Methods.

During July of 2012, we surveyed 14 loggerhead crawl sites on the southern-facing beaches of Horn Island, Mississippi. This island is 1 of 5 managed in Mississippi by the National Park Service's Gulf Islands National Seashore (hereafter GUIS). Historically, Horn Island was a military outpost. However, it is now undeveloped and uninhabited excluding a single housing facility for Park Rangers. Horn Island is approximately 13 miles in length. The beachfront is patrolled several times per week as part of routing Ranger activities. The island is situated in a region routinely impacted by large storms, heavy winds, and strong currents (Morton 2008), resulting in highly dynamic barrier island habitats (Claudino-Sales et al. 2008; Houser et al. 2008). Additionally, this area was heavily impacted by the Deepwater Horizon oil spill and was subjected to extensive beach cleanup activities. During our study there existed variation in beach topography along the coastline, with variable distance from the high tide line to the back dune, changes in slope, and the presence of small berms. The entire beachfront is sandy habitat and potentially suitable for nesting for the loggerhead, green (Chelonia mydas), and Kemp's ridley (Lepidochelys kempii) sea turtles that nest here.

Once a crawl site was located, we determined what species made the crawl and if it resulted in a nest. A crawl was determined to have resulted in a nest if there was evidence of active egg covering and site “camouflaging”. We did not check for, or disturb, eggs during this study. If the turtle returned to the water without laying, we designated the location as a false crawl site. At the locations of nests (n = 11) and false crawls (n = 3), we measured the slope and length of a line perpendicular to the high tide line (HTL). Slope, presented in this article as a ratio (in units of m · m⁻¹), was measured with standard topographical survey equipment consisting of a leveled tripod-mounted laser aimed at a measuring stick. Horizontal distance from HTL was recorded with a Trimble GeoExplorer GeoXT GPS (submeter accuracy). At nest sites, we started measurements from the estimated location of the clutch. At false crawl sites, we measured from the point of the crawl path farthest from the water. Crawls tended to be roughly parabolic, and at false crawls we measured slope and distance to HTL from the vertex of this parabola. Herein we do not focus on changes along the length of the crawl such as berms (see Wood and Bjorndal 2000) and rather examine the overall changes in distance travelled and elevation, comparing start point to endpoint.

For data analysis, we converted the GPS-derived horizontal distance to a straight-line crawl distance using basic trigonometry (Fig. 1). We then ran a linear regression using slope as the predictor variable and crawl distance as the response. We acknowledge concerns for the potential for autocorrelation between the predictor and response inherent in the geometric relationship between the two. Specifically, the response can be thought of as c in the Pythagorean Theorem a² + b² = c², and the predictor (slope) can be calculated using a/b. Thus, in a sense we are regressing √(a² + b²) and a/b. However, the biological underpinning of our hypotheses is that turtles are seeking some imagined elevation threshold, a. If this hypothesis is incorrect, then there would probably not be a significant relationship between slope and crawl distance. Further, if evidence for a target or threshold elevation is present, then we should expect no significant association between elevation and slope or elevation and crawl distance. We ran regressions to evaluate those expectations.

View Figure

• Download as Powerpoint
• Download Figure

To determine if there were morphological characteristics that differed between beach sites that were chosen for a nest versus actively rejected sites (false crawls), we analyzed topographical data using 3 Welch's t-tests to compare slope, crawl distance, and elevation between the two crawl types. Welch's t-test is appropriate in this case because it is robust to unequal sample sizes and variance but gives results very similar to the Student's t-test in the event that those conditions are met (Ruxton 2006).

Results.

Topographical surveys revealed that the loggerheads at our 14 sites crawled an average (± standard error) perpendicular distance of 13.2 ± 1.5 m from the high tide line toward the dune, ranging from 6.20 to 24.8 m. Mean slope for all sites was 0.092 ± 0.010 m · m⁻¹. The minimum slope observed was 0.027 m · m⁻¹, while the maximum was 0.16 m · m⁻¹. Mean final elevation at the crawl sites was 1.06 ± 0.075 m and ranged from 0.533 to 1.64 m.

The linear regression of slope predicting straight-line crawl distance showed a strong association between the predictor and response (R² = 0.62; Fig. 2). Slope was a statistically significant predictor of crawl distance in the linear model fit (p ≤ 0.05) and its estimated coefficient was negative (Table 1). Thus, as slope increased, crawl distance decreased.

View Figure

• Download as Powerpoint
• Download Figure

Table 1. Model fit results for the linear regression with beach slope (m · m⁻¹) as the predictor variable and loggerhead (Caretta caretta) crawl distance (m) as the response. Coefficient estimates are shown with associated standard error and p-values (R² = 0.62). Data were collected from topographical surveys of loggerhead crawls (n = 14) on Horn Island, Mississippi, in July 2012.

View Fullscreen

Linear regressions with elevation as the response and either slope or crawl distance as the respective predictors suggested no relationship exists between variables. Slope (R² = 0.15, p = 0.18) and crawl distance (R² = 0.02, p = 0.64) were not statistically significant predictors of elevation and explained little of the variance in elevation.

Results comparing nest crawls to false crawls showed that elevation did not differ significantly between the site types (p > 0.05; Table 2). Mean elevation for all 14 sites was 1.06 m (SE = 0.075). Slope and crawl distance did significantly differ between nest crawls and false crawls (p ≤ 0.05; Table 2). On average, nest sites had an approximately 5.5-m longer crawl distance and were flatter, or less steep, by 0.04 m · m⁻¹ in slope.

Table 2. Results from t-tests comparing loggerhead (Caretta caretta) nest sites (n = 11) to false crawls (n = 3) using data collected via topographical surveys on Horn Island, Mississippi, in July 2012. Group means are given with standard errors along with resulting t- and p-values. Bold p-values are significant (p ≤ α  = 0.05).

View Fullscreen

Discussion.

Our results are consistent with the hypothesis that sea turtles use beach slope as a nesting cue. The strong negative correlation between beach slope and crawl distance (R² = 0.62) suggests that loggerheads nesting on the barrier islands of Mississippi crawl shorter distances on steeper slopes versus longer distances on flatter slopes to reach a presumed elevation threshold. This relationship is congruous with previous studies that describe how nesting sea turtles interact with beach slope. Hawksbills (Eretmochelys imbricata) in Barbados may also use slope to reach a suitable elevation (Horrocks and Scott 1991), although slope is likely just one of a number of environmental factors affecting nest placement (Kamel and Mrosovsky 2005). Loggerhead turtles in east Florida likewise cue in to slope, though they may seek a beach zone (i.e., the back dune) rather than an elevation threshold (Wood and Bjorndal 2000). The exact nature of slope's role in nest placement may vary with beach-specific morphology and configuration.

For Mississippi loggerheads, we cannot rule out that factors not considered could influence crawl distance in addition to slope, and we acknowledge the constraints of a small sample size. However, our results provide some evidence that slope is a central factor in nest placement. The association between slope and crawl distance makes sense, as it may suggest that loggerheads seek a safe elevation and balance that goal against the energy costs of crawling (i.e., turtles do not crawl longer than needed at steep slopes; Fig. 2). The lack of correlations between elevation and slope and between elevation and crawl distance provides further evidence for this inference. The presumed elevation threshold sought by loggerheads at our sites appears to be approximately 1 m above sea level (mean = 1.06 ± 0.075). Only two sites had an elevation below 0.9 m. The Mississippi Sound has a tidal range of 0.6 m (Moncreiff 2007), so the mean elevation we observed may indeed safeguard against inundation. With no significant difference in elevation between nest crawls and false crawls, our results also suggest that once this suitable elevation was reached, it is likely that other factors ultimately drove a turtle's decision to either nest or return to the water without nesting (Miller et al. 2003). Two factors that we analyzed that may influence this decision are slope and distance crawled. We found that nest sites had a significantly lower mean slope and longer mean crawl distance than did false crawl sites.

A secondary objective of this article is to call attention to sea turtle nesting in Mississippi and the northern Gulf of Mexico. Our study is one of the first to present research on nesting in Mississippi, highlighting the data deficiency for the region. The barrier islands of the northern Gulf of Mexico are habitats that could conceivably increase in importance under projected climate change scenarios described by the Intergovernmental Panel on Climate Change (2014). These islands experience relatively low human impact and represent some of the northernmost nesting sites in the Gulf of Mexico at a time when climate warming and temperature-dependent sex determination threaten to further skew marine turtle sex ratios in the southern United States (Hawkes et al. 2007).

Although regional nesting data exist because GUIS staff opportunistically patrol the Mississippi island beaches, no routine sampling program was in place at the time of our study to document nest locations, hatching success, etc. Data such as ours that specifically describe the effects of beach morphology on regional nesting behavior could be particularly valuable to post-Hurricane Katrina and Deepwater Horizon island restoration planning, especially if techniques such as sand renourishment and topographical manipulation are used. These barrier islands are the subject of a recently approved restoration program expected to last 30–40 yrs (US Army Corps of Engineers 2016). This will be the largest island restoration effort in the United States and may result in significant rebuilding of beach habitats, including the addition of sand to the coastal system. Ultimately, this restoration effort will change the beach habitats and impact sea turtle nesting. How the restoration affects nesting populations will rely on the quality of added sand, survival of nests, and the ability of hatchlings to emerge and reach offshore currents for transport to nursery areas.

Our finding that loggerhead nesting behavior in this region is affected by beach morphology is encouraging with regard to their ability to adapt to predicted changes in weather patterns and the likely resultant increase in barrier island geomorphological variability. If beach slope or other morphological features consistently affect nesting in a manner that helps to ensure nests are sufficiently elevated to protect against inundation, physical changes that occur between nesting seasons should result in less deleterious effects on reproduction than if nest placement were more random or spatially specific. However, we offer only limited data, and we suggest that more research is needed for these unique nesting habitats in the northern Gulf of Mexico, especially to aid the restoration process. Augmenting scientific understanding of sea turtle nesting behavior and nest placement can increase the capacity for beneficial management and conservation of productive nesting habitats in the face of increasing environmental change.