A BAYESIAN APPROACH TO MOOSE POPULATION ASSESSMENT AND HARVEST DECISION

Abstract

Assessments of ungulate populations need to be expressed in probabilistic terms to convey uncertainty about key parameters and the consequences or “risks” of alternative policies for harvest. The use of Bayesian estimation and risk assessment is described and applied to a declining moose (Alces alces) population in north-eastern British Columbia. A simple balance model was used to calculate posterior distributions of probability for population size of adults at the start of the assessment period and recruitment rate of calves. Model inputs included two mid-winter surveys of absolute abundance, a herd-composition survey and a harvest/effort index for adult bull moose. Calf recruitment was positively density dependent at moderate to high densities of moose. Probability distributions were estimated for moose population size in 1988 (95% CI’s: 7,655 - 10,550) and 1995 (95% CI’s: 3,805 - 5,980). Risk functions were used to determine the probability of obtaining various adult sex ratios after 3 years of additional bull harvest. Some of the limitations of the moose assessment were that not all of the model parameters were treated as uncertain, that deterministic assumptions about population dynamics were used, and that the behaviour of this predator-ungulate system at low densities of moose was poorly understood. This can bias the degree of certainty in estimates of parameters and risk associated with various harvest policies.