Regular articles
Special Issues

Caribbean Naturalist
    CANA Home
    Range and Scope
    Board of Editors
    Editorial Workflow
    Publication Charges

Other EH Journals
    Northeastern Naturalist
    Southeastern Naturalist
    Urban Naturalist
    Eastern Paleontologist
    Eastern Biologist
    Journal of the North Atlantic

EH Natural History Home

Assessing Effects of Feral Sheep on Plant Composition and Structure in a Caribbean Tropical Dry Forest
Rachel Granberg, Lori E. Brown, Jessica L. East, Alexis L. Garcia, Alixandra J. Godar, Maria F. Mejia, Clint W. Boal, and Gad Perry

Caribbean Naturalist, No. 33

Download full-text pdf (Accessible only to subscribers. To subscribe click here.)


Site by Bennett Web & Design Co.
Caribbean Naturalist 1 R. Granberg, et al. 22001166 CARIBBEAN NATURALIST No. 33N:1o–. 1383 Assessing Effects of Feral Sheep on Plant Composition and Structure in a Caribbean Tropical Dry Forest Rachel Granberg1,*, Lori E. Brown1, Jessica L. East1, Alexis L. Garcia1, Alixandra J. Godar1, Maria F. Mejia1, Clint W. Boal2, and Gad Perry1 Abstract - Invasive mammalian herbivores impact ecosystems worldwide, often with poorly documented and understood effects. To determine the effect of feral Ovis aries (Sheep) on vegetative structure and composition in the British Virgin Islands, we conducted stratified random vegetation surveys island-wide on Guana Island, and compared vegetation between paired ungulate exclosures and control plots. Our results indicate that plant communities less than 0.5 m in height were significantly altered by Sheep density across the island and by abiotic factors. This was supported by data from experimental exclosures. Island-wide, plants of all height classes were also negatively impacted by Sheep density and select abiotic factors, but this was not supported by exclosure-plot results. Our findings suggest that Sheep herbivory negatively impacts young plants, potentially affecting plant recruitment, forest age structure, and endangered native wildlife species. Introduction Invasion by non-native species is a significant threat to biodiversity, particularly in island ecosystems (Millennium Ecosystem Assessment 2005). Caribbean islands are a biodiversity hotspot (Myers et al. 2000), and the tropical dry forests that occur there are vulnerable, specialized, and highly endangered (Gerhardt 1993, Janzen 1988, Sanchez-Azofeifa et al. 2005). The fragility of the Caribbean islands is attributed to soil erosion, hurricanes, and human disturbances (Vieira and Scariot 2006), all of which contribute to alterations in wildlife populations and species extinctions (Lugo et al. 1981). Invasive species are abundant and broadly distributed (e.g., Powell et al. 2011) throughout the Caribbean and are common contributors to native species declines (Clout and Veitch 2004). We examined possible relationships between feral Ovis aries L. (Sheep) presence, density, abiotic factors, and vegetation community structure and composition on Guana Island, British Virgin Islands, with specific concerns about the potential effects on the critically endangered Cyclura pinguis Barbour (Stout Iguana; Mitchell 1996). Sheep were introduced to Guana during the 1950s, eventually roaming freely and successfully colonizing the island in the absence of predators (Lazell 2005). Periodic control activities have at times limited the area inhabited by them and reduced their numbers. Feral Sheep appear to influence understory vegetation and the distribution of the Stout Iguana on Guana (Skipper 1Department of Natural Resources Management, Texas Tech University, Lubbock, TX 79409, USA. 2US Geological Survey, Texas Cooperative Fish and Wildlife Research Unit, Texas Tech University, Lubbock, TX 79409, USA. *Corresponding author - rachelmgranberg@ Manuscript Editor: Renata Platenberg Caribbean Naturalist R. Granberg, et al. 2016 No. 33 2 et al. 2013). In addition to dietary overlap of leaves and fruit between the 2 species (Mitchell 1999), trampling vegetation and crushing of iguana burrows or nest chambers by Sheep (Bradley and Gerber 2005) are also of concern. Cyclurids are an ecologically important genus as seed dispersers (Iverson 1985); they also improve germination of ingested seeds (Hartley et al. 2000). As the largest native herbivore where it occurs, the Stout Iguana undoubtedly plays an important role in maintaining tropical dry forests. Although historically occurring throughout the Puerto Rican Bank (Pregill 1981), Stout Iguana populations were restricted in geographic distribution to Anegada, as described in 1917 by Barbour, and were later re-introduced to Guana, Necker, Norman, and Little Thatch Islands (Perry and Gerber 2006). Initial decrease in range is attributed to climatic changes (Pregill 1981) and predation by humans (Lazell 2005). Stout Iguanas were re-introduced to Guana island by J. Lazell during the 1980s (Lazell 2002) by moving 7 individuals from Anegada Island. Translocation to suitable islands was (and remains) essential to conservation of the species (Lazell 2002). Though conservation efforts are being made on Anegada, the island has more habitat-degradation issues, larger human impact, and more invasive species to predate upon and compete with lizards than does Guana (Mitchell 1999). The Guana population grew to approximately 130 by 2002 (Perry and Mitchell 2003) and comprised over half of the worldwide population of the Stout Iguana in 1999 (Alberts 1999). Although Stout Iguanas on Guana are well established, research and anecdotal evidence suggest that feral Sheep limit the island-wide distribution of the species (Skipper et al. 2013). Considering that Guana is now a source for individuals to be translocated to new islands (Lazell 2002), increasing carrying capacity of the island is integral to the success of the species. With a focus on vegetation that iguanas as well as other organisms feed on, we tested the hypothesis that increased Sheep density will negatively affect vegetation abundance and structure. This research provides insight for direction of habitat management not only for Guana, but other islands where Stout Iguana occurs. Field-Site Description Guana Island is a ~300-ha privately owned island in the British Virgin Islands located on the eastern edge of the Greater Antilles Islands (Fig. 1). Rainfall is unimodal, averaging 921 mm annually, with peak precipitation occurring September through November. This pattern creates a distinct dry season, and vegetation is adapted to prolonged periods of drought (Lazell 2005). The island is largely (>90%) covered in subtropical dry forest and has a maximum elevation of 264 m (Lazell 1996). Guana is composed of Upper Cretaceous igneous extrusive volcanic rock over 70 million years old (Lazell 2005); soils are shallow, loamy inceptisols and are considered highly erodible (Rankin 2002). Guana has a steep topographic gradient, with severe ridges, narrow drainages, and a salt pond occupying much of the island’s flat, central area. Extensive research on Guana provides an unusually developed understanding of the biota (see Lazell 2005). Caribbean Naturalist 3 R. Granberg, et al. 2016 No. 33 Methods Vegetation surveys We conducted vegetation and fecal-pellet–count surveys in October 2014. We subdivided Guana Island into 4 quadrants representing different levels of perceived Sheep density across the island based on Skipper et al. (2013) (Fig. 2). Because of terrain constraints, we used an established trail system (Fig. 3) as a guide for site locations, surveying 20–22 randomly selected locations within each quadrant with a minimum of 10 m between sites (n = 84). At each site, we established a 20-m2 inventory plot, the center line positioned 2 m away from and parallel to the trail. The inventory survey consisted of a 2 m x 10 m belt-transect and ten 2-m gap-intercept surveys conducted in each plot; we set the gap-intercept lines perpendicular to the established belt-transect line at each meter mark. All plants rooted in the plot were identified to 1 of 5 pre-determined functional groups: grass, forb, vine, epiphyte, succulent, armed (i.e., spined or thorned) woody plant, and unarmed woody plant (Table 1). We omitted grasses and forbs for analysis due to scarcity of samples. Belt transect. Along a 10-m belt-transect line, we enumerated and categorized all plants by functional group (see above) and height class (less than 0.5 m, 0.5 m–1.0 m, 1.0 m–1.5 m, 1.5 m–2.0 m, and >2.0 m) to calculate stem density and approximate forest size structure (a proxy for age). Height classes were defined as plant height from the ground, binned by 0.5-m increments. We considered 2.0 m the maximum height accessible by Sheep. We visually estimated percent canopy cover to the nearest 10% at every meter along the 10-m belt-transect line to generate average canopy cover for each transect. In addition to vegetation data, we collected the following site information: slope, aspect, and position in the landscape relative to sea level (valley bottom, lower third, middle third, upper third, ridge) at origin of each transect line, as well Figure 1. Guana Island is located north of Tortola in the British Virgin Islands. Part of the greater Puerto Rico Bank, Guana is home to several species of interest, including the critically endangered Clyclura pinguis (Stout Iguana). Caribbean Naturalist R. Granberg, et al. 2016 No. 33 4 as litter depth (measured to the nearest tenth of a centimeter using calipers) at 2–3 equidistant locations along the transect. Gap intercept. We surveyed plants less than 0.5 m using a modified gap-intercept method (Herrick et al. 2005) to estimate foliar cover of the youngest, most at-riskto- herbivory plants. At 10 points along the belt-transect line, an observer placed a 2-m line perpendicular to the original transect. The observer measured to the nearest cm the distance each plant, identified to functional group, occupied the 2-m line. Ground cover was also measured to the nearest cm (bare soil, litter, rock, or other) for each gap-intercept survey. Relative Sheep density. Using fecal-pellet surveys, we created an index of relative Sheep density across the island. We also conducted pellet-group searches along the established trail system throughout Guana. For quantification purposes, we considered pellets located >5 cm from another group or pellet as a separate group. Using ArcGIS, we created a raster surface that summed pellet groups contained Figure 2. Guana Island was subdivided into regions of perceived Sheep density, as determined by Skipper et al. (2013). Due to high human concentration and lack of native habitat, we omitted the resort area from surveying. Caribbean Naturalist 5 R. Granberg, et al. 2016 No. 33 Figure 3. Surveyors used the established trail system (black lines) on Guana Island as a guide for vegetation surveys (grey filled circles). Additionally, fecalpellet surveys were conducted along this trail system. Table 1. Examples of plant species present on Guana Island, representing each functional group assessed. Functional group/species Common names Woody Plants Bursera simaruba (L.) Sarg. Gumbo-limbo, Copperwood, or Chaca Guapira fragrans (Dum. Cours.) Little Black Mampoo Eugenia spp. Armed Woody Plants Randia aculeate L. White Indigoberry Vines Passiflora suberosa L. Corkystem Passionflower Dolichandra unguis-cati (L.) Miers Cat’s Claw Creeper, Funnel Creeper, or Cat’s Claw Trumpet Succulents Opuntia repens Bello Roving Pricklypear Pilosocereus royenii (L.) Byles & G.D. Royen’s Tree Cactus, Dildo Cactus, and Pipe Organ Rowley [= Cephalocereus nobilis (Haw. Cactus Britton & Rose)] Epiphytes Tillandsia utriculata L. Spreading Airplant Caribbean Naturalist R. Granberg, et al. 2016 No. 33 6 within a 100-m2–pixel resolution over the entire island. We used this variable to determine correlation between Sheep-density, belt-transect, and gap-intercept results. Sheep-exclosure plots. We surveyed 5 pairs of fenced Sheep-exclusion plots and adjacent control plots. Each plot was 10 m2, and pairs were characterized by identical topographical covariates (slope, aspect, position in the landscape). Gapintercept and belt-transect surveys, as described above, were conducted at the center of each exclusion and control plot, parallel to the nearest trail. The exclosure plots had been fenced for approximately 16 years. Although Sheep have been unable to enter these areas, we assumed insects, birds, and iguanas had not been impeded from accessing the plots. Analysis For both gap-intersect and belt-transect plots located island-wide, we used principle component analysis (PCA; prcomp) in Program R ( to reduce dimensionality of our data. In accordance with the Kaiser criterion, components with an eigenvalue of >1.0 were retained for further analysis, with any factor loading of a coefficient >|0.3| considered important (Tabachnick and Fidell 2001). We then used multivariate analysis of covariance (MANCOVA) to evaluate differences in response of principal components to covariates in the dataset (James and McCulloch 1990), including Sheep density, aspect, slope, position within the landscape, and average litter depth. We followed with pairwise testing to determine which principle components responded to covariates. With belt-transect analysis, we used data for unarmed woody plants, armed woody plants, succulents, and vines, with each grouped by height categories to approximate age class. For gap-transect analysis, we evaluated the presence of total vegetation, canopy cover, unarmed woody plants, armed woody plants, succulents, vines, epiphytes and ground cover (litter, rock, bare soil, or other). We tested the relationship between our raster surface representing Sheep fecal-pellet surveys and pre-determined zones of Sheep density using a general linear model (glm; Program R) with a quasi-poisson distribution. Lastly, we compared Sheep-exclosure plots to respective control plots using one-sided paired t-tests, assessing the hypothesis that a greater density of vegetation will be found within the exclosures. Results Belt transect In terms of Sheep-density index, our 4 stratified sampling zones, delineated based on previous Sheep studies on the island, were upheld by our continuous raster surface of pellet density (P < 0.001). With belt-transect data, we conducted a PCA and retained components meeting the Kaiser criterion (eigenvalue > 1.0), giving us 6 of 16 components (Table 2) with a cumulative 62.9% of the total variance explained (Table 3). After performing a PCA with belt-transect data, we conducted a MANCOVA (n = 84) to evaluate variables that affect observed differences in vegetation (Table 4). We assessed independence, normality, and homogeneity of variance-covariance. Assumption of normality was satisfactory based upon histoCaribbean Naturalist 7 R. Granberg, et al. 2016 No. 33 gram plots and Q-Q plots of the 6 components retained. We rejected homogeneity of variance–covariance based on a Levene’s test (P = 0.005). However, because our sample sizes were roughly equal among treatments, we continued with a MANCOVA (Tabachnik and Fidell 2001). We evaluated differences in the 6 retained principle components based on Sheep density, aspect, slope, average litter depth, and position within the landscape and their interactions. We observed significant differences in principle components based on aspect (P = 0.04), the interaction of aspect and slope (P = 0.02), and the interaction of aspect and Sheep density (P = 0.02). We also observed a trend for the interaction between Sheep density and average litter depth, though it was not statistically significant (P = 0.09). Pairwise testing (Table 5) demonstrated that PC3 (understory vegetation) did not respond to any variables; PC1 (armed/unarmed), Table 2. In accordance with the Kaiser criterion, we retained 6 principle components for belt-transect data. Here, we outline retained belt-transect principle components and a description of the vegetation dynamics for each axis as determined by variable loadings. Principle component Axis description 1 Armed/unarmed plants 2 Succulent plants 3 Understory vegetation 4 Secondary understory vegetation 5 Mid-story vegetation 6 Tall vegetation Table 3. Belt-transect variable loadings for each retained component and variance explained by each component. We considered loading of a coefficient >|0.3| significant and have highlighted them with a *. Principal component axes Variables 1 2 3 4 5 6 Unarmed woody plants less than 0.5 m -0.27 -0.04 0.33* -0.13 0.14 -0.15 Armed woody plants less than 0.5 m -0.44* 0.23 0.15 0.13 0.15 0.20 Succulent plants less than 0.5 m 0.04 0.49* 0.01 -0.15 0.12 0.28 Vines less than 0.5 m -0.22 0.01 -0.56* -0.10 0.20 -0.03 Unarmed woody plants 0.5–1.0 m -0.32* -0.17 0.27 -0.30* -0.11 -0.27 Armed woody plants 0.5–1.0 m -0.34* 0.11 -0.42* 0.18 0.12 -0.03 Succulent plants 0.5–1.0 m 0.04 0.56* 0.08 -0.22 0.13 -0.05 Vines 0.5–1.0 m -0.16 -0.03 -0.44* -0.37 0.03 -0.21 Unarmed woody plants 1.0–1.5 m -0.39* 0.04 0.02 0.00 -0.28 -0.15 Armed woody plants 1.0–1.5 m -0.27 0.10 0.15 0.13 -0.42* 0.08 Succulent plants 1.0–1.5 m 0.07 0.33* -0.02 -0.37* -0.25 -0.07 Vines 1.0–1.5 m -0.13 -0.28 0.11 -0.58* 0.07 0.07 Unarmed woody plants >2 m -0.38* -0.04 0.15 0.28 0.36 -0.07 Armed woody plants >2 m -0.22 0.06 -0.12 0.01 -0.50* 0.46* Succulent plants >2 m 0.01 0.34* 0.17 -0.01 0.21 -0.23 Vines >2 m -0.07 -0.18 0.09 -0.22 0.33 0.65* Variance explained by components 0.15 0.13 0.11 0.09 0.08 0.07 Cumulative variance explained (%) 0.15 0.28 0.39 0.48 0.56 0.63 Caribbean Naturalist R. Granberg, et al. 2016 No. 33 8 PC2 (succulents), and PC5 (mid-story vegetation) responded to 1 interaction of variables each; and PC4 (secondary low vegetation) and PC6 (tall vegetation) responded significantly to 3 variables each. The interaction of aspect and slope was significant for PC1 (P = 0.03), PC4(P = 0.01), and PC6 (P = 0.05). The interaction of all variables was statistically significant for PC2 (P = 0.03) and PC4 (P = 0.01). All other variables were determined significant for 1 or fewer c omponents. Gap intercept We performed a PCA with the gap-transect data, keeping 4 of the 11 components (Table 6) for further analysis for a cumulative explanatory power of 68.2%. (Table 7). Following the PCA, we ran a MANCOVA (n = 84) with retained components (Table 8). Assumption of normality was supported based upon histogram plots and Q-Q plots of the 4 components retained. We rejected homogeneity of variance–covariance based on a Levene’s test (P = 0.005). However, because our sample sizes were roughly equal among treatments, we continued with a MANCOVA (Tabachnik and Fidell 2001). For the MANCOVA, we evaluated differences in our 4 principle components based on our index of Sheep density, aspect, slope, position within the landscape, average litter depth and their interactions. Sheep density (P < 0.001), position in the landscape (P < 0.001), aspect (P < 0.001), and average litter depth (P < 0.001) were significantly different between principle components. The interactions between Sheep density and position in the landscape (P < 0.001), position in the landscape and aspect (P < 0.001), position in the landscape and average litter depth (P less than 0.001), aspect and average litter depth (P = 0.002), aspect and slope (P = 0.002), average litter depth and slope (P = 0.04), and Sheep density, position in the landscape, and slope (P = 0.03) also differed significantly between principle components. Table 4. MANCOVA results for belt-transect surveys testing the effects of Sheep density and abiotic variables on the 6 retained principle components. Fewer variables and interactions were significant with this survey compared to the gap-intersect survey. Statistically significant results indicated with *. Variables df Pillai F P Sheep density 1 0.058 0.337 0.912 Aspect 7 1.116 1.241 0.163 Average litter depth 1 0.435 4.238 0.003* Slope 1 0.249 1.826 0.124 Sheep density + aspect 5 0.892 1.338 0.126 Sheep density + average litter depth 1 0.120 0.748 0.615 Aspect + average litter depth 6 1.114 1.445 0.058 Sheep density + slope 1 0.312 2.500 0.042* Aspect + slope 5 1.266 2.091 0.002* Average litter depth + slope 1 0.175 1.164 0.349 Sheep density + aspect + average litter depth 4 0.573 1.003 0.466 Sheep density + aspect + slope 4 0.615 1.090 0.362 Sheep density + average litter depth + slope 1 0.202 1.396 0.246 Aspect + average litter depth + slope 4 0.410 0.685 0.860 Sheep density + aspect + average litter depth + slope 3 0.964 2.762 0.001* Residuals 38 Caribbean Naturalist 9 R. Granberg, et al. 2016 No. 33 Table 5. ANOVA results for pairwise comparisons of the belt-transect surveys. Sheep density did not appear to have an effect on belt-transect results. Statistically significant results indicated with *. PC1 PC2 PC3 PC4 PC5 PC6 Variables F P F P F P F P F P F P Sheep density 1.40 0.24 0.18 0.67 0.05 0.83 0.37 0.54 0.06 0.81 0.02 0.88 Aspect 2.15 0.06 0.80 0.59 0.53 0.81 1.29 0.28 1.48 0.20 1.14 0.36 Average litter depth 2.92 0.10 2.42 0.13 0.00 0.96 6.97* 0.01* 0.38 0.54 1.38 0.25 Slope 1.06 0.31 0.37 0.54 0.92 0.34 3.09 0.09 0.51 0.48 5.80* 0.02* Sheep density + aspect 1.34 0.27 0.68 0.64 0.17 0.97 0.69 0.64 1.12 0.36 5.97* less than 0.001* Sheep density + average litter depth 0.65 0.43 0.34 0.56 0.01 0.92 1.18 0.28 0.39 0.53 1.22 0.28 Aspect + average litter depth 1.81 0.12 1.08 0.39 0.73 0.63 0.90 0.50 2.79* 0.02* 1.11 0.38 Sheep density + slope 1.84 0.18 0.26 0.62 0.28 0.60 1.91 0.17 3.79 0.06 3.22 0.08 Aspect + slope 2.81* 0.03* 1.88 0.12 0.25 0.94 3.29* 0.01* 1.46 0.23 2.44* 0.05* Average litter depth + slope 0.56 0.46 0.13 0.72 0.02 0.88 2.65 0.11 1.88 0.18 0.29 0.60 Sheep density + aspect + average litter depth 1.59 0.20 0.72 0.58 0.55 0.70 0.99 0.42 0.44 0.78 1.37 0.26 Sheep density + aspect + slope 0.97 0.43 0.72 0.58 0.88 0.49 1.28 0.30 1.83 0.14 1.20 0.33 Sheep density + average litter depth + slope 1.08 0.30 0.14 0.71 0.02 0.88 2.86 0.10 0.58 0.45 0.72 0.40 Aspect + average litter depth + slope 0.95 0.45 1.37 0.26 0.38 0.82 0.88 0.48 0.31 0.87 0.15 0.96 Sheep density + aspect + average litter depth + slope 1.54 0.22 3.34* 0.03* 0.21 0.89 4.41* 0.01* 2.42 0.08 1.48 0.23 Caribbean Naturalist R. Granberg, et al. 2016 No. 33 10 Pairwise testing (Table 9) revealed that only PC2 (a ground-surface axis) was significantly different (P < 0.001) than other components based on Sheep density. However, both PC1 (canopy dynamics) and PC4 (browse-line dynamics) demonstrated trends (P = 0.08) as different based on Sheep density. Overall, PC1 and PC2 responded significantly to more variables than PC3 and PC4. Additionally, PC1 responded strongly (P < 0.001) to the abiotic variables position in the landscape and aspect as well as combinations of these factors with average litter depth and slope. PC2 responded strongly (P < 0.001) to all variables except slope (Sheep density, position in the landscape, aspect, average litter depth) individually; slope only seemed to play an important role when combined with aspect or average litter depth (P = 0.01). PC3 responded to average litter depth when combined with aspect (P = 0.03) and slope (P = 0.01) and only responded to position in the landscape (P = 0.02) when considering individual abiotic variables. Position in the landscape was important for PC4, with the variable considered individually responding strongly (P < 0.001) as well as when combined with Sheep density (P < 0.001) and average litter depth (P = 0.04). Aspect also proved an important variable (P = 0.02) for PC4. Table 7. Gap-intercept (plants less than 0.5 m) variable loadings for each retained component and variance explained by each component. We considered loading of a coefficient > |0.3| significant; these are highlighted with a *. Principal component axes Variables 1 2 3 4 Functional group -0.44* 0.01 -0.19 0.05 Total vegetative cover -0.51* 0.01 -0.14 -0.10 Canopy cover 0.31* 0.17 -0.34* 0.03 Litter -0.05 0.69* 0.12 -0.12 Rock 0.19 -0.57* -0.23 -0.20 Bare soil -0.24 -0.37* 0.21 0.45* Unarmed woody plants -0.08 0.01 -0.58* -0.44* Armed woody plants -0.05 -0.17 0.07 -0.46* Succulent plants -0.35* -0.06 0.31* -0.33* Vines -0.20 0.07 -0.53* 0.46* Epiphytes -0.43* 0.02 -0.02 -0.06 Variance explained by components 0.27 0.18 0.13 0.10 Cumulative variance explained (%) 0.27 0.45 0.58 0.68 Table 6. Retained gap-intercept (plants less than 0.5 m) principle components with a description of the role each axis plays in vegetative relationships, determined by variable loadings. We retained 4 principle components in accordance with the Kaiser criterion. Principle component Axis description 1 Canopy dynamics 2 Ground surface 3 Light availability 4 Browse line dynamics Caribbean Naturalist 11 R. Granberg, et al. 2016 No. 33 Sheep-exclosure plots We found significant differences in density of vegetation less than 0.5 m and no differences in density of belt-transect vegetation (n = 5) between exclosure treatment and control plots. With gap intersect, we found significantly lower density of overall vegetation (P = 0.03) in control plots versus plots protected from Sheep grazing. We also found a strong trend (P = 0.06) for density of unarmed woody plants in control plots. We did not find significant differences in litter depth or density of vines or succulents. For belt transect data, we did not find any significant differences in plant density or structure (armed and unarmed woody plants, vines, and succulents) between control and exclosure plots. Discussion Biotic factors We found that higher Sheep densities were correlated with reduced amounts of vegetation present, most notably at low plant heights that are easily accessible to Sheep. Gap-intercept principle components, which assessed vegetation less than 0.5 m, were responsive to our biotic variable of Sheep density for island-wide surveys as well as our paired exclosure/control surveys. Belt-transect principle components, measuring plants of all height classes, were not responsive to the index of Sheep Table 8. MANCOVA results for gap intersect surveys testing the effects of Sheep density and abiotic variables on the 4 retained principle components. Multiple variables as well as interactions were significantly different based on the MANCOVA. Statistically significant results indicated with *. Variables df Pillai F P Sheep Density 1 0.884 37.950 less than 0.001* Position 3 2.179 14.590 less than 0.001* Aspect 7 2.294 4.420 less than 0.001* Average litter depth 1 0.998 2063.800 less than 0.001* Slope 1 0.203 1.280 0.312 Sheep density + position 2 0.927 4.530 less than 0.001* Sheep density + aspect 5 1.042 1.620 0.064 Position + aspect 6 1.787 3.100 less than 0.001* Sheep density + average litter depth 1 0.162 0.960 0.449 Position + average litter depth 2 1.139 6.940 less than 0.001* Aspect + average litter depth 6 1.489 2.270 0.003* Sheep density + slope 1 0.217 1.380 0.275 Position + slope 1 0.127 0.730 0.584 Aspect + slope 5 1.413 2.510 0.002* Average litter depth + slope 1 0.383 3.110 0.038* Sheep density + position + aspect 3 0.493 1.080 0.389 Sheep density + position + average litter depth 1 0.066 0.360 0.837 Sheep density + aspect + average litter depth 4 0.862 1.580 0.091 Position + aspect + average litter depth 3 0.524 1.160 0.328 Sheep density + position + slope 1 0.409 3.460 0.027* Sheep density + aspect + slope 1 0.043 0.230 0.921 Aspect + average litter depth + slope 4 0.901 1.670 0.067 Residuals 23 Caribbean Naturalist R. Granberg, et al. 2016 No. 33 12 Table 9. ANOVA results for pairwise comparisons of the gap-intercept surveys. Axes 1, 2, and 4 are all significant or near-significant in response to Sheep density, indicating younger plants are impacted by Sheep presence. Sta tistically significant results indicated with *. PC1 PC2 PC3 PC4 Variables F P F P F P F P Sheep density 3.41 0.08 83.56* less than 0.001* 0.84 0.37 3.26 0.08 Position 21.05* less than 0.001* 182.22* less than 0.001* 4.26* 0.02* 9.65* less than 0.001* Aspect 12.52* less than 0.001* 83.91* less than 0.001* 0.99 0.46 3.03* 0.02* Average litter depth 0.29 0.59 3316.90* less than 0.001* 1.90 0.18 2.62 0.12 Slope 2.79 0.11 1.35 0.26 0.38 0.55 0.25 0.62 Sheep density, position 3.13 0.06 20.98* less than 0.001* 1.58 0.23 11.97* less than 0.001* Sheep density, aspect 1.84 0.14 4.65* 0.01* 0.98 0.45 0.86 0.52 Position, aspect 5.00* 0.01* 6.99* less than 0.001* 1.49 0.22 2.38 0.06 Sheep density, average litter depth 0.23 0.64 1.15 0.30 0.13 0.72 0.70 0.41 Position, average litter depth 5.61* 0.01* 16.57* less than 0.001* 0.01 0.99 3.68* 0.04* Aspect, average litter depth 8.87* less than 0.001* 2.06* 0.10* 2.88* 0.03* 2.01 0.11 Sheep density, slope 0.29 0.59 0.05 0.82 0.15 0.70 0.87 0.36 Position, slope 0.19 0.67 2.23 0.15 0.01 0.93 1.13 0.30 Aspect, slope 3.48* 0.02* 4.61* 0.01* 1.72 0.17 1.24 0.32 Average litter depth, slope 5.91* 0.02* 10.11* 0.01* 11.58* 0.01* 0.15 0.71 Sheep density, position, aspect 1.79 0.18 2.66 0.07 2.18 0.12 0.55 0.65 Sheep density, position, average litter depth 0.49 0.49 0.95 0.34 0.29 0.60 0.02 0.89 Sheep density, aspect, average litter depth 3.51* 0.02* 5.84* 0.01* 1.42 0.26 0.36 0.84 Position, aspect, average litter depth 2.82 0.06 0.35 0.79 1.16 0.35 1.22 0.33 Sheep density, position, slope 8.65* 0.01* 2.25 0.15 0.09 0.76 0.04 0.84 Sheep density, aspect, slope 0.41 0.53 0.05 0.82 0.00 0.98 0.02 0.89 Aspect, average litter depth, slope 10.86* less than 0.001* 0.53 0.72 0.52 0.72 0.08 0.99 Caribbean Naturalist 13 R. Granberg, et al. 2016 No. 33 density alone in island-wide or exclosure/control surveys. When considering the interaction of aspect and Sheep density, results for belt-transect principle components were significant for island-wide surveys, indicating that Sheep may decrease vegetation to a greater degree on more exposed aspects and also raises questions regarding biological relevance of this survey, discussed later in this section. In terms of direct effects of ungulate browsing, many of the plant species present in the Caribbean are more sensitive to mammalian herbivory because these herbivores were absent from the ecosystem in recent history (Bowen and Van Vuren 1997, Coley and Barone 1996). Several studies found that feral Sheep have a strong preference for consuming endemic island species over mainland plant species (Bowen and Van Vuren 1997, Van Vuren and Coblentz 1987). Furthermore, rates of mammalian herbivory are up to 25 times higher on young, more-nutritious plants compared to mature leaves (Coley and Barone 1996), which may alter plant community composition and age structure through reduced vigor or induced mortality. Given the fruit- and leaf-based diet of the Stout Iguana, it is clear that feral Sheep may negatively impact available forage via direct competition. Chemical deterrents may have a greater effect on selective browsing than armature. The size and scope of our project circumvented species-specific collection of data and characterization of palatable woody plants. Our results may reflect this, specifically the lack of significance of our proxy for Sheep density with belt-transect surveys. The larger-scale lens with which we approached vegetation identification perhaps failed to determine species-specific effects on more-palatable versus less-palatable plant species. In addition to loss of vegetation biomass due to mammalian herbivores, indirect effects of non-native herbivory are plentiful. Concentrated presence of ungulates, such as feral Sheep along trail systems, leads to increased soil erosion (Cumming and Cumming 2003, Sharrow 2007), soil compaction (Basset et al. 2005, Heckel et al. 2010), nutrient availability reduction (Bradshaw 1969), and reduction in seed germination and plant recruitment (Basset et al. 2005). Effects may be synergistic as well; a reduction in litter accumulation accompanies reduction in vegetation (Sharrow 2007), resulting in increased erosion and reduction in soil-nutrient availability (Bradshaw 1969). Furthermore, there is a clear risk in areas of high Sheep density for trampling of iguana forage as well and crushing of burrows or nest chambers (Bradley and Gerber 2005). Belt transects conducted in exclosure/control plots did not demonstrate differences in the amount of vegetation present, indicating that older age classes of vegetation may not be as heavily influenced by Sheep herbivory as the youngest classes. Older plants may have stronger defenses against mammalian herbivory than younger plants or may simply be out of reach for Sheep to browse. It is also possible surveys of transects 2 m wide by 10 m long did not provide ecological relevance for surveying older, more widely spaced plants. Given the slow rate of growth and recruitment in tropical dry forests (Murphy and Lugo 1986), longerterm and larger-scale data sets may be required to establish relationships between older age classes of plants and Sheep herbivory. Caribbean Naturalist R. Granberg, et al. 2016 No. 33 14 Abiotic factors Overarching limiting factors of light availability, soil nutrition, and soil moisture affect tropical dry forests differently than more-often–studied tropical moist forests (Ceccon et al. 2006). Light availability, which is a major component in plant recruitment within tropical moist forests, does not have as distinct effects in tropical dry forests due to more open and shorter canopy structure (Brienen et al. 2009, Lebrija-Trejos et al. 2010, Swaine et al. 1990, Wright 2002). The effect of nutrient availability is not extensively studied in the tropical dry forest system; however, there is evidence to suggest that nutrition is a not limiting factor for growth (Murphy and Lugo 1986; but see Marrs et al. 1991). High variability in rainfall amount, length of drought, and precipitation-dependent nutrient availability means that tropical dry forests experience more environmental stress than their mesic counterparts in terms of available soil moisture (Bradshaw 1969, Ceccon et al. 2006). For these reasons, this research focused on easily measured abiotic covariates that influenced available soil moisture. Abiotic covariates measured were: slope, aspect and position in the landscape relative to sea level, and average litter depth. With the exception of slope, the youngest age class of plants responded to all abiotic covariates included in analysis as well as interactions between all abiotic covariates. Taller plants assessed in belttransect surveys responded to abiotic variables, but not to the extent that smaller, younger plants responded. Lack of significance of Sheep density and reduced number of significant abiotic covariates with older plants may suggest ontogenetic shifts in growth-limiting factors and causes of plant mortality (Alvarex-Clare and Kitajima 2009, Auspburger 1983) or reflect a lack of ecological relevance for belt-transect surveys with respect to older age classes. Although 2-m–wide belt transects are common, transect length (10 m) was short relative to other forestry studies. Widening or lengthening transects may have altered final results for older age classes; however, research has demonstrated that a shift in factors limiting recruitment is common in forest systems (Clark et al. 1999). Conclusions We observed a reduction in low-growing vegetation in areas where Sheep were most abundant. Smaller vegetation may be more sensitive to herbivory, whereas older, taller vegetation may have more efficient defenses against herbivory and may be out of reach for Sheep. Our results indicate that Sheep herbivory adversely impacts the youngest vegetation on Guana Island, with potential negative effects on seedling recruitment, forest age structure, and native wildlife. Of particular concern is the reduction of low-growing vegetation and its effect on the forage availability of the endangered Stout Iguana, both immediately and in the long-term. Future research on this topic would benefit from species-specific identification of plants, particularly those with known palatability. Furthermore, larger-scale surveys for the tallest plant classes would ensure ecological relevance of the study. Lastly, studies of diet and foraging ecology of co-occuring feral Sheep and Stout Iguana may elucidate much information directly applicable to management of Caribbean Naturalist 15 R. Granberg, et al. 2016 No. 33 Cyclurids on islands with invasive ungulates—a common conservation issue with this genera (Alberts 1999). Ideally, this research would be part of a before–after control study of feral Sheep removal. Given the immediacy of protecting a critically endangered species and demonstrated beneficial effects of removing invasive mammals from island ecosystems (Jones et al. 2016), our results call for action. Goats were successfully eliminated from Guana in 1991; however, there have been multiple failed attempts at eradicating Sheep (Lazell 1996, 1997). Currently, feral Sheep are left unmanaged. Guana Island is a source for Stout Iguana translocation candidates (Lazell 1997); effort should be made to increase the carrying capacity of the island, and Sheep eradication is part of this process. Ortiz-Alcaraz et al. (2016) reported a detailed description of the successful Sheep-eradication program on Soccoro Island, Mexico. Though cover type and use of the Guana as a high-end resort preclude the use of aerial hunting techniques, use of GPS-collared Judas Sheep, terrestrial hunting, leg-hold traps, and hound hunting are all realistic and efficient dispatch methods. With adequate monitoring and data collection, this process could provide a wealth of information in the realm of island ecology informing future management decisions regarding Cyclurid conservation in the Caribbean. Several other species of concern, including Phoenicopterus roseus Pallas (Greater Flamingo), Columba leucocephala (L.) (White-crowned Pigeon), and Geotrygon mystacea (Temminck) (Bridled Quail-dove), may also benefit from invasive Sheep control on Guana Island. Although longer-term and larger-scale exclusion studies may develop a greater understanding of the extent of feral ungulate impact on Guana Island, these results justify a reevaluation of current management practices. Acknowledgments We thank M. Vanlandeghem and R. Rondeau for providing critical insight on various aspects of the project. Funding was provided by The Conservation Agency through a grant from the Falconwood Foundation. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the US Government. Literature Cited Alberts, A. 1999. West-Indian iguanas: Status survey and conservation action plan. IUCN/ SSC West Indian Iguana Specialist Group. IUCN, Gland, Switzerland and Cambridge, UK. 111 pp. Alvarex-Clare, S., and K. Kitajima. 2009. Susceptibility of tree seedlings to biotic and abiotic hazards in the understory of a moist tropical forest in Panama. Biotropica 41:47–56. Augsperger, C.K. 1983. Offspring recruitment around tropical trees: Change in cohort distance with time. Oikos 40:189–196. Barbour, T. 1917. Notes on the herpetology of the Virgin Islands. Proceedings of the Biological Society of Washington 30:97–104. Basset, I.E., R.C. Simcock, and N.D. Mitchell. 2005. Consequences of soil compaction for seedling establishment: Implications for natural regeneration and restoration. Austral Ecology 30:827–833. Bowen, L. and D. Van Vuren. 1997. Insular endemic plants lack defenses against herbivores. Conversation Biology 11:1249–1254. Caribbean Naturalist R. Granberg, et al. 2016 No. 33 16 Bradley, K.A., and G.P. Gerber. 2005. Conservation of the Anegada iguana (Cyclura pinguis). Iguana 12:78–85. Bradshaw, A. 1969. An ecologist’s viewpoint. Pp. 415–427, In I.H. Rorinson (Ed.). Ecological Aspects of the Mineral Nutrition of Plants. Blackwell Publishing, Oxford, UK. Brienen, R.J., P.A. Zuidema, and M. Martinez-Ramos. 2009. Attaining the canopy in dry and moist forests: Strong differences in tree-growth trajectories reflect variation in growing conditions. Oecologia 163:485–496. Ceccon, E., P. Huante, and E. Rincon. 2006. Abiotic factors influencing tropical dry forests regeneration. Brazilian Archives of Biology and Technology 49:305–312. Clark, J.S., B. Beckage, P. Camill. 1999. Interpreting recruitment limitation in forests. The American Journal of Botany 86:1–16. Clout, M.N., and C.R. Veitch. 2004. Turning the tide of biological invasion: The potential for eradicating invasive species. Pp. 1–3, In C.R. Veitch and M.N. Clout (Eds.). Turning the Tide: The Eradication of Invasive Species. Occasional paper of the IUCN Species Survival Commission No. 27, International Union for Conservation of Nature, Gland, Switzerland. Coley, P.D., and J.A. Barone. 1996. Herbivory and plant defenses in tropical forests. Annual Review of Ecology and Systematics 27:305–335. Cumming, D.H.M., and G.R. Cumming. 2003. Ungulate community structure and ecological processes: Body size, hoof area, and trampling in African savannas. Oecologia 134:560–568. Gerhardt, K. 1993. Tree seedling development in tropical dry abandoned pasture and secondary forest in Costa Rica. Journal of Vegetation Science 4:95–102. Hartley, L.M., R.E. Glor, A.L. Sproston, R. Powell, and J.S. Parmer-Lee. 2000. Germination rates of seeds consumed by two species of rock iguanas (Cyclura spp.) in the Dominican Republic. Caribbean Journal of Science 36:149–151. Heckel, C.D., N.A. Bourg, W.J. McShea, and S. Kalisz. 2010. Nonconsumptive effects of a generalist ungulate herbivore drive decline of unpalatable herbs. Ecology 91:319–326. Herrick, J.E., J.W. Van Zee, K.M. Haystad, L.M. Burkett, and W.G. Whitford. 2005. Monitoring Manual for Grassland, Shrubland, and Savanna Ecosystems, Volume II. The University of Arizona Press, Tucson, AZ, USA. 201 pp. Iverson, J.B. 1985. Lizards as seed dispersers? Journal of Herpetology 19:292–293. James, F.C., and C.E. McCulloch. 1990. Multivariate analysis in ecology and systematics: Panacea or Pandora’s box? Annual Review of Ecology and Systematics 21:129–166. Janzen, D.H. 1988. Tropical dry forests: The most endangered major tropical ecosystem. Pp. 130–137, In E.O. Wilson (Ed.). Biodiversity. National Academy Press, Washington, DC, USA. Jones, H.P., N.D. Holmes, S.H. Butchart, B.R. Tershy, P.J. Kappes, I. Corkery, A. Aguirre- Munoz, D.P. Armostrong, E. Bonnaud, A.A. Burbidge, K. Campbell, F. Courchamp, P.E. Cowan, R.J. Cuthbert, S. Ebbert, P. Genovesi, G.R. Howald, B.S. Keitt, S.W. Kress, C.M. Miskelly, S. Oppel, S. Poncet, M.J. Rauzon, G. Rocamora, J.C. Russel, A. Samaniego-Herrera, P.J. Seddon, D.R. Spatz, D.R. Towns, and D.A. Croll. 2016. Invasive-mammal eradication on islands results in substantial conservation gains. Proceedings of the National Academy of Sciences 113:4033–4038. Lazell, J. 1996. Guana Island: A natural history guide. The Conservation Agency Occasional Paper 1:1–20. Lazell, J. 1997. The Stout Iguana of the British Virgin Islands. Iguana Times 6:75–80. Caribbean Naturalist 17 R. Granberg, et al. 2016 No. 33 Lazell, J. 2002. Restoring vertebrate animals in the British Virgin Islands. Ecological Restoration 20:179–185. Lazell, J. 2005. Island Fact and Theory in Nature. University of California Press. Berkeley, CA, USA. 402 pp. Lebrija-Trejos, E., E.A. Perez-Garcia, J.A. Meave, F. Bongers, and L. Poorter. 2010. Functional traits and environmental filtering drive community assembly in a species-rich tropical system. Ecology 91:286–398. Lugo, A.E., R. Schmidt, and S. Brown. 1981. Tropical forests in the Caribbean. Ambio 10:318–324. Marrs, R.H., J. Thompson, D. Scott, and J. Proctor. 1991. Nitrogen minerlization and nitrification in Terre Firme Forest and savanna soils on Ilha de Maraca, Roraima, Brazil. Journal of Tropical Ecology 7:123–137. Millennium Ecosystem Assessment. 2005. Ecosystems and Human Well-being: Biodiversity Synthesis. World Resources Institute, Washington, DC, USA. 100 pp. Mitchell, N.C. 1996. Cyclura pinguis. The IUCN red list of threatened species 1996. Available online at Accessed 10 May 2016. Mitchell, N.C. 1999. Effect of introduced ungulates on density, dietary preferences, home range, and physical condition of the iguana (Cyclura pinguis) on Anegada. Herpetologica 55:7–17. Murphy, P.G., and A.E. Lugo. 1986. Ecology of tropical dry forest. Annual Review of Ecology and Systematics 17:67–88. Myers, N., R.A. Mittermeier, C.G. Mittermeier, G.A.B. da Fonseca, and J. Kent. 2000. Biodiversity hotspots for conservation priorities. Nature 403:853–858. Ortiz-Alcaraz, A., A. Aguirre-Munoz, F. Mendez-Sanchez, and A. Ortega-Rubio. 2016. Feral Sheep eradication at Socorro Island, Mexico: A mandatory step to ensure ecological restoration. Interciencia 41:184–189. Perry, G., and G.A. Gerber. 2006. Conservation of amphibians and reptiles in the British Virgin Islands: Status and patterns. Applied Herpetology 3:237–256. Perry, G., and N. Mitchell. 2003. Guana and Necker Island population assessments 2002. Iguana 10:49. Powell, R., R.W. Henderson, M.C. Farmer, M. Breuil, A.C. Echternacht, G. van Buurt, C.M. Romagosa, and G. Perry. 2011. Introduced amphibians and reptiles in the greater Caribbean: Patterns and conservation implications. Pp. 63–143, In A. Hailey, B.S. Wilson, and J.A. Horrocks (Eds.). Conservation of Caribbean Island Herpetofaunas Volume 1. Brill, Leiden, The Netherlands . Pregill, G. 1981. Late Pleistocene herpetofaunas from Puerto Rico. University of Kansas Museum of Natural History Publication 71:1–72. Rankin, D. 2002. Geology of St. John, US Virgin Islands. US Geological Survey Professional Paper 1631:1–30. Sanchez-Azofeifa, G.A., M. Kalacska, M. Quesada, J.C. Calvo-Alvarado, J.M. Nassar, and J.P. Rodriguez. 2005. Need for integrated research for a sustainable future in tropical dry forests. Conservation Biology 19:285–286. Sharrow, S.H. 2007. Soil compaction by grazing livestock in silvopastures as evidenced by changes in soil physical properties. Agroforestry Systems 71:215–223. Skipper, B., B. Grisham, M. Kalyvaki, K. McGaughey, K. Mougey, L. Navarette, R. Rondeau, C. Boal, and G. Perry. 2013. Non-overlapping distributions of feral Sheep (Ovis aries) and Stout Iguanas (Cychura pinguis) on Guana Island, British Virgin Islands. ICRF Reptiles and Amphibians 20:7–15. Caribbean Naturalist R. Granberg, et al. 2016 No. 33 18 Swaine, M.D., D. Lieberman, and J.B. Hall. 1990. Structure and dynamics of a tropical dry forest in Ghana. Vegetation 88:31–51. Tabachnick, B.G., and L.S. Fidell. 2001. Using Multivariate Statistics, 4th Edition. Allyn and Bacon, Needham Heights, MA, USA. 966 pp. Van Vuren, D., and B.E. Coblentz. 1987. Some ecological effects of feral Sheep on Santa Cruz Island, California, USA. Biological Conservation 41:253–268. Vieira, D.L., and A. Scariot. 2006. Principles of natural regeneration of tropical dry forests for restoration. Restoration Ecology 14:11–20. Wright, S.J. 2002. Plant diversity in tropical forests: A review of mechanisms of species coexistence. Oceologia 130:1–14.