Harold Nagel, Charles Bicak, Linda Spessard-Schueth, Steven Rothenberger, Marvin Williams, Margaret Biddlecome, Jennifer Crawford, Joseph Osterhaus and Troy Walz, Biology Department, University of Nebraska at Kearney
ABSTRACT
A database was developed of native rangeland sites in the Central Platte Natural Resources District. The data collected also served as the basis for developing and testing rangeland health or condition indices.
Species composition was determined on 279 prairies/rangelands during 1993-94. Most of these were located in Buffalo and Dawson Counties, and approximately 90% were upland sites. An ecologically based approach was used in classifying vegetation on the sites, rather than a production approach since the objective was to provide information on biodiversity.
Reference site comparison and clustering methods were used to group, classify and determine the health of the sites. Reference sites were selected which had little or no history of over-utilization and included cemeteries, hay meadows, and lightly grazed pastures. Lowland (located in the Platte River flood plain) and upland reference sites differed very little. Analysis of upland reference sites indicated the potential vegetation was mostly big and little bluestem. Lowland sites were also dominated by big bluestem but had more Indiangrass and switchgrass than upland sites.
Smooth bromegrass and Kentucky bluegrass were prevalent on 40% of the sites. In comparison with the reference sites, most of the upland sites rated poor to fair, averaging only 32% condition. This contrasts with the traditional range condition (SCS) which averaged 54% or low good condition. Two other ecological health indices are presented. One is based upon two threshold vegetation conditions found in the prairie sites and the second one is a modification of the reference site method.
In addition to loss of the tallgrass component, most native rangelands surveyed also had lost much of the native forb component, especially legumes, when compared with the reference sites. Smooth bromegrass and Kentucky bluegrass are serious invaders into central Nebraska uplandprairie rangeland.
INTRODUCTION
The two major objectives of this research were: 1.) to collect baseline data on prairies in the Central Platte Natural Resources District (CPNRD). These data can be used to establish trends in rangeland health through time, and 2.) to develop new methods for determining rangeland health or condition. New methodology is needed to replace the old range condition index, which was primarily developed and used by the Natural Resource Conservation Service (NRCS, previously SCS).
Few quantitative surveys have been published on Central Nebraska prairies. Weaver and Bruner (1954) did extensive floristic and community work in this area but did not identify the locations of prairies they sampled. Without that information, follow-up studies could not be done and rangeland trend couldn' t be documented.
The concept of range condition developed by Dyksterhuis (1949) and based on the concepts of Clements (1916) regarding succession of prairies to climax condition has been used for several decades by several federal agencies in managing private and public rangeland. This method of range condition (referred to as SCS method hereafter) was rejected recently for philosophical and practical reasons (Busby 1994, Joyce 1993). Friedel (1991) and Laycock (1991) have discussed the concept of thresholds in rangelands:
"The concept of thresholds of environmental change appears to provide a reasonable alternative in some circumstances to the concepts of gradual retrogression and secondary succession which are currently accepted. I suggest that environmental change can be discontinuous, with thresholds between alternative states. Once a threshold is crossed to a more degraded state, the former state cannot be attained without significant management effort, such as prescribed burning, ploughing, or herbicide application, rather than simple grazing control." (Friedel 1991).
Laycock (1991) stated: "In order to develop new concepts and models about range condition, we not only need to identify possible stable states, we also need to identify and understand the factors which can force a stable community across a threshold into a transitional phase moving toward another stable state."
An attempt was made to develop range condition methods and site classification procedures which took the concept of thresholds, transitional phases and stable states into consideration.
METHODS
Two hundred seventy nine prairie/rangeland stands were surveyed during August 199 and June and July of 1994. Most of the stands sampled (246) were upland sites, primarily on silty and limy upland range sites. A few sandy and thin loess sites were surveyed. Only 33 sites were from the wetland and subirrigated sites located in Platte River floodplain. More than 90% of the sites surveyed were located in Buffalo and Dawson Counties.
At each site sampled, owners were asked for permission to survey their land and for information about past land use, including past/present stocking rate. We did not survey without permission. This resulted in a patchy distribution of sites surveyed, because a number of owners did not respond. Before site analysis, we obtained data on soil series, range site, and legal description from county soil survey manuals (Brown et al., 1978 and Butler et al. 1974) At each site, we took a longitude-latitude reading, using a Magellen (Global Positioning System) location finder.
Field sampling was accomplished at each site by 2 or 3 trained botanists. Starting 10-30 m from the fence, a 10 minute walk was begun following an oval shaped path through all major range sites on the area. During this timed survey, all species found were either identified or collected for future identification. At the end of the 10 minute walk, the foliage present was estimated for each species as a percentage of the total foliage that would be present at peak growth season (July).
The contribution to total biomass of major species was estimated to the nearest 5 % . Minor species were rated with a 1 = single plant, 2=rare, but more than one plant, 3=common, and 4=abundant, but not making up 5% of total biomass. Ten minutes was not long enough to find all species on all sites but seemed to present a reasonable compromise between being all inclusive and expediting the field survey. Including the time spent to document biomass estimates, approximately 30 minutes per site plus travel time was required.
Range sites were not surveyed separately at a sample site. The aim was to obtain overall floristic composition at the sites. The silty, limy upland and overflow sites in the loess hills of Central Nebraska are relatively small and irregularly shaped, making differential management nearly impossible, and thus their separate analysis unimportant. Although the range sites were not sampled separately, the data was correlated with the SCS range condition. We calculated the range condition for the silty upland and limy upland based upon the overall species composition for the site (Silty overflow or silty lowland sites were found in drainages of the upland loess plains, but were relatively narrow in mostcases, so they were not included in the range condition calculation). The SCS range condition (Anonymous 1981) that were used in correlative analyses were obtained by averaging the silty upland and limy upland range conditions. Although this did not result in exact estimates of the SCS range condition, it was reasonably close, based upon five sites which were sampled by range site and then overall.
Data entered into microcomputer spreadsheets (Excel) were subjected to analysis. Percentage composition for each species was calculated at all sites (Figure 1). Twenty-two parameters were calculated for each site (including the 18 shown on Figure 2), based upon species composition.
Quantitative coefficients of similarity were calculated for each pair of samples (non-standardized) using NTSYS software and the Morisita II method (Horn 1966, Morisita 1959 and Rohlf 1993). Statistical analyses were done with ABstat software (Anonymous 1990). Cluster analyses were done on NTSYS software using SAHN (sequential, agglomerative, hierarchical and nested) clustering algorithm which corresponds to the program TAXON for mainframe computers. The clustering method used was UPGMA (unweighted, pair-group method with arithmetic average, Rohlf 1993).
RESULTS and DISCUSSION
Survey Results
From the 279 sites surveyed over the 2-year period, 213 species of plants were found. Figure 1 shows species composition comparisons for the lowland sites (Platte floodplain) vs. upland sites (loess hills, mostly north of the Platte River). The 246 upland sites are considered a representative sample of upland sites. The 33 lowland sites are not considered representative, since about half of them were wildlife preserves and nature sanctuaries.
Significant differences (p<0.05) were found between upland and lowland sites. Upland sites had more western wheatgrass (Agsm on figure 1 =Agropyron smithii), western ragweed (Amps=Ambrosia psilostachva), little bluestem (Ansc= Andropogon scoparius), buffalo grass (Buda=Buchloe dactyloides), sideoats grama (Bocu=Bouteloua curtipendula), blue grama (Bogr=Bouteloua gracilis) and Kentucky bluegrass (Popr=Poa pratensis). Species which were more abundant on upland, but not significantly so included: smooth bromegrass (Brin=Bromis inermis), Japanese bromegrass (Brja=Bromus japonicus!, and tall dropseed (Spas=Sporobolus asper).
Species which had greater (p<0.05) percentage composition on lowland sites included: redtop (Agst=Agrostis stolonifera), big bluestem(Ange=Andropogon gerardii), sedges (CARE=Carex spp.), Indiangrass (Sonu=Sorghastrum nutans) and prairie cordgrass (Sppe=Spartina pectinata).
Figure 2 compares lowland and upland prairies as to prairie characteristics that were calculated from the species composition data. Several of these comparisons, surprisingly, showed little difference. Lowland sites included 23 subirrigated areas and 10 wetland sites . The upland sites included mostly silty, limy upland, silty overflow, silty lowland areas, with a few thin loess and sandy range sites. Several significant differences were found between the upland (loess hills) and lowland sites (Platte River flood plain), including percentage tallgrass, with lowland having almost 3 times more. There were only minor differences in percent tallgrass between reference sites (Table 1).
Decreaser composition mirrored the tall grass component, since most of the decreaser species were tallgrasses. Upland sites had a greater percentage of increaser species biomass (as a replacement for the tallgrasses, primarily big bluestem). This may reflect the fact that the lowland sites included a higher percentage of prairie preserves, and were either not grazed or were lightly grazed. Although upland sites in excellent condition had as much tallgrass as lowlands, they are more fragile or sensitive to poor management practices, and recovered more slowly after conditions improve than did the subirrigated sites. Three of four (all but the SCS method) of rangeland health indices were significantly greater on lowland. Lowlands (wetland range sites) had more sedges than uplands, but were not significantly different. Other differences in Figure 2 were not statistically significant (p=.05).
Figures 3 to 10 present distribution histograms of several of the characteristics shown on Figures 1 and 2. The data were combined for all sites, but since the lowland sites made up such a small percentage, (12%) they primarily reflect conditions on upland prairies.
Figure 3 shows that 130 of the sites (total number of sites=279) had 0-5% big bluestem. Big bluestem had a mean of 11.8% species composition at all sites. This species was dominant in both lowland and upland mixed prairie sites, in pristine condition (See reference sites, Table 1) but had disappeared in most upland grazed prairies. Big bluestem is the first species to disappear under heavy or continuous grazing with fenced livestock. It returns quickly in Platte River lowland sites (Nagel 1982) but very slowly on upland sites (Nagel, Nicholson and Steuter 1994, and observations on local rangelands in CPNRD).
Figure 4 shows the distribution of smooth bromegrass in the prairie sites surveyed. About 75% of the sites had less than 6% smooth bromegrass. This species, where it has been seeded into ditches or waterways near rangeland invades very rapidly and may completelydominate the rangeland, with more than 65 % of the species biomass found on some sites contributed by smooth bromegrass.
Kentucky bluegrass, another serious exotic invader in CPNRD, is more widespread, as seen in Figure 5. About 25% of the sites surveyed had less than 5% bluegrass, but 70% of the sites surveyed had from 5 to 40% composition of bluegrass. This low production grass is considered undesireable in CPNRD rangeland. Especially disconcerting was the fact that Kentucky bluegrass contributed more biomass than any other species of plant in CPNRD prairies, (Fig. 1) which is not a good indication for future productivity.
Figure 6 shows that about half of the prairies had less than 2% legume biomass present. Since nitrogen is the major limiting nutrient on plant production in native rangeland in Nebraska, and most ranchers/farmers do not fertilize rangeland, this lack of legumes is especially important. The original prairies contained greater amounts, if the reference sites are representative of original prairies (upland reference sites had a mean of 4.2% legumes, while lowlands averaged 7.7%). None of the correlations run between rangeland health indices and legume biomass was significant, however.
Total number of species found on a site produced a normally distributed curve seen in Figure 7. Most sites yielded at least 15 species during the 10 minute survey, and some had more than 50 species. The number of grass species were rather constant throughout the sites. Forb species varied widely from site to site, probably attributable to broadcast spraying of many sites for noxious weeds, mainly musk thistle, Carduus nutans L. Loss of forb diversity from many sites is a serious threat to biodiversity in this region of Nebraska.
SCS range condition also produced a normal curve (Figure 8). Most of the sites fell into the fair to good classification with an average of 54%.
METHODS USED TO DETERMINE RANGELAND HEALTH OR
CONDITION
Reference Site Approach
Reference sites were identified to compare rangeland/prairie condition with sites where past management was unknown. Sites such as infrequently mowed cemeteries, nature sanctuaries, exceptionally wellmanaged rangelands, and hay meadows were identified (See Table 1 for the 5 lowland and 7 upland sites used as reference sites). An ecological index was used rather than a production based index (Wilson 1984), because our primary objective was to identify and evaluate stands of native rangeland for their biodiversity. We followed Wilson (1984), who recommended use of multiple reference stands since there was greatvariability in the potential vegetation at a site.
The composition of big bluestem varied from 35% to 80% among the 7 upland reference sites. Variation between lowland sites was of less magnitude. Potential reference sites were eliminated from consideration if they had significant amounts of exotic species biomass present. Cemeteries varied greatly in species composition on similar sites, perhaps due to different mowing practices.
Coefficients of similarity were compared between reference sites and sites where rangeland health was unknown. Average values were calculated for each site using all reference sites for the location. The coefficient of similarity was based upon the 9 dominant species of grass, and was calculated by the Morisita procedure in NTSYS program (Rohlf 1993). Coefficients of similarity range from 0 to 100%. In setting the dividing points between the reference classes, the traditional 4 class approach was used (where 0-25% = poor, 25-50 = fair, 50-75 =good and 75- 100 = excellent condition). Figure 9 shows the results of this analysis . More than one-third of the sites rated in the lower half of the poor condition class. Sites dominated by the exotic species smooth bromegrass and Kentucky bluegrass scored almost no points when compared with the reference sites. This doesn't mean that this rangeland does not have any grazing value, but it has little value in conserving native species, or natural habitats.
The correlation coefficient between the SCS range condition and the reference site range condition was +0.48 (Table 2). The reference site approach rated rangeland much lower than the SCS method (means of 31 % vs 54%). The reason for this difference was that the SCS method gives partial or full credit for many of the species which replace the dominant species, big bluestem, on the upland site after a disturbance such as overgrazing, drought, etc. The reference site approach counts only what is found at the reference sites.
Maximum Reference Site Condition Method
This is a modification of the Reference Site method given above. The reference site index of rangeland health was determined by taking the average coefficient of similarity for all reference sites compared with the site being evaluated.
In the Maximum Reference Site version, the highest coefficient of similarity scored with a n v of the reference sites was used as the index of health. The point has been made that there is no single "best" combination of plant species on a site. Each site will vary somewhat due to soil, fire, grazing, precipitation and other factors historically affecting succession on that site (Risser 1989). This method, if an adequate number of referencesites are chosen, should allow for site variation in the successional process. The results of this index are shown in Figure 10. Range condition scores calculated by this procedure produced a bi-modal histogram with our sites, reaching peaks at 6% (very poor condition) and 94%(high excellent). These are realistic peaks in that most of the prairie sites were either badly degraded and invaded by exotics (including those classified from 0-25% condition, which made up 47% of all sites) or in excellent condition (those rating 75-94%).
Thresholds and the Decreaser-Invader Range Condition Method
The results obtained from the reference site comparison method depended heavily upon which reference sites were found or selected. By looking at such a large number of sites some excellent condition upland sites were found. Table 1 shows the extent of the tallgrass present at these sites.
A criticism of the reference site method is that it reveals nothing about thresholds present in the rangeland. There seemed to be two thresholds in Central Nebraska upland prairies. The first one which can be identified is when the dominant species are first replaced by increaser species, due to some disturbance. Sideoats grama, buffalo grass, blue grama, western wheatgrass, tall dropseed, and many others replace the originally dominant decreaser species big bluestem and little bluestem.
A second threshold was found when the native vegetation was disturbed to the point where it was replaced by bluegrass, bromegrass and other invaders. Species classified as decreasers (Anonymous 1968) are species occupying the stable environment at the left of the first threshold (Fig. 11). Increaser species lie in the center of the 2 thresholds. Invader species lie to the right of the second threshold.
To use this concept in formulating an index of rangeland health the invader species composition was subtracted from the decreaser composition at a site, then corrected the data so + 100 (all decreaser biomass at the site) was top of the excellent range condition class, and 100 (all invaders) was equal to zero. If there are all increaser species at a site, it will rank 50% or low good to high fair. Likewise, if a site has half decreasers and half invaders, it also ranks 50%. This index still rated the brome and bluegrass dominated sites very low (mostly poor and low fair), but rated the sites which still have a stand of native increasers more fairly than the reference site approach This index rated between the reference site method and the SCS method (Fig.12 vs 8 and 9).
If there are 2 thresholds and 3 stable assemblages it would be logical to use only 3 condition classes, 0-33%=low, 33-67=moderate, and 67- 100%=high range condition. Regardless of the number of classes used,all four indexes of rangeland health are significantly positively correlated (Table 2), probably because they are all ecologically based, not productivity based indexes.
Cluster Analysis
Figure 13 shows a cluster diagram of the 279 sites, and Table 3 gives characteristics of each of 7 main clusters identified as being important. This cluster is based upon a Morisita coefficient of similarity calculated from data for the 9 dominant grass species over all the sites. These species included: big bluestem, little bluestem, side-oats grama, blue grama, buffalo grass, tall dropseed, western wheatgrass, smooth bromegrass and Kentucky bluegrass. Uresk (1990) worked in South Dakota mixed prairie and found the dominant grass species gave the best clustering. After trying many data sets (from factors shown in Fig. 2), we also found this to be true.
The sites were clustered from left to right in Figure 13, with sites separating at the right side being very close to identical species composition (high coefficient of similarity or low coefficient of dissimilarity). Starting at the left bottom (righthand graph in Fig. 13), the first separation is at about 20% similarity when the sites dominated by buffalo grass separate from all other sites. Buffalo grass forms almost pure stands on hilltops under heavy continuous grazing . Next, at about 25%, smooth bromegrass dominated sites are separated from all remaining sites. Smooth bromegrass is extremely competitive and forms almost pure stands, even in native sod, under some conditions. This exotic grass, introduced for soil erosion control, is a severe threat to the continued existence of native prairie plants in central Nebraska.
The next separation comes at 30% similarity where big bluestem and sideoats grama sites separate out. Big bluestem forms the excellent sites, whereas sideoats grama tends to replace bluestem under moderate to heavy grazing. At 35%, sideoats is separated from big bluestem sites. At 38%, little bluestem sites are separated from western wheatgrass and bluegrass sites. Little bluestem is found on upland sites which are somewhat drier than where big bluestem grows (Stubbendieck, Nichols and Roberts 1986) under light to moderate grazing. Finally, at 50% similarity, western wheatgrass sites are separated from Kentucky bluegrass sites. These two species are both cool season grasses, preferring cool moist sites, although growing throughout the prairies in Central Nebraska. A small cluster dominated by tall dropseed is present, but does not separate out from the bromegrass sites until 75% similarity.
When separated at 45% coefficient of similarity, the sites form 7 distinct clusters. To the right of each cluster in Figure 13 is listed thedominant species in that cluster. Cluster 1 contained all the sites dominated by big bluestem (N=61) . All of the reference sites for both upland and lowland were located here. The mean condition (reference site method) averaged 82 percent (S.E. = 1.72) in this cluster. The second group or cluster is the sideoats grama dominated sites, averaging 34% condition (N=13). Group 3 included the sites dominated by little bluestem, a codominant on mixed prairie upland sites. Reference site condition was 32% (N = 44, S.E. = 2.24) in this cluster. Group 4 was dominated by western wheatgrass, an increaser especially abundant on lowland sites. Group 5 was the Kentucky bluegrass dominated sites with a variety of subdominants, including western wheatgrass, tall dropseed, and others. This group had 68 sites and averaged only 13% condition (S.E. = 1.70). Group 6 was the smooth bromegrass dominated sites, with much bluegrass present. This cluster had 55 sites with a mean condition of only 11% (S.E. = 2.00. Cluster 7 sites were dominated by buffalo grass (N=15) and had a condition of 8% (S.E. = 3.3). Three other clusters could have been recognized, but represented such a small number of sites that they were merged into adj acent clusters.
A 25% similarity separation yields only 2 major clusters plus buffalograss. This is a useful splitting of the sites because it identifies the 2 stands on either side of a single threshold, another potential model for Central Nebraska prairie. It divides the sites into good-excellent (dominated by big bluestem), and all the rest, mostly bluegrass and bromegrass dominated sites.
Use of Cluster Analysis as Replacement for Rangeland Health Indices
The rangeland health indices discussed above correlate well with the cluster analysis results. Since there is no obvious "best" index at this time, all four indices were summed for each site. These scores were sorted in descending order, and the site sequence compared with the sequence found in the cluster analysis (Fig. 13).
Of the top 50 sites for total rangeland health score, all but four were located in cluster 1, the best cluster from the ecological condition viewpoint (as determined by the reference site comparison). The four exceptions came from cluster 2 (sideoats grama) and cluster 3 (little bluestem), the next best clusters, ecologically speaking.
Of the bottom 30 sites for rangeland health score, 4 were from cluster 7 (buffalograss), 24 were from cluster 6 (smooth bromegrass) and the remaining 2 were from cluster 5 (Kentucky bluegrass). These 3 clusters represent the "worst" rangeland health when compared to the threshold model, Fig. 12, or the reference sites.
The degree of relationship between the indices and the cluster is sogood that perhaps only the cluster analysis is needed to evaluate rangeland health from an ecological perspective. Decisions about the level of similarity at which to break the tree into clusters is similar to the decision about where to break the classes in rangeland health indices. With rangeland health indices, historically the 25% X 4 categories has been used, but no standardized statistical procedure exists to establish this division.
With cluster analysis, however, there are several ways to determine which level of clusters are likely to be statistically important (although not strictly significant, Uresk 1990, Rohlf 1993, several textbooks on multivariate statistics). Cophenetic correlation (use of a cophenetic matrix compared with the original data matrix using the MSCOMP (Rohlf 1993) matrix comparison showed a very good fit.
Causes for Patterns of Vegetative Composition Observed
Evidence to explain vegetative changes outlined in the previous paragraphs has not been easy to find. It is overly simplistic to blame overstocking or lack of rotation grazing for the degradation seen on many of the sites. Some sites in the CPNRD have had excellent management (or in some cases, just rested, i.e., no management) and have degraded as rapidly or even more rapidly than overgrazed pastures. In fact, heavy grazing in some cases causes the intermediate stage (increaser dominated) to persist against invasion by the exotic species.
As part of the landowner questionnaire, 64 landowners provided their stocking rates, expressed as acres per animal unit. Correlation of these rates with range condition (SCS method) and reference site method yielded an r of +0.40, and +0.25, respectively. At best, this explains 16% of the variation in range condition or as little as 6% in the reference site approach. Stocking rates reported ranged from 2.5 acres per animal unit to 10 acres.
Landowners reported that 31 % of the rangeland had at one time been partially or wholly in cultivation.. The possibility that as much as one-third of the area surveyed may have been cultivated at some time in the past is another possible explanation for the vegetation patterns.
Sample sites were plotted on a map of the CPNRD, which was prepared in 1974 from aerial infrared photography (Seevers 1978). Included in mapping units were rangeland (native vegetation) and pasture (plants or lands which are intensively managed, presumably non-native), cropland, etc.
Ground truthing in 1993 of about 30 each of these range and pasture sites shown on the map showed that almost all of the rangeland areas were native vegetation. The pasture sites were quite diverse,ranging from alfalfa to some native rangeland vegetation. Presumably, in 1974 at least, the native vegetation sites labeled as pasture sites on the map were badly overgrazed.
Forty-five of the sites were on the areas marked as pasture in 1974 while the rest were classed as rangeland. These 45 sites were compared with 45 adjacent upland sites and these differences were found (Table 4). Tallgrasses and decreasers were more plentiful on the rangeland sites while annual plant biomass, forb biomass (especially weedy species) and invader biomass was greater on pasture sites. Condition of the vegetation, as measured by the reference site and decreaser-invader approaches both showed significant reductions on the more disturbed pasture sites
CONCLUSIONS
Two hundred seventy nine upland and lowland rangeland or prairie sites in the Central Platte Natural Resources District were surveyed over a two-year period.
Five reference sites for lowland and 7 for upland sites were selected, based upon past land use. Sites with a history of light grazing pressure, haying, or no use, such as sanctuaries and cemeteries, were chosen as reference sites in that they were representative of the potential vegetation on that site.
The most surprising conclusion was that upland and lowland reference prairies differed very little when comparing dominant plant species. Big bluestem was the dominant plant at both site locations. Other species of indicators differed between the two sites, but in the cluster analysis, the reference sites for both upland and lowland came out in the same cluster.
Rangeland condition or health, as determined by comparing the sites with the reference sites averaged only low fair condition (32%), and more than half of the sites rated poor. With a modification of this method, where the highest coefficient of similarity with any of the reference sites was used as the index, sites averaged 43%. With a condition index based upon species rating as decreaser and invader, the index of condition averaged 45%. This index was based upon perceived thresholds in the vegetative succession. With the SCS range condition method, the same sites averaged good (54%). It was important to identify multiple reference sites to get a representative sample. The 7 upland reference sites ranged from 35% big bluestem to 80%, demonstrating great variation.
Reference site comparison by Morisita coefficient of similarity worked very well to determine rangeland ecological health. The cluster analysis proved to be a useful parallel analytical procedure, especially good for identifying threshold conditions between sites and also indetermining species affiliations. At a 25% coefficient of similarity separation, cluster analysis produced a site distribution demonstrating 2 threshold conditions for the upland sites. These thresholds are: 1. Transition from the potential vegetation dominated by big bluestem to species more resistant to grazing, such as the gramas, wheatgrass and buffalo grass, and 2. With further stress due to overgrazing or drought or both, the invasion of exotic grasses, such as Kentucky bluegrass and smooth bromegrass. These 2 species were the dominant species on about half of the sites surveyed.
Once the prairie has moved past either of the thresholds, it is difficult for the successional process to move back to the previous state. Unfortunately, our understanding of how the multiple factors affecting vegetative composition on an area through time operate is not adequate to understand how these thresholds work. In some cases, the composition of an area moves directly from bluestem to invader-dominated composition, passing 2 thresholds at once. Much needs to be learned about the interspecies dynamics on Central Nebraska prairies before we can predict such changes accurately.
Overgrazing (too high stocking rate, coupled with continuous grazing) has been stated to be the major disturbance causing the vegetation to pass over these thresholds. Based on landowner-provided stocking rates from about one fourth of the sites surveyed, overgrazing accounts for little of the effect. Another factor is if the land was plowed at some time in the past. Indirect analysis of this factor showed several significant responses by the vegetation, especially greatly reduced reference site rangeland health.
ACKNOWLEDGEMENTS
We thank the U.S. Fish and Wildlife Service for the primary funding of this project through Cooperative agreement 14-48-0006-92-922. Financial support was also given by the Research Services Council at University of Nebraska at Kearney, the William Bruner Research Fund and the Center for Great Plains Studies. F. James Rohlf was helpful with advice about how to best utilize NTSYS programs.
LITERATURE CITED
Anonymous. 1968. Common range plants in Nebraska. Extension Circular E.C. 68- 161. Nebraska Cooperative Extension Service. Lincoln, Nebraska.
Anonymous. 1981. Nebraska Range Site Descriptions and Guide for Determining Range Condition and Suggestive Initial Stocking Rates. USDA.Soil Conservation Service. Lincoln, Nebraska.
Anonymous. 1990. ABstat. Anderson-Bell Co., Parker, Colorado 80134.
Brown, L.E., L.L. Butler and F.E.Wahl. 1978. Soil survey of Dawson County, Nebraska. USDA and Conservation and Survey Division University of Nebraska-Lincoln.
Buller, L.L., R.S. Pollock, R.A Boccheciamp, C.L. Hammond, H. Hill and J.A. Elder. 1974. Soil Survey of Buffalo County, Nebraska. USDA and Conservation and Survey Division, University of Nebraska, Lincoln.
Busby, F.E.(Ed.) 1994. Rangeland Health: New methods to classify, inventory, and monitor rangelands. National Research Council.. National Academy Press. Washington.
Clements, F.E. 1916. Plant succession: An analysis of the development of vegetation. Publ. 242. Carnegie Inst. of Washington, Washington, D.C.
Dyksterhuis, E.J. 1949. Condition and management of rangeland based on quantitative ecology . J . Range Mgmt. 2: 104- 115.
Friedel, M.H. 1991. Range condition assessment and concept of thresholds: A viewpoint. J. Range Mgmt. 44:422-426.
Horn, H.S. 1966. Measurement of "overlap" in comparative ecological studies. Amer. Nat., 100:419-424.
Joyce, L.A. 1993. The life cycle of the range condition concept. J. Range Mgmt. 46(2) 132- 138.
Laycock, W.A. 1991. Stable states and thresholds of range condition on North American rangelands: A viewpoint. J. Range Management 44(5): 427433.
Morisita, M. 1959. Measuring of interspecific association and similarity between communities. Mem. Fac. Sci. Kyushu U., Ser. E (Biol.), 3:65-80.
Nagel, H.G. 1982. Ecological baseline study of Lillian Annette Rowe Sanctuary. Report submitted to National Audubon Society.
, R.A. Nicholson and A.A. Steuter. 1994. Management effects on Willa Cather Prairie after 17 years. The Prairie Naturalist 26(3):241-251.
Risser, P.G. 1989. Range condition analysis: Past, present and future. Pp. 143156 In: Secondary succession and the evaluation of rangeland condition. W.K. Lauenroth and W.A. Laycock, Eds. Boulder, Colorado. Westview.
Rohlf, F.J. 1993. Numerical taxonomy and multivariate analysis system (NTSYS-pc). Exeter Software. Setauket, New York.
Seevers, P.M. 1978. Land use in the western half of the Central Platte Natural Resources District, summer 1974. 1 page map. Remote Sensing Center. Conservation and Survey Division, University of Nebraska-Lincoln.
Stubbendieck, J., J.T. Nichols, and K.K. Roberts. 1986. Nebraska range and pasture grasses. Cooperative Extension Service. University of NebraskaLincoln.
Uresk, D.W. 1990. Using multivariate techniques to quantitatively estimate ecological stages in a mixed grass prairie. J. Range Mgmt. 43 (4) :282 - 285.
Weaver, J.E. 1965. Native vegetation of Nebraska. University of Nebraska Press, Lincoln.
_______ and W.E. Bruner. 1954. Nature and plaace of transition from true prairie to mixed prairie. Ecology 35: 117- 126.
Wilson, A.D. 1984. Points of reference in the assessment of change in vegetation and land condition. Australian Rangeland J. 6(2):69-74.
_______. 1986. The monitoring of changes in range condition: A multivariate site potential approach. IN: Rangelands: A resource under seige. Proc. Second International Rangeland Congress. Cambridge Univ. Press. 1986.634 pp.
Return to 1995 Platte River Basin Ecosystem Symposium 
Last updated by Darren A. Jack on 9/16/97