Biogeography of functional trait diversity in the Taiwanese reef fish fauna

Abstract The richness of Taiwanese reef fish species is inversely correlated to latitude as a direct consequence of the abiotic environment and its effects on benthic habitats. However, to date, no studies have investigated the variations in the diversity of traits (FD) linked with the role of these fishes in the ecosystem. FD is usually considered more sensitive than species richness in detecting early changes in response to disturbances, and therefore could serve as an indicator of ecological resilience to environmental changes. Here, we aim to characterize FD in the Taiwanese reef fish fauna and to document its regional variations. Six traits were used to categorize the 1,484 reef fish species occurring in four environmentally contrasted regions around Taiwan. The number of unique trait combinations (FEs), their richness (FRic), their redundancy (FR), their over‐redundancy (FOR), and their vulnerability (FV) were compared among these regions. Overall, 416 FEs were identified. Their number decreased from south to north in step with regional species richness but FRic remained similar among regions. FR and FOR were higher to the south. At the local scale, variations in FEs and FRic are in concordance with the worldwide pattern of FD. High‐latitude, impoverished fish assemblages, offer a range of trait combinations similar to diversified tropical assemblages. Increasing diversity in the latter mainly contributes to raising FR and supports already over‐redundant entities. High vulnerability makes many combinations highly sensitive to species loss, and was higher at intermediate latitudes when using a fine resolution in trait categories. It suggests that the loss of FEs may first be characterized by an increase in their vulnerability, a pattern that could have been overlooked in previous global scale analyses. Overall, this study provides new insights into reef fish trait biogeography with potential ramifications for ecosystem functioning.


| INTRODUC TI ON
Rising seawater temperatures from climate change are pushing tropical marine organisms close to their upper thermal limits (Walther et al., 2002) and transforming coral reef ecosystems Stuart-Smith, Brown, Ceccarelli, & Edgar, 2018). At a global scale, this is commonly associated with a flattening of coral reefs (Alvarez-Filip, Dulvy, Gill, Côté, & Watkinson, 2009;Bozec, Alvarez-Filip, & Mumby, 2015;Wilson et al., 2010), which further reduces the availability of niche spaces upon which a large number of living organisms depend (Graham & Nash, 2013). These modifications notably alter the structure of reef fish assemblages as well as their trophic interactions (Darling et al., 2017;Graham et al., 2007;Pratchett, Hoey, Wilson, Messmer, & Graham, 2011;Richardson, Graham, Pratchett, Eurich, & Hoey, 2018;Wilson et al., 2010). The susceptibility of fish species to extirpation is usually nonrandom  benefits generalists, which today seem able to proliferate on reefs worldwide (Munday, 2004;Richardson et al., 2018). This occurs to the detriment of a small proportion of species with often prominent roles in maintaining ecological processes (Mouillot, Bellwood, et al., 2013a), although generalists can play influential roles too. The loss of functionally important fishes can eventually decrease the capacity of the ecosystems to recover from, or resist transitions to, alternative states, which may have only a limited capacity to sustain original ecosystem services (Chong-Seng, Nash, Bellwood, & Graham, 2014;Fung, Seymour, & Johnson, 2011;Hoey & Bellwood, 2011;Hughes et al., 2017;Pratchett, Hoey, & Wilson, 2014). Consequently, a better understanding of the diversity in functional groups, as well as the functional relationships between coral habitats and associated fish assemblages are of critical importance in managing reefs and prioritizing conservation efforts (Richardson et al., 2018).
Recently, the unique combination of traits (characteristics) describing the role of fish species in the ecosystem has tentatively been proposed as synthetic representation of the functions they performed . This representation is commonly considered a pragmatic description of their potential niche and corresponds to any feature at the individual level that can affect and determine how species use resources and how they interact with each other (Cadotte, Carscadden, & Mirotchnick, 2011). To date, traits responding to environmental variations and focusing on individual performance ("functioning traits"; see Denis, Ribas-deulofeu, Sturaro, Kuo, & Chen, 2017) have been largely overlooked in this definition. Instead, six categorical traits related to food acquisition and locomotion, considered pertinent to representing key facets of fish ecology (size, diet, mobility, gregariousness, period of activity, and vertical position in the water column) have been widely employed (Mouillot, Bellwood, et al., 2013a;Mouillot et al., 2014). The depiction of the diversity of those traits through discrete values has gained in popularity, and sometimes adjusted, for providing a better insight into ecosystem functions beyond that given by species diversity measures (Richardson et al., 2018). Although, several important caveats undermine this relationship (Brandl, Emslie, & Ceccarelli, 2016) and this approach is perfectible (Villéger, Brosse, Mouchet, Mouillot, & Vanni, 2017). However, functional trait-based approach and, more precisely, the characteristics of the unique trait combinations, represents today a benchmark for assessing the diversity of functional traits in the communities (Mouillot, Graham, et al., 2013b).
Diversity of traits (FD), the range of things that organisms do in communities and ecosystems (Petchey & Gaston, 2006), is directly connected to ecosystem processes and stability (Chapin et al., 2000;McCann, 2000;Purvis & Hector, 2000). FD is measured in several ways (Cadotte et al., 2011), and its monotonic response to disturbance usually makes it a better indicator than the traditional taxonomic measurements of early changes affecting reef communities (D'Agata et al., 2014;Mouillot, Graham, et al., 2013b). As an example, focusing on trophic trait, Bellwood, Hughes, Folke, and Nyström (2004) examined fish species richness in 14 pre-determined categories comparing Caribbean and Australian reefs. The Caribbean assemblage was characterized by a lower redundancy (i.e., average number of species among categories) than the Australian, which depicted the higher vulnerability of Caribbean reef ecosystems for the trait of interest.
By including morphological information into a trait morphospace, Bellwood, Wainwright, Fulton, and Hoey (2006b) further confirmed high versatility in trophic categories characterizing coral reef fish diversity. In most feeding guilds, this was supported by a considerable overlap in the occupation of the morphospace by different morphologies. Further extension to behavioral traits reinforced that a small number of key combination of traits could provide the basic ecological structure of reef fish assemblages (Guillemot, Kulbicki, Chabanet, & Vigliola, 2011). This is illustrated on a global scale by a high redundancy that is disproportionately packed into few trait combinations (over-redundancy). In a nonrandom assembly, it leaves a large number of entities supported by only one species, thus increasing the overall vulnerability of trait combinations in the reef fish fauna and the potential sensitivity of FD to the loss of a few species . Surprisingly, FD is revealed to be relatively stable across biogeographical provinces despite their high variability in species richness . Hence, impoverished fish faunas such as those from the Tropical Eastern Pacific and the Atlantic could maintain the range of ecological processes necessary for the growth and persistence of tropical reefs since they share most of the key functions of reefs having richer fauna (Bellwood et al., 2004;Johnson, Jackson, & Budd, 2008). Higher diversity in other provinces mainly contributes to over-redundancy, leaving FD highly vulnerable to species loss.
Those changes can modify the trophic structure of an ecosystem (Ferreira, Floeter, Gasparini, Ferreira, & Joyeux, 2004) and eventually influence regional FD. However, there is always the possibility that a local variation has been largely overlooked in previous studies.
Taiwan is located to the north of the East Indies Triangle, the region hosting the highest marine diversity in the world (Briggs, 2005). It encompasses tropical and subtropical latitudes and spans the transition of two marine realms where three provinces and ecoregions converge (Spalding et al., 2007). The Taiwanese reef fish fauna is highly diverse (Allen, 2008), and species richness decreases steeply from southeast to northwest in correlation to the patterns in sea surface temperatures that shape reef habitats (Dai, Soong, Chen, Fan, & Hsieh, 2005). Therefore, Taiwan constitutes an ideal location for investigating the potential effects of environmental settings on the FD of reef fish assemblages.
The objectives of this study were to (a) compare FD among the main regions of coral development around Taiwan with respect to prevailing environmental conditions, and (b) examine the local spatial patterns of FD of Taiwanese reef fish fauna in comparison to global patterns. To achieve this, the present study compares the richness, redundancy, over-redundancy, and vulnerability of the unique functional trait combinations identified in four coral regions around Taiwan characterized by contrasting environmental conditions.

| Reef fish fauna
Our study focuses on four well-supported reef fish assemblages (so-called regions, Figure 1a): North, East (including Ludao), West (Penghu Archipelago), and South of Taiwan (including Xiaoliuqiu). This distinction is justified by environmental conditions producing F I G U R E 1 (a) The four studied regions around Taiwan. (b) Their respective species richness and number of unique species, unique trait combinations (FEs), range of unique trait combinations (FRic). The y-axis is the percentage of richness in all bar plots contrasting benthic assemblages (Chen, 1999;Chen & Shashank, 2009;Ribas-Deulofeu et al., 2016) and hosting distinct reef fish assemblages (Shao, Chen, & Wang, 1999). The warm waters of the Kuroshio Current flow from the southern point of Taiwan along its east coast toward the Ryukyu Archipelago, pushing tropical organisms (e.g., scleractinian corals; Chen, 1999) northward. Accordingly, fringing reefs only occur on the East and South coasts of Taiwan, where average monthly seawater temperatures remain above 20°C.
In contrast, the frequent occurrence of waters lower than 18°C in winter prevents the accretion of reefs to the West and the North of Taiwan (Chen, 1999;Kleypas et al., 1999;Wang & Chern, 1988).
There, only non-reefal and less diversified coral assemblages develop directly on the basalt substrate that shapes the coastline. Reef fish composition in these four regions were obtained by corroborating information from Taiwan FishBase (Shao, 2018), FishBase (Froese & Pauly, 2018), and additional references documenting local reef fish diversity (Chen, 2004a(Chen, , 2004bChen, Jan, Kuo, Huang, & Chen, 2009;Chen, Shao, Jan, Kuo, & Chen, 2010;Shao et al., 2008). Here, a fish was considered as a coral reef species if it is partly or strictly associated, during its lifetime, to shallow coastal coral habitat, including the non-reefal environments to the West and the North of Taiwan. For each species, its synonymy and status validity were checked in the World Register of Marine Species (WoRMS Editorial Board, 2018), and only valid species were used for further analysis. This step proved to be of critical importance in the preparation of data extracted from databases.

| Trait selection
Our final list of reef fish species among the four regions in Taiwan encompassed 1,484 species. A species was considered as unique to a region when it was absent from the others. Each species was classified into six categorical traits reflecting its possible functions (the role of the species) in the ecosystem: size, diet, mobility, gregariousness, period of activity, and vertical position in the water column (following Mouillot et al., 2014). These traits and their levels are detailed in Supporting Information Appendix S1. Their relevance in describing reef fish functions has been previously evaluated tentatively by Mouillot et al. (2014). The strength of the application of these six traits depends on the extent to which these traits really are indicative of functional attributes (Mouillot, Graham, et al., 2013b). However, their extensive use in the broader field of reef fish functional ecology does not imply their links to ecosystem functioning are clear (see Beauchard, Veríssimo, Queirós, & Herman, 2017;Villéger et al., 2017 for discussion). Species trait information was extracted from databases and/or relevant references (Froese & Pauly, 2018;Shao, 2018). Because trait values can differ among sources, priority was first given to FishBase, followed by Taiwan FishBase, then other references. Missing species trait values were infilled with specific literature on the given species or with the expertise of the authors considering the phylogenetic position of the species. The final dataset provides a species list, regional distributions, and selected traits.

| Trait space
A dissimilarity (trait) matrix was produced by a pairwise comparison of the FEs using Gower distance (S15). S15 has the advantage of being suitable for mixed (continuous and categorical) variables, and is therefore applicable here. A Principal Coordinates Analysis (PCoA) was then computed on the basis of this trait matrix applying a Caillez correction to correct any potential negative eigenvalues generated (Cailliez, 1983). Euclidean distances among FEs on the first four axes of this PCoA were firmly correlated with the initial Gower matrix (Mantel test, r = 0.77, p < 0.001), and the addition of an extra axis only marginally increased this resolution. Therefore, information on the first four axes was considered the most pragmatic representation of variation among FEs. The coordinates of FEs on these axes were used to represent synthetic trait space (to visualize relationship among FEs) and later calculate FD indices.

| Trait diversity indices
where FE is the total number of trait entities, S is the total number of fish species, and is the number of species in a trait entity .
FV and FOR are influenced by the number of species, the num-

| Fish richness and unique trait combinations
Fish species richness ranges between 618 species in the North to 1,278 species in the South, which respectively represent 41.6% and 86.1% of Taiwan's total richness (Figure 1b). The South has the highest number of unique species (306 species, 23.9% of the regional richness), followed by the North and East with 78 (12.6%) and 65 (7.4%) unique species, respectively. The western region hosts only 14 unique species (2.2% of the regional richness).  Figure 1b).

| Trait space and richness
Most of the variation in our Gower matrix comparing the FEs was caught in the first four axes of our trait space (Mantel test, R 2 = 0.58, p < 0.001). It also reduced the mean squared deviation considerably between the initial distance and, the standardized final distance in the trait space (quality_funct_space function, mSD = 0.01).
Therefore, the dispersions among FEs and the computation of

| Vulnerability, redundancy, and overredundancy of trait combinations
The Unique species contribute 8.7% in the South, 3.6% in the East, 0.0% in the West, and 7.0% in the North, to FV (Figure 2).

| D ISCUSS I ON
Around Taiwan, the reef fish fauna encompasses 1,484 species, which is exceptionally high considering the small overall area of reefs (~940 km 2 ) around the island (Allen, 2008). It yields 416 FEs, which correspond to 7.3% of the theoretical maximum possible. In comparison, using the same traits and categories, the world reef fish fauna (6,316 species) yields 646 FEs, and 11.4% of the theoretical combinations . Taiwan is located at the northern limit of the Central Indo-Pacific province, in which the fish fauna yields 468 FEs across 3,600 species . Although there is a two-fold difference in species richness between the overall province and Taiwan, there is an 89% similarity in the total number of FEs observed. This similarity in FEs suggests that the poorer Taiwanese reef fish fauna may be able to maintain the ecological processes necessary for the growth and persistence of reef ecosystems (Bellwood et al., 2004;Johnson et al., 2008).  The four studied regions also had similar and their associated traits, as observed around Singapore (Wong et al., 2018).
Environmental conditions in the North and the West of Taiwan are marginal for corals (Dai & Horng, 2009) and thus limiting for reef accretion (Kleypas et al., 1999).  Mouillot et al., 2014). This supports the notion that the addition of new species does not contribute substantially toward generating new FEs (Halpern & Floeter, 2008). Instead, it tends to make existing functional fish assemblages more robust by offering high "insurance" in a small number of FEs.
The over-representation of species in a few FEs was further robust to a drastic reduction of the possible number of FEs as in the global reef fish fauna . In addition, we demonstrated that the over-representation of a low number of FEs is better preserved than the FR along a decreasing gradient of species richness. Despite a similar range of FEs offered, it makes the functionality of the fish assemblages from northern, western, and eastern regions potentially more vulnerable and less resilient to disturbances than the southern region.
The location of Taiwan at the periphery of the Central Indo-Pacific province is reflected in the relatively high vulnerability of its fish fauna. Regional FVs ranged between 51.2% and 61.7%, which is actually in the range reported from the Tropical Eastern Pacific province (54.2%) and much higher than the FV of the Central Indo-Pacific (38.5%; Mouillot et al., 2014). This suggests that, independently of the location around Taiwan, at least half of the FEs are supported by only one species, leaving many FEs without insurance when facing species loss (Mouillot, Bellwood, et al., 2013a). Rare species, both in terms of local abundance and regional occupancy, often support unique and distinct functions (Mouillot, Bellwood, et al., 2013a) that could be critical for the resilience of coral reefs (Bellwood, Wainwright, et al., 2006b). Unfortunately such species are usually the most susceptible to extraction and extirpation , which may result in the functional homogenization of fish assemblages and a proliferation of generalist species (Munday, 2004;Richardson et al., 2018).
Characteristic of a nonrandom fish assemblage, species loss should result in an increase in FV and/or an increase in FOR (Halpern & Floeter, 2008). Yet, regional patterns observed at large spatial scales support a global mismatch between species richness and FV  A high FV has further been interpreted as a pristine state and a baseline for assemblages untouched by human for priority conservation (Quimbayo, Mendes, Kulbicki, Floeter, & Zapata, 2017). In this scenario, FV could act as an indicator of a functional "tipping-point," a stage prior to more severe functional changes occurring along environmental gradients or from human disturbances. An exciting outcome of these observations will be to test whether FV can be applied to temporal community dynamics and if it could serve as a dimension for assessing ecological shifts. However, FV level is sensitive to the traits considered as well as their categorization .
Overall, FV decreases with a crude categorization of traits and raises with latitude, which makes the FEs recorded in marginal species-poor regions more vulnerable to species loss than in richer assemblages.
This higher FV observed to the North was not caused by unique species. Instead, coarse trait resolution could decrease the susceptibility of identifying FEs that are vulnerable and overlook subtle functional changes in the assemblages. Therefore, the resolution in trait categorization and its outcomes should be considered and interpreted carefully according to the scale of changes expected.
In this study, the Taiwanese reef fish fauna offered new insights into the biogeography of reef fish trait diversity, which has recently raised as an important facet of the diversity. It constitutes the first step toward a better understanding of the fishes' role in the Taiwanese reefal areas. Taiwanese reef fish fauna retains a high proportion of trait combinations defined at the scale of the Central Indo-Pacific province, but remains highly vulnerable to species loss. Possible overfishing and the consequences of human activities on benthic habitats are likely to have already modified historic diversity (Liu et al., 2009;Ribas-Deulofeu et al., 2016). Therefore, this study proposes a baseline upon which the gain and loss of unique trait combinations could be immediately assessed after updating information on species richness. The integration of abundance and/or biomass to the current framework (see Chen et al., 2015) would improve the present trait assessment and the functional interpretation of the Taiwanese reef fish fauna. This would further enhance the prospects for exploring functional relationships between reef fish assemblages and coral habitats.

ACK N OWLED G M ENTS
We thank all members and collaborators of the Functional Reef Ecology Lab (Institute of Oceanography, National Taiwan University) for their useful comments and discussion on this project. This study was funded by the Ministry of Science and Technology of Taiwan

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

AUTH O R CO NTR I B UTI O N S
VD designed the experiment. VD and JWC collected the data.

DATA ACCE SS I B I LIT Y
All data and script for data analysis are available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.838v1j5.