One fundamental concern in conservation biology is species abundance. For many taxa, however, these data are costly to obtain via direct observation and thus limited in geographic or temporal scope. Very high-resolution satellite imagery provides a means to address these limitations and provide remotely-sensed counts of large, colonial species. We used very high resolution satellite imagery paired with field counts of three species of nesting albatrosses (Phoebastria immutabilis, P. nigripes, P. albatrus) at two sites (Torishima, Japan and Sand Island, Midway Atoll, Hawaii) in the Pacific Ocean to test the ability of satellite image-based counts to predict in-field counts with multiple image and habitat covariates. Albatross were identifiable on Torishima using both WorldView-2 and WorldView-3 platforms and on Sand Island using the WorldView-3 platform. Pan-fused images underestimate ground count by about 31%, when taking into account vegetation cover and sun elevation, and their availability is limited to WorldView-3 imagery. Panchromatic images more accurately model in-field count when taking into account platform, species and vegetation cover with errors of -40-25%. We applied the best-performing, panchromatic model to estimate an inaccessible colony of P. albatrus breeding in the Senkaku Islands, from a single satellite image in 2015. We show the colony has expanded to a minimum of 166 adult birds. We demonstrate that with sufficient calibration, robust, multi-species models can be developed to expand the use of very high-resolution satellite imagery to satisfy monitoring objectives constrained by time, funds, or accessibility.