Northern Prairie Wildlife Research Center
In each study area I identified idle WPAs with well-established grassland comprising at least 10 ha. Idle WPAs were not subjected to disturbance from farming activities, although normal management activities continued, such as prescribed burning (1 field, 1 yr) and mowing to control weeds (2 fields, 3 yr). Furthermore, idle WPAs did not experience variation in densities of breeding birds associated with variation in grazing intensity (Kantrud 1981, Kantrud and Kologiski 1982, Renken and Dinsmore 1987, Bowen and Kruse 1993). From the identified WPAs, I selected six in North Dakota and five in Minnesota that were dispersed throughout the study areas. All 11 WPA fields were bordered by potential perch sites (e.g., trees, fences, power lines) for Brown-headed Cowbirds, and some, especially the native-prairie fields, had woody vegetation in the field. Minnesota fields had been seeded with several native warm-season grasses. North Dakota fields were either native prairie (four fields, invaded by cool-season exotic grasses) or had been seeded with cool-season grasses in the 1970s (two fields).
With the assistance of personnel in the Farm Service Agency (formerly Agricultural Stabilization and Conservation Service) and Natural Resources Conservation Service (formerly Soil Conservation Service), both agencies of the USDA, I identified the CRP fields in the vicinity of each selected WPA. I chose the CRP field closest to a selected WPA field that met four criteria: (1) it was not adjacent to a selected WPA field, (2) it comprised 10-16 ha, (3) was enrolled in the CRP prior to 1989, and (4) it had potential cowbird perches on at least one side. In a few cases I had to use a portion of a larger field because a field smaller than 16 ha was not available. All CRP fields were dominated by cool-season grasses.
The same WPA and CRP fields were studied in all 3 yr (1991-1993) except for one CRP field that was replaced by a nearby field in 1992 after the initial field was flooded in 1991.
Density was estimated from 1992 data with program DISTANCE (Buckland et al. 1993) using a polynomial function to model the detectability function. This program emphasizes estimation of parameters rather than tests of hypotheses. The 95% confidence intervals were tabulated for the density estimates. Densities are likely to be truly different if their confidence intervals do not overlap.
For most species, I report estimates from the first count each year because the second count may have been confounded by the inclusion of some fledglings. The exception was the Bobolink, for which I report data from the second count each year. Estimated densities of Bobolinks were notably higher in the first count. Bollinger et al. (1988) noted that Bobolink densities were easily overestimated with transect methodologies. To the extent that this overestimation was related to the frequent long-distance flights of males early in the nesting season, the estimates from the second count probably more accurately reflect the breeding densities for this species. The estimates were intended to represent species nesting in the study fields. Thus, I have not reported densities of Yellow-headed Blackbirds (Xanthocephalus xanthocephalus), which frequently were seen foraging in our study fields in Minnesota.
Nests were visited at intervals of 3-7 d. The final visit was shortly before fledging (i.e., 6-8 d old). Late nest visits (>8 d) were avoided so that field personnel did not induce nestlings to leave the nest prematurely and did not have to judge whether or not any young had fledged from an empty nest. Such judgment would have required analysis of the appearance of the nest and the parental behavior of the adults, both of which might have been subject to error.
Fledging success is the probability that a nesting attempt (i.e., at least one egg laid) will produce at least one fledgling. I estimated fledging success with the Mayfield (1975) method for the entire nesting cycle, that is, not analyzing egg and nestling stages separately. Altricial species that nest on the ground were pooled for statistical analysis, to maximize the power of statistical tests. Ground nesters probably share many of the same nest predators. To estimate fledging success, I raised the estimated daily survival rates of nests to the 25th, 27th, or 29th power (Ehrlich et al. 1988).
Field personnel assigned each nest failure to the most likely cause. If eggs or nestlings were missing or damaged in a manner consistent with predation, failure was attributed to predation. If the nest was unattended and there was no change in the number of host eggs between visits, failure was attributed to abandonment. Abandonment between the first and second nest visit was attributed to investigator disturbance unless cowbird eggs had been added, in which case abandonment was attributed to cowbird parasitism. Other abandonments were not attributed to a specific cause, although some may in fact have been due to cowbird parasitism (see Elliott 1978, Koford et al. in press).
I used a Summary procedure (SAS Institute 1987) to tabulate fledging success by species for each study area and field type. I used a General Linear Model procedure (SAS Institute 1987) to examine variation in daily survival rates of nests of altricial ground-nesting birds between field types (CRP vs. WPA), between study areas (North Dakota vs. Minnesota), and among years (1991, 1992, and 1993). The response variable was an angular transformation (Steel and Torrie 1980) of daily survival rates of nests, weighted by exposure days. Fields were the sampling units. I used a blocked design structure with pairs of fields in each study area as blocks (Milliken and Johnson 1984). A repeated-measures analysis of variance was conducted to account for the non-independence of the fields, which were sampled in all 3 yr (Milliken and Johnson 1984). Statistical significance was indicated by P < 0.05. Least squares means (Milliken and Johnson 1984, SAS Institute 1987) were used in the analysis because samples of nests were not available from all fields in all years. Least squares means are the expected values of class or subclass means that would be expected for a balanced design and may be substantially different from arithmetic means. Nests that had apparently been abandoned because of investigator disturbance (N=6) were not analyzed.