Seals from Outer Space - Population Census of Southern Elephant Seals Using VHR Satellite Imagery

This work aims to assess the utility of World-View 3 satellite imagery of 31 cm resolution for con-ducting population censuses of southern elephant seals (SES). For the first time, we present a comparison of the results of image analyses of a 31 cm resolution satellite image and 1.39 cm res-olution drone-based orthophoto of a SES breeding colony located at Patelnia Point (King George Island, breeding season 2019). The time distance separating both visualisations of the same area was 15 h 4 min, which allowed us to determine the female recognition from the VHR image with an error of 1.5-7.0%, depending on the density of females in the breeding formation. An impor-tant component of this error is false negative matches, which can have a significant impact when selecting the objects needed to train deep learning models. In addition, the pup recognition error was much higher and was 24.6-40.2%, which depended on the density of animals in harems, a large variation in the size of the pups and the diverse substrate on which they were deposited. Counting on VHR images was performed manually by eleven observers familiar with the morpho-logical characteristics of SES. The resolution of the VHR photo allows the identification of fe-males and males involved in breeding due to the clear dimorphism of their sizes. We recommend using VHR satellite imagery to determine the number of breeding formations and spatial relation-ships between them, as well as for counting females and males during the breeding haul-out. The resolution of satellite imagery available at this time is still insufficient to be used to determine breeding success. With increasing access to satellite resources and an ever-expanding database of VHR images, this is a promising tool to fill the gaps in knowledge about the global population of SESs.