Researchers hung camera traps 30 metres up in Lomami to map mammals layer by layer

The design is what makes the height numbers mean anything. At each point the team ran three cameras at once, one facing the ground, one a few metres higher in a tangle of vines, and one 20 to 30 metres up in a tree crown, all watching the same column of forest simultaneously, in lead author Daniel Alempijevic's description ^[9]. The alternative, ground cameras in one set of places and canopy cameras in another, would tangle height up with location. Here the spot and the sampling window are held fixed, so the only thing changing between the three records is how far off the ground the lens sits.

The arithmetic on the strata is where the survey-methods argument lives. Ground cameras logged 32 of the 47 taxa ^[4], which sounds like most of the community until you subtract: 15 taxa, close to a third of the total, never appeared in front of a ground camera ^[16]. The three per-stratum counts also add up to 77, thirty more than the number of taxa recorded ^[17], so multi-layer use is the rule here rather than a curiosity worth a figure caption.

The understory result is the one I would carry into a survey-design meeting. A quarter of the arboreal mammals preferred the understory over the crown ^[6], which means the middle camera is not a redundant copy of the high one in forest laced with vines.

The thing this does not tell you is effort. The Florida Atlantic University summary gives no station count, no camera-nights and no survey duration ^[21], and without a denominator 47 taxa cannot be set against a richness figure from anywhere else. The unit is taxa rather than species, and the material does not say how many were resolved to a species name ^[22]. The companion occupancy model in Ecology, meanwhile, is built on the 25 most-detected species ^[14], roughly half the taxa recorded ^[18], which is the half where the data is thickest and not the half a ground grid misses.

White-bellied pangolins used the canopy more often in mature forest and were detected closer to the ground where land had been cleared for agriculture, a pattern the authors note could raise their exposure to poaching by putting them near people more often ^[13]; pangolins are already among the most trafficked mammals on earth ^[20]. That is a comparison across forest types, not animals watched descending, and cleared land also has fewer tall crowns for a camera 25 metres up to observe.

The bycatch is its own argument for the column. More than 1,500 bird events came off the same cameras, including raptors taking mammals and eastern long-tailed hornbills trailing monkey groups ^[12], along with small insect-eating bats repeatedly approaching other mammals, possibly hunting the insects those animals attract ^[19]. Thousands of videos went to researchers and trained volunteers for identification, which is the hidden labour line in any camera study ^[8].

The undercount is not a correction factor you can carry elsewhere. Fifteen taxa is Lomami's number under Lomami's canopy structure ^[16], across three survey areas in one park ^[2], and a drier or shorter forest would produce a different one.