Urban Woodland Species Inventory 

While work is being done to make it more responsive for small screens, this page is currently best viewed on a desktop or tablet. 

Methodology

METHODOLOGY

3D locator map of the project study woods and surroundings on the DC side of the border

The study used two primary methods for vertebrate species data collection: wildlife cameras and avian bioacoustic recording. For both wildlife cameras and bioacoustic observations, species observed were compared against the 2015 DC State Wildlife Action Plan’s list of Species of Greatest Conservation Need (SGCN) and species observed that are present on that list were flagged, as well as species listed on the International Union for the Conservation of Nature’s (IUCN) Red List. 

Wildlife Camera Monitoring: Four cellular-connected wildlife cameras (three solar powered and one battery powered) were deployed, each equipped with automatic infrared and night-vision capabilities. Two cameras were supplied with water stations (3-gallon rubber water containers), with the water refreshed and replaced approximately every 48 hours, while the other two were left without. The cameras were all placed off of the informal human trails in the woods, but near obvious pinch points and bottlenecks that animals were considered likely to use. The cameras operated continuously throughout the study period, with only two brief interruptions (for two of the four cameras, occurring at different times) due to battery recharging (the cameras are solar-powered, which extends the time between charges considerably, but they still need to be recharged from time to time, particularly the more active cameras). The cameras were attached to trees using straps at approximately one to two feet off the ground in three of the four cases, and the fourth camera was placed at approximately five feet off the ground. 

Camera 1

Camera 2

Camera 3

Camera 4

Birdweather PUC 

Bioacoustic recording device at right, attached to strap of Camera 2

Animals detected were logged by species and count, distinguishing between suspected juveniles and adults. To minimize duplicate counts, observations of the same species on a single camera within a 10-minute window were considered as one event. In cases where multiple events took place within one 10-minute window (raccoons very frequently triggered the camera multiple times within one 10-minute period, to a lesser extent so did some forest floor birds like American robins, as well as eastern gray squirrels), the event with the most individuals present was used. 

Bioacoustic Data Collection: Supplemental avian data was gathered using a Birdweather PUC recording device, which was attached to the camera strap of Camera 2 (see image at left) about one foot above the forest floor. Multiple 24-hour recording sessions were conducted (July 25-26, July 27-28, and July 29-30), limited by the device's approximately 48-hour battery life. The audio data was processed using the BirdNET neural network (see Note 2), a collaborative project between the K. Lisa Yang Center for Conservation Bioacoustics at Cornell Lab of Ornithology and Chemnitz University of Technology.

BirdNET's algorithm differentiates between approximately 6,000 bird species, filtering out human-caused noises, non-avian species, and removing soundscapes containing human vocal detections. A 2023 review of nine different studies evaluating the accuracy of BirdNET’s analysis found that it ranged from 72-97%, depending on the study, with scores on the higher end of that scale for studies of only North American bird species (81-97% - see Note 3)

Limitations. The ground level wildlife cameras were most effective at capturing medium to large terrestrial animals. Consequently, species such as bats, amphibians, reptiles, and very small terrestrial mammals (e.g. chipmunks, mice) were likely undercounted or missed entirely. The cameras’ limited vertical range also means that species spending most of their time in the canopy were unlikely to be detected on the cameras. And while the bioacoustic data was instrumental in expanding the range of species identification into the canopy, capturing vocalizing birds that primarily spend their time there, it was not capable of detecting non-avian species. Additionally, the 24-48 hour bioacoustic recording sessions which were necessitated by the battery life of the recording device might have missed species that vocalize infrequently. Any future studies would benefit from additional techniques to provide a more complete picture of the area's biodiversity, particularly for under-represented groups like bats, amphibians, and reptiles.

Notes

Methodology