A beluga whale distribution dataset for the Canadian sub-Arctic derived from 2021–2022 WorldView imagery

This dataset is derived from WorldView-2/3 satellite imagery acquired in 2021–2022 over key summer habitats of beluga whales in the Canadian sub-Arctic , including northwestern and southeastern Cumberland Sound and the Churchill River estuary. It contains three annotated classes: certain beluga, uncertain beluga, and harp seal. The dataset was developed through a workflow combining manual annotation, deep learning inference, and manual verification. Systematic manual interpretation was first conducted in northwestern Cumberland Sound and the Churchill River to build training samples for object detection. A trained YOLOv8 model was then applied to southeastern Cumberland Sound for large-scale inference, followed by manual review. The final dataset includes geolocations of 721 certain belugas, 477 uncertain belugas, and 551 harp seals. This resource supports studies on the distribution and abundance estimation of beluga whales and harp seals, as well as training and evaluation of object detection algorithms, and provides a useful reference for polar ecological conservation and sustainable tourism planning.