Plant species abundance and species richness in upland regions and seasonal stream channels in Freeman Meadow (Lakes Basin, California)

 

Katrina Strathmann and Edward F. Connor

 

 

Abstract

 

We collected preliminary data using both a within- and a between-subjects sampling design to determine the sampling design and sample size best suited to test the null hypothesis that species richness and percent cover of forbs did not differ between dry stream beds and upland areas in an alpine meadow. Based on our preliminary data, we chose to use a within-subjects sampling design with a sample size of n = 6 subjects (12 total plots). The overall results indicate that both species richness and percent cover of forbs is higher in upland areas than within dry streambeds.

 

Introduction

 

Presence of surface water can strongly influence plant species composition.  Species that are adapted to periodic immersion or saturated soil conditions would be expected to be dominant under those conditions, and species that are unable to tolerate periodic immersion would be expected to be absent or lower in abundance. We examined the species richness and percent cover of plant species in flower in dry streambeds and compared these values to adjacent streamside locations that had never been inundated. The purpose of this experiment was to illustrate the role of preliminary data in determining sample size, and in choosing between a within-subjects or a between-subjects experimental design. Using the sample size and design we chose based on the preliminary data; we then performed the actual experiment.     

 

  Methods

 

Our study site was located in Freeman meadow to the northwest of Yuba Pass, in Tahoe National Forest of the northern Sierra Nevada.  The study site is a high elevation mountain meadow with meandering perennial streams and multiple shallow seasonal streambeds that were dry at the time of sampling.  Seasonal moisture levels were below the annual average.  In July 2001, species composition included various perennial forbs, graminoids (predominantly families Cyperaceae, Poaceae and Juncaceae) and patchily distributed willow thickets (Salix sp.).  We investigated the species richness and abundance (percent cover) of flowering forbs within the dry seasonal streambeds compared to upland areas of the meadow adjacent to the streambeds.

 

We estimated species richness and percent cover of flowering forb species within dry seasonal streambeds and in adjacent upland meadow over two consecutive days (July 11 and 12, 2001).  The region of the study site we sampled occurred to the north of the entrance road and to the south of the perennially flowing stream.  In order to simplify plant identification, we created a list of all flowering species throughout this region of Freeman meadow by an ocular survey prior to sampling.  Only species on this list were noted for species richness and percent cover estimates (see Table 1 below).  We randomly selected seasonally dry streambed locations throughout the study region and placed flags in the center of the streambeds.  For in-streambed sampling locations, we placed a 0.25 m² quadrat over the center of the flag.  For adjacent upland meadow sampling locations, we randomly selected one side of the streambed and measured 1.25 m from the streambed bank at an angle perpendicular to the bank.  At each sampling location, species on our original list were inventoried irrespective of phenology, and percent cover was estimated for each.

 

Table 1. Plant Species Sampled

 

Power calculation and selection of experimental design. In order to determine the sample size necessary to have 80% power against a specified alternative hypothesis with a  = 0.05, and the most efficient sampling design (between-subjects or within-subjects), we collected preliminary data for two subjects per treatment in a within subjects and a between-subjects design.  Two streambed locations were randomly selected to collect within-subjects data, where in-streambed plots and plots in the adjacent upland meadow were paired at the same streambed sampling point.  To collect between-subjects data, the remaining four-streambed locations were randomly allocated to either in-streambed plots or adjacent upland meadow plots.  We used the program PS (Power and Sample Size Calculations Version 1.0.12) to calculate the minimum number of plots required to have 80% power to detect a difference in species richness of 2 species and a difference in percent cover of 25% based on the standard deviations of these initial sample plots. 

 

Statistical analysis.  All 6 streambed locations were subsequently sampled using a within-subjects design. Data was analyzed using paired t-tests in SPSS 11.01.

 

Results

 

For both species richness and percent cover, the within-subjects design required far fewer plots in order to detect the specified effect than did the between-subjects design (Table 2). The sample size required to detect the specified difference in species richness in the within-subjects design was 3 subjects or 6 total plots (3 pairs consisting of one plot in and one plot outside the dry streambed).  To detect a 25% difference in percent cover in the within-subjects design required 8 subjects or 16 total plots.  Based on these preliminary data we decided to use a within-subjects design, and chose 6 replicates as a compromise between the sample sizes required to detect the specified effects for species richness and percent cover. 

 

 

Table 2. Standard deviations and required sample sizes estimated from preliminary data.

* n - Number of experimental subjects required, number of plots required in parentheses.

 

We found that the species richness of forbs and their percent cover was higher in upland areas of the meadow than in seasonal streambed channels (species richness paired - t = 3.264, df = 5 p = 0.022, percent cover: paired - t = 2.956, df = 5, p = 0.032, Figures 1 and 2). 

Discussion

 

Our preliminary data led us to chose a within-subjects design over a between-subjects design.  Within-subjects experimental designs can be more powerful than between-subjects designs when the responses of subjects to each of the treatments are correlated. Such correlations lead to a reduction in the estimated error variation, and increase the probability of rejecting a false null hypothesis. Although our sample sizes were low for the preliminary data, the estimated correlations for both species richness and percent cover of forbs was r = 1.  Winer et al. (1991) point out that for equal sampling effort within-subjects designs have (n – 1) fewer degrees of freedom than an equivalent between-subjects design. Therefore, the reduction in error variation caused by the correlation of responses within each subject must be large enough to more than offset this loss of degrees of freedom for within-subjects designs to be more powerful than between-subjects designs. While Winer et al. (1991) provide no specific recommendations concerning how large these correlation must be, an estimated r = 1 would suggest that a within-subjects design will be more powerful than a between-subjects design.  The sample sizes we report in Table 2 also suggest that to detect the same effect size with the same a (Type I) and b (Type II) error rates, that the within-subjects design is preferred.

 

Overall we found that more species of flowering forbs comprising a greater percentage of the plant cover were observed outside of dry streambeds than within dry streambeds. Presumably some species cannot tolerate extended periods of inundation associated with these shallow streams, or potentially annual species have a shorter temporal window in which to establish and complete development after the streams dry out.

 

Within the channels, a small number of graminaoids in the family Cyperaceae were observed to be dominant, but not flowering. While we sampled during an unusually dry summer, the presence of graminoids and a limited number of forbs in the streambed bottoms suggest that these depressions are not immersed for the complete summer season each year.  It is likely that seasonal stream inundation prevents most species except those adapted to those conditions from persisting.

 

Figure 1. Species richness of forbs in and out

of dry stream beds in Freeman meadow.

 

 

Figure 2. Percent cover of forbs in and out

of dry streambeds in Freeman meadow.

 

 

Literature Cited

 

Dupont, W.D. and W.D. Plummer. 1997. PS: Power and sample size http://www.mc.vanderbilt.edu/prevmed/ps/index.htm (11 June 2002).

 

Winer, B.J., D.R. Brown, and K.M. Michels. 1991. Statistical Principles in Experimental Design. McGraw- Hill, Inc.: New York.