Monitoring bird populations is extremely important to gain an understanding of the overall ecosystem. Additionally, many species distributions are currently moving north, which is a sign of climate change. In order to measure the bird populations, trained ornithologists use the “point count” methodology, in which the observer stands in an area for a given amount of time and tallies the birds that can be heard or seen. The “point count” methodology has a few limitations. Since bird identification is an imprecise science, two very experienced ornithologists can survey the same area yet identify different birds. To account for this, most surveys increase their sample size. In order to reach the high number of survey areas in the given amount of time, ornithologists must start to cut corners, such as finding survey areas that are close to roads. However, birds are very sensitive to noise pollution, so the roadside surveys don’t accurately represent a population within the inner forest.
A new methodology to study bird populations and reduce the limitations associated with “point count” includes the usage of drones. Many ecologists use drones in their research, since they provide a unique and developing technique for studying wildlife. By using drones, we can save and refer back to an audio recording, thereby allowing ornithologists to agree on the birds present. Additionally, drones are fast and easy to maneuver, and will be able to travel to more remote survey areas.
Dr. Andy Wilson recently determined that counting birds from audio recorders on drones yields similar results to the “point count” method (Wilson 2017). Due to these benefits, drones may have a future in bird population monitoring. However, this methodology is still very new, and Dr. Wilson is skeptical about the impacts of drone usage on bird behavior. Because birds are sensitive to noise pollution, could noisy drones affect bird singing behavior? If so, there are ethical issues with using this methodology. Therefore, Dr. Wilson’s lab is currently testing the effects of drones on bird song. Check out video footage from our drone!
We travel to State Game Lands 249, where we have a permit to conduct research in the surrounding fields. We placed a 200m grid on the land, so that all survey points are equally spaced out (Figure 1). At each survey point, we create a 50m square array of SM3 wildlife audio recorders (Figure 2). After placing and turning on the SM3 recorders, we leave the area at least 10 minutes before the experimental data collection starts. This allows the birds to settle and resume normal activity. Next, we allow the recorders to continue gathering audio data for 4 minutes. This period is used as the “before drone” data. After this period, we fly the DJI Mavic Pro drone to the middle of the survey point at 48m high. The drone hovers at this location for 3 minutes. A portable audio recorder is hung from the drone by 8m of fishing line, placing the recorder at 40m from the ground. After the drone returns from the hovering point, we allow the recorders to continue for four minutes, as the “after” period. In this way, we have data from four SM3 recorders before, during, and after the drone flight, as well as data from the recorder suspended from the drone. The recordings are taken back to the lab for further analysis.
Each individual bird has its own distinct call pattern that can be used by a skilled birder or new lab workers to identify the species and location of the bird. After clipping our audio files to the 11-minute segment of interest, we use two high pass filters to eliminate as much of the drone’s noise without sacrificing the birdcalls. The final step in the editing process includes clipping every other minute for analysis. This allows us to identify which birds are singing throughout the experiment. In this way, we can compare the number of birds singing in the before (1st and 3rd minute), during (5th and 7th minute), and after (9th and 11th minute) segments. Raven Pro version 1.5 allows us to simultaneously look at the four SM3 recording spectrograms for a single minute and identify the birds calling (Figure 3). However, there can be many individuals of the same species singing at the same time. Plus, birds may move further away from the drone throughout the experiment, which would be an important result. Therefore, we seek to determine each bird’s exact location. In order for a bird to be localized, the song must be identified on at least three of the four recordings. By using the correlator tool in Raven, we can determine the time of arrival differences to millisecond, in order to determine the location of the bird (Figure 4).
Now that we have three separate correlations, we can enter the data into a downloadable program called Soundfinder. The only information that we need to supply is the temperature of the air when we took the recordings, the three correlation numbers from Raven, and the GPS coordinates of the ground-based recording array. Soundfinder is then able to localize the bird to within 3 meters of the birds’ approximate locations. The UTM (universal transverse Mercator) coordinates provided by Soundfinder can be plugged into Google imagery to double-check the birds’ locations. For example, you would know that you had incorrectly correlated a bird’s call if a woodland bird, such as the White-eyed vireo, was localized in the middle of a field. Therefore, there is some trial by error involved and so it takes a practiced eye to properly correlate a single birdcall.
Raven and Soundfinder allow us to track individual birds throughout the 11-minute long experiment, and determine if the birds are moving away from the drone or if they stop singing. We hypothesize that the drone will not have an effect of the bird behavior, but further analysis of our data are needed to determine our results. If there is no effect, then this methodology can be used in the future. An ornithologist would only need to use the drone and the hanging recorder to gather audio data, thereby quickening and simplifying the bird population monitoring process.
Wilson, A.M., Barr, J. and M. Zagorski (2017). The feasibility of counting songbirds using unmanned aerial vehicles. American Ornithological Society, 134, 350-362.