Data collection and manipulation
We deployed ARUs (Song Meter Model SM2 and SM4, Wildlife Acoustics,
Concord, MA, USA) to collect ambient sound from 1 March – 31 May. We
deployed 15 ARUs on the Webb WMA Complex during 2015 – 2018, 10 on CWMA
during 2014 – 2018, 20 on SRS during 2014 – 2018, 16 on CCWMA in 2017,
and 8 on BFG in 2017. We increased sampling efforts during 2018 in
Georgia by deploying 20 additional ARUs on CCWMA and 10 on BFG. We
placed ARUs >2 km apart to prevent multiple units from
detecting the same call (Wightman et al. 2019, Wakefield et al. 2020).
We attached ARUs to tree trunks approximately 3m off the ground and
placed an external microphone between 6m and 10m above the ground on the
same tree (Wightman et al. 2019). We placed ARUs at locations observed
to have turkey activity based on field observations and global
positioning system (GPS) locations of wild turkeys collected during
previous research (Wightman et al. 2019). We used ambient sound recorded
from 30 minutes prior to sunrise until 150 minutes post sunrise as this
is when > 75% of vocalizations occurred (Wightman et al.
2019, Wakefield et al. 2020).
We used a Convolutional Neural Network (CNN) developed to autonomously
search for turkey gobbles (Wightman et al. 2021). We implemented the CNN
in Python (Python Software Foundation, Wilmington, DE, USA) with the
Keras library (Chollet 2015) using a backend of the open-source
TensorFlow software developed by Google (Abadi et al. 2015). For each
potential gobble selected by the CNN, a record was created containing
call location in the spectrogram, date and time stamp, and a 3 second
sound file of the potential gobble. We auditorily verified all
selections and classified each as a true or false gobble, producing
daily counts of gobbles on all sites.
We collected weather data for SRS and CWMA from 2 weather stations
located on SRS maintained by the U.S. Department of Energy and U.S.
Department of Agriculture Forest Service. We used the most centrally
located weather station on SRS to describe weather metrics associated
with gobbling activity onsite. The second weather station was on the
southern border of SRS, approximately 10.5 km from the center of CWMA,
and was used for gobbling evaluation on CWMA. For the Webb WMA Complex,
CCWMA, and BFG, we collected weather metrics from the closest National
Oceanic and Atmospheric Administration (NOAA) weather station. The
closest weather station to the Webb WMA Complex was located in
Varnville, SC (35 km), whereas the closest station to CCWMA (25 km) and
BFG (35 km) was near Eatonton, GA. Although previous authors have
suggested the potential for placing weather stations at each ARU
(Palumbo et al. 2019, Wightman et al. 2019) such a study design was not
logistically feasible. We offer that using weather data collected on the
same study site or within the distances detailed above is sufficient for
detailing how daily changes in local weather conditions influence
gobbling activity. We calculated mean daily values from 15-minute
weather recordings from 30 minutes prior to 150 minutes after sunrise
for temperature (C°), relative humidity percentage, and wind speed
(kph). For barometric pressure (mb) we calculated the mean for each
morning and then subtracted it from the prior morning to get a change in
barometric pressure. For precipitation, we classified whether rain
occurred (Yes = 1, No = 0) from 30 minutes before to 150 minutes after
sunrise.