Study site and study species
The study was conducted at Cedar Creek Natural History Area (CCNHA) in central Minnesota. For a detailed description of the study site, see Tilman . The four species studied include smooth sumac (Rhus glabra), wild rose (Rosa arkansana), bushy vetch (Lathyrus venosus), and phlox (Phlox pilosa). Smooth sumac is a perennial shrub (1–4 m tall). Wild rose is a short woody perennial shrub (1 m or shorter). Lathyrus venosus is a perennial legume (1.5 m or shorter). Phlox is anherbaceous perennial (30 cm or shorter). These species was chosen because they were abundant inside the fenced area and absent or rare outside the fence (see methods below). There is also evidence that Rhus, Lathyrus, Phlox (Haarstad, personal communication), and Rosa are all browsed by deer. The density of deer in this area has been minimally estimated to be 0.16 deer per ha . This density is similar to other protected areas, where deer herbivory has caused changes in plant composition . Target species were located inside exclosures which kept out large herbivores.
To compare abundance of the study species inside and outside the fenced enclosures, temporary transects (0.5 × 8 m) were established within and outside of each fenced area. For each species, the total number of individuals along the transects were counted. For Rosa and Lathyrus, two transects on either side of the fence were counted. Phlox was counted in four transects inside the fence and four outside. Rhus transects were established at fenced areas in 2 different fields. Two transects on either side of the fences were counted in each field.
To measure the effects of different levels of defoliation, individuals of each species within the exclosures were randomly selected and tagged. Initial height and number of leaves were recorded. Ten individuals of each species (except Rosa, which only had enough for 8 individuals for each treatment level) were randomly assigned to one of the following treatments: 1) control, no simulated herbivory, 2) 20 % of all leaves removed, 3) 40 % of all leaves removed, 4) 60 % of all leaves removed, 5) 80 % of all leaves removed, 6) 100 % of all leaves removed, or 7) meristem + natural leaf removal (called the meristem removal hereafter). This treatment was designed to simulate deer and rabbit browsing in which the entire top of a plant is often removed. Meristems, leaves and flower buds were all removed from the top of the plant and left at the bottom of the plant. The mass of the leaves removed by the meristem removal was determined and converted to the percent of the plant's total leaf biomass.
Removed leaves were dried at 55 degrees C for one week and then weighed. Following the initial damage treatment, the sites were visited twice a week. Flowers were counted on multiple visits. Seeds were collected and dried, and then weighed to give the total mass of all the seeds collected per individual plant. Mesh bags were placed over Phlox flowers because seeds are small and fall off when they ripen. No such bags were needed for Rosa or Rhus, both of which have large seeds, which are retained on the parent plant.
All statistical analysis was performed on SPSS 10.0 for Windows. One-way ANOVAs were used to determine the effect of the enclosures on the abundance of the individual species. Type III GLM analysis was used to test for differences between areas within and outside the enclosure, with abundance as the dependent variable and species, enclosure, and their interaction as the independent variables.
Total leaf biomass was calculated for each plant since larger plants generally produce more biomass and larger and/or more seeds than smaller plants. Using the weight of the leaves collected, the following formula was used to calculate the total leaf biomass per individual:
(dried leaf weight/number of leaves collected) × (total number of leaves on the plant)
This leaf biomass was used to account for plant size in statistical analysis.
Multiple regression was used to examine the relationship between percent leaf removal and flower number with plant size as the covariate. Multiple regression was also used to examine the relationship between percent leaf removal and seed mass with flower number as the covariate. Type III GLMs were run to examine the effect of the different treaments (leaf removal, meristem removal, and controls) on both flower number and seed mass. Plant size was run as a covariate for flower number, and flower number was used as a covariate for seed mass. We also corrected for the actual biomass of the leaves removed since the meristem removal often removed leaves. The level of Type III GLM analysis was also used to test for effects of different levels of leaf removal on flower number and seed mass. Bonferroni tests were performed for multiple comparisons. For all these analyses, seed mass and flower number were square root transformed.