Figure 3
From: Identification errors in camera-trap studies result in systematic population overestimation
How different rates of ghost-producing splitting errors (0.05–0.20) affect the population abundance estimate, based on camera-trapping data from a real-world snow leopard study in Mongolia21 (where the population size = 12, capture occasions = 16, capture probability = 0.16). Estimates are derived from 1000 simulations at each error rate; the solid line is the median error and the grey shading the 95% quantile range. This is shown relative to the expected credible range of splitting errors (50% and 95% CIs as thick and thin lines respectively), from expert and non-expert observers (generated from the binomial likelihood of the model in Appendix S1; for splitting errors that create new individuals).
