Introduction

In recent decades, losses of coral cover have coincided with marked increases in macroalgae on Caribbean coral reefs (Hughes 1994; Gardner et al. 2003; McManus and Polsenberg 2004). Such substantial changes in community composition have raised concerns about the resilience of these valuable ecosystems (McClanahan et al. 2002; Hughes et al. 2007) because reductions in stony coral cover can have far-reaching impacts on coral reef ecosystem services (Wild et al. 2011). Persistent shifts from scleractinian coral to macroalgal cover have been linked to compounding reef degradation factors, such as reductions of herbivorous species and increases in nutrients, often in combination with widespread coral mortality events (Done 1992; Hughes et al. 2010).

While phase shifts from coral to fleshy algal species have been well documented (Done 1992; McManus et al. 2000; Hughes et al. 2010), shifts to encrusting algae are not as widely known. However, over the last decade, reports of abundant encrusting red algae in the genus Ramicrusta (family Peyssonneliaceae) have emerged throughout the Caribbean Sea (Pueschel and Saunders 2009; Ballantine and Ruiz 2013; Eckrich and Engel 2013; Edmunds et al. 2019). Ramicrusta spp. (hereinafter Ramicrusta) have been deemed a potentially severe threat to Caribbean coral reefs due to their rapid growth and strong ability to outcompete other benthic organisms; Ramicrusta overgrow a variety of reef species (Eckrich et al. 2011; Smith et al. 2018) and often quickly dominate the substratum as a highly abundant taxon (Slijkerman et al. 2011; Williams and García-Sais 2020). First described in the Caribbean by Pueschel and Saunders in 2009, three species of Ramicrusta have been identified in the region: R. monensis in Puerto Rico; R. bonairensis in Bonaire, the Lesser Antilles, and Puerto Rico (Ballantine et al. 2016); and R. textilis in Jamaica, Puerto Rico, and St. John (Pueschel and Saunders 2009; Ballantine and Ruiz 2013; Edmunds et al. 2019). Although R. textilis has been identified in the US Virgin Islands (USVI) in St. John, its presence in St. Thomas has yet to be confirmed.

In the USVI, Ramicrusta have expanded to over half of the 34 monitoring sites of the Territorial Coral Reef Monitoring Program (TCRMP) and have quickly become the most abundant benthic taxon at multiple locations (Smith et al. 2018). For example, Savana Island (8 m depth) and College Shoal East (30 m depth) exhibited high prevalence of Ramicrusta in 2016 (occurring in 98.25% and 87.30% of video clips, respectively) (Smith et al. 2018). Both sites are offshore with reduced land-based human impact, and College Shoal is a mesophotic reef within a no-take marine-protected area, suggesting that direct human disturbance is not a necessary driver of Ramicrusta abundance. However, reductions in coral cover may promote expansion of the algae. At Savana Island, the percent cover of Ramicrusta increased fivefold from 2005 to 2007, from 4.64% ± 1.28 SE to 20.2% ± 2.39, after loss of coral cover in the 2005 bleaching event (20.3% ± 4.73 to 13.1% ± 1.82) (Smith et al. 2018).

Ramicrusta are known to overgrow several species of corals in the Caribbean and cause colony mortality (Fig. 1; Eckrich and Engel 2013; Ruiz 2015; Smith et al. 2018). In addition, Ramicrusta may impair coral recruitment (Eckrich and Engel 2013; Bramanti et al. 2017; Wilson et al. 2020). Although the algae have the potential to negatively affect Caribbean coral populations, it remains uncertain which scleractinian species are most at risk. In Bonaire and Puerto Rico, Porites astreoides, Porites porites, and Orbicella annularis had over 30% of their colonies overgrown by Ramicrusta, while Siderastrea siderea, Acropora palmata, and Agaricia sp. were overgrown by the algae in a lower proportion (Eckrich and Engel 2013; Ruiz 2015). Despite nonsignificant differences in overgrowth rates of Ramicrusta among three coral species (Millepora complanata, P. astreoides, and P. porites) reported by Eckrich and Engel (2013), other coral taxa (e.g., Orbicella spp.) may be more vulnerable to Ramicrusta overgrowth. Coral taxa can have differential competitive abilities (Lirman 2001; Nugues et al. 2006; Swierts and Vermeij 2016), underscoring the importance of investigating species-specific competition with Ramicrusta.

Fig. 1
figure 1

Photograph of Ramicrusta textilis overgrowing a star coral (Orbicella sp.)

Full size image

As shifts from coral to algal abundance have become more common on Caribbean reefs, there is a growing need to better understand the dynamics of coral–algal competition to predict ecological impacts and inform management decisions. Recently, advancements in imaging technology have made it possible for non-experts to measure the complex three-dimensional (3D) growth of sessile reef organisms with minimal training and affordable equipment (Gutiérrez-Heredia et al. 2015, 2016). 3D reconstruction of benthic organisms from underwater photographs has gained popularity as a method to capture more meaningful morphological changes (e.g., surface area and volume) because it is less field-intensive, less intrusive, and more informative than other approaches (Burns et al. 2015; Reichert et al. 2016) and mm-scale accuracy can be reasonably achieved (Holmes 2008; Ferrari et al. 2017; Olinger et al. 2019).

The goal of this research was to use 3D photogrammetry to explore variation in competitive interactions between massive Caribbean corals and Ramicrusta. Linear and areal growth rates of Ramicrusta were quantified from 3D models of individual coral colonies competing with the algae across: (1) coral taxa and (2) geographic location. Given the lack of information about Ramicrusta in the Caribbean, additional data were collected from all study sites to identify the species of Ramicrusta present in St. Thomas, examine Ramicrusta–coral interaction prevalence across location and coral species, and assess potential impacts of the algae on coral health.

Materials and methods

Overview

High-resolution 3D models of individual coral colonies were generated to measure the linear and areal expansion of Ramicrusta over time. Coral species with a massive morphology were selected for 3D modeling precision and selection of coral-like control substrates. Coral colonies were tagged and photographed in October 2018 and re-photographed 5 months later to calculate Ramicrusta growth rates using models of the same colony from two time points. To measure algal growth unhindered by coral competition, control replicates were photographed and modeled on dead coral or limestone pavement covered in Ramicrusta and a mixed epilithic algal community. Controls with a macro-morphology similar to massive coral colonies were selected to mimic the structure of the coral replicates. Two studies were conducted using this methodology to investigate algal overgrowth across coral taxa (Study 1) and geographic location (Study 2).

Study 1: Overgrowth across species

To evaluate species-specific differences in competition with Ramicrusta, algal growth rates were compared for five massive coral taxa that were most commonly competing with the algae in the USVI: Orbicella faveolata, Orbicella annularis, Porites astreoides, Siderastrea siderea, and Pseudodiploria strigosa. Four controls and four replicates of each of the selected coral species were marked at three representative study sites (Fig. 2; N = 12 controls and 12 colonies of each coral species).

Fig. 2
figure 2

Map of the southwest side of St. Thomas with study sites. Shape indicates which study corresponds to which sites (square = Study 1 only, triangle = Study 1 and Study 2, circle = Study 2 only). Gray areas show landmasses and color indicates depth. Inset shows average linear growth rates of Ramicrusta textilis (± SE) on controls and Siderastrea siderea colonies across sites from Study 2

Full size image

Study 2: Overgrowth across location

Algal growth rates on S. siderea colonies and controls were compared across eight study sites (Fig. 2) to determine whether Ramicrusta competes more aggressively at different locations. S. siderea was chosen because the species was commonly found interacting with Ramicrusta at all study locations. Four controls and four replicates of S. siderea were marked at each site (N = 32 controls and 32 coral colonies). Data were collected at each site to assess benthic cover, community composition, coral health, and Ramicrusta–coral interaction prevalence (see below: Site characterization).

Study sites

All sites were located on the south side of St. Thomas, USVI, in proximity to the University of the Virgin Islands Center for Marine and Environmental Studies (Online Resource 1, Fig. 2). Each site was limited to a shallow depth range of six-to-nine meters to maximize the natural light necessary for 3D reconstruction. Fortuna Bay (FB), Savana Island (SI), and Flat Cay Windward (FW) were selected as Study 1 sites for species-specific overgrowth rates. FW and SI were also included as sites in Study 2 on location-specific overgrowth rates using separate replicates. Six additional locations were selected for Study 2: Flat Cay Leeward (FL), Buck Island (BI), Perseverance Bay (PB), Water Island (WI), Saba Island Windward (SW), and Saba Island Leeward (SL). Study 1 sites were selected for their high prevalence of coral–Ramicrusta interactions, allowing for sufficient replication of the five coral species. Study 2 sites were selected for their variation in Ramicrusta abundance (Table 2).

DNA sequencing

To identify the species of Ramicrusta present in the USVI and verify whether the same species was under investigation at each study site, representative samples of the algae were collected from each of the nine sites in April 2019 and placed in silica gel for DNA sequencing. Specimens were located near marked coral colonies at each study location. Between one and five samples were taken per site for a total of 23 specimens.

Specimen preparation, extraction, and sequencing were done at the University of North Carolina (UNC) Chapel Hill, following Gabrielson et al. (2011); amplification for rbcL followed Gabrielson et al. (2011); and amplification for psbA followed Adey et al. (2015). Forward and reverse sequence reactions were compiled and edited using Sequencher (Gene Codes Corporation, Ann Arbor, MI, USA), and consensus sequences were accessioned into GenBank.

Growth rates of Ramicrusta

Setup and data collection

Coral colonies and controls exhibiting Ramicrusta overgrowth were selected at each site in a confined area of approximately 900 m2 via haphazard roving diver surveys. Healthy colonies with minimal interactions (e.g., low cover of tissue by fleshy macroalgae) were preferred, but not exclusively selected. The average maximum length of coral colonies included in Studies 1 and 2 was 19.5 cm (± 0.02 SE) and 16.4 cm (± 0.01), respectively. Replicates were at least 1 m apart at each site and within a 4-m depth range across sites, with the exception of SI, which is slightly deeper (Online Resource 1). Prior to photographing the replicates, divers carefully removed any benthic species that could interfere with the growth measurements (primarily Dictyota spp.) by hand and cleared the colony of sediment with brushes to ensure clear visibility of the coral–Ramicrusta border. A minimum of 60 photographs were taken 360° around each colony and from overhead, with minimal rotation between subsequent photographs to ensure overlap. Initial photographs for both studies were taken in October 2018, and final photographs were taken in March and April 2019 using a 12-megapixel Olympus Tough TG-5 digital camera.

Photogrammetry

All 3D reconstructions were created using Agisoft Metashape Professional following methodology similar to Olinger et al. (2019). After photograph alignment, the resulting 3D sparse point cloud was refined using the gradual selection and camera optimization tools. Final sparse point clouds were batch-processed to create high-quality dense point clouds using aggressive depth filtering, and a high-quality textured mesh was built with an unlimited face count. The textured meshes were scaled, keeping the error threshold below 1 mm. Final models were assessed for mesh quality at the Ramicrusta growth border before exporting. See supplementary material for an example of a finished model (Online Resource 2).

Ramicrusta growth was measured using CloudCompare software. For each replicate, two textured meshes (one from each time point) were finely aligned and superimposed using the point picking tool (Online Resource 3). A minimum of four static points (e.g., individual corallites) were matched across the two time points to align the meshes. The alignment error was limited to achieve 1-mm accuracy for all coral replicates; however, the error increased to 3 mm for some control replicates due to the lack of corallites. After aligning the meshes, the segmentation tool was used to isolate the longest “clean” subsection of the coral–Ramicrusta border. A clean border was defined as a clearly reconstructed area where Ramicrusta were overgrowing living coral tissue without any sediment, macroalgae, or blockages obscuring the competitive margin at either time point. A rectangular subsection was simultaneously segmented out of both meshes (Online Resource 3a) and analyzed for linear and areal Ramicrusta growth using the trace polyline and segmentation tools (Online Resource 3b; 3c). There was minimal vertical growth and change in total subsection area between time points, and an attempt was made to keep total subsection area similar among replicates.

Site characterization

At each of the nine study sites, divers assessed benthic cover and coral health along the same six 10-m transects located randomly on the reef near the study area. Benthic cover directly beneath each transect was recorded every 10 cm, and all biotic entries were identified to the lowest possible taxonomic level. To assess coral health, divers assessed as many corals as possible on the left side of 1-m belt transects, which ranged from 4 to 10 m length depending on coral density. Every living coral > 0.5 cm in maximum diameter was identified to the lowest possible taxonomic level. Corals were assessed in situ for interactions with Ramicrusta and signs of mortality, bleaching, and disease using a modified TCRMP protocol (Smith et al. 2018). Ramicrusta interactions were defined as living coral tissue visibly affected by a layer of Ramicrusta overgrowth (Fig. 1).

Statistical analyses

All statistical analyses were performed in R using the stats package; graphics were created using the ggplot2 package. An alpha level of 0.05 was set for all statistical tests. When post hoc tests were conducted with Bonferroni correction, P values were multiplied by the number of comparisons in the analysis. For ANOVA analyses, assumptions of homoscedasticity and normality were investigated.

Growth rates of Ramicrusta

Ramicrusta growth rate analyses were conducted on the two subsections corresponding to each replicate (one subsection from each time point). For each subsection pair, linear algal growth was calculated as the average of 15 measurements of linear extension across the advancing edge of Ramicrusta. This level of replication was selected to achieve a standard deviation below ± 0.0003 mm for average linear growth measurements on five test models. Areal algal growth was calculated as the percent surface area coverage of Ramicrusta in the subsection at the second time point minus the percent surface area coverage of Ramicrusta at the first time point. Both measures of Ramicrusta growth were calculated as a daily rate. Three replicates that could not be properly aligned or clearly measured were omitted from further analyses; one replicate each at PB, SW, and WI for Study 2. For Study 1, two-factor ANOVAs were used to determine the effect of species (control, O. faveolata, O. annularis, P. astreoides, S. siderea, and P. strigosa) and location (FB, SI, and FW) on the linear and areal growth rate of Ramicrusta. For Study 2, two-factor ANOVAs were performed to determine the effect of location (BI, FL, FW, PB, SI, SL, SW, and WI) and substratum (control and S. siderea) on the linear and areal growth rate of Ramicrusta.

Data validation

Using different sets of photographs from the same day, five replicate models of a single S. siderea colony were generated. A single-base model was paired with the replicate models to measure R. textilis growth using the above workflow. SE values for two sets of growth rates (five measurements on the replicate models testing 3D model variability and five measurements on the same model pair testing human measurement variability) were compared to the difference in average growth rate between two significantly different replicate groups (S. siderea and O. faveolata). The standard error values for the average linear (− 0.010 ± 0.0006 mm/d) and areal (− 0.099 ± 0.0100 percent/d, mean ± SE) validation growth rates were between one and two orders of magnitude smaller than the difference in average growth rate between the different replicate groups (+ 0.028 mm/d and + 0.158 percent/d, respectively). Therefore, the statistical differences in the data were due to ecologically relevant growth rates and not measurement variability associated with the 3D reconstruction process.

Ramicrusta interaction prevalence

Using the belt transect data, the prevalence of Ramicrusta interactions was calculated for each study site by dividing the number of corals with overgrowth by the total number of corals assessed. A two-way ANOVA was used to test the effect of coral species (P. astreoides, S. siderea, P. porites, Agaricia agaricites, Siderastrea radians, and Orbicella franksi) and location (FB, FL, and WI) on interaction prevalence. This subset of six coral species and three sites was used to omit data with a low count of coral colonies (N < 10 species per site). Data were squared to meet assumptions of homoscedasticity.

A single-factor ANOVA was used to separately test the effect of location (all sites) on interaction prevalence. In addition, a Kruskal–Wallis rank sum test was used on a subset of data to test the effect of abundant coral species with a total N > 100 across all sites (P. astreoides, S. siderea, P. porites, A. agaricites, S. radians, and O. annularis) on interaction prevalence. A one-way ANOVA was also used to separately test the effect of the five Study 1 coral species on interaction prevalence.

Relationship with coral health

A Pearson’s Chi-square test of independence was conducted to examine the relationship between Ramicrusta interaction (presence/absence) and coral health status (healthy, diseased, bleached, old mortality, and recent mortality).

Results

DNA sequencing

Each of the 23 collected specimens was identified as Ramicrusta textilis, confirming that the same species was under investigation at all study sites. Collections ranged from 4.5 to 17.5 m depths. Field identification based on morpho-anatomy was confirmed by DNA sequencing of each specimen using the plastid-encoded rbcL gene (691 base pairs) and/or the psbA gene (850 base pairs). These gene sequences were identical to sequences obtained from topotype material of R. textilis. Sequences were deposited in GenBank (for the rbcL gene, KC130226; for the psbA gene KM360015) and are publicly available using accession numbers MT215139–MT215161 (https://www.ncbi.nlm.nih.gov/genbank/). All 23 sequences were identical to one another regardless of depth, morpho-anatomy, and substrate. All preserved specimens were deposited in the UNC Chapel Hill Herbarium (NCU 672,224–672,246).

Growth rates of R. textilis

Study 1: Overgrowth across species

Two-factor ANOVAs revealed clear differences in both linear and areal growth rates of R. textilis across coral species, but no significant difference across sites and no interaction (Table 1a). The linear growth rate of R. textilis on S. siderea was significantly slower than growth on all species except controls, and the linear growth rate of the alga on O. annularis colonies was significantly faster than growth on controls (Fig. 3). The areal growth rate of R. textilis on S. siderea was also significantly slower than on all other species (Online Resource 4). In addition, S. siderea was the only species that exhibited negative average overgrowth rates of R. textilis (recession of the algae over time). The alga grew approximately 2.5 times slower on S. siderea than on controls and about 4.2 (areal) to 6.7 (linear) times slower than on O. annularis. The rate of growth of R. textilis was about 1.7 (areal) to 2.6 (linear) times faster on O. annularis than on controls. The fastest average linear growth rate ± SE was + 0.039 ± 0.006 mm d−1 on O. annularis colonies, indicating that projected overgrowth on corals could reach nearly 1.5 cm yr−1 along advancing edges of growth.

Table 1 Results of two-way and one-way ANOVA analyses of Ramicrusta textilis: (a) growth rates (Study 1 and 2), (b) cover (all sites), and (c) interaction prevalence (subsets of belt transect data)
Full size table
Fig. 3
figure 3

Average linear growth rates of Ramicrusta textilis (± SE) by site and species. Study 1 site codes: FB = Fortuna Bay, FW = Flat Cay Windward, SI = Savana Island. Study 1 coral species: Orbicella annularis, Orbicella faveolata, Porites astreoides, Pseudodiploria strigosa, and Siderastrea siderea. Pairwise comparison results shown from Tukey’s honestly significant difference post hoc test; substrata sharing a common letter have no significant difference

Full size image

Study 2: Overgrowth across location

Two-factor ANOVAs revealed significant differences in linear and areal growth rates of R. textilis across both substrata and site (Table 1a). Linear and areal growth rates of R. textilis occurred significantly faster on controls than on S. siderea colonies (Fig. 2; Online Resource 5). The post hoc analysis was unable to detect site-specific differences in linear growth rates; however, the areal growth rates of R. textilis were significantly slower at FW than at PB (Online Resource 5). Notably slow (and sometimes negative) growth rates of R. textilis on S. siderea were suspected to increase variability among the data, making it difficult to detect significant differences among sites. Therefore, to determine whether R. textilis growth rates vary by location, single-factor ANOVAs were conducted on control growth only. Results revealed a significant difference in areal growth rates across sites, but no significant difference in linear growth rates was detected despite mean growth rates that were up to 7 times different (Table 1a; Table 2; Fig. 2). A post hoc test with Bonferroni correction showed a significant difference in areal growth rate between PB and FL, FW, and SL as well as a significant difference between FW and SW. The fastest average linear growth was measured at SW and SI, and the fastest average areal growth was measured at PB; rapid growth rates were observed despite low cover of R. textilis at both PB and SW (Table 2).

Table 2 Ramicrusta textilis cover, growth rates, and interaction prevalence by site (mean ± SE)
Full size table

R. textilis interaction prevalence

The two-factor ANOVA results indicated that R. textilis interaction prevalence did not differ across sites or coral species with no significant statistical interaction, but separate one-way analyses revealed significant differences across both location and taxa (Table 1c). Average interaction prevalence appeared to differ substantially across both sites and species, with means reaching over an order of magnitude in difference. Because the analysis was conducted on a subset of data, insufficient replication of coral species was suspected to be contributing to the results. To address this, single-factor analyses were conducted separately to assess interaction prevalence across all study locations. Results showed a significant effect of location on R. textilis interaction prevalence (Table 1c); prevalence was significantly greater at FW than at WI (Table 2). In addition, there was a significant difference in R. textilis interaction prevalence across coral species with a total sample size of N > 100 across sites, with a significantly lower interaction prevalence for S. radians than all other coral species (Table 1c, Fig. 4). Despite a total sample size of 220 S. radians across all sites, no R. textilis interactions were observed. O. annularis had the greatest recorded average interaction prevalence of all of the coral species, with 24% (± 5.3 SE) of the corals interacting with R. textilis across all sites (Fig. 4). When interaction prevalence was examined separately for the five coral species from Study 1, there was no significant effect. Similar to average growth rate observations, relatively high interaction prevalence co-occurred with low cover of R. textilis at some sites (e.g., SW; Table 2). Furthermore, extremely high interaction prevalence (55% and 64%) occurred with less than 1% cover of the alga on individual transects at SW.

Fig. 4
figure 4

Average interaction prevalence of Ramicrusta textilis (± SE) by coral species. Scleractinian species with total N > 100 across all sites are displayed. Asterisks indicate coral species that were included in the two-way ANOVA analysis across three sites (NOTE: Orbicella franksi also included in the two-way ANOVA, but not displayed because total N < 100 across all sites.) Pairwise comparison letters shown from Tukey’s honestly significant difference post hoc test for one-way ANOVA across coral species with N > 100; coral species sharing a common letter have no significant difference. Coral species: Porites astreoides, Siderastrea siderea, Porites porites, Agaricia agaricites, Siderastrea radians, and Orbicella annularis

Full size image

Relationship with coral health

The Pearson’s Chi-square test of independence revealed a significant relationship between coral health status and R. textilis interaction (χ2 (4, N = 4671) = 440.6, P < 0.001). Variables that contributed the most to the Chi-square statistic were corals with old mortality and healthy corals (both with and without an R. textilis interaction), followed by corals with recent mortality and an R. textilis interaction. For colonies with an R. textilis interaction, the observed frequency was significantly greater than the expected frequency for old and recent mortality, while the observed frequency was significantly less than the expected frequency for healthy corals (Fig. 5). Therefore, corals with an interaction are more likely to have old and recent mortality and less likely to be healthy than corals without an interaction. The total observed frequency of all negative coral health parameters (bleaching, disease, and mortality) was also higher for coral colonies interacting with R. textilis than colonies without interactions (Fig. 5).

Fig. 5
figure 5

Mosaic plot showing relationship between Ramicrusta textilis interaction (YES = interaction present, NO = interaction not present) and coral health status (BL = bleaching, DIS = disease, HEALTHY = no bleaching, disease or mortality observed, OLD = old coral mortality, REC = recent coral mortality). Blue color and solid outlines indicate frequency that is greater than expected, while red color and dashed outlines indicate frequency that is lower than expected. Color indicates significant result, and intensity of the color corresponds to magnitude of deviation from the expected frequency from Pearson Chi-square test of independence

Full size image

Discussion

Overgrowth rates of R. textilis can be strongly influenced by the coral species colonized by the algae. Although most coral taxa were poor competitors against the alga, S. siderea demonstrated a competitive ability to reverse overgrowth by R. textilis. Conversely, O. annularis was an exceptionally poor competitor with R. textilis in rate of overgrowth and the peak prevalence of interactions. Overall, R. textilis poses a severe threat to coral reefs in the USVI and potentially the greater Caribbean. The alga successfully outcompetes four of the five massive scleractinian species most commonly interacting with the alga in the region. Because these species include major reef-building taxa, it is possible that the emergence of R. textilis could have significant impacts on both coral species diversity and the structural complexity of reefs over time.

While only massive-shaped corals were targeted for this research, subtle differences in morphology could have contributed to the differential competitive abilities exhibited across species. Specifically, the lobed structure of O. annularis colonies could facilitate overgrowth by R. textilis by providing space that is favorable for the alga to establish. The alga generally grows from the base of each lobe upwards, presumably engulfing the living upper surface over time. The nonliving bases of O. annularis lobes are inaccessible to large grazers and contain few competitors, which could allow the alga to easily recruit on multiple lobes and aggressively overgrow the coral from multiple contact points. Furthermore, the vertical growth pattern of O. annularis inhibits its ability to overgrow R. textilis, unlike plating or encrusting species which can overgrow the alga horizontally. While these are plausible explanations for the enhanced susceptibility of O. annularis, species with the same morphological traits (namely hemispherical-massive P. strigosa and S. siderea) exhibited significant differences in competitive abilities against R. textilis, suggesting that morphology is not the primary factor influencing these competitive outcomes.

Differences in coral species-specific interaction prevalence with R. textilis were also evident, but did not necessarily reflect the differences in overgrowth rates among coral taxa. An analysis of interactions from the belt transect data indicated that S. radians was not susceptible to R. textilis overgrowth, with no R. textilis interactions observed on any of the colonies assessed. This could reflect a general robustness of corals in the family Siderastreidae to R. textilis interactions; however, differences in interaction prevalence among Study 1 coral taxa were not detected despite differences in the overgrowth rates on O. annularis and S. siderea. While O. annularis did exhibit generally higher interaction prevalence than other species, possibly because of the increased availability of contact points for the alga to establish on multiple lobes, S. siderea did not appear to interact with the alga less frequently than other taxa. Because S. siderea exhibited resistance to algal overgrowth, standoff interactions between R. textilis and S. siderea could have resulted in more prevalent interaction observations for this species. It is also possible that the initiation of interactions between corals and R. textilis does not depend on the competitive ability of the coral species, whereas overgrowth rates show a clear hierarchy.

R. textilis harms corals typically by overgrowing living tissue, but could also impact colony health less directly. Frequency analysis indicated that the presence of R. textilis on corals increases the likelihood that colonies also display signs of bleaching, disease, and mortality. Conversely, evidence of bleaching, disease, and mortality could increase the likelihood that a coral interacts with R. textilis. Regardless, signs of partially bleached tissue past the coral–algal margin were observed on multiple colonies throughout this research, which could be an indication of allelopathic capabilities of R. textilis and/or a result of coral stress. Competitive mechanisms in addition to overgrowth of R. textilis and its indirect impacts on coral health should be investigated (McCook et al. 2001); chemical interactions may be influenced by ecological and environmental factors such as coral species (Bonaldo and Hay 2014) and water flow (Brown and Carpenter 2015; Jorissen et al. 2016).

This research revealed that R. textilis is capable of aggressive growth and competition regardless of its abundance at a site. Some of the fastest average R. textilis growth was measured at PB despite the low average cover of the alga at this site (1.5%, Table 2). Furthermore, cases of extremely high interaction prevalence and moderate overgrowth rates co-occurred with very low cover of R. textilis at SW, suggesting that aggressive interaction and competition with corals occur even with a low abundance of the alga (< 1% average cover, Table 2). This could be a reflection of R. textilis recruitment preferences; it is possible that the alga establishes and/or grows more successfully near corals. Future studies could examine fine-scale spatial patterns of R. textilis occurrence and growth within coral reef patches to identify factors that may drive this pattern. Because R. textilis aggressively interacts with and overgrows corals regardless of its abundance, future research should also take care to focus on measurements in addition to percent coverage of the alga, such as growth rate and interaction prevalence.

This research has successfully expanded on the known physical characteristics of Ramicrusta in the USVI, confirming the presence of R. textilis in St. Thomas using DNA sequencing following its prior identification in St. John (Edmunds et al. 2019). Sequencing results indicate phenotypic differences in coloration (brown to red) and form (flat encrusting to flaky) may be attributed to external influences (e.g., light availability and water motion) rather than genetic variation at the species level, at least for the two plastid encoded markers that were sequenced. The depth range of R. textilis described in this research is 4.5–17.5 m, and average growth rates reached a maximum of + 0.04 mm d−1 and + 0.20 percent d−1 on O. annularis colonies. The maximum linear growth rate (+ 0.04 mm d−1) was 50–67% slower than the average overgrowth rates reported by Eckrich and Engel (2013) (+ 0.06 to + 0.08 mm d−1), which could indicate species-specific differences in the overgrowth rates of R. textilis (USVI) versus R. bonairensis (Bonaire). In making this comparison, it should be noted that the variability reported by Eckrich and Engel (2013) was greater than this study and sample sizes were smaller. Alternatively, these differences could be attributed to different study period lengths or simply reflect greater geographical variations in growth. As this was a 5-month study, seasonal variations in growth could influence competitive interactions and estimations of average growth rate (Renken et al. 2010).

While it still remains unknown whether the genus Ramicrusta is invasive to the Caribbean, a retrospective analysis of video surveys from the USVI revealed a low abundance of apparent R. textilis as early as 2003 (Smith et al. 2018), which precedes its original description by Pueschel and Saunders (2009). Considering that the alga was likely present in the Caribbean years before it was recognized and identified, R. textilis may be a cryptic native species that has only recently been documented due to its population expansion. Furthermore, a recent study reported R. textilis comprising up to 29% of the benthic cover and overgrowing living corals in the South China Sea (Nieder et al. 2019). This was the first report of the alga harming Pacific corals since the first description of the genus in 1981 (Zhang and Zhou 1981; Ballantine et al. 2016; Nieder et al. 2019). Because the species has similarly expanded and demonstrated harm to corals in a region where it is known to be native, it is possible that R. textilis is a competitive indigenous species that emerged in great abundance due to environmental change occurring over the last decade in the Caribbean and more recently in the Pacific.

A thorough understanding of coral interactions with R. textilis is imperative to recognizing the processes that currently shape benthic community structure on coral reefs and predicting future changes. Unchecked expansion of R. textilis across Caribbean reefs may exacerbate unfavorable coral community composition shifts from generalist taxa (e.g., Orbicella spp.) to stress-tolerant species (e.g., Siderastrea spp.), with potentially drastic reductions in coral species diversity due to the superior competitive abilities of the alga. These shifts could have far-reaching impacts on the basic ecological function of reefs and their ecosystem services. This underscores the importance of using informed management strategies to effectively mitigate R. textilis based on information about its ecology and physiology. It will also be important to evaluate the physiology of R. textilis under conditions that were not assessed in this research—particularly varying nutrients and temperature—and consider how the alga may respond to predicted environmental changes.