Survival analysis of wildlife cameras exposed to theft

Setting camera traps along roads is often necessary for ecological research, yet these locations expose cameras to theft leading to substantial data losses. Measures to minimise this risk include placing cameras away from human settlements. However, the effects of this and other measures on camera-trap theft risk are yet to be quantified. Here, we assessed the impact of gates on roads, the frequency of vehicle and human foot traffic, distance to the nearest town, and reduced visibility, on the risk of camera-trap theft, using a four-year, geographically extensive camera-trapping study in Tasmania, Australia. The large dataset covered 564 camera sites operating for 316,372 days (average of 561 camera days per unit), with 112 cumulative thefts. We used Bayesian survival modelling to determine the factors that best explained theft risk. Our results showed a high initial vulnerability to theft that gradually reduced over time, with significant predictors of reduced theft risk being: (i) road sites with lower frequencies of vehicle traffic, (ii) greater distance from the nearest town, (iii) where movement was curtailed by the presence of a gate, and (iv) a temporal trend that likely reflects a selective culling of ‘high exposure’ sites and increased efforts to hide camera units. The frequency of human foot traffic surprisingly did not significantly elevate theft risk. Our study provides important insights into the factors contributing to a higher risk of camera-trap theft on roads and offers a robust analytical framework to identify these factors for application in diverse social and ecological contexts.