Islands around the world support assemblages of endemic land snails (Chiba & Cowie, Reference Chiba and Cowie2016), many of which have been driven to near extinction and rely on intensive management and captive breeding for their survival (Coote & Loève, Reference Coote and Loève2003; Outerbridge et al., Reference Outerbridge, Ovaska and Garcia2019). Poecilozonites is a genus of land snails endemic to Bermuda, with two extant species (Outerbridge & Sarkis, Reference Outerbridge and Sarkis2018). Both are categorized on the IUCN Red List as Critically Endangered (Ovaska & Outerbridge, Reference Ovaska and Outerbridge2019a,Reference Ovaska and Outerbridgeb). The greater Bermuda land snail Poecilozonites bermudensis (Plate 1) was believed extinct by the 1990s as a result of the introduction of invertebrate predators, primarily the rosy wolf snail Euglandina rosea and the flatworm Bipalium vagum (Ovaska & Outerbridge, Reference Ovaska and Outerbridge2019a). In 2014 a subpopulation was rediscovered in an urban alley and ex situ breeding was started (Outerbridge et al., Reference Outerbridge, Ovaska and Garcia2019). A second subpopulation was found in 2017 on Port's Island, a 6.7 ha island unconnected to the rest of the archipelago (Fig. 1). An unmanaged, invasive-dominated, secondary growth forest covers 83% of Port's Island, and the remainder is a camping ground. The rediscovery of this wild subpopulation offered the opportunity to study the habitat preferences of the species, to aid identification of suitable release sites for captive-bred snails and to guide searches for additional wild subpopulations.

Fig. 1 Port's Island in the Bermuda archipelago, indicating locations surveyed for Poecilozonites bermudensis.

Plate 1 Two adult-sized greater Bermuda land snails Poecilozonites bermudensis on Port's Island, Bermuda (Fig. 1).

Observations of zoo-reared P. bermudensis released in 2016 indicated an affinity for limestone (Outerbridge et al., Reference Outerbridge, Ovaska and Garcia2019). Chiba (Reference Chiba2010) suggested that invasion by Casuarina species can reduce densities of native snails on islands under certain conditions. Casuarina equisetifolia, native to South-east Asia, northern Australia and Pacific islands, is invasive on Bermuda's coasts, and previous anecdotal observations suggest P. bermudensis has an aversion to its leaf litter. In this study on Port's Island, we investigated whether P. bermudensis was more abundant around limestone features and less abundant where C. equisetifolia was present. Additionally, we expected P. bermudensis abundance to be higher at sites with greater plant species diversity, and for snail counts to increase with distance from the coast and altitude.

To test our hypotheses we used two approaches to identify sites: (1) random sites were chosen to assess differences in snail abundance and plant diversity at various altitudes and distances from the sea, and (2) targeted surveys to assess if the snails preferred limestone features and avoided C. equisetifolia trees.

The contours of Port's Island were used to stratify 30 random sites by randomly selecting locations along the contour lines using ArcGIS 10.2.1 (Esri, Redlands, USA), ensuring sites were chosen from sea level to the peak of the island at 24 m altitude (Fig. 1). Targeted surveys at 10 C. equisetifolia trees and 10 limestone features were conducted opportunistically when we encountered appropriate features. Limestone features included natural outcrops, ledges, overhangs and rubble piles. An effort was made to select features from different parts of Port's Island (Fig. 1). A paired control site for each feature was selected by moving 3 m away from the feature in a random direction. Control sites were a similar distance from the sea and within the same plant community, but without rocks, C. equisetifolia trees or large amounts of C. equisetifolia litter.

Sites were surveyed during 2 May–14 June 2018. Nested quadrat surveys were conducted at the 30 random sites, using a 0.5 × 0.5 m (0.25 m²) quadrat for snail counts within a 2 × 2 m (4 m²) quadrat for vegetation. In each 4 m² quadrat, plants were identified to species and counted, except for species of Calophyllum, which were only identified to genus. Each 0.25 m² quadrat was searched down to the soil surface, with leaves, rocks and twigs checked for snails. Live and dead P. bermudensis were counted and photographed. Since the invasive-dominated plant assemblage is relatively recent, and the differential preservation of shells in microhabitats is not understood, only live snails were used in data analysis. As it was the dry season, aestivating snails were only dislodged if obscuring others. An SM200 soil moisture sensor with HH2 moisture meter (Delta-T Devices, Cambridge, UK) was used at the centre of the 0.25 m² quadrat once it was cleared. The altitude and distance to the nearest coast were later determined with ArcGIS. Targeted surveys at C. equisetifolia trees, limestone features and paired control sites used the same sized quadrat, soil sensor and methods as for the random sites, but no vegetation survey was performed.

The live snail counts in the 30 random quadrats were analysed with a generalized linear model using the quasi Poisson error distribution, with plant species richness, distance to coast, altitude and soil moisture as predictor variables. Non-significant terms were dropped from the model using the backward elimination method, with probability values estimated from the type II Wald χ ² test. The experimental paired design was used to analyse the response of the number of live snails to the presence of C. equisetifolia, limestone or neither habitat feature (control site). The non-parametric Mann–Whitney test, the paired t test or the paired non-parametric Wilcoxon signed-rank test were used depending on whether the data was normally distributed or not. All tests were conducted in R 3.5.1 (R Core Team, 2018).

On Port's Island, 558 live and 1,179 dead P. bermudensis were found at 70 sites. Our linear model found no significant effect of plant diversity on the number of live snails (χ ² = 0.51, df = 1, P = 0.48), suggesting that plant diversity is not an important predictor of snail abundance on Port's Island. Similarly, we found no effect of soil moisture (χ ² = 0.66, df = 1, P = 0.42), elevation (χ ² = 2.92, df = 1, P = 0.09) or distance from the coast (χ ² = 2.45, df = 1, P = 0.12). Poecilozonites bermudensis were found in both coastal habitat and interior woodlands. The hypothesis that vegetation diversity contributes to higher abundance of snails inland (as fewer plant species tolerate the salt exposure at the coast) can be rejected. Furthermore, a Spearman's rank correlation of the number of plant species in the 30 quadrats against their distance from the coast showed no significant correlation (r = 0.17, df = 30, P = 0.36). As most of the island is covered by invasive plants, this is unsurprising. The number of individual plants in the 4 m² quadrats was 9–105 (mean 31), with a total of 27 species.

If P. bermudensis is differentially drawn to or repelled by limestone and C. equisetifolia, differences in live snail counts would be expected between the two sets of 10 quadrats; a Mann–Whitney test demonstrated a significant difference in medians between them (U = 17.17, N = 20, P = 0.006; Fig. 2). A Wilcoxon matched pairs test showed the number of live P. bermudensis found on limestone was significantly greater than at the control sites (V = 2.10, N = 20, P = 0.025), and the number was significantly less beneath C. equisetifolia trees than at the control sites (t = −4.04, df = 9, P = 0.001).

Fig. 2 Mean ± SE number of live P. bermudensis found in 10 targeted quadrats and 10 paired control quadrats in areas with Casuarina equisetifolia and with limestone.

Our findings suggest that P. bermudensis is not more abundant in places with higher plant diversity and at sites away from the coast, but is more abundant on limestone and less abundant where the invasive tree C. equisetifolia is present. The dry season timing of this survey probably contributed to the strength of the association with limestone, as snails were using overhangs and damp crevices as refugia. Snails concentrated around limestone were possibly seeking calcium for their egg shells. In breeding tanks, large P. bermudensis have been observed eating limestone (R. Marirea, pers. comm., 2018). The breeding season of P. bermudensis in the wild has not been documented, but we found eggs on Port's Island on 4 May, shallowly buried in the soil.

Availability of suitable, protected habitat is critical to the success of snail reintroductions (Coote & Loève, Reference Coote and Loève2003). Our finding that P. bermudensis is common at random sites throughout the invasive-dominated woodland of Port's Island is encouraging, given that all of Bermuda's protected areas are now dominated by invasive plant species. If P. bermudensis does not associate strongly with native plants and therefore requires a managed habitat; the number of potential reintroduction locations is greater. The use of specific indigenous and invasive plant species by P. bermudensis for food and shelter at Port's Island requires further study. Coastal sites above the storm surge zone can be considered as potential reintroduction sites if they are not dominated by C. equisetifolia. At least one shaded limestone feature should be present, preferably a stacked stone wall, rubble pile, or similar rough surfaced feature, to provide refuge in hot, dry weather.

Searches for other wild subpopulations of P. bermudensis are called for in the species recovery plan (Outerbridge & Sarkis, Reference Outerbridge and Sarkis2018). As P. bermudensis uses habitats close to the sea, low elevation islands and coastal sites should be included in future searches.