Empirically testing the effectiveness of thermal imaging as a tool for identification of large mammals in the African bushveldt

Monitoring animal populations often relies on direct visual observations. This is problematic at night when spotlighting can cause misidentification and inaccurate counting. Using infrared thermography (IRT) could potentially solve these difficulties, but reliability is uncertain. Here, we test the accuracy of 24 observers, differing in experience and skill levels, in identifying antelope species from IRT photographs taken in the African bush. Overall, 38% of identifications were correct to species level, and 50% were correct to genus/subfamily level. Identification accuracy depended on the confidence and skill of the observer (positive relationship), the number of animals present (positive relationship), and the distance at which it was taken (negative relationship). Species with characteristic features, horn morphology, or posture were identified with ~80% accuracy (e.g. wildebeest, kudu and impala) while others were considerably lower (e.g. blesbok and waterbuck). Experience significantly improved identification accuracy but the effect was not consistent between species and even experienced observers struggled to identify red hartebeest, reedbuck and eland. Counting inaccuracies were commonplace, particularly when group size was large. We conclude that thermal characteristics of species and experience of observers can pose challenges for African field ecologists, but IRT can be used to identify and count some species accurately, especially