Cost-effectiveness of lures in attracting mammals: results of a large-scale camera-trapping field experiment

While there is widespread agreement that camera trapping is revolutionising the way we study wildlife, the use of attractants is still controversial. In particular, to the best of our knowledge, the cost-effectiveness of different attractants used in camera trap surveys has never been assessed. To contribute in filling this knowledge gap we 1) compare the effectiveness of a suite of attractants in detecting widely distributed mammals in Europe and 2) evaluate the cost-effectiveness of these attractants, by calculating the costs associated to reach a specific management objective. We conducted the field experiment in the spring of 2023 in four study areas in central and northern Italy, encompassing a variety of environments, from lowland forest to alpine beech forest. We deployed camera traps at 60 sites, each consisting of three treatment lure stations (camera trap + treatment) with one control camera station with no lure, placed in a square configuration (side 100 m). The treatments included sardines, a commercial lure for canids and Martes spp. and peanut butter. Cameras were left active at each site for 14-21 days and detection histories of each species were generated using 24h time bins. To model detection probability, we fitted single season multi-method occupancy models for each species, comparing (i) the null model with detection probability constant across treatments and control station, (ii) the ‘method’ model, detectability specific to treatment, and (iii) models with covariates that could affect detection probability (i.e. ‘time since deployment’ and ‘study area’). We further conducted a power analysis to compare the cost-effectiveness of the different treatments. Specifically, we estimated the number of sites to be surveyed and the costs required to detect a decline of 25% in the occupancy of each species. Through a total of 1047 trap nights of activation, we detected 15 mammalian species, of which 13 were included in the analysis. The most widespread species was the roe deer (detected in 51 sites, across all study areas) and the least widespread was the golden jackal (detected in 8 sites, only in the Karst and Alps study area, 30 sites in total). The ‘method’ model was the top ranking model in the case of 7 species, including two canids (golden jackal and red fox), three mustelids (badger and Martes spp.), two Artiodactyla (red deer and roe deer) and a Rodent (porcupine). The results of the power analysis show how differences in required sampling effort (and costs) between the different attractants and no attractants can be dramatic. As an example, in the case of the Martes spp. monitoring protocol, using peanut butter will require 82 sites (6744 €) as compared with a camera without attractant requiring 486 sites (or 39244 €). Through our large scale field experiment we found that for seven species detectability varied with the type of attractant used. Specifically, sardines proved to be the most effective attractant for canids and the porcupine, peanut butter was most effective for mustelids but was avoided by the roe deer, whereas the commercial attractant was the most effective with red deer. Through the power analysis combined with the cost function analysis we were able to show striking differences in the cost-effectiveness of the different methods, which strongly emphasizes the critical importance played by the choice of whether to use an attractant or not and the type of attractant to be used.