Power Analysis of Bait Station Surveys in Idaho and Washington PDF Format - [0.4MB]   View Slide Show

Power Analysis of Bait Station Surveys in Idaho and Washington March 2002 Clifford G. Rice, Jeff Rohlman, John Beecham, Steve Pozzanghera ABSTRACT: We evaluated statistical power for detecting trends of specified magnitude in visitation rate for black bear (Ursus americanus) bait stations in Idaho and Washington. We found evidence for lack of independence due to multiple visits when bait stations were 0.8 km apart and no evidence for this with stations 1.6 km apart. Based on the variability observed in Idaho, we assessed power for several sets of criteria. The minimum criteria were a relative decline of 50% over 3 years at a = 0.20 and power = 0.80. These criteria were met for many of the Idaho surveys, but were generally not met in Washington. More stringent criteria of a decline of 25% over 3 years at a = 0.10 and power = 0.90 were not met in either state. The initial visitation rate had a predominant influence on power, and in areas such as western Washington, where visitation was low but bear populations thought to be substantial, an effective monitoring program is contingent on improving the visitation rate through changes in survey methods. For long-term monitoring (5, 10, or 20 years), we estimated sampling requirements for declines of 50%, 25%, and 10% with a = 0.10 and power = 0.90 and estimated the costs of this sampling. Due to the inherent variability of bait station surveys, substantial sampling is required for detecting trends, and this method is likely to be cost effective only where visitation rates are relatively high. Although power analysis appears to be objective, determining the values for parameters used in its calculation is quite subjective and the results should be interpreted accordingly. INTRODUCTION For effective management of animal populations, some measure of the relative or actual abundance of the population is needed. This measure should be independent of harvest because of potential biases in harvest-based procedures and because not all populations are harvested. Black bears present many problems in deriving population estimates (relative index or actual population estimate) because of their relatively low densities, dense habitats, and solitary nature. Some approaches, such as capture-recapture or resight, are useful, but may be too costly to be conducted routinely over wide areas. Bait stations have been advocated as a potential solution to this problem (Carlock et al. 1983, Johnson 1990, Beecham and Rohlman 1994), and bait-station surveys have been conducted for a number of years in many states (e.g., Tennessee, Georgia, North Carolina, Idaho, Michigan, Wisconsin, Minnesota, and South Carolina; Johnson 1990). These surveys have been used to track populations over time, assess seasonal habitat use (Pelton 1984), and compare bear visitation according to site characteristics (Pitt and Jordan 1996). Despite this extensive use of bait stations, questions remain about the suitability of this technique for monitoring or comparing black bear populations. Much of the evidence supporting the view that bait station visitation rates reflect bear population levels has been anecdotal where differences in visitation rates were comparable to perceived population differences over time or between areas (Johnson 1982, Pelton 1984, Carlock 1986, van Manen 1988, Johnson 1990). Carlock et al. (1983) reported a positive correlation between visitation rate and mark-recapture population estimates over 5 years in Tennessee (r2 = 0.83), but a similar relationship was not evident in Minnesota (Garshelis 1990). There was, however a relationship between visitation and hunter success (Garshelis 1990). In Idaho, bait station visitation dropped from a 3-year average of 24% to 8% after 35 bears were removed from the Priest Lake study population of about 100 (Beecham and Rohlman 1994). At the Council study area (population =135, Beecham and Rohlman, Idaho Department of Fish and Game, unpublished data), Beecham and Rohlman (1994) documented the reversal of a positive trend in visitation following the removal of 19 and 33 bears in succeeding years. Reservations about the relationship between bait station visitation and population levels have focused on 3 areas: lack of independence, the effects of confounding factors, and site-specific influences. Lack of independence is addressed later in this paper. Confounding factors include food availability, weather, and timing of the surveys relative to annual climatic variation and plant phenology. These influences likely affect visitation rate independent of bear population levels. In some cases, these effects can be included in the analysis if their levels are measured, as Garshelis (1993) advocated for food availability. Otherwise, these factors add variability to the visitation rates, which reduces power for the analysis of changes in visitation rate and complicates interpretation of rates for individual years. If a point estimate is desired for visitation rate, it would be better to use a running average over several years as is done in Wisconsin (B.E. Kohn, Wisconsin Department of Natural Resources, Rhinelander, Wisconsin, personal communication, 1998). Site-specific influences likely affect visitation rates. For example, trail versus road (LeCount 1982), elevation, type of road, and forest type (Carlock 1986), and distance from roads and trails (van Manen 1988, Mantey, J. and D.A. Immell 1995. Influence of roads on black bear detections at bait stations. Department of Wildlife, Humboldt State University, Arcata, California, USA.). Because of these potential influences, bait station routes should be fixed between years and trend analysis for visitation rates should employ an analysis of covariance design. These influences also make comparisons among areas problematic unless they are included in the survey design as was done by Powell et al. (1996). We conclude that properly designed bait station surveys can provide useful information for trend analysis, but many extraneous factors add variability to the data. This added variability is likely to mask changes in visitation rate to due small changes in population density. The question is, to what extent is this true? What magnitude of change in visitation rate is likely to be detectable despite this variability? Statistical power analysis provides a framework within which to address these issues. Toward that end, we addressed the following questions: Were sampling levels employed by a monitoring program in Idaho and a pilot survey in Washington sufficient to detect specified magnitudes of change in the visitation rates? In cases where the sampling was deficient, what changes could be made to improve the survey’s performance?