Changes in plant species composition along an elevation gradient in an old-growth bottomland hardwood-Pinus taeda forest in southern Arkansas1

Journal of the Torrey Botanical Society, Jan-Mar 2005 by Grell, Adrian G, Shelton, Michael G, Heitzman, Eric

CALCULATIONS AND DATA ANALYSIS. Depending on stratum, density (number ha^sup -1^), basal area (m^sup 2^ ha^sup -1^), and/or coverage (%) were calculated for each species and plot. A species' frequency of occurrence for seedlings, saplings, and herbaceous layers was the percentage of subplots within a plot that contained at least one individual of that species. For the overstory, frequency of occurrence of a species was expressed as present or absent.

Importance values (IVs) were calculated for each vegetative stratum (Curtis and Macintosh 1951). They were calculated for each species by plot using relative values as follows: overstory (density, basal area, and frequency), saplings (density, basal area, and frequency), seedling (density and frequency), and herbaceous (cover and frequency). Measures of diversity were calculated from species IVs for each stratum by plot as described by Odum (1971). Mean elevation was calculated for each plot by averaging the elevation values for all surveyed points that fell within the plot's boundary; on average, there were four to five surveyed points per plot. In order to reduce the number of soil variables, weighted means were calculated for both depths using bulk density as the weighting factor.

Correlation coefficients among site factors (elevation, forest floor litter cover, and canopy cover), soil physical factors (bulk density, sand, silt, and clay), soil moisture, and soil chemical factors (pH, organic matter, electrical conductivity, total N, P, K, Ca, Mg, S, Fe, Mn, Zn, Cu, and Na) were computed using Pearson correlation analysis (SAS 1989).

Weighted means of soil variables, means for forest floor litter and canopy cover, IVs of vegetation data by species and stratum, diversity, richness, and evenness were analyzed by elevation class. Homogeneity of variances and normality of the data were tested using Levene's test (Levene 1960) and Shapiro-Wilks test, respectfully (SAS 1989). For this and other statistical tests, significance was accepted at α ≤ 0.05. Most of these tests indicated non-homogenous and non-normal data despite several transformations. As a result, the non-parametric Kruskal-Wallis test of group comparisons was used since it makes no assumptions about homogeneity and normality (Lehmann 1975).

Nonmetric multidimensional scaling was used to show the arrangement of sampling units and species. Vegetation data and environmental data were summarized by plot. After eliminating species occurring in

Results. ENVIRONMENTAL VARIABLES. Differences by elevation class. Twenty-four site and soil physical, moisture, and chemical variables were identified in the three elevation classes. Eighteen of these variables differed significantly by elevation class (Table 1). Values for canopy cover, litter cover, organic matter, pH, N, K, and S were greatest at high elevations. In contrast, maxima at low elevations occurred for bulk density, Mg, Fe, Na, Ca, and Zn. Electrical conductivity and P exhibited their lowest values in the mid-elevation class. May soil moisture content decreased with increases in elevation, but the driest soils in July were in the mid-elevation class. All three elevation classes were predominantly sand (49-57%), with 22-26% silt and 21-25% clay; textural classes were sandy clay loams and sandy loams. There was a trend for higher sand and lower silt and clay concentrations as elevation increased.