Wildlife trail or systematic? Camera trap placement has little effect on estimates of mammal diversity in a tropical forest in Gabon

Camera traps (CTs) have been increasingly used for wildlife monitoring worldwide. In the tropics, most CT inventories target wildlife‐friendly sites, and CTs are commonly placed towards wildlife trails. However, it has been argued that this placement strategy potentially provides biased results in comparison to more systematic or randomized approaches. Here, we investigated the impact of CT placement on the remotely sensed mammal diversity in a tropical forest in Gabon by comparing pairs of systematically placed and wildlife‐trail‐oriented CTs. Our survey protocol consisted of 15–17 sampling points arranged on a 2 km2 grid and left for one month in the field. This protocol was replicated sequentially in four areas. Each sampling point comprised a CT pair: the ‘systematic CT’, installed at the theoretical point and systematically oriented towards the most uncluttered view; and the ‘trail CT’, placed within a 20‐m radius and facing a wildlife trail. For the vast majority of species, the detection probabilities were comparable between placements. Species average capture rates were slightly higher for trail‐based CTs, though this trend was not significant for any species. Therefore, the species richness and composition of the overall community, such as the spatial distribution patterns (from evenly spread to site‐restricted) of individual species, were similarly depicted by both placements. Opting for a systematic orientation ensures that pathways used preferentially by some species—and avoided by others—will be sampled proportionally to their density in the forest undergrowth. However, trail‐based placement is routinely used, already producing standardised data within large‐scale monitoring programmes. Here, both placements provided a comparable picture of the mammal community, though it might not be necessarily true in depauperate areas. Both types of CT data can nevertheless be combined in multi‐site analyses, since methods now allow accounting for differences in study design and detection bias in original CT data.