Northern Prairie Wildlife Research Center
|Table 1. Explanatory variable: mean, interquartile range (IQR), and coefficient of variation (CV) for indices of activity of eight predators, indices of two wetland types, and a measure of spring precipitation|
|Franklin's ground squirrel||0.011||0-0.008||232|
Several highly significant correlations existed among the predator indices (Table 2). The strongest was the negative relationship between fox and coyote indices (r = -0.51), a finding consistent with the spatial avoidance and agonistic behavior between these two canids (Dekker 1983; Voigt and Earle 1983; Sargeant et al. 1987a; Major and Sherburne 1987). Also, indices of the badger and skunk were negatively related (r = -0.46). Little is known of relations between these two species, but Sargeant et al. (1982) reported two instances of badgers preying on young skunks. Two other patterns were evident: (i) a general association among indices of the fox, skunk, and raccoon; and (ii) an association among indices of coyote, magpie, and Franklin's ground squirrel. Crow indices were positively related to those of the Franklin's ground squirrel, skunk, magpie, and coyote, and negatively related to raccoon and badger. Raccoons climb trees and may depredate crow nests. Badger indices were positively associated with those of no other species, but negatively associated with indices of the skunk, magpie, and crow.
Table 2. Correlation coefficients between explanatory variables
|Franklin's ground squirrel||0.05||-0.04||-0.00|
|NOTE: Significance levels for |r| are r0.05 = 0.06 and r0.01 = 0.08.|
Correlations with seasonal wetland counts were positive for the raccoon and fox, and negative for the coyote and magpie (Table 2). Correlations with semipermanent wetland counts were positive for raccoon and negative for crow and magpie. Correlations with precipitation indices were positive for coyote and skunk and negative for badger. These correlations between predator indices and pond counts or precipitation indices represent association; we mean to imply no cause-and-effect relationship. The precipitation index was positively correlated with indices of both seasonal (r = 0.35) and semipermanent (r = 0.49) wetlands. The correlation between the two wetland indices was 0.78.
Analysis of daily predation rate on early nests
There were 559 observations (quarters with nests found during the early searches) for this analysis. Included were 1791 nests, mostly mallards (Anas platyrhynchos), blue-winged teal (Anas discors), northern pintails (Anas acuta), and northern shovelers (Anas spatula) (Table 3). The average daily predation rate on early nests was 0.057, with an interquartile range of 0.033-0.158. Correlations between daily predation rate of early nests and the explanatory variables, weighted by exposure, were positive for fox, crow, and skunk, and negative for coyote and both wetland counts (Table 4). The analysis of variance for the full model was significant overall (F = 3.84, df = 11,547, P < 0.0001). Standardized regression coefficients are given in Table 4; these were obtained by subtracting from each explanatory variable its mean and dividing by its standard deviation. Regression coefficients were positive for indices of all predators except coyote and magpie, but because of intercorrelations among the explanatory variables, these coefficients were not all well-conditioned and are difficult to interpret. We successively eliminated nonsignificant explanatory variables, and ultimately obtained an equation involving only indices of fox, crow, badger, and the seasonal wetland count (Table 4). Coefficients of the three predator indices were positive and that of the wetland count was negative. The overall equation was highly significant (F = 8.77, df = 4,554, P < 0.0001) despite a small multiple correlation coefficient (R2 = 0.060). The badger index was marginally significant in the final equation, in spite of its minor importance in the initial one. Conversely, the skunk index was absent in the final equation, although it correlated rather strongly with daily predation rate. These discrepancies resulted from the intercorrelations among the explanatory variables; for example, the pronounced correlation between fox and skunk indices (r = 0.34) made it difficult to distinguish the separate effects of the two species.
|Table 3. Species composition of early and late clutches used in the analysis|
|Early clutches||Late clutches|
|Mallard (Anas platyrhynchos)||735 (41.8%)||294 (22.2%)|
|Gadwall (Anas strepera)||64 (3.6%)||288 (21.8%)|
|American wigeon (Anas penelope)||43 (2.4%)||88 (6.7%)|
|Green-winged teal (Anas crecca)||21 (1.2%)||23 (1.7%)|
|Blue-winged teal (Anas discors)||371 (20.7%)||308 (23.3%)|
|Northern shoveler (Anas spatula)||195 (10.9%)||139 (10.5%)|
|Northern pintail (Anas acuta)||352 (19.7%)||105 (7.9%)|
|Lesser scaup (Aythya affinis)||6 (0.3%)||76 (5.7%)|
|Other||4 (0.2%)||2 (0.2%)|
|Table 4. Correlation coefficients and standardized regression coefficients relating daily predation rates of early and late clutches to explanatory variables|
|Franklin's ground squirrel||0.04||0.013||-0.02||0.000|
|*P < 0.10.
**P < 0.01.
***P < 0.0001.
Analysis of daily predation rate on late nests
We had 424 observations available for this analysis, involving 1323 nests of mostly blue-winged teal, mallards, and gadwalls (Anas strepera) (Table 3). The average daily predation rate was 0.044, with an interquartile range of 0-0.146. Correlations between daily predation rate of late nests and the explanatory variables, weighted by exposure, were positive for fox and skunk, and negative for the wetland indices, especially that of semipermanent wetlands (Table 4). The analysis of variance for the full model was significant (F = 3.69, df = 11,412, P < 0.0001). Regression coefficients are given in Table 4. Regression coefficients for predator indices were positive except for those of badger and Franklin's ground squirrel. After successively eliminating nonsignificant explanatory variables, we obtained an equation involving only indices of fox, skunk, and the semipermanent wetland count (Table 4). Regression coefficients of both predator indices were positive, and that of the wetland count was negative. The overall equation was highly significant (F = 12.98, df = 3,420, P < 0.0001), with R2 = 0.085. The skunk index was retained in this equation despite its marginal significance (P = 0.11) because of its strong first-order correlation with daily predation rate (Table 4).