Northern Prairie Wildlife Research Center
We collected adult, female striped skunks annually during March-July 1979-1981 and 1987-1991. Collections were from scattered sites (each ≤10.4 km²): 12 in Barnes, Griggs, Kidder, MacIntosh, and Stutsman counties of eastcentral North Dakota; seven in Becker and Grant counties of westcentral Minnesota. Collections were made during 1 year at 16 sites and during 2 consecutive years at three sites. Sites in North Dakota were in prairie, and sites in Minnesota were in prairie-forest transition zone. Habitats of all sites included idle grassland managed for wildlife and farmland managed intensively for cereal grains, sunflowers, and livestock. The climate in this area was continental with warm summers and cold winters. Snowmelt generally occurred during mid-Marchearly April.
Most skunks were trapped; a few were dug from dens or found dead on roads. Trapping at each site began in March-April and, once initiated, continued until late June-July to obtain as many skunks as possible (Greenwood, 1986). Skunks captured alive were dispatched by shooting or lethal injection. Carcasses were frozen and necropsied later. Age of each skunk was estimated to the nearest year by counting annuli in canines (Johnston and Watt, 1981), and animals were grouped into two age classes (class 1 = 1 year old, class 2 = ≥2 years old). Grouping reduced errors in estimated ages and statistical problems associated with small samples. Errors in estimation of age tend to increase with an animal's age (Johnston et al., 1987).
For all females collected, we determined those that were pregnant or parous in the year collected by macroscopic examination of uteri for embryo swellings or dark implantation sites (Gilbert, 1987). We counted all visible embryos in all pregnant females and distinguished live embryos from those that appeared to be resorbing. The live-embryo count was used as the estimate of potential litter size at parturition.
For a subsample of females collected in 1979-1981 and 1990, we estimated implantation date, determined sex ratio of embryos, and derived a second estimate of potential litter size at parturition based on number of corpora lutea. Reproductive tracts from females with visible embryo swellings were fixed in AFA (50% ethanol, 38% water, 10% formalin, 2% acetic acid). Later, we sectioned fixed ovaries and macroscopically counted numbers of corpora lutea (Gilbert, 1987). Age, in days, of each fixed embryo believed to be alive when collected was estimated by measuring its crown-rump length (Gilbert, 1987).
To estimate age of embryos we first fit a linear-regression model,
log(crown-rump length + 1) = log a + b age,to mean crown-rump lengths provided by Rakowski (1972) for known-age embryos of striped skunks. The model provided a good fit between actual measurements and predicted values (r = 0.997). Estimated values were = 0.933514 and = 0.054824. We then solved the equation
to obtain an estimated age for each embryo. We calculated a median embryo age for each pregnant female and backdated from collection date to estimate implantation date. Use of the median instead of the mean reduced the effect of aberrant measurements (e.g., runts). Sex of fixed embryos ≥25 mm long was determined from position of genitalia.
For pregnant skunks from all years, we used two-way analysis of variance (ANOVA) to assess effects of female age class, year, and their interactions on estimates of litter size based on counts of live embryos. Individual females served as the sampling units.
For pregnant skunks from the subsample, we used two-way ANOVA to assess effects of female age class, year, and their interactions on implantation date and sex ratio of embryos. We used three-way ANOVA to assess effects of age class, year, estimation method (live-embryo count or corpora-lutea count), and their interactions on litter size. Because both methods of estimation were used on each adult female in the subsample, the design structure was a split-plot type (Milliken and Johnson, 1984). Individual females served as the whole plot. For each female, the method used to estimate litter size was the subplot.
Because samples were not equal for all combinations of treatments, we used least-squares means and their standard errors (Milliken and Johnson, 1984:132) to estimate population means. The least-squares means are unbiased estimates of population means. Fisher's protected least significant difference was used to separate least-squares means following significant effects in all ANOVAs (Milliken and Johnson, 1984). We used the general-linear model procedure of SAS (SAS Institute, Inc., 1989) to conduct all statistical analyses. Tests were considered significant at P ≤ 0.05.