USGS - science for a changing world

Northern Prairie Wildlife Research Center

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

Duck Nest Success in the Prairie Pothole Region

Study Area and Methods

Duck nesting data were obtained from MNW and portions of North Dakota and South Dakota east and north of the Missouri River. For analysis, these states were divided into 5 geographical regions: MNW, NDE, NDC, SDE, and SDC (Fig. 1). Regions in North Dakota and South Dakota were divided by county lines nearest the eastern or northern edge of the Missouri Coteau (Stewart and Kantrud 1973, Brewster et al. 1976).

Fig. 1.  Distribution of duck nests sampled in eastern North Dakota (NDE), central North Dakota (NDC), eastern South Dakota (SDE), central South Dakota (SDC), and western Minnesota (MNW) during 1966-84.

Within each region, data were partitioned into 3 time periods and 8 habitat classes. The periods considered in the regions were 1966-74 (NDE, NDC, SDE, and SDC), 1975-79 (NDE and NDC), and 1980-84 (NDE, NDC, MNW, and SDE).

The 8 habitat classes used were grassland, idle grassland, hayland, planted cover, wetland, cropland, right-of-way, and odd area. Grassland is native prairie used for pasture or mowed for hay and includes planted grasses used for pasture. Idle grassland is native upland prairie on which no haying or grazing occurred during the current or preceding growing season. Hayland is planted crops mowed for forage. Herbaceous plantings established for wildlife or soil protection were called planted cover. Wetland includes the wet meadow, shallow marsh, and deep marsh zones of wetlands defined by Stewart and Kantrud (1971). Cropland is annually tilled fields of small grain or row crops. Right-of-way includes the cover between the outside right-of-way boundaries of roads, railroads, and canals. Miscellaneous habitat features such as rock piles, haystacks, gravel pits, and shelterbelts were called odd area.

Data collection followed procedures described by Klett et al. (1986). We defined a nest as a clutch of ≥1 egg. Nests that were terminated when found were not used to compute nest success. Most nests were found by flushing females with a cable-chain device towed between 2 vehicles (Higgins et al. 1977). Some were found with other types of drags or by searching cover on foot. Most units of cover were systematically searched 1-4 times, but some nests were found fortuitously or by locating radio-marked females. Nests were revisited ≥1 time to determine their fate.

Data recorded for each nest were species, year, habitat class, number of eggs and incubation stage when found, complete clutch size if known, fate (successful, destroyed, abandoned, or unknown), and cause of nesting failure if known. Incubation stage was determined by candling eggs (Weller 1956). A nest was considered successful if ≥1 egg hatched. A clutch was recorded as destroyed if no ducklings hatched and there was evidence of broken or missing eggs; usually all eggs were destroyed or missing. Intact clutches no longer tended by a female were considered abandoned. The fate of some nesting attempts was not determined, usually because the nest was not relocated. We also recorded dates needed to estimate nest success: date found, dates of subsequent visits when the clutch was still viable, and date of final visit when fate was determined.

Estimates of Nest Success

We calculated daily nest survival rates (DSR) for each combination of region, period, habitat, and species using the Mayfield (1961, 1975) method as modified by Johnson (1979). Nests were excluded from the analysis if they were from areas where organized predator removal or predator exclusion was likely to have increased waterfowl production.

The variance of an estimated DSR is inversely proportional to the number of exposure days involved (Johnson 1979). Among the various categories of nests, there was great variation in the number of exposure days and in the precision of the estimates. We used a linear model fit by the method of least squares (Snedecor and Cochran 1980) to improve imprecise estimates. The linear model allowed us to examine and exploit various relations among the categories of nests. Each value of DSR was weighted by the number of exposure days. We tested for significant (P < 0.05) main effects and 2-way interactions using analysis of variance and then fit a model with only significant effects included. Significant effects included main effects for region, period, habitat, and species, and interactions between region and species, period and habitat, and habitat and species. Because interactions between period and region and between period and species were not significant (P > 0.05), we concluded that differences in nest success among regions and among species were similar in all periods. We assumed that differences in success among habitats were similar for all regions. This assumption was necessary to estimate nest success for all habitats in all regions. As a consequence, for a given species and period, the habitat rankings were identical in each region. Therefore, the habitat differences presented from combined data for North Dakota apply to all regions.

Nest success was calculated by raising the model's predicted DSR to a power equal to the mean laying plus incubation periods for successful clutches. We used 35 days for mallards and gadwalls, 34 days for blue-winged teal and shovelers, and 32 days for pintails. Combined nest success estimates were obtained from weighted means of the constituent estimates. Weighting was needed to account for the number of nests in each habitat, the preference for each habitat, and the availability of habitats.

For each species in each region and period, nest initiations were apportioned among the various habitats as follows: let i = an estimate of the preference of the species for nesting in habitat i (described below), and Ai = an estimate of availability of habitat i. The proportion of total nests initiated in habitat i is estimated by Pi = iAi / iAi, where Pi is the product of preference and availability scaled so that Pi = 1.

Breeding Populations.--The number of nest initiations (initial and renesting attempts) by a given species in a particular region and period was considered to be proportional to the size of its breeding population. We therefore used estimates of breeding populations as weights for combining regions and periods. We used aerial survey data from the U.S. Fish and Wildlife Service (FWS) (Martin et al. 1979) to estimate mean annual breeding populations of each species in NDE and NDC by period. Estimates of breeding populations in other regions were not required because nest success data were not available for all periods.

The FWS reports breeding populations for 3 zones in North Dakota. We estimated mean pair densities for each zone and multiplied the mean densities by the size of the area that was encompassed by our regional boundaries. We then summed all estimates within each region to obtain estimates for the entire region. Annual estimates for each period were multiplied by the number of years involved to obtain numbers of breeding ducks.

Habitat Preference.--We defined the preference of a particular species for a certain habitat as the probability that a female will select that habitat for nesting, given that all habitats are equally available. Cowardin et al. (1985) examined the relative preference of mallards nesting in central North Dakota for 6 of the 8 habitats used in this analysis. That information and a relation between vegetation height and density, and mallard nest densities (Kirsch et al. 1978), were used in a stochastic model of mallard productivity (Cowardin et al. 1983). We executed the model on a data set with equal availabilities of all habitat classes to estimate relative preferences of the mallard for each of the 8 habitats.

We used the estimates of mallard preference described above in combination with data from the nest record file to estimate preference values for the other species. Because we did not know the area of each habitat that was searched for nests, we could not calculate relative densities directly from the nest record file. Instead, we assumed that within each habitat the proportion of nests found during search activities was similar for all species. To minimize the bias arising from species differences in nesting chronology, we used only nests that were found in habitats that were searched ≥2 times in NDC. Habitat preferences in NDC were applied to other regions that lacked sufficient samples of nests in all habitats to permit similar analysis.

For 2 habitats (e.g., A and B) let θA and θB = the true mallard preference values for A and B, respectively; θA + θB = 1. Let NA and NB = the number of mallard nests found in habitats A and B, respectively. If equal areas of A and B were searched, then E(NA) / E(NB) = θA / θB, where E(N) denotes the expected value. If, however, unequal areas of each were searched, then:

E(NA) / E(NB) = KθA / θB,   (1)

where K is the ratio of the area of A that was searched to that of B. Similarly:

E(N'A) / E(N'B) = Kθ'A / θ'B,  (2)

where the prime denotes a species other than the mallard. Solving equations (1) and (2) for K, and equating the resulting expressions, we arrive at the following expression for θ'B / θ'A:

θ'B / θ>'A = NAθBN'B / NBθAN'A,   (3)

where E(NA), E(NB), E(N'A), and E(N'B) have been replaced by NA, NB, N'A, and N'B, respectively. Substituting the previously obtained estimates of mallard preference for θA and θB in equation (3) and using θ'A + θ'B = 1, we obtain the following estimators for θ'A and θ'B: 'A =

N'AANB / (NABN'B + N'AANB, and 'B = 1 - 'A.

This method for providing estimates of preferences between 2 habitats is easily extended to estimates for ≥3 habitats. For n habitats, there are n - 1 linearly independent preference values that can be estimated by solving a system of n - 1 linear equations such as equation (3).

Habitat Availability.--Availability of nesting habitats in the 5 regions for 1980-84 was estimated from a stratified random sample of plots selected for a mallard model developed at NPWRC (Cowardin et al. 1983). Each plot was a 10.36-km² block where the area of each habitat class was determined by interpretation of aerial photographs. Estimates from these samples were adjusted to include planted cover and idle grassland habitats that were not distinguished in the mallard model. The area of these habitats for all periods was derived from unpublished annual reports of the FWS Division of Realty (K. F. Higgins and D. A. Davenport, unpubl. rep., NPWRC, 1977), and U.S. Department of Agriculture program summaries. Changes in amounts of planted cover and idle grassland were caused by government programs designed to establish cover for wildlife or to protect retired cropland from erosion.

Amounts of wetland and grassland during 1966-74 and 1975-79 were extrapolated from the 1980-84 data. Loss rates of wetlands were derived from Cowardin et al. (NPWRC, unpubl. data). Loss of range and pasture land in North Dakota from 1958 to 1977 (U.S. Dep. Agric., N.D. Multiyear Plan for Resour. Conserv., unpubl. rep., undated) was used to estimate the amount of grassland present in North Dakota and South Dakota during earlier periods. We assumed that the amounts of right-of-way, hayland, and odd area habitats were similar in all periods.

Previous Section -- Introduction
Return to Contents
Next Section -- Results

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 19:12:50 EST
Reston, VA [vaww55]