Study area and animals
We conducted experiments at the Smithsonian Tropical Research Institute’s field stations in Gamboa, Panama, from April to July 2018 and on Barro Colorado Island (BCI) from March to May 2019. Copiphora brevirostris were collected from the vegetation along pipeline road in Gamboa and on the forest trails of BCI. Animals were separated by sex, housed in net cages, and fed with apple, dry cat food, and water. Collection of animals and experiments were approved by the Ministry of Environment of Panama, scientific permit No. SE/A-22-18.
Characterizing C. brevirostris tremulatory signal
We recorded the tremulation of a single male C. brevirostris using a Laser Doppler Vibrometer (LDV) (Polytec PDV-100, sampling rate 44.1 kHz). The output of the laser was recorded as a .wav file using an audio recorder (Tascam 60D MKII, 44.1 kHz, 16-bit resolution). We placed a male katydid together with a female in a custom-made nylon mesh cage (20 cm3), and we put a 1cm2 reflective tape on one side of the cage. We recorded male tremulations when the male’s legs were positioned on the mesh containing the reflective tape, and the female was not on the same panel of the cage.
Measuring natural wind variation
We recorded wind data for five consecutive nights from 22:00 at night to 08:00 in the morning, for three different trails on BCI (Donato, Geostation, and Snyder Molino) between March and May 2019 on the forest trails of BCI. These were the same forest trails where katydids were collected in the same year. A wind meter (Kestrel 5500) was attached to a small tree at approximately 1.5 m height. Peak wind velocity was logged every minute, with a 2-s integration time. To explore general wind patterns throughout the night, we plotted the mean peak wind velocity per minute across all nights and trails (n = 15). Additionally, we used the max peak wind velocity (highest wind velocity recorded per minute across all nights and trails) to guide our selection of wind gust levels. Using a LDV, we measured wind-induced vibrations from natural wind on an Oenocarpus sp. palm and wind-induced vibrations from artificial wind on the cage on which the animals were tested. In both cases, we also recorded wind velocity with a wind meter.
Measuring signaling activity
Signaling activity was measured during July 2018 in Gamboa, Panama. Using a camcorder with night vision (Sony DCR-SR45), we recorded the behavior of five katydid couples from 22:00 at night to 06:00 in the morning. We decided to start at 22:00 based on pilot recordings where we found no activity in the earlier hours of the evening. Katydids were placed in a custom-made nylon mesh cage (20 cm3) with food and water. There were no plants inside the cage. The animals were placed in the cage 2–3 h before starting observations to let them acclimatize. Observations took place in a closed lab with an ambient temperature of 29 °C (SD ± 3 °C), where they were protected from wind and rain. There were no observers present in the room during the recordings. Katydids were taken out of the experimental cage the following morning and placed in a separate cage, separate from the animals that had not been tested yet. Individual animals were used only once per experiment. We released the katydids after all experiments were finished to avoid re-catching the same animals. The videos were analyzed in VLC media player version 22.214.171.124. Male tremulations were quantified from the videos by counting the total number of tremulations produced every hour.
Wind exposure experiment
The wind exposure experiment was carried out between March and May 2019 in a closed lab on BCI with an ambient temperature of 29 °C (SD ± 3 °C), protected from natural wind and rain. We exposed 16 katydid couples to artificially created wind using a computer fan (ebm-papst S-Force Series Axial, 200 × 50.88 mm, 1220 m3/h, 103 W, 48 V dc) mounted on a metal base. The fan was positioned in front of the cage (~ 1 m away) but on a separate table to reduce the transmission of vibrational noise from the mechanical engine of the fan. We directed the wind toward (and as a control also away) the cage using a PVC ventilation tube with a diameter of 203 mm. This was done manually by the experimenter who was present in the room at the time of the experiments. Treatment levels were also adjusted by hand by the experimenter.
Katydid couples were placed in the same experimental cage as in the signaling activity experiment, provided with food and water. The pairs were chosen randomly from the communal cages. All individuals were used only once. Behavior was recorded with a camcorder (Sony FDR-AX33) coupled with a Led and IR light (Sony HVL-LEIR1, 1500 lux). There were no plants inside the cage during the wind exposure experiment. Therefore, katydids signaled from the sides of the cage. We chose to make our behavioral recordings between 02:00 and 04:00 h, because based on our signaling activity data, this was the peak period of signaling activity.
In addition to different wind velocities (here after referred to as “wind” treatment), we also exposed katydid couples to two different control treatments. To control for any effect of the acoustic noise of the fan on katydid signaling, we exposed the katydid couples to the sound of the fan only (here after referred to as “sound” treatment). We did this by moving the ventilation tube to the side, so that katydids would be exposed to the sound of the fan, but not the artificial wind. Wind velocity levels at the cage during the sound treatment and the silent control were below 0.1 m/s, which was the detection threshold of the wind meter. Acoustic noise levels at the cage during the wind and the sound treatment varied from 60 to 70 dB SPL (A) measured with a Volcroft SPL meter (set to fast and max) (for individual measurements, see Table 1). Furthermore, using a LDV, we measured the vibrations induced by wind on the cage. The sound treatment added very little, if any vibrational noise to the cage in comparison with measurements was done when the fan was turned off. At the treatment level of 0.3 m/s, vibrational noise for the wind and the sound treatments was low, differing only by ± 1 dB from the silent treatment. At 0.6 m/s, vibrational noise remained relatively low, with wind vibrational noise levels increased by 4 dB, and acoustic noise showing little increase. The wind treatment at 0.9 m/s had vibrational noise 15 dB higher than the silent treatment, while acoustic noise remained similar to the silent treatment. At 1.5 m/s, wind produced vibrations that were almost 30 dB higher in amplitude than the silent levels, while acoustic noise remained similar to the silent treatment (Fig. 1, Table 1). We also included a treatment with no wind or sound (here after referred to as “silent”), which consisted of turning off the fan. Each couple was exposed to wind and sound treatments for four different wind velocities, plus the silent control. Each katydid couple was exposed to four 9-min blocks made up of 3 min of wind exposure (of a certain velocity), 3 min of sound treatment exposure, and 3 min of silent treatment for a total experimental time of 36 min. The treatment and the treatment level (wind velocity) were balanced and randomized per trial. The wind velocities that katydids were exposed to were 0.3, 0.6, 0.9, and 1.5 m/s and were based on the natural wind variation we found in the field. Only 1.5 m/s was slightly higher than our maximum recorded wind velocity.
Signal and statistical analyses
Signal analyses and statistical analyses were done with R version 3.3.1 (R Core Team 2016), run in the RStudio interface (RStudio Team 2015).
We calculated the power spectral density using a Hanning window of 2048 samples to measure the spectral characteristics of the tremulatory signal of C. brevirostris and compared its distribution to the power spectra for natural wind-induced vibrations and our experimental stimuli (artificial wind, the sound of the fan, and silence).
We used a Friedman test to determine the effect of time at night on the production of male tremulations. Our response variable was the number of tremulations produced by males per hour. Our grouping (predictor) variable was time at night, which we treated as a categorical variable making 1-hr bins starting at 22:00 at night until 6:00 in the morning. We included male as our block variable. To test the effect of treatment and treatment level on the number of male tremulations, we fitted a generalized linear mixed effect model from the R package “Lme4” (Bates et al. 2015) with Poisson distribution, including male as a random effect. To test whether the fit of the statistical model was improved by the inclusion of treatment and treatment level, we obtained Wald Chi-square statistics from the “Anova” function in the statistical package “Car” (Fox et al. 2011). We ran pairwise comparisons with Bonferroni correction between the treatments at different treatment levels with the R package “Emmeans” (Lenth and Lenth 2018).