Evaluating the use of thermal imagery to count harbor seals in aerial surveys

Abstract

Aerial surveys of pinnipeds are often used to estimate abundance, a critical component of stock assessments and management decisions. In Alaska, USA, aerial surveys of harbor seals (Phoca vitulina) have historically relied on visual detections by human observers, a method which works well on large groups of seals at predictable haul-out locations, or when seals are located on a visually uniform substrate such as a sandy beach or exposed mudflat. However, regions such as the Aleutian Islands of Alaska, where harbor seals haul out in small numbers at variable locations and are inconspicuous on shore, are challenging to survey accurately. To determine whether the use of thermal imaging techniques would improve detections of harbor seals in the Aleutian Islands, we conducted a study to compare counts derived from visual detections documented by color photographs with those derived from thermal detections documented by infrared images. In 2019, we conducted 15 flights in the Aleutian Islands, completing 129 experimental trials. We manually reviewed color and thermal images to count harbor seals and used a Bayesian analysis to explore the effects of several covariates on seal detections. The thermal method detected more harbor seals than the visual method early in the day, when cloud cover was greater, and when observers had more experience operating the thermal imaging equipment. The relative improvement in performance of the thermal method was particularly notable when surveys occurred four or more hours prior to solar noon. We discuss the costs and benefits of incorporating thermal technology as part of the existing monitoring program for harbor seals in the Aleutian Islands, including the need to control for differences if incorporating new survey methods based on thermal detection.