DOT Project Number:  90-00-LRTF-202

Fiscal Year:  2002

Award:  $25,000.00

Principal Investigator:  Dr. Thomas Rosburg, Department of Biology, Drake University, thomas.rosburg@drake.edu

Summary Report:

SEED BANK POPULATIONS OF A RECONSTRUCTED PRAIRIE

Seed banks, the reserve of viable seeds found in or on top of soil, play an important role in plant communities. Over time the seeds produced by plants move into and downward through the soil by gravity, animal activity, and weather. The accumulation of a seed bank allows plant communities to recover after disturbances. Investigating and analyzing the seed bank can supply information on the vegetative history of the community, help predict the future composition of the community, and provide insight on the long-term viability and biology of the aboveground plant life. Although seed banks of prairie ecosystems have been well studied, very little information exists on seed banks of reconstructed prairie. Identification of the seed bank populations of a prairie reconstruction provides a way to judge the success of restorations and establishes baseline data to track the development of the prairie over time.

Seed Bank Methods
The seed bank analysis was performed by sampling the soil of the reconstructed prairie sites, germinating the seeds, and using the seedlings to determine species identification (seedling assay) of seeds in the seed bank. Soil samples of the seed bank were collected on March 26 and 27, 2000, prior to any chance of natural germination. Each of the 15 study sites was sampled with 20 sub-samples, five sub-samples each were made in a stratified-random manner from each of the four treatment plots. A sub-sample consisted of the soil plug obtained with a bulb planter. Soil plugs were 3 cm deep and 6 cm in diameter. The 20 sub-samples obtained from each site were combined in a bag and mixed to distribute the seeds throughout the complete sample (with a total volume of about 1,700 cm3). Soil samples for each of the 15 sites were placed in cold storage in order to keep the seeds in dormancy until the seed assay in the fall of 2000.

In October, 2000, the soil samples were removed from cold storage and individually sieved through a hardwire mesh screen, approximately 6x6 mm, to remove large matter such as roots or gravel. From each seed bank sample, a 500 cm3 sub-sample of seed bank soil was removed and layered 1½ cm deep into each of three 18x18 cm trays half full of sterilized potting soil. The three trays planted for each site created three replicates for observing seedlings. Additionally, three control trays consisting only of the sterilized potting soil were used to control for seeds that might be present in the potting soil mix. The samples were placed in a greenhouse under optimal light and moisture conditions to encourage germination. Trays were arranged in a block configuration, with the replicates from each site in one of three blocks that controlled for variation in temperature, light and air movement within the greenhouse. Seedlings were allowed to grow until identification was possible and were then removed to allow room for additional seeds to germinate. If seedlings were not identifiable and needed to be removed due to crowding, several specimens were potted in separate containers and allowed to continue growth until they could be identified. By 1 March 2001, nearly all seed germination had occurred. To insure equal amounts of germination time for all trays and time for growth of seedlings for identification, all seedlings germinated after 1 March 2001 were pulled and counted as unidentified dicot or monocot. By 15 May 2001, nearly all of the seedlings still present could be identified and counted. All of the remaining trays were processed. About 25 specimens that needed more time for growth were transplanted into an outdoor bed and labeled. Identification of these specimens was made by 2 July 2002, approximately 20 months alter the seedling assay was started.

Data analysis focused on two types of information. One was a description of the seed bank composition and the representation of seeded prairie species. Raw seedling counts were converted to reflect seed density (seeds/m2/3 cm depth). A second goal was a comparison of the seed bank with the extant vegetation. This was done in two ways. One approach utilized Detrended Correspondence Analysis to ordinate the seed bank samples with the extant vegetation in 1999 and 2000*. Density of seeds in the seed bank and density of stems/caudices in the vegetation were used as measures of abundance, but because there was a large discrepancy in the magnitude of the data (some seed densities exceeded 5000/m2), all data were converted to relative density. A second approach in the comparison of the seed bank with the extant vegetation was to use Pearson Correlation to measure association between seed bank composition and vegetation at each site. To assure that the sample size was limited in size and represented only the most common species, only the top 30 species in the seed bank and the top 30 species in the vegetation were included.

Results
There was no difference in germination rates among the three greenhouse blocks, thus potential variation in light and temperature due to location in the greenhouse was not factor affecting germination (Friedman Repeated Measures ANOVA p=0.651).

A total of 2,696 seedlings were observed in the study. Germination rates of all seedlings was fairly high, over the first three weeks nearly 125 seedlings germinated per day, resulting in over 95% of the total germination. A total of 66 taxa were observed, but 4 of those were accounted for by the potting soil controls, resulting in 62 taxa represented in the seed bank samples. Fifteen species were observed that were seeded in the reconstruction, but three of those are exotic species seeded by mistake, thus 12 prairie species out of 79 species included in the seed mixes have established a seed bank or are still viable four years after the seeding (15%).

The most abundant species in the seed bank was carpetweed, which accounted for about 27% of all seeds counted and attained a mean seed density five times greater than any other species (means calculated for the sites where present). The most widespread species, in terms of presence on the study sites, was frost aster, one of the seeded species and which was present on all 15 sites. However, frost aster is a typical species of old-fields and successional grasslands and most likely was a resident species of the I-35 roadside. It is likely that many of these seeds originate from a resident seed bank. The seeded prairie species with the most success in establishment of a seed bank is Black-eyed susan, which resides in the seed bank on 12 of the 15 sites and has the ninth highest mean seed density.

Three species in the seed bank are seeded exotics. These three species represent either accidents or mistakes in the reconstruction and were discussed in an earlier report with eight other species that should not have been in the seed mixes. Although their abundance in the vegetation has been decreasing in recent years, the fact that they are among the more common species in the seed bank verifies that these problems have not gone away. Common wormwood is the most widespread and dense of these, followed by decurrent false aster (present at 3 sites) and ashy sunflower (present at 1 site).

Among the remaining 47 species about half are native (24) and half are exotic (22). Among these native species, nearly 20 of them are ruderals or early successional grassland and old-field species. Altogether, about 40 of the 62 species (66%) observed in the I-35 reconstruction are ruderal or early successional species that are favored by disturbance. Such species predominate most seed banks because they have high seed longevity as a mechanism to find disturbances in time. Those species with higher seed longevity accumulate in seed banks over time. Typically, the longer a community is disturbance-free, the more dissimilar the extant vegetation and seed bank become.

The prairie species in the I-35 seed bank are represented by four families, the asters (7 species) grasses (3 species), evening primrose (1 species) and mint (1 species). All three of the principal grasses seeded – big bluestem, little bluestem, and Indiangrass – are present in the seed bank. Their occurrence is limited to three sites or less and their mean seed densities where present are less than 30 seeds/m2/3 cm. Among the aster family, five are Aster species and account for all but one of the asters seeded. New England aster and smooth aster are present at about 35% of the sites and have mean seed density of nearly 60 seeds/m2/3 cm, while sky blue aster and heath aster were observed at one site. Black-eyed susan and tall boneset are the other two aster family species in the seed bank. Common evening primrose and wild bergamot represent the last two families. Both are have mean seed densities of at least 50 seeds/m2/3 cm, but evening primrose is more widespread occurring at nearly 50% of the sites while wild bergamot was observed at only 20% of the sites.

Total mean seed density for the individual sites ranges from 293 seeds/m2/3 cm at 767S to 12,247 seeds/m2/3 cm at 598N. Most sites are in the range of 1500 to 4500 seeds/m2/3 cm. The low density site (767S) is a dry, gravelly environment with relatively low productivity and diversity. It is one of a few sites to have supported prairie species prior to reconstruction (prairie fleabane and switchgrass). Low productivity (and therefore seed production) is possibly one reason for the low seed density in the seed bank. The biomass data collected in 1999 indicate its productivity was 330 g/m2, the second lowest amount among the sites. The high density site (598N) is located on a broad sandy ridge, possibly formed from a dune created from sand blown out of the nearby Skunk River valley. Slope is fairly level and the soil is a loamy sand. Productivity is much higher; about 500 g/m2 was recorded from the 1999 samples. It also supported some resident prairie species, for example Scribner’s panic grass and field horsetail. The main reason however, for the high seed density at 598N, is that the bulk of the seeds (just over 80%) are from one species – carpetweed – a sand loving species that is apparently capable of high seed production and longevity.

The species compositions of the seed bank samples are very different from the extant vegetation. Axis one of the DCA ordination represents the most important gradient in the variation of vegetation and seed bank species composition. The extant vegetation (with low axis 1 scores) is represented by greater abundance of perennial prairie species in the samples. As axis 1 scores increase towards the seed bank samples, perennial prairie species decrease and are replaced by a greater abundance of ruderal, disturbance species. The seed bank samples with the highest ruderal component are sites 598N and 599S (highest axis 1 scores), which both occur on sandy soil and also have the first and second highest total seed density in the seed bank due to very high densities of carpetweed.

The second most important gradient in the variation of species composition is represented on DCA axis 2. For the most part this variation only occurs in the seed bank samples and is most manifest in the difference in seed banks between site 767S (high axis 2) and 788S (low axis 2). The species that are associated with the ends of axis 2 and produce this gradient are more difficult to characterize. One ecological pattern that seems to fit axis 2 is that the high score species appear to be more tolerant of stressful environments (similar to a low level chronic disturbance) while the low score species tend to respond more to disturbances that open up space in crowded communities. Sites 767S, 739N, and 618S, which are located on the upper half of axis 2, can all be characterized as stressful environments. Low fertility, coarse textured soils occur at 767S, very steep and well-drained soils characterize 618S, while flooded conditions often occur at 739N. These three sites had the lowest productivity in the 1999 biomass samples (all less than 360 g/m2, mean of 328 g/m2). At these sites, space for colonization and establishment results from stressful abiotic conditions restraining growth. At sites 788S, 713S, 752S and 754S, which all have low scores on axis 2, fertility and productivity are high (mean of 535 g/m2, highest productivity among all sites of 655 g/m2 observed at 752S). Colonization and establishment require disturbance to open up the canopy and remove existing vegetation. The difference in the colonization and establishment environment – tolerance for stressful environments versus ability to quickly exploit canopy gaps – is reflected in the seed bank composition.

The difference in variation in species composition between extant vegetation and seed banks is very clear. The area encompassed by the seed bank samples is 6 to 7 times greater than the area encompassed by the extant vegetation. The much larger variation in seed banks occurs despite a significantly lower average richness per site in the seed bank than in the extant vegetation (seed bank mean of 18.7 versus vegetation means of 47.8 for 1999 and 43.1 for 2000, One Way Repeated Measure ANOVA, p<0.001). Likewise, the cumulative richness of the seed bank communities (62) is less than half of the cumulative richness of the vegetation samples (149). The seed bank samples realize a much higher level of uniqueness in species composition utilizing a much smaller pool of species than does the vegetation. Each site’s vegetation is fairly similar to vegetation at the other sites, most likely because each site’s vegetation has developed from very similar seed mixes over the last four years. Unlike the vegetation, each site’s seed bank is generally more unique and specific to that site (exceptions would be the trio of 601N, 691S, and 726N and the pair 754S and 752S). Because seed banks are a memory of the vegetation, the varied histories of the sites over the last 20 to 30 years could lead to greater uniqueness of the seed banks. Because a prerequisite of seed banks is that species presence requires adaptation for seed dormancy and longevity, there are fewer species capable of forming seed banks than there are species in the vegetation. A high ratio of heterogeneity to species richness pool seems to be an important ecological distinction between seed bank communities and their associated aboveground vegetation.

Nine of the 62 plant species observed in the seed bank were not observed in the vegetation at the sites in either 1999 or 2000. These included velvet leaf (Abutilon theophrasti), bitter cress (Cardamine species), tansy mustard (Descurainia pinnata), stinging nettle (Urtica dioica), barnyard grass (Echinochloa species), lovegrass (Eragrostis species), carpetweed (Mollugo verticillata), Norwegian cinquefoil (Potentilla norvegica), and mullein (Verbascum thaspis). The last five species have been observed in the project area in inventories during other years, thus there is a known seed source for them in the project area. However, one of two options must exist for the other four species. Either they have an historic presence in the vegetation and have maintained presence only in the seed bank, or else they are dispersing seed rain into the reconstruction from adjacent areas. Velvet leaf and stinging nettle could easily be dispersing in from adjacent areas, whereas bitter cress and tansy mustard are both small mustards that inhabit sandy areas and could have an historic presence on some of the sites.

The position of the vegetation samples in the ordination are somewhat aligned along axis 2 in a pattern that reflects the seed banks. Sites 767S and 739N have the highest axis 2 scores in the vegetation samples and sites 788S and 752N have the lowest scores, a pattern which is also reflected in the seed banks. The same species composition gradient pulling the seed bank samples apart along axis 2 is also acting on the vegetation, with the exception that these species do not have as high a relative density in the vegetation as in the seed bank so it’s not as important. Thus there is evidence for some association between the vegetation and the seed bank. Seeds germinate and become plants and plants reproduce and form seed, so some association between seed banks and vegetation is natural. The more recruitment there is from the seed bank, the more similar the seed bank and vegetation become (positive association). The longer a community is disturbance free allowing seed bank and vegetation to diverge, the dissimilar the seed bank and vegetation become (negative association).

A correlation analyses between the seed bank and the two vegetation samples for each site demonstrated no association exists for any site. Since these sites are reconstructed prairie and were disturbed and planted with seed in 1996, the lack of any positive associations has a couple interpretations. One is that there is little if any viable seed remaining from the seed mixes planted. In theory, seed bank samples during the first years of the reconstruction should have been somewhat similar to the vegetation, as recruitment of the planted seed took place and established the vegetation. As the planted seed is lost from the seed bank, either through recruitment or death, the seed bank would become more dominated by non-seeded resident species and lose similarity with the vegetation. The lack of association between seed bank and vegetation could result if there is very little if any of the planted seed left. It is impossible to know the source of the seed of prairie species that was found in the seed bank. It may have originated on site from seed rain by established plants. In this study that is likely true for the grasses (big bluestem, little bluestem, indiangrass) and some of the forbs quick to establish (wild bergamot, New England aster, black-eye susan, evening primrose). Or, for slower to establish species that have not produced seed, the presence of their seed indicates that viability in the planted seed has been maintained. Smooth aster, sky blue aster and heath aster may be examples in this study. Nevertheless, we can be reasonably sure that most of the planted seed has either germinated (good) or has lost viability (not so good).

Since seed banks tend to be dominated by ruderal, disturbance-adapted species (and the I-35 sites are very good examples of that), another implication in the lack of association between the seed bank and vegetation is that the reconstructed prairie has developed without too much negative input from the seed bank. The necessity and reliance on seed germination to achieve results in prairie reconstruction does expose the area to the whole seed bank and potential problems with undesirable species. At four years of age, it is good to see that none of the sites has a positive association with its seed bank. If there was a weedy stage, the lack of an association between the seed bank and vegetation is evidence that the prairie has matured beyond it. Careful inspection of the vegetation samples in the ordination reveals that all but two of the sites exhibited vegetation change between 1999 and 2000 that made them less like the seed banks (movement along axis 1 from right to left, or decreasing axis 1 score). In fact there was a significant difference in the mean axis 1 scores for 1999 and 2000 (1999=36.1, 2000=21.9, paired t-test, p=0.001). This is good evidence that the prairie is continuing to diverge from the seed bank. Since the reconstruction does not contain a significant prairie component in the seed bank, the divergence is a positive sign of a maturing prairie. Eventually it would be ideal to see a greater representation of prairie species in the seed bank as assurance of long term stability in the community and its populations.

Summary
Although representation of prairie vegetation in the seed bank is minimal, there is evidence that some species have either become established in the seed bank through growth and reproduction or are maintaining viable seeds from those planted in 1996. All 12 of the prairie species observed in the seed bank have also been observed in flower, so for most of them seed production has occurred and is likely the source of the seed observed in the seed bank. Viability of planted seed is also a possible source of the seed since most plant species should have seed viability of 4 years. Some of the asters (azure aster, heath aster, smooth aster), which have not flowered as profusely as the other species, may be examples. These 12 prairie species account for just over 19% of the species observed and 15% of the seeds counted. Compared to native prairies, the establishment of prairie seed in the seed bank seems respectable especially for a four year old reconstruction. Seedling assays do not find every species in the seed bank due to the lack of germination by some species, so these data represent the minimal levels of seed bank composition.

The importance of avoiding the inclusion of non-native species in the seed mix is highlighted by the unfortunate fact the three non-natives in the seed mix have established long-term viability in the project area through presence in the seed bank (common wormwood, false decurrent aster, and ashy sunflower). Although these species have decreased in abundance in the prairie community since planting, they are still present and among the most common species in the seed bank. Depending on their seed longevity and the amount of disturbance that occurs allowing for periodic growth and seed production, mistakes like this in planting may not go away.

Overall, the seed banks are very dissimilar to the vegetation and contain a great deal more variation than the vegetation. Each site’s seed bank is more unique and specific to the site than the vegetation. One species gradient that helps explain some of this greater variation is that some sites have seed bank species that tolerate stressful environments while other sites are dominated by species that are adapted to disturbances. This characteristic is more readily seen in the seed bank than in the extant vegetation.

The seed bank analysis provided evidence that the extant prairie communities at the study sites are becoming more mature and less like their seed banks, which are mostly represented by resident species. Although seed banks in the project area still reflect the non-prairie pre-reconstruction roadside community, the extant vegetation of the prairie reconstruction is continuing to improve and diverge away from the pre-reconstruction vegetation.

*Note: To view the DCA ordinations referred to in this study, contact the principal investigator.