Northern Prairie Wildlife Research Center
Upland areas without tall shrubs or trees were searched by two investigators using vehicles to pull a chain drag (similar to Higgins et al. 1977). Habitats with trees or other features that precluded use of the chain drags were searched by investigators on foot beating vegetation with switches. Nest searches were conducted between 06:00 and 14:00. Each nest was marked for relocation and the age of the clutch was determined by candling all eggs (Weller 1956). Most nests were visited every 7-10 days until the clutch hatched, was destroyed, or was abandoned.
Nests were classified as successful if at least one egg hatched. Unsuccessful nests were those in which all eggs were destroyed or abandoned. Nests abandoned because of apparent investigator influence (3%) were not used.
Predator species targeted for assessment were five carnivores (coyote, Canis latrans; red fox, Vulpes vulpes; raccoon, Procyon lotor; striped skunk, Mephitis mephitis; and badger, Taxidea taxus), one rodent (Franklin's ground squirrel, Spermophilus franklinii), and two corvids (American crow, Corvus branchyrhynchos, and black-billed magpie, Pica pica). These species were selected because each occurred on at least some of the study areas and is a major predator of duck eggs (Kalmbach 1937; Jones and Hungerford 1972; Sargeant and Arnold 1984). These predators were believed to be responsible for virtually all depredations of duck eggs in the study. Indices of mink (Mustela vison) and long-tailed weasel (Mustela frenata) indicated that these species were scarce or absent on all study areas, so we did not consider them in the analysis.
Indices of carnivore activity
Investigators searched for tracks of coyote, red fox, raccoon, striped skunk, and badger during early to mid May and mid to late June on each quarter of each study area, except for a few (< 1 %) quarters where access was denied. Searches were conducted by personnel who were experienced in track identification. During each search period investigators spent up to 30 min on each quarter examining the best sites (e.g., field edges, marsh edges, trails) for tracks of each species. Investigators categorized the abundances of tracks of each species on each quarter with a value ranging from 0 for absent to 3 for abundant; specific guidelines were given to investigators. If conditions for finding tracks on a quarter were poor (e.g., recent rains had erased tracks, soil was not conducive to forming recognizable tracks), we believed the observed indices were biased low. When these conditions occurred we increased an activity code of 1 to 2, and of 2 to 3; no changes were made for codes of 0 (absent) or 3 (abundant). For each quarter in a year, we used the maximum activity index from the two search periods.
Index of Franklin's ground squirrel activity
In early July, four live traps were set for 1 day in a site with brush or dense upland vegetation, or another site if no such vegetation was available, in each square mile (2.59 km²) along and within 200 m of the 16-km road that ran the length of each study area. A site was defined as an area up to 2 ha in size. Traps were moved at least 200 m each day to another site in the same square mile; this process continued until five sites were covered in each square mile. If trapping was interrupted by inclement weather, it was resumed when conditions improved. Traps were single-door, wire-mesh live traps (25 × 81 cm) baited with sardines and set a minimum of 20 m apart. Traps were set each day before 09:00 and checked the next morning before 09:00. Franklin's ground squirrels and other animals caught were released at capture sites when traps were checked.
We calculated for each square mile the fraction of the 20 trap-days that resulted in capture of Franklin's ground squirrels. This value was apportioned to the four constituent quarters on the basis of the area of ground squirrel habitat (woody, brushy, or other dense vegetation) in each quarter.
Indices of crow and magpie activity
The corvids were surveyed by a line transect method, with supplementary data obtained from searches for crow nests. The transect survey involved an investigator driving the 16-km center road and stopping at the midpoint of each quarter to look for crows and magpies in that quarter during a 1-min period. In addition, investigators recorded any crows or magpies seen while driving between stops. Birds observed moving from one quarter to another were counted only in the quarter in which they were first seen. Most counts were made between 10:00 and 15:00; periods of inclement weather (i.e., fog, rain, snow, or winds in excess of 28 km/h) were avoided. Three counts were made on different days during each of four sample periods (early to mid May, late May to early June, mid to late June, and early July) for a total of 12 counts on each study area.
In addition, all personnel working on each study area recorded the locations of tended crow nests. In 1984 and 1985 a systematic search was also conducted in late May and early June of nearly all wooded habitat on each study area. Locations of tended magpie nests were recorded when found, but because it was difficult to determine which nests were tended, we did not use that information.
For each quarter we calculated the fraction of the 12 counts on which crows or magpies were seen. For crows, we added to this fraction the number of tended crow nests found in the quarter to obtain the index of crow activity. This adjustment was made because a tended crow nest in a quarter greatly increased the amount of crow activity there.
Weather and habitat information
Precipitation statistics were obtained from Atmospheric Environment Service, Environment Canada, for recording stations nearest to study areas. We used the precipitation received during 1 April-30 June, the primary waterfowl breeding period. Wetland basins on one-fourth of each study area were surveyed in May of each year by the United States Fish and Wildlife Service's Office of Migratory Bird Management. A basin was classified as wet if water was visible. The wetland indices used were the percentage of seasonal wetlands containing water and the percentage of semipermanent wetlands containing water (type III and type IV wetlands of Shaw and Fredine, 1956). Because measures of precipitation and wetlands pertained to an entire study area, values were the same for each quarter in a study area.
Smoothing predator indices
Because most predators, including the Franklin's ground squirrel (Choromanski-Norris 1983), range widely, the risk of predation by a particular species for a nest located in a quarter is best reflected by a weighted average of activity indices for the species from that and adjoining quarters. We obtained a smoothed index for a quarter as the weighted average of the indices for that quarter and for the three or five adjoining ones. Weights were calculated to reflect inversely the average distance between points in two quarters. Figure 2 shows weights used for an interior quarter and for one at the end of a study area.
|Fig. 2. Weights used in smoothing indices of predator abundance for a quarter at the end of a study area (left) and for an interior quarter (right). Weights are based on the average distance between points in target quarter (bold line) and adjoining quarters.|
Daily nest predation rates
We calculated the daily predation rate of upland duck nests, all species combined, according to the Mayfield (1961) method with the 40% modification of Johnson (1979). We considered as depredated only those clutches destroyed by predators or abandoned because of predation. We determined the number of days clutches were exposed to destruction and the number of clutch destructions occurring during those days. The ratio of destructions to exposure gave the daily predation rate. The number of exposure days used in calculating the daily predation rate was retained for use as a weighting factor in analysis. Nests were segregated by quarter and according to whether they were found on the first or second search (conducted early May through early June) or on the third search (mid to late June), because we anticipated differences in predation rates between early and late nests.
We conducted univariate correlations and analyses of variance, using SAS for personal computers (SAS Institute Inc. 1985). Correlations were calculated among explanatory variables (the eight predator indices, the precipitation index, and the two wetland indices) and between the daily predation rate of nests and each explanatory variable. Because sample sizes were so large, many small correlation coefficients were statistically significant, so we considered important in the correlation analysis only those that exceeded 0.10 in absolute value. Sample units were quarters, and for analyses involving daily predation rate, observations were weighted by the number of exposure days. A linear model was fit with SAS PROC GLM (SAS Institute Inc. 1985) using all explanatory variables, and nonsignificant (P > 0.10) ones were successively removed based on an examination of the reduction in residual sums of squares. Because of substantial inter-correlations among some explanatory variables, some uncertainty exists about which combination of variables is the most meaningful. We present results of both the initial fitting, with all explanatory variables, and the final fitting, with a minimum of significant ones included. Separate analyses were done for early and late nests.