UAV imagery of wildlife colonies, South Georgia and the South Sandwich Islands, 2019-2020
During the 2019/20 austral summer, an unoccupied aerial vehicle (UAV; drone) was used to collect imagery of breeding populations of seals, penguins and albatross. Two extensive southern elephant seal breeding sites were surveyed with complete counts made around the peak pupping date. A total of nine...
| Main Authors: | , , , |
|---|---|
| Format: | Still Image |
| Language: | English |
| Published: |
NERC EDS UK Polar Data Centre
2021
|
| Subjects: | |
| Online Access: | https://dx.doi.org/10.5285/cc083575-2cb8-4477-be76-8d81d44cfdc5 https://data.bas.ac.uk/full-record.php?id=GB/NERC/BAS/PDC/01483 |
| Summary: | During the 2019/20 austral summer, an unoccupied aerial vehicle (UAV; drone) was used to collect imagery of breeding populations of seals, penguins and albatross. Two extensive southern elephant seal breeding sites were surveyed with complete counts made around the peak pupping date. A total of nine islands historically recorded as breeding sites for wandering albatross were surveyed. Populations of Adelie and chinstrap penguin colonies at the South Sandwich Islands, an extensive king penguin colony on South Georgia and a macaroni penguin colony, also on South Georgia, were also included in the surveys. The South Georgia aspects of this work were funded by grants to the British Antarctic Survey (BAS) from the Government of South Georgia and the South Sandwich Islands and UK Foreign Commonwealth and Development Office. The South Sandwich Islands expedition was funded by the John Ellerman Foundation and donations from passengers on Quark Expeditions' ships. : The study took place at five locations on South Georgia and 1 location at the South Sandwich Islands between October 2019 and January 2020. At South Georgia southern elephant seal populations were surveyed along the coastline within King Edward Cove on the Thatcher Peninsula, and at Hound Bay and St Andrews Bay on the Barff Peninsula. Nine islands within the Bay of Isles were surveyed with a focus on the wandering albatross population. Additionally, a king penguin colony at St Andrews Bay and a macaroni penguin colony at Rookery Bay were surveyed. As part of an expedition to the South Sandwich Islands a survey of a mixed Adelie and chinstrap penguin colony at Beach Point on Thule Island was carried out from a yacht. All surveys utilised a commercially available quadcopter, DJI Mavic 2 Pro (DJI Inc., Shenzhen, China), equipped with the original 20MP RGB Hasselblad L1D-20c camera. The UAV was remotely piloted using the DJI GO 4 app (DJI Inc., Shenzhen, China; version 4.3) run on an iPhone 8 (Apple, Cupertino, CA) or DJI smart controller (DJI Inc., Shenzhen, China). Being small and light the UAV was portable and did not require specialised launching or recovery equipment. The majority of flights involved terrestrial take-off and landing, apart from those to survey the Bay of Isles, Hound Bay and Beach Point on Thule Island, which were piloted from a vessel at sea within close proximity of the study site. Initial plans of using software to compile predetermined survey flight paths at South Georgia were dismissed, primarily due to limited internet access which would have been required to download the maps necessary for offline use when in the field, but also due to the imprecise location of targets within maps. Breeding colonies may expand, contract or shift location making it impractical to rely on predefined survey paths. The UAV was therefore controlled manually with the pilot configuring the camera settings to achieve consistent sets of imagery and estimating the level of overlap between images with attempts made to achieve at least 70 percent forward and lateral overlap. The majority of the surveys were flown at a horizontal speed of 5 m/s with the UAV set to capture a photograph every 2 seconds. An observer was present during flights to assist the pilot by keeping visual contact of the UAV and making observations of potential disturbance to wildlife. Briefings were given to observers by the pilot to ensure they were aware of emergency procedures, should an issue arise. The UAV pilots underwent training prior to the study and gained a Remote Pilot Qualification-Small (RPQ-S) and a Civil Aviation Authority Permission for Commercial Operation (CAA PfCO). Flights were only conducted under suitable flying conditions and were made in accordance with the Air Navigation (Overseas Territories) Ordinance, following regulations set out by the Government of South Georgia and the South Sandwich Islands (GSGSSI) under Regulated Activity Permits RAP 2019/020, RAP 2019/024 and RAP 2019/050. Images were recovered from the UAV after each flight and saved on external storage media, in duplicate with one copy as a backup, for further processing. The images were initially organised and processed using Adobe Photoshop Lightroom Classic (Adobe Systems Inc, San Jose, California, USA) with keywords assigned to each unique survey, including additional metadata about the flights. Exposure corrections were made where necessary and altered image files were exported as georeferenced JPEGs for further processing. Agisoft Metashape (Agisoft 2020, Version 1.6.2, https://www.agisoft.com) was used to align and stitch the images from each unique survey to form orthomosaic images. These were exported as georeferenced TIFF files for use in mapping software, and as lower resolution JPEG files for counts. The resolution of the orthoimages ranged from 1.2 cm/pixel in surveys co...(2) : UAV - DJI Mavic 2 Pro UAV flight software - DJI GO4 app Adobe Lightroom CC 10.0 was used for image management Agisoft Metashape (Agisoft 2020, Version 1.6.2, https://www.agisoft.com) was used for image processing Counts of southern elephant seals from the aerial imagery were all made by a single observer using opensource software, DotDotGoose - Ersts, P. J. (2019). DotDotGoose (version 1.5.1). American Museum of Natural History, Center for Biodiversity and Conservation. Available from https://biodiversityinformatics.amnh.org/open_source/dotdotgoose. Accessed on 2020-11-11. : Georeferenced UAV imagery is accurate to within 5 m of actual ground position. UAVs used in the study were calibrated to the manufacturing standards. Imagery may have altered slightly during stitching with minimal distortion due to changes in ground elevation and movement of animals between overlapping captured images. Resolution: Variable - spatial resolution between 1.2 and 3.2 cm per pixel for much of the imagery. |
|---|