Scientists and conservation advocates are well aware of the many human causes of population declines in many animal species, including habitat loss, exploitation, and climate change. However, diagnosing and measuring such threats for particular species has proven to be something of a challenge. This is because many proposed diagnostic methods are impractical or inconclusive, and population data have traditionally been sparse.

This paper, published in Conservation Biology, reports on the development of a new framework for diagnosing the nature of human threats to animal species. It uses a Bayesian analysis of data obtained from publicly available census information on animal populations. The authors note that Bayesian methods are “especially well suited to the task of threat diagnosis because they can be used to consider multiple sources of information within a common analytical framework and to specify a broad range of models…and sources of uncertainty.”

They first developed a Bayesian model selection method to diagnose varying levels of three conditions: no threat, exploitation, and habitat loss. They then tested the effectiveness of their framework on two models. One model was created using artificial data that simulated different threats. The other was a time series model that used real data from four well-studied species for which threats are relatively well known: the bluefin tuna and Atlantic cod, which are threatened primarily by exploitation; and the red grouse and Eurasian skylark, which are threatened primarily by habitat loss.

The simulated model correctly diagnosed the true threat for 88% of severe exploitation scenarios, 92% of severe habitat loss scenarios, and 82% of no-threat scenarios. However, the model was slightly less accurate in diagnosing the threat at moderate severity levels and much less accurate at weak severity levels. Additionally, the model performed significantly better when the duration of threat was longer (30 years or more) and when populations were not simultaneously affected by multiple threats.

The model created with real data correctly identified the presumed threat for all four case studies. The exploitation model received 100% support for bluefin tuna and 97% support for Atlantic cod, and the habitat loss model received 99% support for red grouse and 70% support for the Eurasian skylark, the latter of which increased to 92% with the incorporation of additional outside data.

In conclusion, the authors state that their findings “suggest that patterns of population decline embodied in long-term census records can provide valuable information for threat diagnosis.” They note that their model “would be especially beneficial for populations wherein the most effective conservation or recovery action requires knowing which threat… is the dominant driver of the population’s decline”. But they suggest that their model could also be used for scenarios involving multiple simultaneous threats if information from additional sources is included.

Additionally, although the method frequently failed to identify weak threats, the authors point out that weaker threats are much less urgent than moderate or severe threats, which the model often accurately identified. While the tool requires further refinement and is not yet available, the authors indicate that they will continue to develop and test the model. They express hope that it could ultimately “greatly assist conservation organizations in documenting threatening processes and planning species recovery.”