Use of site occupancy models for targeted monitoring of the cheetah

The cheetah Acinonyx jubatus has suffered dramatic range contractions and population declines as a result of habitat degradation, prey depletion and conflict with humans. Of further concern is that many of Africa’s remaining cheetah populations persist in human-dominated and highly fragmented landscapes, where their ecology is poorly understood and population data are lacking. Presence–absence surveys may be a practical means to collect these data; however, failing to account for detection error can lead to biased estimates and misleading inferences; potentially having deleterious consequences for species conservation. The goal of this study was to identify how an occupancy modelling technique that explicitly accounts for detectability could be used for quantifying cheetah status in humanimpacted landscapes. Replicated camera-trap and track surveys of 100-km2 sample units were used to estimate the proportion of area occupied by cheetahs and to determine the survey effort required to inform conservation planning. Based on our results, 16 km [±standard error (SE) = 12–22] of walking or 193 camera-trap nights (±SE = 141–292) are required to confirm cheetah absence at a given 100-km2 grid cell (with 95% certainty). Accounting for detection resulted in an overall cheetah occurrence estimate of 0.40 (SE = 0.13), which is 16% higher than the traditional presence–absence estimate that ignores detection error. We test a priori hypotheses to investigate factors limiting cheetahs using an occurrence probability model of their preferred prey. The results show that both cheetahs and their prey were strongly negatively influenced by human settlements. Our study provides an unbiased estimate of occurrence that can be used to compare status across different sites and as a basis for long-term monitoring. Based on our results, we suggest that track and/or camera-trap surveys coupled with site occupancy models may be useful for targeted monitoring of cheetahs across their distribution.