Known on campus alternatively as the “drone people” or the “bird people,” we (Megan, Brian, Janine, and our professor, Dr. Andy Wilson) have been spending this summer flying our DJI Phantom quadcopter drone, affectionately named LARRY. However, we’re not using LARRY to spy on our fellow researchers, but rather to count birds.
How do we count birds?
Traditionally, ornithologists estimate bird populations by counting all the birds they see and hear within a pre-determined sample area. These estimates are reliable because they are performed during the breeding season, when males sing to establish and defend their territory, and are therefore much easier to detect. Territorial birds remain in the same small area for several weeks during the breeding season, which helps to ensure that bird counts are as accurate as possible. Point counts involve counting all the birds seen or heard within a pre-determined radius (e.g. 100m) for a set length of time, noting sex, distance from observer, and behavior. Transect counts are similar, but are performed by walking a straight line and allow for a larger sample area. These techniques are widely used and convenient for researchers and are usually performed on or near roadsides. However, over-sampling of road sides leads to biases, whereby less accessible habitats, such as forest interiors and marshlands, are under-sampled.
Feasibility of Using Drones to Estimate Population Density
To overcome the habitat sampling bias, we’re testing the feasibility of using a microphone suspended 8 meters below our drone, LARRY, to replace human observers. The drone is flown at approximately 58m above the ground, so we are able to fly over important interior habitat without disturbing the flora and fauna of an area. The recording can then be analyzed in the lab, to identify all birds singing or calling in the area sampled.
Most of our fieldwork this summer has been done on State Game Lands 249 in Biglerville PA, a few miles north of campus. When the weather is good (no rain and low wind), we leave campus at 5:30am to perform point counts. Using a grid system, we plotted 50 point count stations, each 200 meters apart, within the state game lands. At each point, LARRY hovers for three minutes while recording bird song in the area under the drone, before moving onto the next point. The flight time for the DJI Phantom drone is approximately 20 minutes, so we are able to perform three points per battery before landing to replace the battery. Either before or after each aerial recording, Dr. Wilson performs a traditional ground-based point count at the same location on the same day. The counts from the recordings will then be compared to the traditional field count methods to see whether the drone recordings are a suitable alternative to traditional sampling methods.
Where we’ve Been this Summer
- State Game Lands 249—Adams County, PA
- Our main research site, here we are looking to record any and all singing bird species. The most common of which include Grey Catbird, Red-winged Blackbird, Song Sparrow and Common Yellowthroat.
- State Game Lands 330—Clarion County, PA
- This site is the location of a reclaimed surface mine where most of the overlying land was stripped away for coal extraction. Now this site is managed as a crucial habitat for many grassland species. At this location, we were the most interested in the ability to detect quiet grassland species such as the Henslow’s Sparrow and Clay-colored Sparrow. At this site we flew 11 line transect surveys, each about 400 m long. We also conducted traditional ground-based surveys, with which we can compare our aerial recordings.
- State Game Lands 313—Tioga County, PA
- Our final field work location was a large wetland in northern Pennsylvania. Here we recorded secretive marsh bird calls in order to estimate their populations. To do this, we played a recording of a variety of marsh bird songs while the drone was hovering over the wetland. Some of the birds seen at these sites were Great Blue Herons, Green Herons, and Virginia Rail.