Revealing the drivers of taxonomic and functional diversity of nearshore fish assemblages: Implications for conservation priorities

Understanding patterns and drivers of biodiversity are essential towards developing effective conservation strategies for shallow marine habitats broadly. However, little is known about the natural and anthropogenic factors that structure fish biodiversity of sandy beaches, one of the largest and most socio‐economically valuable nearshore habitats due to their endemic fauna, economic importance and cultural relevance. Here, we investigated how environmental variables and urbanization affect taxonomic and functional diversity of nearshore fish assemblages to provide general biodiversity patterns that can serve as baseline information for management plans when considering conservation prioritization.


| INTRODUC TI ON
Understanding spatial patterns in biodiversity and the processes that shape them is a critical step towards the design and implementation of effective conservation strategies for marine ecosystems, particularly in coastal zones where people and the sea meet. Sandy beaches occupy one third of the world's ice-free coastline (Luijendijk et al., 2018) and provide important ecosystem services such as tourism and recreational economies (Defeo et al., 2009) and function as important nurseries (Lombardi et al., 2014), foraging sites (Tatematsu et al., 2014) and spawning grounds (Hirose & Kawaguchi, 1998;Lasiak, 1986) for fishes. Many of the fish species found near sandy beaches occupy surf zones only during their juvenile stages before migrating to adjacent ecosystems. Therefore, surf zones are fundamental habitats that contribute to the successful development and recruitment of adult populations of commercially important species and ultimately support 40-60% of coastal fisheries (Gillanders et al., 2003). Potential changes in fish diversity of sandy beach surf zones may consequently have serious ecological and economic and social implications. Nearshore and artisanal fisheries are still very important for many communities, either for subsistence or for the local economy (Begossi et al., 2011), and ensuring the sustainability of this activity is of great socio-economic relevance.
Sandy beaches face increasing pressures from urbanization (e.g. construction on the sand, urban runoff) and climate-related environmental changes (e.g. erosion, storminess) on both the terrestrial and marine sides (Schlacher et al., 2007). These pressures are expected to strongly modify the beach environment and resident biodiversity, likely compromising the key ecosystem services noted above. To safeguard beach biodiversity and services, it is necessary to analyse the main drivers of local biodiversity. Understanding how changes in the environment act on the local species can provide essential information to construct tools that support conservation and can help mitigate or even reverse these impacts. Despite their high intrinsic value, research in conservation and management of sandy beaches remains one of the most limited and further rarely considers maintaining adequate ecosystem functioning (Nel et al., 2014). This knowledge gap is even more critical when we consider the current state of knowledge of surf zone fishes, which has rarely integrated observations across diverse morphodynamic and anthropogenic conditions (Olds et al., 2018).
The composition of surf zone fish assemblages can be influenced by several environmental factors such as beach exposure (Borland et al., 2017;Nakane et al., 2013) and freshwater input (Araújo et al., 2002, Pessanha & Araújo, 2003. Surf zone assemblages may also be exposed to multiple stressors linked to human activities, such as urbanization, tourism and fishing (Costa et al., 2017a). However, the identification of these combined effects on surf zone fishes have gone largely unexplored with focus instead on fish communities in other adjacent aquatic environments such as rivers (Tejerina-Garro et al., 2005), estuaries (Yabsley et al., 2020) and reefs (Taira et al., 2018).
More recently, studies have begun to explore the functional consequences of human impacts on natural communities through the use of organismal traits . By estimating the diversity of organismal traits, or functional diversity (FD), it is possible to understand how species respond to and modify their environments given their characteristics (traits). This approach therefore provides a more mechanistic understanding of how biodiversity is structured in response to changes in biological and physical-chemical drivers and how changes in species influence the functioning of ecosystems (Gagic et al., 2015;Petchey & Gaston, 2006). Decreases in FD reduce the potential of communities to respond to changing environmental conditions in marine (Mouillot et al., 2011) and terrestrial ecosystems (Cadotte et al., 2011;Flynn et al., 2011). Sandy beaches are one of the most vulnerable coastal systems, as prior evidence suggests that few species fulfil key functional roles and may also be the least able to withstand eventual disturbances . Most decision-making processes, however, have been based on species, habitat or socioeconomic metrics (D'Agata et al., 2014;Mouillot et al., 2011), and FD is often left out of management and conservation assessments (Laikre et al., 2016;von der Heyden, 2017).
To address these various knowledge gaps, we investigated how a combination of environmental variables and urbanization affect fish taxonomic and functional diversity of nearshore fish assemblages by sampling 77 sandy beach sites across the southeastern Brazilian coast with different natural features and degrees of urbanization.
This approach delivers a more comprehensive picture of the fish diversity at these sites across a gradient of impacts, which together can provide general biodiversity patterns that can serve as baseline information for management plans when considering conservation prioritization.
Editor: Jorge García (AQE) Molinos affecting species that possess specific functional traits differently. These drivers should therefore be considered simultaneously in appointing protected areas in order to preserve a diversity of organisms and functional traits integral to productive beach ecosystems.

K E Y W O R D S
anthropogenic effects, ichthyofauna, integrated coastal management, sandy beach, surf-BRUVS, urbanization 2 | ME THODS

| Study area
This study was performed in 77 sites along 27 sandy beaches on the North Coast of the state of São Paulo, Southeast Brazil. Our sampling sites (Figure 1) stretched over 150 km of coastline and included most beach types, from reflective to dissipative beaches (McLachlan et al., 2018) and from nearly pristine to highly urbanized beaches.

| Sampling and processing
Each beach was divided into three sites parallel to the beach edge: two close to the endpoints of the beach and one at the middle of the beach. These sites were established to account for the considerable spatial variation in abiotic features along a single beach, such as degree of exposure and beach slope. Surf fish assemblages may display high spatial (e.g. depth) and temporal (e.g. tide, season) variability (Olds et al., 2018); therefore, we conducted surveys only during high-tide, around midday of spring tides during 3 mo of the austral autumn of 2018 (i.e. when the impact of coastal tourism is lower and in order to avoid drastic changes in the fish assemblage caused by seasonal changes and the entry of cold fronts, which are more common in winter (Pianca et al., 2010)) and at similar depths (1-1.5 m) at each site.
Fish biodiversity was sampled using a combination of two methods: surf-BRUVS and beach seine nets, since synergistic use allows more accurate estimates of total fish diversity (Shah Esmaeili et al., 2021). Surf-BRUVS consist of a GoPro 6+ © Camera attached to a 10 kg weight and equipped with a bag containing 500 g of selected bait (Sardinella brasiliensis) approximately 0.5 m in front of the camera (Vargas-Fonseca et al., 2016). At each site, two surf-BRUVS were placed 100 m apart. Each BRUV recorded fish for one hour and fishes were identified and counted using the standard Max N statistic (Murphy & Jenkins, 2010).
Seine sampling was done 60 minutes after the video recording using a beach seine net of 20 m wide, 2.6 m high with 5 mm mesh and a 2 m high, 4 m long funnel. Two hauls were performed per site for approximately 20 m parallel to the shore (approximately 400 m 2 for each haul). Fishes were sorted, counted and conserved for further identification to species level in the laboratory. For each site, richness was based on the number of species combining both methods, while for abundance only the highest number of individuals registered between the two methods was considered, thereby eliminating double counts.
For environmental variables, we measured chlorophyll-a (μg/L), significant breaker wave height (cm), beach slope (tan β), distance to and size of the nearest river mouth (m) for each site. Sediment mean grain size was also evaluated based on five sediment samples that were collected along the beach profile using a 5-cm wide cylindrical corer, up to 20-cm deep. Samples were sieved into twelve granulometric fractions, and the mean calculated using the formulae provided by Folk and Ward (1957). However, this variable was not included in further analyses due to the strong correlation with beach slope. Samples for chlorophyll-a were collected using a 1-L dark bottle of both bottom and superficial water and filtered through GF/F filters (25 mm diameter, 0.7 μm pore size). Chl-a was determined spectrophotometrically in acetone extracts before and after acidification (Strickland & Parsons, 1972) and total phytoplankton biomass (μg/L) estimated according to Lorenzen's (1967) equation. Beach slope was determined by across-shore profiling of the supralittoral to the surf zone at each site (Emery, 1961). River influence was estimated for each site by calculating the over-water distance to and width of the closest river mouth using satellite images (CNES/Airbus, Maxar Technologies) obtained from Google Earth software. In beaches without a river, the distance between sampling sites and the nearest river was replaced by an arbitrarily large value equal to four times the longest distance F I G U R E 1 Study area on the North coast of São Paulo, 77 sites (symbols) along 27 sandy beaches in the municipality of São Sebastião, Caraguatatuba and Ubatuba connecting a site and river within the same beach, a modification of the approach proposed by Borcard and Legendre (2002) for metacommunity studies. The width of the closest river mouth was considered null in beaches without a river. Significant breaker wave height was measured as the average breaking height of 1/3 of the highest waves registered of 20 consecutive waves (McLachlan et al., 2018), evaluated after the end of fish sampling.
Urbanization was estimated by attributing values from 0 (absence) to 5 (highest level) to each of the following variables: (1) Proximity to urban centres, (2) buildings on the sand (i.e., those located within the beach perimeter), (3) beach cleaning (i.e., periodicity of beach cleaning by public agencies), (4) solid waste on the sand, (5) vehicle traffic on the sand and (6) frequency of visitors. An 'urbanization index' (González et al., 2014), was then calculated as the mean across these five categories, ranging from ''0'' (least urbanized) to ''5'' (highest urbanization). Scoring for these indicators was defined either in situ (averaging the scores from three observers, to avoid individual bias) and/or with information provided by local authorities (e.g., beach cleaning, frequency of visitors).
Then, we calculated five metrics of functional diversity: number of singular species, functional richness (FRic), functional dispersion (FDis), functional evenness (Feve) and community-weighted mean trait value (CWM). The number of singular species is related to the number of functionally unique species in each community. If all species are functionally different (i.e., occupy completely orthogonal functional states), the number of functionally singular species will be identical to the total number of species (Laliberté et al., 2015).
FRic, Fdis and Feve are complementary and describe the distribution of species and their abundances within the functional space (Mouchet et al., 2010). FRic estimates the total volume of multivariate functional space occupied by a given species assemblage and can be used as a proxy of the range of functional traits represented in an assemblage (Mouchet et al., 2010). Feve represents the regularity of the distribution of species abundances in the functional space, with low Feve indicating that some niches have a reduced number of individuals performing those functions (Mason et al., 2005;Mouchet et al., 2010). Fdis refers to the mean distance of all species in a given assemblage from the centre of the functional space weighted by their abundances, with low functional dispersion suggesting a convergence of individuals and/or functional traits in multivariate trait space. CWMs summarize the average value of each individual trait within a community (Garnier et al., 2004) and were used here to identify responses of individual traits that contribute to the multivariate indices. These analyses were conducted in the R environment (R Development Core Team, 2013) using the function 'fd' of the FD package (Laliberté et al., 2015).
We used the 'gam' function of the R package mgcv (Wood, 2012) to run generalized additive models (GAMs) to investigate the relationship between environmental variables and urbanization, and taxonomic (i.e., species richness and abundance) and functional diversity of fish assemblages. Data were log10-transformed and models were checked for homogeneity of variance and normality of errors. As no substantial deviations were found, we fit all models to a Gaussian distribution. Variables were checked for multicollinearity using the variance inflation factor (VIF) using VIF >2 as a cut-off value (Zuur et al., 2010), and for concurvity using the largest (worst) value >0.7 as a cut-off. We accounted for the presence of spatial structure including latitude and longitude of sampling sites as a smoothed, interaction term in the model (Wood, 2006).
Models best predicting taxonomic and functional diversity of fishes were selected using an information theory approach (Burnham & Anderson, 2002). First, models were run comprising all combinations of predictors without interactions. Then, we ranked multiple models using sample-size corrected Akaike information criteria (AICc) and excluded models with ΔAICc >4. The relative importance of each variable was then estimated as the sum of the Akaike weights (AICcw) over all of the models in which the variable appears from the model averaging (Burnham & Anderson, 2002). We considered only predictors with a probability greater than 50% (i.e. sum of the AICw >0.5) as important. Relationships between selected response variables and predictors were analysed through the graphical output of GAMs. All analyses were conducted in the R environment (R Development Core Team, 2013) using the mgcv (Wood, 2012), MuMIn (Barton & Barton, 2015), vegan (Oksanen et al., 2013) and (Laliberté et al., 2015) packages.

| Environmental characteristics
The environmental characterization performed at each sampling site reinforced the wide range of natural features and degrees of urbanization investigated in this study. Significant breaker wave height ranged from 3.86 to 160 cm (mean ±SD: 56.04 ± 35.42), chlorophyll-a concentration ranged from 0.92 to 81.1 μg/L (mean ± SD: 5.79 ± 35.81), the slope ranged from 0.008 to 0.12 (mean ±SD: 0.051 ± 0.028) and average grain size from 1.252 to 3.46 ɸ (mean ± SD: 2.37 ± 0.552). Sampled sites also differed in regard to river influence, with the width of the nearest river ranging from 0 to 280 m (mean ± SD: 45.55 ± 55.90) and the distance to the nearest river between 18 and 39,480 m (mean ± SD: 12377.7 ± 18026.51).
A complete characterization of the sampling sites can be found in

| Richness and Abundance
A total of 8766 fishes was sampled across all beaches, belonging to 64 species and 26 families (Table S1)

| Functional diversity
As with species richness and abundance, the functional diversity of fish assemblages was also strongly influenced by breaker wave height, chlorophyll-a levels and urbanization (Table 1; Figure 5).
Additionally, the distance to and size of the mouth of the closest river also exerted a significant effect on the selection of fish functional traits.
Association (solitary/schoolforming) was also not related to any of the environmental variables or degree of urbanization.

| DISCUSS ION
Despite surf zones having ecological importance for many fishes and supporting significant fisheries , few studies to date have addressed how environmental and anthropogenic variables shape surf zone taxonomic and functional fish biodiversity patterns on a regional scale (Andrade-Tubino et al., 2020).
We show that surf zone fish communities are mainly structured by wave breaker height and proximity to human impacts and that these variables tended to reduce taxonomic and functional diversity to a few generalist predatory fish species, a consequence of their ability to explore multiple resources, which allows them to effectively forage in less diverse or impoverished conditions caused by anthropic impacts. In addition to wave height and urbanization, we also found that chlorophyll-a levels and proximity to rivers may significantly influence the taxonomic and functional patterns of fish species in sandy beach surf zones.
Abundance and number of species of surf fish has been hypothesized to be higher in areas with intermediate wave action since it may enhance schooling (Valesini et al., 2004) and liberate benthic prey species from the sediment (Pessanha and Araújo,

F I G U R E 4 Smoothers curves (S)
showing the relationship (solid line) between (a) distance to river, (b) waves and benthic species, (c) chlorophyll-a, (d) urbanization and demersal species, (e) chlorophyll-a, (f) waves, (g) distance to river, (h) urbanization, (i) beach slope and pelagic species, (j) waves and detrivores, (k) waves, (l) urbanization and phytoplanktivores, (m) waves, (n) distance to river and zooplanktivores, (o) waves, (p) chlorophyll-a and zoobenthivores, (q) waves, (r) river size and piscivores, (s) distance to river and small fishes (L med 5-20 cm). Shaded areas indicate standard errors of the smooth curve. The 'rug plots' on the x-axis indicate the range of variables over which measurements were taken

(s)
2003). Our results, however, showed that the number of species and individuals of nearshore fish assemblages was linearly and negatively related to wave height. As a consequence, fish richness and abundance are likely higher in sheltered conditions, possibly due to a calmer surf climate which may favour the occurrence of several species, and in dissipative beaches, areas with wider surf zones where wave energy is dissipated through. Dissipative beaches are also known to have larger number of microhabitats such as runnels and bars along the benthic environment (Mosman et al., 2020), characteristic that tends to increase fish richness and abundance (Layman, 2000;Marin Jarrin & Miller, 2016). Higher fish richness and diversity in sheltered or dissipative beaches have been previously reported by Clark (1997) and Andrade-Tubino et al. (2020), reinforcing that habitats with lower wave energy on the beach face may be the most beneficial for nearshore fish diversity (Beyst et al., 2001).
Importantly, we found urbanization to be a fundamental driver of nearshore fish assemblages with an overall negative effect on their number of species and individuals. Similarly, Araújo et al. (2017) registered a decrease in species richness and abundance of fishes in a sandy beach after coastal industrialization while Reyes-Martínez et al. (2015) and Costa et al. (2017a) found low fish richness with high levels of visitation and beach maintenance, suggesting that impacted beaches are avoided by surf zone fishes. Human disturbances may also negatively affect fish assemblages by reducing water quality  and food availability (Costa et al., 2017b;Schlacher & Thompson, 2012) and simplifying food webs in comparison to non-urbanized beaches (Reyes-Martínez et al., 2015).
Nevertheless, we found the influence of urbanization on fish richness and abundance to have a non-linear relationship, with stabilization at moderate levels of urbanization, suggesting that reduction of just a few of these direct or indirect impacts is unlikely to generate TA B L E 1 Overall Akaike weights (AICcw) for each variable (Waves=significant breaker wave height, Chl-a= Chlorophyll-a, UI=Urbanization Index, Distance river=distance to closest river mouth, River size=size of the closest river mouth, Slope=beach inclination, Lat, Long=Latitude and Longitude) in the GAM models for richness and abundance, singular species (Sing.sp), functional richness (FRic), functional dispersion (FDis), functional evenness (FEve) and community-weighted mean trait value (CWM) for the 4 traits with 14 modalities. Values in bold represent significant overall Akaike weights substantial change. Besides wave height and urbanization, chlorophyll-a levels in sandy beach surf zones also significantly influenced the species richness of fish assemblages. Reflecting the role of phytoplankton production as an important food resource for many surf-zone fishes, number of species was higher in sites with higher chlorophyll-a levels, likely due to increased food resources and more links in the local trophic web (Bergamino et al., 2011).
As with taxonomic diversity, wave height, urbanization and chlorophyll-a levels were also the main drivers of functional diversity of nearshore fish assemblages. The higher number of species in sites with lower waves and higher chlorophyll-a levels resulted in higher values of FRic and singular species, showing that fish assemblages in these sites have a wider range of functional traits.
The urbanization effect, however, was not so clear. We found higher values of Fdis at sites with low (between 1 and 2) and high (>3) UI, suggesting that, despite having lower species richness and abundance of individuals, fish assemblages in urbanized sites may maintain high functional diversity (Sousa Gomes-Gonçalves et al., 2020;Teichert et al., 2018). Also, beaches with a steep slope have short, deep surf zones that consist of less microhabitats (like runnels and bars) (Wright & Short, 1984), which can explain the decrease of functional dispersion, since less ecological niches are available to be filled by distinct functional groups. Fdis was also lower at sites close to river mouths, likely reflecting the lower and more variable salinity which could prevent the occurrence of marine/stenohaline species. In fact, variations in salinity resulting from local freshwater discharges have shown to strongly modify the structure of sandy beach macrofaunal assemblages (Lercari & Defeo, 2006) and our results suggest that this effect may be extended to the nektonic diversity.
The analyses of the single-trait CWM enabled us to further dissect how environmental variables and human actions affect the functional diversity of nearshore assemblages. Higher waves reduced the abundance of pelagic, detritivorous, phyto-and zooplanktivorous species, likely due to the lower primary productivity and deposition of organic matter linked to higher hydrodynamics in the shallow areas of the surf zone (Brown & McLachlan, 2010;McLachlan et al., 1981;Odebrecht et al., 2014). Conversely, the abundance of piscivores (e.g. Strongylura timucu (Walbaum, 1792) and Carcharhinus limbatus (Müller & Henle, 1839)), increased with wave height, highlighting the ability of predatory species to withstand or migrate away from these harsher environmental conditions as necessary (Stewart & Jones, 2001). Interestingly, the abundance of benthic and zoobenthivorous species was higher in areas with intermediate wave height, suggesting that the hypothesis of higher fish abundances in areas with moderate wave action may be true for these species traits.
We also observed that pelagic fishes were more abundant in sites with lower Chl-a levels, while demersal and zoobenthivorous fishes showed the opposite trend. Although studies show that natural Chl-a levels for sandy beaches are commonly found to be low (0.05-9.16 μg L −1 , Menéndez et al., 2016), some isolated values registered in our study are quite high: up to 80 μg L −1 .
Many sites in the study area (e.g. Itagua & Itamambuca beaches) experience sewage discharges from neighbouring areas (CETESB -Environmental Agency of the State of São Paulo), either from local outfalls or from rivers contaminated by anthropic discharges along its path, resulting in an ongoing nutrient enrichment process, which has been shown to profoundly affect phytoplankton species composition and production (Aktan et al., 2005) and fishes (Guidetti et al., 2002). Since the majority of the pelagic fish in our study were planktivorous, this contrasts with other studies that found these groups to be more abundant in areas with nutrient enrichment associated with sewage discharge (Palacios-Sánchez et al., 2019). This discrepancy highlights that even though the input of nutrient-rich water creates favourable phytoplankton growth conditions (Grimes & Finucane, 1991) 1863)) that enter estuaries during the initial or reproductive phases of their lives, which may explain the occurrence of smallbodied species closer to rivers. Furthermore, rivers modify the functioning of the coastal system processes by discharging large amounts of nutrients and sedimentary material into the water column, which are then distributed by marine currents and alongshore transport (Herrera & Bone, 2011). These nutrient inputs can be facilitated by the numerous small river tributary mouths (0.006-0.3 km) present near almost all of the beaches in our study.
These river tributaries may also serve as a source of organic and inorganic contamination derived from human activities, for especially due to discharges along its course, such as found in rivers in the study area (CETESB 2019), which can benefit some but limit other functional groups.
In summary, our results show that surf-zone fish assemblages are structured by an interplay between environmental and anthropogenic factors and highlight the importance of considering both drivers in ecological assessments. They also reaffirm the importance of considering functional traits to complement taxonomic assessments, since environmental and anthropic variables affect species with specific functional traits differently in ways that are not always reflected in the total species richness. Targeting low-wave sites with low urbanization is an important strategy to protect communities taxonomically and functionally more distinct. In particular, species which are of critical conservation importance (e.g. rays of the genus Dasyatis), were only found at sites with low waves and urbanization.
However, this strategy would miss trophic categories, such as piscivorous fishes (e.g. blacktip reef shark, Carcharhinus limbatus or several species of the genus Trachinotus), which were most abundant in high-wave and urbanized settings. We also recognize that extending surveys into, for example, other seasons or places might further encompass rare or unique species that comprise regional biodiversity.
Therefore, our recommendation is more nuanced and involves an integrated strategy that aims at the conservation of pristine beaches with low wave height, protecting those from anthropic intervention, together with high-wave beaches to maximize taxonomic and functional diversity across the region. To this end, it is important that managers identify the natural and anthropic characteristics of multiple beaches, which are getting progressively accessible with the use of remote tools (Harris et al., 2011) to identify target beaches with these characteristics and evaluate which sites should be prioritized by conservation actions to preserve the integrity of surf-zones and the function of this ecosystem as a nursery for fish assemblages.

ACK N OWLED G EM ENTS
This work was supported by the Fundação de Amparo à Pesquisa do Network.

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

PEER R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/ddi.13453.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data is made available as supplementary material.

O RCI D
Yasmina Shah Esmaeili https://orcid.org/0000-0002-2343-1888 F I G U R E 5 Split bargraphs illustrating the relative importance of explanatory variables on the fish species richness, abundance singular species, functional richness, dispersion and evenness