This chapter addresses Objective 5 -- Design field experiments based on data from objectives 1- 4 that could assess the impact of wildlife grazing on water considering, (a) seasonal and diurnal movements, (b) presence and absence of livestock grazing, (c) various riparian habitat types present in Wyoming -- and objective 6 -- Recommend potential study sites in different riparian habitats that could be used to carry out field experiments. The purpose of this section is to describe and evaluate potential study sites throughout Wyoming. As stated in Objective 5, the design of field experiments is based on information gathered for Objectives 1-4, as well as consideration of the following (Objective 5):
Wild ungulates will utilize different elevations and habitat types during the year. Animals generally migrate to lower elevations during winter in search of food and thermal cover. Many of these natural and man-induced wintering areas have high concentrations of animals which are located in valley bottoms and basins with perennial streams. The possibility of degradation to riparian habitat and water quality in this scenario is high. Animals will generally disperse and migrate to higher elevations during late spring and early summer. Riparian habitat may not used as intensively during this time of year.
Information considered when evaluating potential study sites included:
Personnel from the Wyoming Game and Fish Department, U.S Bureau of Land Management, U.S. Forest Service, National Park Service, University of Wyoming, and the U.S Fish and Wildlife Service were contacted to discuss potential study sites. Criteria needed for consideration as a potential field study site or experimental site included: (1) site must be at least 5 acres in size, and (2) site must have minimal or no livestock use. Visits were made to identified sites with agency personnel. Notes were taken from observation and discussion about the general landscape, species of wildlife that utilize the site, seasons of use by wildlife, approximate numbers of ungulates that utilize the site, grazing history of the site, types of land treatments implemented on the site, vegetative composition, and types of wildlife impacts occurring on the site. Photographs of the sites were taken and can be used as references for future planning.
Photographs and notes regarding location, accessibility, species of wild ungulates on site, season of use, livestock grazing history, wetland classification, vegetative structure, rangeland improvements on site, and observed impacts were taken at each of the thirty sites visited. A brief description is given for each site visited (Appendices I-XXX). Each appendix is listed in alphabetical order by site by county. Each site was evaluated and given a ranking as a field study site and experimental site (Table 1). Table 1 lists each potential study site in alphabetical order by site and county and gives it a ranking. The ranking system used is as follows: 0 = no potential due to several limiting factors, such as, degraded riparian habitat and water quality due to flooding, drought, natural erosion, past livestock grazing impacts and other factors, such as, poor accessibility, limited size, topography, fencing, and human activity; 1 = some potential, but some limiting factors mentioned above may generate uncontrolled variability; 2 = high potential, none or very few limiting factors exist.
DESIGN OF FIELD EXPERIMENTS
Determination of the relations between livestock and wildlife grazing and water quality can be approached using two different experimental designs: (1) comparison of multiple sites with differing levels of wildlife and livestock grazing, or (2) controlled experiments with known kinds and numbers of animals held in confined areas. The basic experimental design, problems, and feasibility of both approaches are discussed.
Comparison of Multiple Sites
This approach would involve field study of numerous sites. The design assumes a functional relationship between livestock and/or wildlife grazing and water quality. Measures of livestock and/or wildlife grazing intensity are evaluated for their relation to measures of water quality among several study sites.
Linear-regression and multiple-regression analyses are the statistical methods for determining if statistically significant relationships may exist between grazing intensity (independent variable) and water quality (dependent variable). The null hypothesis is that no relation exists between independent variables) and the dependent variable. Regression analysis may indicate relationship, but it does not confirm cause and effect or define physical/biological mechanisms causing the observed relationship. Additionally, regression analysis is based on four assumptions (Sokal and Rohlf 1981):
(1) The independent variable (grazing intensity) is measured without error,
(2) the relation of the dependent (water quality) and independent variable is a linear function (straight line),
(3) for any given value of the independent variable, the measurements of the dependent variable are normally distributed, and
(4) the variance of both the independent and dependent variables is independent of their magnitude.
In order for a regression analysis approach to be used, several sites with similar climatic, geomorphic, geologic, and vegetative features would have to be found. If the influence of wildlife grazing on water quality were being examined, an accurate measure of the magnitude of grazing would be needed at each site (Assumption 1 above).
There are numerous problems associated with a regression analysis design for determining the relation between grazing and water quality:
(1) Identification of several sites with similar features is extremely difficult. For example, if the focus was to assess water quality in second order streams within a riparian area dominated by willows, then study sites with many similarities would be needed. The sites would have to have similar climates (elevation, precipitation, etc.) and be in drainages of similar sizes and with similar geomorphic features. Additionally, the history of management and use would have to be similar.
(2) The assumption that the independent variable (magnitude of grazing) is measured without error would be violated. Accurate estimates of wildlife use (numbers, forage utilization, season of use, etc.) are difficult to obtain with extensive amounts of work.
(3) Other uses of the various drainage areas besides wildlife and livestock grazing are difficult to measure and are likely to confound a regression approach. Much unaccounted for variability in water quality is likely to occur due to other anthropogenic activities such as construction of roads and buildings, timber harvest, water development, or past histories of mining, logging, and livestock grazing.
Comparison of several sites to determine relations between grazing and water quality is not a reasonable approach to research. Sufficient numbers of suitable sites will be very difficult to locate. The most severe limitation is the ability to measure the past and present grazing intensity in potential study areas. Determination of the magnitude of grazing by wildlife cannot be measured without error, the first assumption necessary for regression analysis.
This approach would involve large-scale experiments at specific study sites. The basic experimental design would entail measurement of the water quality-under known, controlled levels of grazing. While simple in statistical design, this approach would provide numerous logistic hurdles due to the magnitude of an experiment site.
Two statistical approaches could be used to design and analyze data from controlled experiments: (1) linear/multiple regression, or (2) analysis of variance. Both designs have unique advantages and disadvantages.
Within this design, several know levels of grazing intensity (independent variable) could be assessed for their relation to a measure of water quality. For example, 10 identical pastures could be stocked each with a different number of mule deer. The suspended solids in runoff from each pasture could be measured. Linear-regression analysis could be used to determine if the density of mule deer was related to suspended solids in runoff.
Because this is a controlled experiment, the assumptions required for regression analysis can be met through the research design. For example, the assumption that the independent variable (magnitude of grazing) is measured without error can be met by the researchers' manipulation of the number of mule deer in each pasture.
Analysis of Variance
This design also entails controlled manipulation of a single variable, such as grazing intensity. Within this design specific treatments (magnitude of grazing) are defined and the effect of the treatments on a variable (water quality) are assessed with replicates of each treatment.
A example of a simple experiment might entail the influence of mule deer grazing on suspended sediment in runoff from upland prairie. To conduct the experiment a gently sloping grassland site is selected and nine 1-acre pastures are fenced to hold the deer. There levels of grazing will be evaluated -- 0,5, and 10 mule deer per acre for the same 3-month summer period -- with three replicates of each. Surface runoff will be collected during storm events within sumps at the lowest elevation point within each pasture and suspended sediment in the water will be determined.
Analysis of variance is the statistical method for testing if the null hypothesis should be rejected. In the example, the null hypothesis is that there is no difference in suspended sediment among the three levels of grazing intensity. The example is a one-way analysis of variance, but more complex designs, such as two-way analysis of variance could be used.
A fundamental set of assumptions for analysis of variance includes: (1) Sampling of individuals is random (For example, if three levels of grazing by mule deer were to be measured on three pastures each, the three levels of grazing would be randomly allocated among nine pastures.), (2) The error term of each expected value of a variate is a random normal variable, and (3) The error terms have identical variance. Within controlled experiments these assumptions can be met in most cases; therefore, the statistical approach is valid in this situation.
Controlled experiments do not pose the statistical problems that are encountered in comparison among multiple sites using regression analysis. While the experimental design is sound, the physical construction of an experimental facility poses many problems. Fencing, water collection and water quality monitoring devices, and other components of the physical facility are expensive. Siting of a facility is difficult because an area of sufficient size with similar topography, vegetation, and grazing history may be difficult to find. The actual experiment will likely alter vegetation and soil features at a site so that repeated use following completion of the initial experiment may not be feasible.
Controlled experiments will require long periods of time to complete. It is likely that several years will be required to complete an experiment at a particular site because treatment effects are likely not to be observed until vegetation changes occur.
Numerous problems in obtaining and holding large mammals will be encountered. Permits will have to be obtained from the state game and fish agency.
Because the animals are held in confinement an unnatural situation is created. Behaviors of animals associated with seasonal movements, reproduction, diurnal variation in habitat use, and seasonal availability or use of specific forage plants can be simulated, but the complexity of the experimental design is increased. The limited mobility of the animals is likely to have effects that might not be seen with free-ranging animals.
A particular experiment will be limited to a single species in a particular habitat type, such as elk in mountain meadows or mule deer in upland sagebrush habitat. Numerous experiments will be needed to ascertain the effects of all the large mammal species on the array of different habitat types found in the state.
The limitations associated with finding suitable sites, the cost of establishing a facility, and the availability of test animals, will limit both the number of treatments and replicates that can be managed in any given experiment. Simple experimental designs will have to be used as a result.
The use of controlled experiments will be very costly and time consuming. It is estimated that a single experiment on one species in one habitat type will cost in excess of $500,000 and require more than 4 years to complete.
The use of controlled experiments seems to be the only scientifically sound approach to determining the relations between livestock and wildlife grazing and water quality. Such experiments cannot be conducted with ease because they will require extensive funding and long time periods to complete, but they are the only feasible approach to obtaining factual information. Controlled experiments can be manipulated to simulate seasonal and diurnal movements in presence of livestock, but the complexity of the work is increased. Separate experiments will have to be conducted on various riparian habitat types to assess the effects of wildlife in each.
POSSIBLE STUDY SITES
Only seven field study sites were identified that had high potential. Each site is listed in order of greatest potential based on criteria used to evaluate each site: (1) Flat Creek/National Elk Refuge, (2) Horse Creek, (3) Beaver Creek, (4) Fence Creek, (5) Blacktail Creek, (6) Wagonhound Creek, and (7) Camp Creek.
The potential for using a multiple-site, regression analysis approach is quite limited because only seven sites were identified. This is probably not a sufficient sample size to allow statistical inference to be obtained.
Only five experimental study sites were identified that had high potential. Each site is listed in order of greatest potential based on criteria to evaluate each site and relative pristine condition of the site: (1) Torrey Creek, (2) Labonte Creek, (3) Johnson Creek, (4) Bear Creek, and (5) Green River/Seedskadee National Wildlife Refuge. All of these sites have good potential for controlled experiments.
92-02 Table of Contents
Water Resources Publications List
Water Resources Data System Library | Water Resources Data System Homepage