Northern Prairie Wildlife Research Center
I drew a 9.7-km (6 mi) diameter circle around each selected WPA. I hoped that the area within circles would be small enough to yield study fields with roughly similar complements of mammalian predators, yet large enough to contain a representative group of CRP fields from which to sample. In each circle I mapped and numbered all CRP fields of at least 16 ha (40 ac) and enrolled in the program for at least two years. I drew a random sample of usable CRP tracts from the circles to obtain an area of CRP cover approximately equal to the area of WPA cover in each physiographic region. I defined usable CRP tracts as those not more than 1.6 km (1 mi) from a semipermanent wetland (23) and on which I could obtain permission to work for three years. All plots in the Great Plains were in McIntosh and Stutsman counties, ND; all plots in the Central Lowlands were in Clay and Ottertail counties, MN (Figure 1).
I used the nest searching technique described by Klett et.al. (12). Two teams of two or three persons searched each field three times per season with a chain drag towed between two Jeeps. Search periods were May 1-21, June 1-21, and June 22-July 14. Search hours were 7 a.m. to 1 p.m. daily, except during the third search, when we sometimes searched until 4 p.m., assuming that the hens would be more reluctant to leave nests during warmer weather. Nests were marked with flagged willow switches and checked on later visits to determine fate.
Numbers and sizes of fields searched each year changed slightly because of inadvertent mowing, landowner decisions to change CRP practices, or access problems (Table 1). Planted nesting cover on WPAs was well-established stands of native or introduced grasses or mixtures of introduced cool-season grasses and legumes. CRP fields in the central lowlands had well-established stands of legumes or mixtures of introduced grasses and legumes. Most CRP fields in the Great Plains were similar although two had hilltops or slopes nearly devoid of vegetation or dominated by annual weeds throughout the study. In a few CRP fields, landowners clipped small patches of perennial weeds to prevent their spread.
|Table 1. Number of fields and areas searched for duck nests, CRP study, 1989-1991.|
|Number||Total Area (ha)|
I defined a nest as an excavated depression with one or more eggs or young for the year. I considered a nest in the laying stage usable for calculating modified Mayfield nest success (10) if the number of eggs was recorded and revisits showed an increase in clutch size. Nests in the incubation stage were usable if the estimated number of incubation days was recorded and revisits showed advanced incubation of one or more undamaged eggs. I considered nests in either stage unusable if damaged by observers or if the hen was not flushed during one or more revisits, and clutch size or incubation stage did not increase.
A three-way ANOVA was used to compare daily survival rates (DSRs) (10) of clutches of duck eggs between physiographic regions between covers (WPA and CRP), and among years (1989, 1990, and 1991). Each of these three factors involved a different-sized experimental unit. The unit for physiographic region was a study area or circle, for cover an individual field, and for year an individual field during a particular year. The three experimental units required an ANOVA model with three error terms. Furthermore, some fields were not searched all three years or were searched and no nests were found, yielding an unbalanced design; therefore, the techniques of Milliken and Johnson (17) were used for the analysis. Computations were made by the GLM procedure of the SAS Institute (20).
I compensated for differences in precision of DSR estimates among fields by weighting the estimate for all clutches in each field by exposure days (i.e., number of days a clutch survives while under observation) with the weighted least squares technique of Snedecor and Cochran (22). Data were too sparse to permit such comparisons among individual species. Unbiased estimates of the marginal means of the DSRs and their standard errors were computed from the LSMEANS statement of the procedure. Nest success was calculated by raising the DSR estimates to the 34th power because 34 days is the average number of days elapsed from the onset of laying until hatching (12).
Data was used from this study and the Northern Prairie Wildlife Research Center's nest file to compare duck nest success in CRP cover with success on WPA cover during the pre-CRP era. I also compared success on WPAs during the pre-CRP era with that on WPAs during 1989-1991. DSRs for these analyses were calculated by combining nest file data for mallards, northern pintails, blue-winged teals (A. discors), northern shovelers (A. clypeata), and gadwalls (A. strepera) from the four counties searched during my study. Because the sampling structure of the nest file sample is largely unknown, DSRs were calculated for each year and cover type from the nest file data and from my data. Two two-way weighted ANOVA's were performed on these data. The first compared success on WPAs during 1989-1991 (this study) with that on WPAs during 1979-1988 (nest file), and the second compared success on CRP cover during 1989-1991 (this study) with that on WPAs during 1979-1988 (nest file data). Effects in these ANOVAs included period, physiographic region, and period by physiographic region interactions. The weighted least squares technique described above was used with weights equal to the number of exposure days. Least squares means were used to estimate marginal means.
We measured height-density of vegetation during June of each year at 25 stations along transects bisecting the long axes of all fields using apparatus described by Robel et al. (19). Four measurements were taken at each station and were averaged prior to analysis. Results were tested with two-way weighted ANOVAs to compare cover height-density measurements among years, between cover types, and between landforms. Orthogonal polynomials (22) were used to test for annual trends within landforms. We examined possible relations between duck nest success and vegetation height-density by calculating weighted correlation coefficients for each plot, with weights equal to number of egg exposure days for each field. Data were treated separately by cover type for each year and for pooled cover types.