Home | Publications | Lighthouse study

METHODS

Recording sites

Acoustic monitoring sites were selected that had an open window to the sky in at least a 60-degree expanding cone overhead, grid power, shelter for the recording equipment, and as quiet an ambient environment as possible. Given these specific requirements, sites were sought to fulfill an array design with three stations near the Lake Ontario shoreline, three roughly two miles inland, and three roughly four miles inland. The stations in each distance-from-lake stratum were aimed to be widely spaced across the length of the proposed Lighthouse Wind Energy Project. This acoustic monitoring station array was a preliminary investigation toward assessing migration rate differences as reflected by calling rate.

Recording equipment

Old Bird 21c microphones (Old Bird, Ithaca, NY) oriented skyward were used to receive sound from the night sky. This microphone has a hypercardioid sensitivity pattern with acute directional sensitivity in the 2-10 kHz range in roughly an inverted 60-degree cone expanding out from the direction it is aimed. Four-conducter telephone cable was used to transfer the audio signal to the microphone input of a Turtle Beach Amigo II usb sound card connected to Dell Latitude 2120 laptops running Windows 7 or 10 operating systems. Easy HiQ audio recording software was used to automatically record sound on the computer nightly from 21:00 to 05:00 edt at 22050 Hz sampling rate and 16 bit resolution. The resulting recording system, combined with the relatively quiet recording locations, was sufficient for acquiring a consistently gathered sample of avian night flight calls for 8 hours a night during the peak of landbird migration from late April through early June. All time noted in this report is edt (utc -4 hours).

Call detection & analysis

Tseep and Thrush flight call detection software (Old Bird, Ithaca, NY) were run on the all night recordings to automatically extract bird calls -- Tseep for high frequency calls (5-10 kHz) and Thrush for mid-frequency calls (2-5 kHz). This step was carried out in the versions of these detectors contained within Vesper software (ver. 4, Harold Mills, Ithaca, NY). W. Evans then used Vesper to visually analyze spectrograms of the extracted call clips, separate calls from non-calls, and classify calls to species categories based on Evans and O'Brien (2002).

The Old Bird detectors work by detecting signals relative to average background noise. So, for example, if the background noise is louder at one station compared to another and they are tasked with detecting the same stream of flight calls, which vary across the spectrum of possible loudness, then the station with lower background noise is likely to detect more calls. Specifically, the detector would notice weaker calls that did not meet the threshold above background noise during the louder background. In order to quantitatively compare calling between stations, variance in background noise needs to be considered. To help evaluate the impact of background noise differences on the relative number of calls detected, hourly background noise measurements were made. For each hour of the night a 10-second seqment was measured for average power in the 5-10 kHz range using the bioacoustic software Raven Pro (ver. 1.5; Bioacoustics Research Program, Ithaca, NY). The sample was taken in the window plus or minus 1 minute of the first through seventh hour point after recording began, therefore missing the evening and dawn bird chorus. Care was taken in selecting a 10-second sample window to avoid any short transient sounds. The seven 10-second samples were averaged to come up with a background noise level to represent each night at each station.

To assess whether background noise might be a factor when comparing call totals among stations, norms were established during quiet periods with no detectable sound from wind or surf, the primary sources of sustained (non-transient) backgound sound in the local environment of the acoustic stations.