Abstract:
Loggerhead turtle populations have been on the decline in recent years throughout much of their range, partially as a result of mortality in shrimp trawling nets. To curtail this decline, regulations have been enacted that require the use of devices that allow turtles to escape unharmed from the trawl nets. These devices are known as turtle excluder devices or TEDs. Little data have been published that demonstrate the effectiveness of TEDs in enhancing turtle populations. In this report, loggerhead turtle strandings (i.e. dead turtles that wash ashore) from South Carolina are examined for possible relationships to the use of these devices. A linear statistical model is proposed that includes components representing aspects of the observed South Carolina strandings record, and allows the inclusion of a TED parameter. We fit this model by ordinary least squares, but the residuals from this model were found to be both autocorrelated and heteroscedastic. Fitting a regression model with a time-series error structure using the natural log transformed (strandings + 1) provided a very good statistical fit to the data. But interpretation of the parameters is not straight forward on the log scale. We used several methods to enable interpretation on the original scale. First, we interpreted the parameter estimates based on the log transformation approximately as percentage changes. Second, we used an approximation based on a Taylor series expansion to back-transform the parameters of the log transformed model to the original scale. Finally, the model was fit using ordinary least squares and the standard errors adjusted for serial correlation and heteroscedastic variance. These various methods each have their drawbacks which are discussed. The TED effect is estimated to be -16 turtles/bi-weekly period based on the log-strandings analysis, indicating that TEDs reduce strandings. This is roughly a 44 percent decline in strandings as a result of TED use, and was highly significant. The estimated trend in strandings is found to closely mimic what is believed to be occurring in the SC loggerhead population based on aerial swveys. Finally, we are able to detect spring and fall peaks in strandings in addition to that occurring in the summer months. Future analyses pertaining to these data are suggested.
Keywords:
Loggerhead turtles, turtle-excluder devices, TEDs, shrimp fishery.
tr13.pdf