USGS Home
Contact USGS
Search USGS

1. Ground Counts

Many species of waterfowl frequent areas where they are hidden by vegetation. Biologists have coined the term "beat out" to describe a survey in which birds are flushed from cover in order to tally them. This method involves walking through all areas of vegetation and often includes making noise to assist in flushing the birds. On deeper water areas similar techniques have been used from canoes (Stoudt 1938, Collins 1974). Besides being an arduous task, ground counts lead to another problem, for which the term "roll up" has been used. Birds flushed from one wetland may settle on another, and the same birds may be counted a second time. In addition, the act of flushing may make the behavioral interpretations discussed in the previous section difficult.

Counting birds from stationary observation points and assuming that all birds will move into openings and become visible is an alternative to "beat outs" (Kadlec 1966). Rumble and Flake (1982) compared active and passive counting techniques for broods, which are more difficult to observe than pairs. Kirby (1980) used passive counts from elevated blinds for both pairs and broods in forested areas of Minnesota where direct counts were difficult or impossible. In some habitats passive methods will not work because birds may never move into areas where they are observable. Passive counts also present sampling problems because of the difficulty in determining whether birds counted from one observation point are the same birds as those counted from another. For prairie habitat, Dzubin (1969a) recommended a combination of methods that included counting pairs from a vehicle parked at a good observation point for each pond and using noise to flush birds not seen from the observation point.

There can be no hard-and-fast rules about the use of active flushing versus passive counting from observation points. The method must vary depending on the nature of the habitat, the size of the wetlands, and the ease of access to them. Under some habitat conditions a direct method may be virtually impossible, especially for broods, and one of the less direct methods listed below must be used.

2. Counts from Aircraft

Aerial surveys for ducks have generally been conducted from fixed-wing aircraft flying at low level, but helicopters have also been used especially for arctic-nesting geese (Reed and Changnon 1987). Helicopters have advantages over fixed-wing aircraft because they can be flown at low speeds and low elevations, and they furnish excellent visibility. The noise from a helicopter may disturb birds and move them. The cost of helicopters is much higher than for fixed-wing aircraft, and they have not been used as the primary tool in continental surveys. Ross (1985) found that helicopter and ground surveys for conducting waterfowl counts in the boreal forest had comparable costs when salaries were considered. In addition, he found that the Black Ducks moved into open water where they could be counted when approached by a helicopter. Kaminski and Prince (1984) used a helicopter to count pairs in marsh habitat and found that 50 64.8-ha plots could be censused in four hours.

Conspicuous species such as swans can be inventoried from aircraft with reasonable certainty that all individuals will be counted (King 1973, Lensink 1973, King and Conant 1981, Reese 1982). Other species are difficult to observe or to identify, especially females (McLaren and McLaren 1982). Aircraft counts generally miss some individuals for most species of waterfowl even when the survey is designed to obtain a total count. Furthermore, the proportion of the population that is missed will vary with species, habitat, year, and to some extent the observer (Diem and Lu 1960, Tacha et al. 1978, Broome 1985). Caughley (1974) discussed various biases that can occur from aircraft transect censuses of big game, and Caughley and Goddard (1972) suggested a method for estimating the number of animals missed by using repeated censuses, but Routledge (1981) presented evidence that their method was not appropriate. Burnham and Anderson (1984) discussed methods of estimating numbers from strip transect data where the distance to each animal observed is known. If the proportion of birds not seen remains constant, an assumption that is open to question, uncorrected aerial counts can serve as an index to population size. Normally, however, interest is in an estimate of size of the population.

One source of bias in aerial waterfowl inventories can be overcome by estimating the proportion of birds that are not seen by comparing birds seen from the aircraft with birds seen on the ground, which provides an airground correction ratio (U.S. Fish and Wildlife Service and Canadian Wildlife Service 1987). Nearly simultaneous counts of sample areas are made from an aircraft and by ground surveys. An air-ground correction ratio is then used to expand aerial counts by species on areas of similar habitat. These ratios are calculated each year during the cooperative breeding ground surveys in North America on those transects in the prairie habitat.

On northern transects in the forested region, constant ratios are used. Although many species are found here, they often occur in low densities, and the logistic problems and associated expenses in labor make extensive ground verification of aerial counts unfeasible. Consequently, aerial counts by species have been adjusted annually with long-term constants derived principally from historical data gathered in prairie-parkland areas. However, the necessity of obtaining annual adjustments in the boreal region has long been recognized. Beginning with preliminary trials in northern Manitoba and northern Saskatchewan in the late 1970s, and continuing with a more expanded effort across all northern areas in 1985 and 1986, helicopters are being used in place of ground counts to construct representative visibility rate estimates for the region.

The use of helicopters for censusing waterfowl has been considered for many years, and various studies have explored their feasibility in a variety of habitats and conditions (Lotter and Cornwell 1969, Bateman 1970, Malecki et al. 1981). The advantages of helicopter surveys include the speed with which the counts may be obtained, thus increasing sample size, and their potential use in remote areas where there is little chance of obtaining ground counts. However, the substitution of helicopters for ground counts also assumes that all birds are counted from the air, an assumption that may not be satisfied in many situations. In addition, helicopter flight time is expensive, which may limit its application. Ratios of helicopter to ground counts and of helicopter to fixed-wing counts have varied from study to study, but in some cases the results suggest that this survey method has promise.

Several assumptions are required for the corrected estimate to be without bias: (1) all individuals are counted in the ground count, (2) the population sampled from the aircraft is the same as the population sampled from the ground, and (3) the air-ground ratio derived from the sample plots is the same as that for unsampled plots to which it is applied. In practice, some of these assumptions may not be met. For example, the ground subsample of the cooperative breeding ground surveys is purposely selected and may not be representative of the aerial sample. In addition, ground access to some transects is not practical, and some of the usual assumptions regarding randomness and subsequent independence among the various estimators are not met. Nevertheless, some degree of bias is accepted and tempered by the high precision attained by very large sample sizes that are obtained with the aerial survey. Sorensen and Ibister (1970), Surrendi (1972), and Bowden (1974), among others, have discussed possible sources of bias associated with aerial surveys of ducks in prairie-parkland areas of Canada and the United States. Species identification from aircraft can also be a source of error in surveys, although experienced observers are very skillful in identifying birds from the air. However, in most aerial surveys of breeding pairs some birds will not be identified to species and are placed in an unidentified category. Species identification is even more difficult for broods, and in many surveys, including the cooperative breeding ground survey, no attempt is made to separate species.

3. Counts from Aerial Photographs

Erwin (1982) used aerial photographs experimentally in a test of the ability of different observers to count large numbers of ducks in flocks. He found that without reinforcement through training, experienced observers were more accurate than inexperienced observers. Eight of nine observers tested estimated number of birds within 10% of the actual number.

4. Indices to Breeding Population

Extensive surveys by cooperating professional and amateur ornithologists are used for estimating waterfowl numbers in Europe (Milstein 1968, Yarker and Atkinson-Willes 1972). In North America a breeding bird survey is conducted by cooperating amateur ornithologists along roadside transects (Erskine 1978, Robbins et al. 1986). The surveys do not estimate the size of breeding populations but do yield an index to abundance that has been used to establish population trends for all species. A summary of population trends from the first 15 years of the survey has recently been published (Robbins et al. 1986).

Because of their selection of woodland habitats, Wood Ducks are one of the most difficult waterfowl species to inventory, and special survey techniques have been developed. Smith and Flake (1985) found that on a prairie river Wood Duck broods were restricted to the river and adjacent oxbow lakes. Several states have calculated population indices for Wood Ducks based on counts made from a boat floating stretches of river, but these indices are subject to many biases and lack precision (Cink 1977).

Postbreeding Wood Ducks use roosting areas at night and fly to them at dusk. Counts of birds entering these areas have been used as an index to Wood Duck abundance (Luckett and Hair 1979). Hein and Haugen (1966) concluded that roost counts were sufficiently precise to detect a 15% change in population size. The validity of the technique has been questioned because of daily (Tabberer et al. 1971) and yearly variation in use of roosts. Parr and Scott (1978) reviewed use of roost counts as a population index and concluded that the technique was not practical. Furthermore, roost counts, like winter counts, include both surviving adults from the breeding population and recruits from that year's production and can furnish only a rough index to breeding population.

5. Estimates Based on Habitat Availability

6. Indirect Estimates Based on Mathematical Models

Waterfowl have been banded in large numbers and furnish an opportunity for development of indirect population estimates based on band recovery and harvest data. Crissey (1957) pointed out that if the size of the harvest can be estimated from harvest surveys and the rate of harvest can be estimated from banding data, then the fall population producing the harvest can be estimated by harvest divided by harvest rate. If summer survival rate is also known, then calculation of the size of the breeding population that produced the harvest is possible. These techniques have been used to estimate North American populations of Mallards (Crissey 1957) and Wood Ducks (Geis 1966).

Mark-recapture models have also been used for estimating population parameters (Boyd 1956, Anderson and Sterling 1974), but as Martin et al. (1979) pointed out, these methods are more appropriate for species that breed in relatively discrete populations than for continental populations. Sulzbach and Cooke (1979) used mark-recapture methods to estimate population parameters, including population size, for a Lesser Snow Goose population. Coulson (1984) banded nesting Eiders and recaptured them in subsequent years. He then used Jolly-Seber models (Jolly 1965, Seber 1965) to estimate population size. Barratt (1966) suggested the possibility of banding Wood Ducks by night-lighting, recapturing them by drive-trapping, and then using mark-recapture methods to estimate population size. D. G. Raveling (pers. comm.) used the Petersen index to estimate breeding population by treating sightings of neckbanded geese as recaptures on the wintering ground.