Detecting mass mortality events in wildlife populations.

Reports in the literature of mass mortality events (MMEs) involving diverse animal taxa are increasing. Yet, many likely go unobserved due to imperfect detection and infrequent sampling. MMEs involving small, cryptic species, for instance, can be difficult to detect even during the event, and degradation and scavenging of carcasses can make the window for detection very short. Such detection biases make it difficult to understand trends in MMEs across time, regions, or taxa. Thus, we developed a simple modeling framework to clarify key aspects (e.g., sampling frequency, dynamics of detectability) of the problem and spur future work. Our framework describes the probability of detecting an MME as a function of the observation frequency relative to the rate at which MMEs become undetectable. Although simple, this framework is useful for developing an intuition about how the probability of detecting a randomly occurring MME increases with peak detectability, with slower rates of decay in detectability, and with more frequent observations. It can also facilitate the design of surveillance programs. To illustrate its utility, we applied it to Ranavirus-related MMEs in 35 populations of an endangered salamander subspecies. We found that the probability of detecting an MME was <50% and that the frequency of MMEs in this system was likely much greater than the one MME observed in the 35 ponds. The limitations of this framework (e.g., assumption that surveys occur regularly and with equal effort) may help set an agenda for future research in this area.