Northern Prairie Wildlife Research Center
All data were examined and records of unusual occurrences (e.g., sighting of locally rare species) were excluded unless discussion with the observer(s) convinced us of their authenticity. All calculations of species abundance are based on portions of study areas that were actually searched. We averaged annual results for study areas with >1 year of data to obtain single index values of abundance for each species and species-group. Averaging of annual results was necessary to avoid the problem of nonindependence of repeated observations in the same study area. Correlation and other statistical analyses were performed on averages of the annual values of >1 year of data per study area and on the single annual value of 1 year of data per study area. The term "study area- year" is used to refer to data gathered in one study area during 1 year. Two-sample l-tests were used to compare means (Snedecor and Cochran 1980) between data sets from various geographic areas. Before performing such tests, we verified equal variances (Snedecor and Cochran 1980). Chi-squared tests of homogeneity were used to compare percentages between data sets from various geographic areas (Snedecor and Cochran 1980). Differences were declared significant if P is less than or equal to 0.05. We provide standard deviations (SD) for means, but not for percentages (*) where sample sizes (N) are provided and readers can calculate values.
To assign abundance, we established a range of index values for each species and species-group from our data and from published information and adjusted the index to a scale of 0-100. The abundance scale was partitioned as follows: 0 = undetected, 1-5 = scarce, 6-25 = uncommon, 26-75 = common, 76-100 = numerous. All ratings of undetected were elevated to scarce or uncommon if other evidence revealed presence of the species. The rating of scarce was for species detected only once or twice and at only one or two locations. The albeit subjectively established abundance categories provided criteria for quantitatively describing species abundance in familiar terms.
Partitioning of data sets for comparisons was primarily by habitat zone (prairie versus aspen parkland), country, state or province, and combinations of these. This partitioning reflects major habitat subdivisions and provides tailored data summaries for government agencies responsible for management of waterfowl and predators in the prairie pothole region.
We investigated possible effects on abundance of certain carnivores and on spacing of nests by some avian predator species (certain raptors, American crow) from interspecific relations. For selected pairs of carnivore species, we used correlation analysis (Snedecor and Cochran 1980) to examine the relation between percentage of quarter sections of each study area with tracks of one predator species and percentage of quarter sections of each study area with tracks of the other species. Negative correlations were interpreted as indicating avoidance of one species by the other and positive correlations as indicating interspecific tolerance or attraction of at least one species by the other. In each study area, we examined the relation between the proportions expected by chance and observed proportions of occupied nests of selected pairs of avian predator species that were near (less than or equal to 0.5 km) each other. We assumed that adult birds tending nests within 0.5 km of each other had overlapping home ranges and high potential to interact, and that if two species avoided nesting near each other, at least one of the species tended to exclude the other. Conversely, two species nesting near each other inferred at least one of the species benefited or that local habitat features facilitated nesting in close proximity.
We used SAS software (SAS Institute Inc. 1988) to simulate expected distributions of minimum distances between occupied nests for each combination of pairs of avian predator species. We examined whether nest locations of the species most likely to be repelled or attracted (affected species) were independent of those of the other species (effecter species). Most raptors are portrayed as interspecifically aggressive (McInvaille and Keith 1974; Schmutz et al. 1980; Rothfels and Lein 1983), but no clear dominance hierarchy has been established among species we tested. Therefore, we labeled the later nesting species of each pair as the affected species. For tests of raptors and American crows, we labeled the American crow the affected species, even though American crows have been known to depredate clutches of Swainson's hawks (Dunkle 1977). For each pair of species, we compared the expected with the observed number of occupied nests of affected species that were less than or equal to 0.5 km of occupied nests of effector species. Our rationale was that, if affected species avoid nesting near the effecter species, fewer than the number of expected nests are within 0.5 km of each other. Similarly, if more than the expected number of nests are within 0.5 km, the affected species is selectively nesting near the effecter species.
The simulations resulted in two-by-two contingency tables for each species pair in each study area with nests of both species. The results in each study area were combined to assess the tendency of the affected species to avoid or be attracted by the effector species. We used the Mantel and Haenzel method (Fleiss 1973), which also provided a test for how consistent the tendency was by pairs among study areas. In addition, the odds ratio (Fleiss 1973) compared the expected and observed chance of a nest being within 0.5 km of a nest of the other species. Odds ratios significantly >1.0 provided evidence that the affected species selectively avoided nesting near the effecter species, and ratios significantly <1.0, that the affected species selectively nested near the effecter species.
We also examined intraspecific spacing of nests of certain raptors and the American crow. To do this, we compared the expected with the observed number of nests of a given species within 0.5 km of a conspecific's nest and assumed that all breeding pairs positioned nests independently of one another. Simulations were again used to generate the expected distributions of minimum distances.