Dr. Kimberly Andrews
University of Georgia, Odum School of Ecology
We are in Q1 of our project investigating the vocalization of gopher tortoises (Gopherus polyphemus), so we have not yet received results from our study although we maintain confidence in our ability to succeed at our proposed goals. The occurrence of tortoise vocalizations simply has been documented but not characterized; hence, we are delving into when and under what circumstances these vocalizations occur. We are currently launching and testing our methodologies and exercising an adaptive approach to our study design accordingly. We initially deployed the Wildlife Acoustics units in a captive setting with young (~1 yr) gopher tortoises as a pilot setting. Following those tests, we then deployed the units in the field at a Georgia Department of Natural Resources Wildlife Management Area with the project goal of monitoring communication in wild populations. Specifically, we are interested in whether animals who were relocated from a mining site are interacting with resident males. This project uses Wildlife Acoustic units as a novel means of validation and is being used in concert with wildlife camera data. As directed by these two technological methods, we will collect genetic samples from hatchlings that are produced from nests where we heard and observed male-female interactions using the cameras and acoustic units.
The exploration into low-frequency (Less than 5 kHz) vocalizations in turtles and tortoises is a new frontier. These research endeavors add uniquely to our understanding of these animals' natural history. Further, this technology from Wildlife Acoustics allows us to quantitively and qualitatively study interactions in a manner that is biologically meaningful to the tortoises. By describing meaningful interactions among individuals, we can better interpret intra-populational interactions as indicators of true population health.
This work is conducted by the Andrews Applied Wildlife Conservation Lab (AWCL) at the UGA Odum School of Ecology. We operate out of a partner facility, the Marine Extension in Brunswick, GA. This Wildlife Acoustics project is a partnership with Southern Ionics Minerals, who executes a strong stewardship and wildlife conservation mission of translocating tortoises to lands where they can receive long-term protection. These recipient sites are managed by the Georgia Department of Natural Resources, who additionally engage in establishing our research priorities and providing our scientific collection permits that are necessary to conduct the work.
Our findings will absolutely influence existing conservation priorities and policy. We will use these data as a key method in an integrated approach to assess whether translocated and resident tortoise individuals are interacting and reproducing. If so, we can confirm that our mitigation and management efforts to translate tortoises are effective in meeting our goal of augmenting the sizes of resident populations on state lands. If population sizes meet those of the Minimum Viable Population classification set by the US Fish & Wildlife Service, these populations will count toward our goal of the species receiving a secure, rather than a threatened or endangered status, in the current petition review of the species for the Endangered Species Act.
As part of our research in the UGA Applied Wildlife Conservation Lab, we are continuing to monitor the activities of adult gopher tortoises with our SM4 recorders received from Wildlife Acoustics. These units are placed at the burrows of female gopher tortoises, some of which have been relocated from a heavy mineral mining site. We are interested in whether these newly established females are interacting with the resident males. Our monitoring process is adaptive as the level of tortoise activity varies in space and time and among individuals. As these radio-tracked individuals move among burrows, UGA researcher, Oscar Thompson, moves the recorders accordingly to "follow" these females. By determining which females have the highest number of male interactions, we increase our probability of capturing vocalizations between the sexes. We set the game cameras to look directly at the burrow to capture any social interactions that occur in front of the burrow, and then place our SM4 recorders to the side, where they will not obstruct our cameras but still capture any vocalizations made during interactions. By coupling our acoustic output with game camera data collected at the same burrows, we can isolate which vocalizations are tortoise-talk by pairing them with the images of tortoise interactions.
As we are early on in our acoustic research, this past quarter involved "training" our Kaleidoscope software to recognize tortoise banter. Initially, each of our 5 SM4 recorders was recording approximately 45 hours of audio files a week for a total of about 220 hours of audio recordings to analyze every week. Since gopher tortoise vocalizations have not been characterized previously, we are on the frontier of identifying the frequency and behavior of their vocalizations! Exciting stuff! To begin streamlining our searches, we referenced the camera data where we could confirm an interaction between two individuals and homed in on the audio recording for that date and time segment to identify vocalizations between individuals. Based on that characterization and using the Kaleidoscope Cluster Analysis function, we have been able to identify several different individual calls, including those that occur during mating. Our next step is to differentiate whether the differences in calls are attributable to differences in types of activities or based on individual variation.
In addition to capturing gopher tortoise vocalizations, we have been able to record numerous species of birds, as well as several different species of frog in the areas around the burrow.
This past quarter, we have continued to collect and analyze our acoustics data, all the while refining the cluster analysis classifier to increase the accuracy in our data sets. Based on the frequencies and patterns of the gopher tortoise vocalizations that we are attempting to isolate, we experience a high rate of false positives where Kaleidoscope is classifying other sounds with similar characteristics as gopher tortoise vocalizations. To ensure the accuracy of our data sets and eliminate the false positives, we are still manually identifying all vocalizations that Kaleidoscope classifies as tortoises, using the photos collected from our game cameras that are trained on the apron of each burrow to verify interactions. In a typical week’s data during the active tortoise season, we record and identify about 800-900 gopher tortoise vocalizations from our five SM4 recorders.
Additionally, this past fall, we conducted a trapping session to recapture the telemetered tortoises for equipment replacement. This telemetry equipment is central to the operation of this study as we move the cameras and acoustic units among burrows based upon movements confirmed through the GPS and VHF telemetry. Finally, we sent genetics samples to Dr. Stephen Spear, the lead geneticist at The Wilds in Columbus, OH for analysis. These samples will confirm which individuals are reproducing. We then can compare those data against the animals that we have seen on cameras and heard on the acoustic units for a more comprehensive integration of the behavioral and reproductive ecology of these gopher tortoises.
With lower temperatures in the 2nd half of November, the gopher tortoises at our study site have greatly reduced their activity levels and are settling into their burrows for the winter. Hence, we are receiving very few interactions or vocalizations on the cameras and recorders. Due to this reduction in activity, we decided to suspend our acoustic sampling until spring while maintaining our effort on further analyzing the almost 1,800 hours of audio data that we have collected with our five SM4 recorders. We are continuing with our project of matching individual vocalizations to game camera photos so that we can classify different vocalizations with specific behaviors. To date, we have been able to classify three vocalizations with specific behaviors. In addition, we have identified approximately 8-10 other vocalizations from the audio data, but still need to determine whether they are associated with specific behaviors. We hope that we can create an even more accurate Kaleidoscope-Pro cluster analysis file to better classify our data next year and reduce the number of false positives that are classified as gopher tortoise vocalizations.
In October, UGA researcher and project co-PI, Oscar Thompson, presented a poster on the Wildlife Acoustics project. This presentation outlined our methodology for deployment and analysis along with some preliminary results at the 40th annual Gopher Tortoise Council meeting at Archbold Biological Station, in Venus, Florida. The poster was very well received, and many people are looking forward to our results and the potential application in their studies with the species. A pdf copy of this poster is being submitted with this report. We are also targeting a publication to submit in 2019 outlining our application and methods.