Response of Nesting Ducks to Predator Exclosures and Water Conditions During Drought

Introduction

Low recruitment is a problem for managers attempting to increase populations of mallards and other upland-nesting ducks in the Prairie Pothole Region of North America (Cowardin et al. 1985, Klett et al. 1988, Sargeant and Raveling 1992). Low nest success has been cited as a major cause (Klett et al. 1988, Johnson et al. 1992). In 1985, biologists at Northern Prairie Wildlife Research Center (NPWRC) proposed the hypothesis that management to increase nest success on a small area could improve recruitment from females in a larger surrounding area and result in a local increase in breeding population. The hypothesis was derived from evidence that (1) successfully nesting mallard and gadwall females tend to return to the same sites in subsequent breeding seasons (Lokemoen et al. 1990), (2) female offspring of these species tend to return to their natal sites to breed (Doty and Lee 1974, Lokemoen et al. 1990), and (3) females of these and other duck species sometimes tolerate high nesting densities where predators are excluded via management (Duebbert and Lokemoen 1980) or by natural barriers (Duebbert et al. 1983). A Small Unit Management Project (SUMP) designed to test the hypothesis was initiated in 1987 and continued through 1991.

Our ability to test the original SUMP hypothesis was hampered by severe drought that began in 1988 and continued until the study ended in 1991. The drought caused a dramatic decrease in available water which, in turn, reduced numbers of ducks attracted to the study areas to nest (Fig. 1) and reduced survival of broods that were produced (G. L. Krapu, NPWRC, unpublished data). To compensate for this setback, we also tested a second hypothesis closely related to the original: If exclosures increased nest success, the proportion of the females from the entire study area that nested in the exclosure would increase with time.

Intensive ground surveys and complete habitat inventory by aerial video developed for this study also provided an opportunity to compare the performance of a regression ratio estimator, used in this study and in previous surveys (Cowardin et al. 1995a), to that of 2 other frequently used ratio estimators.

In this paper, we present results of hypothesis tests, evaluate the methodology used to determine breeding populations, and document the effect of drought on breeding populations of 5 duck species: mallard, gadwall, blue-winged teal (hereafter, teal), northern shoveler (hereafter, shoveler), and northern pintail (hereafter, pintail).

Study Areas

In 1985, we selected 3 51-km² circular study areas in each of 3 geologic areas of the Prairie Pothole Region: glacial moraine near Kulm, North Dakota; glacial drift near Jamestown, North Dakota (Bluemle 1977); and glacial terminal moraine and ground moraine near Detroit Lakes, Minnesota (Leverett 1932). Each study area contained a moderate to high density of natural wetlands and a centrally located Waterfowl Production Area (WPA). We refer to the study areas as Kulm, Jamestown, and Detroit Lakes, as did Pietz et al. (1993).

Wetlands covered 18.7% of Kulm, 12.9% of Jamestown, and 17.5% of Detroit Lakes (Table 1). Much of the wetland area at Detroit Lakes had a permanent water regime (50.8%), whereas permanent water was negligible at the Kulm and Jamestown sites. Cropland and fields enrolled in the Conservation Reserve Program (CRP) constituted the most common landcover on all study sites (Kulm = 34.7%, Jamestown = 74.9%, Detroit Lakes = 62.7%).

Sargeant et al. (1993, 1995; NPWRC unpublished data) obtained an index of predator populations from track surveys conducted throughout each study area in 1988, 1989, and 1990. These surveys indicated that red foxes (Vulpes vulpes) and raccoons (Procyon lotor) were numerous and striped skunks (Mephitis mephitis) were common on all 3 study areas. Mink (Mustela vison) occurred on all 3 study areas but were common only at Detroit Lakes. Badgers (Taxidea taxus) were common only at Kulm and Jamestown, and coyotes (Canis latrans) were common only at Kulm.

Methods

We assigned 3 treatments (predator-exclosure, predator-removal, control) at random to study areas within each geologic area. We selected predator-exclosure sites on the WPAs to meet several criteria: suitable for establishing a 25-ha block of upland nesting cover, relatively flat terrain, fertile soil, and limited competing nesting cover.

The control treatment served to indicate natural fluctuations in breeding populations, and the predator-removal treatment was evaluated by Sargeant et al. (1995). In this paper, we evaluate the predator-exclosure treatment.

Mapping Land Cover and Wetland Basins

The National Wetlands Inventory (NWI) mapped our study areas from high-altitude (1:63,360) color-infrared photographs taken in the early 1980s. Maps were similar to standard NWI maps, except that upland polygons were added, and wetland polygons were identified as belonging to discrete wetland basins. Classification of wetland polygons followed Cowardin et al. (1979); classification of uplands and assignment of wetland polygons to basins followed Cowardin et al. (1988b). The NWI digitized the maps and furnished us with vector files. We updated the vector files by adding polygons for CRP cover obtained from county offices of the Agricultural Stabilization and Conservation Service (currently, Farm Service Agency). We updated other land-cover changes since the time of photography via ground observation and aerial video obtained during the study.

Predator Exclosures

In 1985 and spring 1986, proposed predator exclosure sites were repeatedly tilled and twice sprayed with herbicide to remove existing vegetation. In spring 1986, we seeded all sites to dense nesting cover (DNC; Duebbert et al. 1981), via a pure, live-seed mixture containing 6.7 kg/ha of tall wheatgrass (Agropyron elongatum), 4.5 kg/ha of intermediate wheatgrass (A. intermedium), and 1 kg/ha of alfalfa (Medicago sativa). During 1987, vegetation became established as seeded, except that switchgrass (Panicum virgatum) developed on one-half of the Detroit Lakes cover plot. Planted cover occupied 96.3% of the area inside exclosures (Table 1).

Predator exclosures were constructed in 1986-87 by Ducks Unlimited and became fully operational in 1988. Exclosures were designed to exclude red foxes, striped skunks, raccoons, badgers, and coyotes and were described in detail by Pietz and Krapu (1994). We added 8 pairs of ground-level openings to each exclosure during the 1988 nesting season to provide exits for female ducks with broods. In 1990, the number of brood exits was expanded to 32 (Pietz and Krapu 1994). Beginning in 1988, we maintained quick-kill traps inside each exclosure throughout the nesting season to ensure exclosures remained as predator-free as possible. We maintained track plots inside exclosures to monitor the effectiveness of predator control.

Duck Nest Data

We located nests inside predator exclosures by dragging a 61-m chain between 2 vehicles to flush nesting females (Klett et al. 1986). We marked each nest location with a willow stick and numbered flag placed 4 m north of the nest bowl. In 1987-91, we conducted 6-7 nest searches per year in each predator exclosure, beginning in late April or early May. We checked nest status every 2 weeks during subsequent nest searches.

We calculated nest densities using all duck nests found inside an exclosure, divided by the total number of hectares in that exclosure (all cover types combined). We estimated nest success rates with the Mayfield (1961, 1975) method as modified by Johnson (1979); we excluded nests likely affected by the investigators. Rates reported throughout this paper are for Mayfield nest success unless otherwise stated.

Area and Numbers of Ponds

We determined the number of ponds (wetland basins containing water) and area of each pond from color aerial video taken during the first week of May each year (Cowardin et al. 1988a). We obtained video at an elevation of approximately 1,830 m above ground level. Details of the camera, other hardware, and their configuration in the aircraft were described by Cowardin et al. (1988a). Video scenes were captured as raster files by a microcomputer equipped with a TARGA 16 image-capture board and Map and Image Processing System (MIPS) software (MicroImages, Lincoln, Nebraska, USA). (Use of trade names does not imply endorsement by the U.S. Geological Survey.) Aerial video did not uniquely identify individual ponds, but sizes of specific ponds were required to estimate a correction factor, γ, described in the next section. We determined sizes of specific ponds from NWI maps by multiplying the area of each pond's wetland basin by a visual estimate of the proportion covered with water.

Estimating Duck Populations

We counted breeding pairs on sample tracts delineated within property boundaries of individual landowners. Approximately one-half of each study area was included in sample tracts. Most tracts were legal quarter sections (65 ha), which facilitated the identification of boundaries, although smaller tracts were delineated when property lines or study area boundaries subdivided quarter sections. Each tract was assigned a random number. We contacted landowners of tracts in numerical order to obtain permission to access wetlands. If permission was denied, we selected the next tract. We surveyed wetland basins that overlapped property boundaries only when we could obtain permission to access both properties.

We conducted 2 pair counts per year on each wetland basin in a selected tract. We conducted the first count from 1 to 15 May, and conducted the second count from 20 May to 5 June. We visited tracts in numerical order. The first count was used for estimating mallard and pintail populations, the second count for gadwall and teal populations, and the count nearest 15 May for shoveler populations. Birds were tallied by social groups (Dzubin 1969). We used counts of observed pairs, lone males, and males in flocks ≤ 5 to estimate numbers of breeding pairs, except that flocked males were not included for shovelers.

We computed 3 different ratio estimates (Cochran 1977:150-156) of breeding population for each species, study area, and year: a land-area ratio estimate (_L), a water-area ratio estimate (_W), and a regression ratio estimate (_R). To test hypotheses about changes in population, we compared results of all 3 methods before using regression ratio estimates, which we preferred for theoretical reasons (Cowardin et al. 1995a).

The land-area ratio estimator divided the ground count of pairs (y) for a species, study area, and year by the proportion of land area (a_L/A_L) surveyed to produce an estimate of the total breeding population. The water-area ratio estimator operated similarly but divided the ground count by the proportion of water area (a_W/A_W) surveyed. The regression ratio estimator was

where γ corrected for annual and geographic variation, a_p was the area of pond p, N was the number of ponds on the study area, and f(a_p) was an uncorrected estimate of breeding population, f(a_p)= A × (Area) + B × . For A and B, we used regression coefficients provided by Cowardin (1995a:7). We computed the correction factor for each study area and year as

where y is the number of ducks counted and n is the number of ponds in the sample. We based uncorrected population estimates on pond areas derived from aerial video. We assumed land and water area were measured without error.

Relations between Pairs and Ponds

We used multiple regression to relate numbers of pairs to numbers or area of ponds. Our regression model was y_ij = β_0i + β_1ix_ij, where y denoted numbers of pairs, x denoted either number or area of ponds, subscript i distinguished 3 study areas, and subscript j distinguished 5 years. We first tested equality of slopes among study areas (H₀: β₁₁ = β₁₂ = β₁₃ = β₁). When slopes were not different (P < 0.05), we reduced the model to y_ij = β_0i + β₁x_ij and compared the common slope to zero (H₀: β₁ = 0). We calculated the proportion of total variation explained by changes in pond numbers or area (denoted by R^2*) for the most detailed model that contained only statistically significant slope terms.

We expected philopatric species such as mallards and gadwalls to crowd onto remaining ponds as drought reduced available water. To test this expectation, we regressed the number of pairs per study area and year against the number of ponds for each of the 5 species. We used a 1-tailed t-test to compare the resulting slope estimate to the mean number of pairs per pond. If the mean number of pairs per pond exceeded the slope estimate, we concluded birds concentrated on remaining ponds when pond numbers declined. Prior to performing t-tests, we used F-tests to confirm homogeneity of variances for slope estimates and means.

Effects of Predator Exclosures

To determine whether populations increased on treatment areas relative to control areas (the original SUMP hypothesis), we compared breeding duck populations on control study areas to those on study areas with predator exclosures. For each species and for all 5 species combined, we calculated the ratio of pairs per treated study area/pairs per control study area, and averaged the ratios across geographic areas. We used Spearman's rank-order correlation coefficient (hereafter, rank correlation, ρ) to test for positive associations between ratios and years (i.e., for increasing trends in ranks over time). This procedure precluded the need to make distributional assumptions or to specify the form of increases that potentially occurred.

We also used rank correlation to test for increases in an index to the proportions of ducks nesting within predator exclosures (the second SUMP hypothesis). As an index to the proportion of nests within the exclosure, for each study area and year, we used the number of nests inside the exclosure, divided by the regression ratio estimate of the breeding population. When testing the hypothesis for individual species, we ranked within study area to adjust for differences unrelated to treatment effects. We excluded the Detroit Lakes study area for shovelers, gadwalls, and pintails because few pairs nested within the exclosure. We regarded our measure as an index because we could not determine the proportion of observed pairs that nested on the study area, and because some females may have initiated more than 1 nest inside an exclosure.

Results

The pattern of change in ponds was similar among study areas (Fig. 1), although water loss was more pronounced at the North Dakota sites (Kulm, Jamestown) than in Minnesota (Detroit Lakes). Wettest conditions occurred in 1987, a near-normal year for pond abundance in the northcentral United States (U. S. Fish and Wildlife Service and Canadian Wildlife Service 1987). Pond numbers and area of ponds declined sharply in 1988; pond numbers recovered in 1989 (due to temporary water in small basins), but area of ponds did not. Pond numbers were lowest in 1990, with little improvement in 1991.

Population Estimates

On average, regression ratio estimates of breeding population were slightly lower than land-area ratio estimates and slightly higher than water-area ratio estimates. However, all 3 methods produced similar results (Fig. 2). Thus, conclusions were robust to our choice of method for estimating populations and we used only 1, the regression ratio estimator, for reasons described in Cowardin et al. (1995a).

Breeding populations showed a general decline on the North Dakota study areas (Kulm, Jamestown) and reached lows in 1990 (Fig. 1). Teal populations dropped most sharply, especially on Kulm. In North Dakota, populations of the other 4 species remained relatively stable from 1987 through 1989 but declined in 1990 and 1991 as drought intensified. On the Minnesota study area (Detroit Lakes), populations of all 5 species remained relatively stable.

Relations of Pairs and Ponds

Pair numbers were positively related to area of ponds for all 5 species (Table 2). Pair numbers were positively related to number of ponds for mallards, teal, shovelers, and pintails, but not gadwalls. Area of ponds explained more variation in numbers of pairs than did number of ponds for all species except shovelers.

As the number of ponds decreased, the number of pairs per pond increased for mallards (t₂₇ = 3.27, P = 0.001) and gadwalls (t₂₅ = 2.93, P = 0.004), but not for teal (t₂₇ = -0.45, P = 0.672), shovelers (t₂₇ = -0.72, P = 0.760), or pintails (t₂₅ = -1.54, P = 0.932).

Nesting in the Predator Exclosures

Average nest density inside the 3 exclosures was 1.2 nests/ha for all species for the 5 years of study. Mean nest densities inside individual exclosures were 1.7 nest/ha for Jamestown, 0.9 nest/ha for Kulm, and 0.7 nests/ha for Detroit Lakes. Overall means were 0.3 nests/ha for mallards and 0.6 nests/ha for gadwalls. Mallards and gadwalls together composed 76% of the nesting population, followed by teal (15%), pintails (5%), and shovelers (4%; Table 3). Gadwalls contributed the greatest number of nests on each of the 2 North Dakota study areas, whereas teal contributed the greatest number of nests on the Minnesota study area.

Nest success of all 5 species was high inside the predator exclosures. Nest success in 1987 averaged only 48% for the 3 exclosures, probably because they were not fully operational that year. Nest success varied from 70 to 84% during 1988-91, and averaged 72% over all 5 years. Nest success averaged 45% at Detroit Lakes, 76% at Kulm, and 80% at Jamestown for all years and species combined. Nest success for mallards across years varied from a low of 45% (95% CI = 28-70) at Detroit Lakes to a high of 66% (95% CI = 50-87) at Kulm, with a mean of 60% (95% CI = 50-72) across study areas. Mean nest success for gadwalls was 81% (95% CI = 75-88) and varied little among study areas. Although nests within exclosures were protected from most mammalian predation, 75% of unsuccessful nests were destroyed by predators.

Effects of Predator Exclosures

From 1987 through 1991, there was no evidence of an increase in pairs per treated area/pairs per control area for mallards (ρ = 0.40, P = 0.252), gadwalls (ρ = 0.50, P = 0.196), teal (ρ = 0.20, P = 0.374), shovelers (ρ = 0.40, P = 0.252), pintails (ρ = -0.20, P = 0.626), or all species combined (ρ = 0.20, P = 0.374). However, results for our second hypothesis indicated 4 of 5 species did respond to the predator-exclosure treatment. The index to the proportion of females nesting in the predator exclosures increased with time for mallards (ρ = 0.9, P = 0.0004), gadwalls (ρ = 0.8, P = 0.009), teal (ρ = 0.54, P = 0.021), and pintails (ρ = 0.8, P = 0.009), but not for shovelers (ρ = 0.15, P = 0.333). Although the proportion of nests within exclosures increased on all 3 study areas (P ≤ 0.045), the treatment effect was much greater at Jamestown than at the other 2 study areas (Fig. 3).

Discussion

Our test of the hypothesis that duck populations on study areas with predator exclosures would increase relative to populations of control study areas was inconclusive. Variability among study areas, drought, and an inadequate number of study areas probably resulted in our inability to detect treatment effects under the original SUMP hypothesis. We were successful, however, in testing the second hypothesis related to the exclosure treatment and demonstrated that an index to the proportion of females nesting in exclosures increased with time.

Waterfowl biologists have observed that ducks settle in a breeding area in response to wetland conditions, which are usually measured as the number of ponds present (Crissey 1969). Johnson and Grier (1988) reviewed literature on the subject and proposed 3 different patterns of settling including opportunistic settling in response to habitat conditions. Our data confirm the relation between breeding population and wetland conditions, as measured by area and number of ponds, and demonstrate the dramatic effects of the drought that extended from 1988 through 1991.

We have shown that mallard and gadwall pairs per pond increased as numbers of ponds decreased. Smith (1971), Duebbert and Lokemoen (1980), and Duebbert et al. (1983) all found that mallard and gadwall pairs per pond were greater in areas where breeding success was high and wetland area was stable or declining. Duebbert and Lokemoen (1980) suggested that breeding pairs may have adapted to higher densities by reducing the size of the area they defended.

Little is known about the reproductive consequences of increased pair densities. Presumably, limiting factors (e.g., availability of food or access to food) must start to operate at some population level. Many authors have suggested that undisturbed feeding time is important for female ducks during nesting (e.g., McKinney 1973, Dwyer 1974, Seymour and Titman 1978, Stewart and Titman 1980, Mjelstad and Saetersdal 1988). Increased levels of agonistic interactions have been associated with higher pair densities (McKinney 1965, Duebbert 1966, Jackson et al. 1985). Such interactions frequently involve females being chased from occupied feeding sites. On a wetland where mallard pair numbers were as high as 26/ha, Lokemoen et al. (1984) observed >1 aggressive interaction every 5 min. In the vicinity of a 2.8-ha island on which >100 gadwall nests were found, Duebbert (1966) often observed up to 50 gadwall "pursuit flights" per hour. Nevertheless, we do not know if such high rates of agonistic interactions negatively affect reproduction.

The average of 1.2 nests/ha and 72% nest success in predator exclosures demonstrated their effectiveness for nesting waterfowl. Working in the same regions and years on study areas without predator management, Sargeant et al. (1995) reported 301 nests on 1,771 ha and a mean nest success of 5.6%. Working mainly on cultivated lands in similar terrain, Milonski (1958) and Higgins (1977) both found 0.02 nests/ha, and Higgins noted 25% apparent nest success. In a more diversified landscape, Duebbert and Kantrud (1974) found 0.14 nest/ha and 51% nest success on an area where extensive predator control had been practiced. Nest densities and nest success in the SUMP exclosures were more similar to those found in other predator exclosures in North Dakota (3.4 nests/ha, 65% success) and western Minnesota (1.5 nests/ha, 55% success) by Lokemoen et al. (1982). However, the nest densities found in the SUMP exclosures were well below levels attainable by these species. Densities of 107.8 nests/ha were observed on Miller Island in North Dakota where apparent nest success was 79-98% (Duebbert et al. 1983).

Ducks may select nest sites inside predator exclosures in response to attractive nesting cover, high rates of daily nest survival, homing, or other behavioral cues. Mallards and gadwalls probably preferred nesting in the tall, dense grasses and forbs within the exclosures, as this habitat is their favored nesting cover (Duebbert and Kantrud 1974). Mallards and gadwalls could concentrate nests in preferred cover locations because they can move several kilometers from wetland feeding sites to nesting locations (Lokemoen et al. 1984). Also, adult females of these 2 species return at a high rate to successful nesting sites of the previous year, and young females return, albeit at a somewhat lower rate, to natal sites. Blue-winged teal, northern shovelers, and northern pintails all home at lower rates than mallards and gadwalls (Poston 1974, Johnson and Grier 1988, Lokemoen et al. 1990). Blue-winged teal and northern shovelers both have small home ranges (e.g., Nudds and Ankney 1982) and do not nest far from wetlands.

Renesting may allow nest densities of mallards and gadwalls to increase rapidly, even within a breeding season (Lokemoen et al. 1984). Some renesting females may nest in predator exclosures by chance, after moving from habitats where their previous nests were destroyed. It is possible that some females select nest sites in exclosures in response to behavioral cues from females already nesting there.

In 1990, the number of gadwall nests in the Jamestown exclosure (41) exceeded the estimated number of gadwall pairs on the study area (24). This disparity may represent some combination of sampling error, unusually long movements by females during a drought year, a geographic redistribution of pairs during the breeding season, and renesting. Based on home range estimates from this study (G. L. Krapu, NPWRC, unpublished data) and others (e.g., Gates 1962, Drewien 1968, Poston 1974, Derrickson 1978, Dwyer et al. 1979, Cowardin et al. 1985), study areas were large enough to encompass typical home ranges of ducks nesting within exclosures.

Our results suggested that the regression, land-area, and water-area ratio estimators produce similar estimates for pond conditions and species encountered in this study. The regression ratio estimator has several advantages because its use of remote sensing data (1) often allows complete coverage of large areas rather than sampling, (2) enables estimates to be made for areas where ground access is not possible, and (3) facilitates comparison of duck estimates to habitat characteristics. Furthermore, the regression ratio estimator is biologically more realistic than the water-area ratio estimator because it assumes a curvilinear rather than linear relation between pair numbers and pond size (Cowardin et al. 1995a).

Management Implications

We demonstrated that ducks nested in high densities inside predator exclosures, and that the proportion of the breeding populations nesting in exclosures increased with time for mallards, gadwalls, teal, and pintails. These results suggest that management applied to small areas can increase recruitment from females in a larger area, if brood survival is adequate. This concept can help managers increase the efficiency and cost effectiveness (Cowardin et al. 1995b) of techniques designed to increase recruitment. The same concept probably applies to unmanaged populations where highly productive segments of the population (e.g., those nesting on predator-free islands) may compensate for unproductive segments.

Information from this study of breeding population responses to exclosures was limited by severe drought during the period of study. Previous work with exclosures during wetter periods has shown that higher duck nest densities can be achieved. The question of how high a local population might rise due to the successful operation of predator exclosures during good water conditions in the Prairie Pothole Region remains unanswered. We believe that further work designed to understand how small, locally productive populations of ducks function during periods of wet as well as dry conditions is essential to planning duck management and efficient use of scarce management funds.