Recognize nuance when interpreting monitoring results

Abstract

We recently published a study discussing the pitfalls of non‐probability sampling when selecting monitoring sites. We demonstrated that selecting sites based on abundance can often lead to biased inference, and we suggested that researchers use probability sampling. We also called for nuance when interpreting results of monitoring programs that use non‐probability sampling. We suggested that inference from sites of great abundance might still be useful for inference into population dynamics of long‐lived species such as raptors. Perret et al. seem to misinterpret our call for nuance as advocating for non‐probability sampling. They state that we concluded the general recommendation of using probability sampling should be revised. We did not conclude this. In fact, we agree with their recommendation. Perret et al. implemented simulations that are unrealistic within the context of our study. We use empirical data for 12 raptor species to demonstrate that our previous results are valid and that simulations implemented by Perret et al. do not reflect the biology of long‐lived raptors. The time series simulated by Perret et al. fluctuated greatly in abundance with populations often more than doubling within a year. This is extremely unlikely for populations of long‐lived species having high site fidelity. Many historical programs monitor sites of great abundance and thus risk biased results. We demonstrate that this risk is minimal under some important conditions and our results likely apply to other long‐lived species. Acknowledging this nuance could rescue many long‐term monitoring programs and their data thereby preserving efforts of costly conservation programs. Consistent with our original study, these exceptions do not invalidate the general recommendation to avoid non‐probability sampling; however, they do support our call for nuance when interpreting results of studies that monitored animals at sites of great abundance.