Northern Prairie Wildlife Research Center
Fixed-wing aircraft are used in the aerial counts and are flown along the transects at a height of 30-46 m (100-150 feet) above ground and at a ground speed not to exceed 169 km/hr (105 miles/hr). Dates of the survey, in May, are standardized to conform to geographical differences in phonological events. Time of the survey and flight conditions are also prescribed.
The aircraft pilot and another observer record all identifiable ducks within transect boundaries (200 m = 0.125 mile on either side of the flight line). Unidentified ducks are not recorded. Factors influencing aerial counts were described by Martinson and Kaczynski (1967).
To estimate the proportion of ducks present that are seen and identified by the aerial survey crew, ground counts are conducted on certain segments of some transects. Ground counts are not made in the northern areas (strata 1-25) because of inaccessibility. Crews of 2-4 biologists make ground surveys, preferably the day after the aerial survey. Efforts are made to census all ducks within the boundaries of the sample unit.
The ratio of birds of each species counted on the ground to the number counted from the air, taken over several segments within a group of strata, is termed the "visibility rate," although its reciprocal might more aptly bear that name. Visibility rates are calculated by species and year for groups of strata. For the northern strata, where ground surveys are not conducted visibility rates are taken as the long-term average from strata where they are determined.
During May aerial surveys, a count of pond basins containing water is made in strata 26-49. Included in the tally are natural wetlands of types III, IV, and V (Straw and Fredine 1956), which are basins that are inundated seasonally, semipermanently, or permanently, respectively. Streams and rivers that meander across the transect are counted each time they enter the transect. Also included in the pond count are man-made wetlands such as stock dams, dugouts, and large ditches that maintain water into summer. Water areas excluded from the count are temporary wetlands and sheet water (expected to persist <3 weeks), roadside or borrow ditches, and muskeg areas. Water bodies in the northern strata (1-25) are more permanent than those in prairie and parkland strata and therefore, are not counted. Pond counts for strata 26-49 were initiated at various times Breeding-season Distribution, during the 1955-65 period (Table 2).
Counts of ducks (by species) and areas surveyed are summed across all transects within a stratum, and their ratio, multiplied by the visibility rate, is termed the adjusted density. Adjusted density, multiplied by area of the stratum, yields the estimated population in a stratum. The sum of these values across all strata is the estimated population within the surveyed area. A similar method is used to determine pond count, except that no correction for visibility is made.
Pond counts provide the only index to breeding habitat available for most of the surveyed area. Counts in a stratum for a particular year are divided by the area surveyed within that stratum to produce an estimate of pond density. Pond densities varied by year and even more so by stratum.
Because pond counts in different strata were initiated at various times during the 1955-65 period, the data were unbalanced, as not all strata were surveyed in all years. As a result, the usual averages by year or by stratum may be misleading. To overcome problems resulting from this imbalance, we calculated Least Squares (LS) means, in addition to the usual means. LS means are based on an assumed linear model, in this case representing pond density as an additive combination of effects due to stratum and to year. LS means by year, as an example, are calculated by adding the intercept of the linear equation, the effect for the desired year, and the average of all stratum effects. This method is employed by the SAS GLM procedure (SAS Institute, Inc. 1982). The linear model of pond densities, with 26 year terms and 23 stratum terms, produced a coefficient of determination R2 = 0.75 for the 601 observations.
Breeding-season distributions of ducks were examined by computing the average adjusted density of each species for each of the 49 strata (averaged across all years that strata were surveyed). Densities were ranked and plotted on maps according to the quartile in which they fell: highest 12, second highest 12, third highest 12, and lowest 13. Plotting according to quartiles facilitated the comparison of distributions among species despite differences in average density.
The direction from which ducks approach their potential breeding ground was determined by their wintering area and migration route. The paths they traverse on their spring trips converge within rather well-defined migration corridors (Bellrose 1980). Most are generally north-south or northwest-southeast in orientation, but lateral movements are fairly common.
Patterns of use of migration corridors for the 10 species were determined from information summarized by Bellrose (1980). We estimated the proportions of members of a particular species that return to their major breeding grounds through (a) southern or southeastern corridors, entering through the Dakotas, and (b) western or southwestern corridors, entering through a northwestern Montana or Alberta portal. The importance of migration corridors to our analysis is that they affect the way the population may be distributed relative to the habitat; for example, a pintail migrating directly from California to Alberta is less likely to be affected by wetland conditions in South Dakota than would one returning from a wintering area in Texas.
To determine how local wetland conditions affected the distribution of ducks, we calculated the correlation coefficient between the density of each species and the density of ponds among years within each transect and stratum. This coefficient indicates how duck numbers and pond numbers varied together from year to year within a transect. For analysis, we averaged correlation coefficients over all transects within a stratum. We calculated medians, means, and weighted means, where the weight was the number of years on which the correlation coefficient was based. All 3 measures performed similarly, and medians are used in the following analyses.
Duck distributions may depend not only on local habitat conditions but also on conditions elsewhere in the breeding range, especially during drought. To assess this effect, we calculated a measure of total ponds for the pond-surveyed area (strata 26-49). For any year, this measure was the pond density for a stratum multiplied by the area of that stratum, and then summed over all strata. We used only the data for 1966-81, the period during which ponds were counted throughout strata 26-49. Correlation coefficients between the density of each species and the total pond count were calculated for each species over all years within a transect and stratum. The median of these coefficients across transects within a stratum was used as a measure of how a species in that stratum varied in relation to total ponds.
We attempted to measure several life history characteristics for each species. We quantified the stability or predictability of each species' habitat by developing an index to the permanence of basin wetlands used by the species in an extensive study by Stewart and Kantrud (1973) in North Dakota. Cropland ponds were scored 2 seasonal ponds were scored 3, and semipermanent ponds and reservoirs were scored 4. The weighted average was multiplied by -0.1; high negative values of the resulting index thus suggest considerable use of stable habitats (hypothetically, a K-selected attribute). The relative extent of opportunism in the response to habitat change was indexed by the median correlation between duck densities and pond densities; high values here suggest an adaptation to unpredictable environments (hypothetically, a r-selected attribute). We predict that these 2 measures are related to each other.