Misleading biases in methods for estimating wolf abundance using spatial models

Abundance, the primary criterion used for management and conservation decision-making, is very difficult to estimate for carnivores, especially highly mobile, wide-ranging wolves. Due to their economic and ecological importance to game management, society, and ecosystem health, reliable methods are necessary for wolves. The recently developed integrated patch occupancy model (iPOM) claims to provide improved accuracy of carnivore abundance at finer scales while reducing cost. We evaluate its input data models, potential biases, precision, sampling design, model coherence, validation, and reproducibility. Testing their method’s sensitive and vital assumptions to estimate the area occupied by wolf territories revealed that three sources of bias: false positive errors, closure violations, and resolution, combined to cause a substantial overestimation of abundance. The crucial confirmation step to independently determine individual wolf territorial centroids was eliminated, leading to a severe overestimation due to three sources of double-counting errors and mortality that occurred just before and during the survey period. iPOM lacks an inferential sampling design and relies upon a flawed survey of hunters’ inadvertent recollections of wolf sightings. It also includes a static and deficient covariate model, which limits the ability to correct for the sources of overestimation bias. The variance method used to report a confidence interval is incorrect and omits multiple components of variation, resulting in substantial underreporting of uncertainty. With many recommendations for improvement, we conclude that iPOM produces unreliable predictions, is irreproducible, and is misleading by reporting accurate and precise abundance predictions when abundance is severely biased (overestimation) and imprecise.