Counting Wolves for Decision-Making
Montana has recently adopted two new methods to determine the population size (abundance) of wolves for statewide management decisions, including where and how many wolves can be hunted each year. The methods are called POM (Patch Occupancy Model) and iPOM (Integrated Patch Occupancy Model) which is a combination of several sub-models including POM. But scientists have voiced skepticism over their use for management as both published and unpublished papers have pointed out many likely problems with these methods. To evaluate their utility for important management decisions, we examined the models’ accuracy and precision, using the State’s own methods, models and data. We evaluated the effects of one intuitively obvious problem and demonstrated that their spatial submodels are highly sensitive to the area that an individual wolf observation represents. Thus, their methods can overestimate wolf abundance by as much as two and half times larger than the actual number of wolves. This presents a difficult problem to fix and reveals other weaknesses in their overall approach. Among other potential problems is their use of hunter observations to confirm if a wolf observation is (1) from a territorial pack, and (2) representative of a much larger area that is fully occupied by a stationary wolf territory during the fall survey period. Furthermore, their method incorrectly overestimates the confidence (substantial underreporting of the confidence interval) of their estimate of abundance. Given the current design of POM and iPOM, there is no ability to detect change let alone determine wolf population size. Nonetheless, we recommend numerous ways to improve Montana’s methods in a detailed paper available on the internet for peer review and comment and briefly summarized below. After collaborative improvement by peers, the paper will then be submitted to a refereed scientific journal. Our findings are transparent and repeatable so that both scientists and practitioners can help improve how to reliably estimate wolf abundance—a species that has important economic costs and benefits.
Occupancy modeling (OM) is used within a multi-model method called iPOM used to estimate the number of wolves in Montana and make sound management decisions. OM uses individual observations of wolves from hunters and agency personnel to determine the area occupied by all territorial packs (not individuals) across Montana. This illustration depicts a realistic scenario where four territorial packs assumed to be stable are observed during the fall iPOM survey period. The four packs range entirely within an average territorial area of 600km² with an average pack size of 5 individuals. OM uses a simple 600km² grid overlay (depicted here) as their sampling unit. OM developed as a method to determine a species distribution and is not typically used to estimate population abundance. As such, OM labels an entire 600km² grid cell as fully occupied (see shaded grid cells) if any observation anywhere in the grid cell is reported. In this case, the total area occupied by assumed stable wolf packs equals 7,200km² when the actual area occupied by the four territories is 2,400km² which results in an overestimation bias that is three times larger than the actual value for area occupied, the number of packs, and overall estimate of abundance. For the entire statewide estimate of wolf abundance, our simulation testing of the iPOM using their models and their data, resulted in a severe overestimation bias: 2.5 times larger than the actual number. Researchers in Montana and elsewhere have all confirmed that the area occupied from OM is not the true occupancy, especially when the assumption of closure is violated, that is, there are no changes due to movements or demography (eg, births or deaths) during the fall survey. Wolves as entities are highly mobile and range over large areas. In iPOM, territories are assumed to be stable during the fall yet wolf packs during the fall are highly dynamic as (1) individuals shift spatially outside their normal territorial area in response to changes in prey and other factors, (2) undergo dispersal of young adults in the fall, and (3) dissolve (break apart) in a high percentage of cases when a breeding adult is killed. All three factors affect neighboring territorial packs. A simple case is illustrated on the far left where the remaining pack of 4 breaks apart after the breeder is killed with one leaving south of the area depicted. Another wolf is accurately recorded as a lone wolf not belonging to a pack and two siblings leave together and could easily be recorded as a pack in that cell which would lead to a further increase in the overestimation bias.