Aims
Transformations of species and functional compositions on subtropical reefs are ongoing due to poleward range shifts of some tropical species, with largely unknown consequences to ecosystem functioning. Trait-based approaches are powerful tools to quantify such changes. Here, we evaluated changes in the trait composition of coral-associated fish communities along a tropical to a temperate environmental gradient of ca.1400 km in southern Japan with abundance-weighted trait expression to assess how trait complexity changes with increasing latitude.

Location
Ryukyu Islands and southern Pacific coast of Japan

Taxon
Reef fish

Methods
We tested for shifts in trait space and functional redundancy, based on five morphological, life history, and behavioural traits: maximum length, pelagic larval duration, trophic level, substrate preference, and reproductive mode. Our trait database was coded with two approaches, first, by attributing a single value to each trait per species, and second, by fuzzy coding that allows more than one value per trait and hence considers some intraspecific trait variation.

Results
We found a reduction in specialist habitat traits (coral substrate preference, nesters, herbivores) and an increase in generalist traits (predators) with increasing latitude, along with a contraction in trait space from tropical to temperate reefs. Functional redundancy declined with increasing latitude. These trends were closely linked with latitudinal gradients in temperature, along with changes in other environmental factors such as turbidity and photosynthetically active radiation.

Main Conclusion
Functional turnover and contractions are thus likely due to the marginal conditions for coral-associated fishes at higher latitudes, favouring generalist species, whereas increased resources at lower latitudes favour high redundancy and niche partitioning. Accounting for intraspecific trait variation indicated the same trends but highlighted increased functional vulnerability across all sites. We show that trait complexity in coral-associated fish communities decreases from tropical to temperate reefs, highlighting the reduced functional scope that comes with marginal environmental conditions.

Study sites & data collection
Visual census fish surveys were carried out at 31 sites along the Pacific coast of Japan, located between 24^o and 35^o northern latitude, with climates ranging from tropical to temperate. This thermal gradient creates several areas with different environmental properties which were clustered into regions: tropical, warm subtropical, cold subtropical and warm temperate. Surveys were carried out along belt-transects (25m x 5m) in coral habitat, at a standardised depth of 8-12m during the northern summers in 2015 for all sites except sites 20-31, which were surveyed in 2016. At our most northern site (Tateyama), surveys were conducted at 3m depth as the coral habitat did not extend to deeper depths. We recorded individual fish lengths estimated to the nearest cm of all target species along the transects, whilst swimming along 5 replicate transects per site. The abundance of each species was counted visually and standardised using log(1+x) to meet normality assumptions.

Fuzzy set theory
Unlike traditional trait-based approaches, the fuzzy coding method allows species to be placed in more than one category per trait to account for uncertainty in ecological characteristics (Chevene et al., 1994; Cheung et al., 2005; Jones and Cheung, 2018), and some degree of intraspecific trait variation. For example, it is difficult to know whether a fish species is of medium or large size, as size may vary with environment or habitat. Cheung et al. (2005) therefore used fuzzy set theory when describing the maximum length of fish, allowing fish species to be placed in more than one category, whereby in some communities a species is classed as a medium fish and as a large fish in others, with the degree of membership to each category (i.e., small, medium or large) defined by relative abundance. For example, a fish species with a maximum size of 68 cm can be classified as large or medium, with a degree of membership of 0.7 (large) and 0.3 (medium), therefore within 70% of hypothetical populations the species is classed as large but in the other 30%, it is medium (Cheung et al., 2005). Here, we used this fuzzy approach such that fishes whose traits vary within a species could be placed into several categories for the same trait.

Species Traits
Species traits were compiled from the primary literature and FishBase (Froese and Pauly, 2021). Five traits were selected to describe the morphological, behavioural, and biological functional niches of the reef fish species surveyed (Brandl et al., 2019; Miller et al., 2023). Two different trait databases were created from the trait information: a categorical single trait value database, and a fuzzy-coded trait database. The categorical single-trait database followed traditional trait-based techniques and allowed for one value per trait per species. These values were obtained from FishBase, where the mean (numerical traits) or the dominant reported value (categorical traits) were assigned (Cook et al., 2022; Anderson et al., 2022; Clay et al., 2023; Miller et al., 2023). Traits for the categorical trait database included two numerical (maximum length and pelagic larval duration (PLD)) and three categorical (trophic level, substrate preference, and reproductive method) data formats. The maximum length and PLD were assigned to four categories to make them comparable to the fuzzy database. The fuzzy trait database applied the same categories as the categorical database but allowed species traits to be placed in more than one category (Chevene et al., 1994, Cheung et al., 2005, Jones and Cheung, 2018). Values for the fuzzy database were obtained from FishBase (Froese and Pauly, 2021) and other literature (Craig, 1996; Shibukawa and Suzuki, 2002; Wilson et al., 2008), capturing a range of trait values per species. Each trait value was assigned a weighting between 0 and 3, relating to the affinity of the species for the trait value, where 0 indicated no affinity and 3 complete affinity. These were then standardised so that the sum of each trait was equal to 1. These weightings were assigned depending on the abundance of the trait value within the literature and thus were not site-specific, e.g., where a secondary trait value was available, the most dominant reported value was assigned a 2, and the secondary value 1. This strategy allowed for some intraspecific variability along the temperature gradient in response to different environmental pressures to be estimated based on abundance-weighted trait expression.  

Community Weighted Means
To calculate community weighted means (CWM) a site-by-species abundance and species-by-trait matrices were combined through the “functcomp” function from the “FD” package in R (Laliberté et al., 2014). For categorical data, the CWM represents the abundance of each trait at each site (Lavorel et al., 2008). 

References
Anderson, L., McLean, M., Houk, P., Graham, C., Kanemoto, K., Terk, E., McLeod, E. and Beger, M. 2022. Decoupling linked coral and fish trait structure. Marine Ecology Progress Series. 689, pp.19–32.

Brandl, S.J., Rasher, D.B., Côté, I.M., Casey, J.M., Darling, E.S., Lefcheck, J.S. and Duffy, J.E. 2019. Coral reef ecosystem functioning: eight core processes and the role of biodiversity. Frontiers in Ecology and the Environment. 17(8), pp.445–454.

Cheung, W.W.L., Pitcher, T.J. and Pauly, D. 2005. A fuzzy logic expert system to estimate intrinsic extinction vulnerabilities of marine fishes to fishing. Biological Conservation. 124(1), pp.97–111.

Chevene, F., DolÉAdec, S. and Chessel, D. 1994. A fuzzy coding approach for the analysis of long-term ecological data. Freshwater Biology. 31(3), pp.295–309.

Clay, C.G., Reimer, J.D., Cook, K.M., Yamagiwa, H., Gravener, E., Theodora, L.H.Y. and Beger, M. 2023. Temporal functional changes in coral and fish communities on subtropical coastal coral reefs. Marine and Freshwater Research.

Cook, K.M., Yamagiwa, H., Beger, M., Masucci, G.D., Ross, S., Lee, H.Y.T., Stuart‐Smith, R.D. and Reimer, J.D. 2022. A community and functional comparison of coral and reef fish assemblages between four decades of coastal urbanisation and thermal stress. Ecology and Evolution. 12(3), p.e8736.

Craig, P. 1996. Intertidal territoriality and time-budget of the surgeonfish, Acanthurus lineatus, in American Samoa. Environmental biology of fishes. 46, pp.27–36.

Froese, R. and Pauly, D. 2021. FishBase.

Jones, M.C. and Cheung, W.W.L. 2018. Using fuzzy logic to determine the vulnerability of marine species to climate change. Global change biology. 24(2), pp.e719–e731.

Laliberté, E., Legendre, P., Shipley, B. and Laliberté, M.E. 2014. Package ‘FD’. Measuring functional diversity from multiple traits, and other tools for functional ecology.

Lavorel, S., Grigulis, K., McIntyre, S., Williams, N.S.G., Garden, D., Dorrough, J., Berman, S., Quétier, F., Thébault, A. and Bonis, A. 2008. Assessing functional diversity in the field–methodology matters! Functional Ecology. 22(1), pp.134–147.

Miller, M., Reimer, J.D., Cook, K.M., Pandolfi, J.M., Sommer, B., Obuchi, M. and Beger, M. 2023. Temperate functional niche availability not resident-invader competition shapes tropicalisation in reef fishes. Nature communications.

Shibukawa, K. and Suzuki, T. 2002. Asterropteryx atripes, a new gobiid fish from the Western Pacific Ocean (Perciformes: Gobioidei). Ichthyological Research. 49, pp.274–280.

Wilson, S.K., Burgess, S.C., Cheal, A.J., Emslie, M., Fisher, R., Miller, I., Polunin, N.V.C. and Sweatman, H.P.A. 2008. Habitat utilization by coral reef fish: implications for specialists vs. generalists in a changing environment. Journal of Animal Ecology., pp.220–228.

All data files can be opened in Excel.