Northern Prairie Wildlife Research Center
|Fig. 2. Statewide trends in carnivore scent-station indices for Minnesota, 1986-93. Results are shown for stations and for lines of 10 stations placed at 480-m intervals. Annual means are weighted averages of means for sections, with weights proportional to the geographic area of sections.|
Our regression analysis resulted in estimates of trend (increasing or decreasing) and associated P-values that readers should use to judge statistical significance for themselves. Weighted statewide station indices increased from 1986 to 1993 for wolves (F1,6 = 37.56, P < 0.001), coyotes (F1,6 = 13.64, P = 0.01), red foxes (F1,6 = 11.41, P = 0.01), raccoons (F1,6 = 37.56, P < 0.001), and bobcats (F1,6 = 4.23, P = 0.09), but declined for skunks (F1,6 = 2.27, P = 0.18). However, these statewide analyses obscured differences in section trends for coyotes (F8,54= 2.18, P = 0.04) and raccoons (F8,54= 2.35, P = 0.03), but not other species (P ≥ 0.13) that would have been significant at conventional probability levels (P = 0.05). Station indices for coyotes declined in the West Superior and Peatland sections but increased elsewhere (Table 1). For raccoons, station indices increased in all but the Peatland section, where they declined (Table 1).
Weighted statewide line indices increased for wolves (F1,6 = 20.79, P < 0.01), coyotes (F1,6 = 4.23, P = 0.09), red foxes (F1,6 = 37.56, P < 0.001), raccoons (F1,6 = 16.62, P < 0.01), and bobcats (F1,6 = 4.80, P = 0.07), but declined for skunks (F1,6 = 3.29, P = 0.12). Statewide analyses obscured statistically significant differences in section trends for skunks (F8,54 = 2.24, P = 0.04) and raccoons (F8,54 = 2.07, P = 0.05) but not other species (P ≥ 0.23). Line indices for skunks increased in the West Superior, North Superior, and Driftless sections, but declined elsewhere (Table 1). Raccoon indices declined in the Peatland section but increased elsewhere (Table 1). Tests could have failed to detect important nonmonotonic trends, so we checked statistical results by plotting index values against time. Undetected trends were not evident in the Minnesota data.
Table 1: Monotonic trends (increasing  or decreasing ) in station and line indices by section of Minnesota, 1986-93, for species showing differences in trends among sections.
|Station index||Line index|
For all species except raccoons, line indices were lower and multiple visits per line more frequent than expected (P < 0.001) if visits to stations occurred independently at a constant rate (Table 2). Thus, for most species, the binomial distribution was not an appropriate model for visits to stations. For most species, average squared differences between results for stations increased with the distance between stations until stations were separated by 2.0-2.5 km, but they seemed to decrease again at distances >3000 m (Fig. 3). Similar results for closely spaced stations are indicative of multiple visits by individual carnivores or a tendency for closely spaced stations to be placed in similar habitats. Stations near the opposite ends of lines, however, also produced surprisingly similar results. We could not determine the cause of this phenomenon. Locations of stations within lines did not seem to affect visitation rates; other possible explanations include observer bias and systematic, short-scale variation in use of the landscape by carnivores.
Table 2. Results of chi-square goodness-of-fit tests for independence of visits by carnivores to scent stations within scent-station lines, and for independence of visits to scent-station lines within sections of Minnesota, 1986-93.
Whether or not a line received 0 or ≥1 visit did not appear to depend on results for other lines in the same section: goodness-of-fit tests failed to reject the hypothesis of independence for any species (P ≥ 0.17; Table 2). Fisher's inverse chi-square test failed to reveal any deviation of P-values from uniformity (X212 = 8.18, P = 0.84), providing further evidence the binomial distribution was a reasonable model for numbers of lines visited within sections.
|Fig. 3. Indicator variograms for detections at scent stations within lines in Minnesota, 1986-91. Variograms are restricted to lines with ≥2 stations visited within the same year.|
Abundance explained 73% of variation in inverse visitation rates of bobcats on Cumberland Island (F1,12 = 32.52, P < 0.001). However, within time periods, visitation rates did not increase with population size (Fig. 6). A model that adjusted for differences between time periods explained 79% of variation in inverse visitation rates (F1,12 = 45.69, P < 0.001), left no statistical evidence for an additional effect of abundance (F1,11 = 0.39, P = 0.55), and produced normally distributed, homoscedastic, independent residuals. Moreover, we found nearly significant differences between time periods that could not be explained by changes in abundance (F1,11 = 3.98, P = 0.07). Assigning data to 1 of 2 time periods was superior to regressing them on abundance with respect to both explanatory value and adherence to model assumptions. Both models were biologically plausible.