Vertical and temporal variability in the probability of detection of fruit‐feeding butterflies and moths (Lepidoptera) in tropical forest

One important source of variation present in all biodiversity sampling protocols is species detection probability (i.e. the probability of observing a species, given that it is present at a site). In tropical forests, many species have very low probability of detection, and thus they can be easily overlooked. Measuring the detection probability (detectability) of these species is important to determine the sampling effort required to accurately record them. In the present study, the variation of detectability of fruit‐feeding butterflies and moths between strata (understory vs. canopy), temporally across sampling months, and in relation to observed abundance were examined in the Amazon and Atlantic forests using models of logistic regression. These results enabled the estimation of the sampling effort required to detect a given fraction of the total assemblage. Species detectability was positively correlated with observed abundance for both butterflies and moths. In the Amazon, most species were more detectable in a specific stratum (canopy or understory). Biblidinae, Charaxinae and Satyrini showed temporal variation in detection probability in Atlantic Forest, with distinct peaks during July–August, November, and March. In contrast, Brassolini and Coeini showed peak detectability in January and March, respectively. The vertical and temporal variation in detection probability means that sampling effort must be applied strategically to the correct seasons and strata to improve the accuracy of results. The estimated minimum sampling effort to detect 70% of the species present in these tropical forests is 428 trap‐days in Atlantic Forest and 1435 trap‐days in the Central Amazon forest.