Estimating total human-caused mortality from reported mortality using data from radio-instrumented grizzly bears

 Tracking mortality of the Yellowstone grizzly bear (Ursus arctos horribilis) is an essential issue of the recovery process. Problem bears removed by agencies are well documented. Deaths of radiocollared bears are known or, in many cases, can be reliably inferred. Additionally, the public reports an unknown proportion of deaths of uncollared bears. Estimating the number of non-agency human-caused mortalities is a necessary element that must be factored into the total annual mortality. Here, we describe a method of estimating the number of such deaths from records of reported human-caused bear mortalities. We used a hierarchical Bayesian model with a non-informative prior distribution for the number of deaths. Estimates of reporting rates developed from deaths of radio-instrumented bears from 1983 to 2000 were used to develop beta prior probability distributions that the public will report a death. Twenty-seven known deaths of radio-instrumented bears occurred during this period with 16 reported. Additionally, fates of 23 radio-instrumented bears were unknown and are considered possible unreported mortalities. We describe 3 ways of using this information to specify prior distributions on the probability a death will be reported by the public. We estimated total deaths of noninstrumented bears in running 3-year periods from 1993 to 2000. Thirty-nine known deaths of non-instrumented bears were reported during this period, ranging from 0 to 7/year. Seven possible mortalities were recorded. We applied the method to both sets of mortality data. Results from this method can be combined with agency removals and deaths of collared bears to produce defensible estimates of total mortality over relevant periods and to incorporate uncertainty when evaluating mortality limits established for the Yellowstone grizzly bear population. Assumptions and limitations of this procedure are discussed.