014a) for mean genetic p-distances in mtMutS (0.5%) (Kessel et al. 2022), although mtMutS often lack the resolution needed to discriminate between congeneric species (e.g., Sánchez et al. 2003b; Wirshing et al. 2005; Cairns and Bayer 2005; Cairns and Baco 2007; McFadden et al. 2006, 2009). Additionally, nominal species of Alcyonium have been shown to share identical haplotypes for both mtMutS (McFadden et al. 2011) and 28S (McFadden et al. 2014a). Perhaps most importantly, in cases where morpho-molecular comparisons are inconclusive, clear and consistent morphological differences are commonly weighted more highly than a lack of phylogenetic resolution in new species descriptions (e.g., van Ofwegen et al. 2007; Moore et al. 2016; Núñez-Flores et al. 2020).

Limitations and future research

Ideally, predictive models such as those discussed here should be trained and then tested on separate data sets. However, in the present study, available material is not abundant enough for this method to be effective, and virtually all available specimens of K. aurantiaca and K. lobata were instead used in the building of the logistic regression models to include as much morphological information as possible.

The novel techniques presented here may be useful in discriminating between other sets of problematic soft coral species where molecular and sclerite comparisons are impractical or uninformative. This approach also represents a potential method for discriminating soft coral species in situ, without the need for collecting specimens. However, future collection of additional specimens will be necessary to test the performance of the models and to compare the results obtained with preserved material against fresh samples or measurements taken from in situ photographs, for example. Larger sample sizes based on targetted collections from a broader range of depths and locations will also improve the ability to discern patterns relating to intraspecific variation and could allow for examination of other environmental factors that may play a role in determining the morphologies of these species. Reciprocal transplant experiments (e.g., West et al. 1993; West 1997; Prada et al. 2008; Calixto-Botía and Sánchez 2017) or phylogenetic, phylogeographic, and population genetic analyses (e.g., Andras et al. 2012; Holland et al. 2017; Yesson et al. 2018) utilising new markers and additional specimens could be especially informative in re-evaluating intraspecific variation in both species, as well as in confirming their separate species status.

Conclusions

While phenotypic plasticity with depth was only observed in K. lobata, both K. lobata and K. aurantiaca overlap considerably in their colony morphologies. Despite this, specimens can be assigned to either K. aurantiaca or K. lobata with a high degree of confidence based only on measurements of colony morphology and their ratios, demonstrating the taxonomic utility of colony-scale morphometrics in soft corals and supporting the separate taxonomic status of both species. This study marks the first use of ratios to control for the variable state of expansion/contraction among soft coral specimens, and because these ratios performed slightly better at species classification, their use is recommended over direct measurements. It is hoped that this species discrimination approach will stimulate further research on these New Zealand endemics, as well as on other sets of similar soft coral species for which species discrimination has so far been problematic.