Winners and losers of reef flattening: An assessment of coral reef fish species and traits

Anthropogenic stressors are causing widespread coral mortality, leading to loss of coral cover and decreased structural complexity that threatens reef biodiversity, functioning, and ecosystem services. Reef fishes are intimately linked to coral reef complexity, but we lack a generic understanding of which species are particularly affected by reef flattening and what traits make them susceptible. We used extensive species- and trait-based analyses to build a framework for western Atlantic fish association with both structural complexity and coral cover to better understand the implications of reef degradation. These analyses also investigated the relative importance of live coral versus the value of the structure it provides to reef fishes. We modeled how 25 biophysical and anthropogenic factors correlated with the densities of 109 fish species across 3292 Floridian reef sites. The importance of a metric of structural complexity and coral cover to the abundance of each species was then isolated. Species with positive associations were categorized as likely future ‘losers’ and negative associations as ‘winners’. We showed that structure loss was more critical than loss of coral cover, as 53% of species were predicted as losers on low-relief reefs, while only 11% were losers with decreased coral cover. We found morphological, behavioral, and ecological traits mediate species’ responses to reef degradation, and shared evolutionary history is unlikely to be a strong driver of trait relationships. Eight traits explained 79.7% of variation in species’ associations with relief and six traits explained 27.8% of associations with coral cover. Smaller, streamlined, habitat and trophic generalists are more likely winners on flattened reefs and large-bodied predators and species with deeper bodies and intermediate caudal fin shapes are likely losers. Identifying these important traits provides insight into mechanisms that may link fish and complex habitats, which allows us to better predict assemblage-wide responses to future reef flattening.

Methods
Reef and fish survey data

Since 1979, NOAA has conducted fishery-independent coral reef fish and benthic surveys across southeast Florida as part of the Reef Visual Census (RVC) and National Coral Reef Monitoring Program (NCRMP) (Bohnsack et al. 1999; Smith et al. 2011; NOAA 2017). Sites were surveyed annually (before 2014) or biennially (2014 to present) using a two-stage stratified random sampling scheme in conjunction with a 100x100 m resolution grid covering Florida’s Coral Reef. The first stage stratified grid cells by habitat type within three sub-jurisdictions (Florida Keys, Dry Tortugas/Marquesas, and Southeast Florida), and survey sites were randomly selected within each 100x100 m cell. One or two diver pairs conduct stationary point counts and size assessments of over 500 species of fish (see Bohnsack and Bannerot 1986 for details of this approach), visually estimate benthic coral cover, and measure rugosity based on maximum vertical relief of the substrate within individual 7.5m radius cylinders. Briefly, each diver created a list of fish identified to species level that entered the cylinder in the first five minutes, counting and sizing highly mobile species. The following ten minutes were used to count and size individuals remaining in the cylinder while adding any new fish entering the cylinder. The rest of the dive was used to assess the benthos, including measuring maximum height of hard relief and visually estimating live hard coral cover from an overhead view within the cylinder (NOAA 2017). Maximum hard relief is a metric of reef complexity measured as the average of the highest rigid point within 8 segments of the 7.5m survey cylinder. Coral cover was measured as the average of live hard coral visually estimated within 8 segments of the 7.5m survey cylinder. Data from the pair of divers (or from multiple pairs of divers where appropriate) at the same site were averaged. We used survey data from 2005 to 2018, excluding 2010 when a significant cold snap caused fish kills (Kemp et al. 2016). Out of 7046 unique sites in the region surveyed since 2005, using data from the most recent surveys if surveyed multiple times, 3292 sites were designated as coral reef or hardbottom by the FWC UFRTM Level 2 classifications. We included 109 reef fish species from 25 families (Table 1), all of which were present at 4% or more of sites, which was experimentally determined as the minimum number of non-zero values necessary to run the density models. We excluded cryptic species that are difficult to survey with point counts (e.g. species in families Gobiidae, Blennidae, and Muraenidae) (Willis 2001).

Biophysical and Anthropogenic Predictors

To model species densities at each survey site, we compiled biophysical and anthropogenic explanatory variables (Table 1) from in situ, remotely sensed, and published data sources. Justification for inclusion and full derivation of each predictor is available in the supplementary materials (SI). Maximum hard relief, coral cover, and depth were all measured in situ by NOAA NCRMP divers. Maximum hard relief provides a proxy for large scale complexity, capturing boulders and drop offs, while coral cover generally captures small-scale complexity. Month and year of surveys were included as categorical variables to account for seasonal and longer-term trends in species densities.

The remaining predictors were extracted (using the coordinates of the underwater surveys) from continuous geospatial data layers with a grid size of 100 x 100 m in ArcGIS Pro (v10.7 ESRI). The biophysical variables used to predict individual species’ densities from these layers included UFRTM habitat level 2 class, the total area of coral reef and hardbottom habitats within 20 km of each site, connectivity to mangrove and seagrass nursery habitats, distance to deep water habitats, lowest monthly mean sea surface temperature (SST), net primary productivity (NPP), larval connectivity, wave exposure, and distance to the nearest mapped fish spawning aggregation (FSA). We also included 10 anthropogenic variables to account for human impacts, specifically impacts from fishing which has major effects on populations of reef fishes like snappers and groupers (Zuercher et al. 2023). These predictors included the number of recreational anglers within 50 km, human population within 50 km, human population per reef area within 50 km, the number of marina slips for boats under 14 m within 25 km, gravity of all potential fish markets (within 500 km), the number of federal commercial (within 50km) and charter (within 25 km) snapper-grouper permits, metrics of community fishing engagement and reliance, the estimated number of tourist fishers, and the protected status of reefs.

Fish trait data compilation

We assembled 13 morphological, behavioral, and life history traits to identify which traits were predictive of a species’ relationship to relief or coral cover (Table 2). Justification for inclusion and full derivation of each set of fish traits is available in the supplementary materials (SI). Trait values were collected from a combination of published literature, online databases (particularly FishBase, Froese and Pauly 2022), and measurements from publicly available photographs. For morphological traits derived from photographs, a single value was obtained from the mean of three lateral images of adults of the species of interest. Trait values from local specimen images (i.e., in Florida or northern Caribbean) were used where available. When possible, continuous, quantifiable analogs were used in place of categorical variables to produce higher quality functional spaces (Maire et al. 2015). Maximum total length and body fineness (total length to body depth ratio), were included based on the known importance of morphology in determining predation risk and availability of spatial refuge (Green and Côté 2014). We included the presence of physical or chemical defenses as categorized by Green and Côté (2014) or as described in the ‘Biology’ or ‘Threat to humans’ sections of FishBase (Froese and Pauly 2022). Defenses such as sharp spines or barbs that make capture or ingestion by predators difficult (see Price et al. 2015) or toxins that harm predators or decrease palatability (see Harris and Jenner 2019) may reduce the need for physical refuge. Swim mode (fin and body region and movement combinations used for propulsion, e.g. labriform or subcarangiform) and aspect ratio (ratio of height squared to surface area) of the caudal fin were used to highlight the importance of swimming performance for predator avoidance and maneuverability for navigating complex reefs (Fulton 2010). We included schooling (Green and Côté 2014) as an important anti-predator defense (Magurran 1990) that may reduce reliance on physical refuge. Additionally, nocturnality (Green and Côté 2014) was included as a potential behavior to reduce the risk of visual predators (Kronfeld-Schor and Dayan 2003); however, nocturnal species may also rely on structure for diurnal refuge (Ménard et al. 2008). Home range size, depth range, and the use of multiple habitat types (hereafter referred to as multihabitat) were included as proxies for specialization to shallow coral reef habitats (Luiz et al. 2013). Position in water column and spawning mode were included to identify specific associations with the benthos (Luiz et al. 2013). Finally, trophic level (Froese and Pauly 2022) was included because reef structure has a variety of effects on feeding, for example, complexity may increase the surface area for herbivores to graze (González-Rivero et al. 2017) or provide hiding places for ambush predators (Harborne et al. 2022).