Restoration in the Colorado Desert: Management Notes

Methods for Plant Sampling

Prepared for the California Department of Transportation
District 11, 2829 Juan Street, San Diego, CA, 92138
as part of the Desert Revegetation Project
October 1993 
Matthew W. Fidelibus and Robert T.F. Mac Aller
Biology Department
San Diego State University
San Diego, CA 92182 

Introduction

The goal of most rstoation and revegetation projects is to recreate the plant cover, distribution, and species compostion of the site prior to disturbance, or of a comparable less disturbed reference site. Accurate data on community compostion is desirable for the planning and evaluation of these projects. While it is impractical to take a complete census of even a relatively small site; cover, density, and frequency of plant species can ve accurately estimated from as little as 1% of the community (Barbour et al., 1987).

The selection of sample site can be based typical sites (releve), random samples, systematic samples in a regular pattern, or by a combination of random and systematic selection (Greig-Smith, 1983). Although the releve method uses subjective choice of sample locations, the process of recording data is relatively rapid and nonmathmatical. Systematic and random methods, which are commonly used in the United States, are more conductive to statistical analysis. In the field, random sampling may be much less convenient than systematic sampling, but the regular sampling of a population showing periodic variation would not bvbe representative of a population as a whole (Eberhardt and Thomas, 1991). The selection of an appropriate sampling technique depends upon the type of data needed, the size of the sampling site and the number of available workers.

Quadrat Sampling

Sampling with quadrats (plots of a standard size) can be used for most plant communities (Cox, 1990). A quadrat delimits an area in which vegetation cover can be estimated, plants counted, or species listed. Quadrats can be established randomly, regularly, or subjectively witin a study site. Since plants often grow in clumps, long, narrow plots often include more species than square or round plots of equal area; especially if the long axis is established parallel to environmental gradients (Cox, 1990; Barbour et al., 1987;Greg-Smith, 1983). However, accuracy may decline as the plot lengthens because, as the perimeter increases, the surveyor must make more subjective decisions about the placement of plants inside or outside the plot. Round quadrats can be most accurate because they have the smallest perimeter for a given area. Round quadrats are also simple to define in the field, requiring only a center stake and a tape measure (Cox, 1990).

The appropriate size for a quadrat depends on the items to be measured. If cover is the only factor being measured, size is relatively unimportant. If plant numbers per unit area are to be measure, then quadra size is critical. A plot size should be large enough to include significant numbers of individuals, but small enough so that plants can be separated, counted and measured without duplication or omission of individuals (Cox, 1990; Barbour et al., 1987). Large quadrats with many plants may require two or more people to obtain an accurate census, while one person may be sufficient for smaller plots or those with sparse vegetation.

An accurate estimate of the necessary number of quadrats can be determined by plotting data for a given feature (i.e. percent cover) vs. number of quadrats. The appropriate quadrat number will correspond to the point at which the curve plateaus (Figure 1) Barbour et al., 1987). Some field researchers sample until the standard error of the quadrat is within a previously decided, acceptable boundary. A standard error of + or - 15-20% of the mean (i.e. two thirds of all quadrats supply data that fall within this range about the mean) is sometimes used (Barbour et al., 1987)

Cover, density, and frequency are important aspects of the plant community which can be measured by quadrat sampling. Cover is the percentage of quadrat area beneath the canopy of a given species, FOr the practical estimation of cover, holes in the canopy can be thought of as nonexistant, and the canopy is mentally "rounded out" (Barbour et al., 1987). Plants rooted outside the quadrat are included in cover measurements to the extent that their canopy projects into the quadrat space (Barbour et al., 1987). It is sometimes difficult to accurately estimate cover, especially if the plants are at or above eye level. In these cases, an aerial photo often proves useful. Canopy overlap can further complicate cover measurements. Overlap of the same species should not be counted twice, but recorded as continous cover between two or more plants. If two or more plant spcies overlap, the cover of each should be tallied independently (Barbour et al., 1987). If ther are many ovelapping canopies, it is possible to estimate more than 100% cover and still have open ground.

Density is determined by the number of plants rooted within each quadrat. Relative density is the density of one species as a percent of total plant density. Area per plant, or mean area, is plot area per density.

Frequency is the percentage of total quadrats containing at least one rooted individual of a given species. Relative frequency of one species as a percentage of total plant frequency. Frequency is affected by quadrat size and my ble less meaningful than other measurements.

Relev»

In the relev», or "sample stand" method, a person knowledgeable with a region's vegetation develops concepts about certain community types that are repeated in similar habitats and then chooses several representative stands for a community (Barbour et al., 1987). The community is name based on the most abundant species composition (i.e. "Larrea scrub"). The surveyor then walks though each stand recording all of the species encountered and describing the habitat and soil profile. The stand that best represents the community vegetation and soil profile is selected. In this stand, data from a series of nested quadrats is plotted on a species vs. area curve (Figure 2) to determine the smallest area within which the species of the community are adequately represented (the minimal area).

The presence of addtional species in each larger quadra is recorded. A point is reached where increasing quadrat size does not significantly increase the numer of species encountered. The minimal sample area can then be determined from the species/area curve where the slope is nearly horizontal. This resulting quadra area is called a relev» (Barbour et al., 1987). The minimal area is thought to be an important community trait that is just as characteristic as the species that make it up.

In the relev» method, cover is measure as a category (a number between zero and seven denoting 0-100% cover, respectively) rather than a precise number (Barbour et al., 1987). An exact estimate of percent cover is thought to give a false sense of precision and cover estimates from mulitple observers rarely agree. Although some precision is lost, categorical classification has good repeatability.

The relev» method is quick, nonmathimatical, and should detect nearly all plant species in a given community. Unfortunately, it is sometimes difficult to characterize a community. This method may be most efficient and useful for large scale ecological restoration projects, provided the biologists performing the initial analysis are sufficiently knowledgeable with the regions vegetation.

Plot-less Sampling

Plot-less sampling has partly replaced quadrat sampling in North American studies. In this technique, communities are not sampled with delimited plots, but with sampling points, one dimensional transect lines, or certain distances within the stand (Knapp, 1984). Plot-less methods could be thought of as quadrats shrunk to a line or a point of no dimension (Barbour et al., 1987). The advantages of plot-less sampling are: 1) a sample plot does not need to be established, saving time and 2) elimination of subjective error associated with the sample plot bounderies.

The point method determines the number of points, distributed randomly or regularly in the survey area, where parts of a plant are above ground. The points can be established with a regular grid, randomly chosen coordiante pairs, or regular or random points along a meter tape (transect line). For coordiante pairs and regular grids, an x and y axis are established along the edges of the study area. Random points are selected using a random number generator (on most calculators) or from a random number table. Points can also be established using stratified random sampling. In this method, the area to be sampled is divided into sections with and equal number of random points in each, insuring an adequate distribution of sample points throughout the site. A tape measure and a compass are used to locate the ordered pairs (random or regular) designating field sample points. With transect lines, only a single baseline is required. A series of points along this baseline are selected using a random, stratified random, or systematic procedure (Cox, 1990). These points serve as starting points for transects throughout the area. With a sufficient number of points, an exact measurement of percent cover is possible (Knapp, 1984).

The number of points necessary for an adequate assessment is partly dependent on the species cover within a survey area. A graph depicting percent cover vs. number of points can help determine the number of necessary points. With only a few points, the curvatures in these graphs oscillate highly. As the points approach a sufficient number, the curvatures become flatter, only fluctuating slightly around the average percent cover (Knapp, 1984). If a plant provides > 20% cove, a smallnumber of points is needed. A much greater number of points is needed for assessing percent cover for species occupying less than 5% of the study area (Knapp, 1984).

In addition to point methods, one dimensional transect lines can also be used to determine cover. The line intercept methods is just accurate as the quadrat method, but much less time consuming. In this technique, only plants which cross the imaginary veritcal plane of the transect line are recorded. Two measurements should be taken for each plant: the length of intercept (I) and the maximum width of the plant perpendicular to the transect line (M). The amount of bare gournd within each transect segment should also be measured and recorded in the same manner (Cox, 1990). In the case of communities with two or more distinct strata (such as herb, shrub, and tree strat in a forest) it is usually necesarry to sample each level separately.

Cover is calculated as the percent of transect line covered by each species. The chance of an indivual being encountered by the transect line is proportional to its width perpendicular to the transect line, so density can be calculated by multiplying the total recipricals of maximum plant width by the unit area/total transect length (Cox, 1990).

Distance Methods

Distance methods measure the distance from a sampling point (or plant) to the nearest plant or nth nearest plant. The results of such a technique cna provide important information about the relationships between plants. Distance methods can help determine whether plants are growing in discernible (and often ecologically important) patterns or are randomly dispersed. Many inter and intra-specific plant relationships are difficult to observe without using distance based sampling techniques.

With the nearest individual method, random points are located in a stand, the distance from each sampling point to the nearest plant is measured, the species is identified, and its basal area is measured. Only one measurement is made from each random point, and all distances for all species are summed and divided to yield one average distance (Barbour, 1987).

In the nearest neighbor method, random points are located in a stand and the nearest plant is located. The distance from this individual to its nearest neighbor is measured. Density is calculated as in the nearest individual technique. An advantage of the nearest neighbor method is that it can be used to determine whether plants of the same species are distributed randomly, regularly, or are clumped (Barbour et al., 1987).

In the random pairs method, a line is taken from a point along a transect to the nearest plant. Perpendicular to the line and passing through the point is an exclusion line. The distance from this individual to the nearest neighbor that is on the same side of the exclusion line is measured. A difficulty in using these methods is that if the density is low, it may be impractical to search beyond a certain distance for the nearest individual.

Sampling With Photographs

Traditional sampling methods may be too labor intensive to provide accurate information over large areas. Aerial photography using large scale (1:200) color or infra-red photographs are useful for mapping and recording individual plant species in a range of vegetation types (Hacker et al., 1990). Acceptable estimates of plant cover can be made, and the condition of the soil surface is clearly recognizable, particularly on color infra-red film. Neither film type permits accurate identification of all species and the presence of understory plants may be obscyred by foliage or shadow. Furthermore, the best interpretaton, photographs need to be taken within four to six weeks of effective rain. Unfortunately, the difficulty and expense of establishing permanently marked flight line, and acquiring and printing the photographs, precludes using the technique for some projects.

Helium filled balloons could be a less expensive method of aerial photography. A ground anchored helium balloon could support a camera to photograph a quadrat or permanet plot. Grids placed over the photographs could simplify percent cover estimates. Balloons could also be used with small video cameras to record "video-tansect" lines. Wind patterns must be considered with such data recording techniques. There are obvious difficulties in establishing transect lines against prevailing winds or photographing in high wind areas.

Photgraphic records of permanet experimental sites, or photo points, can be a simple, rapid, and cost effective laternative to aerial photos. Photo points are obtained with a hand held camera from an elevated position, such as the roof of a vehicle (Hacker et al., 1990).

The Western Austrlian Rangeland Monitoring System (WARMS) was developed to integrate the most desirable features of the photo print method with soil and vegetation data. WARMS consists of a photographed area or "photoplot", and a series of fixed "belt" tansects within which shrubs are recorded by species and canopy width and height (Hacker et al., 1990). The belt transects are divided into blocks to allow for a precise estimate of community composition. The density of shrub seedlings, herbs, and grass species, is scored on an interval or "category" scale (as a releve). Plant counts within the photographed area supplement transect data. In areas of low shrub density, the photographs can provide sufficient assessment of change and additional transect data is not needed. Quadrats established in regular intervals between transects are used to assess soil surface condition.

Summary

Quantitative information on the structure of a plant community is desirable for planning, and evaluating the success of restoration and revengetation projects. Traditional plant sampling methods such as quadra sampling, plot-less methods, and distance methods can provide accurate estimates of cover, density, and frequency. However, the extensive labor and preparation needed may make these techniques poorly suited for characterizing large mitigation sites. "Completely randomized sampling will inevitably under sample rare but interesting and ecologically informative kinds of vegetation" (Barbour et al., 1987). The appropriate method of community sampling is dependent upon the project. The terrain investigated, the research goals, and available capital must be considered prior to choosing a sampling technique.

References

Barbour, M.G., J.H. Burk, and W.D. Pitts. Terrestrial Plant Ecology. Chapter 9: Method of sampling the plant community. Menlo Park, CA: Benjamin/Cummings Publishing Co.; 1987.

Cox, G. Laboratory manual. of general ecology 6th Ed. Dubuque, Iowa: WIlliam C. Brown; 1990.

Eberhardt, L.L. , and J.M. Thomas. Designing Environmental Field Studies. Ecological Monographs. 1991; 61(1): 53-73.

Greig-Smith, P. Quantitative Plant Ecology. London: Butterworths; 1964.

Hacker, R., D. Beurle, and G. Gardiner. Monitoring Wester Australia's rangelands. Journal of Agriculture. Western Australia Department of Agriculture. No.1, 1990.

Knapp, R. Ed. Sampling methods and taxon analysis in vegetation science. Handbook of Vegetation Science 1. Part 4. Hague: Junk; 1984.

Equations

 

Quadrat Sampling Density = # of individuals / area sampled

Relative Density = species density / total density for all species x 100

Frequency = # of quadrats in which species occur / total # of quadrats sampled

Relative Frequency = species frequency / total of frequency values for all species x 100

 

Point Sampling

Cover = number of points covered by a species / total number of points x 100

 

Line Intercept Sampling

Cover = sum of intercept lengths for a species / total length of transect x 100

Density = (¬ 1/M)(unit area / total transect length) where M = maximum width of plant perpendicular to the transect line

Relative Density = species density / total density for all species x 100

Distance Methods

Density = square meters / 2 * (average distance in meters)^2

Relative Density of a Species = # of individuals A encountered / # of all species encountered * (density for all species)