Skip to main content
Dryad logo

Data from: Vocal characteristics of prairie dog alarm calls across an urban noise gradient


Shannon, Graeme et al. (2019), Data from: Vocal characteristics of prairie dog alarm calls across an urban noise gradient, Dryad, Dataset,


Increasing anthropogenic noise is having a global impact on wildlife, particularly due to the masking of crucial acoustical communication. However, there have been few studies examining the impacts of noise exposure on communication in free-ranging terrestrial mammals. We studied alarm calls of black-tailed prairie dogs (Cynomys ludovicianus) across an urban gradient to explore vocal adjustment relative to different levels of noise exposure. There was no change in the frequency 5%, peak frequency or duration of the alarm calls across the noise gradient. However, the minimum frequency – a commonly used, yet potentially compromised metric – did indeed show a positive relationship with noise exposure. We suspect this is a result of masking of observable call properties by noise, rather than behavioural adjustment. In addition, the proximity of conspecifics and the distance to the perceived threat (observer) did affect the frequency 5% of alarm calls. These results reveal that prairie dogs do not appear to be adjusting their alarm calls in noisy environments but likely do in relation to their social context and the proximity of a predatory threat. Anthropogenic noise can elicit a range of behavioural and physiological responses across taxa, but elucidating the specific mechanisms driving these responses can be challenging, particularly as these are not necessarily mutually exclusive. Our research sheds light on how prairie dogs appear to respond to noise as a source of increased risk, rather than as a distraction or through acoustical masking as shown in other commonly studied species (e.g. fish, songbirds, marine mammals).


Prairie dog alarm calls were recorded across an urban gradient at three distinct colonies from 28 August to 6 December 2014. Alarm calls were elicited by the observer approaching a randomly selected prairie dog. Once the prairie dog began alarm calling the observer remained stationary and recorded 30 seconds of vocalization while the animal was in situ. A band-limited automated detector was used in Raven Pro v1.5 to select each of the individual barks in the 30-second calling bouts and to optimize extraction of call parameters. Before measurements were extracted on the individual barks, all detections were examined manually for accuracy and adjusted to maximize the detection of all barks within a recording period and to ensure the entire bandwidth and duration of calls were selected. Random selections of half of the barks in a calling bout (n = 4516) were then measured. Four acoustic metrics were calculated for each bark: (1) minimum frequency (Hz) – the lower frequency limit of the call, a commonly used metric in previous studies; (2) frequency 5% (Hz) – the frequency where the summed energy equals 5% of the total, a measure of lower frequency properties; (3) peak frequency (Hz) – the frequency with the highest concentration of energy; and (4) bark duration (milliseconds). Ambient sound levels were measured using a calibrated Larson-Davis 831 sound level meter (frequency weighting = A) over a 2-minute period as soon as the vocalization recording was completed. Sound pressure levels were measured as 1-second frequency weighted (12.5Hz - 20kHz) equivalent continuous levels (LAeq, 1s). 

Usage Notes

Spreadsheet columns:

Colony: 1 = Pineridge, 2 = Coyote Ridge & 3 = The Coterie

BeginTime & EndTime: Specific start and end time of each bark on the recorder

MinFreq, Freq5, PeakFreq & BarkDuration: Response variables - details of these provided in methods section above

Individual: ID for each individual that was recorded

Laeq: Ambient sound level recorded over a two minute period

Dist_Rec: Distance to the focal animal

Dist_NPD: Distance from the focal animal to nearest neighbour

Windspeed & Julian Day: Additional explanatory variables

Normalized values are provided for all explanatory variables - these are given the 'n' prefix