Skip common site navigation and headers
United States Environmental Protection Agency
Great Lakes
Begin Hierarchical Links EPA Home > Great Lakes > EcosystemsUpland Ecosystems > Oak Savanna Conferences > 1993 Midwest Oak Conferences > Steve Glass   
Aquatic Ecosystems
EPA Region 5 Critical Ecosystems
Ecosystem Funding
Great Lakes Basin Ecosystem
Great Lakes Biological Diversity
Green Landscaping
Rivers and Streams
Upland Ecosystem



1993 Proceedings of the Midwest Oak Savanna Conferences

1993 Proceedings of the Midwest Oak Savanna Conferences

Steve Glass
University of Wisconsin
Madison Arboretum
1207 Seminole Highway
Madison, WI 53711

Evelyn Howell
Department of Landscape Architecture 
University of Wisconsin-Madison
25 Agriculture Hall
Madison, WI 53706 


The use of natural re-vegetation to restore overgrown former prairie or savanna sites may be an alternative to traditional methods. Knowledge of the sources of re-vegetation may help restorationists choose sites that have the potential for restoration through this technique. Few studies have investigated all the possible sources of re-vegetation at a single site and the relative contributions of the seed bank and seed rain to re-vegetation in the prairie/forest province are unknown.

We sought to determine: 1) the likely origin of new plant recruits that colonized an overgrown former southern Wisconsin oak barrens after removal of the 40+ year-old canopy; and 2) if a natural grassland could be restored to the site by this management technique. We identified the reserves of viable, available seeds in the soil; monitored the seed rain onto the site; and compared the available propagates with the vegetation that established itself after the disturbance.

Eighteen species increased in frequency or were newly sampled on the site after treatment; eleven were woodland species, two prairie/savanna species and five non-native species. Sixteen other species were observed on the site for the first time after treatment: one woodland species, seven prairie species and eight non-native species typical of disturbed or weedy habitats. Of these 34 species, five were represented in the seed bank, six in the seed rain, 13 existed on the site previously, and ten were of unknown origin. 

INTRODUCTION  back to top

Restoration of remnant or degraded natural areas may be viewed as an alternative, or supplement, to the preservation of intact natural systems that are becoming increasingly scarce. The use of natural re-vegetation, in whole or in part, to restore overgrown former prairie or savanna sites, in particular, may be an alternative to traditional restoration techniques (Holtz 1985). 

Manipulation of vegetation to simulate natural perturbations has been used to restore prairie or savanna from forest in Wisconsin and elsewhere. Anecdotal reports are common and there are a few documented cases of the dramatic change from forest to prairie or savanna simply by cutting and burning the ground layer. Crex Meadows, a sand barrens area in northwestern Wisconsin, is a site dominated by 20 to 30 year old jack pine (Pinus banksiana) and Hill’s oak (Quercus ellipsoidalis). After two burns within four years the savanna returned. In another area of the site, a crown fire destroyed the tops of the trees. The result was the appearance of over seventy prairie species, most of them in bloom, within three months (Curtis 1959; Vogl 1964). Curtis (1959) referred to this phenomenon as the “instantaneous prairie response.” 

White (1983) reported another example in which annual burning of a Hill’s oak community in central Minnesota decreased canopy density and shrub cover and increased the presence of typical prairie grasses and forbs. 

Holtz (1985) recounts the experience of a landowner in Sauk County, Wisconsin who achieved a dramatic return to savanna (42 forb and 13 grass species) within one year after he cleared an oak woods and ran a spring burn through the site. Holtz (1985) documented a similar transformation on an adjacent site at the Aldo Leopold Memorial Reserve. 

What is not known are the mechanisms responsible for such sudden shifts of vegetation in the prairie/forest province of Wisconsin. Curtis (1959) suggested several possibilities: viable seeds in the soil or the presence of dormant roots or stunted plants in the understory. Lippert and Hopkins (1950) and Shimek (1930) thought invasion by seed to be the primary source. Holtz (1985) discovered that several of her invading species were present in the soil seed bank; noted the presence of several holdover prairie species in the understory before treatment ; and speculated on the origin of the other species. 

Holtz and Howell (1984) suggest that it would be “helpful for restorationists and preservationists to be able to identify sites suitable for prairie or savanna restoration,” and that knowledge of the mechanisms responsible for sudden vegetation shifts might provide this predictive ability. They state that “if holdovers from a former prairie are more responsive to this restoration method than is invading seed rain from a nearby prairie source, then woodedsites which have these holdovers in the understory would have more potential for a rapid prairie restoration than wooded sites without them.” 

THE STUDY SITE  back to top

The study site is located on the Grady Tract of the University of Wisconsin-Madison Arboretum in Madison. The original land survey records indicate that at the time of settlement this area was an oak barrens, defined by Curtis (1959) as occurring manly on sandy soil with a dry-mesic prairie understory dominated by the black oak (Quercus velutina)/Hill’s oak complex without the presence of jack pine. 

Since settlement and the cessation of fire the site has grown up to a southern dry forest (Curtis 1959) dominated by the black oak/Hill’s oak complex, black cherry (Prunus serotina) and a few open-grown bur oaks (Quercus macrocarpa) with a typical southern dry forest groundlayer. In addition, the site has a thick understory of honeysuckle (Lonicera x bella) and scattered buckthorns (Rhamnus cathartica). The site was grazed periodically until the early 1940s and the last wildfire went through it about 1946 (Zimmerman, personal communication). 

The study site is a narrow slice (182m by 274m) in a triangularly-shaped wooded area .18 hectares in size. The site is gently sloped four to eight degrees to the south and east. Soils are of the Boyer and Spinks series (Soil Survey of Dane County, Wisconsin, 1978). The site is bordered on the east by extensive oak woods, similar to that found on the study site. To the south of the site and separated from it by a thin strip of woods and two fire lanes lies the 14 hectare Greene Prairie. On the northern and western edge of the study site and separated from it by only the width of a fire lane is the 4.5 hectare West Grady Knoll Savanna. At least 256 species of prairie grasses and forbs are known to grow on the Grady Tract. 

METHODS  back to top

Year One - Pre-disturbance

The study began in the summer of 1985. Vegetation sampling on July 23 consisted of gathering presence/absence data in 30, 1-meter square quadrats laid out in a stratified random pattern. Species on the site but outside the quadrats were noted as they were observed throughout the summer. 

Soil Seed Bank 
Soil samples were gathered on July 27 and 28, 1985 for the purpose of determining the kinds and numbers of viable seeds in the soil and the presence of dormant roots. One sample was collected from the outside southwest corner of each of the 30 vegetation sampling quadrats. We sampled a surface area 7.4cm in diameter to a depth of 10cm (including the surface litter). 

On July 28, the soil samples were spread on standard nursery flats (28cm x 54cm) filled with Jiffy-Mix, a commercial, sterile, soil-less potting mix. The flats were placed in a shaded, unheated greenhouse, watered thoroughly initially and as needed thereafter to keep the soil evenly moist. The samples were checked daily for three months for germinants. If the germinants could not be identified as they emerged they were transplanted as soon as possible and grown to an identifiable stage. Three control flats, filled with Jiffy-Mix, were placed alongside the soil sample flats to check for contamination. 

The number of contaminant germinants of a particular species in the control flats was subtracted from the total germinants of that species in the field flats on a proportional basis (because there was one control flat for every ten field flats, it was assumed that one germinant of a given species in a control flat could equal up to ten germinants of that species in the field flats) unless this would have given a negative number, in which case a value of zero was assigned. For example, if species A produced six germinants in the field flats and one in a control flat, the raw contaminant count would be one, the contaminant correction factor would be seven and the total for species A would be 0. 

The Seed Rain 
The seed rain was measured by trapping seeds in standard nursery flats filled with Jiffy-Mix. The southern outside edge of the 30 vegetation sampling quadrats was used as a seed rain sampling point. The traps were placed on the ground and seeds deposited in them by any means (except by human intervention, where noted) were considered to constitute the seed rain. 

Trapping began on July 25, 1985. Each trapping period lasted two weeks. At the end of the two week period the traps were removed to the greenhouse and replaced in the field by a new set of traps. This procedure was repeated six other times, the last set of traps being retrieved on October 31, 1985. The procedures employed in the greenhouse were identical to those used to measure the soil seed bank. 

Year Two--Post Disturbance

On May 3, 1986, the canopy was removed from the western one half of the study site. The original research design then called for burning the ground layer and herbiciding the resprouts to prevent an oak resprout and bramble thicket from developing; in practice this was not able to be carried out. 

Vegetation sampling, using the procedure employed in 1985, was conducted on September 16. Species not in quadrats but growing on the site were noted throughout the season. 

Soil Seed Bank and Seed Rain 
Examination of the seed bank and monitoring of the seed rain continued as in 1985. 


The Seed Bank 
The 1985 seed bank produced 100 germinants identified to 19 species (Table 3). Ten of the species (52.6%) were represented by one or two individuals each. The contamination correction factor was nine for five contaminants. Of the remaining 91 germinants, 11 (12.2%) died before they could be identified, leaving a total of 80 identified germinants. Discounting contaminants, the mean number of germinants per soil sample (surface area of 21.5cm2; soil volume of 215cm3) was 3.03 (S.D. 2.24). The number of germinants per sample ranged from zero to eight. Only five soil samples failed to produce at least one germinant. 

The species composition of the 1985 seed bank bore little resemblance to the 1985 standing vegetation (Table 1); only three of the species in the seed bank had representatives in the standing vegetation (38 species). These three species--strawberry (Fragaria virginiana), raspberry/blackberry (Rubus sp.), and honeysuckle--were 15.8% of the 19 seed bank species and accounted for 16.5% of all germinants. 

Perennial forbs and grasses (11 species) accounted for 57.9% of species and 62.5% of the 80 germinants identified; annual and biennial forbs and grasses (5 species) accounted for 26.3% of the species and 18.75% of the germinants; trees and shrubs (3 species) accounted for 15.8% of the species and 18.75% of the germinants. 

The Seed Rain 
In 1985 seven trapping periods produced 178 germinants identified to 14 species (Table 4). The contamination correction factor was 59 for 12 raw contaminants, leaving a total of 119 true field-flat germinants. Of these, 42 (35.3%) died before they could be identified, leaving a total of 77 identified germinants. 

The 1985 seed rain also bore little resemblance to the 1985 standing vegetation. Six of the 14 seed rain species had representatives in the standing vegetation. This was twice the representation of the soil seed bank. These six species (42 germinants) accounted for 55% of the 77 germinants identified. 

Perennial forbs and grasses accounted for one half of the species and 30.7% of the germinants; annual and biennial forbs and grasses (three species) accounted for 21.4% of the species and 12.5% of the germinants; trees and shrubs (4 species) accounted for 28.6% of the species and 56.8% of the germinants. 

Dormant Roots/Stunted Plants 
Examination of soil samples used to measure the seed bank failed to detect any dormant roots. A systematic search of the study site throughout the 1985 growing season failed to detect any stunted prairie/savanna plants. 

The Vegetation Change  
Virtually no changes occurred on the wooded (uncut) portion of the study site. Only one new species, Great Solomon’s seal (Polygonatum biflorum), was recorded on the untreated portion of the site and no species disappeared from this area. The remainder of the discussion therefore is concerned only with the treatment portion of the study site. 

Because the canopy was removed from the treatment (clearcut) portion of the study site, the area no longer had the structure of a southern dry forest although it retained the species composition typical of one. No species disappeared from this area although several species decreased in frequency (Table 7); all the woody species produced vigorous resprouts; none of the herbaceous or shrubby groundlayer species were lost and several new species invaded the area. 

Four categories of vegetation change were noted on the treatment (clearcut) portion of the study site: unchanged, decreasers, increasers and a set of new or invader species that, although they were not picked up in the sampling, nonetheless were observed on the site for the first time after treatment. 

There were those species such as Gragaria virginiana and Prunus serotina (Table 7) whose percent frequency were unchanged, while ten species (eight woodland and two alien) decreased in frequency (Table 7). 

Eighteen species fell into the increaser class (Table 5). That is, after treatment, either their percent frequency increased from 1985(6), they moved from the observed-only category to the sampled category in 1986(7), or they were entirely new to the site in 1986(5). The species that increased in frequency the most was Rubus sp., up from 60% to 86.6%. 

Eleven of the increasers were woodland species, two were prairie/savanna species and five were aliens (Curtis 1959). It should be noted, however, that eight of the 11 are considered by some to also be savanna species (Fifield-Murray 1993, these proceedings). 

There were 16 new, or invader, species that were observed on site for the first time in 1986, after treatment (Table 6). This group includes one woodland species, seven prairie species, and eight exotic species typical of disturbed or weedy habitats. 

By summer’s end the treatment portion was a nearly impenetrable tangle of brambles and waist-high oak resprouts and only a handful of prairie forbs or grasses had appeared. By the end of the 1987 growing season, two years after treatment, the area looked much the same except that the oak resprouts were nearly head-high. The simple treatment of removing the canopy had not led to the instantaneous reappearance of a sand barrens-like community. 

Sources of Invading Species 
The majority of the increasers (13) were species growing on the site at the start of the study and likely increased either vegetatively or from seeds produced on site. Five species were new to the site. One, Chenopodium, was represented in the soil seed bank, and another, Lactuca, was represented in the seed rain; nothing can be said of the likely origin of the other three species: Hackelia virginiana, Cirsium vulare, and Solanum nigrum

Eight of the new or invading species were of unknown origin; although four of the remaining species were represented in the soil seed bank and five in the seed rain (Table 6) we can only say that these seed reservoirs were their likely origin. 

Although the actual contributions of the seed bank and seed rain were not significantly different, the 1985 seed rain made a greater relative contribution to the 1986 standing vegetation than did the 1985 seed bank. Eight of the 19 seed bank species from 1985 (42.1%) were found growing on the site in 1986, whereas 11 of the 14 species from the 1985 seed rain (78.6%) were found growing on the site in 1986. 

What Did the Study Tell Us About Natural Revegetation Following a Disturbance? 
Although a handful of prairie plants appeared, removing the forest canopy of this former savanna community was not enough to stimulate the sudden reemergence of a prairie community. On the study site the vegetation that established itself immediately after the disturbance was composed primarily of those species growing on the site before the disturbance or present in the seed bank, plus those species that were able to make their way to the site through the seed rain during, or immediately after, the disturbance. 

Although an abundant source of prairie propagules was in close proximity to the study site, few prairie species and propagules found their way to the study site disturbance. Most detected in the seed rain were from species growing on the site. 

The management tool of cutting initially created a community dominated by opportunistic early successional species and resprouts of the existing woody species. 

Why Was There No Instantaneous Prairie Response? 
Although several prairie species--Antennaria plantaginifolia, Cyperus filliculmis, Solidago nemoralis, Conyza canadensis, Oenothera biennis, Rhus glabra, and Verbena urticifolia (Table 6)--invaded the site and an existent species, Tradescantia ohiensis, visibly increased in numbers, the immediate change was not enough to classify the site as a savanna nor to hope that without sustained manipulation of the vegetation one will emerge. 

While there is no reason to expect that every site should undergo the rapid transformation to prairie/savanna documented by others, it is interesting and perhaps instructive to speculate why this particular site did not do so. 


At our study site a number of factors may have worked against such a transformation. The study site was relatively small compared to the height of the surrounding vegetation. This increased the amount of shade on the site and may have prevented the germination and establishment of a number of prairie/savanna species. The vigorous resprouting of brambles and other woody species increased competition for other resources. 


The reason for the failure of prairie/savanna to reappear immediately is probably not due entirely to the absence of propagules in the soil seed bank. Although the seed bank did not have a particularly diverse species composition nor a great number of prairie/savanna species, it did show the potential to make a significant contribution to the standing vegetation after the disturbance. In terms of both number of germinants and number of species, prairie/savanna species were the largest category of species in the soil seed bank (Table 3), comprising 45% of the germinants and 42% if the species. 

The degree to which the site characteristics described above aided or prevented recruitment from the soil seed bank are unknown. The role of fire (absent in the study) in influencing recruitment from the soil seed band should be investigated in further studies of this type. 


Despite the close proximity of the West Grady Knoll savanna which has an abundant source of prairie/savanna propagules, few off-site prairie/savanna species were detected in the seed rain study. Discounting the seeds of Quercus (an on-site species), only two off-site prairie/savanna species (14% of seed rain species) and 13 off-site prairie/savanna germinants (17% of seed rain germinants) were detected in the seed rain. Why? Was this a sampling artifact, ineffective seed traps, or something else at work? A separate seed rain study (Glass 1988) sought to find out.  

A series of seed traps identical to those used in this study and covering the comparable sampling period in 1986, were set out on a transect from the West Grady Knoll savanna into the study site. The study was designed, in part, to measure the movement of seeds along the transect from Knoll into the study site. It was documented that the seed traps were capable of catching a variety of prairie species propagules. The seven seed traps in the Knoll savanna portion of the transect caught 12 prairie/savanna species that yielded 841 germinants. Pertinent here is the number of prairie/savanna species and germinants detected in the study site portion of the transect. Here, eight seed traps caught nine of the 12 species and yielded only 50 germinants. 

Variables such as a species dispersal ability and prevailing winds at dispersal time obviously played a role in the movement of seeds along the transect into the study site although these factors were not investigated. One factor that may have played a great role in preventing the penetration of more species and greater numbers of propagules into the study site was the dense vegetation wall created by the vigorous resprouts of woody species. 

HOLDOVERS  back to top

The absence of prairie holdovers may be the major reason that the disturbance-induced vegetation changes did not result in a savanna-like community. Holtz (1985) was able to document the transformation of an oak woods dominated by black/Hill’s oak and black cherry to a plant community similar to dry-mesic prairie and oak barrens by cutting and burning. 

Holtz concluded that the vegetation change was due to “the invasion of new species, the elimination of some species and the continued presence of others.” She speculated that some species may have emerged form the seed bank, and others invaded from nearby communities via the seed rain. An important component of the vegetation change appeared to be the presence of eight grass and 21 forb species typical of prairie or savanna communities (Curtis 1959) that were growing on the site while it was a woods and before the cutting and burning. The fact that our site had only two forbs typical of prairie or savanna communities, Rosa sp. and Tradescantia ohiensis, may account for its failure to be transformed to a prairie-like community. In fact, as measured by number of prevalents and major dominants (Curtis 1959), the Holtz site was similar to an oak barrens and dry mesic prairie while our site was similar to a southern dry forest. 


Curtis, John T. 1959. The vegetation of Wisconsin: an ordination of plant communities. University of Wisconsin Press, Madison, Wisconsin. 

Fassett, Norman G. 1951. Grasses of Wisconsin. University of Wisconsin Press, Madison, Wisconsin. 

Glass, Steve. 1988. An investigation of the mechanisms responsible for the re-vegetation of a degraded oak barrens after removal of the canopy. Masters Thesis, University of Wisconsin, Madison. 

Gleason, H.A. and Arthur Conquist. 1963. Manual of vascular plants of Northeastern United States and adjacent Canada. Van Nostrand Company, New York. 

Holtz, Signe L. 1985. cutting and burning a degraded oak barrens: management techniques that simulate natural disturbance. Masters Thesis, University of Wisconsin, Madison. 

Holtz, Signe L. and E.A. Howell. 1983. Restoration of grassland in a degraded woods using the management techniques of cutting and burning. Proceedings of the Eighth North American Prairie Conference. 

Lippert, Robert D. and H.H. Hopkins. 1950. Study of viable seeds in various habitats in mixed prairie. Transactions of the Kansas Academy of Sciences 5: 355-364. 

Shimek, Bohumil. 1931. The relation between migrant and native flora of the prairie region. University of Iowa Studies in Natural History 14: 10-16. 

Soil Survey of Dane County, Wisconsin. 1972. U.S.D.A. Soil Conservation Service and Research Division. College of Agriculture and Life Sciences, University of Wisconsin, Madison, Wisconsin. 

Vogl, Richard J. 1964. Vegetational history of Crex Meadows, a prairie savanna in northwestern Wisconsin. American Midland Naturalist 72: 157-175. 

White, Alan S. 1983. The effects of thirteen years of annual prescribed burning on a Quercus ellipsoidalis community in Minnesota. Ecology 64: 1081-1085. 

Zimmerman, James. 1986. Lecturer, Department of Landscape Architecture, University of Wisconsin, Madison, Wisconsin. 


Begin Site Footer

EPA Home | Privacy and Security Notice | Contact Us