Track surveys do not provide accurate or precise lion density estimates in serengeti

Abstract

More than 60% of the world's large carnivore species are threatened with extinction. Enumerating species abundance is critical for assessing their conservation status in response to anthropogenic threats and environmental stochasticity. Track surveys are commonly used to estimate abundance and density of large carnivore species, including lions (Panthera leo), but their suitability for estimating species abundance has been challenged. Recently developed regression models for track surveys of African large carnivores offer improvements over previous density estimators but have not been independently validated. We conducted weekly track surveys for lions in southeastern Serengeti National Park during 2015–2016 and applied one of these recent regression models, comparing corresponding lion densities to an independent density estimate in 2015 derived from a repeated call-in survey conducted during the same season. We surveyed 3289 km for tracks in total with overall lion densities of 41.2 (95% confidence limits [CL] = 31.9–57.9)/100 km2 in 2015 and 34.6 (26.8–46.0)/100 km2 in 2016. Within year point estimates of lion density varied up to 56% among weeks, though 95% CLs overlapped. Overall annual CLs from the track survey in 2015 did not overlap with the 95% credible interval from the estimate of lion density using a repeated call-in survey (14.36 lions/100 km2; 95% CrI = 9.04–29.31), suggesting overestimation of lion densities using track surveys in 2015. High between-year and among-week variation in density estimates from track surveys suggests that lion use of roads for movement varied over time and that other factors (e.g., prey distribution, luminosity) influenced lion movements independent of road distributions. We recommend caution when applying current track survey methods for estimating lion density and support application of survey designs that include direct observation of lions (e.g., call-in surveys), account for imperfect detection in a spatially-explicit framework, and incorporate environmental variables (e.g., prey, land cover) that can influence lion space use and movements.