3  Conclusions

There was no statistical relationship between sea turtle nesting on Pensacola Beach and either foreshore slope or beach slope. It is possible that the nesting habits of loggerhead sea turtles are influenced by factors not included in our study and unavailable in our dataset.

When comparing the two modeling approaches, binary logistic regression and multinomial logistic regression, we also found no meaningful difference. Both methods resulted in similar conclusions regarding the relationship between nesting status and the beach characteristics. While the regression coefficients are different in the resulting models, they do not differ in a meaningful way. The story of the data remains the same, regardless of the modeling approach.

When directly comparing the use of false crawl points versus pseudo absence points as the “no nest” observation, we found no meaningful difference between the two methods. Both methods struggled to accurately classify the nesting status of the locations, especially false crawls. None of the modeling approaches predicted a single false crawl point. Thus, we saw low accuracy, sensitivity, and specificity. This is likely due to the lack of a relationship between nesting status and the beach characteristics as well as a small sample size.

3.1 Limitations

There are several limitations in this study and analysis.

First, we are limited by the data collection methods. This analysis relies on a combination of human collected data and randomly generated data from ArcGIS. For the observed nests and observed false crawls, FWC employees recorded information during turtle patrol; human error will inherently play a role. The pseudo absence points are randomly generated by ArcGIS, profile lines are then created by our environmental science collaborators, then beach slope and foreshore slope are calculated from those profile lines.

Next, there is not a large population of loggerhead sea turtles nesting on Pensacola Beach. This leads to a smaller sample size, limiting the statistical power of our analysis, and results may not be generalizable outside of Escambia County. Similarly, this analysis is only for one nesting season and may not be representative of other years.

The true relationship between loggerhead sea turtle nesting status and either the beach slope or foreshore slope of the location is unknown. In order to truly compare the methods, a full simulation study is needed so that bias and mean squared error can be examined as well.

3.2 Suggestions for Further Study

Limitations can be addressed in future studies. While data collection methods cannot be changed, the Computational Geomorphology and Modeling Lab plans to expand upon this project.

First, more data will be collected over multiple nesting seasons to increase the sample size and statistical power of the analysis, adding a temporal component to the analysis. Similarly, we can expand the spatial study area to include other beaches in the Florida Panhandle or throughout the State of Florida. This would increase the generalizability of the results.

A full Monte Carlo simulation study can also be conducted to truly compare the methods. By specifying the true relationship between nesting status and the beach characteristics, we can evaluate bias and mean squared error of the resulting models. This will provide more insight into which method is “better” in terms of estimating the true relationship.

Finally, other environmental predictors can be included in the analysis, if observed and available. There are many factors that may influence sea turtle nesting behavior, such as sand temperature, vegetation, human activity, and artificial lighting. Including additional predictors may help to better understand the nesting habits of loggerhead sea turtles on Pensacola Beach.