Effects of migratory geese on plant communities of an Alaskan salt marsh

Authors


Amy Zacheis (fax +907 474 6967; e-mail:ftabz@uaf.edu).

Summary

  • 1We studied the effects of lesser snow geese (Anser caerulescens caerulescens) and Canada geese (Branta canadensis) on two salt marsh plant communities in Cook Inlet, Alaska, a stopover area used during spring migration. From 1995 to 1997 we compared plant species composition and biomass on plots where geese were excluded from feeding with paired plots where foraging could occur.
  • 2Foraging intensity was low (650–1930 goose-days km−2) compared to other goose-grazing systems.
  • 3Canada geese fed mainly on above-ground shoots of Triglochin maritimum, Puccinellia spp. and Carex ramenskii, whereas the majority of the snow goose diet consisted of below-ground tissues of Plantago maritima and Triglochin maritimum.
  • 4Plant communities responded differently to goose herbivory. In the sedge meadow community, where feeding was primarily on above-ground shoots, there was no effect of grazing on the dominant species Carex ramenskii and Triglochin maritimum. In the herb meadow community, where snow geese fed on Plantago maritima roots and other below-ground tissues, there was a difference in the relative abundance of plant species between treatments. Biomass of Plantago maritima and Potentilla egedii was lower on grazed plots compared with exclosed, whereas biomass of Carex ramenskii was greater on grazed plots. There was no effect of herbivory on total standing crop biomass in either community. The variable effect of herbivory on Carex ramenskii between communities suggests that plant neighbours and competitive interactions are important factors in a species’ response to herbivory. In addition, the type of herbivory (above- or below-ground) was important in determining plant community response to herbivory.
  • 5Litter accumulation was reduced in grazed areas compared with exclosed in both communities. Trampling of the previous year’s litter into the soil surface by geese incorporated more litter into soils in grazed areas.
  • 6This study illustrates that even light herbivore pressure can alter plant communities and affect forage availability.

Introduction

Goose herbivory can significantly impact production, composition and species abundance of plant communities, and this may in turn affect forage availability. The effects of geese on plant communities are variable, and depend on factors such as foraging intensity, plant community composition, whether geese feed on above- or below-ground plant tissues and the ability of plant species to re-grow following damage. For example, goose grazing on above-ground plant material may increase (Cargill & Jefferies 1984), decrease (Gauthier et al. 1995), or have no effect (Person et al. 1998) on the primary productivity of a plant community, whereas grubbing for below-ground tissues usually decreases productivity or the spatial extent of vegetation (Smith & Odum 1981; Kerbes et al. 1990; Bélanger & Bédard 1994; Ganter et al. 1996; Miller et al. 1996; Esselink et al. 1997; Jano et al. 1998). Within an ecosystem, individual plant species may show different re-growth patterns and productivity under similar grazing intensities (Kotanen & Jefferies 1987; Bazely & Jefferies 1989; Hik & Jefferies 1990; Zellmer et al. 1993; Gauthier et al. 1995). In addition, the timing, intensity and frequency of defoliation, as well as the fertilizing effect of faeces, alters the re-growth potential of some species (Prins et al. 1980; Hik & Jefferies 1990; Hik et al. 1991), while having no effect on others (Beaulieu et al. 1996). Finally, the response of a plant species to herbivory may vary depending on the plant community in which it is growing, and who its neighbours are (Crawley 1990).

Increasing populations of some goose species, including the North American mid-continent population of lesser snow geese (Anser caerulescens caerulescens L.), have raised interest in the effects of geese on their environment (Ankney 1996; Abraham & Jefferies 1997). By grubbing for below-ground plant tissues, snow geese can degrade habitat in wintering and nesting areas, and along migratory routes (Jefferies 1988; Kerbes et al. 1990; Ganter et al. 1996; Miller et al. 1996; Srivastava & Jefferies 1996; Jano et al. 1998). The degradation of Arctic and sub-arctic wetlands are of particular concern as are the effects of this loss of habitat on snow goose and other waterbird populations (Cooch & Cooke 1991; Williams et al. 1993; Ankney 1996; Ben-Ari 1998). Studies of the impacts of geese on their environment have focused on areas where herbivory is intense, and where the effects of geese on plant communities are visible, and sometimes destructive (e.g. Giroux & Bédard 1987; Miller et al. 1996; Srivastava & Jefferies 1996; Esselink et al. 1997; Person et al. 1998). Less research has been conducted in systems with less intense herbivore pressure, although such cases can provide information on threshold levels of herbivory that plant communities can sustain before the habitat is degraded (Archer & Smeins 1991).

We examined the effects of snow and Canada goose (Branta canadensis L.) activity on two plant communities in a salt marsh in south-central Alaska used during spring migration. Geese used the area only during a brief period each spring, and did not return to feed until the following spring. Thus, our study provides information on the response of a sub-arctic salt marsh to relatively light intensities of goose foraging. Also, because Canada geese and snow geese simultaneously used the same habitats, our study provides information on whether forage removal affected subsequent forage abundance for one or both species. Our objectives were (i) to determine which plants were used as foods by Canada and snow geese and thus whether diet differed between species; (ii) to estimate feeding intensity and forage removal within the marsh; and (iii) to determine if plant response to goose herbivory varied between two major plant communities.

Study area

This study was conducted in a 20.82-km2 portion of Susitna Flats, a 100-km2 salt marsh located in Upper Cook Inlet, Alaska (61°15′ N, 150°30′ W) (Fig. 1). Elevation of this coastal wetland gradually increases by 1 m from the shore to the beginning of woody vegetation, a distance of 1.5–3 km (A. Zacheis, unpublished data). Plant communities show a pronounced zonation parallel to the shore (Vince & Snow 1984). Geese use Susitna Flats in April and early May and at this time there is little above-ground vegetation, except for overwintering shoots of Carex spp. and small amounts of new shoot growth.

Figure 1.

Location of plant communities and study sites at Susitna Flats, Alaska. Transect locations are denoted with an x.

Up to 34 000 lesser snow geese use Cook Inlet salt marshes during spring migration to nesting areas on Wrangel Island, Russia (Butler & Gill 1987). The Wrangel Island snow goose population is less than 50% of its historical level and unlike other snow goose populations has not increased in recent years (Kerbes et al. 1999). Snow geese arrived in our study area in mid-April. Their numbers peaked in late April, and they departed during the first week of May, so that residence time was 10–20 days. We estimated goose use of the study area from aerial and ground surveys done in 1993–1997 (J. Hupp, W. Eldridge, unpublished data). Estimates of goose-days km−2 were made by summing the number of geese observed each day, and correcting this by a sampling fraction (number of days of observation/total number of days geese were in Susitna Flats) (Giroux & Bédard 1988). This was divided by the area of the study site to give estimates of 450–1300 goose-days km−2 for 1993–1996. Counts done in 1997, which were limited to a brief period in April when snow geese were present, indicated that snow goose numbers were much lower than in previous years (30 goose-days km−2) (J. Hupp, unpublished data).

Up to 100 000 Canada geese also use Upper Cook Inlet as a spring migration habitat (Butler & Gill 1987). Most are Taverner’s (B. c. taverneri Delacour) and cackling Canada geese (B. c. minima Ridgway) that nest in western Alaska (King & Derksen 1986). Reliable population estimates for Taverner’s Canada geese are not available; however, the population of cackling Canada geese has increased approximately 12% annually since 1988 (Wilkens & Cooch 1999). During 1995–1997 Canada goose use of the study area was between 480 and 830 goose-days km−2 (J. Hupp, W. Eldridge, unpublished data). Canada geese tend to arrive earlier and stay later than snow geese so that residence times were between 4 and 7 days longer.

Virtually all goose feeding in our study area occurred in early spring. Snow geese do not nest in Cook Inlet and rarely use the area during the autumn migration. Although small numbers of Canada geese nest in Cook Inlet, we observed no nesting pairs or broods on our study area. Some Canada geese do use Cook Inlet wetlands as staging areas in autumn but numbers are smaller than in spring and most flocks remain along the outer coastal fringe of wetlands. We observed no evidence of autumn feeding on our plots.

We conducted our research in two plant communities (Fig. 1) used by geese (properties described in Table 1). The sedge meadow community was composed mainly of Carex ramenskii Kom. and Triglochin maritimum L. (nomenclature follows Hultén 1968). The herb meadow community was dominated by Triglochin maritimum, Potentilla egedii Wormsk., Plantago maritima L. and Carex ramenskii. Soil moisture, soil salinity and frequency of tidal flooding were all higher in the sedge meadow community.

Table 1.  Plant species composition, soil characteristics and flooding regime of the sedge meadow and herb meadow communities in Susitna Flats, Alaska. Plant species composition is based on biomass measured on grazed plots in 1994, before exclosure treatments were applied (n = 16 for the sedge meadow and n = 72 for the herb meadow). Soil properties were measured in 1997 on grazed plots (n = 16 for the sedge meadow and n = 24 for the herb meadow). Soil moisture and salinity are the means of three and two measurements, respectively, taken throughout the growing season. Other soil properties were measured in May. Data are means ± 1 SE
 Sedge meadow Herb meadow 
Plant speciesCarex ramenskii42.4 ± 6.7%Triglochin maritimum50.9 ± 2.5%
 Triglochin maritimum56.9 ± 6.8%Plantago maritima13.0 ± 2.2%
 Other forbs and grasses 0.7 ± 0.7%Potentilla egedii 8.2 ± 1.1%
   Carex ramenskii20.5 ± 2.7%
   Other forbs, grasses and sedges 7.4 ± 1.2%
Soil moisture35.67 ± 0.41% 30.40 ± 0.47% 
Soil salinity14.97 ± 0.82 12.07 ± 0.45 
Soil organic matter 3.39 ± 0.15%  3.51 ± 0.17% 
Soil pH 7.25 ± 0.04  7.33 ± 0.02 
Plot distance from coast25–350 m 75 m – 1 km 
Flooding regimeMonthly high tides Storm tides and extreme high tides 

Methods

Goose diets

From 18 to 28 April in each of 3 years (1996–1998), we used a rifle or shotgun to collect snow geese and Canada geese from flocks that we observed feeding in coastal wetlands throughout Upper Cook Inlet. Geese were collected under permits authorized by the US Fish and Wildlife Service (permit number PRT-789758) and the Alaska Department of Fish and Game (permit numbers 96–51, 97–021 and 98–031). Habitats where geese were collected were similar to those where we studied the effects of goose foraging. Oesophageal contents were removed following collection, washed in fresh water in a fine-mesh sieve and frozen. We later thawed samples and separated material by species and plant part (above-ground vs. below-ground tissue). We identified forage items by comparison to reference material. Although freezing of samples softened oesophageal contents, individual forage items were clearly recognizable. Samples were dried to constant mass at 60 °C and weighed (±0.01 g). We calculated aggregate percentage of dry mass of each item in the diet as the average of the proportions of each food item within each bird (Swanson et al. 1974). Analysis was based only on individuals that had > 0.05 g (dry mass) of forage in their oesophagi. Geese collected at the same time and location usually fed on similar items. Therefore, we pooled data (within goose species) from birds that were collected at the same time and location before we calculated aggregate percentage of dry mass.

Effects of geese on plant communities

To examine the effects of herbivory on plant species composition and litter accumulation, we set up paired grazed/exclosed plots in the two plant communities. Eight pairs of plots were located at approximately 15-m intervals along 115 m long transects. Plots were 1 m2. Members of a pair were usually separated by 5 m and were in similar vegetation, based on visual cover estimates. One plot of each pair was randomly selected to receive an exclosure treatment. We established two transects in the sedge meadow community and nine transects in the herb meadow community (Fig. 1). Transects were separated by a minimum of 200 m.

We established plots in August of 1994. In early April of 1995, 1996 and 1997, before snow had melted, we erected 0.5 m tall fences around exclosed plots, and left the other plots open to goose foraging. Fences contained the 1 m2 plot and a border of at least 25 cm. Exclosures were constructed of 2 cm mesh herring seine stretched around reinforcement rods at the corners, except in 1995 when we encircled plots with three levels of twine wrapped around corner posts. Use of the herring seine allowed the sides of the exclosure to fall to the ground as the snow melted. Twine was crossed over the tops of exclosures in all years. Exclosures were taken down in late May, after geese had left the study area.

We clipped subsamples of above-ground vegetation from each plot each year in August from 1994 through 1997. The sample in 1994 was to determine if there were pre-treatment differences between paired plots. In 1994 and 1995, all above-ground biomass in one randomly selected 25 cm × 25 cm quadrat in each plot was clipped. In 1996 and 1997, two randomly selected, smaller quadrats (25 cm × 12.5 cm) were clipped in each plot. We clipped two quadrats in 1996 and 1997 to incorporate more of the spatial variation in plant cover within plots into our sample. Clipped quadrats and areas directly adjacent to them were not sampled in subsequent years. In all years, clippings were washed in fresh water, standing dead material was separated from live, and live biomass was sorted by species. Puccinellia species were combined, because Puccinellia phryganodes (Trin.) Scribn. & Merr. and Puccinellia nutkaensis (Presl) Fern. & Weath. may be ecophenes (Snow 1982). Total above-ground live biomass was the sum of the biomass of all species. In 1996 and 1997, litter lying on the soil surface was collected from quadrats where vegetation was clipped. All samples were frozen for transport to the University of Alaska, where they were dried at 60 °C to constant mass and weighed (±0.01 g).

Beginning in 1995, we counted the number of inflorescences or flowers for the dominant species (Triglochin, C. ramenskii, Potentilla and Plantago) from the sampled quadrats, except that C. ramenskii inflorescences were not counted until 1996. In 1996 and 1997, we counted plant densities of the dominant species within two 9 cm × 12.5 cm quadrats within each plot. For Triglochin, C. ramenskii and Plantago, we counted where plant bases emerged from the soil (basal meristem locations), and for Potentilla we counted the locations where petioles and roots grew from stolons. Estimates of total plant and inflorescence densities did not include Puccinellia or other species that were relatively rare on plots.

In August 1997 we sampled below-ground biomass by taking one soil core (5.5 cm diameter by 10 cm deep) in each of the 25 cm × 12.5 cm quadrats where above-ground biomass had been harvested that year. Collections were made from plots along three transects in the herb meadow community, and from all plots in the sedge meadow community. Soil cores were stored in polyethylene bags, frozen within 3 days, and transported to the University of Alaska, where roots and rhizomes were washed free of soil, dried at 60 °C to constant mass, and weighed (±0.01 g). No attempt was made to separate dead from live biomass or to separate biomass by species.

Foraging intensity

Offtake plots

To estimate the amount of vegetation removed by geese (offtake), we established a series of small (25 cm × 25 cm) paired grazed/exclosed offtake plots in April of 1996 and 1997. Four pairs of offtake plots were located near each transect of 1 m2 plots, generally within 0.5 m to 2 m of alternating grazed plots along the transect. Paired offtake plots were separated by 0.5 m to 1 m. Offtake plots were set up at different locations along transects in each of the 2 years. In 1996 offtake exclosures were cylinders of chicken wire anchored with reinforcement rods, whereas in 1997 we used herring seine to surround offtake plots.

Offtake plots were sampled as soon as possible in May, usually 1 to 3 weeks after geese had fed on them. We removed all soil and vegetation in the offtake plots to a depth of approximately 10 cm, and washed most of the soil from the vegetation in the field. Vegetation was re-washed in the field laboratory, and shoot material sorted to species. Plantago roots and C. ramenskii rhizomes were also sampled. Samples were immediately frozen, and later dried at 60 °C to constant mass before weighing (±0.001 g).

Faecal counts

We also attempted to estimate foraging intensity by counting goose faeces on the grazed 1 m2 transect plots. In 1995 and 1996, high tides and goose trampling of wet faeces made accurate counts impossible. However, 1997 was a dry spring with early snow-melt, and we were able to count faeces on plots reliably. We counted faeces along each transect once, as soon as we could access plots after birds left an area. We could easily differentiate between fresh faeces and those from previous years, as year-old faeces were considerably decayed and desiccated.

Statistical analysis

We analysed the pre-treatment (1994) biomass data as a doubly blocked anova, with pairs of plots (inner blocks) nested within transects (outer blocks). Plant communities and each major species within a community were analysed separately. This anova model is equivalent to a paired t-test (Sokal & Rohlf 1995), but is preferable to a t-test, as it incorporates an additional level of blocking (transects). We tested for treatment effect (exclosure, no exclosure) with the residual mean square as the error term (Newman et al. 1997). We used the same anova model to analyse offtake data, with the data separated by year, plant community and species.

Biomass data from 1995 through 1997 were analysed as repeated measures manovas, with the same doubly blocked design. We used an autocorrelation function analysis to confirm that plots pairs (inner blocks) were independent and could be considered replicates on which the repeated measures were made. Plant communities were again analysed separately; individual species biomass and total biomass were analysed in separate repeated measures. The sum of the two 25 cm × 12.5 cm quadrats sampled within each plot in 1996 and 1997 was used in the repeated measures, so that the data structure was comparable to that in 1995. The same repeated measures manova model was used to analyse litter biomass, standing dead material, species richness and inflorescence and plant densities. In cases where a year × treatment effect was significant, but an overall treatment effect was not, years were analysed separately using the univariate anova model.

All data were transformed (either loge+ 1 or square root + 3/8) to correct for non-normality and non-constancy of error variance (Neter et al. 1990; Zar 1996). However, most analyses were also run on ranked data, as transformations did not guarantee that the assumptions of anova were met (Conover 1980; Johnson & Wichern 1992). We report the results of the analyses using transformed data, as in most cases results on the ranked data were similar (Conover 1980). Exceptions are noted in the text. Non-transformed means ± 1 SE are reported throughout.

Results

Goose diets

We examined oesophageal contents of 45 snow geese and 28 Canada geese. Because we pooled data from individuals that were collected at the same time and location, analysis was based on 28 and 25 samples from snow geese and Canada geese, respectively.

Sixty-nine per cent of the snow goose diet was below-ground forage, whereas only 8% of the Canada goose diet consisted of below-ground plant parts (Fig. 2). Snow geese fed primarily on roots of Plantago that were 1–4 mm in diameter, and also on Triglochin root crowns. Snow geese also consumed above-ground shoots of Triglochin, Carex lyngbyaei Hornem. and C. ramenskii, including basal meristematic tissue. They rarely ate green tips of vegetation. In contrast, Canada geese grazed primarily on green tips of Puccinellia, Triglochin and C. ramenskii shoots. They did not usually consume the meristem at the base of shoots. Most birds (61% and 56% of snow geese and Canada geese, respectively) had consumed more than one forage species.

Figure 2.

Aggregate percentage of dry mass of forage items in oesophagi of snow and Canada geese (n = 28 and 25, respectively) collected in Upper Cook Inlet, Alaska, April 1996–1998.

Foraging intensity

Offtake plots

In the sedge meadow community, there were no significant differences between grazed and exclosed offtake plots in 1996 or 1997 for either C. ramenskii biomass (1996: F1,5 = 0.0002, P = 0.99; 1997: F1,7 = 0.29, P = 0.61), Triglochin biomass (1996: F1,5 = 0.05, P = 0.83; 1997: F1,7 = 1.47, P = 0.27) or total biomass (1996: F1,5 = 0.03, P = 0.87; 1997: F1,7 = 0.44, P = 0.53).

In the herb meadow community in 1996 there was 25% less Plantago shoot biomass in grazed offtake plots (0.98 ± 0.24 g m−2) than in exclosures (1.32 ± 0.27 g m−2; F1,33 = 3.98, P = 0.05). Plantago roots showed a similar (26%) reduction in biomass in grazed plots (28.24 ± 7.93 g m−2) compared to exclosures (38.18 ± 8.24 g m−2), although in this case it was not statistically significant (F1,33 = 2.75, P = 0.11). We included all Plantago roots in our offtake samples, although snow geese fed on only smaller diameter roots and not on thicker taproots. The large biomass of the taproots may have swamped any statistically significant removal of smaller roots. Potentilla showed the greatest response to use by geese, with 45% less biomass in grazed offtake plots (0.08 ± 0.02 g m−2) compared to exclosed (0.14 ± 0.03 g m−2; F1,33 = 9.10, P = 0.005). There were no differences in biomass of Triglochin (F1,33 = 0.60, P = 0.44), Puccinellia (F1,33 = 0.57, P = 0.45) or C. ramenskii (F1,33 = 0.05, P = 0.83) between exclosed and grazed offtake plots in 1996. The 18% reduction in total biomass in grazed offtake plots (49.96 ± 7.46 g m−2) compared to exclosed (60.70 ± 7.97 g m−2) was not statistically significant (F1,33 = 2.82, P = 0.10).

In the herb meadow community in 1997, a year of lower foraging intensity, there were no significant differences between exclosed and grazed offtake plots for any of the dominant species, or for total biomass (Triglochin: F1,35 = 0.01, P = 0.93; Plantago shoots: F1,35 = 0.72, P = 0.40; Plantago roots: F1,35 = 0.07, P = 0.79; Potentilla: F1,35 = 1.00, P = 0.32; Puccinellia: F1,35 = 1.26, P = 0.27; C. ramenskii: F1,34 = 0.11, P = 0.74; total biomass: F1,35 = 0.001, P = 0.98).

Faecal counts

Faecal counts in 1997 indicated slightly greater use by geese of the sedge meadow community (1.8 ± 0.4 faeces m−2) than the herb meadow community (1.3 ± 0.2 faeces m−2).

Effects of geese on plant communities

Sedge meadow community

There were no pre-treatment differences in the biomass of the dominant species within the sedge meadow community between designated grazed and exclosed plots in 1994 (C. ramenskii: F1,15 = 2.82, P = 0.11; Triglochin: F1,15 = 0.30, P = 0.59).

Exclosures had little effect in the sedge meadow community. There was no difference in the biomass of the dominant species C. ramenskii and Triglochin between exclosed and grazed plots, and no difference in total live biomass for any of the 3 years of treatment (Fig. 3a–c, Table 2). In addition, there was no effect of exclosures on the density of either dominant species, or on total plant density (Fig. 4a–c, Table 2). C. ramenskii had more inflorescences in grazed plots (Fig. 4d, Table 2). This resulted in more total inflorescences in grazed areas, despite a lack of exclosure effects on Triglochin (Fig. 4e,f; Table 2). There was no difference in root biomass between exclosed (212.41 ± 15.56 g m−2) and grazed (236.19 ± 12.48 g m−2) plots in 1997 after 3 years of fencing (F1,31 = 2.08, P = 0.16).

Figure 3.

Response of the sedge meadow community to goose herbivory, Susitna Flats, Alaska 1995–1997. Comparison of plots grazed by geese and those exclosed from geese for (a) Carex ramenskii biomass, (b) Triglochin maritimum biomass, (c) total above-ground live biomass, (d) standing dead material, and (e) litter. Litter was measured in 1996 and 1997 only. Note different y-axis scales. ▪= grazed; □= exclosed.

Table 2.  Response of the sedge and herb meadow communities to goose herbivory, Susitna Flats, Alaska 1995–1997. F-values from repeated measures manova analyses are reported. Treatment effect compares plots fed on by geese (grazed) with plots exclosed from goose use. * P ≤ 0.05, ** P ≤ 0.01, *** P ≤ 0.001
   F-values
Community/plant speciesSource of variationd.f.Above-ground biomassPlant densityInflorescence density
  • a

    d.f. 1, 15;

  • b

    b d.f. 1, 71.

Sedge Meadow community
Carex ramenskiiTreatment1, 15 0.04  0.28  4.94*
 Year2, 14 1.10  4.57*a 10.88**a
 Year × Treatment2, 14 0.10  0.67a  0.42a
Triglochin maritimumTreatment1, 15 0.14  1.45  2.71
 Year2, 14 8.91**  0.41a170.91***
 Year × Treatment2, 14 0.71  0.04a  0.12
Total liveTreatment1, 15 0.02  2.19  8.27*
 Year2, 1414.77***  6.79*a 23.88***a
 Year × Treatment2, 14 0.31  0.13a  0.80a
Standing deadTreatment1, 15 0.69
 Year2, 14 2.41
 Year × Treatment2, 14 1.43
LitterTreatment1, 15 8.61**
 Year1, 1510.67**
 Year × Treatment1, 15 0.14
Herb Meadow community
Triglochin maritimumTreatment1, 71  0.02  2.93  2.40
 Year2, 70 22.85***  5.69*b281.39***
 Year × Treatment2, 70  0.18  2.66b  1.04
Plantago maritimaTreatment1, 71  4.20*  5.01*  3.12
 Year2, 70  9.28***122.79***b 34.47***
 Year × Treatment2, 70  0.47  0.76b  4.55*
Potentilla egediiTreatment1, 71  9.94** 21.58*** 21.01***
 Year2, 70  9.83*** 85.37***b  6.63**
 Year × Treatment2, 70  6.15**  0.22b  4.75*
Puccinellia spp.Treatment1, 71  4.60*
 Year2, 70 25.02***
 Year × Treatment2, 70  0.14
Carex ramenskiiTreatment1, 71 11.95*** 36.34*** 14.10***
 Year2, 70 44.40***  4.57*b 17.33***b
 Year × Treatment2, 70  2.10  3.59b  8.77**b
Total liveTreatment1, 71  2.53  1.17  0.07
 Year2, 70 42.18*** 92.94***b 12.13***b
 Year × Treatment2, 70  0.41  1.82b  4.89*b
Standing deadTreatment1, 71  9.40**
 Year2, 70189.42***
 Year × Treatment2, 70  3.23*
LitterTreatment1, 71 38.95***
 Year1, 71 16.18***
 Year × Treatment1, 71  3.27
Figure 4.

Response of the sedge meadow community to goose herbivory, Susitna Flats, Alaska 1995–1997. Comparison of plots grazed by geese and those exclosed from geese for (a) Carex ramenskii plant density, (b) Triglochin maritimum plant density, (c) total plant density, (d) Carex ramenskii inflorescence density, (e) Triglochin maritimum inflorescence density, and (f) total inflorescence density. Plant density and Carex ramenskii inflorescence density were measured in 1996 and 1997 only. Note different y-axis scales. ▪= grazed; □= exclosed.

Litter accumulation was greater within exclosures, although there was no effect of exclosures on standing dead material (Fig. 3d,e, Table 2). There was no treatment effect on species richness per 25 cm × 25 cm quadrat for all years (F1,15 = 0.03, P = 0.86). Mean species richness for the 3 years of the study was 2.1 ± 0.1 in grazed plots and 2.2 ± 0.1 in exclosed plots.

Herb meadow community

In the herb meadow community there were no pre-treatment differences in the biomass of any dominant species between designated exclosed and grazed plots in 1994 (Triglochin: F1,71 = 1.10, P = 0.30; Plantago: F1,71 = 0.0002, P = 0.99; Potentilla: F1,71 = 3.21, P = 0.08; Puccinellia: F1,71 = 0.13, P = 0.72; C. ramenskii: F1,71 = 2.73, P = 0.10).

Plant species responded differently to fencing. There were no treatment effects on biomass, plant density or inflorescence density for Triglochin (Figs 5a and 6a,f, Table 2). In contrast, Plantago had less biomass and lower plant densities in grazed plots (Figs 5b and 6b, Table 2). Plantago inflorescence density had a significant year × treatment interaction, with lower densities in grazed plots not evident until the third year of fencing (univariate anovas: 1995: F1,71 = 2.29, P = 0.13; 1996: F1,143 = 1.18, P = 0.28; 1997: F1,143 = 10.56, P = 0.001) (Fig. 6g, Table 2). Puccinellia also had less biomass in grazed plots (Fig. 5d, Table 2).

Figure 5.

Response of the herb meadow community to goose herbivory, Susitna Flats, Alaska 1995–1997. Comparison of plots grazed by geese and those exclosed from geese for (a) Triglochin maritimum biomass, (b) Plantago maritima biomass, (c) Potentilla egedii biomass, (d) Puccinellia spp. biomass, (e) Carex ramenskii biomass, (f) total above-ground live biomass, (g) standing dead material, and (h) litter. Litter was measured in 1996 and 1997 only. Note different y-axis scales. ▪= grazed; □= exclosed.

Figure 6.

Response of the herb meadow community to goose herbivory, Susitna Flats, Alaska 1995–1997. Comparison of plots grazed by geese and those exclosed from geese for (a) Triglochin maritimum plant density, (b) Plantago maritima plant density, (c) Potentilla egedii plant density, (d) Carex ramenskii plant density, (e) total plant density, (f) Triglochin maritimum inflorescence density, (g) Plantago maritima inflorescence density, (h) Potentilla egedii flower density, (i) Carex ramenskii inflorescence density, and (j) total inflorescence density. Plant density and Carex ramenskii inflorescence density were measured in 1996 and 1997 only. Note different y-axis scales. ▪= grazed; □= exclosed.

Potentilla had greater biomass, plant density and flowering within exclosures (Figs 5c and 6c, h, Table 2), representing the greatest response to grazing among species. There was a year × treatment interaction for biomass and inflorescence density (Table 2), due to annual variability in the difference between exclosed and grazed plots (Figs 5c and 6h).

C. ramenskii was the only species with greater biomass, plant density and inflorescence density in grazed plots (Figs 5e and 6d,i,Table 2). There was a steadily increasing difference in biomass between treatments (16% more biomass in grazed plots in 1995, 44% more in 1996, and 59% more in 1997), resulting in a weak year × treatment interaction, significant on the ranked data only (F2,70 = 3.15, P = 0.05; compare with Table 2). Similarly, treatment differences in plant and inflorescence densities increased between 1996 and 1997 (Fig. 6d,i), resulting in a significant year × treatment interaction for inflorescences (Table 2). However, analysis on the ranked inflorescence data only showed a marginally significant year × treatment interaction (F1,71 = 3.18, P = 0.08).

For each of the 3 years of exclusion of geese, total live biomass in the herb meadow community in August did not differ between treatments (Fig. 5f, Table 2), although there was a substantial shift in the proportion of Plantago, Potentilla, Puccinellia and C. ramenskii within the plant community (Fig. 5b–e). Total plant and inflorescence density also did not differ between exclosed and grazed plots (Fig. 6e,j, Table 2). Total inflorescence density had a year × treatment interaction because densities were slightly higher in exclosures in 1996, but higher in grazed areas in 1997 (Fig. 6j). There was no difference in root biomass between exclosed plots (168.05 ± 13.38 g m−2) and grazed plots (173.40 ± 13.15 g m−2) upon completion of the study in 1997 (F1,47 = 0.04, P = 0.85).

We found more litter within exclosures compared to grazed plots in 1996 and 1997 (Fig. 5h, Table 2). There was also more standing dead material within exclosures, although treatment differences were not as pronounced as the litter effect (Fig. 5g, Table 2). There was a significant year × treatment interaction for standing dead material due to annual variability in treatment differences (Fig. 5g).

We found no difference in species richness per 25 cm × 25 cm quadrat between exclosed and grazed plots for all years (F1,71 = 1.17, P = 0.28). Mean species richness for the 3 years of the study was 4.6 ± 0.1 in grazed plots and 4.8 ± 0.1 in exclosed plots.

Discussion

Goose diets

With the exception of Potentilla, geese consumed all major wetland plant species in the sedge meadow and herb meadow communities. The presence of more than one forage item in most oesophagi indicated that geese probably consumed whichever plants were present at feeding sites. Potentilla was probably not a forage item because it exists as a tiny corm in early spring and was a relatively minor part of the total biomass available to geese.

Although Canada geese and snow geese often exploited the same habitats and frequently fed in mixed flocks, there was little dietary overlap. Plantago roots were 40% of the snow goose diet but only 2% of the Canada goose diet. C. ramenskii and Triglochin combined were 40% and 53% of snow goose and Canada goose diets, respectively. However, snow geese primarily exploited below-ground parts or non-photosynthetic basal portions of shoots, whereas Canada geese fed almost exclusively on above-ground green shoots of these species. Thus diets were largely partitioned by plant part. Prevett et al. (1985) also found that snow geese consumed more below-ground material than Canada geese did during spring staging at coastal habitats in James Bay, Canada.

Foraging intensity

Estimates of combined use by Canada and snow geese for our study area were 930, 1930, and 650 goose-days km−2 for 1995, 1996 and 1997, respectively, several orders of magnitude less than estimates in other goose staging areas. Use of the 1.47-km2 Montmagny sanctuary along the St Lawrence River, a spring and fall staging area for greater snow geese (Anser caerulescens atlantica), was 34 000 goose-days km−2 in the spring and 197 300 goose-days km−2 in the fall (Bélanger et al. 1990). The 7.6-km2 Dutch portion of the Ems Dollard estuary, bordering the North Sea, is used by greylag geese (Anser anser) as a spring and fall staging area, with some geese remaining throughout the winter (Esselink et al. 1997). Estimated goose use ranged from 32 900 to 80 300 goose-days km−2 for 1983–1994 (Esselink et al. 1997).

Lesser snow geese use wetlands along the coast of Hudson Bay for staging, nesting and brood-rearing, where they grub for roots and rhizomes in the spring, and graze above-ground vegetation in the summer (Jefferies 1988; Kerbes et al. 1990). At the McConnell River colony on Hudson Bay the 1985 estimate of breeding pair density was 132 299 pairs in a 339.7-km2 area (Kerbes et al. 1990). Assuming birds arrived by 1 June and departed by 15 August (exact dates were not published), this translates into 58 000 goose-days km−2 (Kerbes et al. 1990). This is an underestimate of herbivore pressure, however, as it does not include use by goslings, non-breeders and spring migrants. Foraging intensity is also very high at the La Pérouse Bay colony, with 1990 estimates of 22 500 nesting pairs of geese, their goslings, and additional geese stopping over during migration (Jefferies 1988; Cooke et al. 1995). Nest density can reach as high as 2000 nests km−2 (Kotanen & Jefferies 1997).

Not only is herbivore pressure relatively light in Susitna Flats, but the return time of geese to a specific portion of the marsh is a minimum of 1 year, as annual distributions of geese may vary. Snow goose flocks tend to feed along, and follow, the edge of the melting snow pack, so that the distribution of available habitat may vary among years depending on patterns and timing of snow-melt (Hupp et al. 2001). For example, in 1997 snow-melt was unusually early, and snow goose use of the largely snow-free herb meadow community was light. However, the sedge meadow community, which became snow-free later, tended to have more use.

There were no significant differences in the biomass of Triglochin, C. ramenskii or Puccinellia between grazed and exclosed offtake plots in 1996 or 1997, although all were forage items found in the oesophagi of geese. In addition, we frequently found evidence of herbivory on these plants (e.g. grazed shoot tips). These plant species may be quite tolerant to herbivory and may have replaced lost tissue between the incidence of herbivory and the time we sampled plots 1 to 3 weeks later. Alternatively, the small size and limited number of offtake plots, and the variability in goose distribution, may have made it unlikely that all plots were fed on, or fed on to a sufficient extent that significant differences between grazed and exclosed plots could be detected.

There was 25% less Plantago root and shoot biomass in grazed offtake plots as compared with exclosed in 1996, although the difference in root biomass was not statistically significant. In 1997 there was 2% less Plantago root biomass in grazed plots, again not a significant difference. We may have been unable to detect significant reductions in biomass of below-ground forage due to high spatial variability in plant biomass and in snow goose feeding sites. Similarly, in the St Lawrence staging area, Giroux & Bédard (1987) were unable to document differences in below-ground biomass before and after greater snow geese fed in area, due to high variability in biomass. However, the larger reduction in Plantago root biomass in 1996, when there was greater use of the herb meadow community by snow geese, compared to 1997, suggests that geese probably had an effect on the abundance of this forage species. The 45% reduction of Potentilla in grazed offtake plots in 1996 suggests that this species was also affected, although it is not a forage item.

Effects of geese on plant communities

Sedge meadow community

The impact of geese varied between the two plant communities. In the sedge meadow community, feeding was primarily on above-ground shoots, although snow geese may also have fed on C. ramenskii rhizomes and the upper portion of Triglochin rootstocks, and removed meristematic tissue in both species. Herbivory had no effect on total live biomass in August, or on the biomass of the dominant species C. ramenskii or Triglochin. The dominant species in this community have characteristics conferring tolerance to herbivory, such as below-ground carbohydrate reserves, multiple meristems and carbohydrate production in undamaged tillers or shoots (Youngner 1972; Archer & Tieszen 1986; Rosenthal & Kotanen 1994; Crawley 1997). C. ramenskii has been shown to tolerate clipping, so that productivity in repeatedly clipped plots is equal to or greater than that in unclipped plots (Ruess et al. 1997), and Triglochin has a large rootstock from which carbohydrate reserves may be mobilized. The stability of this plant community’s response to herbivory is probably due to minimal amounts of below-ground feeding and tolerance of the dominant species to above-ground herbivory. In addition, herbivore pressure is light, and herbivore return time is a minimum of one growing season. In contrast, other staging and wintering areas used by snow geese have longer residence times, shorter return times, and greater amounts of below-ground grubbing, resulting in reduction of productivity, reduction in biomass of forage species, and, at extreme goose densities, destruction of wetland vegetation (Smith & Odum 1981; Giroux & Bédard 1987; Kerbes et al. 1990; Bélanger & Bédard 1994; Ganter et al. 1996; Miller et al. 1996; Srivastava & Jefferies 1996).

The sample size of 16 paired plots within the sedge meadow community allowed detection of only relatively large differences between treatments. In 1997 treatment differences in C. ramenskii biomass had to be greater than 30% of the average biomass on plots to be statistically significant. Similarly, treatment differences in Triglochin biomass and total biomass had to be 36% and 21%, respectively, to be significant. However, mean differences between biomass on grazed and exclosed plots in 1997 were actually very small (6.5% more C. ramenskii biomass in exclosures, 3.7% less Triglochin, and 0.4% less total biomass). Therefore, we believe that our failure to detect a treatment effect is because herbivory did not substantially affect biomass within this community.

Herb meadow community

Herbivory significantly reduced the biomass of several plant species within the herb meadow community. Plantago was unable to compensate for root loss, and grubbing by snow geese resulted in lower biomass, plant density and inflorescence density in grazed plots. Dormann et al. (2000) similarly found Plantago biomass to be greatly reduced by brent geese (Branta bernicla bernicla) feeding on leaves and roots. Other saltmarsh perennials such as Scirpus spp. and Spartina spp. have also shown reductions in production, cover or spatial extent due to feeding by geese on below-ground tissues (Smith & Odum 1981; Giroux & Bédard 1987; Esselink et al. 1997). Potentilla also appears unable to tolerate use by geese. Although it is not a forage species, grubbing for Plantago roots by snow geese may incidentally damage Potentilla. Because of its small size in the spring, Potentilla has little storage capacity or above-ground growth that would enable it to recover from tissue loss. Potentilla corms are shallowly rooted near the soil surface where disturbance may cause high rates of mortality. This would explain the 45% reduction in Potentilla in grazed offtake plots as compared with exclosed plots, and the strong reductions in standing crop biomass, plant density and flowering at the end of the growing season in grazed areas.

Competitive interactions, along with different tolerances to herbivory among plant species, are probably responsible for the shifts in relative species abundance within the herb meadow community. Whereas the decreased abundance of Plantago and Potentilla in grazed areas is likely to be due to the inability of these species to re-grow following below-ground herbivory or disturbance, greater growth of C. ramenskii in grazed areas may be due to reduced competition from Plantago and Potentilla (Crawley 1983), as well as tolerance of above-ground herbivory (Ruess et al. 1997). Although we do not have data on competitive interactions within this plant community, negative correlations between species may indicate competitive relationships. There was a significant negative correlation between Plantago and C. ramenskii biomass for all years of this study (Table 3), which persisted after the effects of herbivory had been removed (exclosed plots only in Table 3). We also found a negative association between Potentilla and C. ramenskii, although this correlation was not significant in exclosed plots only (Table 3).

Table 3.  Pearson correlation coefficients between Carex ramenskii biomass and Plantago maritima, Potentilla egedii, and Puccinellia spp. biomass within the herb meadow community, Susitna Flats, Alaska, for 1994 (pre-treatment year) through 1997. Significance values are based on log transformed data. * P ≤ 0.05, ** P ≤ 0.01, *** P ≤ 0.001
   Plantago maritimaPotentilla egediiPuccinellia spp.
1994Carex ramenskiiAll plots−0.618***−0.347***−0.260**
  Exclosed plots only−0.609***−0.174−0.212
1995Carex ramenskiiAll plots−0.505***−0.227**−0.363***
  Exclosed plots only−0.588***−0.201−0.387***
1996Carex ramenskiiAll plots−0.465***−0.226**−0.297***
  Exclosed plots only−0.503***−0.195−0.332**
1997Carex ramenskiiAll plots−0.493***−0.215**−0.292***
  Exclosed plots only−0.504***−0.195−0.188

Canada geese grazed above-ground shoots of Puccinellia, which had lower biomass in grazed compared to exclosed plots. In contrast, Puccinellia phryganodes grazed by snow geese at La Pérouse Bay tolerated grazing, and under some conditions exhibited greater biomass and/or production in grazed plots (Cargill & Jefferies 1984; Hik & Jefferies 1990; Hik et al. 1991). In particular, P. phryganodes showed increased production only when goose faeces were present (Hik & Jefferies 1990). The reduced growth of Puccinellia in grazed plots in Susitna Flats may be due to herbivory combined with low faecal nitrogen return. Alternatively, the negative response of Puccinellia to grazing may be due, at least partially, to increased competition with C. ramenskii.Puccinellia and C. ramenskii biomass was negatively correlated, suggesting a competitive relationship (Table 3).

Shifts in community composition or relative species abundance under herbivory have often been hypothesized to be caused by alterations of the competitive hierarchy within the plant community (e.g. Inouye et al. 1980; Crawley 1990; Furbish & Albano 1994). Shifts in species composition were documented on staging grounds of the St Lawrence estuary (Giroux & Bédard 1987). Grubbing of Scirpus americanus reduced its abundance, whereas Zizania aquatica, whose seeds and stems are only eaten in the autumn, increased on grazed relative to exclosed plots (Giroux & Bédard 1987). Spartina patens and Scirpus robustus responded in opposite directions to grubbing by snow geese in a North Carolina wintering area (Smith 1983). At La Pérouse Bay, Carex subspathacea and dicotyledons increased in abundance within exclosures, at the expense of Puccinellia phryganodes (Bazely & Jefferies 1986).

Our results from two different communities suggest that the interaction of herbivory and competition can cause shifts in relative species abundance if grazing intolerant plants are present, but may have no effect on a community if they are not. The type of herbivory (i.e. above- or below-ground) is important in determining a plant species’ ability to tolerate herbivory or to re-grow following damage. In addition, the effect of herbivory on a particular plant species may depend on the community in which it is found (e.g. C. ramenskii increased in biomass and density in grazed plots in the herb meadow community but did not increase in the sedge meadow community). Therefore, the response of an individual species to herbivory is partly dependent on plant community composition.

Effects of geese on litter accumulation

Although we did not measure litter-fall directly, litter production was likely the same on grazed and exclosed plots in the sedge meadow community. This is because neither total standing crop biomass at the end of the growing season nor relative species abundance differed between treatments. In the herb meadow community, standing crop biomass did not differ between grazed and exclosed plots but relative species abundance did, making inferences about litter production more difficult. However, we suggest that litter production may have been slightly greater, and certainly not less, in grazed plots than in exclosed plots in this community. More C. ramenskii grows in grazed areas, and this species produces large amounts of litter compared to the other plant species.

Although litter production was the same or greater in grazed plots compared to exclosed, we found less litter on the soil surface in grazed plots. We attribute the disappearance of litter in grazed areas to trampling by geese. Trampling of previous years’ litter into the soil surface means that more litter is incorporated into the soil in grazed areas and less accumulates on the soil surface. Trampling may also reduce litter loss through tidal export. Litter accumulation within exclosures commonly occurs in vertebrate grazing and browsing systems (e.g. Fuller et al. 1985; Bazely & Jefferies 1986; McNaughton et al. 1988; Pastor et al. 1993; Biondini et al. 1998; Evers et al. 1998; Van Wijnen et al. 1999), although in some cases litter accumulation is greater in grazed areas (Ford & Grace 1998). With high herbivore pressure, grazing or browsing tends to reduce above-ground standing biomass (Fuller et al. 1985; Bazely & Jefferies 1986; Evers et al. 1998) and litter-fall (Pastor et al. 1993) such that litter inputs to the soil are reduced in grazed areas. In contrast, at Susitna Flats litter inputs to the soil are greater on grazed areas due to trampling, which results in less litter accumulation on the soil surface.

Effects of geese on forage availability

Feeding by Canada and snow geese had no effect on forage availability in the sedge meadow community, but reduced the availability of some forage species in the herb meadow community. Of particular importance to snow geese was a 29% reduction in Plantago in grazed areas compared to exclosed after 3 years of fencing, as this species was 40% of the snow goose diet. The 21% reduction in biomass of Puccinellia in grazed areas after 3 years probably impacted on Canada geese, as this forage composed 28% of their diet.

In contrast, Triglochin, comprising 20% and 30% of snow goose and Canada goose diets, respectively, was not affected by feeding. There was 59% more C. ramenskii, an important part of snow (20%) and Canada (16%) goose diets, in grazed areas compared to exclosed after 3 years. The species on which herbivory had the greatest effect, Potentilla, was not a forage plant for either goose species. However, the decrease in Potentilla on grazed plots may have reduced competitive pressure on other species (particularly C. ramenskii), and indirectly resulted in increased forage availability.

C. ramenskii and Triglochin may fill gaps in both snow goose and Canada goose diets left by reduced availability of Plantago and Puccinellia. However, C. ramenskii, Triglochin and Plantago differ in nitrogen, acid detergent fibre and non-structural carbohydrate content in early spring (Hupp et al. 2001). Therefore, although the quantity of forage available to geese may be unaltered by goose activity, we do not know how the overall quality of the diet is affected.

Negative effects of geese on forage availability do not appear to be cumulative. The differences in Plantago and Puccinellia biomass between grazed and exclosed plots did not increase over the 3 years of the study. This indicates that the marsh is not rapidly deteriorating as habitat for either goose species due to grubbing, probably because foraging pressure is light and spatially variable among years.

Conclusion

This study illustrates that even light or ephemeral herbivore pressure can alter plant communities and the availability of forage plants for the consumer. However, communities respond to herbivory differently, and the tolerance of plant species to herbivory, as well as the type of herbivory, may be important in determining community response. In this marsh, grubbing for below-ground biomass of Plantago by snow geese appears to cause shifts in relative species abundance in the herb meadow community. Finally, the response of an individual plant species to herbivory is dependent on the community in which it is growing, suggesting that herbivores alter competitive interactions between plant species.

Acknowledgements

We thank A. Banks, M. Boswell, T. Bowman, P. Busteed, L. Butler, K. Chapin, J. Farnam, C. Hinshaw, S. Reidsma and C. Talus for assistance with laboratory and fieldwork. We thank the Alaska Department of Fish and Game for their support of our work on a State Game Refuge and for the use of their truck. J. Clinton and B. Stewart and C. Stewart kindly lent the use of their cabins at Susitna Flats. W. Eldridge provided aerial survey data for Cook Inlet. K. Doran, B. Person, J. Sedinger, R. Jefferies and two anonymous reviewers provided valuable comments and discussion on earlier drafts of this manuscript, and R. Barry and D. Thomas gave statistical advice. This research was supported by the Alaska Biological Science Center, US Geological Survey, a University of Alaska Office of Global Change and Arctic Systems Research grant, and a UAF Thesis Completion Fellowship to A. Zacheis.

Received 11 January 2000 revision accepted 20 June 2000

Ancillary