Satellite surveys prove a reliable monitoring method for high latitude southern right whale habitat
Caroline H?schle, Kelly Macleod, Amel Ben Mahjoub, Vladislav Kosarev, Grant Humphries, Julika Voss, Emma L. Carroll, Rochelle Constantine, Simon Childerhouse, Dave Lundquist, Leena Riekkola, Georg Nehls
The application of Very-High-Resolution (VHR) satellites to survey cetaceans has gained considerable traction over the last decade. Large whale species in particular lend themselves to detection by VHR imagery of ~0.50mresolution or less. Processing of satellite images can be manually intensive, and consequently artificial intelligence
methods are underpinning progress in this field. We have developed a method that uses a Faster-R-CNN (Regionbased Convolutional Neural Network) object detection algorithm to process VHR imagery. This has been coupled with a manual species identification stage and packaged as the SPACEWHALE service. The service was used to acquire and process WorldView-2 archival images (~0.50 m resolution) of Port Ross, Auckland Island - Maukahuka in August 2020 to investigate the detection of southern right whales during the austral winter on this well-known breeding ground. The period of image acquisition coincided with boat-based surveys of the area which were used to compare the effectiveness of SPACEWHALE. The number of whales detected by both methods were equivalent, despite the 12-hour difference between the timing of data collection. The number of calves detected by satellite was slightly lower than the boat-based surveys, likely due to the algorithm missing calves that were completely obscured by their mothers during the snapshot survey. Use of higher resolution images (~0.30 m) may improve detection of calves that are not entirely hidden by their mothers. The SPACEWHALE survey did provide additional coverage and possible whale detections in a secondary area that has received little on-the-water survey effort; whilst this area is more exposed than Port Ross and is likely less suitable habitat, as the population continues to grow, areas beyond Port Ross may warrant research effort to determine whether this is a potential extension of core habitat or serves some other biological function for the whales (e.g., foraging). Satellite surveys are an effective alternative to monitor large whales in remote areas and can be used to augment existing data and to help explore and fill data gaps.