Identification of African antelope species: using thermographic videos to test the efficacy of real-time thermography

Real-time thermography using the live-view function of a thermal camera has considerable potential to improve surveys of nocturnal wildlife relative to traditional spotlighting, while also decreasing disturbance. However, ability to identify species accurately is paramount. We use video as a proxy for real-time thermography to test African antelope identification accuracy among 34 observers of differing experience. Overall accuracy was 41% but there were substantial species-specific differences (e.g. wildebeest (Connochaetes taurinus) = 81%; reedbuck (Redunca arundinum) = 12%). Observer experience was significantly positively related to accuracy (inexperienced = 30%; expert = 61%) with experienced observers being better able to use subtle movement and behavioural information to identify animals. However, the effect of experience was inconsistent between species: even experts found some species challenging (e.g. waterbuck (Kobus ellipsiprymnus) where coat patterning was invisible thermographically). Self-assessed confidence did not correlate with accuracy. Observers who were good at identifying species were also good at assessing group size. We conclude that real-time thermography is not a "magic bullet" and the understanding of species-specific effectiveness is vital. However, for some species and some groups of observers, accuracy can be extremely high (e.g. 100% for expert observers viewing wildebeest). Tailored training is essential for real-time thermography to be a reliable field technique.