Assessing Breeding Populations of Ducks by Ground Counts
Duck Census Techinques
Spring and summer duck population estimates, whether based on direct air or ground counts, remain relatively inexact. Even more inexact is the accurate assessment of absolute seasonal populations of pairs attempting to breed in a stratum, along a transect, or on a sample block. It is generally conceded that counts of absolute numbers of birds breeding in a particular habitat are not feasible, and therefore workers are forced to estimate populations on the basis of various sampling procedures. The early works of Nicholson (1931), Leopold (1933), Lack (1937), and Kendeigh (1944) reviewed many of the problems inherent in conducting censuses, while more recent reviews by Fisher (1954) and Davis (1963) discussed sampling problems of various population estimate methods and the assumptions on which census techniques are based. The following discussion covers some of the difficulties inherent in any dabbler duck population determinations.
Moore (1955) described many of the problems involved in using strip-transect methods for estimating upland game birds. A number of these were discussed by Stewart et al. (1958) and Diem and Lu (1960) for transect counts of waterfowl. Yapp (1956) discussed the theory of line transect counts and suggested that the number of animals a census taker would see, walking at a constant speed over a straight-line course, depended on (1) the density and speed of the animal, (2) the walking speed of the census taker, and (3) the effective distance of recognition or visibility. Skellam (1958) questioned the method because of inexact derivation of the average speed of the animal and its relationship to the speed of the observer. Many of the problems involved in obtaining increasingly accurate and reliable passerine bird estimates have been examined by Taylor (1965) and Snow (1965) and apply equally well to census of many other bird groups. Seasonal replication of counts and intimate knowledge of species ecology and behaviour tend to make census techniques more valuable in construction of valid population indices.
Many of the techniques used for ground and aerial census of waterfowl in spring and fall have been summarized for Europe by Isakov (1961, 1963), Formosov and Isakov (1963), Matthews (1960), Tamisier (1965), and Grenquist (1965) and for North America by Smith (1956), Crissey (1957), Stewart, Geis and Evans (1958), Diem and Lu (1960), Smith (1964), and Hammond (1966). Population estimates have been calculated by using marked to unmarked ratios for ducks by Lincoln (1930), for geese by MacInnis (1964, 1966) and Fisher (1954), and for immature ducks by Cowardin and Higgens (1967). Photographic methods of estimating number and density of birds were presented by Chattin (1952), Cowardin and Ashe (1965), and Van Tets (1966).
Recent fluctuations in absolute numbers of the continental mallard population, their yearly production of young, and correlations of production ratios and number of July ponds were discussed by Crissey (1969). Continental trends of all duck species from 1955 to 1966, based on breeding population indices from aerial counts, were given in Martinson and Henry (1966: USBSFW unpublished Administrative Report 119, 10 p.).
Components Counted as Indicated Pairs: A Review
Estimates of abundance of waterfowl breeding pairs in various North American habitats have been based on a variety of population components. There has been a remarkable lack of standardization as to categories of each species censused. Differences of census technique have reflected objectives of the census-whether yearly trends, indices, estimates, or absolute numbers were desired. Comparing results of studies based on different components with variable errors of estimation is, at best, difficult.Counts of lone drakes on waiting sites as reliable indicators of breeding pairs of blue-winged teal was first proposed by Bennett (1938). Hochbaum (1944) expanded the concept of lone drake census to include all dabbler species but pointed out that there were variations in length of localization of the drakes' activity on "territory". Low (1947) concluded that numbers of nesting pairs on an area could be ascertained more accurately by a count of pairs or drakes on their territories than by a nest count. Generally, censuses of breeding pairs of waterfowl were based on the premises of drake and pair isolation, activity localization, and conspicuousness during the prelaying, laying, and early incubation periods. Smith and Hawkins (1948) noted that "unique in the spring inventory are the definiteness of the territory, tendency of the drake to display, tameness of ducks in the spring, sparseness of cover and scattered arrangement of birds". They describe a breeding pair census as the tallying of all water areas and breeding birds (as evidenced by pairs or single drakes) falling within a designated strip. For a number of years after 1948 waterfowl pair censuses were based on "the number of breeding birds by species as evidenced by pairs and single drakes and the total number of ducks by species" (Williams, 1948). Reeves, Lundy, and Kreller (1955), Ellig (1955), and Ordal (in Moyle, 1964) utilized only lone drakes plus pairs to equal breeding pairs. Smith (1953) did not include lone drakes on his counts of territorial pairs on artificial reservoirs in eastern Montana, as counts were conducted only, once a month. No workers included grouped males as indicative of breeding pairs.
On 12 Manitoba transects, a single hen component was added to the lone drake and pair categories by Kiel (1949) to arrive at an estimate of indicated diver and dabbler pairs. Groups of males or females or mixed parties of both were recorded as ratios and not considered breeding pairs. Evans and Black (1956) also included lone hens on their Waubay Study Area censuses as did Leitch (1952) for the Caron Study Area. In Alberta, Smith (1957) included lone females, especially as an index of pairs in late season counts. Stoudt (1952, 1964) also included this component in southeastern Saskatchewan. Brood hens were tallied as indicated pairs in late season counts on a Manitoba study area by Evans (1949) and Evans et al. (1952).
In Maine, Mendall (1949) used brood counts and a nesting study to estimate pair populations. He noted that because of the spread of the nesting season, accurate waterfowl census by the "territorial count" method had to be supplemented by counts of broods. "When this is done the number of breeding pairs is calculated by using the total number of broods and maternal or "broody" females and adding to this a proportional number of pairs (to represent the unsuccessful breeders ) as based on the annual nesting success study." Later, Mendall (1958) determined population trends through a combination of three techniques: (1) counts of pairs and territorial males, (2) sample nesting studies, and (3) brood counts. In 1957, Rogers(1964) used brood counts divided nests to estimate resident lesser scaup pairs. Jessen, Lindmeier, and Farmes (1964) also used nest counts, nesting success, and brood counts to estimate populations of ducks breeding on a Minnesota study area. They discuss at some length (p. 83-85) the problems involved in estimating numbers of breeding pairs using three components-pairs, lone drakes, and males in groups up to five, especially during protracted breeding seasons.
Lynch (1951), recognizing the behavioural significance of drake groupings, suggested that lone mallard and pintail drakes and all grouped drakes of three or four be enumerated as indicated cased pairs by aerial crews. Calculated ducks-per-square-mile figures should then be adjusted for the laying or incubating but unseen hens, associated with drakes. He stressed that hens of pairs were more difficult to observe from the air while most drakes were clearly visible. In essence, he recommended enumeration of all apparently resident drakes as pairs. Yearly variations in percentages of lone drakes to all ducks seen were to be utilized as an index to successful first nesting attempts.
Bue (1952) has conducted the most intensive analysis of breeding population dynamics based on weekly censuses of 50 stock ponds in western South Dakota. Counts were made from April to August of 1950 and 1951 and included pairs, lone drakes, lone hens, grouped pairs, drakes and hens, and postbreeding groups. He used four methods to arrive at seasonal breeding pair populations of mallards, pintails, shoveler, gadwall, and blue-winged teal: (1) the weekly indicated breeding pair population represented by lone drakes and pairs; (2) the weekly potential breeding pair population by enumerating all females seen in (1) above plus all hens observed as lone hens, in grouped courting parties, or as brood hens; (3) the indicated breeding pair populations by "accumulative calculated desertion of males" by weeks. This method accounted for a shift of lone and paired males to grouped drakes and courting parties but the population could not be tallied until the last drake deserted the home range; (4) a weekly breeding pair population by adding the population from (3) to all hens observed as pairs and in courting parties. Females seen as lone hens and with broods were not used. Each method gave fairly comparable results but for different weeks of the breeding season. Bue (1952: 13) noted that the then current single-count census techniques which enumerated only pairs and lone drakes did not account for pairs in which drakes had deserted territories, pairs which arrive late, or pairs which leave the area without attempting to nest.
In South Dakota, Murdy (1953) counted pairs, lone drakes, unpaired males, and unpaired females of five species of ducks during an entire breeding season. Using ratios of pairs to lone drakes plus a knowledge of nesting phonology and migration chronology, he was able to establish optimum census periods for each species for the state. He concluded that (1) various percentages of each species were paired on arrival, with pintails the least paired, (2) lone drake indices may have been affected by presence of unpaired males, (3) lone drake indices fluctuated throughout the season.
In England, Boyd and King (1959) estimated potential breeding pairs of mallards on four reservoirs from frequent direct counts and sex ratio counts made from February to August. They point out that "a nest count is in theory the best measure of the breeding population" but recognized the problems involved in finding all nests and the effects of finding nests on increased predator loss. In Alberta, Keith (1961) used an average of seasonal counts of adult ducks to estimate the number of breeding pairs of 11 species on his study area impoundments. Gates (1965) used Keith's (1961: 66) data on average seasonal nests per pair to calculate breeding pair populations of mallard and blue-winged teal on Wisconsin farmlands. He assumed that renesting rates were similar in the two areas.
In their intensive evaluation of ground transect census methods in Alberta, Diem and Lu (1960) separated species into four groups based on observed mobility, viz. (1) sedentary puddlers, (2) mobile puddlers, (3) sedentary divers, and (4) mobile divers. They tested the influence of time of day on three components (the indicated population, single drakes and single hens) of three species-mallard, blue-winged teal, and lesser scaup-but made no mention of enumerating grouped drakes. On the basis of an intensive study of black duck breeding biology and behaviour by Stotts and Davis (1960), Chamberlain and Kaczynski (1965) utilized four components - pairs, single drakes, groups of three drakes, and large groups of five or more drakes - to determine stage of nesting season. The data were used to better predict optimum aerial census periods for black ducks in eastern Canada. In Wisconsin, Jahn and Hunt (1964) enumerated lone males, lone females, pairs, flocked males, and flocked females, but used lone males and pairs only to compute pair densities on ponds. Hammond's (1959) recommendations of censusing only pair and lone drake dabblers but pairs plus extra female divers were used by Burgess, Price, and Trauger (1965) for censuses in Iowa. Martz (1967) also used Hammond's recommendations for censusing waterfowl, mostly gadwall and blue-winged teal, at the Lower Souris National Wildlife Refuge. He excluded flocks of three or more males or three or more pairs from the counts as representing nonbreeding or postbreeding birds. To obtain estimates of breeding populations of blue-winged teal, Glover (1956) recommended counts during spring migration and while males were on waiting stations, with associated nest counts on sample study areas. He utilized a series of seasonal censuses, chiefly of pairs, lone males on waiting stations, and nests per unit of habitat.
Stewart, Geis, and Evans (1958) described how pairs, lone drakes, groups of mixed sexes, and unidentified birds were recorded by aerial survey crews. Each drake on a breeding area, except drakes in groups of mixed sexes, was assumed to be mated to a hen for calculations of the index of ducks per square mile for each provincial stratum. Unidentified birds were allocated to species and sexes on the basis of the proportions in the identified segment. Aerial surveys were designed so that the sampling error of the total duck index, for each province, would be less than 20 per cent at the 0.05 probability level. Smith (1964) in his recommendations for waterfowl breeding ground aerial surveys noted that pairs, lone drakes, and flocked drakes should be enumerated for mallard, pintail, and canvasback in the May pair surveys, but that groups of two or more drakes or groups of three or more birds of mixed sexes should not be recorded in the July production surveys. In July, late nesting indices should be arrived at by enumerating pairs or single drakes only. Hammond (1966) suggested enumeration of drakes in groups of up to five on large study blocks but only those of two or less on small study areas, i.e., less than 640 acres. Widgeon and shoveler grouped males were not to be tabulated. Lone males were to be enumerated but not lone females, except for lone diver females and hens on artificially constructed ponds, where the waiting drake may be located on a large nearby marsh.
For divers, with heavily distorted sex ratios, counts of all males would naturally lead to over-estimation of breeding populations. Murdy (1964) enumerated only observed pairs of lesser scaup and ring-necked ducks on the Yellowknife Study Area, Northwest Territories. In Manitoba, Rogers (1964) considered pairs and lone females as indicated pairs of lesser scaup. In western Montana, Lokemoen (1966) considered only pairs of redheads in estimating breeding populations and discounted lone drakes and lone hens.
European census methods
In Finland, Koskimies (1949), Hilden (1964), and Grenquist (1965, 1966) utilized counts of males, pairs, and the number of nests found in relation to pair numbers to estimate yearly changes in waterfowl population numbers. Most of the counts were conducted around islands of various archipelagos and included velvet scoters (Melanitta fusca), mergansers (Mergus serrator, M. merganser), tufted duck (Aythya fuligula), mallard, pintail, goldeneye, shoveler, and teal (Anas crecca, A. querquedula). Koskimies (1949) discussed the methodological aspects of hourly, daily, and seasonal estimates of island-nesting velvet scoter and red-breasted merganser (M. serrator) populations. He concluded that time of day, stage of summer, weather conditions, personnel experience, and other environmental factors all affected the validity of estimates obtained. With velvet scoters, morning and afternoon counts gave comparable results during the early spring. Summer counts showed greater variability, as males left the breeding area and fewer pairs were enumerated. Nonbreeding yearlings were separable by plumage. Difficulty was experienced in separating nonbreeding and breeding components of merganser populations.In Iceland, Bengtson (1967) used a composite of methods to arrive at breeding populations of 15 duck species on Lake Myvatn. The lake covered 14 square miles, had an irregular rocky shore line of over 20 miles and contained 30 islands. Direct counts were used in the latter half of May and continued until egg laying was well advanced. Preliminary counts were adjusted when results of nest studies, moulting area counts, and brood studies were available. He felt there was an error of only 15 per cent in his estimates on a population that lay between 13,500 and 18,500 pairs.
For grassland areas I recommend, in a following section, the periodic enumeration of all dabbler pairs, lone drakes, and grouped drakes as indicated breeding pairs before specific dates, based on nest phonology, and the correction of the data with a prelaying season sex ratio to account for unmated males. Lone hens are not enumerated unless they are from an uncommon breeding species. Because of highly distorted sex ratios and aggregation of pairs on certain preferred ponds, diver population estimates are better taken from nesting studies supplemented with periodic counts of pairs and lone drakes on waiting sites.
In all, a number of different population components have been utilized to estimate breeding pair abundance or to arrive at some population index. Although most authors recognize that their estimates are crude, few have attempted to show the magnitude of errors of estimate or describe biases encountered in the use or rejection of a population component. The need for an evaluation of what components to count for each species, what time of year they should be counted, and the standardization of censused components between workers and areas is obvious.
Previous Section -- Migration, Inventory, and Population Components
Return to Contents
Next Section -- Appraisal of Factors Influencing Inventories
