ESR prepress abstract   -  DOI: https://doi.org/10.3354/esr01396

ABSTRACT: The Tristan albatross Diomedea dabbenena is a Critically Endangered species that breeds exclusively on remote islands in the South Atlantic Ocean. Although the population continues to decline, regular on-the-ground monitoring of Tristan albatross populations is logistically challenging and costly. If this monitoring is reduced in the future, then alternative methods would be necessary to track long-term population trends. Here, we assessed the viability of using 31 cm resolution satellite imagery to count Tristan albatrosses during the breeding season on Gough Island. Counts of birds in a satellite image by 2 wildlife remote-sensing specialists were compared with GPS coordinates of active nests recorded in the field. Birds were detected at 103 (67.8%) of the 152 active nests in the cloud-free regions of the satellite image. Acquiring suitable imagery is challenging because upland nesting sites are prone to low-lying orographic cloud, with only one cloud-free image obtained across 8 seasons of archived and one year of tasked imagery. Our research demonstrates that due to incomplete detection, and the limited availability of suitable imagery resulting from persistent cloud cover over the island, Tristan albatrosses cannot be reliably counted or monitored with available satellite imagery. Differences in detection probability were not explained by nest attributes or bird plumage colouration. More commercial satellites in orbit may improve chances of obtaining cloud-free imagery across the island in the future, but until then, on-the-ground monitoring is required if we are to obtain accurate population counts and for the UK to meet its commitments to monitor this species.