Aerial Photo Imagery from Fall Waterfowl Surveys, Izembek Lagoon, Alaska, 2017-2019
Summary
The imagery and annotations presented here were generated while testing an aerial photographic survey design to improve repeatability, transparency, and estimation of variance for annual population estimates of geese staging at Izembek Lagoon, Alaska. This dataset includes 1) 131,031 .JPG images captured from a small fixed-wing occupied aircraft, usually at an altitude of about 457 m, over Izembek Lagoon; 2) annotations of objects (geese, ducks, and gulls) assigned by an automated algorithm, and 3) manually corrected annotations that were edited by project staff for only those photos automatically identified as containing geese of any species. The photos were captured at predetermined points to systematically sample the waterfowl present [...]
Summary
The imagery and annotations presented here were generated while testing an aerial photographic survey design to improve repeatability, transparency, and estimation of variance for annual population estimates of geese staging at Izembek Lagoon, Alaska. This dataset includes 1) 131,031 .JPG images captured from a small fixed-wing occupied aircraft, usually at an altitude of about 457 m, over Izembek Lagoon; 2) annotations of objects (geese, ducks, and gulls) assigned by an automated algorithm, and 3) manually corrected annotations that were edited by project staff for only those photos automatically identified as containing geese of any species. The photos were captured at predetermined points to systematically sample the waterfowl present on the open water of the lagoon. The systematic sample included photo points over land around the periphery of the lagoon (at the beginning and end of each transect), but only photos that were triggered over open water were evaluated for the presence of waterfowl. The overland photos are included here for completeness, but were not processed and do not have associated annotations or count data. The photos and annotations are organized by year (2017-2019), date (one replicate of the survey [usually complete, but sometimes partial] was conducted on each date, with a total of three surveys in September/October of each year), and camera (two full-frame cameras were used on each replicate, providing non-overlapping images). The full dataset totals 1.82 TB. The EXIF data for each .JPG include the camera parameters and latitude, longitude, and altitude at which the image was captured. Most photos were captured at standard parameters (approximately 457 m altitude and 200 mm focal length, producing an average photo footprint of 4555 sq m [83.1 m × 54.8 m]), but some were captured at other altitudes or shorter focal lengths to evaluate the effect on identifying objects in photos. Photo locations were spaced 62 m apart along each transect; the footprints of photos taken at standard parameters rarely overlapped by approximately 1–2 m when the camera triggering system experienced a brief delay, but photos at lower altitudes or shorter focal lengths sometimes overlapped extensively. Annotations are provided as the coordinates (with pixels as the unit) of each object detected in the photo, with the automatically or manually assigned identification of each object (i.e., bird taxon), in one file per photo (.csv and .json versions). Summary data from the images, including all pertinent metadata and the automated and manual waterbird counts for each image, are available as an accompanying USGS data release (Weiser et al. 2022).
