Effects of herbivory on arrowgrass: interactions between geese, neighboring plants, and abiotic factors.
Herbivores can affect plant fitness and population dynamics in several ways: (1) directly, through biomass removal (e.g., Morrow and LaMarche 1979, Louda 1984, Crawley 1989); (2) indirectly, by altering morphological traits that determine competitive ability (e.g., Dirzo and Harper 1980, Parker and Salzman 1985, Cottam 1986, Louda et al. 1990); and (3) indirectly, by altering the environment that the plant experiences (e.g., Bazely and Jefferies 1985, 1986, Huntly and Inouye 1988, McNaughton et al. 1988, Pastor et al. 1988, Whicker and Detling 1988, Ruess et al. 1989, Prins and Nell 1990, Srivastava and Jefferies 1996). Herbivory often is selective (e.g., Crawley 1983, 1989, Archer and Detling 1984, Kinsman and Platt 1984, Sedinger and Raveling 1984, Ingham and Detling 1986, Pastor and Naiman 1992, Brown and Stuth 1993), which makes it difficult to extrapolate from the responses of individual plants to effects at the community level (Brown and Stuth 1993). The effect of herbivory on a particular species of plant should depend on the differential effects on competing species (Fox and Morrow 1992) and the ability of competitors to respond to changes in the resource environment. The effect of herbivory on competition between plants should be greatest when the environment severely limits the opportunities for compensatory regrowth in the consumed species (Louda et al. 1990).
The subarctic salt marsh is one system in which we may expect to find strong interactions between direct and indirect effects of herbivory. Plants in subarctic salt marshes face a number of stress factors (Adam 1990), including high or variable soil salinity, waterlogged soils, flooding, and, in some areas, high levels of herbivory by geese (Cargill and Jefferies 1984a, Bazely and Jefferies 1985, Gauthier et al. 1995, Prevett et al. 1985). In addition, arctic and subarctic soils frequently have low availability of nitrogen due to low rates of mineralization and decomposition (Cargill and Jefferies 1984a, Nadelhoffer et al. 1991, Kielland and Chapin 1992). Geese may have large effects on community and ecosystem processes in subarctic salt marshes (e.g., Cargill and Jefferies 1984b, Bazely and Jefferies 1986, 1989, Ruess et al. 1989, Kerbes et al. 1990, Belanger and Bedard 1994). Geese can increase rates of nutrient cycling (Cargill and Jefferies 1984b, Bazely and Jefferies 1985, 1989, Ruess et al. 1989), affect net aboveground primary productivity (NAPP; Cargill and Jefferies 1984b, Hik and Jefferies 1990), and change species composition and successional rates (Bazely and Jefferies 1986, Hik et al. 1992). Most studies of goose herbivory in subarctic systems have concentrated on relatively homogeneous "grazing lawns," but in many other communities, geese are highly selective foragers (Sedinger and Raveling 1984, Prevett et al. 1985, Prins and Ydenberg 1985, Thomas and Prevett 1986), and their preferred forage species are not always dominant. Recently, numerous studies have shown that competition and facilitation play a major role in the structuring of saltmarsh communities (Snow and Vince 1984, Bertness and Ellison 1987, Ellison 1987, Bertness 1992, Pennings and Callaway 1992, Bertness and Shumway 1993, Bertness and Hacker 1994, Castellanos et al. 1994), but it is not well understood how herbivory, abiotic factors, and interspecific interactions combine to affect establishment and growth of individual saltmarsh plants.
We examined the effect of herbivory by geese on Triglochin palustris L. (arrowgrass; Juncaginaceae), a small, herbaceous perennial plant, in a subarctic salt marsh. This species is a highly preferred forage for several species of geese (Sedinger and Raveling 1984; C. Mulder, personal observation). Arrowgrass has a high protein and low fiber content (Sedinger and Raveling 1984, Thomas and Prevett 1986) and it is of particular importance to goslings with growth rates limited by protein. Foraging for this species is highly selective: 44-98% of the diet of Cackling Canada goslings (Branta canadensis minima) prior to fledging may consist of arrowgrass (Sedinger 1984, Sedinger and Raveling 1984), but arrowgrass represents only a minor component of the biomass on the Yukon-Kuskokwim (Y-K) Delta ([less than] 4% where abundant; Mulder et al. 1996). Pacific Black Brant (Branta bernicla nigricans), the main herbivores in our study, are similar to Cackling Canada Geese in size and foraging habits, and appear to exhibit the same selectivity for arrowgrass (C. Mulder, personal observation).
On the Yukon-Kuskokwim Delta, the number of nesting pairs of Brant Geese has steadily increased for the past decade (Sedinger et al. 1993, 1994, Anthony et al. 1995). This increase in the goose population should increase grazing intensity and competition for high-quality forage, and might potentially increase breadth of the diet as preferred species become sparse (Pyke et al. 1977). On the Y-K Delta, Cackling Canada Geese ate less arrowgrass and more graminoids in years when brood densities were higher, as well as later in brood rearing, when vegetation quality declined (Sedinger and Raveling 1984). Reduced consumption of arrowgrass also was associated with greater time spent foraging, suggesting that greater search times were needed to fill the gut (Sedinger and Raveling 1988). These observations suggest that a reduction in availability of arrowgrass may have a negative impact on goose nutrition. In order to predict the effects of increased grazing intensity, coupled with decreased selectivity, on future arrowgrass availability, we need to understand the mechanisms by which geese affect arrowgrass.
The study had two principal goals: (1) to test for direct and indirect effects of goose herbivory, including selective foraging, on growth, abundance, and distribution of arrowgrass; and (2) to examine whether changes in the plant community, as a result of goose presence, altered the probability that an individual arrowgrass plant would be grazed. The latter could result simply from changes in individual or population characteristics of arrowgrass, such as plant size, nutrient content, abundance, or distribution. A change in community species composition could, however, also alter the probability of consumption of an individual arrowgrass plant by providing an associational refuge, either by reducing the rate at which herbivores encounter their prey items, by lowering the plant's "visibility" or attractiveness (Atsatt and O'Dowd 1976, O'Dowd and Williamson 1979, Hay 1986), or by increasing the availability of alternative food sources (Atsatt and O'Dowd 1976, Prins and Nell 1990). Because of the potential for other plant species to provide protection from grazing and, thus, to interact positively with arrowgrass, we will refer to them as "neighbors" rather than as "competitors."
We investigated direct effects of herbivory (removal of arrowgrass biomass), indirect effects (removal of biomass from neighboring plants and deposition of feces), and interactions between direct and indirect effects on the performance of arrowgrass by testing two sets of hypotheses generated from previous experiments and observations (Mulder et al. 1996).
Set one: the effect of geese on individual arrowgrass plants
1. Deposition of goose feces results in smaller plants, greater percentage of biomass in leaves, and lower percentage of biomass in bulbs. - Results from previous experiments suggested that feces deposition may result in a negative effect on arrowgrass, due to an increase in interspecific competition for light (Mulder et al. 1996). Artificial fertilization resulted in smaller plants with a greater percentage of biomass in leaves. Although arrowgrass was primarily nutrient limited in most communities, it appeared to be primarily light limited under fertilization (Mulder et al. 1996).
2. Feces deposition has a greater negative effect on ungrazed than on grazed arrowgrass. - Grazed arrowgrass plants are expected to be severely light limited, so that increased competition for light following feces deposition should not change their competitive environment as much as it would for ungrazed plants.
3. Effect of feces deposition are less negative when aboveground biomass of neighbors is simultaneously reduced; the effect of neighbor biomass removal on arrowgrass is more positive with than without feces deposition. - This follows from the hypothesis that the underlying mechanism for negative effects of feces deposition on arrowgrass is increased competition for light.
4. Grazing of neighbors benefits ungrazed more than grazed arrowgrass, and results in lower percentage of biomass in leaves. - Ungrazed arrowgrass may be able to reallocate carbon to storage or reproduction when more light is available, whereas grazed arrowgrass individuals are highly carbon limited and will reallocate carbon to leaves, regardless of light environment.
Set two: effect of characteristics of arrowgrass and neighbor species on the probability that an arrowgrass individual will be grazed
1. Large arrowgrass plants are more likely to be grazed than small arrowgrass plants.
2. Arrowgrass plants with high nitrogen concentrations are more likely to be grazed than arrowgrass plants with low nitrogen concentrations.
3. The grazing rate on arrowgrass is independent of arrowgrass density: where arrowgrass density is high, the probability that an individual will be grazed is low.
4. As percent cover of other species increases, the probability of an arrowgrass individual being grazed decreases. - Arrowgrass is more difficult to find when percent cover is high.
5. As the contribution of alternative food sources (other preferred species) to species composition increases, the probability of an arrowgrass individual being grazed decreases. - Selectivity for arrowgrass is lower when other good food sources are available.
All of these hypotheses assume that a goose chooses to graze in the patch of interest; it is, of course, possible that a change in vegetation may result in geese bypassing a patch altogether, but we were unable to test that hypothesis.
This study was conducted near the Tutakoke River Black Brant colony during June and July of 1993 and 1994. The site is located on the Yukon-Kuskokwim (Y-K) Delta (southwestern Alaska; 61 [degrees] 15 [minutes] N, 165 [degrees] 30 [minutes] W), and encompasses an area of [approximately]8 [km.sup.2] on both sides of the Tutakoke River. The vegetation is described in detail in Kincheloe and Stehn (1991). The Y-K Delta contains high concentrations of nesting Pacific Black Brant Geese and Cackling Canada Geese.
Arrowgrass grows on slough levees and along the edges of small ponds ("slough levee habitat"). This habitat contains a mixture of species including graminoids (e.g., Carex ramenskii, Elymus arenarius, Deschampsia caespitosa), herbaceous species (e.g., Potentilla egedei, Stellaria humifusa), and several species of dwarf willow. Slough levee habitat borders on mudflat habitat (containing primarily Carex subspathacea and Puccinellia phryganodes) on the seaside end, and on Carex wet meadows (dominated by C. ramenskii and C. glareosa) on the upland end. The following information is contributed by C. Mulder (personal observation). Arrowgrass is a stoloniferous perennial that, in this habitat, is small (usually 3-15 cm high, [less than]40 mg dry mass) and seldom reproduces sexually. Arrowgrass produces new bulbs during the growing season, but these do not normally emerge until the following spring. At the end of the growing season, the plant produces a new bulb directly above the bulb of the previous year. Arrowgrass initiates summer growth at least one week later than the dominant graminoid species; it started emerging during the first days of June in both 1993 and 1994. This species is short compared with most of the surrounding vegetation (its mid-July 1994 mean height was 3.7 cm), and where it is abundant it comprises only [approximately]4% of aboveground biomass (C. Mulder, unpublished data). This small size simplifies the interpretation of interspecific interactions; neighbors can affect resources available to arrowgrass, but the effect of arrowgrass on growth of neighbors probably is insignificant and unlikely to result in a change in the competitive ability of neighbors (Goldberg 1990).
Brant and Cackling Canada Geese forage in the slough levee habitat during the later stages of brood rearing (C. Mulder, personal observation). They consume the aboveground portions of arrowgrass only, although the much less numerous Emperor Geese (Anser canagicus) and White-fronted Geese (Anser albifrons) also may grub for bulbs before aboveground growth has begun (Budeau et al. 1991).
We conducted three experiments: the "grazing experiment," the "clipping experiment," and the "gosling experiment." Goals of the grazing experiment were: (1) to evaluate the overall effect of the presence of geese and feces deposition on arrowgrass biomass allocation, abundance, and distribution; and (2) to understand relationships between the intensity of grazing pressure and characteristics of plant growth. The grazing experiment involved a blocked design with three treatments per spatially separated block: a plot from which geese were excluded (EXCLOSE), an unexclosed plot from which feces were removed (REMOVE), and an unexclosed plot (CONTROL). This design was replicated nine times. The "triplets" (sets of three plots) experienced a wide range of grazing pressures, so that a relationship between grazing pressure and the effect of interest could be established.
The clipping experiment was designed to evaluate the importance of three aspects of goose presence (removal of arrowgrass biomass, removal of neighbor biomass, and fecal deposition), and particularly their interactions, in controlling biomass allocation, abundance, and distribution of arrowgrass. This experiment consisted of enclosed plots ("single" plots) subjected to 12 treatments in a 3 x 2 x 2 full-factorial design: three levels of feces deposition (none, FECES = 0; a single load, FECES = 1; and a double load, FECES = 2), two levels of arrowgrass clipping (not clipped, AGCLIP = 0; and clipped, AGCLIP = 1), and two levels of neighbor clipping (not clipped, NBCLIP = 0; and clipped, NBCLIP = 1). The single load of feces represented the high end of the natural range of feces deposition, whereas the double load contained more feces than would normally be deposited naturally. This design was replicated three times.
In the "gosling experiment," we investigated the effect of changes in species composition on the probability that arrowgrass would be grazed by placing captive Black Brant goslings on the premanipulated clipping experiment plots at the end of the second field season. We chose manipulated plots, rather than plots under a natural range of grazing intensities, because it is likely that, under natural conditions, correlations already exist between the size of the forage species, species composition, and forage quality (Ward and Saltz 1994). For example, a heavily grazed plot may have a high percentage of bare ground and may contain a few small arrowgrass plants with high nutrient content. These correlations make it difficult to separate causal factors. By manipulating forage species size, forage quality, and species composition independently, we diminished this problem.
Plot set-up. - Plots were distributed over an area of [approximately]4 [km.sup.2] on both sides of the Tutakoke River on 5-8 June 1993. At this time, arrowgrass was 1-2 cm high at most locations. Because arrowgrass distribution across the marsh was patchy, we selected nine general areas based on availability of arrowgrass, grazing pressure (a range from low to high), and accessibility. In each area, one triplet (for the grazing experiment) and four single plots (for the clipping experiment) were set up. The triplets consisted of three adjacent plots spaced 0.5 m apart. Four single plots were placed within 150 m of each triplet, [greater than or equal to] 20 m apart. EXCLOSE, REMOVE, and CONTROL treatments were randomly assigned to plots within each triplet, whereas manipulative treatments were randomly assigned to plots across all areas (i.e., they were not blocked by area). All plots were 1.5 x 1.5 m, and geese were exclosed from all clipping experiment plots and EXCLOSE plots by 0.3 m high chicken wire (2.5-cm mesh) and flagging tape crossed over the top of the plots.
Treatments. - Treatments were applied four times in 1993 and three times in 1994 [ILLUSTRATION FOR FIGURE 1A OMITTED]. Feces were removed at the start of the experiment (9-14 June 1993) from all plots except CONTROL plots. During each treatment period, all feces were removed from the REMOVE plots and were counted in the CONTROL plots. Collected feces were dried immediately at 50 [degrees] - 60 [degrees] C.
Feces for fecal additions in the clipping experiment were collected from a Carex subspathacaea grazing lawn (they were not abundant enough in the slough levee community). We collected only fresh feces and kept a subsample for determination of dry mass [ILLUSTRATION FOR FIGURE 2C OMITTED]. Feces were stored in a plastic bag and applied within 24 h of collection.
For AGCLIP = 1 plots, we clipped arrowgrass plants individually with scissors at [approximately]1-1.5 cm in height, avoiding inflorescences. This treatment removed 50-80% of arrowgrass plant biomass. For NBCLIP = 1 plots, we initially clipped neighbors (all species except arrowgrass) with garden shears at a height of 3-5 cm and removed the clippings. This avoided simultaneously clipping arrowgrass, but it also left most of the "understory," primarily Salix, intact. Later clippings of neighbors [ILLUSTRATION FOR FIGURE 2 OMITTED] were done with scissors at heights varying from 5 cm to 10 cm (depending on the time of year) to avoid clipping arrowgrass.
Measurements. - Measurements were conducted four times in 1993 and three times in 1994 [ILLUSTRATION FOR FIGURE 2B OMITTED], and they were, with few exceptions, identical for plots in the clipping and grazing experiments. We excavated 6-12 plants per plot, counted leaves and stolons, classified each leaf as clipped, grazed, or whole, and measured plant height, length of each leaf, bulb height and width, "stem height" (the length of the nonphotosynthetic part of the leaves, a measure of depth of the bulb in the soil), stolon length, and (on flowering plants) inflorescence height and number of flowers or fruits. In late July 1995, we collected five flowering plants per plot where available, took aboveground measurements on additional flowering plants in the plot (up to 15 plants), and counted and collected seeds from all remaining flowering plants. All plants were dried at 50 [degrees]-60 [degrees] C for 48 h in the field laboratory, divided into leaves, roots, bulbs, stolons, fruits, and remainder of the inflorescence, and weighed. Additional aboveground measurements (plant height, longest leaf length, clipping or grazing status) were recorded in the field. Plants collected in late July 1994 were transported to Fairbanks and were frozen until the time of measurement; some plants thawed prematurely, which accounts for the lower sample sizes for that time period.
We estimated arrowgrass abundance and percentage of plants clipped or grazed by counting clipped or grazed and unclipped arrowgrass in 3-5 subplots, 10 x 10 cm, in each plot. Estimates of the distribution and abundance of arrowgrass were obtained for the grazing experiment plots once a year (17-30 June 1993, 18-22 July 1994) by placing a 1 x 0.5 m Plexiglas board in the corner of each plot and marking the location of each arrowgrass individual on an acetate sheet. Measures of dispersion were obtained at two scales by randomly sampling each of the acetate maps 40 times, using a 5 x 5cm quadrat and a 10 x 10 cm quadrat, and calculating the mean: variance ratio for the number of plants located in samples for each size quadrat.
Species composition was estimated in two ways. We removed two or three 10 x 10 x 2 cm subplots from each plot, cut them to 8 x 8 x 2 cm in the field laboratory, and removed all aboveground biomass. Clippings were sorted (to species for most dicotyledons, arrowgrass, Elymus arenarius, and for Puccinellia phryganodes in 1994; to genus for Salix and Carex; most grasses were lumped), dried at 50 [degrees]-60 [degrees] C for 48 h, and weighed. A visual estimate of percent cover in 5% increments (plus a category for [less than]5%) was obtained in four adjacent 10 x 10 cm blocks at three locations in each plot. In early July, we noticed that, in almost every disturbed spot where a subplot had been removed, arrowgrass had emerged within 10 d of the disturbance, so we counted emerged arrowgrass and all other newly emerged species in these areas. On CONTROL and REMOVE plots, grazing intensity was estimated by visually estimating the proportions of plants of each species that were grazed in three 10 x 10 cm subplots per plot (in early June, late June, and late July in 1993, and in early June in 1994).
Two soil cores (10 cm deep, 5 cm in diameter) per plot were obtained on 28-29 July 1994. Cores were brought to Fairbanks and stored at 5 [degrees] C until processed. Soil subsamples (50 g wet mass) were placed in 500-mL Mason jars at 15 [degrees] C in the dark for 21 d to determine rates of net nitrogen mineralization. We measured the rate of soil respiration weekly by gas chromatography (Shimadzu 8 A, Shimadzu, Tokyo, Japan); jars were vented after each measurement. Both unincubated soils (on day 1) and incubated soils (day 22) were extracted with 2 mol/L KCl. We analyzed the filtered extracts for N[[H.sub.4].sup.+]-N using a phenol hypochlorite assay, and for (N[[O.sub.2].sup.-] + N[[O.sub.3].sup.-])-N using the Griess-Illosvay procedure in combination with a Cd-reduction column on a modified Technicon AA II (Whitledge et al. 1981). Net nitrogen mineralization was the difference between the mineral nitrogen content of soils after and before incubations.
We measured depth of thaw at three points per plot in mid-June and early July 1993, and maximum plant height (of any species) for 12 points in a grid (10 cm between points) for three areas within each plot on 1-5 July 1993.
Gosling experiment. - For the gosling experiment, we used hand-reared, 6-wk-old Black Brant goslings, which were accustomed to feeding freely on vegetation similar to that in the plots. We used a total of four goslings on 23 of the 36 clipping experiment plots. Prior to the experiment, all leaves on each arrowgrass plant in five 10 x 10 cm subplots were measured, and subplots were marked with two tongue depressors at opposite corners. Three of the four goslings were fasted for a minimum of 15 min and were placed on a plot for an adjustment period of 30 s. The behavior of each gosling was noted once each minute and classified as foraging (eating or searching for food), drinking, walking (head up), standing, preening, or grubbing (in mud or water bowl, without head tilting). When the cumulative number of "foraging minutes" (feeding or searching for food) for the three goslings reached 28 min, they were removed from the plot. This ensured that the grazing pressure on each plot was identical. Goslings were fasted for 15 min between plots, and the order in which plots were used was random with respect to previous treatment of the plots. After goslings were removed from the plot, the subplots were removed and brought back to the field laboratory. We counted and measured clipped, grazed, and whole leaves, dried the plants at 50 [degrees]-60 [degrees] C for 48 h, and estimated arrowgrass biomass per plot. We determined three indices of grazing: the proportion of plants completely grazed (1 - (number of plants remaining/number of plants before grazing)); the proportion of plants partially grazed (number of plants partially grazed/number of plants before grazing); and the total proportion grazed (complete grazing + partial grazing). The main purpose of separating complete and partial grazing was to provide insights into goose foraging behavior; differences in effects on arrowgrass individuals or populations are not clear. Complete and partial grazing were expected to be negatively correlated: an increase in complete grazing may be the result of increased ability to locate arrowgrass or preference for arrowgrass, whereas an increase in partial grazing may indicate a switch to a preferred or more visible plant.
All data were analyzed using SAS statistical packages (SAS Institute 1995). Data are referred to as significant for P [less than] 0.05 and as marginally significant for 0.05 [less than] P [less than] 0.1, and are expressed throughout as mean [+ or -] 1 SD.
Where data were available for multiple time periods (e.g., for mass and size of arrowgrass parts), we used a repeated-measures MANOVA (von Ende 1993) for mid-July and late July 1993 and 1994 to examine the effects of month, year, and their interactions with the main effects of the clipping and grazing experiments on arrowgrass size and biomass allocation. We then performed univariate ANOVAs for two time periods (late July 1993 and late July 1994) separately. For the clipping experiment, we used the full model (FECES, AGCLIP NBCLIP, and all interactions). For the grazing experiment, the ANOVA included triplet (as a blocking variable) and treatment. Significant effects were followed by contrasts between pairs determined a priori. Biomass allocation variables (percentage of mass in leaves, bulbs, roots, and stolons) were analyzed by MANOVA for two time periods (late 1993 and late 1994); where significant effects were detected, they were followed by univariate ANOVA. Biomass allocation to one plant part is not independent of allocation to other plant parts. Therefore, for any analysis in which biomass was significantly different between treatments, we ran a univariate ANOVA on the biomass allocation of the most affected plant part. We then subtracted the mass of that plant part from the total mass, calculated allocation of the remaining plant parts to the new total mass, and performed another ANOVA (for which the results are reported only where significant).
Gosling experiment. - Both previous counts of arrowgrass and results from this experiment demonstrated that plants were undercounted in the field compared with laboratory conditions, and that undercounting was proportional to density. We therefore applied a correction factor to the counts taken in the field (correct count = initial count x 1.15) and used corrected values in the calculations of both indices. Stepwise regression models using percentage of biomass or percent cover of all species as independent variables were performed to identify those species that best explained the proportion of plants grazed.
A measure of dispersion was obtained by taking the mean:variance ratio, which was used in the ANOVA for treatment effect. Significant deviations from random dispersion were detected through comparisons to a Poisson distribution using a [[Chi].sup.2] goodness-of-fit test (Pielou 1977).
Most data were log transformed (all mass data and feces counts) or square-root transformed (the proportion flowering) to meet model assumptions, but we use untransformed numbers in figures for ease of interpretation.
Comparison of grazing and clipping experiments
In June 1993 and 1994, the arrowgrass clipping treatment resulted in a higher mean proportion of clipped arrowgrass than the mean proportion of plants grazed in control plots, but by July of both years, the proportions were similar [ILLUSTRATION FOR FIGURE 2A OMITTED]. The proportion of leaves affected was similar throughout the season in both years [ILLUSTRATION FOR FIGURE 2B OMITTED]. In 1993, the FECES = 1 treatment resulted in the addition of a greater amount of feces than would normally be experienced, particularly in the early part of the season; the cumulative feces mass at the end of the season was [approximately]2.5 times that of the mean for grazing plots and was similar to that of the most intensely grazed plot [ILLUSTRATION FOR FIGURE 2C OMITTED]. Addition of feces in 1994 was within the normal range for natural deposition in 1993, but was high compared with 1994 levels of natural deposition. Both grazing levels and feces deposition suggest that grazing plots experienced lower levels of grazing in 1994 than in 1993. The neighbor clipping treatment resulted in a smaller difference in total biomass per square meter between clipped and unclipped plots than did grazing [ILLUSTRATION FOR FIGURE 2D OMITTED].
Change in plant mass and biomass allocation over time
In both the clipping and grazing experiments, plants differed in mass between months (10-13 July = "early July" 1993 and 15-20 June = "late June" 1994 vs. 20-23 July = "late July" 1993 and 21-22 July = "late July" 1994) and between years. All mass values are reported as mean [+ or -] 1 SD, in milligrams of oven-dry mass. For all plant parts in both experiments, there were significant month and year effects or a significant month x year interaction. [ILLUSTRATION FOR FIGURE 3A, B OMITTED]. Percentage of biomass in bulbs and roots decreased over the course of the season, whereas percentage biomass in leaves and stolons increased [ILLUSTRATION FOR FIGURE 3C, D OMITTED]; all values are reported as mean [+ or -] 1 SD.
Effects of clipping experiment treatments on arrowgrass: individual level
Details of the statistics associated with treatment effects are listed in Table 1. Addition of feces alone had no effect on the mass of any plant part for all time periods together (P [greater than] 0.1 for all mass variables). Clipping arrowgrass did not have an effect on the mass of any plant part over all time periods (P [greater than] 0.1), although in late June 1994, root mass in plots where arrowgrass was clipped was significantly greater than where arrowgrass was not clipped (2.29 [+ or -] 0.2 vs, 1.71 [+ or -] 0.2 mg, mean [+ or -] 1 SD). Not surprisingly, clipping arrowgrass did affect biomass allocation: it resulted in a lower percentage of biomass in leaves (40.0 [+ or -] 1.3 vs. 45.3 [+ or -] 0.2% for unclipped plants) and a higher percentage of biomass in bulbs (48.8 [+ or -] 1.5 vs. 42.1 [+ or -] 1.9% in unclipped plants). In early June 1994, prior to the first clipping treatment in that year, plants in plots where arrowgrass was clipped had significantly shorter longest leaf lengths than plants in plots where arrowgrass was not clipped (2.4 [+ or -] 0.1 vs. 2.7 [+ or -] 0.1 cm), but by late July 1994, the mean number of live leaves per plant was greater (3.0 [+ or -] 0.1 vs. 3.5 [+ or -] 0.1 leaves), so that total leaf length did not differ between clipped and unclipped plots (P [greater than] 0.1).
Clipping neighbors resulted in larger arrowgrass plants. Bulb mass and root mass (but not leaf mass) were increased when all time periods were considered simultaneously. In late June 1994, clipping neighbors resulted in greater bulb mass (8.4 [+ or -] 0.5 vs. 6.6 [+ or -] 0.5 mg), and marginally greater root mass (2.3 [+ or -] 0.2 vs. 1.7 [+ or -] 0.1 mg) and total mass (21.4 [+ or -] 2.7 vs. 15.1 [+ or -] 1.1 mg). By late July 1994, differences were no longer significant, although trends were in the same direction. The effects of clipping neighbors on root and stolon mass were significant only when arrowgrass was not clipped [ILLUSTRATION FOR FIGURE 4 OMITTED]. In early June 1994, before the first clipping treatments for that year, clipping neighbors resulted in shorter arrowgrass plants (1.3 [+ or -] 0.1 vs. 1.7 [+ or -] 0.2 cm).
Feces deposition and arrowgrass clipping interacted in their effects on both arrowgrass mass and biomass allocation, but the direction of the interaction varied. In late July 1993, the addition of feces resulted in smaller bulbs, increased percentage of biomass in leaves, and decreased percentage of biomass in bulbs only in plots where arrowgrass was not clipped [ILLUSTRATION FOR FIGURE 5 OMITTED]. In contrast, the percentage of biomass in roots was not affected by addition of feces when arrowgrass was not clipped, but it decreased with the addition of feces when arrowgrass was clipped (late July 1994; [ILLUSTRATION FOR FIGURE 5C OMITTED]).
Feces deposition and neighbor clipping interacted in their effects on vegetative reproduction in arrowgrass. Clipping neighbors resulted in increased stolon mass and length only under feces deposition [ILLUSTRATION FOR FIGURE 6A, B OMITTED].
Clipping arrowgrass negatively affected its sexual reproduction. In plots where arrowgrass was clipped, the percentage of plants that flowered was lower than [TABULAR DATA FOR TABLE 1 OMITTED] in plots where arrowgrass was not clipped (1.85 [+ or -] 1.7 vs. 0.54 [+ or -] 1.07%). Among those arrowgrass plants that flowered, clipping resulted in lower bulb mass (12.10 [+ or -] 1.1 vs. 15.41 [+ or -] 1.1 mg) and lower total fruit mass (13.52 [+ or -] 1.0 vs. 16.24 [+ or -] 1.1 mg). Feces deposition resulted in greater total fruit mass and mean fruit mass. The two clipping treatments interacted in their effect on reproduction: clipping neighbors resulted in a smaller number of fruits only in plots where arrowgrass was clipped.
Effects of grazing experiment treatments on arrowgrass: individual level
Details of the statistics associated with treatment effects are listed in Table 2. In general, plants in EXCLOSE plots were larger than those in REMOVE or CONTROL plots. When examined over all time periods simultaneously, grazing treatments affected total mass, leaf mass, and stolon mass [ILLUSTRATION FOR FIGURE 7 OMITTED]. In late June 1994, total mass, leaf mass, and bulb mass of EXCLOSE plot plants were significantly greater than those of CONTROL and REMOVE plots, but by late July, there were no differences among treatments [ILLUSTRATION FOR FIGURE 7 OMITTED]. In late July 1994, the percentage of biomass in leaves in the EXCLOSE treatment (43.7 [+ or -] 2.4%) was significantly greater than in the CONTROL treatment (38.5 [+ or -] 1.7%). Longest leaves on plants in early June 1994 in EXCLOSE plots were significantly longer (2.73 [+ or -] 0.1 cm) than those in CONTROL plots (2.28 [+ or -] 0.1 cm), and mean leaf length was greater in EXCLOSE plots than in either CONTROL plots or REMOVE plots.
To examine the effects of grazing intensity on plant mass, we regressed plant size variables against grazing intensity, as measured by the number of feces removed from the REMOVE plots in each triplet. In late July 1993, the relationship between grazing intensity and total mass or leaf mass of REMOVE plots was negative [ILLUSTRATION FOR FIGURE 8 OMITTED], but by late July 1994 there was no relationship. When we examined grazing intensity in 1993 and total mass, we found a positive, but not statistically significant, relationship ([R.sup.2] = 0.50, [F.sub.1,3] = 5.07, P = 0.11). Correlation between the number of feces removed in 1993 and 1994 (an index of use by geese) was weak (r = 0.454, P = 0.22).
Treatments in the grazing experiment had a significant effect on the percentage of plants flowering: EXCLOSE plots had a significantly higher percentage of plants flowering (2.1 [+ or -] 2.9%) than did REMOVE (0.02 [+ or -] 0.13%) or CONTROL plots (0.1 [+ or -] 0.26%).
Total bulb nitrogen in late July 1994 was significantly greater in EXCLOSE plots (0.18 [+ or -] 0.03 mg) than in CONTROL (0.12 [+ or -] 0.01 mg) or REMOVE plots (0.13 [+ or -] 0.01 mg). Total bulb carbon also was significantly greater in EXCLOSE plots (2.83 [+ or -] 0.3 mg) than in CONTROL (2.00 [+ or -] 0.2 mg) or REMOVE plots (1.87 [+ or -] 0.2 mg). There were, however, no differences in nitrogen or carbon concentrations between the treatments (P [greater than] 0.05).
Effects of treatments on arrowgrass: population level
The rate of population growth (number of plants in late 1994/number of plants in late 1993) was not significantly affected by clipping experiment treatments. For stolon productivity (total stolon mass per 100 [cm.sup.2]) in late 1994, there was an interaction between feces addition and clipping neighbors ([F.sub.2,15] = 4.93, P = 0.023): where neighbors were not clipped, feces addition decreased stolon productivity (1.10 [+ or -] 0.2 vs. 3.78 [+ or -] 1.2 mg); where neighbors were clipped, feces addition had no effect on stolon productivity.
In the grazing experiment, treatment had no effect [TABULAR DATA FOR TABLE 2 OMITTED] on the rate of population growth, but it did have a strong effect on standing biomass of arrowgrass in early July 1994: standing biomass of arrowgrass in EXCLOSE plots (1.52 [+ or -] 0.3 g/100 [cm.sup.2]) was greater than in REMOVE plots (1.09 [+ or -] 0.2 g/100 [cm.sup.2]; [F.sub.1,7] = 19.74, P = 0.003), whereas standing biomass of CONTROL plots was intermediate (1.20 [+ or -] 0.2 g/100 [cm.sup.2]).
The dispersion of arrowgrass was usually significantly clumped or not distinguishable from random; few plots exhibited a hyperdispersed distribution (Table 3). Although treatments had no significant effect at any time-scale combination (P [greater than] 0.8 for all), there were some intriguing patterns (Table 3). For example, four of five plots that showed significant hyperdispersion at some time or scale were control plots. There was a marginally significant relationship between grazing intensity in 1993 (as measured by feces deposition for REMOVE and CONTROL plots) and dispersion at the small scale for both 1993 ([F.sub.1,15] = 3.28, P = 0.09, [R.sup.2] = 0.13) and 1994 ([F.sub.1,15] = 3.98, P = 0.06, [R.sup.2] = 0.16; [ILLUSTRATION FOR FIGURE 9 OMITTED]).
Effects of treatments on community and ecosystem levels
Species composition was too variable between plots and between years to evaluate effects of the clipping and grazing experiments. The maximum height of vegetation (including all species) in the clipping experiment was significantly reduced by clipping neighbors (3.9 [+ or -] 0.2 vs. 5.7 [+ or -] 0.5 cm; [F.sub.1,23] = 14.6, P = 0.009). In the grazing experiment, vegetation height was greater in EXCLOSE plots (5.5 [+ or -] 0.2 cm) than in CONTROL (4.9 [+ or -] 0.2 cm) or REMOVE plots (4.7 [+ or -] 0.2 cm; Table 2). Treatments in the clipping experiment had no effect on the rate of net mineralization or on the rate of soil respiration (P [greater than] 0.1 for all variables). The grazing experiment treatment had no effect on respiration rate, but it did affect rates of net mineralization: net mineralization in REMOVE plots, based on dry mass, was positive (2.26 [+ or -] 0.9 [[micro]gram] N[multiplied by][(g soil).sup.-1][multiplied by][d.sup.-1]) [TABULAR DATA FOR TABLE 3 OMITTED] and significantly greater than in CONTROL plots, where it was negative (-0.67 [+ or -] 0.8 [[micro]gram] N[multiplied by][(g soil).sup.-1][multiplied by][d.sup.-1]; Table 4). Change in depth of thaw between mid-June and early July 1993 was not significantly affected by any treatment (P [greater than] 0.1 for all).
The proportion of arrowgrass completely grazed was higher for FECES = 2 plots than for FECES = 1 plots (0.39; [+ or -] 0.07 vs. 0.18 [+ or -] 0.04; [F.sub.1,12] = 7.04, P = 0.021). The total proportion grazed was lower in plots where arrowgrass was clipped than where it was not (0.43 [+ or -] 0.07 vs. 0.72 [+ or -] 0.06; [F.sub.1,13] = 6.58, P = 0.024). Clipping treatments had no effect on the proportion of arrowgrass completely grazed (P [greater than] 0.1 for all treatments).
We conducted stepwise multiple regressions to find the model that best explained the proportion of plants grazed, by species biomass or percent cover, with 11 candidate species or taxonomic groups (Table 4). In all models (proportion partially grazed, proportion completely grazed, and total proportion grazed), arrowgrass biomass or percent cover were retained in the model (Table 4). Other explanatory variables retained were [TABULAR DATA FOR TABLE 4 OMITTED] moss biomass, Chrysanthemum arcticum biomass, Potentilla egedei biomass, Carex percent cover, and Salix percent cover (Table 4). No relationship existed between mean vegetation height and the proportion of plants completely grazed ([F.sub.1,20] = 2.9, P = 0.10), but there was a negative relationship between mean vegetation height and the proportion partially grazed ([F.sub.1,20] = 8.62, P = 0.008, [R.sup.2] = 0.27).
The proportion of plants completely grazed was affected by arrowgrass characteristics. We found a significant positive relationship between the proportion completely grazed and mean total leaf length per plant ([F.sub.1,21] = 4.49, P = 0.046, [R.sup.2] = 0.14), and a marginal positive relationship with the mean number of leaves per plant ([F.sub.1,21] = 4.21, P = 0.053, [R.sup.2] = 0.13), but none with mean leaf length ([F.sub.1,21] = 1.88, P = 0.18). As expected, there was a strong negative relationship between the proportion of plants completely grazed and the sum of all leaf lengths in the subplots ([F.sub.1,21] = 10.22, P = 0.004, [R.sup.2] = 0.30). The proportion of plants completely grazed was negatively related to the mean number of arrowgrass plants in the subplots: the higher the plant density, the lower the proportion of plants completely grazed ([R.sup.2] = 0.41, F.sub.1,21] = 16.05, P = 0.0006; [ILLUSTRATION FOR FIGURE 10A OMITTED]). A regression of the number of arrowgrass plants completely grazed vs. arrowgrass density was also negative ([F.sub.1,21] = 13.29, P = 0.0015), but this was due entirely to one data point [ILLUSTRATION FOR FIGURE 10B OMITTED]. When this outlier was removed, this relationship disappeared ([F.sub.1,20] = 0.12, P = 0.73; [ILLUSTRATION FOR FIGURE 10B OMITTED]), but removing the data point had little effect on the relationship between the proportion of plants grazed and arrowgrass density ([F.sub.1,20] = 9.20, P = 0.007, [R.sup.2] = 0.28). In other words, the number of arrowgrass plants removed was independent of arrowgrass density. In contrast, the number of remaining arrowgrass plants partially grazed was strongly related to initial arrowgrass density ([R.sup.2] = 0.57, [F.sub.1,21] = 27.32, P [less than] 0.0001; [ILLUSTRATION FOR FIGURE 10C OMITTED]): as arrowgrass density increased, so did the number of plants partially grazed. Removing the "outlier" had no effect on this relationship ([R.sup.2] = 0.54, [F.sub.1,20] = 25.97, P [less than] 0.0001).
The proportion of plants partially grazed also was affected by arrowgrass characteristics, but often in the opposite direction. It was negatively related to mean total leaf length per plant ([F.sub.1,21] = 6.20, P = 0.021, [R.sup.2] = 0.19), and (marginally) related to mean length of leaves ([F.sub.1,21] = 3.73, P = 0.067, [R.sup.2] = 0.11), but showed no relationship with mean number of leaves ([F.sub.1,21] = 1.67, P = 0.21). The proportion of partially grazed plants was not related to arrowgrass density ([F.sub.1,21] = 0.67, P = 0.42) or to the total sum of all leaf lengths in the subplots ([F.sub.1,21] = 0.015, P = 0.91). The proportion of plants grazed (partially or totally) was unrelated to nitrogen concentration in leaves (P [greater than] 0.1)
Evaluation of hypotheses regarding effects of geese on arrowgrass
Relationships among herbivory, abiotic factors, and interspecific factors have been well studied in intertidal and marine systems (e.g., Lubchenco and Gaines 1981, Paine and Levin 1981, Moreno and Sutherland 1982, Duggins and Dethier 1985). In terrestrial systems, however, the importance of herbivory for changes in competition is often implied (e.g., Crawley 1983, Ellison 1987, Hik et al. 1992), but is seldom explicitly examined in natural systems (Louda et al. 1990; but see Fowler and Rauscher 1985, Bergelson, 1990). Our study strongly suggests that the presence of geese affects arrowgrass both directly and by altering the competitive environment it experiences. Moreover, the indirect effects of geese on arrowgrass may be greater than the effects of biomass removal alone. The hypotheses that deposition of goose feces results in smaller plants, greater percentage of biomass in leaves, lower percentage of biomass in bulbs, and that the effects would be greater for ungrazed than for grazed arrowgrass were fully supported by the clipping experiment. For unclipped plants only, feces deposition resulted in reduced bulb mass, reduced percentage of biomass in bulbs and roots, and increased percentage of biomass in leaves; there was no effect of feces deposition on clipped plants (Table 1, [ILLUSTRATION FOR FIGURE 5 OMITTED]). These results imply that feces deposition stimulates the growth of neighbors, leading to increased competition for light. This, in turn, results in increased demand for photosynthetic tissue, but that demand cannot be met when plants are severely carbon limited. Effects of clipping arrowgrass in the absence of other treatments were minimal: although biomass allocation was affected, plant size and total leaf length were not (Table 1). The primary cost to arrowgrass of biomass removal was a decrease in sexual reproduction: in the clipping experiment, clipped plants were less likely to flower (even though we avoided removing inflorescences when clipping). When clipped plants did flower, their bulb mass and total fruit mass were lower than for unclipped flowering plants (Table 1). However, rates of sexual reproduction were low ([less than] 4%) in all plots, so that this effect may be minor at the population level.
We expect that, although deposition of feces on plants during grazing generally will have a negative effect on arrowgrass, such effects will be less visible for plants that have been grazed. This smaller effect of feces deposition on grazed plants may explain some of the results from our grazing experiment: although plants in CONTROL plots tended to be smaller than those in REMOVE plots, differences were not significant, possibly because high feces deposition was associated with high rates of grazing. Feces deposition had a positive effect on flowering arrowgrass; the addition of a single load of feces resulted in a greater total mass of fruit (Table 1). Flowering plants were, however, larger than nonflowering plants, and they may, therefore, represent those plants that suffered relatively little competition (e.g., because they were in an open location) and, therefore, experienced only the advantages of feces deposition. If this is the case, then feces deposition may increase the variance in plant size. Addition of a single load of feces also resulted in a greater standing biomass for arrowgrass than either addition of a double load or no feces, which may have resulted from an increased percentage of biomass in leaves.
We observed less support for the hypothesis that the effect of feces deposition would be lessened when aboveground biomass of neighbors was simultaneously reduced. Such an effect was seen only for vegetative reproduction: stolon length and mass increased under feces deposition when neighbors were clipped, but decreased when neighbors were not clipped (Table 1; [ILLUSTRATION FOR FIGURE 6 OMITTED]). The last hypothesis, that the grazing of neighbors would benefit previously ungrazed arrowgrass more than grazed arrowgrass, was partially supported by the clipping experiment. Clipping neighbors resulted in increased root mass and stolon mass only for unclipped arrowgrass (Table 1), suggesting that, for grazed plants, the advantage of simultaneous grazing of neighbors was minimal.
Our data demonstrate that the presence of geese affects arrowgrass by altering the competitive environment for all plants and by changing biomass allocation and sexual reproduction in grazed plants. Results from these experiments thus support the hypothesis of Louda et al. (1990) that herbivory can change a plant's ability to acquire limited resources by altering its morphology. Although we did not directly measure effects of our treatments on competitors, our results also support the hypothesis that effects of herbivory depend on differential impacts on competing species (Fox and Morrow 1986, Louda et al. 1990). In our case, the differential impact of geese on plant species may be due both to selectivity for arrowgrass and to the greater ability of competitors to take advantage of the altered nutrient environment following feces deposition.
The overall negative effect of the presence of geese on arrowgrass is clearly negative, as evidenced by the greater size, greater C and N mass, and higher probability of flowering of plants in exclosed compared to unexclosed plots (Table 2, [ILLUSTRATION FOR FIGURE 7 OMITTED]). In other words, effects of biomass removal plus increased competition following feces deposition are more negative than any effects of increased competition for light (through elimination of trampling and herbivory for all species) in the absence of geese. The effects of feces or urine deposition in grazing systems have been studied primarily in systems where either grazing is not selective or the preferred species account for a large percentage of the biomass (e.g., Stenseth 1978, Woodmansee 1978, McNaughton 1979, Cargill and Jefferies 1984a, b, Bazely and Jefferies 1985, 1986). In a subarctic saltmarsh community along La Perouse Bay (Manitoba, Canada), where geese graze on monospecific stands of preferred forage species such as Carex subspathacea and Puccinellia phryganodes, moderate levels of herbivory increased aboveground productivity and forage quality through increased rates of nitrogen cycling, resulting in a positive feedback cycle between herbivory and food availability (Cargill and Jefferies 1984a, b, Bazely and Jefferies 1985, Ruess et al. 1989, Hik and Jefferies 1990). Yet, in many communities, geese are highly selective foragers (Sedinger and Raveling 1984, Prevett et al. 1985, Prins and Ydenberg 1985, Thomas and Prevett 1986). Our data indicate that in areas where foraging is selective and the preferred species represent a small proportion of the total biomass, such a positive effect of goose presence may be absent. This raises the possibility that for herbivores such as Brant Geese and Cackling Canada Geese, which depend on monospecific "grazing lawns" for some parts of the growth season, but on selective foraging in communities with many species during other time periods, there may be no positive relationship between forage availability over the whole season and level of herbivory.
Evaluation of hypotheses about effects of arrowgrass and neighbor species on probability of arrowgrass being grazed
Associational resistance (a reduction in herbivory on one plant species in the presence of another) has been reported in a number of systems, many of them involving specialized insect herbivores (e.g., Tahvanainen and Root 1972, Bach 1980, Risch 1980, 1981, Kareiva 1982, Ellison 1987), but also in some systems involving more generalist, wide-ranging herbivores (e.g., McNaughton 1978, Hay 1986). The repellent-plant hypothesis (that physical or chemical characteristics of an unpalatable species interfere with the herbivore's ability to find or use the palatable species; Atsatt and O'Dowd 1976, McNaughton 1978, Hay 1986, Ellison 1987) and the attractant-decoy hypothesis, (that other species provide more attractive alternative food sources; Atsatt and O'Dowd 1976) have both been used to explain this effect. These hypotheses are difficult to test, particularly for non-insect, generalist species, because correlations between rates of herbivory and the presence of other species in natural situations can be the result of both direct and indirect effects of past herbivory (Ward and Saltz 1994). For example, past herbivory may have increased the palatability of the preferred species and reduced the presence of other palatable species, resulting in a negative correlation between the percentage of preferred plants grazed and the biomass of other species. We reduced this problem by using premanipulated plots in the gosling experiment, in which correlations between arrowgrass size, community composition, and arrowgrass quality (N concentration) had been decreased.
Both arrowgrass characteristics and community composition affected the probability of arrowgrass being grazed. Plant size was a good indicator of grazing probability: the percentage of plants completely grazed increased as the mean number of leaves increased; plants that had been clipped previously were less likely to be grazed than those that had not been clipped. On the other hand, there was no evidence that plants in plots with a higher mean N concentration were more likely to be grazed. Because we pooled plants per plot to obtain a sample large enough for analysis, we do not know the within-plot variance in N concentration, which makes it impossible to evaluate the power of our test.
Comparing complete vs. partial herbivory provided some insights into the foraging behavior of the goslings. The number of arrowgrass plants completely removed was unrelated to arrowgrass density [ILLUSTRATION FOR FIGURE 10B OMITTED], but the number of arrowgrass plants partially removed increased with density [ILLUSTRATION FOR FIGURE 10A OMITTED], suggesting a change in the effectiveness with which arrowgrass was consumed. Additional species, both palatable and unpalatable to geese, were included in the models explaining the percentage of plants grazed. In general, relationships between species presence (biomass or percent cover) and the percentage of arrowgrass removed were negative, whereas relationships with the percentage of plants partially grazed were positive (Table 4). Moss biomass, which is correlated with patches of bare ground (C. Mulder, personal observation), was the exception. Therefore, it appears that in plots with little vegetation, the percentage of completely grazed arrowgrass is very high, whereas the percentage of partially grazed arrowgrass declines. Potentilla egedei and Carex are potential alternative food sources, whereas dwarf Salix species provide good ground cover, suggesting that both increased crypsis and the availability of alternative food sources decrease the probability of arrowgrass being grazed. The significant correlations between the percentage of plants partially grazed and biomass or percent cover of other species indicate that changes in the probability of grazing are not simply the result of a shift to other food sources, which should lead to changes in complete grazing only. Results of this experiment are consistent with the observation by Sedinger and Raveling (1984) that arrowgrass is grazed more heavily by Cackling Canada Geese on the mud-flats, where it is surrounded only by very short Puccinellia phryganodes and Carex subspathacea, than in the mixed-species slough levee communities.
Our data support both the repellent-plant hypothesis and the attractant-decoy hypothesis, raising the possibility that, at least for wide-ranging generalist herbivores, both mechanisms may operate at the same time. As was found in work on a temperate salt marsh on the east coast of the Unites States, interactions between plant species in the slough levee community range from positive to negative, depending on conditions. Neighboring plants can act as competitors, but they can also have positive effects by buffering harsh physical conditions (Bertness and Shumway 1993, Bertness and Hacker 1994), or, as in our case, reducing levels of herbivory.
Implications of experimental results for effects of increasing grazing intensity
Given a steady increase in the number of nesting pairs of Brant Geese on the Y-K Delta in the past decade (Sedinger et al. 1993, 1994, Anthony et al. 1995), potential effects of increased grazing intensity and competition for high-quality forage, particularly if coupled with greater consumption of other plant species, are of interest. Predictions regarding the effects of an increase in grazing intensity, coupled with a decrease in selectivity, on the future availability of arrowgrass differ, depending on whether they are drawn from the clipping and grazing experiments or from the gosling experiment. Results from the clipping experiment suggest that if an increase in grazing pressure is accompanied by an decrease in selectivity, it may initially benefit arrowgrass: plants will be more likely to be grazed, but light availability should increase as neighbors are also consumed; hence, the overall effect on arrowgrass may be neutral. The grazing experiment provides mixed evidence for this hypothesis. In 1993, a negative relationship between grazing intensity and plant mass existed, contradicting our hypothesis. In 1994, however, this relationship was not found, and a slightly positive relationship occurred between 1993 grazing intensity and 1994 plant mass, which we would expect if, as predicted from the clipping experiment, the percentage of biomass in vegetative reproduction in 1993 was greater in heavily grazed plots. Thus, both clipping and grazing experiments suggest that a moderate increase in grazing intensity is not necessarily detrimental to arrowgrass.
In contrast, the gosling experiment suggests that any increase in grazing pressure accompanied by a decrease in selectivity will exacerbate the negative effects of goose grazing. If the presence of some neighboring species provides arrowgrass with a measure of protection from goose grazing (either by providing alternative foods or by decreasing detection of arrowgrass), greater consumption of neighbors should result in greater probability of arrowgrass being grazed. Plants that have been grazed previously will have shorter leaves, but more of them; this should decrease the probability of partial grazing, but increase the probability of presumably more detrimental complete grazing. Overall, the gosling experiment results point toward a rapid decline in arrowgrass under increased grazing pressure accompanied by decreased selectivity for arrowgrass by geese.
The effects of increased grazing pressure on arrowgrass populations will depend, of course, on the relative magnitudes of these different predicted effects. Such effects cannot be estimated without more detailed information on the foraging behavior of geese at both small and large scales when faced with different foraging situations. Extrapolating from the gosling experiment to larger scale effects has several limitations. First, we used hand-reared goslings that, although raised on natural vegetation, may not show exactly the same foraging behavior as wild geese. Second, we do not know at what point goslings would have left the plot and sought other foraging areas had they not been prevented from doing so. A better understanding of how geese select a foraging area is essential to extrapolating our short-term, small-scale results. Although the increase in standing biomass of arrowgrass following feces deposition may result in greater attractiveness of previously grazed areas, both observations and feces correlations between 1993 and 1994 suggest that there is little consistency in the specific areas that are grazed. Questions regarding the way in which geese locate arrowgrass within a patch also remain. For example, the grazing experiment provides evidence that goose herbivory results in a decrease in clumping, but it is unclear how this would affect the probability of arrowgrass being grazed. Finally, we currently have very little information on how selectivity changes with goose population size, which is crucial for predicting the feedback of changes in species composition to goose populations.
When we examine results from the clipping and grazing experiments only, we are tempted to conclude that the overall effect of geese on arrowgrass is negative, in part because, under increased feces deposition, arrowgrass is outcompeted for light by its neighbors. Nevertheless, the gosling experiment suggests that the presence of other species also can have a positive effect on arrowgrass. Thus, the way in which we view other species in the system (as competitors or potential protectors) changes the predictions we make regarding the effects of an increase in goose numbers when accompanied by an increase in diet breadth. We cannot predict the net result of these opposing effects; however, they demonstrate that, in examining the effect of herbivory on a plant in the context of a community, we need to focus not just on direct and indirect effects of the herbivore on the plant, but also on feedback from the plant community to the herbivore through changes in forage plant size, forage quality, and species composition. Understanding such feedbacks will require the coupling of clipping or controlled grazing experiments with experiments that can independently assess the effects of these factors (plant size, forage quality, and species composition) on herbivore behavior, forage preference, and foraging efficiency.
We thank S. Clark and the Tutakoke field crews of 1993 and 1994 (particularly B. and D. Person, T. Olson, and T. Obritschkewitsch) for help with field work well beyond the call of duty, and L. Oliver for technical support. We are particularly grateful to M. Chambers for the "loan" of goslings that she spent an enormous amount of time raising, and M. Eichholtz for his many nights tending to the aforementioned goslings. J. Bryant, E. Rexstad, K. Schwaegerle, and J. Sedinger provided helpful discussion through all stages of the project and we thank them and two anonymous reviewers for their helpful critiques of earlier versions of the manuscript. Financial support was provided by a University of Alaska Fairbanks Office of Global Change and Systems Research grant and a UAF Dissertation Completion scholarship to C. Mulder, and by NSF grant OPP-9214970.
Adam, P. 1990. Saltmarsh ecology. Cambridge University Press, Cambridge, UK.
Anthony, R. M., W. H. Anderson, J. S. Sedinger, and L. L. McDonald. 1995. Estimating populations of nesting Brant using aerial videography. Wildlife Society Bulletin 23:80-87.
Archer, S. R., and J. K. Detling. 1984. The effects of defoliation and competition on regrowth of tillers of two North American mixed grass prairie graminoids. Oikos 43: 351-357.
Atsatt, P. R., and D. J. O'Dowd. 1976. Plant defense guilds. Science 193:24-29.
Bach, C. E. 1980. Effects of plant diversity and time of colonization on an herbivore-plant interaction. Oecologia 44:319-326.
Bazely, D. R., and R. L. Jefferies. 1985. Goose faeces: a source of nitrogen for plant growth in a grazed saltmarsh. Journal of Applied Ecology 22:693-703.
Bazely, D. R., and R. L. Jefferies. 1986. Changes in the composition and standing crop of salt-marsh communities in response to the removal of a grazer. Journal of Ecology 74:693-706.
Bazely, D. R., and R. L. Jefferies. 1989. Lesser Snow Geese and the nitrogen economy of a grazed salt marsh. Journal of Ecology 77:24-34.
Belanger, L., and J. Bedard. 1994. Role of ice scouring and goose grubbing in marsh plant dynamics. Journal of Ecology 82:437-445.
Bergelson, J. 1990. Spatial patterning in plants: opposing effects of herbivory and competition. Journal of Ecology 78:937-948.
Bertness, M.D. 1992. The ecology of New England salt marsh plant communities. American Scientist 80:260-268.
Bertness, M.D., and A.M. Ellison. 1987. Determinants of pattern in a New England salt marsh plant community. Ecological Monographs 57:129-147.
Bertness, M.D., and S. D. Hacker. 1994. Physical stress and positive associations among marsh plants. American Naturalist 144:363-372.
Bertness, M.D., and S. W. Shumway. 1993. Competition and facilitation in marsh plants. American Naturalist 142:718-724.
Brown, J. R., and J. W. Stuth. 1993. How herbivory affects grazing tolerant and sensitive grasses in a central Texas grassland: integrating plant response across hierarchical levels. Oikos 67:291-298.
Budeau, D. A., J. T. Ratti, and C. R. Ely. 1991. Energy dynamics, foraging ecology, and behavior of prenesting greater white-fronted geese. Journal of Wildlife Management 55:556-563.
Cargill, S. M., and R. L. Jefferies. 1984a. The effects of grazing by lesser snow geese on the vegetation of a subarctic salt marsh. Journal of Applied Ecology 21:669-686.
Cargill, S. M., and R. L. Jefferies. 1984b. Nutrient limitation of primary production in a sub-arctic salt marsh. Journal of Applied Ecology 21:657-668.
Castellanos, E. M., M. E. Figueroa, and A. J. Davy. 1994. Nucleation and facilitation in saltmarsh succession: interactions between Spartina maritima and Arthrocnemum perenne. Journal of Ecology 82:239-248.
Cottam, D. A. 1986. The effects of slug-grazing on Trifolium repens and Dactylis glomerata in monoculture and mixed sward. Oikos 47:275-279.
Crawley, M. J. 1983. Herbivory: the dynamics of plant-animal interactions. University of California Press, Berkeley, California, USA.
-----. 1989. Herbivores and plant population dynamics. Pages 367-392 in A. J. Davy, M. J. Hutchings, and A. R. Watkinson, editors. Plant population ecology. Blackwell, Oxford, UK.
Dirzo, R., and J. L. Harper. 1980. Experimental studies on slug-plant interactions. II. The effect of grazing by slugs on high density monocultures of Capsella bursa-pastoris and Poa annua. Journal of Ecology 68:999-1011.
Duggins, D. O., and M. N. Dethier. 1985. Experimental studies of herbivory and algal competition in a low intertidal habitat. Oecologia 67:183-191.
Ellison, A.M. 1987. Effects of competition, disturbance, and herbivory on Salicornia europaea. Ecology 68:576-586.
Ende, C. N. von. 1993. Repeated-measures analysis: growth and other time-dependent measures. Pages 113-137 in S. M. Scheiner and J. Gurevitch, editors. Design and analysis of ecological experiments. Chapman and Hall, New York, New York, USA.
Fowler, N. L., and M.D. Rausher. 1985. Joint effects of competitors and herbivores on growth and reproduction in Aristolochia reticulata. Ecology 66:1580-1587.
Fox, L. R., and P. A. Morrow. 1992. Eucalypt responses to fertilization and reduced herbivory. Oecologia 89:214-222.
Gauthier, G., R. J. Hughes, A. Reed, J. Beaulieu, and L. Rochefort. 1995. Effect of grazing by greater snow geese on the production of graminoids at an arctic site (Bylot Island, NWT, Canada). Journal of Ecology 83:653-664.
Goldberg, D. E. 1990. Components of resource competition in plant communities. Pages 27-49 in J. B. Grace and D. Tilman, editors. Perspectives on plant competition, Academic Press, San Diego, California, USA.
Hay, M. E. 1986. Associational plant defenses and the maintenance of species diversity: turning competitors into accomplices. American Naturalist 128:617-641.
Hik, D. S, and R. L. Jefferies. 1990. Increases in the net above-ground primary production of a salt-marsh forage grass: a test of the predictions of the herbivore-optimization model. Journal of Ecology 78:180-1995.
Hik, D. S., R. L. Jefferies, and A. R. E. Sinclair. 1992. Foraging by geese, isostatic uplift and asymmetry in the development of salt-marsh plant communities. Journal of Ecology 80:395-406.
Huntly, N., and R. Inouye. 1988. Pocket gophers in ecosystems: patterns and mechanisms. BioScience 38:786-793.
Ingham, R. E., and J. K. Detling. 1986. Effects of defoliation and nematode consumption on growth and leaf gas exchange in Bouteloua curtipendula. Oikos 46:23-28.
Kareiva, P. 1982. Experimental and mathematical analyses of herbivore movement: quantifying the influence of plant spacing and quality on foraging discrimination. Ecological Monographs 52:261-282.
Kerbes, R. H., P. M. Kotanen, and R. L. Jefferies. 1990. Destruction of wetland habitats by lesser snow geese: a keystone species on the west coast of Hudson Bay. Journal of Applied Ecology 27:242-258.
Kielland, K., and F. S. Chapin III. 1992. Nutrient absorption and accumulation in arctic plants. Pages 321-335 in F. S. Chapin III, R. L. Jefferies, J. F. Reynolds, G. R. Shaver, and J. Svoboda, editors. Arctic ecosystems in a changing climate: an ecophysiological perspective. Academic Press, New York, New York, USA.
Kincheloe, K. L., and R. A. Stehn. 1991. Vegetation patterns and environmental gradients in coastal meadows on the Yukon-Kuskokwim delta, Alaska. Canadian Journal of Botany 69:1616-1627.
Kinsman, S., and W. J. Platt. 1984. The impact of herbivores (Hellodines nyctaginella: Lepidoptera) upon Mirabilis hirsuta, a fugitive prairie plant. Oecologia 65:2-6.
Louda, S. M. 1984. Herbivore effect on stature, fruiting and leaf dynamics of a native crucifer. Ecology 65:1379-1386.
Louda, S. M., K. H. Keeler, and R. D. Holt. 1990. Herbivore influences on plant performance and competitive interactions. Pages 413-444 in J. B. Grace and D. Tilman, editors. Perspectives on plant competition. First edition. Academic Press, San Diego, California, USA.
Lubchenco, J., and S. D. Gaines. 1981. A unified approach to marine plant-herbivore interactions. I. Populations and communities. Annual Review of Ecology and Systematics 12:405-437.
McNaughton, S. I. 1978. Serengetic ungulates: feeding selectivity influences the effectiveness of plant defense guilds. Science 199:806-807.
-----. 1979. Grazing as an optimization process: grass-ungulate relationships in the Serengeti. American Naturalist 113:691-703.
McNaughton, S. J., R. W. Ruess, and S. W. Seagle. 1988. Large mammals and process dynamics in African ecosystems. BioScience 38:794-799.
Moreno, C. A., and J. P. Sutherland. 1982. Physical and biological processes in a Macrocystis pyrifera community near Valdivia, Chile. Oecologia 55:1-6.
Morrow, P. A., and V. C. LaMarche. 1979. Tree ring evidence for chronic insect suppression of productivity in subalpine Eucalyptus. Science 201:1244-1245.
Mulder, C. P. H., R. W. Ruess, and J. S. Sedinger. 1996. Effects of environmental manipulations on Triglochin palustris: implications for the role of goose herbivory in controlling its distribution. Journal of Ecology 84:267-278.
Nadelhoffer, K. J., A. E. Giblin, G. R. Shaver, and J. A. Laundre. 1991. Effects of temperature and substrate quality on element mineralization in six arctic soils. Ecology 72:242-253.
O'Dowd, D. J., and G. B. Williamson. 1979 Stability conditions in plant defense guilds. American Naturalist 114:379-383.
Paine, R. T., and S. A. Levin. 1981. Intertidal landscapes: disturbance and the dynamics of pattern. Ecological Monographs 51:145-178.
Parker, M. A., and A. G. Salzman. 1985. Herbivore exclosure and competitor removal: effects on juvenile survivorship and growth in the shrub, Gutierrezia microcephala. Journal of Ecology 73:903-913.
Pastor, J., and R. J. Naiman. 1992. Selective foraging and ecosystem processes in boreal forests. American Naturalist 139:690-705.
Pastor, J., R. J. Naiman, B. Dewey, and P. McInnes. 1988. Moose, microbes, and the boreal forest. BioScience 38:770-777.
Pennings, S. C., and R. M. Callaway. 1992. Salt marsh plant zonation: the relative importance of competition and physical factors. Ecology 73:681-690.
Pielou, E. C. 1977. Mathematical ecology. Wiley, New York, New York, USA.
Prevett, J. P., I. E Marshall, and V. G. Thomas. 1985. Spring foods of snow and Canada geese at James Bay. Journal of Wildlife Management 49:558-563.
Prins, A. H., and H. W. Nell. 1990. Positive and negative effects of herbivory on the population dynamics of Senecio jacobaea L. and Cynoglossum officinale L. Oecologia 83:325-332.
Prins, H. H. T, and R. C. Ydenberg. 1985. Vegetation growth and a seasonal habitat shift of the barnacle goose (Branta leucopsis). Oecologia 66:122-125.
Pyke, G. H., H. R. Pulliam, and E. L. Charnov. 1977. Optimal foraging: a selective review of theory and tests. Quarterly Review of Biology 52:137-154.
Risch, S. 1980. The population dynamics of several herbivorous beetles in a tropical agroecosystem: the effect of intercropping corn, beans and squash in Costa Rica. Journal of Applied Ecology 17:593-612.
-----. 1981. Insect herbivore abundance in tropical monocultures and polycultures: an experimental test of two hypotheses. Ecology 62:1325-1340.
Ruess, R. W., D. S. Hik, and R. L. Jefferies. 1989. The role of lesser snow geese as nitrogen processors in a sub-arctic salt marsh. Oecologia 79:23-29.
SAS Institute. 1995. SAS/STAT user's guide. Release 6.1 edition. SAS Institute, Cary, North Carolina, USA.
Sedinger, J. S. 1984. Protein and amino acid composition of tundra vegetation in relation to nutritional requirements of geese. Journal of wildlife Management 48:1126-1136.
Sedinger, J. S., C. J. Lensink, D. H. Ward, R. M. Anthony, M. L. Wege, and G. V. Byrd. 1993. Current status and recent dynamics of the Black Brant (Branta bernicla) breeding population. Wildfowl 44:49-59.
Sedinger, J. S., and D. G. Raveling. 1984. Dietary selectivity in relationship to availability and quality of food for goslings of cackling geese. Auk 101:295-305.
Sedinger, J. S., and D. G. Raveling. 1988. Foraging behavior of cackling Canada goose goslings: implications for the roles of food availability and processing rate. Oecologia 75:119-124.
Sedinger, J. S., D. H. Ward, R. M. Anthony, D. V. Derksen, C. J. Lensink, K. S. Bollinger, and N. K. Dawe. 1994. Management of Pacific Brant: population structure and conservation issues. Transactions of the 59th North American Wildlife and Natural Resources Conference, 1994:50-59.
Snow, A. A., and S. W. Vince. 1984. Plant zonation in an Alaskan salt marsh. II. An experimental study of the role of edaphic conditions. Journal of Ecology 72:669-684.
Srivastava, D. S., and R. L. Jefferies. 1996. A positive feedback: herbivory, plant growth, salinity, and the desertification of an Arctic salt-marsh. Journal of Ecology 84:31-42.
Stenseth, N. C. 1978. Do grazers maximize individual plant fitness? Oikos 31:299-306.
Tahvanainen, J. O., and R. B. Root. 1972. The influence of vegetational diversity on the population ecology of a specialized herbivore, Phyllotreta cruciferae (Coleoptera: Chrysomelidae). Oecologia 10:321-346.
Thomas, V. G., and J. P. Preyett. 1986. The nutritional value of arrowgrasses to geese at James Bay. Journal of Wildlife Management 44:830-835.
Ward, D., and D. Saltz. 1994. Foraging at different spatial scales: dorcas gazelles foraging for lilies in the Negev desert. Ecology 75:48-58.
Whicker, A.D., and J. K. Detling. 1988. Ecological consequences of prairie dog disturbances. BioScience 38:778-786.
Whitledge, T. E., S. C. Mallow, C. J. Patton, and C. D. Wirick. 1981. Automated nutrient analysis in seawater. Technical Report of the Oceanographic Sciences Division, Department of Energy and Environment, Brookhaven National Laboratory, BNL 51398, Upton, New York, USA.
Woodmansee, R. G. 1978. Additions and losses of nitrogen in grassland ecosystems. BioScience 28:448-453.
|Printer friendly Cite/link Email Feedback|
|Author:||Mulder, Christa P.H.; Ruess, Roger W.|
|Date:||May 1, 1998|
|Previous Article:||Scale-dependent variation in local vs. regional effects on coral species richness.|
|Next Article:||Experimental studies of species-specificity in Cecropia-ant relationships.|