Statistical power and sea turtle nesting beach surveys: How long and when?

Identifying long-term population trends is essential for the conservation of species of concern. With money and personnel often in short supply, the success of a monitoring program to provide an estimate of population trend requires an efficient survey design that can detect biologically important changes within some prescribed budgetary framework. In this paper we illustrate how costs can be reduced while maintaining an efficient design using a nesting beach survey of female hawksbill sea turtles (Eretmochelys imbricata) in the Eastern Caribbean as a specific example. We compared the statistical power of the current, intensive protocol with those of shorter duration and differing start date to generate general guidelines on sampling. Counts of individual turtles from 1987-2003 were analyzed using Poisson log-linear regression and models with both fixed and random effects describing the temporal variability were explored. We estimate that the current 20 week saturation-tagging survey design could detect a 3% per annum growth with acceptable statistical power (>0.8) in less than 14 years. However, by changing the current survey start date and counting for only 10 weeks, it is possible to detect similar trends with little loss of power while reducing monitoring costs by 25-50%. Our analyses indicate that the statistical power of surveys for hawksbills is influenced by the timing and duration of the survey because of their effect on the proportion of the annual number of nesting females that the survey captures. This simplifies estimating the power of a prospective survey program.