USGS - science for a changing world

Northern Prairie Wildlife Research Center

  Home About NPWRC Our Science Staff Employment Contacts Common Questions About the Site

Mallard Recruitment in the Agricultural Environment of North Dakota

Definitions, Assumptions, and Statistical Tests

There is considerable variation in use of terms applicable to studies of nesting and production of waterfowl. We used a number of specific definitions in this study. A nest was defined as any depression containing 1 or more eggs (Miller and Johnson 1978), a destroyed nest was any nest in which 1 or more eggs had been destroyed and to which the hen did not return and continue laying or incubating, a successful nest was any nest in which 1 or more eggs hatched, and an abandoned nest was any nest where the hen failed to return but the eggs remained undamaged. If the hen failed to return after the initial visit, as determined by telemetry, or if we observed aberrant behavior of the hen caused by the radio, such as erratic flight or failure to resume normal daily movement patterns, the abandonment was classed as observer-caused and the nest was not used in calculation of nest success. Several abandoned nests had missing eggs or were left uncovered suggesting disturbance, probably by predators, of the hen and nest. For calculation of nest success both destroyed and abandoned nests were considered unsuccessful.

Recruitment, defined as the number of hens recruited to the fall population per hen in that year's spring population, can be subdivided into several components: nest success, the probability that a nest will be successful; hen success, the probability that a hen will have 1 successful nest in 1 or more attempts; brood survival to water, the probability that at least 1 brood member will survive from nest to early age class IIc (Gollop and Marshall 1954); and average brood size at fledging.

The definition of recruitment above requires definition of a population of hens available on the study area in spring. We defined a member of the spring population as any bird captured before 15 May and remaining in the area at least 1 week after capture. This date is comparable to the date on which the U.S. Fish and Wildlife Service conducts its annual census of the breeding population in the study area. Some radioed birds were rejected from the population for a number of reasons that related to radio failure or effect of the radio on the bird: (1) observed aberrant behavior (In 2 cases, we observed birds with a broken transmitter harness that could interfere with flight. We also had 2 cases where hens remained in dense cover and did not flush, which was unlike behavior of other radio-marked as well as unmarked hens.), (2) loss to a predator within 3 days of marking, (3) loss of transmitter, (4) loss of signal within 7 days of marking, (5) loss of signal after the signal became weak or intermittent, and (6) nest abandonment judged to be caused by the observer. We could tell when a hen never returned to a nest by carefully observing the material used to cover the nest and by monitoring presence or absence of the radio signal at the nest. If the hen never returned after being flushed by the observer, we considered the abandonment observer-caused. By these definitions, 177 hens were available for calculating hen success (Table 2).

Table 2. Numbers of radio-marked mallard hens, nests, and radiolocations obtained in central North Dakota, 1977-80.
Year Hens equipped with radios Selected population for hen success Total nests Nests for calculating nest survivala Nests for calculating hen success Telemetry locations
1977 69 34 17 17 16 805
1978 77b 38 38 34 32 1,374
1979 93c 59 45 41 41 1,935
1980 99d 59 45 41 41 1,935
All years 338 177 142 129 126 5,948
aNests for hens that were trapped too late for inclusion in the defined breeding population can be used in calculating nest   survival rate but not hen success.
bIncludes 1 bird captured and not included for hen success in 1977, but included in 1978.
cIncludes 1 bird captured and included for hen success in 1978 and in 1979.
dIncludes 1 bird captured and included for hen success in 1977, but not included in 1980.

Four assumptions must be met for the estimates presented here to be without bias: (1) the radio packages did not influence the breeding behavior of the birds, (2) the birds captured were representative of the population being sampled, (3) birds included under the definition of hens available to breed did not leave the study area and breed elsewhere, and (4) no successful nest of a marked bird went undetected.

Medians and 10 and 90% quantiles were used for analyses dealing with nest initiation date and length of the nesting season. Estimates were derived by univariate procedure described by SAS Institute, Inc. (1982). Medians were preferable to means because of the skewed distribution of nest initiations, and quantiles were preferable to the date of first and last nest because the method reduces the variability inherent in small samples.

Data on nest success presented in many published reports contain serious biases that can invalidate the relation of nest success to habitat. The bias is caused by the fact that destroyed nests are not found because the hen is no longer present. Mayfield (1961) corrected for this bias by calculating the daily nest mortality rate as total nest losses observed ÷ total days under observation. Daily nest survival rate (1 - daily nest mortality) can then be raised to a power equal to days required for laying and incubation and thus obtain an estimate of nest survival rate. Johnson (1979) developed a method for estimating the days that a nest was under observation, as well as procedures for estimating daily mortality rates and their standard errors. Confidence intervals for nest success were derived by first setting limits on the daily survival rate and then raising the limits to a power equal to the average length of exposure for successful nests, a procedure that gives asymmetrical limits. Tests of differences among habitats, years, and ages were based on an extension of the method developed by Johnson (1979) for comparing daily nest survival rates. This expanded method, also developed by Johnson (pers. commun.), allows for comparisons of more than 2 classes and assessment of interaction as in analysis of variance.

Statistical comparisons and tests of differences among years, ages, and habitats were based on the general linear model procedures of SAS Institute, Inc. (1982). Details of the applications, as well as other test statistics, are discussed under the appropriate section where data are presented. Significance of statistical tests were based on the 95% probability level unless stated otherwise in the text.

Previous Section -- Habitat Conditions and Climate
Return to Contents
Next Section -- Results and Discussion

Accessibility FOIA Privacy Policies and Notices

Take Pride in America logo logo U.S. Department of the Interior | U.S. Geological Survey
Page Contact Information: Webmaster
Page Last Modified: Friday, 01-Feb-2013 19:50:43 EST
Sioux Falls, SD [sdww55]