Fish functional diversity is modulated by small-scale habitat complexity in a temperate ecosystem

Rocky reefs and kelp forests form conspicuous habitats that promote species diversity and support livelihoods in coastal communities. However, taxonomic approaches often disregard biological identity and differences between species. In this study, we explore the relationship between functional diversity of fish communities and habitat complexity in temperate reefs and test if greater habitat complexity, such as rugosity or kelp three-dimensional structure, would result in higher functional diversity. We conducted fish surveys using SCUBA in four kelp forest sites and rocky reef sites. Although the rocky reef and kelp forest sites showed differences in habitat complexity, no significant differences in fish taxonomic or functional diversity were found between the two habitat types. However, we did find differences at smaller spatial scales for the kelp forest sites, where number of stipes, stipe bundle diameter, and kelp density influenced functional richness, but not species richness, highlighting the importance of functional approaches in certain ecosystems. The differences found among kelp forest sites may be linked with small-scale spatiotemporal oceanographic drivers of productivity such as upwelling exposure or nutrient availability. We recommend considering small-scale spatial drivers when aiming to understand how habitat characteristics link with functional diversity.


Introduction
Habitat complexity is one of the ecological factors that explains the spatial variability in fish taxonomic diversity (Jones, 1991;Lyon et al., 2019). The fact that habitat structural complexity is a determining factor for local abundance and taxonomic diversity may be due to the influence on ecological processes, such as competition and predation (Hixon & Menge, 1991;Almany, 2004). Complex habitats can reduce competition via the availability of resources such as food (García-Charton & Pérez-Ruzafa 2001), as well as reduce encounter rates between predators and prey by providing refuge (Murdoch & Oaten, 1975). Understanding the driving factors of species diversity is a central goal in ecology and conservation (Ding et al., 2015), and although there are many factors that can influence species diversity, its positive relationship with habitat complexity appears universal (Kovalenko et al., 2012).
Habitat characteristics and environmental factors may influence biotic interactions and limit the capacity of certain species to persist in a community and modify the ecosystem (Hoeinghaus et al., 2007). Refuge would then be a key physical resource that could limit the abundance and taxonomic richness of fish (Ebeling & Hixon, 1991). However, there is a growing body of literature recognizing that not only is the number of species important, but also the actual species that are present. This is critical in determining the nature and strength of the relationships between species diversity and the range of their ecological functions (Stuart-Smith et al., 2013). In addition to changes in taxonomic structure, it is important to understand how changes in habitat affect ecosystem functions through changes in the functional diversity of communities (Díaz et al., 2007).
Considering the importance of functional composition in addition to taxonomic diversity has led to the concept of functional diversity (FD), which is increasingly used to address changes in diversity (Pla et al., 2012). Although there is a wide variety of definitions, FD can be defined as those components of diversity that influence the functioning of an ecosystem (Tilman, 2001). In contrast to the indices of taxonomic diversity, which are based on species richness and species relative abundances within the community, FD summarizes various aspects of biological composition and thus the role of populations in the community. Taxonomic diversity unveils the distribution patterns of species, meanwhile, FD reveals the assemblages based on the role of the species as well as their relationships with the habitat (Hoeinghaus et al., 2007). The FD approach takes natural selection-influenced traits into account, whereas the traditional taxonomic diversity approach does not (Hitt & Chambers, 2014), and FD is often more directly linked to the functioning of an ecosystem than biodiversity or taxonomic indices (Wong & Kay, 2019).
Fishes have been proposed as one of the most suitable groups to study the FD because (1) most biomass flows through them, (2) they are the most diverse vertebrate group, and (3) they perform a wide array of functions and are reasonably well-known taxonomically as well as functionally (Stuart-Smith et al., 2013;Villéger et al., 2017). Shelter provided by habitat complexity is a key physical resource that can limit the abundance and taxonomic richness of fish (Ebeling & Hixon, 1991). However, the links between habitat complexity and fish functional diversity might respond to biological and environmental factors differently than species richness (Hoeinghaus et al., 2007;Leduc et al., 2015). Understanding how habitat modulates ecosystem function through changes in functional diversity is important for identifying which habitats to conserve and protect (Díaz et al., 2007).
Rocky reefs and kelp forests from the temperate North-eastern Pacific coast are conspicuous ecosystems with economic importance as they support aquaculture, tourism, and fisheries (Graham, 2004;Foster & Schiel, 2010;Parnell et al., 2010). Both ecosystems have a complex yet different habitat structure, with rocky reefs having crevices providing refuge and kelp forests adding a three-dimensional structure as a habitat provider for shelter. Nevertheless, the functional diversity of the fish assemblages inhabiting these ecosystems and the link with habitat characteristics have only recently become a focus on a larger scale (Stuart-Smith et al., 2013), yet they have not been compared between these ecosystems. At a local scale, such as in marine reserves, the recovery of species richness coincides with increased functional richness in kelp forests (Micheli & Halpern, 2005), suggesting that small changes in species diversity can have profound impacts on functional diversity and showing that both functional and taxonomic diversity are important. Furthermore, sites with higher-relief reefs show greater functional and taxonomic diversity due to the increased availability of refuges for shelter (Tuya et al., 2019). However, very little is known about how the different habitat complexity of the two ecosystems of rocky reefs and kelp forests can influence functional diversity.
In this study, we evaluate whether taxonomic and functional diversity in temperate fish communities increase as a function of habitat complexity across two marine ecosystems. Specifically, we used taxonomic and functional diversity indices to see if the fish communities from kelp forests and rocky reefs are different, considering kelp forests as the more complex habitat due to the water column habitat complexity it provides. We proposed that greater habitat complexity would result in higher functional diversity. To test this hypothesis, we conducted surveys on sites where both rocky reefs with no kelp and kelp forests existed nearby at the same depth.

Study site
The study sites were located off the coast of Ensenada, Mexico, inside and outside of Bahía Todos Santos (BTS). This bay is on the Pacific coast of Baja California (about 100 km south of the US-Mexico border) and covers approximately 180 km 2 (Mateos et al., 2009). This area is influenced by the Eastern Boundary California Current, dominated by equatorward flow and upwelling favourable winds. Upwelling is evident in spring (showing a maximum in May), surface warming and increased variability in summer, and strong wave action, greater mixing, and less variable water column temperatures in winter (Ladah et al., 2012).
Sampling was conducted at eight sites around BTS; four kelp forest (KF) dominated sites and four rocky reef (RR) sites (Fig. 1). Three sampling events were carried out in each study site during May and June of 2014. These sites were chosen because of their accessibility and similar depth range (10-12 m), and each study site showed both ecosystems in proximity exposed to similar physical drivers. Herein, we define a kelp forest as an environment dominated by canopy-forming giant kelp Macrocystis pyrifera (Linnaeus) C. Agardh, 1820, and rocky reefs as a hard substrate-dominated environment without the presence of giant kelp.

Study design: fish surveys and habitat measurements
Underwater visual censuses (UVC) were performed using SCUBA diving along 30 × 2 × 2 m band transects. Three sampling campaigns were carried out on all sites, and three transects were surveyed each time. A total of nine transects were conducted in each study site, covering 4320 m 2 in the study area. Fish were identified to the species level and abundance was estimated. Fish surveys were focused on assessing the conspicuous fish community; however, all divers were able to include cryptic fish, thus achieving a better representation of the community. Abundance within 5-cm size classes was quantified for all species and we calculated species biomass using abundance data and the allometric length-weight conversion ( Biomass = a × Length b ) with parameters (a and b) obtained from FishBase (Froese & Pauly, 2020). In kelp forests, fish species and abundance were also evaluated in the middle water column associated with the kelp fronds. In all study sites, substrate type was assessed every five meters, ranking the substrate type as sand, rock-boulders, or a mix of both. A reef rugosity index (as a measure of habitat complexity) was obtained using the length of a metal chain, carefully placing it as closely as possible to the bottom contours, and the length of the measuring tape used to run the transect: For rocky reefs, cave depth was measured using a PVC stick, considering this is an important refuge for fish. In kelp forests, the number of M. pyrifera individuals along the transect was counted to estimate density, and stipe numbers of nine random individuals along the transect were counted and their combined stipe bundle diameter was measured with a vernier.

Functional structure of fish assemblages
Five functional traits providing information about species functionality were used: maximum length, diet, habitat preference, body morphology, and gregariousness (see Table 1 for more details). The maximum length was a quantitative trait, the diet was a nominal trait, and habitat preference, body morphology, and gregariousness were coded as ordered variables. Species traits were obtained from a comprehensive literature survey and field annotations (see Table S1 for the full trait matrix and citations). We used species richness as a proxy of the variety of taxa present in each fish community. To obtain functional diversity, we first built a multidimensional functional space using a Principal Coordinates Analysis (PCoA; Villéger et al., 2008). The first four dimensions were used as they have the minimum mean absolute deviation (mAD) between traitbased distances and distances in the functional space (mAD = 0.038, Fig. S1, Magneville et al., 2021). In this four-dimensional functional space, we found that the changes in species positions on PC1 are driven by all the traits; PC2 is driven by maximum length, diet, and habitat preference; PC3 is only driven by body morphology, and PC4 is driven by diet and habitat preference (Table S2 and Fig. S2). Using this multidimensional functional space and the biomass data, we calculated three complementary indices: functional richness (FRic), functional evenness (FEve) and functional divergence (FDiv). FRic indicates the amount of functional space occupied by the community (Villéger et al., 2008) and was calculated as the volume of the convex hull that includes all the species (Cornwell et al., 2006). FRic only accounts for species with extreme trait values and it does not account for species biomass (Villéger et al. 2008). FEve measures regularity in the distribution of species weights in the functional space (Villéger et al., 2008). High values of FEve indicate an even distribution (Schleuter et al.,  (Mouillot et al., 2013), and it is closer to 1 when species are near the border of the convex hull. The mFD package was used to compute functional diversity metrics and obtain multidimensional space (Magneville et al., 2021) using R 4.1.1.

Data analysis
We compared the fish assemblages between kelp forests and rocky reefs, and between the sites of each habitat type. The comparison between habitats was conducted with a Student's t-test using species richness and functional diversity indices as the response variables. To see if there were differences between sites of each habitat, we conducted a oneway ANOVA with a Post hoc Tukey HSD-test to elucidate patterns. In both analyses (Student t-test and ANOVA), QQplots were performed previously to graphically test for normality. To test variance homogeneity, a Bartlett test was conducted. All tests were performed using the R packages nortest (Gross &  Ligges, 2015), ggpubr (Kassambara, 2020), and car (Fox & Weisberg, 2019). If normality was not met, a Wilcoxon signed-rank test was used.
To determine how the different habitat complexity measurements influenced the structure of fish assemblages, we ran generalized linear mixed models (GLMM) separately for each habitat using the R package glmmTMB (Brooks et al., 2017). S, FRic, FEve, and FDiv were the predictor variables, site was used as a random effect, and the different habitat measurements were used as explanatory variables. We examined possible correlations between the habitat measurements, and those that were highly correlated (Zuur et al., 2009) were used in different models. Normality and homoscedasticity were checked using the R package dHARMa (Hartig, 2021) which supports diagnostic checks for models of GLMM. A Poisson distribution was used for the models including S as a predictor variable, and a Beta distribution was used when the predictor variables were FRic, FEve and FDiv.

Results
Taxonomic and functional structure of fish assemblages Forty species were recorded during our study in both habitats, with 35 species recorded in the kelp forest ecosystem and 29 for rocky reefs. We found 24 species (60%) of the total recorded species that were shared between kelp forests and rocky reefs. For kelp forests, 35 species of fish belonging to five orders and 15 families (Table S3) were recorded. The most abundant species was the señorita Oxyjulis californica (Günther, 1861) and the families with the highest species richness were Embiotocidae and Sebastidae, with eight and six species, respectively. At the genus level, Sebastes showed the highest representation with six species, followed by Embiotoca with three species, and Paralabrax with two species.
The fish species found in rocky reefs belong to four orders and 14 families (Table S3). The most abundant species was the blacksmith Chromis punctipinnis (Cooper, 1863) and the families with the highest species richness were Embiotocidae and Sebastidae, with four and five species, respectively. The genus Sebastes showed the highest representation with five species, followed by Embiotoca with three species, and Lythrypnus and Paralabrax with two species each.
We found higher species richness, functional richness, and functional evenness in kelp forests (Fig. 2); however, these differences were not significant (Wilcoxon test, species richness: P = 0.46; FRic: P = 0.74, FEve: P = 0.23). Even when functional richness did not show differences between habitats, we found that species like salema and black rockfish found exclusively in kelp forests were the species with extreme traits like being associated with kelp fronds (salema) and forming pairs (black rockfish). These species are occupying the vertices of convex hull (Fig. 3), driving the higher functional richness in this ecosystem (bigger green convex hull). On the other hand, moray and scorpionfish were driving the slightly bigger convex hull of rocky reefs in some of the combinations (i.e., bigger brown convex hull in PC1 vs. PC4).
FDiv was the only variable that differed between habitats (Wilcoxon test, P = 0.02), being higher in kelp forests (Fig. 2). Furthermore, we did not find differences for any of the indices (taxonomic or functional) when comparing sites within each habitat (ANOVA).

Habitat complexity
In kelp forests, Macrocystis pyrifera stipes and stipe bundle diameter were the only explanatory variables that were highly correlated (r = 0.94). Two separate models were generated: one removing the diameter of the stipes from the explanatory variables (model 1) and the other removing the number of stipes (model 2). None of the explanatory variables for the kelp forest habitat was significant for S, FEve or FDiv in either model. However, both models for FRic showed significant explanatory variables. Kelp density and number of stipes were the significant explanatory variables in model 1, while kelp density and stipe bundle diameter were the variables explaining differences in model 2 (Table S4).
Cave depth and rugosity were not correlated to fish communities in rocky reefs, and the substrate type was not added to the model as it was the same type across all transects and sites. None of the explanatory variables were significant for models with the different taxonomic and functional indices (Table S4).

Discussion
Our study did not reveal statistical differences in fish taxonomic and functional diversity between kelp forests and rocky reefs, even when these two habitats showed different habitat complexity. Nevertheless, we did find differences at smaller spatial scales within kelp forest sites, where kelp density and other morphological kelp characteristics influenced higher functional richness. Our findings contrast with other studies showing habitat complexity did enhance fish taxonomic richness in both kelp forests (Angel & Ojeda, 2001;Cole et al., 2012;Efird & Konar, 2014) and in rocky reefs (García-Charton & Pérez-Ruzafa 2001;Parsons et al 2016). Furthermore, rather than at the larger scale of ecosystem-based habitat modulation as proposed, our results suggest smaller scale drivers within the kelp forest ecosystem beyond a taxonomic approach. Recent work shows that environmental filtering of functional traits can lead to complex and unexpected functional diversity patterns (Wong & Kay, 2019) and the expected relationship between taxonomic diversity and habitat complexity  (Tuya et al., 2019;Wong & Kay, 2019;Cáceres et al., 2020).
Although there are no comparative studies of diversity between kelp forests ecosystems (KF) and rocky reefs (RR) in our study area, previous records of diversity in the biogeographic region have reported a lower overall species richness compared to our results (Hammann & Rosales Casian, 1990;Pondella et al., 2005). There are 32 fish species reported for the rocky-algal subtidal of Bahía Todos Santos (Hammann & Rosales Casian, 1990)  The two-coloured shapes represent the functional space of the species. The dots inside represent the different species. Even when there were no differences between both ecosystems, some species are only present in one of the ecosystems (see, for example, green dots that are not inside the brown polygon). Some of the species' names with extreme traits are shown. Salema: Brachygenys californiensis, black rockfish: Sebastes melanops, scorpionfish: Scorpaena guttata, moray: Gymnothorax mordax. Fish illustrations by Larry G. Allen the rocky bottoms of the Islas Coronado, in northern Baja California (Pondella et al., 2005). Overall, we found more species: 35 species for kelp forests and 29 species for rocky reefs.
Neither species richness, functional richness nor functional evenness showed significant differences between habitat types in the Student's t-test (rocky reefs vs. kelp forests), although we originally predicted that the habitat types would differ. It is important to consider that these species are mobile and can move between reefs, that their distribution can vary seasonally, and that our surveys may have not captured the entire diversity in both ecosystems due to diver presence (Dickens et al., 2011). In addition, the variability between sites was high (Fig. 2), which may be another reason masking the ability to detect differences between habitats if they did exist. However, FDiv did show differences between habitats, reinforcing the importance of considering several complementary functional indices. FDiv was higher in kelp forests (see Fig. 2) indicating the increase of unique and specialized species in this habitat (i.e., those with unique traits, Villéger et al., 2010). Some species such as the salema are associated with kelp forests, inhabiting the canopy, or hidden between fronds and are rarely found in rocky reefs. Furthermore, this species has extreme functional traits (see the position in the vertices of the convex hull, Fig. 3) like inhabiting kelp fronds (see Fig. S2, negative values in PC1), which may enhance functional richness in this habitat (i.e., bigger convex hull). These results highlight the importance of the habitat for these species, and how the loss of habitat (i.e., kelp loss because of climate change) may lead to the disappearance of species. California moray and California scorpionfish, only found in rocky reefs in our study, also have extreme traits like eel-shaped (moray) and roving invertivore (scorpionfish), yet they are also residents of kelp forests (Higgins & Mehta 2018;House & Allen, 2022). The camouflage of scorpionfish and the use of caves as extended habitat for moray eels may hinder their detection in the surveys in kelp forests (Gilbert et al., 2005), which may have led to the lack of differences in functional diversity between habitats.
Our surveys did not record top predator fishes with geographic distribution in the study region, such as the giant sea bass Stereolepis gigas Ayres, 1859, or shark species like the swell shark Cephaloscyllium ventriosum (Garman, 1880), the leopard shark Triakis semifasciata Girard, 1855, and the bullhead shark Heterodontus francisci (Girard, 1855) (Horn et al., 2006). These species are considered residents of the kelp forests and rocky reef of the San Diegan biogeographic province (Horn et al., 2006;Pondella et al., 2005) and have been recorded in Bahía Todos Santos in the past (Hammann & Rosales Casian, 1990;Ramírez-Valdez et al., 2021). The absence of higher trophic level species has been widely documented to result from fishing pressure (Pauly et al., 1998) and it generates important changes in the structure of communities (Sala et al., 2004), eliminating whole functional groups from marine ecosystems (Micheli & Halpern, 2005). The absence of these predators from our surveys with unique functional traits certainly sets current results separate from past findings. An important caveat, however, is that we surveyed each site three times, making it hard to find these species. In addition, there have been reports about some of these species on social media and from fishermen on some of our sites regarding these species, suggesting our survey may have missed these species. We then acknowledge the need to study these ecosystems at a broader spatial and temporal scale.
Contrary to our expectations, none of the functional indices were different between sites in rocky reefs, indicating that the species among this habitat type have similar traits (Villéger et al., 2008). However, site-specific differences did occur for the kelp forests of this study, but only for FRic. This site-specific difference found in the kelp forests might be due to habitat differences between sites and small-scale oceanographic conditions modulating local productivity in the kelp forests, and as stated before, due to the presence of exclusive species with extreme traits (e.g., salema). Other studies have found that fish distribution can also be driven by fine-scale oceanographic conditions like stratification (McInnes et al., 2017), and our results at a much smaller scale support this finding.
Among the kelp forests sites, Rincón de Ballenas was the site with the highest FRic (mean FRic = 0.55), with a maximum of 111 kelp individuals/60 m 2 (one transect), ~ 7.5 times higher than in La Bufadora. A decrease in FRic values is often generally attributed to habitat degradation processes such as eutrophication or urban development (Villéger et al., 2010). In the other kelp forest sites, fishing and proximity to the city centre might explain their reduced FRic. For example, lobsters and urchins are fished with traps in the rural Campo Kennedy site, and it is a popular site for diving and spearfishing as well. Even when fishing could not be accounted for as a process of habitat degradation, the process of deploying and collecting traps, and diving can be disruptive for the ecosystem. The other sites (La Bufadora and Las Rosas) are also close to urbanized or tourist-impacted areas, also suffering from some sort of habitat degradation.
Macrocystis pyrifera stipes number, stipe bundle diameter, and kelp density were the habitat variables that influenced functional richness in our study but did not impact species richness, whereas none of the habitat variables explained the differences in taxonomic or functional indices in rocky reefs. There are several habitat variables deemed important in regulating fish abundance and diversity in the literature, such as greater kelp density and bottom rugosity (Angel & Ojeda, 2001;Cole et al., 2012;Efird & Konar, 2014). Furthermore, M. pyrifera is by itself habitat-forming and is an ecosystem engineer, providing three-dimensional structure and habitat availability throughout the entire water column (Steneck et al., 2002). Rocky reefs, on the other hand, present a lower habitat complexity when compared with the three-dimensional structure provided by kelp forest fronds and surface canopies. However, it is important to highlight that the surveyed rocky reef sites in this study were very similar to each other in habitat structure, which may have led to the lack of spatial patterns in our results. Furthermore, fine-scale habitat complexity along with microhabitat selectivity has been shown to modulate fish functional groups in coral reefs (Eggertsen et al., 2020), and it is known that modulating factors can act on separate traits, thereby either concealing or amplifying expected patterns (Ford & Roberts, 2020), with overfishing further complicating expected functional diversity patterns (Cáceres et al., 2020). The fact that similar species richness within sites in kelp forests results in an increase in functional richness indicates that the species may be playing different functional roles in the ecosystem. On the contrary, the similarities in species richness and functional richness within sites in rocky reefs show that those species may be functionally similar.
In our study area, small-scale structure in oceanographic conditions is known to occur, with significant differences in zooplankton communities and nutrient fields over just a few kilometres (Ladah et al., 2012;Filonov et al., 2014), however spatial differences in environmental forcing at these smaller scales are rarely considered. In seagrass communities, for example, the structuring of functional diversity by habitat complexity can be confounded by site-level patterns such as wave exposure (Wong & Kay, 2019). Further studies considering the physicochemical environment of the nearby occurring habitat types may help reveal the modulating factors of the small-scale spatial differences in functional diversity patterns.
It is important to note that between 2013 and 2015, there was a mix of cold periods followed by considerable ocean warming along the Baja California coast (Cavole et al., 2016;Arafeh-Dalmau et al., 2019). We did not account for this in our study, but there is evidence that the increase in temperature was followed by declines in kelp density and a shift in the fish community (Arafeh-Dalmau et al., 2019). As our sampling occurred in 2014, these ecosystems may have been undergoing warming impacts at that time. Because kelp forests are highly sensitive to warming (Wernberg et al., 2015), which essentially reverts them to rocky reefs devoid of kelp, the understanding of how these three-dimensional biogenic habitats might structure fish assemblages may hold even greater importance as in a warmer ocean climate.