Study Site

The study was conducted in 5 sites within the Culebra Archipelago (lat 18°19′N, long 65°179′W) located 27 km east of Puerto Rico where juvenile hawksbill turtle aggregations occur at varying densities (van Dam and Diez, unpubl. data; Fig. 1). Surveyed sites were the coral reefs at Carlos Rosario (CR), Luis Peña Muelle (LPM), Luis Peña Avión (LPA), Cayo Lobo (CL), and Punta Soldado (PS). Sites included coral and rocky reef areas, and some were located inside a no-take Marine Protected Area. CR is mainly a linear reef surrounded by colonized bedrock and pavement, scattered coral rock, and a limited strip of continuous seagrass at its northern side. LPM is mainly a colonized pavement surrounded by scattered coral rock, bedrock, sand, and a small area of linear reef. CL is basically colonized bedrock surrounded at the east side by colonized pavement. The LPA and PS sites are characterized by having large extensions of colonized bedrock and pavement with channels. In PS, bedrock and pavement are surrounded mainly by small areas of scattered coral rock and spur and groove reefs (Kendall et al. 2001).

View Figure

• Download as Powerpoint
• Download Figure

Habitat Characterization

Benthic photo transects (n  =  166) were done between June 2008 and June 2009 in the 5 study sites to measure cover of live and dead coral, gorgonians, sponges, macroalgae, bedrock, sand, pavement, and rubble. A total of 46 benthic transects was conducted in CR, 33 at LPM, 24 at LPA, 37 at CL, and 26 at PS. Differences in the number of benthic transects per site respond to differences in the coastal length of each site. Random points generated with the Hawktools of ArcMap were used to place the initial point of transect sets inside study areas. Benthic surveys consisted of 5 10 × 1-m photo transects separated by 10 m in 2 depth zones (2–3 m and 7–8 m). Pictures of 1-m² quadrates were calibrated using the Coral Point Count with Excel Extension Program (http://www.nova.edu/ocean/cpce), and 20 random points were generated inside each quadrate and classified as belonging to one of the benthic features described above calculating their percentage frequency (Kohler and Gill 2006).

Turtle Surveys

Daytime snorkeling censuses were conducted at all sites between April 2008 and June 2009. Within a given a period of 3 months, all sites were visited at least once. Censuses were conducted by 2–5 observers, swimming parallel to each other and keeping a distance of 10 m between observers for duration of 1 hour. The average distance covered for each survey sample was 1.316 ± 0.426 km. A total of 15 1-hour surveys were conducted at each site during a 14-month period, with a total of 81.6 km surveyed during the study. Turtles captured by hand were brought on board a boat to be measured (maximum straight carapace length [SCL]) and tagged, to have esophageal content sampled, and released. Only turtles greater than 35 cm SCL were used to obtain diet samples. Individuals smaller than 65 cm SCL were considered immature (Witzell 1983; van Dam and Diez 1998). There was no difference in the number of swimmers per hour among sites (Kruskal-Wallis test statistic  =  5.078, p  =  0.279, df  =  4). The relative sighting frequency of sea turtles in the different feeding grounds was evaluated by dividing the number of turtles captured and sighted by the time spent during the snorkeling censuses. The time unit was defined as 1 hour of in-the-water census employing the capture methods described in Leon and Diez (1999), Diez and van Dam (2002), and Diez et al. (2002). We attempted to capture all sighted turtles; those that escaped were classified as sighted and also counted. An estimated size of sighted turtles was recorded together with the species identification. Turtle sighting locations were recorded with a global positioning system receiver, as are all survey transects' start and end points. Sightings were used in the catch per unit effort (CPUE) together with captures as an index of relative abundance of hawksbills. To prevent double counting of sighted turtles, we conducted the same transect once a day. Animals sighted but not captured swam away from the transect, making it difficult to be observed twice. The locations of captured and sighted turtles were plotted using ArcGIS 9.3 of ESRI. For turtles captured more than once, distance between the first and last capture was used as an index of relative movement within the study areas.

Diet Analysis and Food Availability

We use the diet content and food prey items availability data of Rincon-Diaz et al. (2011) to quantify the abundance of food resources in study sites and then related it with abundance of turtles. Percent cover of prey items was calculated in this study as the ratio of prey items' cover to the total surveyed area per study site. We exclude the algae Lobophora variegata from this quantification because it was present in diet contents of only 1 study site and not recorded in all of them. The number of colonies of the sponges Chondrilla caribensis, known previously as Chondrilla nucula (Rützler et al. 2007), and Cinachyrella kuekenthali and of individuals of Ricordea florida and Lebrunia danae were taken as an additional measurement of their abundance.

Statistical Analyses

The analysis of similarities (ANOSIM) and its pairwise comparisons was used to test for differences in percent cover of bottom types among and between study sites (Clarke 1993). Percent cover of benthic features of phototransects were arcsine square root transformed and standardized by the maximum value of the column following the recommendation for multivariate analysis (McCune et al. 2002; Quinn and Keough 2002). Nine hundred and ninety-nine permutations were used for all ANOSIM comparisons. A similarity percentage (SIMPER) analysis based on benthic features (Clarke and Gorley 2006) with a cutoff of 70% was used to determine the average dissimilarities between study sites. A nonmetric multidimensional scaling ordination (MDS) based on a similarity matrix created from the Bray-Curtis similarity analysis was used to plot similarities between different benthic transects in 2 dimensions. Kruskal-Wallis 1-way analysis of variance and Mann-Whitney U-tests were used to test for differences in the relative abundance of hawksbills (i.e., CPUE), and cover of prey items in benthic transects. All the analysis was performed using Systat 12 and PRIMER 6 Programs.

Habitat Characterization at 10-m² Scale

ANOSIM for benthic variables using study sites as factor revealed more differences among sites than within them (global R  =  0.391, p  =  0.001). Pairwise tests show that all possible pairs are different except for the LPM and PS reefs (p  =  0.494; Table 1; Fig. 2A).

View Figure

• Download as Powerpoint
• Download Figure

Table 1. Analysis of similarities pairwise tests for benthic variables between study sites. Asterisks indicate significant differences, p < 0.05.

View Fullscreen

SIMPER pairwise analysis for study sites suggests separation of CR from other study sites. This difference is due to the low cover of sand, pavement, rubble, and bedrock and high cover of gorgonians, coral, and dead coral with algae in this reef (Table 2). Differences between LPM with LPA, CL, and PS appear to be related to low cover of bedrock, gorgonians, and sponges and high cover of sand and rubble. Separation between LPA with CL and PS is due to low cover of bedrock and rubble and high cover of pavement, gorgonians, and sponges. Finally, differences between CL and PS are due to high bedrock and rubble cover, respectively. All mentioned variables explain together more than 70% of dissimilarity between study sites.

Table 2. Similarity percentage pairwise analysis for benthic variables in study sites.

View Fullscreen

The MDS analysis shows 4 groups with similarities in benthic variables with a stress of 0.21, which suggests a good configuration (Fig. 2B). The first group is formed by CL transects, the second one by LPM and PS transects, the third one by CL and LPA transects, and the last one by CR, LPM, LPA, CL, and PS transects. The CL transects contrast to other groups by the high frequency of bedrock and gorgonians. The second group is characterized by having high frequency of sand, dead coral and rubble, and gorgonians. The third group is characterized by high frequency of gorgonians and pavement. The last group is characterized by high frequency of gorgonians, live and dead coral, rubble, and sponges.

Turtle Abundance

A total of 80 captures of 58 tagged individuals since 1997, plus 39 sightings of hawksbill turtles, was recorded for all study sites (Table 3). CR showed a CPUE value of 3 turtles per hour, significantly higher than values found in the other study sites with CPUE less than 2 turtles per hour. In this study, 47 new individuals were marked, and of these, 12 were subsequently recaptured in the study sites. The mean within-site distance between the first and last capture of these 12 recaptured turtles was 0.292 ± 0.213 km (range  =  0.115–0.845 km). One individual was excluded from the mean distance calculation because it was the only one to move between sites (from CR to LPM, a distance of 4.28 km). Tagging of hawksbill turtles at the Culebra Archipelago started in CR in 1997. During this study, we recaptured at CR 3 turtles that had initially been tagged at this site during 2001, 2003, 2004; 2 that had been tagged in 2006; and 6 that had been tagged in 2007. These data suggest that some individuals are resident in the area for a number of years.

Table 3. Relative abundance of hawksbill turtles among study sites (n  =  15).

View Fullscreen

Relative abundance of hawksbill turtles significantly differed among sites (Kruskal-Wallis statistic  =  20.981, p  =  0.0001, df  =  4). CR has significantly more turtles than all the other sites (Mann-Whitney U-statistics  =  160.500, 189, 171, 198.500, p < 0.05, df  =  1), while no significant differences exist among the remaining sites (Fig. 3A).

View Figure

• Download as Powerpoint
• Download Figure

Availability of Food Resources

Abundance of diet items differed significantly among sites (Fig. 3B). PS is the study area with the highest cover of food resources, followed by CR, CL, LPA, and LPM. Percent cover of the sponge prey species C. caribensis and C. kuekenthali were different among sites (Kruskal-Wallis statistics  =  34.741 and 20.258, respectively, p  =  0.000, df  =  4) as well as number of colonies per transect (Kruskal-Wallis statistics  =  20.258 and 18.448, respectively, p < 0.0001, df  =  4).

Percent cover of C. caribensis in the 10-m² transects ranged from 0% to 4.99%, with a mean ± SD of 0.10 ± 0.42%. This sponge was very abundant in the PS reef (mean ± SD, 0.29 ± 0.99%), followed by CR (0.13 ± 0.22%), CL (0.06 ± 0.21%), LPA (0.02 ± 0.05%), and LPM reefs (0.01 ± 0.05%; Fig. 3C). The percent cover of the sponge C. kuekenthali ranged between 0% and 0.32%, with a mean ± SD of 0 ± 0.05%. Cinachyrella kuekenthali was abundant in CL (mean ± SD, 0.04 ± 0.08%), followed by LPA (0.03 ± 0.06%), CR (0.02 ± 0.05%), and PS (0 ± 0.01%). The other 2 prey species, the corallimorph R. florida and the sea anemone L. danae, were rare and did not show differences among study sites.

Turtle Abundance

Our estimates of relative abundance (CPUE) of juvenile hawksbill turtles indicate that with the exception of the CR site (3.07 turtles per hour), the other sites of the Culebra Archipelago have relatively low hawksbill density (1.73–0.53 turtles per hour) compared to other juvenile aggregation areas in the Caribbean, such as Mona and Monito Islands (van Dam and Diez 2008) and the Dominican Republic (Leon and Diez 1999). These studies used the same technique of daytime snorkeling censuses and calculation of CPUE. Capture and sighting frequencies of juveniles and subadults hawksbill turtles in feeding grounds around the Caribbean vary from 3.86 and 19.33 turtles per hour for Mona and Monito Islands, respectively, in Puerto Rico (van Dam and Diez 2008); 3.43 and 3.28 turtles per hour in the Dominican Republic (Leon and Diez 1999); 3.15 ± 0.98 turtle sightings per hour (mean ± SD) in the Cayman Islands (Blumenthal et al. 2009); to 0.51 turtle sightings per hour in the Los Roques Archipelago, Venezuela (Hunt 2009). Mean CPUE of hawksbill turtles in the CR site during this study was similar to values obtained in previous monitoring activities at this site. In 1997 and between 2000 and 2007, the mean CPUE was estimated to be 2 turtles per hour (1997), 1.75 (2000), 4.5 (2001), 3.44 (2003), 1.88 (2004), 2.66 (2005), and 2.87 (2007). This temporal variation in CR is comparable to observed variation in other places in the Caribbean with the exception of Monito and Los Roques sites. The relatively high turtle abundance in CR and the high site fidelity exhibited by juvenile hawksbills at these capture locations point toward the importance of the Culebra Archipelago as a developmental habitat of juvenile hawksbills turtles and for the need of continuing the monitoring and protection activities for turtle aggregations in these study sites.

Effect of Food on Hawksbill Turtle Abundance

The availability and distribution of certain prey species can regulate predator population sizes and their distribution patterns, as demonstrated by studies on crustaceans (Boscarino et al. 2007), tortoises (Doody et al. 2002), sharks (Heithaus et al. 2002; Wirsing et al. 2007), and felines (Weckel et al. 2006; Laundre 2010; Wilson et al. 2010). In the Caribbean, however, patterns of abundance of hawksbills appear not always to strictly follow the overall availability of preferred prey items. Only 2 studies have estimated food availability for adults and juvenile individuals or have inferred it using hawksbills' growth rates and diving duration in the Caribbean. While one study shows that variation in hawksbill relative abundances is related to the abundance of preferred food prey species based on diet composition (Diez and van Dam 2002), the others suggest that food availability of preferred prey items is not that important (Leon and Diez 1999; Leon and Bjorndal 2002). A comparison of hawksbill sea turtle diets between 2 types of habitats in Puerto Rico, the cliff wall of Monito Island and coral reef areas of Mona Island, showed that the cliff wall was a nutritionally more favorable habitat than the reef areas (Diez and van Dam 2002). Turtles feed on a narrow range of sponges in the Monito Island and use a broader assortment of sponges in Mona (van Dam and Diez 1997). Turtles utilizing the walls of Monito Island spent less effort foraging and exhibited more diet specialization, suggesting that the sponge Geodia neptuni, their main prey item, appeared to be very abundant in that habitat and possibly explaining the high density and growth rates of hawksbill turtles in this site (van Dam and Diez 1997; Diez and van Dam 2002). The other study shows no concordance between hawksbill and food abundance (Leon and Diez 1999; Leon and Bjorndal 2002). Hawksbills were more abundant at the Cabo Rojo site than at Bahia de las Águilas (Leon and Diez 1999) despite the fact that the abundances of C. caribensis and Myriastra kalitetilla, the 2 most preferred items in the diet, were higher at the Bahia de las Águilas site (Leon and Bjorndal 2002).

Our data reject the hypothesis that spatial variability in the abundance of juvenile hawksbill turtles is related to the availability of preferred prey species. In our study, the site with the highest abundance of food resources, particularly the sponge C. caribensis (PS), showed the lowest hawksbill CPUE among study sites. Even though C. nucula has been established as one of the most nutritious prey items for hawksbills in the Caribbean because of its high energy and nutrient content and low spicule content (Leon and Bjorndal 2002), turtles showed a low preference for this sponge in the Culebra Archipelago, suggesting that this is not a limiting resource (Rincón-Diaz et al. 2011) and explaining partially why turtle abundance is not responding to its availability. Similarly, R. florida, the most preferred item by hawksbills in the Culebra Archipelago, was very abundant in CL, the site with the second-lowest turtle CPUE. Our results show that food availability alone, even of preferred items, does not explain much of the variability in spatial turtle abundance.

Turtle Abundance and Habitat Characteristics

Our results of benthic features suggest that high cover of gorgonians and stony corals are important structural features associated to the abundance of juvenile hawksbills. Structure of benthic bottoms differed among sites with the exception of LPM and PS sites as the global and pairwise ANOSIM analysis showed. The MDS analysis showed CR transects as part of the fourth cluster, and this clustering was explained because of its high cover of gorgonians and live and dead corals. The fact that CR has the highest turtle abundance among study sites suggests that high cover of stony corals and gorgonians makes up preferred turtle habitats.

The presence of coral reef areas composed of stony corals and gorgonians is used as an indicator to identify in-water habitats for hawksbill sea turtles (Diez and Ottenwalder 2000), and these areas are associated as feeding and shelter grounds (van Dam and Diez 1998; Blumenthal et al. 2008, 2009; Hunt 2009). We suggest that complex structures such as dead and live corals are important because they may offer refuges and assisted resting sites for juvenile turtles, which were observed during our turtle surveys hiding in little caves or under heads of corals in reefs (Rincon-Diaz and Diez, pers. obs.). In addition to the physical structure of the benthos, the sessile organisms covering these surfaces are important in determining how changes in benthic cover can affect the abundance and distribution of turtles. Reefs of CR and LPM are part of the no-take zone of the Luis Pena Channel and have experienced a dramatic decline in coral cover and species richness because of bleaching events in the last 13 years (Hernandez-Delgado and Sabat 2003; Hernandez-Pacheco et al. 2011). It is not clear how changes in coral cover and habitat availability, as a result of climate change, are going to impact turtle abundance and distribution. Blumenthal et al. (2008) pointed out that hawksbills in the Cayman Islands show a partitioning of vertical habitat by turtle size; this fact can balance competition of animals in degraded shallow areas, suggesting that deep habitats need to be protected. The finding of Blumenthal et al. (2008) and the low live coral cover observed in our study sites underscore the need to evaluate 1) the effects benthic community shifts on the abundance and feeding ecology of sea turtles in inshore habitats and 2) the real alternatives to conserve and protect turtles and their habitats in view of current scenarios of reductions in coral cover of inshore areas.