USGS Home
Contact USGS
Search USGS

Dzubin (1969a) presented a detailed discussion of problems involving field methods for estimating breeding population, which is the first step in estimating recruitment rate. Dzubin's work remains the foundation of most currently used methods for estimating recruitment of ducks, but a number of new methods have been proposed and used. We will review the advantages and disadvantages of currently used methods with regard to precision, accuracy, and practicality.

1. Pair and Brood Counts

The advantages of the pair count-brood count index are that it is simple, direct, and easily understood. If both pairs and broods can be accurately counted, the method can furnish a good estimate of recruitment rate. Its most serious disadvantage is that brood counts for dabbling ducks are frequently suspect. Dabbling duck young often seek dense emergent cover where they are difficult or impossible to see. In addition, the young of some species like the Mallard may escape into the uplands when disturbed during a brood count (Smith 1971). Some investigators have employed morning or evening counts of broods from fixed observation points. Kirby (1980) suggested that this method could furnish an index but not an estimate of young produced in forested habitat. Rumble and Flake (1982) compared counts from fixed observation posts and counts made by searching ponds ("beat outs"). They estimated total broods present by the Petersen index (Seber 1982:59) by treating broods seen on the first count as marked broods and broods of the same size, age, and species seen on a second count as recaptures. They then used that estimate to determine the proportion of broods seen by the passive and active methods. Sixty-nine percent of the broods were observed by passive counts and 74% by active counts. Their study was in stock pond habitat; the proportion seen would undoubtedly be much smaller in habitat with densely vegetated ponds. Visibility of broods also varies by age of the brood and time of day (Ringelman and Flake 1980). Minser and Dabney (1973) compared float counts for Wood Ducks made during the day with counts made at night with the aid of searchlights. They found that 29.5 broods were counted at night, as opposed to 7.3 broods during the daytime on the same area. Diving ducks, on the other hand, tend to move out into open water where it is possible to count both broods and the number of young (Stoudt 1982). The pair count-brood count method may be the best available method for obtaining estimates of diving duck recruitment. Because the method requires estimation of numbers of pairs and broods on large sample plots, it is time-consuming and expensive.

2. Nesting Studies

Because recruitment rate is highly correlated with nest success, nest success is a good index to recruitment (see chapter 14 of this volume). In addition, if information on renesting rate is available, hen success can be estimated from equation 2, and if information on brood survival and average brood size is available, it is possible to estimate recruitment rate from equation 1.

Nest success data are frequently misused in recruitment estimation. Nest success rates typically vary among habitats and years (Greenwood et al. 1987, Klett et al. 1988). Hens select nest sites for one or more nests from among one or more habitats. Recruitment is defined in terms of breeding hens and because breeding hens may use more than one habitat, habitat-specific recruitment rates are not appropriate.

Use of equation 2 requires an overall estimate of nest success. In practice, waterfowl biologists usually obtain samples of nests from each habitat. To obtain an overall estimate, it is necessary to weight estimates from the individual habitats by the total number of nests initiated in each habitat. This number should not be confused with the number of nests in the sample because the total number of nest initiations is seldom known. An alternative procedure is to weight the individual habitat estimates by the product of availability and nest density. However, estimates of nest density, like nest success, are influenced by inability to find unsuccessful nests. The bias of the unweighted estimate is illustrated in Table 13-1. If the purpose of the study is to compare success among habitats, it is necessary to obtain some minimal sample of nests in each habitat. Klett and Johnson (1982) suggested 50 nests per habitat as a target. In the example, the unweighted estimate (22.0%) from equal samples of 25 nests per habitat seriously underestimated true overall nest success (38.7%). On the other hand, if the purpose of a study is to estimate nest density, equal areas may be sampled from each habitat. In the example, this procedure led to an overestimate (46.0%) of overall nest success. Unbiased estimates can be obtained by sampling in proportion to the number of nests in each habitat, not to be confused with the areal extent of each habitat. In reality, the number of nests in each habitat is not known. An approximation can be developed by sampling in proportion to habitat availability and then weighting by some measure of habitat preference (Klett et al. 1988).

Nesting studies have a number of distinct advantages: (1) nest survival rates are relatively easy to estimate from field data; (2) nests can be related to a habitat, so that nesting studies are well suited to evaluating nesting habitats or habitat management (Livezey 1981); (3) nest density provides a measure of habitat preference, if all nests are found or if the proportion of nests found is the same among habitats; (4) nesting studies provide information on the cause of nest loss; and (5) nest success alone provides an index to recruitment rate.

The disadvantage of nesting studies is that they provide no information on renesting rate or brood survival. Estimates of recruitment rate therefore require independent knowledge of those parameters and the use of a model like the one shown in equation 1. Nest searching may lead predators to the nest and thus influence the success rate. The presence of such a bias is difficult to prove or disprove, but Livezey (1980) and other authors (Gotmark and Ahlund 1984, Cowardin et al. 1985) found no evidence that nest searching caused nest failure. On the other hand, Strang (1980) presented evidence that Parasitic Jaegers (Stercorarius parasiticus) can be led to waterfowl nests by people. MacInnes and Misra (1972) presented similar evidence for Canada Geese.

Nesting studies also furnish a direct, though time-consuming, method of estimating recruits produced. Two assumptions must be met: (1) all habitats are searched either completely or on a random sample of plots, and (2) all successful nests are found. If these assumptions are met and an estimate of brood survival (Z) is available, an unbiased estimate of recruits produced = successful nests × Z × B. Brood survival (Z) and average size of broods at fledging (B) are usually treated as constants, a procedure that underestimates variation in estimates of the number of recruits produced.

3. Radiotelemetry

The disadvantages of radiotelemetry studies are (1) they require large outlays of both money and personnel; (2) it is necessary to capture a sample of hens, a sometimes difficult task in spring; and (3) unknown biases may exist due to the effect of the radio.

4. Social Index

Advantages of the social index are (1) it is easy to conduct, (2) it is cheap and not highly demanding of personnel compared with other techniques, and (3) it may furnish an index to recruitment that is available early enough in the year to be used in decisions on regulations for that year. If models relating recruitment rate to social index can be developed and validated, then the social index has the potential for becoming an economical method of estimating recruitment rate. At present, the method is highly experimental and based on limited data.

5. Late Nesting Index

6. Methods Based on Fall Age Ratio

Most waterfowl species can be aged and sexed from plumage characteristics (wings of ducks and tails of geese) (Carney 1964). In North America a sample of duck wings and goose tail fans is obtained each year from hunters, and the sex and age of the birds in the sample are determined. To calculate a fall age ratio, it is necessary to determine the differential vulnerability of the age and sex classes to hunting. In practice the sexes are often pooled. Differential vulnerability is estimated by dividing the age ratio in the sample of wings or tails by the relative recovery rate (recovered adults/recovered young) of leg-banded birds.

Enough Mallards are banded each year in major portions of the breeding range in Canada and the north-central United States to provide sufficient recoveries for adjusting harvest data. For other species, adequate and well-distributed banded samples are not obtained each year. If a sufficient sample of band recoveries is not available, the age ratio from the wings and tails must be used as a relative index of fall age ratio. Estimates of unadjusted fall age ratio of many waterfowl species are published annually for each flyway by the USFWS. If an independent estimate of adult female survival during the summer is available and the age ratios in the harvest are adjusted for differential vulnerability, it is possible then to estimate recruitment rate (female fall age ratio x female summer survival rate) for these species as well. The procedure of using harvest and recovery data to estimate fall age ratio has been reviewed recently by Munro and Kimball (1982), and their concerns will be discussed later.

Age of some goose species and swans can be determined from observation in the fall because the younger birds have different plumage than the adults. Chamberlain (1966) used 35-mm aerial photographs to determine ratios of young to adults as well as size of family groups for assessing productivity of Tundra Swans. In theory, it should be possible to determine the fall age ratio by observing samples of these birds during the fall and winter. Age of Canada Geese can be distinguished by observing behavioral traits just prior to the hunting season (Raveling 1981). Nilsson (1970) concluded that age ratios obtained from flocks of several species were subject to many biases and doubted that they could be used to measure recruitment in the previous summer because of differential migration by the age groups. Family groups of geese remain together during fall and winter, and successful adult geese can be distinguished from unsuccessful ones during fall migration and winter (Lynch and Singleton 1964). Raveling (1968) demonstrated that group counts of geese made when the birds are landing but not when they are flying overhead can furnish an index to recruitment. Prevett et al. (1982) reviewed the use of winter counts of Snow Geese that had been conducted in Louisiana since 1937. They found that age ratios obtained in winter overestimated actual production because of more rapid migration of successful than unsuccessful females.

Various models have been used to predict fall age ratio and fall flight. Reeves et al. (1976) used satellite data from LANDSAT and TIROS to evaluate snow conditions in the Arctic. They showed a relation between this variable and subsequent fall age ratios of arctic-nesting geese. Russell et al. (1979) suggested similar procedures for predicting waterfowl productivity on the Old Crow Flats of Alaska. Boyd (1981) developed models for predicting the number of ducks in the prairies of Canada from precipitation data from the current and prior years.

In order to set hunting regulations for a season, an estimate of size of the fall flight is needed in August, before all results from the current breeding season are available. Geis et al. (1969) developed a regression model for estimating the number of young Mallards in the fall from four explanatory variables: the number of July ponds, the estimated size of the continental breeding population, the percentage of ponds remaining from May to July, and an index (unadjusted for visibility) to the number of broods. The model was constructed from all previous years' data and used to predict the fall age ratio for the current year. Each year the model was fitted to a new data set that included the past year's data. Assumptions and potential biases in the regression procedure will be treated in detail in the following examples.

Prediction of size of the fall flight requires information on summer survival. Fall flight may be the estimate of most interest to the waterfowl manager. Summer survival rate is also essential for predicting annual change in population size. However, no realistic year-round models or sufficient data for constructing such year-round models have been developed (Johnson et al. 1987b), although simple deterministic models for closed populations have been used (Cowardin and Johnson 1979, Martin et al. 1979).