Page  41 ï~~2004 THE MICHIGAN BOTANIST 41 A REANALYSIS OF 1960S MICHIGAN OLD-FIELD DATA Erica A. Corbett Gail A. Corbett Department of Biological Sciences Department of Biological Sciences, Southeastern Oklahoma State University Illinois State University, Durant, OK, 74701-0609. Normal, IL 61790-41201 580. 745. 2082; ecorbett@sosu.edu (Author for correspondence) ABSTRACT Data were collected from four old-field areas on the grounds of the Matthaei Botanical Gardens, Ann Arbor, Washtenaw County, Michigan, as part of a dissertation research project. Four different sites were studied, with different treatments applied to each site: control, clipping, deep-tilling, scalping, and manuring with clipping. The data were originally collected in the early 1960s and were originally presented in differential tables. We have reanalyzed these data using detrended correspondence analysis, an ordination technique. Although the four sites were geographically fairly close to one another, they differed in site conditions and species composition. The first ordination axis separated sites on the basis of moisture availability. There were relatively few differences between the different treatments, with deep-tilling showing the greatest influence on species composition, possibly because the tilling treatment brought buried seeds to the surface and allowed their germination. The individual sites differed significantly in species richness. Among treatments, the control treatment had a significantly higher species richness than most other treatments but was not statistically significantly greater than the deeptilled treatment. Environmental differences were the major factors affecting species composition, with treatments having less of an effect. INTRODUCTION There may be a wealth of ecological data that have been collected in dissertations or theses in the past that have not been analyzed and published. These data are still valid, and in some cases represent sites that no longer exist. These data can be reanalyzed using more modemrn methods. We are presenting a reanalysis and interpretation of Michigan old-field data collected by one of the authors (GAC) in the early 1960s as part of dissertation research. These data were originally presented in differential tables (following the methods of Braun-Blanquet 1932), but were never published. In 2002, we abstracted the field data from the dissertation. These data were reanalyzed using Detrended Correspondence Analysis (Gauch 1982). Old fields have long been studied as a model for examining secondary succession, community assembly, and competitive interactions among plants (e.g., Maycock and Guzikowa 1984). Succession is a complex process involving interactions between plant species, the abiotic environment, and other species (e.g., herbivores). Initially, succession was thought to follow a consistent and deterministic pattern (e.g., Clements 1916), or to follow one of several possible pathways (Connell and Slatyer 1977), but more recent reviews have suggested

Page  42 ï~~42 THE MICHIGAN BOTANIST Vol. 43 that there may be elements of facilitation, tolerance, and inhibition (sensu Connell and Slatyer 1977) in each succession, with the different patterns representing different species interactions (Pickett et al. 1987) or differences in site conditions. Succession includes a wide variety of effects and interactions, and is a more complex process than originally thought. Factors that can affect species composition of a site include seed size and amount of litter present (Facelli and Pickett 1991, Reader 1993), soil characteristics (Wilson and Tilman 1993, Baer et al 2003), herbivore activity (Carson and Root 2000), and site history or "abandonment patterns" (Myster and Pickett 1994). Old fields have long been used as study sites. They are widespread and easily accessed, and succession proceeds more rapidly than on some other sites. One of the characteristics typical of old fields is a high abundance of non-native (usually Eurasian introductions) agricultural weeds in the flora (Maycock and Guzikowa 1984). Michigan has experienced these introductions, as it is in the hay and dairy region of the United States. The naturalization of certain introduced forage crop species has affected the composition of grassland and old-field communities. The most important species introduced into the region during the period of early agricultural development were Melilotus alba (white sweet clover), Melilotus officinalis (yellow sweet clover), Trifolium pratense (red clover), T hybridum (alsike clover), T repens (white clover), Phleum pratense (timothy), Poa pratensis (Kentucky bluegrass), P. compressa (Canada bluegrass), and Agrostis alba (redtop) (Klages 1942, Wheeler 1950). Many of these forage species were present on the Michigan plots studied as part of this research. The sites studied were former farm fields with varied histories. Beckwith (1954) observed that early stages of secondary succession on former agricultural land in Michigan were strongly influenced by the last crop planted, with lesser influences from soil and fire history of the site. A general pattern described by Beckwith (1954) for Michigan old-field succession is an annual-biennial stage, followed by a perennial grass stage, followed by a mixed herbaceous perennial stage, a shrub stage, and a shade-intolerant tree stage. Hayfields showed a slower progression than crop fields, perhaps because there was a grass community already established. Pickett et al (1987) suggest that competitive inhibition may be a major factor affecting succession in old fields, and that disturbance can "break the hold" that a group of species has on the site and allow for new invasions. In this study, we examine the effects of different sites and treatments on the change in species composition over a threeyear period. MATERIALS AND METHODS Field sites The research sites were located on the grounds of the Matthaei Botanical Garden, Ann Arbor, Washtenaw County. Four different sites were located within the grounds of the Botanical Garden (see figure 1 for map of the Matthaei Botanical Gardens and site locations contemporaneous with the time of sampling). Each of the sites differed slightly in site history and species initially present. However, extensive site histories were not available or were not collected; we are presenting the available information on site-history here.

Page  43 ï~~2004 THE MICHIGAN BOTANIST 43 University of Michigan Botanical Gardens Dixboro Site soo Dinbere Rood Scale' oow' to.. os II - Legend - FIGRE1:Ma o Mttae Btaicl aren sownloions ofheeursapest.Mp shows onditonsocan1962

Page  44 ï~~44 THE MICHIGAN BOTANIST Vol. 43 Site 1 (referred to as Matteson Field in the dissertation) was first occupied in 1824 and was farmed extensively until the land was deeded to the University of Michigan. Unfortunately, no data are available on the exact date of abandonment and on the types of crops previously grown there. The site had "imperfectly drained" soils of the Matherton Sandy Loam series and a maximum slope of 2%. Before treatments were applied, the most abundant species were Phleum pratense, Trifolium pratense, Plantago lanceolata, Potentilla recta, and Cerastium vulgatum. Site 2 was located in an embayment of grassland in a conifer plantation in the northeast portion of the Matthaei Botanical Gardens. Soils on this site were similar to those of site 1, but perhaps with better drainage. This site was smaller than the others in this study, and only the clipped, control, and deep-tilled treatments were applied. Additionally, this site was destroyed before the completion of research. Before treatments were applied, the most abundant species were Sonchus arvensis, Poa compressa, Solidago canadensis, and Daucus carota. Site 3 was located in the southeastern portion of the Garden in an area known as Sanford Hill. The soil on this site was described as "well-drained" and was a Boyer Sandy Loam. Slopes on this site ranged from 6% to 12% and these soils were known to be susceptible to erosion. Experimental plots were located near the top of a small rise parallel to the contour of the hill. Before treatments were applied, Agropyron repens and Tragopogon major were the dominant species. Site 4 was located in John Dix Field in the eastern part of the Botanical Gardens. This site was excavated for sand and gravel prior to the original study, which probably altered the soil profile. Slopes in this area were around 7%. Before treatments were applied, the most abundant species were Bromus tectorum, Melilotus officinalis, Ambrosia artemisiifolia, Plantago lanceolata, and Arenaria serpyllifolia. All of these fields had been mowed at least once each growing season for more than ten years prior to the study, but the precise number of years is not known. During the study, mowing of vegetation in the areas surrounding the experimental plots was continued. Within each of the four sites, different treatments were applied (see figure 2). The treatments included a control (no treatment applied), deep-tilling (sites were tilled once and then allowed to recover), clipping (sites were clipped yearly), scalping (sites were scalped on a yearly basis), and manuring with clipping (sites were treated yearly). In the control quadrats, the vegetation was left undisturbed. Actually, the control quadrats did constitute another experimental treatment because the fields in which the experimental quadrats were located had been mowed once or twice annually for a number of years prior to the beginning of the research project. The deep-tilling treatment consisted of spading the soil to a depth of approximately 20 cm (8 inches) with a shovel. The vegetation was completely turned under. This treatment was carried out only once, in the summer of 1961, whereas the other treatments were performed twice in 1961 and once in 1962. In the clipped quadrats, vegetation was cut to a height of approximately 6.35 cm (2.5 inches) using a pair of garden shears. The clippings were left on the plot. This treatment was done to simulate mowing and, because the experimental study had been mowed at least annually for a number of years, this was actually similar to the previous management practices. Quadrats subjected to scalping were completely denuded of green shoots and litter with a sharpened trowel. The aerial parts of vascular plants, including stolons, were severed at the soil level, leaving the root systems intact. The aboveground biomass was removed from the quadrats. This process disturbed the soil surface. The quadrats subjected to the manuring and clipping treatment had vegetation clipped to about 3.8 cm (1.5 inches) using garden shears. The clippings were removed from the site and 500 grams of sterilized cow manure were spread evenly on each of these quadrats. This treatment was designed to simulate grazing effects. The use of clipped quadrats to simulate grazing is subject to certain limitations: 1) clipped vegetation is cut uniformly at a certain height, whereas livestock pull and break off some of the shoots, 2) livestock are selective in what they eat, and some plants are left untouched, and 3) the species they prefer are often eaten to a height of 2 to 4 centimeters, whereas other species will be less damaged (Culley et al. 1933). In addition to these factors, the trampling by hooves increases the amount of disturbance in natural pastureland. Despite these restrictions, it has been shown that such simulation of grazing provides useful results when compared with actual pastureland (Aldous 1930). The area set aside for sampling was 1 m wide by 10 m long (site 2) or 20 m long (all other sites).

Page  45 ï~~2004 THE MICHIGAN BOTANIST 45 2004 THE MICHIGAN BOTANIST 45 MC Manured and Clipped C Control DT Deep Tilled Cl Clipped Sc Scalped Scale I i 0 4 Meters C DT DT C Cl C Sc Sc C Sc c DT C C C C C1 c Cl Cl C FIGURE 2: Diagram of experimental treatment layout. Treatments were assigned randomly. Each treatment plot was separated by a control plot, and treatments were assigned such that no two replicates of the same treatment were on either side of a control plot (figure 2). Sites 1, 3, and 4 had all treatments applied; site 2 was smaller and had only clipping and deep-tilling treatments. Experimental plots were established as either 1 m2 or 2 m2 areas. The larger plots were initially used to test for edge effects: in the center of these plots, a 1.6 m by 0.6 m area was marked off and sampled after the initial application of treatments. No edge effects could be detected, so both the larger and smaller plots were used in later samples. Data collection Each site was sampled during each month of the growing season in 1960, 1961, and 1962. A onemeter square quadrat was used to collect samples from each of the twenty (ten, for site 2) one-meter square plots. Data were collected on abundance and sociability following the methods of BraunBlanquet (1932); for the current analysis, the abundance data were converted to percent relative frequency based on occurrence in quadrats. Data analysis Data given here were abstracted from the dissertation (Corbett 1967). These data represented numbers of occurrences per site per sampling date, and so were easily converted to relative fre

Page  46 ï~~46 THE MICHIGAN BOTANIST Vol. 43 quency. Relative frequency was computed as the total number of occurrences of a species at a site on a particular date divided by all occurrences of all species on that site on that date. Values were then converted to a percent. These data were entered into a database (a Lotus-style,.wkl database generated using Microsoft Excel) for data analysis. The computer package PC-ORD (McCune and Mefford 1997) was used to analyze data. We used detrended correspondence analysis to examine patterns in the data. Detrended correspondence analysis is a multivariate analysis technique. It is a form of ordination which examines the abundance of all species at all sites simultaneously (Gauch 1982). The end result of detrended correspondence analysis is that sites are arranged in "ecological space" based on similarities in species composition. The ordination diagram that results shows the relationships between sites. The sites may form a continuum that represents an environmental gradient (ter Braak and Prentice 1988). It is then up to the researcher to interpret the results. We chose this method of analysis for several reasons. First, a multivariate technique is capable of examining a large number of independent variables (sites, treatments, sampling dates) and response variables (species relative frequencies) simultaneously. Of the various ordination techniques available, we chose detrended correspondence analysis because it is known to work well on field data and is effective where there may be long environmental gradients (Peet et al. 1988). Information that can also be calculated from these data are measures of species richness (s, the number of species per sample), Shannon diversity (calculated as -E(pi (ln pi)) and evenness (Shannon diversity/In(s), where In (s) is the maximal Shannon diversity possible). These values were calculated for each sample using the "summary" function of PC-Ord (McCune and Mefford 1997). We statistically tested the effect of site and treatment on species richness at the sites. However, the data were not normal (and could not be made normal by transformation), so it was necessary to conduct nonparametric tests on the data. We used the JMP version 5.0 statistics program (SAS Institute 2002). There was no option for a two-way nonparametric analysis of variance in JMP. Instead, we used Kruskal-Wallis to test site and treatment effects separately. We did test for the presence of an interaction using a standard, parametric two-way ANOVA and found no interaction. We performed follow-up tests for the analyses of variance using standard parametric methods (the Tukey-Kramer "honestly significant difference" test). This is not an ideal technique. However, few good non parametric multiple comparison tests exist and they are not available in many standard statistical packages, including JMP (S. A. Juliano, pers. comm., 2003). RESULTS When we examined an analysis of all sites together, we observed a separation among the four sites on ordination axis 1 (see figure 3). Samples from field site 3 consistently received the lowest ordination axis 1 scores, sites 1 and 4 received similar (and mid-range) ordination axis 1 scores, and site 2 received the highest scores. This result was regardless of date and treatment, demonstrating that differences among the various field sites was the most important factor affecting ordination results. This separation may be the result of differences in moisture availability among the sites. We used the "overlay main matrix" function in PC-Ord to examine relationships between sites and species abundance. Table 1 contains lists of species associated with each of the four sites. Site 3, the site receiving the lowest Axis 1 scores, was the only site with samples having Euphorbia corollata, Lepidium campestre, Hedeoma hispida, Lactuca canadensis, and Holosteum umbellatum present. In Illinois prairies, Euphorbia corollata is abundant on drier sites, such as hill prairies (Corbett 1999). None of these species had high abundance before treatment.

Page  47 ï~~2004 THE MICHIGAN BOTANIST 47 Q Site number "fA 1 Q 2 *" 3 " VV 4 V A A * * A * ** v 0 * O O * 0 Axis 1 FIGURE 3: Detrended correspondence ordination of all treatments for all sites at all dates. Ordination axis 1 separates sites on the basis of moisture availability. Separation on axis 2 is unclear and may not be meaningful. Site 2 was the only site to have Carex spp., Cornus stolonifera, Equisetum, and Juncus tenuis present. All of these species are typically thought of as being present on wetter sites. Site 2 also had several species of Aster and of Solidago, as well as Acalypha rhomboidea, which are sometimes found in prairies. Sites 2 and 3 were the most different from one another in terms of species composition. Again, these species were not present in high abundance prior to treatment. Sites 1 and 4 shared many species (see table 1). Many of these species, such as Aster ericoides, are considered to be species present on mesic sites (e.g., Curtis, 1971). Sites 1 and 4 did differ somewhat in terms of species composition; however these differences did not show a clear pattern and may more reflect the differing histories of these sites than any environmental difference between them. We examined the relationships between individual species and the ordination axes. The species showing high correlation with axis 1 tended to be species having either high abundance in site 3 (Agropyron repens, r = -.714 and Lepidium campestre, r = -.727) or site 2 (Sonchus arvense r =.756, Solidago canadensis r

Page  48 ï~~48 THE MICHIGAN BOTANIST Vol. 43 TABLE 1: Lists of species associated with each of the four sites. Species-site relationships Site 1: Barbarea vulgaris Chrysanthemum leucanthemum Dianthus armeria Lepidium virginicum Medicago sativa Mollugo verticillata Plantago rugelii Potentilla argentea Rumex crispus Trifolium hybridum Verbascum thapsus Site 2 Acalypha rhomboidea Agrimonia parviflora Agrostis alba Aster laevis Aster novae-angliae Carex spp. Cirsium arvense Cornus stolonifera (= sericea) Equisetum arvense Erigeron philadelphicus Hypericum spathulatum (= prolificum) Juncus tenuis Solidago canadensis Solidago graminifolia Solidago juncea Solidago rugellii Sonchus arvensis Stachys arvensis Site 3 Hedeoma hispida Holosteum umbellatum Lactuca canadensis Lepidium campestre Tragopogon major Site 4 Arenaria serpyllifolia Bromus tectorum Centaurea maculosa Erigeron (= Conyza) canadensis Polygonum arvense Silene dichotoma Trifolium pratense Trifolium repens Verbascum blattaria In sites 1+4 Potentilla intermedia Setaria glauca Achillea millefolium Aster ericoides Cirsium vulgare Phleum pratense Lychnis alba Erigeron strigosus =.743, or Daucus carota, r =.813). On this basis, it seems that axis 1 reflects differences in site conditions, and possibly, site history. Axis 2 of the ordination analysis is less informative and shows less separation. In general, the deep-tilled treatments, which received a single application of a more severe disturbance and then were allowed to revegetate naturally, showed higher axis 2 scores than the other treatments. Typically, also, the highest axis 2 scores were shown by deep-tilled sites early in the study, suggesting that deeptilled sites became "more like" the other sites as time progressed. There was no clear separation among the other treatments, and no clear separation by sampling date. Axis 2 generally separated the deep-tilled treatments from the other treatments. The only species showing a strong relationship with axis 2 of the ordination is Ambrosia artemisiifolia (r=.615), which has higher abundance in deeptilled treatments than in other treatments. We also analyzed each site individually to determine if there were any pat

Page  49 ï~~2004 THE MICHIGAN BOTANIST 49 Site 1 data alone v treatment * " A Clipped " E Control 0 Deep-tilled 0 V Manured and clipped 0 1 * < Scalped.N v0 X f 0 0 Axis 1 FIGURE 4: Detrended correspondence analysis of Site 1 data alone. The major separation appears to be between the deep-tilled treatment and the other treatments. terns of treatment or time within each site. For site 1 (see figure 4), the deeptilled samples separated on axis 1 from other treatments. Deep-tilling apparently had the largest effect on community composition, with other treatments (especially clipping) having very similar axis 1 and 2 scores to the control treatment. There was no clear progression of change over time; this may be partly because of variability inherent in the site and partly because several of the treatments were reapplied during the course of the study. Site 2, which was a smaller site with fewer treatments applied, showed a clearer pattern of response to treatments (see figure 5). Deep-tilled samples received higher axis 1 scores than either the control or clipped samples. And among the clipped and control treatments, separation was more on the basis of sampling date than on the basis of treatment, with clipped and control plots sampled at the same date receiving similar axis 1 and axis 2 scores. Site 3 data again showed a separation of the deep-tilled treatment, with no clear pattern among the other treatments. However, the samples for date 4, regardless of treatment, received higher axis 2 scores than other dates (see figure 6).

Page  50 ï~~50 THE MICHIGAN BOTANIST Vol. 43 50 THE MICHIGAN BOTANIST Vol. 43 Site 2 data alone treatment A Clipped 1 Control Â~* Deep-tilled (N" C> A * A 0 0 Axis 1 FIGURE 5: Detrended correspondence analysis of Site 2 data alone. The major separation appears to be between the deep-tilled treatment and the other treatments. The ordination diagram for site 4 is difficult to interpret because of two outlier points (see figure 7). Date 2 for the "scalped" treatment and for the "clipped" treatment are similar in species composition, and are markedly different from the rest of the samples. Examination of the data for date 2 for the clipped and scalped treatments showed that they had very low species diversity: the clipped treatment had only Cirsium vulgare, Erigeron strigosus, Lychnis alba, Oxalis stricta, Plantago lanceolata, Potentilla recta, and Silene dichotoma present. The scalped treatment had even lower species diversity, with only Plantago lanceolata and Potentilla recta present. When the data are re-analyzed with both outlier samples removed, the deeptilled treatment for date 2 becomes an outlier, but not so severely as the clipped and scalped treatments were. The main separation on axis 1 is between the manured and clipped treatments, which received high axis 1 scores, with the other treatments generally intergrading. For the statistical analysis of species richness, sites differed significantly (p

Page  51 ï~~2004 THE MICHIGAN BOTANIST 51 Site 3 data alone v treatment A Clipped Q Control 0 Deep-tilled V Manured and clipped o Scalped CN A Ul) 0 "" X VA Axis 1 FIGURE 6: Detrended correspondence analysis of site 3 data alone. <.0001). Site three had a significantly lower species richness than all other sites (mean = 7.71), and site 1 had a significantly greater species richness than all sites (mean = 24.86). Sites 2 and 4 did not differ significantly from each other in species richness. There were significant differences among treatments in species richness (p =.0043). The control treatment had significantly higher species richness (mean = 22.81) than the clipped, manured and clipped, or the scalped treatment. It was not significantly different from the deep-tilled treatment (mean = 17.81), but the deep-tilled treatment was not significantly greater than the clipped, manured and clipped, or scalped treatments. DISCUSSION Based on the ordination analyses, the major difference among sites and treatments were environmental differences between the four sites surveyed. Ordination axis 1 separated the four sites, with site three receiving the lowest axis 1 scores, site two receiving the highest axis 1 scores, and sites one and four receiving similar (mid-range) scores. This difference is related to moisture avail<I vA V] c[ I Axis I FIGURE 6: Detrended correspondence analysis of site 3 data alone. <.000 1). Site three had a significantly lower species richness than all other sites (mean = 7.71), and site 1 had a significantly greater species richness than all sites (mean = 24.86). Sites 2 and 4 did not differ significantly from each other in species richness. There were significant differences among treatments in species richness (p =.0043). The control treatment had significantly higher species richness (mean = 22.81) than the clipped, manured and clipped, or the scalped treatment. It was not significantly different from the deep-tilled treatment (mean = 17.81), but the deep-tilled treatment was not significantly greater than the clipped, manured and clipped, or scalped treatments. DISCUSSION Based on the ordination analyses, the major difference among sites and treatments were environmental differences between the four sites surveyed. Ordination axis 1 separated the four sites, with site three receiving the lowest axis 1 scores, site two receiving the highest axis 1 scores, and sites one and four receiving similar (mid-range) scores. This difference is related to moisture avail

Page  52 ï~~52 THE MICHIGAN BOTANIST Vol. 43 L Site 4dtaon treatment A Clipped A V Control 0 Deep-tilled A Av 7 Manured and clipped o AO Scalped *V N so V0 X V " Axis 1 FIGURE 7: Detrended correspondence analysis of site 4 data alone. Outlier points (the clipped and scalped treatment for date 2) were removed. ability-site three is the driest of the four sites and site two is the wettest. Sites 1 and 2 belonged to similar soil associations but differed in degree of drainage, with site 1 supposedly showing poorer drainage than site 2 (this contradicts the results of the detrended correspondence analysis). Site three, the driest site, had a sandy loam soil that was considered to be "well-drained". This site also showed the highest percentage slope of all four sites. Site 4 had an altered soil profile resulting from sand and gravel removal, but the alterations were not great, as this site showed similarities in ordination scores to site 1. All four sites received similar precipitation and had similar temperature regimes, so the differences observed were most likely related to topographic position and substrate conditions. Agropyron repens had high relative frequency in site three samples. In some samples, this species achieved high levels of relative frequency and dominated the samples. This may have affected the species-richness result; site three had the lowest species richness of all sites. This species was present on site 3 before the treatments were applied; it is likely that disturbance of the site allowed further spread of this species. Among the individual treatments, the deep-tilling treatment had the greatest effect on the species composition, with deep-tilled treatments receiving the highest axis 2 scores. The species having the greatest relationship with this axis was

Page  53 ï~~2004 THE MICHIGAN BOTANIST 53 Ambrosia artemesiifolia, which had highest abundance in deep-tilled sites. This result was somewhat similar to that found by Miller (1994), who established a five-species system on a tilled site to test competitive and indirect effects. He found that Ambrosia had the greatest competitive effect on other species, and was relatively insensitive to competition from the other species planted. Ambrosia was not abundant in any of the other treatments in this study, including the scalped treatment. Possibly, the deep-tilling treatment (or the plowing, in Miller (1994)) brought buried seeds to the surface and allowed for their rapid establishment and growth. Interestingly, Armesto and Pickett (1985) found that Ambrosia artemisiifolia was one of the species that responded positively to a clipping disturbance, and that clipping increased species richness. We did not find either to be the case in the current study. It is possible that we did not find more differences between treatments because the sites were sampled at different times across the season for 3 years, rather than sampled yearly for a longer span of years. Maycock and Guzikowa (1984) note that the species abundance on their study sites changed over the course of the season, because plants display different phenology. Earlier studies of Michigan old-fields demonstrated that moderately fertile, cleared land developed high population density of Poa pratensis or Phleum pratense, whereas Agrostis alba and Poa compressa invaded areas with lower fertility and thin upland soils (Wheeler 1950). Benninghoff et al. (1961) studied communities that redeveloped on silty soil and gravelly sand or till following agricultural abandonment. They found that Agrostis alba, Phleum pratense, Trifolium pratense, and Trifolium repens were abundant on these sites. However, none of these species were major species in the current study. In old-field succession studies in the literature, various species of Solidago (esp. the Solidago canadensis-altissima complex) are frequently listed as species having high abundance or responding positively to the disturbances that give rise to old fields. Maycock and Guzikowa (1984) describe it as forming large clones and being able to remain competitive for long periods of time. Solidago was also the most speciose genus found in their study. Removal or reduction of Solidago through herbivory or clipping increased diversity in fields where it had dominated (Armesto and Pickett 1985, Carson and Root 2000). In our study, Solidago species only reached high abundance in site 2. This may in part be a result of different site histories. Armesto and Pickett (1985) describe Solidago canadensis as colonizing old fields by seed early in succession, then forming large clones that are resistant to invasion by other species. Agropyron repens had highest relative frequency in Site 3 samples. In fact, this species was the species with highest relative frequency in a number of samples from this site. Maycock and Guzikowa (1984) list Agropyron repens as a species present on drier sites in their study, which is also apparent in the current study. Agropyron is also thought to be allelopathic (Werner 1975), which could in part explain its high abundance on site three. It is not clear whether environmental differences at the other sites prevented it from growing there, or if it merely had not spread to the other sites yet. Another concern is woody invasion. Tree seedlings were present in the Michigan sites, specifically Acer negundo, Quercus alba, and Ulmus americana,

Page  54 ï~~54 THE MICHIGAN BOTANIST Vol. 43 but they showed closer relationship with site (Acer seedlings were present in site 3, Quercus in site 1, and Ulmus in site 2), rather than with treatment. This pattern probably reflects what tree species were closest to each site. When each site was examined independently, similar patterns of reactions to treatments were revealed. In all sites, deep-tilling had the greatest effect on species composition, with deep-tilled treatments receiving higher ordination axis scores than the other treatments. Deep-tilled sites also had significantly lower species richness. Other treatments seemed to have similar effects on species composition and richness. The greatest differences we observed in this study were related to environmental differences among the different sites. These differences resulted from substrate and location differences. Additionally, site history may affect the composition of the different sites. Individual treatments had less of an effect on species composition. It seems that old field succession is largely affected by individual site conditions, with less of an effect from the type of disturbance that most recently took place. ACKNOWLEDGMENTS We acknowledge the much-appreciated guidance of the late William S. Benninghoff, who was the dissertation advisor of the second author. LITERATURE CITED Aldous, A. E. 1930. Effect of different clipping treatments on the yield and vigor of prairie grass vegetation. Ecology 11: 752-759. Armesto, J. J. and S. T. A. Pickett. 1985. Experiments on distubance in old-field plant communities: impact of species richness and abundance. Ecology 66(1): 230-240. Baer, S.G., J. M. Blair, S. L. Collins, and A. K. Knapp. 2003. Soil resources regulate productivity and diversity in newly established tallgrass prairie. Ecology 84(3): 724-735. Beckwith, S. L. 1954. Ecological succession on abandoned farm lands and its relationship to wildlife management. Ecological Monographs 24: 349-376. Benninghoff, W. S., F. Bevis, and R. Wieger. 1961. Differential table of grassland communities on moraine upland, southeastern Michigan. Unpublished manuscript. Braun-Blanquet, J. 1932. Plant Sociology, the Study of Plant Communities (English translation by G. D. Fuller and H. S. Conard). McGraw-Hill Co., New York. 429 pp. Carson, W. P. and R. B. Root. 2000. Herbivory and plant species coexistence: community regulation by an outbreaking phytophagous insect. Ecological Monographs 70 (1): 73-100. Clements, F. E. 1916. Plant succession: an analysis of the development of vegetation. Carnegie Institute of Washington Publication no. 242. Connell, J. H. and R. O. Slatyer. 1977. Mechanisms of succession in natural communities and their role in community stability and organization. American Naturalist 111: 1119-1144. Corbett, G. A. R. 1967. Field vegetation development and responses to management in Washtenaw County, Michigan. Unpublished doctoral dissertation, University of Michigan, Ann Arbor, Michigan Corbett, E. A. 1999. Environmental and geographic correlates of Illinois prairie. Unpublished doctoral dissertation, Illinois State University, Normal, Illinois. Culley, M., R. Campbell, and R. Canfield. 1923. Values and limitations of clipped quadrats. Ecology 14: 35-39. Curtis, J. T. 1971. The Vegetation of Wisconsin. University of Wisconsin Press, Madison, Wisconsin. Facelli, J. M. and S. T. A. Pickett. Plant litter: Light interception and effects on an old-field plant community. Ecology: 72(3): 1024-1031.

Page  55 ï~~2004 THE MICHIGAN BOTANIST 55 Gauch, H. G. 1982. Multivariate Analysis in Community Ecology. Cambridge, UK: Cambridge University Press. Klages, K. 1942. Ecological Crop Geography. New York: The Macmillan Co. 615 pp. Maycock, P. F. and M. Guizkowa. 1984. Flora and vegetation of an old field community at Erindale, southern Ontario. Canadian Journal of Botany 62: 2193-2207. McCune, B. and M. J. Mefford. 1997. PC-Ord: Multivariate analysis of ecological data, Version 3.0. Glenedon Beach, OR: MjM Software Design. Miller, T. E. 1994. Direct and indirect species interactions in an early old-field plant community. American Naturalist 143(6): 1007-1025. Myster, R. W. and S. T. A. Pickett. 1994. Initial conditions, history and successional pathways in ten contrasting old fields. American Midland Naturalist 124(2): 231-238. Peet, R. K., R. G. Knox, J. S. Case, and R. B. Allen. 1988. Putting things in order: the advantages of detrended correspondence analysis. Ecology 74: 2215-2230. Pickett, S. T. A., S. L. Collins, and J. J. Armesto. 1987. Models, mechanisms, and pathways of succession. Botanical Review 53(3): 335-371. Reader, R. J. 1993. Control of seedling emergence by ground cover and seed predation in relation to seed size for some old-field species. Journal of Ecology 81(1): 169-175. SAS Institute. 2002. JMP version 5.0. Cary, North Carolina: SAS Institute. Ter Braak, C. J. F. and I. C. Prentice. 1988. A theory of gradient analysis. Advances in Ecological research 18: 271-317. Werner, P. A. The effects of plant litter on germination in teasel, Dipsacus sylvestris Huds. American Midland Naturalist 94: 470-476. Wheeler, W. A. 1950. Forage and Pasture Crops. New York: D. van Nostrand Co. 752 pp. Wilson, S. D. and D. Tilman. 1993. Plant competition and resource availability in response to disturbance and fertilization. Ecology 74 (2): 599-611.