Disentangling the links between habitat complexity and biodiversity in a kelp‐dominated subantarctic community

Abstract Habitat complexity is one of the most important factors modulating species diversity. This feature comprises several interrelated attributes, such as number, size, and spatial arrangement of complexity‐forming elements. However, the separate and joint effects of these attributes on diversity and community structure are still not well understood. Here, we assess the relationships between several structural‐complexity attributes of the subantarctic kelp Lessonia flavicans and species richness, total abundance, and structure of kelp‐associated macrobenthic communities. We predicted that longer thalli and larger holdfasts favor greater species richness and total abundance of invertebrate organisms. To test the prediction, an observational sampling program was established in two sites of the Strait of Magellan. Uni‐ and multivariate analyses revealed both positive and negative effects of kelp structural‐complexity attributes on diversity. Holdfast diameter and maximum frond length, followed by thallus wet weight, had the strongest positive fits to species richness and total abundance; the number of stipes, on the other hand, was negatively associated with both response variables. Longer fronds were associated with greater abundances of spirorbid polychaetes. Larger holdfasts supported larger abundances of Nereididae and Terebelidae polychaetes and the limpet Nacella mytilina. Contrarily, kelps with longer fronds and more stipes supported fewer amphipods. In this way, we demonstrate that different dimensions of habitat complexity can have contrasting effects on diversity and community structure, highlighting the fundamental role of multiple dimensions of kelp habitat complexity for local biodiversity.


| INTRODUC TI ON
Species diversity, or biodiversity, is perhaps the most striking feature of our biosphere. However, the unprecedent anthropogenic climate crisis poses severe threats to biodiversity across scales and realms (Trisos et al., 2020). Anthropogenic impacts on biodiversity can be particularly severe when affecting species with special functional traits, such as foundation species. Kelps, mussels, and coral are examples of foundation species that support local biodiversity and define entire ecological communities by enhancing habitat's physical complexity (e.g., Ellison | 1215 VELASCO-CHARPENTIER . The loss of foundation species results therefore in habitat destruction and simplification, one of the main drivers of community alteration worldwide (Luypaert et al., 2020). However, our understanding of the relationship between species diversity and the structural complexity of biogenic habitats is still incipient (Miller et al., 2018).
Since the first publication of the effects of habitat complexity on species diversity (MacArthur & MacArthur, 1961), many studies have evinced a positive relationship between the structural complexity of habitats and species diversity (Dean & Connell, 1987;Loke & Todd, 2016;Luckhurst & Luckhurst, 1978). The mechanisms underpinning this relation include, but are not limited to, an increasing number of niches due to increased microhabitat availability, higher food web productivity and stability, and enhanced protection from physical disturbances (reviewed by Kovalenko et al., 2012). However, the use of multiple definitions of complexity has limited our understanding of its role in influencing species diversity (Kovalenko et al., 2012). For example, habitat complexity can be described as the variation in the size of the living spaces and structural components (Loke & Todd, 2016), or just as the number of structural components of the habitat (known as "heterogeneity"; McCoy & Bell, 1991).
Habitat complexity is a multifaceted concept, encompassing five interrelated dimensions (Tokeshi & Arakaki, 2012): scale of observation of the structural elements (e.g., global, regional, local, or microhabitat); diversity of elements (i.e., surface geometry or topography); spatial arrangement of elements (the elements may be scattered randomly, clustered, in patches of different elements, or in zonation); size of the structural elements; and their density (i.e., number of elements in a determined area). These criteria provide a more accurate view of how habitat complexity is related to species diversity and community structure. Within a given habitat, for example, a greater size range of structural elements can support a greater species diversity and abundance due to enhanced niche availability (e.g., Hacker & Steneck, 1990;Loke & Todd, 2016;Tsuchiya & Nishihira, 1986).
As foundation species, mature thalli comprise multiple habitatcomplexity attributes, such as number of stipes, total height, holdfast diameter, foliage cover, and kelp density (Dayton, 1985; Figure 1). Kelp forests enhance local species diversity through increasing primary productivity, nutrient supply, shelter, and secondary settlement space (Arkema et al., 2009;Duggins & Eckman, 1997;Steneck et al., 2002). Despite their important role as complex habitats, kelp's complexity attributes are usually analyzed separately (Villouta & Santelices, 1984;Vega et al., 2014), and most studies focus on complexity without a clear distinction of attributes (but see Trujillo et al., 2019). So, little is known about which complexity attribute(s) are playing the most relevant role in determining how the kelp-associated community is structured.
The kelp Lessonia spp. have major ecological roles structuring benthic marine communities on Chilean temperate coasts (Vásquez & Santelices, 1984;Villouta & Santelices, 1984). The available information about this genus in the Chilean Pacific coast is mainly based on studies of Lessonia berteroana and L. spicata in northern and northern-central Chile (Vásquez & Santelices, 1984;Vásquez et al., 2012;Vega et al., 2014;Vega, 2016). Lessonia species are currently under strong harvesting pressure for the extraction of alginic acid in these latitudes (Steneck et al., 2002). Illegal fishermen use pry bars to detach the entire kelps from the substratum, an activity that threatens to move southward soon (Rosenfeld et al., 2019). Yet, the functional roles of this group are largely unknown in southern cold-temperate and subpolar regions, where Lessonia flavicans (Bory 1825) is one of the most conspicuous macroalga Marambio et al., 2016).  In this study, we assess the relationships between several structural-complexity attributes of L. flavicans and species richness, total abundance, and structure of kelp-associated macrobenthic invertebrate communities. We hypothesized that longer fronds and larger holdfasts will favor greater species richness and total abundance due to enhanced secondary settlement space. These effects should be reflected in strong associations between frond size, holdfast diameter, and community structure (i.e., the combination of species identities and abundances).

| Study sites and sampling design
The Magellan region is located in the South East Pacific, encompassing a diversity of habitats that include fjords, inland seas, glaciers, gulfs, and channels-a complex landscape resulting from the combined effect of tectonic processes and glaciation. The oceanographic features and diverse environmental conditions determine a particular marine biogeographic unit (Camus, 2001) dominated by benthic invertebrates and extensive kelp forests (Friedlander et al., 2018).
Both sites harbor L. flavicans forests of around 400 m 2 , located between 0.5 and 4 m depth (C. Velasco-Charpentier pers. obs.). Both sites are moderately wave-exposed, with a hard substratum consisting of boulders and large rocks covered by crustose coralline algae.
The patchy rocky reefs extend from the intertidal to ~8 m depth. and inclination (~0°). In each transect, ten 0.25 m 2 quadrants separated by ~3 m were photographed with a Nikon D7200 camera equipped with an Ikelite housing and a rectilinear Tokina 11-16 lens.
The unit of replication was each kelp (N = 10 per site and season) since we selected the quadrants with one L. flavicans specimen.
Each thallus was wrapped with a mesh bag (1 mm 2 pore size). Then, the holdfast was detached from the rock with a knife and the bag was swiftly closed to minimize the escape of mobile macrobenthic organisms. Within one hour after collection, samples were transported to the Laboratorio de Ecofisiología y Biotecnología de Algas, Universidad de Magallanes to measure the thallus morphology attributes and species sorting.

| Estimation of kelp structural complexity and associated biodiversity
Once in the laboratory, we measured five kelp structural-complexity attributes. The number of stipes, maximum length (i.e., thallus length, from holdfast base to the apex in cm), holdfast diameter (i.e., maximum width of the holdfast base in cm), and total wet weight in g (0.01 g precision) were directly measured from the collected kelps.
Foliage cover was estimated from each digital photograph taken in situ from above, at a zenith angle. This variable was categorized as 1 (fronds cover between <10 and 30% of the quadrant), 2 (fronds cover between 31% and 50%), 3 (fronds cover between 51% and 80%), and 4 (frond cover between 81% and 100%).
For each L. flavicans thallus, macrobenthic (>1 mm) mobile and semi-sessile organisms (e.g., mussels) in addition to tube-dwelling sessile organisms (e.g., spirorbid polychaetes) were collected from thallus surface with aid of forceps and scalpels. In addition, the stipes and disks were dissected to collect organisms living in crevices and galleries. We used specialized literature and field guides to identify each individual to the lowest taxonomic level possible, usually species (González-Wevar et al., 2018;Häussermann & Försterra, 2009;McLean, 1984;Osorio et al., 1979;Pastorino, 2005). Species-specific abundance was estimated as counts of individuals. These data were used to estimate species richness, defined as the number of taxonomic identities; total abundance, defined as the total number of individuals per taxonomic identities; and community structure, defined as the combination of species identities and abundances.
Before the analyses, Pearson-product moment correlations were calculated among the explanatory variables in order to account for collinearity.

| Univariate analyses
Generalized linear models (GLM) were used to analyze species richness and total abundance separately. Since the explanatory variables (number of stipes, maximum length, holdfast diameter, total wet weight, and foliage cover) were highly correlated ( Figure S1), we first combined these variables in a principal component analysis (PCA). In the GLMs, the explanatory variables were the PC scores of the first two axes (64% of explained variance), in addition to season and site.
Since one site was sampled only in two seasons, we did not include the site by season interactive effect in the models. Thus, the sites were compared by averaging the seasons, while the seasons were compared by averaging both sites. For each categorical variable, we used a "treatment" contrast in which the mean value of each group is compared against a reference group: Bahía Buzos was the reference site because was the site with more sampling events, while spring was the reference for season because it is usually the season with larger species abundances. The use of this contrast type allowed us to keep the number of comparisons below the number of groups in the case of season and thus to prevent the inflation of Type-I error type.
Due to the difference in measurement scales, the continuous response variables were centered and standardized before the analyses (Becker et al., 1988). Model parameters were estimated through maximum likelihood. For both models, that is, species richness and total abundance, we assumed a Poisson distribution of errors and we estimated a marginal coefficient of determination (R 2 ) according to Nakagawa and Schielzeth (2013).

| Multivariate analyses
Multivariate community structure was analyzed with canonical analysis on principal coordinates (CAP, Anderson & Willis, 2003). CAP is a constrained multivariate method in which an a priori prediction (e.g., maximum length and holdfast diameter strongly relate with community structure) is used to produce an ordination plot. The CAP axes are linear combinations of the response variables that maximize the between-to within-group variation, allowing the detection of patterns that could be masked by overall dispersion in unconstrained methods (Anderson & Willis, 2003). The model was based on species abundances, which were used to estimate a matrix of Bray-Curtis dissimilarities. In addition to kelp's structural-complexity attributes, season and site were included as fixed effects in the model. The statistical significance of the CAP model fit was assessed by means of an analysis of variance.
All analyses were conducted in the R statistical environment

| RE SULTS
A total of 41 taxonomic identities were identified associated with L. flavicans kelps across both sites (Table 1) Figure S2). These results demonstrate that both, species richness and total abundance, were higher in the kelps with a greater holdfast diameter, maximum length, wet weight, but with fewer stipes.
The site Bahía Buzos, on average, presented a greater species richness and total abundance than Carrera (Figure 4a,b, respectively; Figure S2). Regarding seasonality, species richness and total abundance decreased from spring to winter, which was represented by increasingly negative effect coefficients (Figure 4a Xymenopsis (cf) muriciformis 0 3

| D ISCUSS I ON
In this study, we showed that different kelp morphological at-

| Positive associations between species richness and abundances with kelp complexity attributes
Holdfast diameter was strongly and positively related to species diversity and abundance in our study. While this result may conform to a specific form of the general species-area relationship (Preston, 1962), part of the pattern may be related to the availability of interstices (Dibble et al., 1996)  Note: The first and second PC axis explained 64% and 19% of the variance, respectively. PC loadings with absolute values > 1/√5 are in bold.
would provide greater niche opportunities in terms of habitable space and food for associate fauna.
The holdfast structure offers protection from predators and adverse environmental conditions, accumulates food sources, and increases the area and volume of habitable space for colonization (e.g., Christie et al., 2003;Tuya et al., 2011;Vásquez & Santelices, 1984). For example, each holdfast of the kelps Saccorhiza polyschides and Laminaria hyperborean can harbor more than 400 individuals of amphipods, molluscs, and polychaetes (Tuya et al., 2011). The latter two groups also dominate the holdfast communities of the complex Lessonia berteroana-spicata in Chile (Vega, 2016). Another study in L. berteroana-spicata found 43 species on the kelp holdfasts, over 70% of the individual were juveniles (Cancino & Santelices, 1984). This is consistent with our

| Negative diversity-kelp complexity relationships
Surprisingly, the number of stipes had a negative effect on species richness and total abundance, and kelp frond length negatively affected the abundance of the numerically dominant peracarids Cassidinopsis emarginata and Peramphithoe femorata. These negative effects could be related to the indirect effects of enhanced shading and altered water flow regimes on juveniles. In addition, thallus scour and whiplash-like movements of longer fronds could well reduce the abundance of amphipods using these structures for attachment (Kiirikki, 1996;Teagle et al., 2017).
An alternative and nonexclusive explanation to these negative  (Cerda et al., 2010). However, and to our best knowledge, there is no detailed description of amphipod nest-building behavior in relation to L. flavicans' growth rates. Despite this limitation, our study suggests that the role of canopy-forming species in structuring the local communities could be the result of a balance between positive and negative effects on individual species (see Valdivia et al., 2012 for an example from intertidal communities).

| Spatiotemporal variation in kelp-associated communities
The Magellan region is characterized by strong seasonal changes in abiotic factors, which may significantly influence species richness and biomass (Ojeda et al., 2014). In this study, species richness and total abundance tended to increase in the periods of high productivity (spring and summer), which agrees with previous studies of kelp communities in the Magellan region (e.g., Asensi & Küpper, 2012).
Species richness and abundance gradually varied over the year: The highest species richness and abundance were observed in spring-summer, and the lower numbers were observed in winter. In this region, the strong seasonal changes in light availability  can be reflected into temporal variations in kelp growth rates (Blain & Shears, 2020;Gendron, 1989). Moreover, the numerically dominant amphipods, in addition to A. atra and the keyhole limpet Fissurella picta, increased in abundance during the spring season.
This reflects the importance of seasonality in abiotic environmental conditions for community structure in this region. The detected patterns suggest that the current climate change-related climatic anomalies in Patagonia (e.g., Garreaud, 2018;Aguayo et al., 2019) could have severe effects on the diversity, structure, and functioning of these communities. conserve these natural ecosystems that are facing multiple anthropogenic threats.

ACK N OWLED G M ENTS
The authors thank the staff of the Parque del Estrecho de Magallanes for allowing access to "Bahía Buzos" and for providing logistical support and Vicente Villalobos for providing

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.