MEASURING CHANGES IN OAK SAVANNAS: A REVIEW AND
RECOMMENDATIONS FOR A MONITORING PROTOCOL
Alan Haney, Professor of Forestry
Steven I. Apfelbaum, Chief Research Ecologist
Monitoring involves measurement of ecological characteristics that can be used to evaluate the status or change of one ecosystem relative to another, or at one time relative to another. In this paper, we address monitoring of savannas, although the protocol suggested is applicable to a wide variety of ecosystem types, including prairies.
Questions that should be addressed in developing a monitoring protocol include (Kent and Coker 1992):
These questions suggest a need for a standardized protocol that will provide comparable data from similar ecosystems yet be flexible enough to accommodate local variations in ecosystems, monitoring resources, and knowledge of field workers. This paper lays out a protocol we have used for over ten years (see Haney and Apfelbaum 1990). It incorporates methods described by Lindsay (1956), the Wisconsin Natural Areas Program (unpublished report of the Bureau of Endangered Resources, Wisconsin Department of Natural Resources), and the U.S. Army Land Condition-Trend Analysis (Tazik et al. 1991).
The protocol consists of four parts: 1) identifying the ecosystem to be monitored, 2) establishing random points, 3) sampling and documenting, and 4) data summary and analysis.
Most monitoring is done at the community level. Landscape-scale approaches, however, should be used in management of ecosystems. Savannas are typically interspersed with forests, prairies, and wetlands. The connections between community types is important to maintain diversity and other ecological processes associated with savannas (Haney and Apfelbaum, in press). The landscape on which monitoring is focused, therefore, is usually a matter of ownership, management, disturbance history, and barriers. Large areas are more diverse (Cain 1938), and therefore, require more time and effort to adequately sample. It is better to sample smaller areas adequately than large units superficially. In reaching a decision about the landscape to be included, it is useful to gather all pertinent information available, including aerial photographs, evidence of disturbances, topographic maps, and land survey notes. Generally, a monitoring unit should contain communities with similar disturbance histories.
Once a decision is made about the landscape to be included in the monitoring unit, a sampling plan must be developed. In nearly all instances, it is necessary to extrapolate the sample data to the entire community, or even to all comparable communities within that landscape. If the landscape contains more than one community type, sampling should be stratified such that each community type of interest is studied separately. Data from each community type can then be extrapolated to other areas of that community type within the sampled landscape. In order to have confidence in the extrapolation, sampling points must be established such that there is equal opportunity for each area within each community to be included in a sample.
Use of a geographic information system (GIS) will help not only in the sampling design, but also later as vegetation is mapped. First order data bases can be developed by digitizing cover from aerial photographs, soil surveys, available from your local Soil Conservation Service (SCS), or plats showing ownership. Each of these data bases eventually should be incorporated into landscape maps. Topographic, hydrologic, and various cultural information contained on U.S. Geological Survey maps should be included.
A map that overlays cover type with ownership or other cultural information is sufficient to develop random sampling points. GIS software allows the identification of study units. This is usually done by cover type (one type of savanna, for example). Within the designated areas, GIS software will create as many random points as specified and list the latitude and longitude of each. The number of points should be determined by variation in the vegetation and resources for sampling: the greater the variation, the larger the number of points needed for an adequate sample. A buffer zone to remove edge effects also can be specified.
In the absence of a GIS system, points can be located by use of a dot grid overlaid on a recent aerial photograph. By numbering the vertical rows and the horizontal columns of dots, you can use a random number table from any introductory statistics book to specify a column and a row, thereby selecting a random point. In selecting points, identify the cover type over which they occur, thereby stratifying the sample design. This approach suffers from difficulty in locating random points in the field.
The quickest way to locate random points in the field is with use of Global Positioning Satellites (GPS). Without a reference base, accuracy is usually only plus or minus 20 to 30 meters. As long as no bias is introduced into location of each point, and points can be relocated, this error is not a concern.
Without GPS, points must be located by surveying. Usually a tape and a good compass are sufficient, even for location of random points, but this can be quite time consuming. In the absence of a GPS, a systematic approach will save time both in initial establishment and relocation of points. A systematic approach usually is more efficient and meets most statistical assumptions; at most, there are only minor statistical compromises.
In applying a systematic approach, random points are selected from either a horizontal or a vertical scale overlaying a map or aerial photographs of the landscape. Ideally, the scale should follow a property boundary, or road that is well marked. From these points along the edge of the landscape, determine the compass bearings that lead across the area. Sampling can be at regularly established intervals along these bearings or at random points located using a random numbers table.
It is important that each point be established so that it can be relocated, often after fire or successional changes have altered the landscape, and often by someone other than those who established the point. We suggest marking points with a metal pipe driven 0.5 to 1.0 m into the ground. Leave the top few centimeters above the ground and paint it a bright color to facilitate relocation. With GPS, or careful surveying, and a metal detector, we have seldom failed to relocate a point. Use of witness trees can also help, along with other landscape features, such as large rocks, streams, or prominent hills.
At each point, a 50 m line is established with a measuring tape, along a compass bearing. The bearing can be randomly determined, but we use a systematic bearing, usually a cardinal direction. The tape should be stretched tight at ground level. A second pipe is used to mark the end of the 50 m line. Tree (woody plants 5 cm or more diameter breast height (DBH)) cover is recorded by species by line intercept along the tape, or by point sampling at one or one-half meter intervals. Shrub (woody plants over 1.0 m tall and less than 5 cm DBH) cover is recorded in the same way. Tree size distribution is determined by recording all trees rooted within 1.0 m of the line by species and diameter, and indicating whether they are alive or dead. Shrub density is measured by tallying stems by species rooted within 1 m. of the right side of the line. We use circular 1 m2 quadrats to estimate cover by species in the herbaceous layer (all living vegetation less than 1.0 m tall). The circular quadrat frame is centered on the line at 10 m intervals, beginning at the 5 m mark. In addition to estimating cover by each species, we also estimate per cent cover of fine litter, coarse litter (woody stems greater than 2.5 cm in diameter), and cover of exposed mineral substrate. The circular frame can also be used to record or plot species of special interest.
A composite soil sample is taken from along the 50 m. line. The slope and aspect of the line is noted, and photographs are taken from each end looking along the transect. A thorough search is made within 50 m. of the line for any additional species. Sampling for various faunal groups, such as birds or invertebrates can also be conducted in the vicinity of each line.
Estimated parameters can be summarized for each community type using means (averages) and standard deviations. These summaries provide ground truth to refine original interpretations of aerial photos. Summaries can also be used to compare two or more communities, or the same area at different times. Ordination also can be used to analyze data. The use of abiotic parameters such as soil characteristics, disturbance history, slope, and aspect allows both an environmental ordination and a vegetation ordination (Austin 1968). ter Braak (1988) developed an excellent software that simultaneously computes both. For an discussion of ordination, see ter Braak and Prentice (1988).
Austin, M. P. 1968. An ordination study of a chalk-grassland community. Journal of Ecology 56:739-757.
Cain, S. A. 1938. The species-area curve. American Midland Naturalist 19:573-581.
D'Erichia, F. Geographic information systems and remote sensing applications for ecosystem management. In: A. Haney and M. Boyce, eds. Ecosystem Management: Application for Sustainable Forest and Wildlife Resources. Yale University Press, New Haven, Connecticut. In press.
Haney, A., and S. I. Apfelbaum. 1990. Structure and dynamics of Midwest oak savannas. In: J.M. Sweeney, ed. Management of Dynamic Ecosystems. North Central Section, The Wildlife Society, West Lafayette, Indiana.
Kent, M., and P. Coker. 1992. Vegetation Description and Analysis. CRC Press, Ann Arbor, Michigan. 363 p.
Lindsay, A. A. 1956. Sampling methods and community attributes in forest ecology. Forest Science 2:287-296.
Tazik, D. J., S. D. Warren, V. E. Diersing, R. B. Shaw, R. J. Brozka, C. F. Bagley, and W. R. Whitworth. 1991. U.S. Army land conditions-trend analysis: Field methods. (Draft). Unpublished report.
ter Braak, C. J. F. 1988. CANOCO - an extension of DECORANA to analyze species-environment relationships. Vegetation 159-160.