Discussion
A variety of human activities can impact coral reefs directly and indirectly, resulting in their degradation (Pandolfi et al. 2003) and a decline of fish diversity associated with this habitat (Graham et al. 2011). In particular, the coral cover of Curaçao has been steadily decreasing over the last decades (Gardner et al. 2005, Jackson et al. 2014) although to a lesser extent than most other islands in the Caribbean and with variations between different coastal stretches of the island (Waitt institute 2017). Here, using eDNA, we showed differences in species functional and phylogenetic compositions between two coastal areas on the Southern side of the island of Curaçao. By comparing two reef stretches under different degrees of anthropogenic pressures, we showed how eDNA combined with species features from associated databases and advanced modelling approaches can deliver ecological indices that can inform ecosystem status. Management toward the preservation of coral reef ecosystems requires monitoring approaches that can be quickly deployed in the field (Obura et al. 2019), and we demonstrated that eDNA metabarcoding provides as rich fish assemblage information as UVC in term of taxa samples, but which requires significantly less sampling time and resource in the field. Yet, remaining gaps in the reference database still limit the information provided by eDNA. Building on increasing evidence of the monitoring capacity of eDNA metabarcoding (DiBattista et al. 2017, West et al. 2021, Polanco Fernández et al. 2021), our study illustrates how this technique could evolve toward a general approach for the monitoring of fish communities on coral reefs.
Functional and phylogenetic characteristics are expected to offer higher dimensions of information to describe and manage ecosystems (Strecker et al. 2011). Coupling ecological indices with eDNA can provide more complete ecosystem information for coral reefs (Aglieri et al. 2020, Marques et al. 2021). As demonstrated previously with UVC (D’agata et al. 2014), we found that functional and phylogenetic indices better discriminate between the two inventoried coastal areas than taxonomic information alone. While the two reefs were similar regarding the fish species richness recovered from eDNA, we found more marked differences in their functional and phylogenetic properties. Specifically, the Valentijnsbaai reef area contained larger species such asAetobatus narinari , more pelagic species (e.g. Thunnus sp,Istiophorus sp.) with higher trophic levels (e.g.Acanthocybium solandri ). In addition, crypto benthic species are also present such as the mimic cardinalfish (Apogon phenax) or the pale cardinalfish (Apogon planifrons) increasing the functional diversity. In contrast, Willemstad presented higher phylogenetic diversity, mainly driven by a few phylogenetically distinct species associated with soft bottoms (Albula vulpes , Elops smithi ) or the water column (Anchoa colonensis ,Opisthonema oglinum ). The higher occupancy of this coastal stretch by sandy bottom and pelagic species could reflect the higher state of degradation of the coral reefs near the city. These findings suggest that environmental filtering under high levels of coastal development near Willemstad and high levels of sediments is associated with distinct fish assemblages as previously documented using UVC in Singapore (Wong et al. 2018). Hence, even if the difference between the fish assemblages in two coastal areas is subtle, the combination of eDNA metabarcoding surveys, functional and phylogenetic information allow their discrimination. Díaz-Pérez et al. (2016) proposed that the estimation of coral reef health indices should be complemented with fish community indices, to improve the accuracy of the estimated health status of coral reefs in the western Caribbean Sea. In future research, indices such as the Reef Health Index (RHI) could be complemented with multidimensional information including functional and phylogenetic indices from eDNA to inform policy makers about reef health status (Obura et al. 2019).
With the combination of eDNA metabarcoding including all MOTUs and novel statistical approaches (i.e., HMSC), we reveal a greater power of eDNA to discern species occupancy across the two coastal stretches in comparison with traditional UVCs. The application of joint species distribution models to eDNA was suggested to increase the ecological interpretation of the molecular signal (Burian et al. 2021). For similar sampling effort, eDNA metabarcoding outperformed UVCs in its capacity to identify the contrast between the two coastal areas and detected more negative responses to the more anthropogenically stressed reef area. Importantly, some of the strongest responses of MOTUs to the spatial contrast were assigned to species that are elusive, highly mobile, and cryptic. In contrast, UVCs could fail to detect the occurrence of those species, thus increasing uncertainty in their estimated responses to the environment in the distinct coastal areas. Additionally, eDNA metabarcoding generates more identifications of taxa as MOTUs than UVCs does. When we combine this richer data with HMSC, a statistical framework that reduces parameter uncertainty (via shrinkage) across similarly responding species, we can obtain greater confidence in species responses. MOTU response was further associated with a phylogenetic signal, indicating a strong distinction between clades with a positive response (Apogonidae, Murenidae) and those with more negative responses (Labridae) towards more anthropogenic stressed areas. We expect that, assuming that MOTUs are true diversity units acting as a species proxy, the generation of more data (MOTUs) to feed statistical models will lead to more robust indicators of ecological status (with a higher certainty of responses). That said, key sources of uncertainty still exist in using eDNA to assign species and a better coverage within reference databases will yield more information on the taxonomic units recovered from eDNA (Valdivia‐Carrillo et al. 2021), to the point where generating MOTUs as a species proxy will become unnecessary if almost all regionally occurring species are genetically referenced.
Increasing evidence suggests that eDNA metabarcoding offers higher species detection abilities compared with traditional surveys (Polanco Fernández et al. 2021, Valdivia‐Carrillo et al. 2021), which was confirmed in our study with the greater number of MOTUs detected with eDNA (129 MOTUs) than fish species in UVC (120 species). We found overlap in species composition between eDNA and UVC, but also differences. While several species of relatively high abundance and easy to detect visually such as Bodianus rufus andMicrospathodon chrysurus were detected with both methods, the UVC detected more shallow reef species (e.g. Acanthurus spp), which were not detected with eDNA. The shallow reef of Curaçao is characterized by a very thin stretch averaging 40 m and the eDNA transects were conducted slightly further away from the coast at approximately 100 m of distance, which could explain why some of the reef fish species were not detected. Our results suggest that the eDNA signal could be spatially localized (e.g. as in West et al. 2021), stressing the need for careful eDNA sampling to capture the entire signal of a habitat. Nevertheless, both methods of observation detected distinct fish composition between the two areas. While eDNA metabarcoding can provide a rapid inventory of species composition (Polanco Fernández et al. 2021) and can better detect small and cryptic species, eDNA surveys cannot entirely replace UVC. In addition to generating species lists, UVC transects can provide fish ontogenetic stage, body size structure and abundance information that, at present, eDNA does not provide at all or not accurately (Rourke et al. 2021). These sources of information are key ecological indicators so that future surveys might integrate, when possible, the strengths of both survey approaches.