US Geological Survey


Date of this Version



Published in The JOURNAL OF WILDLIFE MANAGEMENT 67(2):289-298.


Scent stations usually are deployed in clusters to expedite data collection and increase the number of stations that can be operated for a given cost. Presumed benefits of cluster sampling may not be realized, however, unless cluster sizes are chosen with respect to sampling variation within and among clusters. To encourage and facilitate the use of efficient designs and reporting standards, we used data collected in Minnesota, USA, during 1986-1991 to (1) compare the performance of survey designs with various numbers of stations/cluster; (2) estimate relations between required sample sizes and visitation rates, changes in visitation rates, and error rates; and (3) compare 2 measures of carnivore response: proportions of scent stations (station index) and proportions of clusters (line index) visited by red foxes (Vulpes vulpes) and striped skunks (Mephitis mephitis). Despite broad ecological differences between the species, results were similar for foxes and skunks. Foxes visited 2-21% of stations and 15-84% of lines. Skunks visited 1-16% of stations and 3-54% of lines. Station and line indices were closely related (r2 > 0.86) and were similarly sensitive indicators of change in visitation rates. Low visitation rates greatly limited the potential usefulness of scent-station surveys because required minimum sample sizes increased exponentially as visitation rates decreased. For visitation rates below 5-10%, required minimum sample sizes were very large and difficult to anticipate. Relative to single-stage sampling, cluster sampling with 10 stations/cluster inflated sample variances, hence sample sizes required to achieve a fixed level of precision, by a factor of 1.6-2.2. Cluster sampling is advantageous only when cost savings permit increases in sample sizes that outweigh concomitant increases in sampling variability. Costs and sampling variation both should be considered when choosing survey designs, and designs should be evaluated and refined as data accumulate.