USGS Home
Contact USGS
Search USGS

Vegetation measurement

Baselines were marked with metal posts to aid in relocation. A tape measure stretched between end points of the baseline allowed us to resample the same plots in different years. Plant species nomenclature follows the Flora of the Great Plains (Great Plains Flora Association 1986). Our analyses were limited to a subset of species and groups of species that we regarded as potential indicators of change (Table 1).

Independent variables

Forty-seven elk were introduced into the South Unit of TRNP in 1985. Numbers increased to approximately 400 by 1993 (Theodore Roosevelt National Park, Unpubl. data). In January 1993, 200 elk were removed from the population. The estimate in 1996 was 300 to 400. Between 1985 and 1992, elk distribution in TRNP expanded from the introduction site to cover all areas of the South Unit that were inside the boundary game fence (approximately 95% of the South Unit). Potential duration of use by elk in stands we sampled at the time of the 1996 vegetation measurements varied from 0 to 12 years.

White-tailed and/or mule deer had access to all of our sample sites from 1985 to 1996. Fall surveys conducted by the North Dakota State Game and Fish Department indicated that mule deer numbers declined by more than 50% between 1985 and 1996 (North Dakota State Game and Fish Department, Unpubl. data). No data were available for white-tailed deer numbers in TRNP, but white-tailed deer were much less common than mule deer in areas where our stands were located.

Weather, especially precipitation and temperature, exert major influences on green ash stands (Girard et al. 1987). Temperatures during 1985 to 1996 were well within historic norms. The warmest and coolest summers of the monitoring period were 1988 and 1993, respectively (Fig. 2). Winter temperatures were generally milder than average with the mildest winter before the 1992 growing season and the coldest before the 1996 growing season. Precipitation during the growing season was greater than 20% below the 30-year average in 1984, 1985, 1988, 1989, and 1990 (Fig. 3). Precipitation was greater than 20% above average in 1986, 1987, and 1993. We included 11 variables related to weather (Table 2) in our analyses. Weather data were obtained from a weather station maintained by TRNP in the South Unit. We attempted to measure all stands before herbaceous plants became desiccated, but in some years we were forced to measure one or more stands after the optimal mid-July to mid-August sampling period. We included month of vegetation measurement as an index to phenology to determine if deviations from our optimal sampling period had detectable impacts on values.

Figure 2.  Mean temperatures for the summer (June - August) and winter (November - March) recorded at the South Unit of Theodore Roosevelt National Park, 1984 to 1996.

Figure 3.  Precipitation during the growing season (March - September) and winter (November - March) recorded at the South Unit of Theodore Roosevelt National Park, 1984 to 1996. Means for 1950 to 1996 are displayed as dashed lines in the figure.

Analysis

We could not do a simple test of vegetation measurements in 1985 versus vegetation measurements in 1996 to identify changes that might be associated with elk impacts on green ash stands because the opportunity for elk to impact green ash communities varied among stands. Elk distribution within TRNP expanded from an area that included four of our sample sites in 1985 to 11 of the 12 sites by 1996. We hypothesized that changes in vegetation and ground cover associated with elk use of stands should increase with the number of years elk had access to a stand. We developed 18 vegetation and ground cover categories from our vegetation data that, based on food habits studies in TRNP (Sullivan 1988, Westfall 1989) and impacts on vegetation reported in other areas (Lyon and Ward 1982, Singer and Harter 1996, Singer et al. 1998, Wambolt 1998, White et al. 1998), we expected to decline following intensive use of a site by elk. Differences between values recorded for these categories in 1985 and 1996 were converted to percent change (V₁₉₉₆ - V₁₉₈₅ /V₁₉₈₅ x 100) and truncated to a range of -100% to +101% to normalize the data and decrease the influence of large proportionate changes in low absolute cover or density measurements. Truncation affected only 11% of the vegetation measurements but reduced the average skew of variables by 53% to -0.38 and the average kurtosis by 21% to 3.69. Truncated percent change between 1985 and 1996 was tested using multiple t-tests on categories paired across sites. If elk did have impacts on green ash stands in TRNP, we expected to see significant differences in mean percent change among a set of stands with zero to 12 years of elk use, and we expected stands with the most years of elk use to have lower average values (i.e. greater negative changes between 1985 and 1996) than stands with the fewest years of elk use.

The changes in vegetation we measured between the 1985 and 1996 sampling periods could also reflect the impacts of short-term (one month to two years) rather than long-term factors. We used two approaches to determine if recent weather patterns or recent ungulate use influenced the changes we observed from one sampling period to another. The strength of association between individual explanatory variables (weather indices and percent ground coverage of elk and deer fecal pellets in stands) and response variables (truncated percent change between measurements in vegetation/ground coverage categories taken at two-year intervals between 1986 and 1996) was determined using Pearson's correlations. The second approach, multivariate analysis, was an attempt to develop a more realistic model for explaining changes in vegetation. Changes in vegetation are highly variable, usually not linear, seldom occur at consistent rates among different species, and frequently involve responses to multiple causative factors. Of the multivariate analytical methods available, we selected logistic regression as the most appropriate for our data based on guidelines presented by Morrison et al. (1998). We used logistic regression to determine if combinations of weather variables and ungulate fecal pellet indices would be useful in predicting the direction of change between sampling periods. Percent changes in vegetation categories between sampling periods were converted to class variables (1 = no change or increase, 2 = decline), and a step-up procedure (P < 0.05 to enter) was applied to identify the variable sets that could best distinguish between positive and negative changes. The MSUSTAT package (Lund 1989) was used for multiple t-tests and correlation analysis. SAS (1994) was used in the logistic regression analysis.