Northern Prairie Wildlife Research Center
Our goal each year was to obtain data from ≥10 study areas, with half of the areas occupied exclusively or predominantly by coyotes (coyote areas) and half occupied exclusively or predominantly by red foxes (fox areas). To accomplish this, we established a pool of ≥30 candidate study areas each year from information provided by managers. Criteria for candidate study areas were (1) each would be ≥60 ha with ≥40 ha of upland cover that could be searched with vehicles for nests; (2) there would be no manipulation of vegetation in uplands searched for nests during the evaluation period; and (3) there would be no organized predator removal within 5 km during the evaluation period.
We also established criteria to increase the probability that each study area would be occupied by different individual canids. Coyote areas had to be ≥10.5 km apart and fox areas had to be ≥5.4 km apart, unless separated by a major water barrier, which we assumed would separate family territories (Sargeant 1972, Sargeant et al. 1987). These distances are the diameter plus 2 standard deviations of average home-range sizes of these species in the study region (Sargeant et al. 1987).
We randomly selected study areas each year from the pool of candidate areas on the basis of canid occupancy. Sargeant et al. (1993) showed that track surveys could be used to determine the presence of canid species in the PPR. Their results conformed to that expected from the spatial distribution of sympatric coyotes and red foxes (Sargeant et al. 1987, Harrison et al. 1989). Therefore, we determined the principal canid species of each study area from tracks. We conducted 2 systematic track surveys. To select study areas, we conducted a track survey of candidate areas in April-early May. We randomly ordered candidate areas to establish the sequence in which they would be searched for tracks. The first 5-7 candidate areas classified as coyote areas and the first 5-7 classified as fox areas became study areas. Candidate areas not meeting the criteria of a coyote or fox area (defined below) were rejected. The number of areas selected annually depended on resources available for data collection. After selecting areas and while nest success data were being collected, we conducted a second track survey in late May-June. If second survey results concurred with results of the first survey, we considered the area occupied by the canid species identified. If the results of the second survey indicated both species occupied the study area, we changed the status of the area to mixed canid. We used additional information from incidental observations of tracks on study areas during the field season and from sightings of canids on or within 1.6 km of study areas to support the final determination of canid status. We classified areas with repeated evidence of both species as mixed-canid areas for all analyses.
We modified track survey methods described by Sargeant et al. (1993). We first established 200 × 200-m plots throughout and around the outer perimeter of each candidate area and searched each plot ≤15 minutes for tracks. We used all-terrain vehicles to enable rapid assessment of each plot. We used only results from plots with ≥1 site suitable for finding tracks (e.g., muddy wetland edges, gravel road edges, crop field edges) in analyses. Our goal was ≥10 usable plots for each area during each survey.
To initially classify a candidate area as a coyote area we had to observe coyote tracks on ≥10% of plots and fox tracks on ≤10% of plots. To initially classify an area as a fox area we had to observe fox tracks on ≥20% of plots and dispersed throughout the candidate area and coyote tracks on ≤10% of plots. Criteria for identifying the principal canid occupying a candidate area differed for coyotes and red foxes because of differences in their territory size, movements, and densities. Coyotes, in contrast to foxes, usually have much larger territories (Sargeant et al. 1987, Harrison et al. 1989), do not travel as extensively throughout the territories each day (Sargeant 1972, Johnson and Sargeant 1977, Harrison and Gilbert 1985), and occur in lower densities (Sargeant et al. 1987, Harrison et al. 1989). Thus, coyote tracks are less abundant and less easily found in areas they occupy than are fox tracks in areas fox occupy.
To estimate duck nest success, we conducted 3 systematic searches of uplands on each study area at 3-week intervals from early May through June. We allowed a maximum of 1 day for each search period on each study area. Crews of 2-3 persons conducted nest searches using all-terrain vehicles or 4-wheel-drive jeeps to drag a chain to flush hens (Higgins et al. 1977). We recorded all information for each nest as suggested by Klett et al. (1986). We candled eggs to determine stage of incubation (Weller 1956). For relocation, we marked nests with a flagged willow stake placed 4 m away and plotted locations on aerial photos. We visited each nest at 7-10-day intervals until the nest was successful (≥1 egg hatched), the clutch destroyed, or the nest was abandoned by the hen. We did not include in analyses nests abandoned within an estimated 2 days of discovery or damaged by investigators.
We used multivariate analysis of variance (MANOVA) to test for differences in species composition of nests. Main effects tested were canid, year, and canid-year interaction. Dependent variables tested were percent mallard, northern pintail, blue-winged teal, gadwall, northern shoveler, and other (American wigeon [Anas americana], green-winged teal [A. crecca], and lesser scaup [Aythya affinis]).
Using the Mayfield method (Mayfield 1961) as modified by Johnson (1979), we calculated daily survival rates (DSRs) of nests. We combined data from nests of all duck species for each study area because of small sample sizes for individual species. The variance of estimated DSRs is inversely proportional to the number of exposure days (Johnson 1979). Therefore, because of inequity in number of exposure days among study areas, which influences precision of DSR estimates, we used weighted least-squares (Snedecor and Cochran 1980) to improve precision of estimates. We weighted each estimated DSR by the number of exposure days to estimate parameters for analysis of variance (ANOVA) models. We computed confidence intervals (95%) following Johnson (1979), using standard error values generated by the least-squares means statement of the GLM procedure (SAS Inst. Inc. 1990).
We used a 2-way factorial treatment structure with randomized design in analyses of DSRs. We used ANOVA to test for differences in DSRs among coyote and fox areas and to test effects of year and canid-year interaction. The design was randomized, except canid classification could not be assigned randomly to individual study areas. Except when we report nest success, we excluded mixed-canid areas from all analyses because they were not included in the hypothesis and because sample size was small. For ease of interpretation, we converted DSRs to nest success by raising the weighted least-squares means estimates of DSR to the thirty-fourth power ( days of laying plus incubation period for duck species in our evaluation; Klett et al. 1986).
We collected information on predator species responsible for depredations of nests by quantifying evidence found ≤3 m from each depredated nest, using procedures developed at the Northern Prairie Science Center. Evidence at nests depredated by foxes is distinctive because foxes usually remove and cache all eggs, leaving no shells (Kruuk 1964; Tinbergen 1965; Sargeant and Sovada, unpubl. data). Coyotes and nearly all other predators of duck nests in our study areas leave shell, whole eggs, or both at nests (Sooter 1946; Rearden 1951; Einarsen 1956; Sargeant and Sovada, unpubl. data). Using only nests with ≥6 eggs (large-clutch nests) on the last visit prior to hatch or prior to depredation, we examined occurrence of predation by foxes for each area with ≥10 large-clutch nests. We assumed depredation by foxes if no whole egg or eggshell was at the nest. We calculated percentages of total nests and of total depredated nests that foxes destroyed. We used ANOVA (least-squares means estimates, GLM PROC, SAS Inst. Inc. 1990) to test for differences between coyote and fox areas in percent of nests depredated by foxes.
We determined habitat composition (%) of a 10-km² block (approx) including and surrounding each study area (study area in center of block) with aerial photos and ground observation. We categorized habitat in each block as grassland, hayland, cropland, wetland, or right-of-way and farmstead. We conducted MANOVA to test for differences in the percent area of grassland, hayland, wetland, and cropland between coyote and fox areas.
We used observation surveys (Sargeant et al. 1993) to assess presence on each area of predator species with potential to prey on duck eggs (Sargeant and Arnold 1984, Sargeant et al. 1993). We monitored striped skunk (Mephitis mephitis), badger (Taxidea taxus), raccoon (Procyon lotor), long-tailed weasel (Mustela frenata), Franklin's ground squirrel (Spermophilus franklinii), American crow (Corvus brachyrhynchos), and black-billed magpie (Pica pica). Additionally, we used track surveys to assess presence of carnivores and live trapping to assess presence of Franklin's ground squirrels (Greenwood 1986, Choromanski-Norris et al. 1989).
We conducted observation surveys of each predator species during May-June. Each person who worked on a study area for >0.5 hours in a day kept a record of the number of 0.4-ha places where ≥1 individual of each predator species was sighted. We used these data to assess presence of predator species for each study area. We trapped for Franklin's ground squirrels in each area by setting 1 trap/4 ha of habitat searched for duck nests. Trapping was conducted for 2 days (24-hr periods) between 26 June and 27 July during the year that nest success in the area was being evaluated. With Chi-square analyses, we tested for differences in the proportion of coyote and fox areas in which each of the other monitored species was observed, detected by tracks, or trapped. We did not analyze species detected on <25% of both coyote and fox areas for differences in presence.