Northern Prairie Wildlife Research Center
Breeding pair was the unit of record; the criteria used to indicate pairs, after Hammond (1969), were (1) observed pairs, (2) lone drakes, (3) lone hens only for diving ducks and only when males were not nearby, and (4) groups of up to five males or mixed sexes (except for northern shovelers and American wigeon, for which only pairs and lone males were counted). \
Censuses were ordinarily conducted between 0700 and 1700 hours. In most years, censuses were repeated during the breeding season. Estimates of mallards, northern pintails, northern shovelers, and canvasbacks were based on censuses made in early May; estimates of gadwalls, blue-winged teal, redheads, lesser scaup, and ruddy ducks were based on early June censuses; estimates of American wigeon and green-winged teal were taken as the greater of the counts recorded on these two census dates.
Pond counts. -- Wetlands in the Woodworth Study Area were surveyed several times each year (Higgins et aL, 1992). The wetland basins were classified as containing water if at least 5% of the area was inundated by 25 mm or more of water. For the analyses reported here, the wetland surveys conducted during 1 -15 May of each year are used, as they consist of measurements made closest to the time waterfowl settled in the area.
Weather data. -- Monthly precipitation and monthly mean temperatures were obtained from weather stations near the study site. These included Steele and Pettibone, North Dakota, as well as the Woodworth Study Area itself.
Conserved soil moisture index. -- The conserved soil moisture (CSM) index is a weighted average of precipitation during the 21 months preceding May of a particular year. It was developed by Williams and Robertson (1965) for agronomic purposes and popularized for waterfowl biologists byBoyd (1981), who suggested that it mirrored variation in wetlands.
Continental population indices. -- Annual indices of duck abundance were obtained by the U.S. Fish and Wildlife Service and the Canadian Wildlife Service from aerial surveys of the primary breeding ground of most duck species. The survey region encompasses about 3.03 million km2 in the north-central United States, the prairie provinces, the Northwest Territories of Canada, and Alaska. In typical years more than 80% of the North American populations of the duck species considered in this chapter breed in this survey region (Johnson and Grier, 1988). The survey utilizes aerial and ground surveys conducted each May, with adjustments made for visibility; the survey methods were described by Martin et al. (1979).
Waterfowl distributions and philopatry. -- Information on the overall breeding distribution tendency of species to return to the same breeding taken from Bellrose (1980) and Johnson and Gri both published and unpublished information o~ Grier (1988) also identified areas (strata as used previously described) according to the average der highest to the lowest quartile.
Analytic methods. -- The densities of the most common duck species were related to several explanatory variables, including the number of ponds in containing water during early May ("May ponds") ture (CSM) index, both of which reflect long-term more proximate climatic features were examined: precipitation in April and in May. Other explanato the index of continental population size and the worth Study Area during the previous year.
Statistical tools used [and SAS Institute, Inc. (1989, 1990) procedures] included Pearson product-moment correlation (Proc CORR), analysis of variance (Proc GLM), and stepwise regression (Proc REG), as well as descriptive methods (Proc UNIVARIATE).
In an effort to examine the suite of variables that might influence populations of ducks in the Woodworth Study Area, I attempted to model the numbers of each species there in relation to May ponds, CSM, mean temperature, and total precipitation in each of April and May, the count of each species in the previous year, and the continental population index. Five models were fitted for each species, which included, respectively: (1) only proximate habitat variables (May ponds, CSM, and temperature and precipitation values), (2) proximate habitat variables plus the previous year's count, (3) proximate habitat variables plus the continental index, (4) all variables mentioned above, and (5) the "best" model as determined by stepwise regression, with significance levels for including and deleting variables set at 0. 15. Significance of the multiple correlation coefficient was determined for all but the "best" model, for which the stepwise selection of variables renders the R2 values noncomparable.
We sought interactions among species above and beyond those related to similar responses by different species to abiotic or environmental features. This was done by calculating residuals from the best-fitting models for each species and relating them with regression to the counts of other species. Positive associations would be consistent with either the hypothesis that species actively associate with each other or the notion that the species respond similarly to certain environmental conditions that are not directly represented by the explanatory variables. Negative associations would be consistent with the hypothesis that the species compete with each other.