Calibration of Aerial Thermal Infrared Imagery for Walrus Population Assessment

Concurrent aerial photography and emitted thermal infrared (10.6 imagery were acquired over walrus hauled out on sea ice in Foxe Basin, Northwest Territories, Canada. Digital thermal infrared data from a Forward Looking Infrared (FLIR) imager provides a method for estimating walrus numbers, since the objects (walrus groups) are considerably warmer than the background (ocean and sea ice). Coincident photographic counts and thermal infrared pixel counts are regressed by means of a least squares linear regression and an estimate of group size predicted from the number of pixels represented by each group. The results indicate that analog thermal imagery provides an effective means for obtaining a stratification variable that can subsequently be used in survey design. The FLIR walrus estimation approach is evaluated in the context of replicability, both through the physical mechanics and within a limited range of environmental conditions. Precision of FLIR estimated walrus counts is evaluated relative to the precision of photo counts. Walrus numbers extracted from the digital thermal imagery at a sample swath of 1778 m are as precise as those obtained from three independent photo counts at a sample swath of 686 m. In this configuration the FLIR provides a 160% increase in the sampling (area) fraction. Based on these results, we recommend a stratified approach to estimating walrus abundance using a thermal infrared sensor coupled with visual and photographic censusing techniques. We conclude with recommendations for continued analysis of this infrared censusing technique.