Introduction

Rhodoliths are free-living, long-lived, and slow-growing non-geniculate calcareous algae (Rhodophyta: Corallinaceae, also known as maërl). They accumulate as rhodolith beds in subtidal areas characterized by coarse, detritic sediments consisting of gravels and clean sands (Illa-López et al. 2023). They are distributed globally (Riosmena-Rodríguez et al. 2017) in environments where bottom currents prevent their burial by finer sediment, and where irradiance is low but still adequate for photosynthesis (Peña et al. 2014; Carvalho et al. 2020). In consequence, the depth limit of rhodolith beds is determined by the optical properties of the water body (transparency, turbidity, sediment load), affecting the quality and intensity of light reaching the seabed (Basso et al. 2017). In Mediterranean islands, rhodoliths thrive at great depths, reaching 100 m in the Mediterranean western basin (Ballesteros 1988; Barberá et al. 2012a) and reported at a maximum depth of 137 m for the Mallorca channel in the Balearic Islands (Massutí et al. 2022).

Rhodolith beds build hard, three-dimensional complex structures that provide substrate for other algae and invertebrates (Bordehore et al. 2003; Teichert 2014). These organisms are often referred to as ecosystem engineers (Nelson 2009). Rhodoliths can acquire various complex morphologies, from highly branched to massive round thalli (Bosence 1983). Therefore, they provide a greater array of microhabitats at different scales, from the interstitial spaces between branches (Bernard et al. 2019; Fredericq et al. 2019; Boyé et al. 2019) to the seascape level by increasing habitat heterogeneity (Amado-Filho et al. 2017). In consequence, the accumulation of rhodoliths over sandy sediments contributes to increase the local and regional biodiversity (Bernard et al. 2019; de Juan et al. 2023).

Previous studies identified a relationship between rhodolith morphological complexity and the associated biodiversity. In particular, the diversity of cryptofauna is related to thallus size, growth form and structural complexity (Kamenos et al. 2004; Steller et al. 2003; Figueiredo et al. 2007), with higher diversity associated with highly branched forms (Steller et al. 2003; Foster et al. 2007; Meihoub-Berlandi et al. 2012; Méndez Trejo et al. 2021). High density of rhodoliths also favors higher benthic fauna richness (Veras et al. 2020), while larger sizes of rhodoliths seem to facilitate reproduction and feeding in fishes and invertebrates (Steller et al. 2003; Gagnon et al. 2012).

Physical processes such as sedimentation or turbulence can affect rhodolith morphology and, therefore, their associated communities (Pérez-Peris et al. 2023). Trawl or dredge fishing contribute to increased sedimentation rates (Puig et al. 2015) and to physical damage of the rhodolith (e.g., breaking or chipping) (Cabanellas-Reboredo et al. 2017; Gabara et al. 2018). Rhodolith beds have a high conservation value; they are included in the European “Habitats Directive” (Council Directive 92/43/EEC). However, in the Mediterranean, they continue to face threats from trawling activities (de Juan et al. 2013) and current protection measures are often inadequate. Due to their slow growth (of the order of 1 mm yr−1, Blake et al. 2007) rhodoliths exhibit exceptionally low renewal rates, which can result in slow recovery trajectories (Tauran et al. 2020).

This study focuses on rhodoliths from a well-preserved rhodolith bed in the Menorca Channel (Balearic Islands, Western Mediterranean, 3°45′ E, 39°75′ N) (Barberá et al. 2012a). The main objective of the study is to describe and measure the morphology of rhodoliths as a proxy for habitat structural complexity. By analyzing samples from the well-preserved rhodolith bed, this study (1) analyzes the diversity of rhodolith morphologies; (2) determines the allometric growth curve for each of the major morphologies; (3) produces a discriminant rule to classify rhodoliths based on few morphometric measurements; (4) compares 2D with 3D measurements of branched rhodoliths to examine whether simple, relatively easy to measure 2D metrics would be sufficient to describe the morphometry of a naturally occurring 3D object. This is relevant because, given the intricate three-dimensional growth pattern of rhodoliths, 2D metrics may lack the precision and accuracy needed to effectively describe their shape or interstitial space. This estimation is important, as it is directly linked to the ecological functions facilitated by rhodoliths, which inherently exist in three dimensions.

The understanding of the structural complexity of rhodolith beds, achieved by exploring objectives 1–4, is essential for deducing their functional importance (Stelzer et al. 2021) and to improve their protection and preservation status globally (del Río et al. 2022).

Methodology

Study area

The study area is located in the Menorca Channel between 45 and 80 m depth (Fig. 1) and was selected based on previous studies (Moranta et al. 2014) that identified a well-developed and preserved rhodolith bed that alternates with detritic bottoms and biogenic assemblages formed by other species of algae (de Juan et al. 2023). The Menorca Channel was declared a European Site of Community Importance (LIC ESZZ16002) in 2016, including a trawling ban for most of the channel. Moreover, the presence of underwater communication cables crossing the channel delimits an exclusion area that has been protected from trawling for more than five decades.

Fig. 1
figure 1

Source: INDEMARES (Moranta et al. 2014)

Map of the study area in the Menorca Channel, showing the positions of the dredges (green dots). Area limited by the cables zone (“Cables zone” in the figure, established in 1971) and the “Trawl Exclusion Reserve” (declared 2016, green cross-hatch).

Full size image

The presence of the cables’ exclusion zone, a gently sloping continental shelf (reaching a maximum depth of 100 m), and the prevalence of strong bottom currents create optimal conditions for the development of rhodolith beds. As a consequence, the study area includes the largest identified rhodolith bed in the region, which has an extension of 304 km2 (Fig. 1). As the distance from this bed increases, the depth increases toward the south-east and decreases toward the coast (west), with the north-west area being influenced by the sandy platform of the Bay of Alcudia (Fig. 1). In addition to the variation of these environmental factors, the density of rhodoliths decreases as the distance from the center of the rhodolith bed increases, hereafter referred to as Core Habitat (Fig. 1).

We assessed the variation of rhodolith morphology in relation to the distance to the Core Habitat. Five sampling stations were deployed from the center to the periphery of the rhodolith bed (or Core Habitat). Sampling stations were thus classified into: (1) “Core Habitat”, including samples within the cable zone at 69 m depth (S13) and in the trawl exclusion area close to the cable zone at 58.8 m depth (S15), where rhodolith cover is typically > 50%; (2) “Adjacent Habitat”, including samples far from the cables zone at 56 and 74 m depth (S18 and L1, respectively), where rhodolith cover is less than 50%, and “outside the reserve area” at 48.5 m depth (S5), where rhodoliths are scarce (Fig. 1). The percentage cover of rhodoliths at each station was extracted from Moranta et al. (2014), where a spatially structured survey was conducted to estimate the habitats distribution in the channel. To verify the rhodolith coverage at a small scale (1–10 m), additional video images were obtained using still cameras (see Supplementary Material—Fig. S1).

Data acquisition

We collected four samples per station using a “Naturalist” Dredge (frame 45 × 18.5 cm, fitted to a 1 mm mesh bag) in summer 2022. Nineteen valid samples were recovered, as only 3 dredges could be collected in S5 (example in Supplementary Material—Fig. S2). For each sample, a volume of approximately 7000 cm3 of the content was systematically extracted, distributed in a tray and photographed. The proportion of live and dead rhodoliths was visually estimated on board for each tray.

The rhodoliths were classified into nine morphotypes (A to I), described based on the six used in Barberá et al. (2012b) with some modifications that further divided heterogeneous groups (Table 1). These nine morphotypes belong to the three main rhodolith categories or growth forms according to Basso et al. (2016): “Branched” (Br), “Praline” (PR) and “Bowork” (Bo). The rhodoliths were grouped by morphotypes, photographed on a graduated white board and dry stored. A total of 3877 rhodoliths were counted in the photographs (see Supplementary Material Table S1). The number and frequency of the morphotypes (A to I) and growth forms in each sample were estimated from the photographs.

Table 1 Rhodolith growth forms (top) and description of morphotypes defined in the finer classification (A to I)
Full size table

A sample of 100 g of sediment extracted from each dredge was stored for granulometry analysis, while the remaining content of the dredges was returned to the sea. A subsample of 20 g of sediment was used for the analysis of grain size. The granulometry was analyzed using a laser diffraction particle size analyzer (HORIBA LA-950V2) after disaggregation and removal of the biogenic fraction (10% H2O2 solution) and dispersion in an ultrasonic bath.

2D measurements

From the rhodolith samples (i.e., 3877 rhodoliths photographed on board), five individuals for each morphotype per station (when possible) were selected and set aside. This selection resulted in a total of 189 rhodoliths belonging to the nine morphotypes (Supplementary Material Table S2). These 189 rhodoliths were photographed individually to obtain 2D morphological metrics. A total of 15 metrics were obtained for each rhodolith. The maximum and minimum axis of the fitted ellipse (Amax and Amin, mm), the surface of the projected area (A, mm2), convex hull area (CHA, mm2), shape perimeter (P, mm), the angle (a, °), the Feret length (Lmax, mm) and Feret’s angle (Fa, °) were obtained measuring the rhodolith shape and estimating the convex hull of the shape. These metrics were obtained using the default measurement routines in ImageJ 1.52a software (Schneider et al. 2012), except for CHA which was obtained with the plugin “Hull and Circle” v.2.1b contributed by A. Karperien (http://imagej.net/ij/plugins/hull-circle.html).

In addition, the dimensionless metrics circularity (C), roundness (R), aspect ratio (AR) and solidity (S) were automatically determined by ImageJ program according to the following formulas:

$$C = 4\pi \frac{A}{{P^{2} }},$$
$$R = 4\frac{A}{{\pi A_{\max }^{2} }},$$
$$AR = \frac{{A_{\max } }}{{A_{\min } }},$$
$$\it S = \frac{A}{{{{CHA}}}}.$$

Finally, three additional metrics were obtained:

1. the interstitial space (IS, mm2) as the missing space (area) between the convex hull area and the rhodolith surface area:

$${\text{IS}} = {CHA} - A,$$

2. the individual dry weight (W, g) and

3. the shape coefficient (D, Kooijman 2010) extracted from:

$$D = \sqrt[3]{{\frac{W}{{2.71\,{L_{\text{max}}}^{3} }}}},$$

with \(d\) = 2.71 g/cm3 being the density of calcite.

2D/3D measurements in branched morphotype A

Given that the difference between 2 and 3D metrics is likely to be more noticeable in the branched growth form, this analysis focuses only on morphotype A. We subsampled a new set of 30 rhodoliths (from the stored dry sample) classified as branched (morphotype A) and collected in the stations S13 and S15 (Supplementary Material Fig. S3). The 30 rhodoliths were photographed and dry weighed individually. Then, the images were processed through ImageJ as described in the previous section. To compare the 2D and 3D information extracted, the measurements assessed are subscripted “2D”: the maximum ellipse axis (Amax-2D), the maximum Feret diameter (Lmax-2D), the perimeter (P2D), the surface area (A2D), the convex hull area (CHA2D), the roundness (R2D), the solidity (S2D) and the interstitial space (IS2D).

For 3D measurements, digital twins of selected rhodoliths were created using high-precision blue light 3D scanner (Revopoint MINI) with a precision of up to 0.02 mm and the associated programs Revoscan and Revostudio (www.revopoint3d.com). For each 3D model twin, the surface area (A3D, mm2), the maximum length from the center of mass (L3D, mm), the volume (V3D, mm3) and the convex hull volume (CHV3D, mm3) were calculated using an ad hoc R script based on libraries Rvcg, geometry and MeshesTools, while the maximum Feret diameter (Lmax-3D) and the roundness (R3D) were obtained using Slicer program v. 5.2.2 (Fedorov et al. 2012). With these measurements, the 3D interstitial space (IS3D, mm3) was calculated as the difference between the volume of the convex hull and the volume of the rhodolith:

$$\it {\text{IS}}_{{3{\text{D}}}} = {{CHV}}_{{3{\text{D}}}} - V_{{3{\text{D}}}} .$$

The solidity in 3D (S3D) was calculated with the relation between the 3D rhodolith volume and the 3D convex hull volume:

$$\it S_{{3{\text{D}}}} = \frac{{V_{{3{\text{D}}}} }}{{{{CHV}}_{{{\text{3D}}}} }}.$$

Finally, we obtained a matrix of 17 metrics (8 from 2D, 8 from 3D and the weight) measured for the 30 rhodolith samples (see Supplementary Material Table S3).

Statistical analysis

Morphological rhodolith composition of the study area

Differences in the proportion of growth forms (Br, Pr and Bo) between Core Habitat and Adjacent Habitat stations were analyzed with PERMANOVA, with a model design including “Station” nested within Habitat factor with two levels (Core and Adjacent Habitat). Both factors were treated as fixed factors. The proportions were arcsin transformed for this analysis. The routine adonis2 of R package vegan was used (Oksanen et al. 2022).

2D: General measurements analysis

After removing nine outliers (determined through exploratory data analysis with the aid of boxplots), 180 measured rhodoliths were used to fit several models. Firstly, a regression model was performed to assess the W/L relation, with L corresponding to the Feret’s diameter, by linearizing the allometric growth equation:

$$W = aL_{\max }^{b} .$$

This was done for the pool of 180 rhodoliths and for each category or growth form (Br, Bo, Pr) separately. Then, to evaluate the significance of the difference in the values of the a and b parameter of this relation between the three growth forms, we performed a generalized linear model (GLM) from package stats (R Core Team 2022) fitted with a Gaussian distribution with logarithm link. To examine the possibility of overdispersion in the data, we performed a dispersion test from the package AER (Kleiber and Zeileis 2022).

2D: Multivariate analyses

A PCA was performed on the full 2D metrics data set to reduce the number of variables while preserving the maximum variance possible of the original data set. This reduction in dimensionality facilitates visualization and data interpretation. Function prcomp from the R stats package was used. To accomplish that, we standardized all 15 morphometric variables measured from the 180 rhodoliths (Table 2).

Table 2 Graphical representation of the 15 morphometric variables measured
Full size table

To examine how these variables perform as morphology classifiers, we selected the less correlated variables based on the Pearson correlation test from package corrplot (Wei and Simko 2021) to perform a linear discriminant analysis (LDA) from package MASS (Venables and Ripley 2002). The numeric variables used in the model were: maximum Feret diameter (Lmax), roundness (R) and solidity (S) and the classification factor was the growth form (Br, Pr and Bo). To fit the model, the variable Lmax was log transformed, and solidity and roundness were arcsin transformed. With these transformations, the assumptions for LDA of multivariate normality and homoscedasticity between groups were met. Multivariate normality of the metrics by morphology was not rejected according to the Royston’s multivariate normality test (p value = 0.05 for branched; p value = 0.40 for pralines and p value = 0.88 for boxwork), using the MVN package (Korkmaz et al. 2014). The hypothesis of homoscedasticity could not be rejected based on Box’s M-test for homogeneity of covariances matrices (X2 = 18.562, df = 12, p value = 0.0996) of the biotools package (da Silva 2021). The matrix of 180 individuals was randomly re-sampled to split the data into a training data matrix (70%) to train the model, and a test data matrix (30%) to test the model. To achieve better prediction accuracy, this process was repeated a total of 100 times.

2D/3D morphotype A measurement analysis

With the matrix of 17 morphological variables (8 in 2D, 8 in 3D and the weight), a Pearson correlation test was performed from package corrplot (Wei and Simko 2021) to assess the correlation between the two measurements. Furthermore, considering the total available surface area and the volume as key morphological measurements of high biological interest for metabolic or growth studies, linear models from package ‘stats’ were fitted between the surface area measured in 2D (A2D) and in 3D (A3D) and between the surface area measured in 2D (A2D) and the total volume measured in 3D (V3D).

All statistical analyses were computed in R (R Core Team 2022) under version 4.2.1.

Results

Morphological rhodolith composition of the study area

The proportion of dead (bleached) rhodoliths was predominantly low (< 20%) with exceptions found in three samples from station L1 and two samples from stations S5 and S13 that had no dead rhodoliths, while three samples from S18 had higher percentages, one reaching 50% coverage by dead rhodoliths (Supplementary Material Fig. S4).

All stations exhibited similar sediment compositions, primarily gravel (85% or higher), with minimal silt and clay (never exceeding 15%). Branched forms predominated in 50–90% of the samples, with lower proportions of pralines (5–40%) and boxwork (0–30%). However, variations in rhodolith morphologies were observed within samples collected at the same station, suggesting high small-scale heterogeneity. The Adjacent Habitat station L1 was very heterogeneous, and boxwork forms were more prevalent in several samples (Fig. 2).

Fig. 2
figure 2

Ternary plot of sediment composition showing the proportion of gravels, sand, silt and clay (right) and ternary plot of rhodolith growth forms composition showing the proportion of branched, pralines and boxwork rhodoliths (left). Each dredge was represented and color coded following the stations sampled

Full size image

Of the 3877 rhodoliths photographed, 69.2% were branched, 19.1% pralines and 11.8% boxwork (Fig. 3, left panel). From these, the most abundant morphotypes were the branched types C (37.53%), B (14.70%) and A (10.14%), and praline E (14.24%) (see Table 1 for reference to the morphotypes). The proportions and representation of rhodolith morphotypes remained consistent across the two habitats (PERMANOVA p(F) = 0.211); however, the PERMANOVA test identified significant effects of the factor station within habitat (PERMANOVA p(F) = 0.003).

Fig. 3
figure 3

Total proportions of each growth form (Br, Pr and Bo) considering all stations together (left). Percentage of morphotypes (A to I) coverage in each station (right)

Full size image

At S5, the dominant morphotypes were branched (D) and praline (F), which were scarce at other stations. These differences likely stem from the habitat, as S5 is characterized by sand with scattered rhodoliths, and it is considered outside the rhodolith bed. Consequently, S5 was excluded from the statistical analyses conducted in this section and the subsequent one.

The stations within the Core Habitat exhibited more consistent proportions of morphologies compared to those in the Adjacent Habitat. Specifically, the dominant branched morphotype A decreased in abundance from station S13 to S15 and further to L1, where it was absent, while the branched morphotype C remained consistently abundant across all stations, with the highest representation observed at adjacent station S18. The non-branched morphotypes were never dominant, but appeared relatively frequently in station L1. The praline morphotype E was represented equally in all stations. The praline morphotype G occurred mostly in the stations from Core Habitat. The boxwork growth forms (morphotypes H and I) were proportionally represented in all core and adjacent stations except in S15 where mophotype H was predominant.

2D morphological measurements

The PCA revealed the main morphometric variables describing the 2D rhodolith shape. These results showed that 46.7% of the variance could be explained by the first dimension, while the second dimension explained 25.2% of the variance. Figure 4 shows that the most differentiated group is the one formed by the branched forms. Praline and boxwork overlapped in the biplot.

Fig. 4
figure 4

Principal components analysis (PCA) of individual rhodolith metrics from the Menorca Channel. A area, a angle, Amax maximum ellipse axis, Amin minimum ellipse axis, AR aspect ratio, C circularity, CHA convex hull area, D shape coefficient, Fa Feret angle, IS interstitial space, Lmax maximum Feret diameter, P perimeter, R roundness, S solidity, W weight

Full size image

The variables maximum Feret diameter (Lmax), area (A) and convex hull area (CHA), weight (W) and the maximum and minimum ellipse axis (Amax and Amin) were highly correlated and their variability is represented by the first dimension (PCA X-axis). Shape coefficient (D), circularity (C), solidity (S) and roundness (R) were also correlated and defined by the second dimension (PCA Y-axis). Aspect ratio (AR) and roundness (R) were negatively correlated, indicating that branched forms have lower roundness (and higher aspect ratio) than pralines and boxwork rhodoliths.

Because of the high correlation between several shape measurements, only a representative subset (Lmax, IS, W, A, D and P) was used to examine differences between samples recovered from the Core Habitat and the Adjacent Habitat (Fig. 5). Boxwork growth forms tended to be bigger, longer and heavier than praline and these than branched forms (Fig. 5). Moreover, there was a clear effect of the Core Habitat in terms of interstitial space, maximum Feret diameter and area. These metrics had greater values in all growth forms in the stations within the Core Habitat compared to the stations from the Adjacent Habitat.

Fig. 5
figure 5

Boxplots for a subset of six morphological variables both in Core Habitat and in Adjacent Habitat for the three growth forms of rhodoliths. From left to right at the top: interstitial space (IS), maximum Feret diameter (Lmax) and perimeter (P). From left to right at the bottom: area (A), shape coefficient (D) and weight (W). All variables were log transformed

Full size image

The allometric growth curves (W = a Lb) were estimated for each rhodolith growth form based on the linear models fitted for the relationship between weight and the Lmax (Fig. 6). The parameter a was significantly different between boxwork and branched curves (p value < 0.001), but not between praline and branched curves (p value = 0.097). The parameter b showed no significant differences between growth forms. Therefore, the three groups showed a similar allometric growth rate, despite the groups having significantly different weights for the same length. The test for overdispersion was significant (p value < 0.001), but with a low value of dispersion 1.80.

Fig. 6
figure 6

Rhodolith allometric growth curves, representing the relation between Lmax and W for each morphotype or growth form

Full size image

Discriminant analysis

A linear discriminant analysis was used to build a classifier to predict growth forms from a set of relatively easy to measure variables. The results of the LDA showed that, although the three groups had a large degree of overlap, they can be statistically distinguished. The variables with the greatest weight in defining the first dimension of the LDA were the roundness (R) and the solidity (S), while the maximum Feret diameter (Lmax) had the greatest weight in the second dimension. The first LDA axis explained 96.6% of the variance. This first direction separates the rhodoliths with greatest solidity (less branching), roundness and relatively large Lmax in the positive axis, from the rhodoliths with less S (more branching) and less R in the negative axis. The second LDA dimension explained the rest of the variance (3.37%), driven by Lmax, with longer rhodoliths in the positive axis and shorter in the negative one. Thus, the first dimension separates the largest, most rounded and solid rhodoliths (both pralines and boxwork growth forms) from the branched forms (Fig. 7). The classification equations based on maximum Feret diameter, roundness and solidity are:

$${\text{LDA}}1:1.996 \, \left[ {L_{\max } } \right] + {\mathbf{0}}{\mathbf{.252}}\;[{\mathbf{R}}] + {\mathbf{0}}{\mathbf{.274}}\;[{\mathbf{S}}],$$
$${\text{LDA}}2:{\mathbf{21}}{\mathbf{.232}} \, [{\mathbf{L}}_{{{\mathbf{max}}}} ] + 0.123\;\left[ R \right] + 0.147\;\left[ S \right].$$
Fig. 7
figure 7

Representation of the linear discriminant analysis (LDA)

Full size image

Applying the estimation model to the data test, we found that the classification rate was correct in 63.15% of the cases. After reproducing the analysis 100 times and running the predictive model, the classification rate for a new rhodolith was 69.5%, with the percentages of success for each growth form being different (Table 3). The best classified form was always branched with 88.13% of success. The more common misclassification was obtained when classifying a boxwork as a praline (40.87% of times) or a praline as a boxwork (30.73% of times).

Table 3 Table of predictions when classifying rhodoliths according to the variables considered in the LDA
Full size table

3D morphological measurements in a subset of morphotype A

A 3D model twin of the complete rhodolith shape (Fig. 8) and its corresponding measurements were obtained for 30 rhodoliths. The correlation test showed that both sets of metrics (2D vs. 3D) are highly correlated (Supplementary Material Fig. S5), especially a metric on two different dimensions (e.g., area 2D and area 3D), but also between metrics in the same dimension (e.g., perimeter 2D with area 2D).

Fig. 8
figure 8

Example of 3D model twin from a morphotype A rhodolith. 3D model view scanning the QR

Full size image

The linear models (with intercept at 0) fitted between 2 and 3D area and between 2D area and the volume showed that both sets were highly correlated (R = 0.99 and R = 0.98, respectively) (Fig. 9). The 3D area resulted in 4.27 times the 2D area (Fig. 9a), while the volume was 6.45 times the 2D area (Fig. 9b). These equations can be useful in estimating the ecologically meaningful quantities (area, volume) in 3D from their equivalent quantities in 2D, which are much easier to measure.

Fig. 9
figure 9

Linear models and equations between area measured in 2D and (a) superficial area measured in 3D and (b) volume measured in 3D

Full size image

Discussion

The present study has characterized the morphological composition of a rhodolith bed in the continental shelf of the Menorca Channel (western Mediterranean). This area was selected due to its good conservation status, with a well-developed rhodolith bed (i.e., above 50% rhodolith coverage over 304 km2), and adjacent locations with decreasing rhodolith cover alternating with detritic sands (Barberá et al. 2012a; de Juan et al. 2023). It is common to find in rhodolith beds a variable quantity of dead rhodoliths and their fragments accompanying the live rhodoliths (Bracchi and Basso 2012). Our results generally show proportions of dead rhodoliths ranging from 5 to 20% in most samples (except for those from the Adjacent Habitat station S18). These values are similar to those found by Rendina et al. (2020) and support the hypothesis that the rhodolith bed in the Menorca Channel is in a good environmental condition. Despite that branched forms predominate in all stations, we observed small-scale heterogeneity in the proportions of growth forms and of morphotypes. The heterogeneity observed agrees with previous studies in the Mediterranean (i.e., Bracchi et al. 2022) and particularly in the Menorca Channel (Farriols et al. 2022). In general, the proportion of branched growth forms, and particularly the proportion of morphotype A, was highest in the Core Habitat.

The differences in the proportion of morphotypes between stations cannot be attributed to differences in sediment granulometry. The sediment below the rhodolith bed was coarse, with the sand fraction (grain size > 63 µm) comprising at least 50% of the sample in all stations and over 80% in S13, L1 and S5. In S15 and S18, the gravel fraction (grain size > 2 mm) comprised around 40% of the sample. This granulometry is typical of Mediterranean rhodolith beds (de Juan et al. 2013), although dominance of fine sediments (grain size < 63 µm, Bracchi et al. 2022; Ingrassia et al. 2023) or coarser sediments, including fractions larger than 20 mm (del Río et al. 2022), have been reported in Mediterranean rhodolith beds. Our findings corroborate the conclusions drawn by Cabanellas-Reboredo et al. (2017), indicating that, at the scale of a rhodolith bed, sediment granulometry has limited explanatory power regarding variations in rhodolith morphology. However, at larger spatial scales, coarse sediments are indicative of low sediment input from terrestrial origin, which is expected around Mediterranean islands. High currents predominating in the area (Moranta et al. 2014) would also prevent the deposition of fine sediments.

The specimens from the Core Habitat were larger and with more interstitial spaces, even considering the praline morphologies. It is likely that rhodoliths at the edge of the Core Habitat exhibit lower complexity, due to suboptimal environmental conditions combined with being exposed, directly or indirectly, to trawling. Interestingly, a decrease in morphotype A from S13 (in the Core Habitat and within the cables zone) to the edge of the Core Habitat and exclusion zone in S18 was observed suggesting that this morphotype is a potential candidate indicator of the conservation status of the habitat. Both natural and anthropogenic processes may be responsible for increased fragmentation of branched growth form, resulting in smaller and less ramified (more solid) branched rhodoliths (Barberá et al. 2012a, b). Branched rhodoliths exhibited a greater degree of branching and had higher interstitial space (IS) in samples from the Core Habitat, suggesting that good environmental condition of the rhodolith bed contributes to maintain the architectural complexity of this habitat. Habitat complexity is significant for enhancing local species richness (de Juan et al. 2013). The spaces between branches provide microhabitats for diverse assemblages of other algae, epiphytes and invertebrates including sponges, cnidarians or bryozoans (Phillips 2023) (see examples found in Supplementary Material—Fig.S6). As such, the rhodolith bed community in the protected Core Habitat, characterized by high diversity of rhodolith morphotypes, could be viewed as the reference for a good environmental condition. This holds significant implications for local conservation management, as often, such reference or baseline points are scarce (Lindeboom 1995).

Owing to legislation prohibiting trawling within the cables exclusion zone, there is a high degree of certainty that the Core Habitat has experienced no physical disturbances since the early 1970s. Initial examination of the data from both Core and Adjacent Habitat stations suggest a correlation between the condition of rhodolith beds and their level of protection. However, our study was not designed to assess this relationship. Extending sampling over a gradient of trawling intensity, with enough spatial coverage to incorporate environmental variability (e.g., see Hinz et al. 2009), would contribute to assess potential variability in rhodolith morphology linked to trawling activity. While our study supports that trawling has not impacted the rhodolith bed in the Core Habitat within the cables exclusion zone, the composition of Adjacent Habitat cannot be attributed solely to anthropogenic impacts. This recognition is important, considering that high small-scale heterogeneity within rhodolith beds (de Juan et al. 2023) may not exclusively result from fishing pressure (Jardim et al. 2022). Natural environmental factors such as depth, sediment grain size, bottom geomorphology and bottom currents (Illa-López et al. 2023; Pérez-Peris et al. 2023) can influence the density, dimensions and composition of rhodolith beds, independent of human intervention (Peña et al. 2014; Veras et al. 2020).

The morphological complexity of the rhodolith is a key driver for the high faunistic and floristic diversity of rhodolith beds (Fredericq et al. 2019; Gabara et al. 2018; Teichert 2014). Our results also support the idea proposed by Jardim et al. (2022) for using the rhodolith interstitial space (or solidity as equivalent indicator) as proxy for habitat complexity. The high correlation identified between 2 and 3D metrics for morphotype A demonstrates that 2D metrics may yield useful indicators that are faster to obtain and potentially effective to be incorporated in the routine monitoring of rhodolith beds. 3D measurements resulted to be time-consuming and can be estimated to a good approximation through 2D metrics if 2D to 3D conversions have been estimated previously. Moreover, we recommend using the linear relations between measures as a method to rapidly obtain a total photosynthetic surface. This measure is important for calculating the photosynthesis rates for the branched growth forms, as well as to obtain the rhodolith volume and quantify the calcium carbonate production (Dulin et al. 2020). Similar conclusions were reached by House et al. (2018) comparing 2D and 3D metrics across different morphologies of scleractinian corals: although 3D metrics of surface area and volume are more precise, 2D metrics scaled consistently with 3D metrics in corals, providing convenient and fast measurements of ecological function.

The study of algal physiology commonly uses dry weight to standardize indicators such as biomass of epiphytes by unit mass of rhodoliths or to estimate respiration and production (photosynthesis). However, in multi-specific rhodolith beds, as in our case study, it is often difficult to individualize different morphologies (as shown in the results of the discriminant analysis). Individual weight may vary among morphologies, for the same longitudinal size, and even within individuals of the same morphology because of the weight of organisms forming the nucleus of some rhodoliths (bryozoans, polychaetes, some boring organisms) (see an example in Fig. 10) (Pérez-Peris et al. 2023). For these reasons, we propose using the area corresponding to the thallus surface in 3D as a more realistic measurement to estimate metabolic rates or to standardize cryptofaunal abundance. Rhodolith volume is an important measure related to carbon retention and is a metric that can only be obtained with sufficient precision with 3D measurements. Simple methods based on the equation of an ellipsoid (Teichert 2014) or water displacement (Amado-Filho et al. 2007) have low precision and limited general application.

Fig. 10
figure 10

(Left) Red calcareous algae growing and completely covering a bryozoan. (Right) broken branch of a rhodolith from morphology A, with a bryozoan found inside

Full size image

In conclusion, the intricate morphological diversity observed between and within rhodolith species, their role as habitat for benthic organisms, coupled with their potential significant role as carbon reservoirs, make them suitable candidates for biomonitoring the environmental condition of deep biogenic habitats. Key measures of morphological complexity, as well as assessing parameters such as photosynthetic surface and carbonate volume, could be valuable indicators of the rhodolith bed condition. This research contributes to our comprehensive understanding of how the structural complexity of rhodoliths enhances ecosystem functioning, emphasizing the ongoing importance of advancing our knowledge and developing additional morphological indicators for these ecologically valuable species.