Effects of Food Supplementation on Depredation of Duck Nests in Upland Habitat

Abstract

We examined provision of supplemental food as a method for reducing depredation of upland-duck nests, especially by striped skunks (Mephitis mephitis). Predators negatively influence duck recruitment in grassland ecosystems. Managers are in search of methods, particularly nonlethal methods, for reducing nest depredation. We conducted this study on 24 areas managed for wildlife production in the Prairie Pothole Region of central North Dakota during 1993-1994. We provided a mix of fish offal and sunflower seeds on 12 areas; no food was provided on the remaining 12 control areas. Although we observed a tendency during both years for higher nest success rates on provisioned areas ( = 46%, 1993; 36%, 1994) than on control areas ( = 27%, 1993; 31%, 1994), mean nest success rates (Mayfield 1961) overall did not differ significantly between food-provisioned areas ( = 41%) and control areas ( = 29%). Striped skunk depredation rate was lower on food-provisioned areas (11%) than on control areas (24%), suggesting that skunks reduced their consumption of eggs when provided with a food supplement. In 1994, habitat conditions were optimal, and ducks nested persistently into the summer when nest success rates of food-provisioned areas and control areas differed by only 5 percentage points. That year American badgers (Taxidea taxus) and Franklin's ground squirrels (Spermophilus franklinii) apparently compensated for reduced depredation by skunks. Thus, although skunks and other mammalian predators seem to have responded positively to food provisioning, nest depredations overall did not change. Provision of supplemental food apparently has limited value for managing depredation of upland duck nests in the Prairie Pothole Region where predator communities are complex.

Key Words: duck, Mephitis mephitis, nest depredation, nest success, North Dakota, prairie, striped skunk, supplemental food, waterfowl.

Introduction

Predators severely limit recruitment in prairie-nesting ducks (Cowardin et al. 1985, Klett et al. 1988, Sargeant and Raveling 1992, Greenwood et al. 1995). Although a variety of methods for reducing predation have been tested, many of those that have proved to be the most cost-effective, especially the lethal methods, are controversial (Greenwood and Sovada 1996). This controversy stimulated our interest in alternative means of managing depredation.

In Arctic and sub-Arctic ecosystems, bird populations fluctuate in relation to naturally occurring prey (Larson 1960, McInvaille and Keith 1974, Pehrsson 1986, Summers 1986, Beintema and Müskens 1987). Byers (1974) suggested a similar positive relationship between nest success of ducks in grassland and abundance of small rodents in temperate regions. Crabtree and Wolfe (1988) reported a decrease in depredation of duck nests in Utah, where they supplemented the food supply of striped skunks (Mephitis mephitis). We attempted to determine if providing food to predators would increase nest success of ducks in the Prairie Pothole Region, where the predator community is markedly different than that in Utah (Sargeant and Arnold 1984).

We provided supplemental food for striped skunks during the nesting season in managed uplands of the Prairie Pothole Region to test whether food availability would influence nest success of ducks (Pietruszewski 1996). We targeted striped skunks because they are common in the Prairie Pothole Region, are important predators of eggs of upland-nesting ducks (Sargeant and Arnold 1984, Greenwood 1986, Johnson et al. 1989), and respond readily to food provisions (Greenwood et al. 1997).

Study Areas

We conducted the study on areas managed for wildlife production in the Missouri Coteau and Drift Plain physiographic regions of McLean and Stutsman counties, North Dakota. The Missouri Coteau is moderate to steeply rolling glacial moraine and outwash plain; the Drift Plain is moderately rolling to relatively flat (Stewart 1975). Pastures, hayfields, and cultivated land interspersed with numerous wetland basins comprise landscapes of both regions. During years of our study, much formerly cultivated land was enrolled in the Conservation Reserve Program (CRP; Young and Osborn 1990) and planted to perennial grasses and forbs; remaining cultivated land was used for cereal grain and row-crop production.

Climate of North Dakota is continental with warm summers and cold winters. Snow melt is generally complete by mid- to late April. Drought had negatively impacted wetlands and uplands in much of North Dakota for several years before our study (Sovada et al. 1995), but conditions improved during summer 1993, and by 1994, ponds were abundant and uplands were lush throughout the state (Greenwood and Sovada 1996). Total annual precipitation in the vicinity of the study locations was 57 cm in 1993 and 59 cm in 1994; the long-term mean for these areas is about 46 cm (Natl. Oceanic and Atmospheric Administration 1993, 1994).

We annually established a pool of candidate areas, each ≥65 ha, in localities with sufficient ponds nearby to support breeding populations of ducks. Each area contained ≥30 ha of upland nesting cover established >5 years earlier. We rejected areas that were subject to grazing or burning during the study year, or where organized predator removal to benefit wildlife production was conducted within 8 km. Candidate areas selected for study could not be within 6.5 km (a distance exceeding the home-range size of striped skunks in prairie regions [Greenwood et al. 1997]) of each other, thus ensuring independence among treatments. We waived the last criterion for areas separated by water barriers, because skunks are reluctant to swim (R. J. Greenwood, pers. observation).

Methods

Because canids have strong influence on nest success of ducks (Sovada et al. 1995), we did a preliminary track survey of all candidate areas in April each year to determine occupation by red foxes (Vulpes vulpes) or coyotes (Canis latrans). Based on results of this survey and information from wildlife managers and residents, we randomly assigned treatments (food provisioned or control) each year within the pool of candidate areas until we had equal numbers of each treatment in areas occupied by foxes (hereafter called fox area) or coyotes (hereafter called coyote area). Study areas totaled 10 in 1993 and 14 in 1994. We conducted 2 additional systematic track surveys (1 in late April-early May; 1 in late May-early June) each year to confirm preliminary canid assignments following methods of Sovada et al. (1995). Final determinations also utilized tracks found on study areas at times other than during systematic surveys and observations of canids within 0.8 km of a study area.

Food distribution

During 16 April-15 July 1993 and 20 April-13 July 1994, we distributed 90-100 kg of a mixture of chopped fish offal and sunflower seeds (10 parts offal: 1 part seeds by weight) in food plots between 0700-1200 hours every 3-4 days. This mixture was readily consumed by striped skunks in previous tests (Greenwood et al. 1997). Food plots were >100 m from improved roads with public access, >400 m from occupied residences, and accessible by vehicle. In 1993 each provisioned area contained 2 50- x 200-m food plots along opposite sides of the area. In 1994 each provisioned area contained a single 50- x 300-m food plot along 1 side. We discontinued the second food plot in 1994 because of concern about predators being drawn into nesting habitat when they traveled between plots on an area.

We tossed the food mixture by shovel from the back of a truck into dense grass and brush to conceal it from ring-billed gulls (Larus delawarensis) and California gulls (L. californicus) that were attracted to the sites. Because we feared gulls might not leave enough food for skunks, in 1993 we also suspended monofilament line 1 m above a 30- x 50-m area of each food plot (Ostergaard 1981, Blokpoel 1984) and placed a great horned owl (Bubo virginianus) decoy nearby to discourage gulls from feeding. In 1994 we attempted to haze gulls from plots during daylight hours with cracker shells. We discontinued use of monofilament line because we found it to be unnecessary.

The chopped offal (primarily walleyed pike [Stizostedion vitreum], sauger [S. canadense], and several species of salmonids [R and R Feeds, Ottertail, Minn.]) was delivered frozen in 22.7-kg blocks to freezer storage facilities near study locations. Offal costs were $0.034/kg; storage costs were $125 per month. We purchased oil sunflower seeds locally in 45.4-kg bags for $0.06/kg. Offal was thawed in a large tank and mixed with sunflower seeds the day before distribution.

We constructed 8-10 1-m² dust plots within food plots by removing sod and replacing it with sifted soil to monitor use of plots by striped skunks and other mammals. We inspected dust plots for presence of tracks before each food distribution. Tracks were identified to species, which was recorded, after which we raked the soil smooth in preparation for the next observation. At the same time, we also noted availability of food mixture remaining in the food plot.

Assessment of nest success

We conducted 3 searches at 3-week intervals each year beginning in early May, using a vehicle-towed chain drag (Higgins et al. 1969) to locate duck nests in uplands of each study area. A nest was defined as a bowl containing ≥1 egg tended by a female when found (Klett et al. 1986). We recorded the following data for each nest: date found or visited, duck species, location, number of eggs, and incubation stage (Weller 1956, Klett et al. 1986). Nests were marked individually using numbered willow (Salix sp.) sticks with a small piece of pink plastic flagging attached, placed 4 m away; locations were plotted on aerial photographs.

We visited nests at intervals of 6-10 days in 1993 and 21 days in 1994 until ≥1 egg hatched or eggs were destroyed or abandoned; we reduced number of visitations in 1994 to devote more resources towards increasing number of study area replicates. At each visit, duck species and recorded number of eggs were verified. We recorded fate at the final visit. A nest was successful if ≥1 egg hatched, as determined by presence of shell membranes (Klett et al. 1986) or ducklings. Suspected cause of failure (e.g., predator, abandonment) was determined for unsuccessful nests. Evidence (e.g., broken eggshells, digging) at depredated nests was recorded by methods of Sargeant et al. (1998) to aid in determination of predator species responsible for nest failure.

We estimated daily survival rates (DSRs) of nests by Mayfield's (1961) method, as modified by Johnson (1979). All species were pooled for each study area to increase sample sizes. We used weighted least-squares to improve precision of our DSR estimates (Johnson 1979, 1990), with weights equal to number of exposure days. Nests thought to have been abandoned due to investigator influences (i.e., abandoned on day of discovery) were excluded. For ease of interpretation, after analyses were performed, we converted DSR to nest success (P), where P = (DSR)^I and I is the mean laying interval plus incubation period in days.

We examined effects of treatment (food provision or control), dominant canid (fox or coyote), year (1993 or 1994), and their interactions on DSR using a 3-way ANOVA (GLM Procedure, SAS Inst. Inc. 1989a). Statistical significance was defined at P ≤ 0.05 for all analyses. We computed 95% confidence intervals (CI) on nest success (Johnson 1979).

Assessment of nest depredation

Differences in proportions of depredated nests attributed to striped skunks, red foxes, and American badgers (Taxidea taxus) between provisioned and control areas, between red fox and coyote areas, between years, and for all interactions were tested using a weighted least-squares approach for categorical data (CATMOD Procedure, SAS Inst. Inc. 1989b) and chi-square analyses. Determinations were based on evidence found at nests containing ≥6 eggs (hereafter called large-clutch nest) observed ≤21 days before being depredated. We excluded Franklin's ground squirrels (Spermophilus franklinii), because our nest visitation interval was insufficient in 1994 (Sargeant et al. 1998), and raccoons (Procyon lotor) because they forage mostly in wetlands and farmsteads (Greenwood 1982). Available criteria were inadequate for coyotes (Sargeant et al. 1998).

Assessment of predator community

We assessed the composition of the predator community thought to prey on duck eggs, which included, in addition to red fox and coyote, raccoon, American badger, Franklin's ground squirrel, long-tailed weasel (Mustela frenata), American crow (Corvus brachyrhynchos), and black-billed magpie (Pica pica; Sargeant and Arnold 1984). We utilized observational surveys during May-July (Sargeant et al. 1993), track surveys coincident with canid track surveys, and live traps baited with food mixture in late July (Greenwood 1986) to determine predators present within the study area. In 1993, 8 live traps were set for 4 consecutive 24-hour periods and checked once daily; in 1994, 10 traps were set and checked twice daily. Striped skunks and Franklin's ground squirrels were eartagged to permit a tally of total captures; striped skunks were aged (adult or juvenile <6-mo-old).

We used 2 × 2 frequency tables and likelihood ratio chi-square tests to estimate whether striped skunks, raccoons, American badgers, or Franklin's ground squirrels were equally likely to occur on provisioned versus control study areas (Agresti 1990), adding a constant of 0.5 to each cell frequency to accommodate values of zero. This analysis excluded long-tailed weasels, American crows, and black-billed magpies, which we rarely detected.

Assessment of vegetation density

We estimated density of residual vegetation in all study areas during late April or early May each year, using a visual obstruction pole, similar to methods of Robel et al. (1970) and Higgins and Barker (1982). Differences in mean height-density values between provisioned and control areas or years were tested using ANOVA.

Results

We found 1,046 duck nests; 1,008 were used in analyses (Table 1). In 1993, 267 nests were of gadwalls (Anas strepera; 33%), blue-winged teals (A. discors; 22%), mallards (A. platyrhynchos; 18%), northern shovelers (A. clypeata; 13%), northern pintails (A. acuta; 10%), and other species (5%). In 1994, 741 nests were of blue-winged teals (40%), mallards (20%), gadwalls (19%), northern shovelers (11%), northern pintails (7%), and other species (3%). More than twice as many nests were found per hectare searched in 1994 (0.98/ha) as in 1993 (0.47/ha). We distributed about 13,000 kg of food mixture in 1993 (2,600 kg/area) and 16,000 kg in 1994 (2,300 kg/area). Direct cost of provisioning 7 areas on 25 occasions in 1994 averaged about $796 per area (food mixture and storage = $3,390; labor @ $10/hr = $1,500; 25 days vehicle rent = $138; 3,400 km mileage @ $0.16/km = $544). Cost per area was similar in 1993.

In 1993 a preliminary survey indicated we had 6 fox areas and 4 coyote areas; based on a later systematic survey, we reclassified 1 coyote area as a fox area and 1 fox area as a coyote area. In 1994 a preliminary survey indicated we had 12 fox areas and 2 coyote areas; these did not change. In the final classification for both years combined, 10 provisioned areas and 8 control areas were fox areas, and 2 provisioned areas and 4 control areas were coyote areas (Table 1).

Nest success

We detected no difference (F = 1.32; df = 1,16; P = 0.27) in daily survival rates among years due to treatment. Mean daily survival rate was 0.973 (SE = 0.005) on provisioned areas and 0.964 (SE = 0.007) on control areas. Converted to nest success rate, the mean was 41% (95% CI = 29-57) on provisioned areas and 29% (CI = 18-47) on control areas. We detected a significant effect due to canid community (F =12.13; df = 1,16; P = 0.003). Mean daily survival rate was lower (P = 0.003) on fox areas ( = 0.954, SE = 0.006) than on coyote areas ( = 0.984, SE = 0.006). Mean nest success rate on fox areas was 20% (CI = 13-32) and on coyote areas was 57% (CI = 40-82). Effects of year and all interactions were not significant (P ≥ 0.20).

Predator community

We detected no overall differences in presence of striped skunks (χ² = 0.49, df = 1, P = 0.48), American badgers (χ² = 1.56, df = 1, P = 0.21), raccoons (χ² = 0.49, df = 1, P = 0.48), or Franklin's ground squirrels (χ² = 0.01, df = 1, P = 0.99) between provisioned and control areas. Striped skunks were present on 23 of 24 areas, but ≤2 adults were captured on any area. Franklin's ground squirrels were detected on only 1 of 10 areas in 1993 versus 12 of 14 areas in 1994.

Nest depredation

We used 325 large-clutch nests to estimate proportions of depredations due to striped skunks, red foxes, and American badgers. Fewer nest depredations were attributed to fewer striped skunks on provisioned areas (11%) than on control areas (24%; χ² = 3.83, df = 1, P = 0.05). Likewise, there was a tendency for fewer depredations by striped skunks on coyote areas (11%) than on fox areas (24%; χ² = 3.43, df = 1, P = 0.06). Fewer nest depredations were attributed to fewer red foxes on coyote areas (6%) than on fox areas (32%; χ² = 28.93, df = 1, P = 0.001), and we attributed fewer depredations to American badgers in 1993 (19%) than in 1994 (41%; χ² = 10.09, df = 1, P = 0.002). We detected no other significant relationships or interactions (P = 0.27).

Table 1. Dominant canid, size of area searched for nests, number of nests found, percent nest success^a (Mayfield 1961), and CI^b (in parentheses) for treatment (supplemental food provided) and control (no food provided) areas in a study of nest depredation by striped skunks of upland ducks in North Dakota during 1993 and 1994.

	Supplemental food provided						Supplemental food not provided
	Area	Canidc	Size (ha)	Nests	% success		Area	Canid	Size (ha)	Nests	% success
1993	1	F	68	27	5(2-17)		13	C	45	20	83 (65-100)
	2	C	83	72	84 (74-96)		14	F	49	13	10 (2-41)
	3	F	47	14	48 (25-92)		15	C	47	39	55 (38-79)
	4	F	59	28	13 (5-31)		16	F	35	17	13 (4-42)
	5	F	62	28	67 (48-93)		17	C	75	9	28 (8-98)
Annual , 1993					46 (22-77)						27 (23-56)
1994	6	F	48	31	23 (12-46)		18	F	68	26	34 (18-63)
	7	F	33	62	42 (30-61)		19	C	67	81	45 (33-60)
	8	F	57	57	36 (23-54)		20	F	60	29	43 (26-72)
	9	F	64	80	16 (10-27)		21	F	47	28	16 (7-36)
	10	C	52	79	47 (35-63)		22	F	47	41	8 (3-19)
	11	F	35	25	15 (6-37)		23	F	55	36	34 (20-56)
	12	F	70	106	29 (21-41)		24	F	58	60	18 (10-31)
Annual , 1994					36 (23-56)						31 (20-49)
Overall , 1993 and 1994					41 (29-57)						29 (18-47)
aLeast-squares means estimate weighted by exposure days. b95% confidence interval computed by method of Johnson (1979). cF = red fox; C = coyote.

Predator activity in food plots

We found striped skunk tracks in dust plots on an average of 35% of our visits to food plots in 1993 and 38% in 1994. Red fox tracks were found in dust plots on 9% of visits in 1993 and 35% in 1994. Both species visited food plots soon after we began distributing food. In 1994, 9 days after food mixture was first distributed, evidence in snow revealed striped skunks foraged at >70% of food plots and red foxes at >55% of plots. Striped skunks also excavated dens at 3 food plots in 1993 and 2 food plots in 1994. Raccoons and American badgers were uncommon visitors to food plots; combined, we detected their tracks on only 4% of visits in 1993 and 3% in 1994. Franklin's ground squirrels were detected, based on observations and vocalizations, at food plots only in 1994.

We observed gulls at food plots on 45% of our visits in 1993 and 42% in 1994. Although we were only marginally successful in keeping gulls away from food plots, ample food mixture appeared to be available at all times during both years. Although not tested, we found no evidence either year that measures used to discourage gulls from visiting food plots affected striped skunks or other nocturnal mammalian predators.

Vegetation density

We did not detect a difference (F = 1.57; df = 1,20; P = 0.22) in height-density measurements of residual vegetation between provisioned areas ( = 1.15 dm, SE = 0.04) and control areas ( = 1.23 dm, SE = 0.04). However, mean height-density measurements in 1993 were greater ( = 1.31 dm, SE = 0.05) than in 1994 ( = 1.07 dm, SE = 0.04; F = 16.36; df = 1,20; P < 0.001). We detected no year-by-treatment interaction (F = 0.0; df = 1,20; P = 0.99).

Discussion

Striped skunks regularly visited food plots throughout the study; red foxes were common visitors also, as were Franklin's ground squirrels in 1994. We did not detect an overall increase in nest success on provisioned areas, but observed an annual tendency towards higher nest success on provisioned areas than on controls. The difference was 19 percentage points in 1993 and 5 percentage points in 1994. The proportion of depredated nests attributed to striped skunks on provisioned areas was reduced by >50% of that on control areas. We interpreted this as evidence that skunks depredated fewer nests when supplementary food was provided. Inspection of locations of depredated and hatched nests did not reveal a pattern of depredations concentrated near food plots. Although skunks probably discover nests incidently while searching for other prey (Vickery et al. 1992), plot locations did not appear to affect fates of nearby nests in our study.

We believe lack of a detectable effect on nest success overall may have been due to excellent habitat conditions in 1994, availability of alternative foods, and compensatory predation. Recovery from prolonged drought in North Dakota began in 1993, and by 1994, record high numbers of breeding ducks were in the state. Conditions for breeding ducks in 1994 were the best ever observed in some parts of the Prairie Pothole Region (Krapu 1994, U.S. Fish and Wildl. Serv. and Can. Wildl. Serv. 1995, Greenwood and Sovada 1996). We believe that in 1994, in response to abundant and persistent shallow wetlands that encourage persistent nesting and numerous renesting attempts (Krapu et al. 1983, Cowardin et al. 1985), the sheer numbers of duck nests available may have had a swamping effect on predators, thereby affecting nest success. In 1994 success rates of upland-nesting ducks were high throughout much of North Dakota (Reynolds et al. 1994), higher than reported by Klett et al. (1988) for the same region in earlier years.

During both years our study was conducted in regions of the state where much cropland had been restored to grassland by the CRP. Conservation Reserve Program fields, besides supporting abundant populations of nesting ducks (Kantrud 1993, Reynolds et al. 1994), also support numerous species of alternative prey such as passerine birds (Johnson and Schwartz 1993), rodents, and invertebrates. Microtines, for instance, were especially abundant in CRP fields near our study sites in 1994 (R.J. Greenwood, pers. observation). Availability of alternative prey can influence depredation rates on nesting birds (Larson 1960, Byers 1974, McInvaille and Keith 1974, Pehrsson 1986, Summers 1986, Beintema and Müskens 1987). When prey increase to a level beyond which they can be utilized by predators, the predator-prey relationship becomes nonregulatory (Newton 1993). We believe the total biomass of prey reached this level in 1994.

We suspect that increased depredation by American badgers and Franklin's ground squirrels also may have compensated for reduced depredation by striped skunks in 1994. The proportion of depredations attributed to badgers was 19% in 1993, but increased to 41% in 1994. Franklin's ground squirrels, which are important predators of duck eggs (Sowls 1948, Sargeant et al.1987), were virtually absent from study areas in 1993, but were relatively abundant in 1994. This species previously has been implicated in compensatory depredation on duck nests in other studies (Greenwood 1986).

Presence of coyotes on some of our study areas may have impacted populations of striped skunks and further enhanced duck nest success. The proportion of nest failures attributed to striped skunk depredation tended to be less on coyote areas (11%) than on fox areas (24%). Johnson et al. (1989) suggested that coyotes may negatively affect striped skunks, and Sovada et al. (1995) also reported lower striped skunk activity on coyote areas than on red fox areas. Baker (1978) suggested that exclusion of coyotes resulted in an increase in striped skunks.

Although large gulls were common both years and were attracted to the food mixture, we believe it is unlikely that they influenced success of duck nests. We did not observe gulls on the ground other than in food plots, and their response to duck eggs was weak (Sargeant et al. 1998).

We did not evaluate long-term benefits of food provisioning on predator populations, but expect it might vary by species. Prairie striped skunks, for instance, are highly fecund (Greenwood and Sargeant 1994), and it seems unlikely that supplemental food in the spring would lead to enhanced fecundity. It also seems unlikely that a constant food supply would enhance summer survival of adult skunks, because typically their mortality is influenced by disease, predation, or interactions with humans (Sargeant et al. 1982, Fuller and Kuehn 1985, Greenwood and Sargeant 1994, Greenwood et al. 1997). Juvenile skunks, however, might benefit temporarily from food provisioning in early July when they become independent of adult females. Likewise, Franklin's ground squirrels might benefit from food provisioning. They accumulate their largest fat reserves just before entering hibernation, as early as mid-summer (Choromanski-Norris et al. 1986). An enhanced diet preceding hibernation might increase over-winter survival of ground squirrels and also might enhance fecundity of this species the following spring.

Management Implications

Our study did not clearly explain effects of providing food to reduce depredation of eggs by striped skunks. Although results from 1993 suggested a positive effect, we detected little benefit in 1994. Conditions for nesting ducks were excellent in 1994; however, expectations of further enhancement that year from an expanded food base for skunks may have been unrealistic. Beneficial effects of supplemental feeding to manage nest depredation might be greater under conditions less favorable to production of alternative forms of prey, in areas with less diverse communities of mammalian predators, and when used in combination with other forms of management.

Costs of providing food were about $800 per area, relatively low compared with some other management practices to enhance waterfowl production (Lokemoen 1984). Our results suggest, however, that benefits of providing supplemental food to reduce depredation of eggs by striped skunks will be quite variable, difficult to predict, and probably limited in extent to the immediate area of treatment. We doubt that additional investment of resources (more food or more frequent application) would have improved nest protection in our study. Although skunks and some other mammalian predators likely will respond to food provisioning, this method appears to have limited value as a treatment for managing depredation of upland duck nests in the Prairie Pothole Region.