---

---

Introduction

Upland nesting waterfowl have low rates of reproduction in the prairie pothole region because of severe habitat change and increased predation (Sargeant and Arnold 1984, Cowardin et al. 1985:33-35). The destruction of upland nests by predators has generated interest in management techniques that protect nesting hens, such as electric fences (Lokemoen et al. 1982), straw bales (Giroux et al. 1983), and islands (Willms and Crawford 1989). High nest densities and high nest success have been found on predator-free natural islands (Hammond and Mann 1956, Newton and Campbell 1975, Duebbert et al. 1983). Managers have attempted to duplicate successful natural island habitats by constructing earthen and rock islands (Giroux 1981, Giroux et al. 1983). However, island construction is expensive (Lokemoen 1984). An alternative strategy is to manage existing islands on government and private property.

The objectives of this study were to determine the extent of waterfowl nesting and rate of nest success on islands in North Dakota, South Dakota, and northeastern Montana and to identify island habitat features associated with nesting waterfowl and water birds.

Study Area

The study was conducted in the prairie pothole region of North Dakota, South Dakota, and northeastern Montana during 1985 and 1986. Islands were defined as any land mass (≥0.04 ha) completely surrounded by water during the survey period. Islands were found by examining aerial photographs or satellite imagery; their locations were plotted on county highway maps. We and cooperating biologists from the 3 states visited island locations during the waterfowl nesting season. Seventeen islands examined were in impoundments and the remainder were in natural wetlands. Forty-nine of the islands were human-made and the remainder were natural.

Methods

Nest and Island Surveys

Two nest searches (1 each during May and June) were conducted on every island. From 2-6 people walked through the vegetation in an attempt to locate nests. On 5% of the nest searches, a weighted rope was dragged over the cover by 2-3 people. The location of the nest, the stage of egg incubation, and nest success were determined (Klett et al. 1986). Apparent nest success, rather than the Mayfield estimator (Johnson 1979), was used to estimate nest success. Johnson and Shaffer (1990) indicated that the apparent estimator was the best method for islands because investigators were likely to find all or most nests and losses tended to be catastrophic.

During the first visit to each island, observers recorded wetland class and subclass (Stewart and Kantrud 1971:5-10), distance to nearest shore, number of wetlands with water within 1.6 km, maximum water depth between shore and island, and island and wetland size. Cover on the islands was separated into percent low shrubs (≤1.25 m), tall shrubs (>1.25 m), grasses, forbs, and bare ground. Each cover type was ranked by averaging 20 visual obstruction readings (VOR) from a modified Robel pole (Kirsch et al. 1978). Vegetation on the islands was not altered by haying, grazing or tillage.

Statistical Analyses

We used stepwise logistic regression to identify habitat variables that could explain the presence or absence of ≥1 nest on an island (SAS Inst. Inc. 1990:175-229). Models were developed for each species and year (1985 and 1986). Islands were coded as 1 or 0 if the species did or did not nest on the island, respectively. Variables were deemed significant if they entered and stayed in the model at α ≤ 0.10. We chose logistic regression over multiple regression because of the large number of islands having no nests. The large number of 0's violates the assumptions of normally distributed variables and homogenous variances necessary for multiple regression. Simple correlations among habitat variables were examined to guard against multicollinearity that can cause regression coefficients to be unstable and difficult to interpret (Snedecor and Cochran 1980:352-353). The 2 highest correlations were between low shrub and VOR (r = 0.67) and between grass and forbs (r = −0.67). All remaining correlations were <0.5, and many were near 0. When variables were significant in the models in both years for a species and the sign of their regression coefficient was the same, we calculated the means of these variables for islands with and without nests.

Mammalian predators were removed from 24 islands using mainly 220 conibear traps set in boxes plus a few leg-hold traps. Trapping extended from early April to mid-July. To determine if nest success was affected by predator control, we used analysis of variance, fitted by the method of weighted least squares, to compare nest success between islands with and without predator control (Snedecor and Cochran 1980:230-232). Weights were calculated as the square root of the total number of nests on each island. This weighting procedure was chosen as a compromise between weighting by actual number of nests and not weighting at all. The former provided an estimator of average nest success that has minimum variance but may be biased. The latter was unbiased but more variable than the former.

Results

We made 253 surveys of habitat features and waterfowl nesting on 209 islands. Forty-four islands were visited during both years. The mean size of the islands was 1.36 ha and ranged from 0.04-21.45 ha.

Species Composition

Gadwalls (Anas strepera) and mallards (A. platyrhynchos) were the most abundant species on the islands and together composed 74% of the waterfowl nesting population (Table 1). We compared the species composition of breeding waterfowl in the Dakotas during 1985 and 1986 (U.S. Fish and Wildl. Serv. 1990) with that for nesting waterfowl on islands and found use by nesting mallard, gadwall, lesser scaup (Aythya affinis), and Canada goose (Branta canadensis) disproportionately higher than the Dakota breeding population. Blue-winged teal (Anas discors), northern shoveler (A. clypeata), northern pintail (A. acuta), and redhead (Aythya americana) were proportionally less common on islands than in the Dakota breeding population. American wigeon (Anas americana), green-winged teal (A. crecca), and ruddy duck (Oxyura jamaicensis) were rare.

Besides waterfowl, 12 other species nested on the islands. These included the ring-billed gull (Larus delawarensis [728 nests on 4 islands]), common tern (Sterna hirundo [109 nests on 9 islands]), double-crested cormorant (Phalacrocorax auritus [796 nests on 6 islands]), American white pelican (Pelecanus erythrorhynchos [1,384 nests on 4 islands]), and American avocet (Recurvirostra americana [446 nests on 32 islands]).

Nest Density

Waterfowl nests were found on 66% of the islands studied. We found 2,274 duck nests and 115 Canada goose nests on the islands. An average island had 9.0 nests (SE = 1.82), and 3 islands contained >100 waterfowl nests/year. Overall duck density was 6.6 nests/ha (SE = 1.54) for both years, including mean densities of 2.3 mallard, 3.0 gadwall, 0.4 blue-winged teal, 0.3 northern pintail, and 0.5 lesser scaup nests/ha. There was a mean density of 0.3 Canada goose nests/ha.

Nest Success

The weighted average nest success for 11 waterfowl species was 60% on islands without predator control and 76% on islands with predator control (Table 2). Nest success for mallards, gadwalls, all ducks, and all waterfowl was higher (P < 0.01) on islands with predator control than on islands without predator control. High nest success rates were attributed to the natural low density of mammalian predators on many islands and their removal from 24 islands. Nest success on islands was correlated with nest density ranging from a low for blue-winged teal (r = 0.65, n = 73, P ≤ 0.001) to a high for lesser scaup (r = 0.99, n = 46, P ≤ 0.001).

Habitat Correlates of Nesting Waterfowl and Water Birds

Mallard. — The only significant association for mallards in both years was VOR (Table 3). The mean VOR for islands with mallard nests in both years was 2.3 (SE = 0.18, n = 99) compared to 1.1 (SE = 0.10, n = 154) for islands without mallard nests. In 1 of 2 years, the regression model indicated that mallards tended to nest on islands with higher salinity and more surrounding wetlands and avoid islands with tall shrubs.

Gadwall. — Presence of gadwall nests during both years was positively associated with wetland salinity, distance to shore, and low-shrub cover but was negatively associated with tall-shrub cover. Mean wetland salinity for islands with gadwall nests was 6,166 micromhos/cm³ (SE = 649.25, n = 72) compared to 2,485 micromhos/cm³ (SE = 346.14, n = 142) for islands without gadwall nests (see Stewart and Kantrud 1971 for salinity descriptions). For both years, islands with gadwall nests were an average of 134 m from shore (SE = 11.50, n = 89), whereas islands without nests averaged 79 m from shore (SE = 6.17, n = 164). Islands with gadwall nests contained 19% low shrubs (SE = 2.38, n = 89), which was primarily Wood's rose (Rosa woodsii) and snowberry (Symphoricarpos occidentalis), and 0.8% tall shrubs (SE = 0.35, n = 89), primarily wild plum (Prunus americana) and northern hawthorn (Crataegus rotundifolia). The mean density of tall shrubs was low because it was calculated from islands with nests, and only a few islands contained tall shrubs. For islands containing tall shrubs, the mean percent gadwall occurrence was 17%, indicating that where tall shrubs occurred, they were an important plant component with a strong negative influence on nesting. Islands without gadwall nests contained 7% low shrubs (SE = 1.05, n = 164) and 3% tall shrubs (SE = 0.68, n = 164). Gadwall nesting on islands was positively related to the number of wetlands within 1.6 km in 1985 and was negatively associated with bare ground in 1986.

Blue-winged Teal. — In both years, the likelihood of nesting by blue-winged teal on islands increased with increasing distance to shore. The mean distance to shore for islands with nests was 125 m (SE = 12.59, n = 73) compared to 88 m (SE = 6.47, n = 180) for islands without nests. In the 1985 model, blue-winged teal nesting was positively related to wetland salinity and the number of wetlands within 1.6 km and negatively related to tall shrub and bare ground. In 1986, The probability of blue-winged teal nesting increased on larger islands, on islands with a higher VOR, or in Class V wetlands.

Northern Pintail. — We did not detect any relationships between habitat and the probability of northern pintail nesting that were consistent between years. Significant single-year relationships indicated an association with islands in saline and Class V wetlands and islands with increased numbers of wetlands within 1.6 km and low-shrub cover. There were negative associations with tall shrubs and bare ground.

Lesser Scaup. — In both years, islands with lesser scaup nests were located in more saline wetlands ( = 6,671 micromhos/cm³ [SE = 896.09, n = 38]) than islands without scaup nests ( = 3,106 micromhos/cm³ [SE = 377.23, n = 176]). Also, significantly more islands in Class V wetlands (35.3%) had lesser scaup nests than did islands in seasonal and semipermanent wetlands (19.3%).

American Avocet. — Island nesting by avocets was associated with increasing amount of bare ground in both years. Islands with and without avocet nests had 16% (SE = 4.25, n = 32) and 5% (SE = 1.09, n = 221) bare ground, respectively. In 1986, American avocet nesting was associated with saline wetlands.

Discussion

Species Composition

High island use by nesting Canada geese was understandable given their known tendency to readily seek insular nest sites (Giroux et al. 1983). Mallards, gadwalls, and lesser scaup probably dominate island populations because adult hens and some of their female young return to successful natal areas (Lokemoen et al. 1990:17-20; A. D. Afton, Louisiana Coop. Wildl. and Fish. Unit, Baton Rouge, pers. commun., 1991). Also, mallards, gadwalls, and lesser scaup have large home ranges and can move long distances between feeding, loafing, and nesting sites (McKinney 1965). Blue-winged teal and northern shoveler may not be common nesters on islands because they have small home ranges and do not move far (e.g., across open lakes) to nest. Also, blue-winged teal have low homing tendencies (McKinney 1965). Giroux (1981) and Duebbert (1982) recorded mallards, gadwalls, and lesser scaup as common island nesters (67-75%).

Nest Densities

Duck nest densities on islands in our study were higher than the 0.3 nests/ha found by Duebbert et al. (1986) and 0.5 nests/ha reported by Higgins et al. (1992) in North Dakota upland habitats. However, nest densities on islands in our study were similar to those observed in other island studies. In the prairie region of Canada, Giroux (1981) found an average of 13.8 nests/ha. In North Dakota, Duebbert (1982) found 10.9 and 13.2 nests/ha, and Willms and Crawford (1989) found 8.4 and 9.0 nests/ha.

Nest Success

Duck nest success on islands in our study was 4 times higher (62%) than the average nest success (16%) in recent studies conducted in uplands. Klett et al. (1988) reported an average Mayfield nest success rate of 15.2% for 5 duck species during 1980-1984, and Higgins et al. (1992) found 16.3% nest success in North Dakota upland habitats. Our observed nest success was intermediate to the Mayfield nest success of 24% recorded by Willms and Crawford (1989) and the apparent nest success rates of 48 % observed by Giroux (1981) and the 83 and 91% recorded by Duebbert (1982). Our ability to locate destroyed nests was reduced on some large islands; thus, our estimate of nest success may be inflated.

Nest success on islands was high and increased only moderately with predator control. However, predators can have a striking effect on waterfowl nest success. On 9 islands near Devils Lake, North Dakota, nest numbers and nest success increased from 52 and 8% before predator control to 851 and 87% during the first year of control (Lokemoen et al. 1987). These islands averaged 3.4 ha, and predators (e.g., red fox [Vulpes vulpes]) were residents. Smaller islands have fewer resident predators but attract mink (Mustela vison) and raccoons (Procyon lotor), which have wetland affinities.

Habitat Correlates of Nesting Waterfowl and Water Birds

Saline wetlands were universally associated with waterfowl nesting on islands. Saline wetlands may be beneficial for waterfowl because they provide poor habitat for mammalian predators, with few large invertebrates and small vertebrates as food, poor water quality, and/or inadequate cover (few emergent plants [Swanson et al. 1988]). Other highly successful islands have been in large, saline lakes or large, open lakes far from shore (Hammond and Mann 1956, Newton and Campbell 1975, Duebbert et al. 1983, Simpson and Limmer 1990). Similarly, the occurrence of duck nests was more likely on islands farther from shore because the expanse of open water acted as a deterrent to mammalian predators. Giroux (1981) found more nests located on islands farther from the mainland than on islands close to shore. He recommended that islands be constructed at least 170 m from shore and spaced at least 100 m apart. Nesting may have been enhanced on islands in Class V wetlands because they were permanent and generally provided a predator barrier. Also, islands in permanent wetlands were always surrounded by water; thus, homing females were able to return to successful nest sites (Simpson and Limmer 1990). However, there can be negative ramifications to nesting on islands in large lakes. Vermeer (1968) recorded California gulls (Larus californicus) killing newly hatched ducklings in Alberta. Gulls require bare ground for nesting, and they might be deterred from specific islands by maintaining some form of tall, dense cover on the entire site.

Islands with a high density of nearby flooded wetlands were more likely to contain nests, perhaps because these wetlands attracted additional breeding pairs. Flooded wetlands were probably more important near islands in 1985 because conditions were drier that year (U.S. Fish and Wildl. Serv. 1990). Increased nesting associated with low shrubs and a higher VOR was consistent with results from previous studies (Duebbert 1982, Hines and Mitchell 1983, Lokemoen et al. 1984). Dwernychuk and Boag (1973) noted that nest numbers declined when cover (mainly forbs) was reduced by herbicides. Though nest predators were uncommon on islands, dense nesting cover was required to attract hens and conceal nests from avian predators, visiting mammalian predators, and agonistic conspecifics (Lokemoen et al. 1984). Ducks probably avoid nesting in tall shrubs, because thick foliage and branches inhibit growth of low, dense vegetation needed for nest concealment. Also, tall shrubs and associated trees create nesting and perching sites for avian predators and cover for mammalian predators.

Island size was not strongly associated with nesting by ducks. Willms and Crawford (1989) found that island size was not an important variable influencing mallard and gadwall nest numbers, but they found an association between island size and blue-winged teal nest abundance. The selection of islands with increased amounts of bare ground in saline wetlands by American avocets was similar to previously reported habitat selection (Hamilton 1975).

Management Implications

Islands in the Dakotas and northeastern Montana had higher waterfowl nest densities and higher nest success than the surrounding uplands. Islands were most important to gadwalls, mallards, lesser scaup, and Canada geese, but were used by other waterfowl species, shorebirds, and colonial water birds.

An effective method of producing waterfowl and nongame birds in prairie environments is to locate and manage naturally occurring islands. Managing existing islands is efficient because there is usually little need to obtain long-term leases on land or building permits, hire expensive earth-moving machinery, or plant nesting cover. Our data suggested that there are hundreds of potentially manageable islands present in the prairie pothole region of the United States and Canada. Landsat or Spot data could be used to inventory island locations and numbers for waterfowl managers in the prairie region of North America.

Waterfowl nest success can usually be improved on existing islands, particularly those >1.5 ha, by removing mammalian predators. Trappers should visit islands soon after icemelt to remove resident animals. Islands <1.5 ha normally have few resident mammals. Establishing dense cover on islands should deter nesting and occasional use by gulls. Conversely, maintaining bare ground on islands should favor nesting by gulls, terns, and certain shorebirds, such as American avocets.

Nesting cover removed on islands by grazing can be restored by excluding livestock with fences. Islands lacking suitable nesting cover (e.g., VOR ≤1) for ducks can be improved by adding Wood's rose and snowberry (Snyder 1982) or mixtures of tall grasses and legumes. Low shrubs require hand labor to plant and maintain weed-free but once established will provide excellent nesting cover for many years. The vigor and attractiveness of grass-legume plantings decline over time. Also, removing tall shrubs and associated trees from islands should positively influence nesting by ducks. Management rights may have to be leased on private islands where fencing, predator removal, or cover plantings are implemented.

New islands should be built only at suitable situations to maximize the high cost of investment. Superior island locations are far from shore (≥100 m) in open, saline lakes (≥6,000 micromhos/cm³), with a high density (≥40) of peripheral wetlands within 1.6 km. Preferred waterfowl nesting cover should be established immediately on newly constructed islands. We have seen excellent results when grass-legume mixtures were planted (Duebbert et al. 1981:4-6) at the completion of construction during winter. Low shrubs should be planted the first spring following construction on a portion of the island not planted to grass and forbs. New islands should be visited regularly and predators removed. In some situations, new islands can be created most efficiently by removing a land bridge or by building on a shoal area in a lake.

---

Literature Cited

---

---

NPWRC Home | Site Map | About Us | Staff | Search | Contact | Web Help | Copyright