Soil seed banks and the potential restoration of forested wetlands after farming

Authors


*Current address and correspondence: National Wetlands Research Center, 700 Cajundome Boulevard, Lafayette, LA 70506, USA (tel. +1 337 266 8618; fax +1 337 266 8586; e-mail beth_middleton@usgs.gov).

Summary

  • 1Changes in farming practice provide an opportunity to restore once extensive forested wetlands on agricultural land. In some parts of the world, however, it has proved difficult to restore the full complement of plant species through natural regeneration. Similarly, the restoration of forested wetlands by replanting has often resulted in ecosystems of low diversity. Better methods of restoring these important ecosystems are now required and baldcypress swamps provide an opportunity to investigate alternative approaches to the restoration of forested wetlands. This study examined the composition of seed banks of farmed fields to determine their value in restoring swamps in the south-eastern United States.
  • 2A seed bank assay of soils from baldcypress swamps was conducted to determine the extent to which seeds are maintained during farming for various lengths of time. Soils from swamps that were farmed for 0–50 years were collected near the northern boundary of the Mississippi Alluvial Valley along the Cache River, Illinois. Soils were placed in a glasshouse setting in flooded and freely drained conditions, and the numbers and species of seeds germinating were recorded.
  • 3Woody species including trees, shrubs, and vines were poorly represented in seed banks of both farmed and intact sites (51 and 9 sites, respectively). Missing dominants in the seed banks included tree species with short-lived seeds such as Taxodium distichum and Nyssa aquatica. Cephalanthus occidentalis constituted the most abundantly dispersed seed of all woody species.
  • 4Herbaceous species were well represented in the seed banks of both farmed and intact swamps (species richness of 207 vs. 173 species, respectively) suggesting that herbaceous species may live longer than woody species in seed banks. Few of the herbaceous species decreased in seed density in seed banks with time under cultivation, although seed density was lower at sites that had not been farmed. Species that relied on vegetative organs for dispersal were absent in the seed banks of farmed sites including Heteranthera dubia, Hottonia inflata, Lemna minor, Lemna trisulca and Wolffia columbiana. These species may require active reintroduction during restoration.
  • 5Synthesis and applications. Both restoration ecologists and managers of nature conservation areas need to be cognisant of seed bank and dispersal characteristics of species to effectively restore and manage forested wetlands. In the case of baldcypress swamps, critical components of the vegetation are not maintained in seed banks, which may make these floodplain wetlands difficult to restore via natural recolonization. Ultimately, the successful restoration of abandoned farm fields to forested wetlands may depend on the re-engineering of flood pulsing across landscapes to reconnect dispersal pathways.

Introduction

Vast tracts of the wettest agricultural fields have been abandoned world-wide, hence knowledge of their restoration potential as nature conservation areas is becoming important (Haynes & Moore 1988; Galatowitsch & van der Valk 1995; Jensen 1998; Middleton 1999). Since 1937, 11·8 million ha of forested wetland in the Lower Mississippi River Alluvial Valley have been converted to agriculture (MacDonáld, Frayer & Clauser 1979). Similarly in Europe, more than 50% of original wetlands have been lost in France, Germany, Greece, Italy, the Netherlands and Spain, mostly due to drainage for farming (Jones & Hughes 1993). Some of these converted floodplains are no longer farmed, at least partly due to recent changes in agricultural market demands (Newling 1990). In the south-eastern United States, 89 000 ha of these floodplains will be reforested from 1999 to 2003 (King & Keeland 1999). The study of seed availability from remnant seed banks and dispersal is particularly relevant because of the need to better understand the dynamics of forest regeneration on abandoned farmland.

The success of converting these farmlands into forested wetlands with functional qualities and biodiversity levels similar to those originally present depends on the availability and recruitment potential of seeds and propagules within the hydrological framework provided to these landscapes (Middleton 1999). The self-design approach to restoration suggests that species re-establishment is a fate determined by the environment and that succession ultimately determines species composition (Mitsch et al. 1998). Self-design fails to acknowledge the problem that pivotal species may be absent in restoration sites due to a lack of seeds or propagules. Sedges (Carex spp.) are missing from the seed and propagule banks of sedge meadows after farming and seeds of these species can not disperse to sites in the current hydrologic setting (Galatowitsch & van der Valk 1996a). Overwhelmingly, European and North American studies suggest that important components of the vegetation are commonly missing from ex-arable fields or grazing lands, so that formerly dominant species may not spontaneously reappear after cultivation or grazing ceases (Wienhold & van der Valk 1989; Bullock et al. 1994; Prach & Pyšek 1994; Galatowitsch & van der Valk 1995, 1996a,b; Bakker et al. 1996; Hutchings & Booth 1996; van der Valk 2000; Coulson et al. 2001; Verhagen et al. 2001). Flood pulsed hydrology may be the missing factor that can promote aquatic dispersal and other landscape-level processes in floodplain restoration projects (Middleton 1999).

Forested wetlands that resemble the original forests have naturally regenerated on some abandoned farmland. Riverine forests along the Kinabatangan River in North Borneo became diverse rain forests after the Dutch abandoned tobacco plantations in the 1850s (Spencer 1995). Coastal mangrove restoration is often more successful without rather than with planting, but only if the hydrology is properly restored (Buchanan 1989). Riparian vegetation and semi-arid Populus forests of the western North America have an ability to naturally revegetate, but only under flood pulsed conditions (Ellis, Crawford & Molles 2002; Scott & Auble 2002; Stromberg & Chew 2002).

Whether or not species will naturally re-establish in formerly farmed baldcypress Taxodium distichum (L.) Rich. swamps depends to a large extent on the availability of seeds in the seed bank and/or their ability to disperse to sites. Because most of the dominant species in these systems are aquatically dispersed (Schneider & Sharitz 1988; Middleton 1999), the re-establishment of the flood pulse on these formerly farmed sites may be a key issue in their restoration (Middleton 1999). Indeed, restoration failure is often blamed on inadequate seed supply (van der Valk & Pederson 1989; Allen 1990; Newling 1990; Reinartz & Warne 1993; Galatowitsch & van der Valk 1995, 1996a; Middleton 1999, 2000, 2002).

Restored floodplain forests in the south-eastern United States generally do not resemble those that were originally on the sites (Allen 1997; Ouchley et al. 2000). Seasonally flooded baldcypress forests are species-rich, with as many as 48–62 woody species (trees, shrubs and vines; Nixon, Willet & Cox 1977; Robertson, Weaver & Cavanaugh 1978), but the forests that spontaneously redevelop on these sites after agricultural abandonment are dominated by a few wind-dispersed species (Newling 1990; Allen 1997; Harmer et al. 1997; Jensen 1998). Many abandoned farmed floodplains have been replanted with species that are useful to wildlife species, such as Quercus palustris Muenchh. While it is generally assumed that diversity will increase over time, instead these forests have permanently low diversity because new species cannot invade where planted trees have already established. A much higher number of species should be seeded at sites for planting to result in diverse forests (Allen 1997).

A critical theoretical issue in restoration ecology is to determine if the presence of specific species is necessary for successful restoration. Without question, dominant species are important because these exert a controlling force in wetland function (e.g. productivity, trophic dynamics, carbon and nutrient cycling; Junk, Bayley & Sparks 1989; Smock & Gilinsky 1992; Bayley 1995; Nielson & Chick 1997; Middleton 1999). Whether or not specific dominants are important is a part of the developing debate over the functional equivalence of restored wetlands particularly in regard to production, standing crop biomass, nutrient cycling and soil carbon storage (Mitsch & Jørgensen 1989; Erwin 1991; Craft et al. 1999; Zedler 2000). The importance of non-dominant species is less certain. However, biodiversity levels also may be related to ecosystem functioning (Loreau et al. 2002).

The objectives of this study were to examine the natural restoration potential of farmland that originally was seasonally flooded baldcypress swamp in the northern portion of the Mississippi Alluvial Valley. If important components of these systems are missing, then it is unlikely that restoration sites will spontaneously develop into baldcypress swamps, unless dispersal vectors (e.g. wind, water and animals) are capable of reintroducing species. The composition (density and species richness) of seed banks in swamps that were either intact or farmed for various lengths of time was examined. The hypothesis that the species richness and density of seeds in the seed bank would decrease as the length of time of farming increased was tested.

Methods

study sites

Baldcypress swamps lie along rivers and streams on floodplains of the Gulf Coastal Plain of North America from south-eastern Texas, north along the Mississippi River to southern Illinois (Conner & Buford 1998), and extend east to the Atlantic Ocean (Mitsch & Gosselink 2002). The Cache River watershed was comprised of approximately 250 000 ha of forested wetland prior to its extensive modification for agriculture (Ugent, Tindall & Doorenbos 1981). The Post Creek Cutoff was constructed by diverting water from the upper watershed of the Cache River to the Ohio River in 1916 to develop floodplains in southern Illinois (Cache River Drainage Commissioners of Illinois 1905; Muir et al. 1995). Agricultural development expanded during the 1950s, and by the late 1980s only about 50% of the original forested wetland remained. The Lower Cache downstream of the Post Creek Cutoff was extensively modified by farming (south of 37·3° N latitude; Fig. 1) (Muir et al. 1995). Much marginal agricultural land was abandoned after the crash of the soybean market throughout the Mississippi Alluvial Valley and became available for restoration (Newling 1990), although 70% of the catchment is presently used for agriculture (Wetlands International 2003). Government and private agencies plan to restore much of the river corridor of the Cache River watershed to baldcypress and bottomland forest as part of the Cypress Creek Wildlife Refuge within the North American Waterfowl Management Plan (USFWS 1990).

Figure 1.

Site locations for seed bank samples locations along the Cache River, southern Illinois. Base map from Illinois Geographic Information Systems (1996) in ArcView 3·2 (1999). Sites 1–30 and 31–60 were sampled in February to April 1992 and 1993, respectively. Sites farmed for 1–3 years were 23, 29, 49 and 58; 5 years: 28, 30, 34, 35, 40 and 43; 10 years: 4, 22, 27 and 31; 13–15 years: 2, 3, 7, 9, 10, 18, 24, 37, 41, 42, 57, 59 and 60; 30 years: 46 and 48; 50 years: 17 and 19–21. Sites never farmed were 1, 5, 6, 8, 25, 32, 33, 36, 38, 44, 45, 47, 50–52 and 56.

In this study, the seed banks of both farmed and intact baldcypress swamps were compared throughout the entire 120-km reach of the Cache River and its tributaries, from West Vienna (tract owned by WestVaco; Site 53) to Mounds at the confluence of the Cache and the Ohio Rivers in Illinois (Fig. 1). The intact baldcypress swamps in this study included Deer Pond, Little Black Slough, Heron Pond, Goose Pond, Section 8 Woods, Buttonland Swamp, Stallings Tract, Bellrose Tract and Horseshoe Lake (Fig. 1). Buttonland Swamp and Horseshoe Lake are intact swamps that lie below the Post Creek Cutoff along the lower Cache, which are both fed by tributaries and have water control structures. For the most part, the rest of the lower Cache was drained at the time of this study, and in many places farmed to nearly the edge of the river. For swamps in the upper Cache above the Post Creek Cutoff, drawdown is common during the late summer (B.A. Middleton, unpublished data). Portions of the channel of the Cache River as well as some of its tributaries are downcut (Sengupta 1995).

Dominant species of these swamps included Taxodium distichum, Nyssa aquatica L., Cephalanthus occidentalis L. and Fraxinus pennsylvanica Marsh. In tracts farmed for various lengths of time (Table 1), the dominant crop species was Glycine max (L.) Merr. or in some cases, Zea mays L. with common field weeds including Cyperus erythrorhizos Muhl., Gratiola neglecta Torr., Ipomoea lacunose L. and Xanthium strumarium L. Fields planted with G. max and Z. mays were ploughed every year, and a variety of chemicals were used including insecticides, herbicides and fertilizers. None of the farmers interviewed indicated that they were using organic farming methods in these fields (pers. obs.).

Table 1.  Mean density m−2 ± SE of species exceeding 14 seeds or spores m−2 as based on the maximum number of seeds germinating in freely drained vs. flooded soils from farmed and unfarmed baldcypress swamps, Cache River Watershed, Illinois. Contrasts (t and P values) are based on anova comparisons of log transformed data for unfarmed vs. farmed sites. For the contrasts, the alpha value (P < 0·05) was adjusted to P < 0·002 to accommodate experimentwise error
SpeciesPercentage frequencyMean DensityContrast
UnfarmedFarmedtP
Amaranthus retroflexus L.0·65   9·4 ± 4·3 84·5 ± 31·9 −1·9    0·052
Ammannia auriculata0·75  27·9 ± 14·9141·6 ± 28·2 −2·0    0·043
 Willd.
Ammannia coccinea0·93 147·1 ± 87·2206·3 ± 48·3 −0·3    0·789
 Rottb.
Callitriche heterophylla Pursh.0·05   7·3 ± 7·3 37·6 ± 37·4  3·1    0·002*
Cyperus erythrorhizos0·55  90·9 ± 32·3 21·6 ± 8·1  2·92    0·004
Cyperus odouratus L.0·63  72·9 ± 37·6 18·6 ± 4·5  2·92    0·004
Digitaria sanguinalis L. (Scop).0·38  10·3 ± 7·2 53·8 ± 21·1 −1·39    0·166
Echinochloa crus-galli0·72  41·2 ± 12·1 95·4 ± 24·4  0·16    0·874
 (L.) Beauv.
Eleocharis obtusa0·87  21·4 ± 11·0 93·4 ± 19·6 −1·5    0·127
 (Willd.) J.A. Schultes.
Gratiola virginiana0·621254·1 ± 434·5414·8 ± 110·9  4·7< 0·001*
Juncus biflorus Ell.0·67  72·5 ± 37·9429·8 ± 110·0 −0·5    0·574
Juncus marginatus Rostk.0·58  79·3 ± 36·2388·8 ± 86·4 −0·8    0·450
Krigia caespitosa0·28   1·3 ± 0·7 18·7 ± 12·2 −0·45    0·654
 (Raf.) Chambers
Leucospora multifida0·22   0·4 ± 0·4 33·5 ± 24·7 −1·56    0·119
 (Michx.) Nutt.
Ludwigia palustris0·47  28·3 ± 11·1 67·5 ± 35·9  1·9    0·057
Ludwigia polycarpa0·70 109·3 ± 61·4 75·0 ± 21·9  2·0    0·042
 (Short & Peter)
Muhlenbergia frondosa0·17  72·9 ± 50·8  2·7 ± 1·5  7·60< 0·001*
 (Poir.) Fern.
Muhlenbergia schreberi0·43  84·9 ± 57·8107·3 ± 49·7  2·1    0·039
 J.F. Gmel.
Neeragrostis reptans0·50   7·7 ± 5·1121·9 ± 34·1 −3·3    0·001*
Oenothera fruticosa L.0·50 165·5 ± 121·5 99·3 ± 23·3  2·5    0·014
Paspalum fluitans0·53 115·8 ± 38·1 68·0 ± 25·5  3·9< 0·001*
 (Ell.) Kunth
Penthorum sedoides L.0·68  92·2 ± 36·3 49·9 ± 20·1  2·3  0·022
Ricciocarpos natans0·45 201·1 ± 97·2 62·1 ± 25·2  4·1< 0·001*
 (L.) Corda
Veronica peregrina0·77 178·8 ± 144·6115·2 ± 16·3< 0·1    0·976
Other1·001307·7 ± 527·8682·5 ± 99·4  

regional seed bank collection

In February to April 1992 and 1993, seed banks were collected from 60 sites that included 9 intact baldcypress swamps and 51 former swamps that had been farmed for 1 to 50 years, as determined by talking with local landowners and air photo interpretation (Fig. 1). Twenty seed bank samples were taken at stratified random positions at 10-m intervals along two randomly located transects of 100 m in length. The soil samples were collected from the top 10 cm of the soil with a shovel and composited (one sample per site); the volume lifted from each field sampling point was 4705 ± 105 cm3. Sites that were underwater were sampled using a posthole digger. The samples were held until May of the year of collection in a cold storage room at 5 °C.

After sieving through a wire screen to remove roots and rhizomes, the composite samples were divided into 10 trays (27 × 27 cm) and filled to a depth of 4·5 cm (no potting soil or vermiculite was used underneath these samples). The volume of the seed bank sample in the tray was 2600 ± 5 cm3. For each of the 60 sampling sites, trays were assigned randomly to blocks within a glasshouse at Southern Illinois University. Of the 10 trays from each site, 5 trays were assigned to freely drained (moist) and 5 to flooded treatments (20 cm above soil level).

Seedlings that emerged from the soil were removed to avoid seed input by mature plants. To check for seed contamination from sources within the glasshouse, trays of sterile vermiculite were interspersed throughout the seed bank tanks; species that germinated in the vermiculite were eliminated from the analysis. After the seed bank assay was completed, soils were sieved to determine if any seeds were left ungerminated. Seeds recovered in this manner were split open with a knife and tested with tetrazolium for viability (Baskin & Baskin 1998). Nomenclature and naming authorities follow USDA (2003).

data analyses

To calculate seed density m−2 for each dominant species, the treatment (flooded vs. freely drained) with the highest number of seeds germinating (Middleton et al. 1991) was multiplied by 13·72. This multiplication procedure standardizes seed numbers germinating on a m−2 basis to make studies using germination trays of various sizes more comparable to one another (van der Valk & Davis 1978). Species with mean seed densities > 14 seeds m−2 were designated as dominants (24 species), with the majority of the remaining less dominant species having a seed density of less than 7 seeds m−2 species. Less dominant species were grouped into a minor category. Percentage (%) frequency was calculated as the number of occurrences for a species divided by the total number of sites (60) (Brower, Zar & van Ende 1998). Mean seed density for sites for each dominant species vs. length of time the site was farmed were regressed using linear, log linear, second order polynomial and exponential models. Because this analysis explained little of the variability in the data, two-way anova was performed on log mean seed density m−2 with dominant species and farm history (farmed vs. intact) as the independent factors. Single degree-freedom contrasts of interest were conducted using the species × farm history interaction with log transformed data (SAS JMP 2002). Bonferroni corrections were used in the contrasts to accommodate experimentwise data, and the alpha value (P < 0·05) was adjusted to P < 0·002 (Kirk 1990). Data were log transformed to meet the normality and homogeneity assumptions of Analysis of Variance.

Results

seed density changes over time

Changes in the seed or spore density were not related to the length of time the site was farmed for the dominant species (species × time interaction; F = 10·1, P < 0·0001). Regression analysis explained little of the variability in log mean seed density (i.e. low r2 values), although log mean seed density had a significant relationship (P < 0·03) with length of time of farming for a few species, including Ammannia auriculata Willd., Gratiola virginiana L., Juncus marginatus Rostk., Ludwigia palustris (L.) Ell., Neeragrostis reptans (Michx.) Nicora, and Veronica peregrina L. None of the other dominant species showed a significant relationship between mean density of seeds and length of time of farming using linear, log, second order polynomial, or exponential models.

Because most of the dominant species lacked a relationship between seed density and length of time of farming, seed density was compared in farmed vs. intact baldcypress swamps. The mean seed density m−2 differed for some dominant species in farmed vs. intact sites (species × farm type interaction: F = 4·4, P < 0·0001) with higher seed density in intact than farmed sites for Veronica peregrina (Table 1). However, some species had higher seed densities in farmed than intact sites, including A. auriculata, Eleocharis obtusa, Juncus biflorus and J. marginatus (1 d.f. contrast; P < 0·002; Table 1). At least one species, Callitriche heterophylla was locally abundant at some sites, but otherwise rare (percentage frequency = 0·05; mean density = 29·5 ± 27·5 seeds m−2).

species richness and farming

Seeds of the tree species of these baldcypress swamps were mostly absent in seed banks except for C. occidentalis, which was present in both farmed and intact sites. Betula nigra L. was present only in intact sites and Acer saccharinum L. only in farmed sites. Live seeds of dominant species of baldcypress swamps, such as T. distichum and N. aquatica, were absent in the seed banks of all sites; however, dead seeds of both species were observed in seed banks.

The seed banks of both farmed and intact baldcypress swamps had a large number of herbaceous species (species richness = 207 and 173, respectively). Several aquatic species were absent (or nearly absent) in farmed sites, including Alisma plantago-aquatica L., Heteranthera dubia (Jacq.) MacM., Hottonia inflata Ell. and Wolffia columbiana Karst. Lemna minor L. and L. trisulca L. were absent in both farmed and intact sites, while Spirodela polyrhiza (L.) Schleiden was present in both sites. The emergent species, Carex lupulina Muhl. and Saururus cernuus L. were present only in intact sites; Typha latifolia L. was present in both farmed and intact sites. Many herbaceous species were present in farmed sites, but were not present in intact sites, including Cyperus pseudovegetus Steud., C. strigosus L., Digitaria ciliaris (Retz.) Koel., Eleocharis palustris (L.) Roemer & J.A. Schultes, Eragrostis hypnoides (Lam.) B.S.P., Eupatorium fistulosum Barratt, Euphorbia hexagona Nutt. ex Spreng., Euphorbia spathulata Lam., Hippuris vulgaris L., Juncus acuminatus Michx., J. interior Wieg., Myosotis verna Nutt., Polygonum pensylvanicum L., Sibara virginica (L.) Rollins and X. strumarium L.

Introduced species were not numerous in the seed banks of either farmed or intact sites (species richness = 12 vs. 10 species) but included Abutilon theophrasti Medik., Amaranthus retroflexus, Echinochloa crus-galli, Poa compressa L., Ranunculus sardous Crantz., Setaria faberi Herrm, Sinapis alba L., Sonchus oleraceus L. and X. spinosum L. Additional species were listed as both native and introduced by the USDA (2003) such as Taraxacum officinale G.H. Weber ex. Wiggers in both farmed and intact sites, as well as Artemisia vulgaris L. and Samolus valerandi L. in farmed sites only.

Discussion

the missing dominants

Woody species, including trees, shrubs and vines are not maintained as long-term components of the agricultural seed banks of baldcypress swamps and this makes the restoration of abandoned farm fields more difficult. Woody species also are under-represented in the agricultural seed banks of former floodplain forests in Europe (Prach & Pyšek 1994; Bossuyt & Martin 2001). In this study, key dominants of baldcypress swamps were absent in the seed banks of farmed agricultural fields including trees such as T. distichum and N. aquatica. Shrub and vine species were also missing with the exception of C. occidentalis, which was present in the seed banks of both intact and farmed sites.

Tree seeds of baldcypress swamps are short-lived (e.g. T. distichum; Middleton 2000) and mainly dispersed by water to the seed bank (Schneider & Sharitz 1986, 1988; Middleton 1999, 2000). The presence of viable seeds of these species in seed banks is dependent on recent winter flood pulses of water. After flooding, seeds settle along drift lines as water levels recede. Deposited seeds must germinate in drawndown conditions, and typically have a germination window of less than 1 year following drawdown (Middleton 2000). The recruitment patterns of these dominants are adapted to the climatically driven flood pulse associated with this region, i.e. high water levels emanating from the channel during the winter rain and/or snow melt period, followed by summer drawdown (Middleton 1999, 2002).

Problematically, while these aquatically dispersed woody species may rely on recurring episodes of flooding to maintain themselves in seed banks, flood pulses of water may no longer occur reliably within landscapes that have been altered for agricultural development. It is not known to what extent hydrologic re-engineering may be hindering restoration efforts in floodplains around the world. However, it has been widely recognized that restoration is served by unimpeded dispersal pathways from natural communities to restoration sites (Allan & Flecker 1993; Henry & Amoros 1996; Middleton 1999, 2000, 2002) to resupply depleted seed banks (Bakker et al. 1996). Riverine wetlands with the highest connectivity via flooding often have the highest species diversity (Bornette, Amoros & Lamoureau 1998).

The origin of the dead seeds of T. distichum and N. aquatica in the fields in this study require some further explanation. These seeds either could have fallen in situ from the forest canopy before the fields were cleared of trees, or the seeds could have floated to the sites via floodwater. The majority of these fields flood during winter and spring flood events (pers. obs.). However, floodwaters in fields may come mostly from runoff from the uplands and not from pulsing from the river channel. The outer coverings of these two species are woody, and likely to persist in the soil long after the seed itself has died, so that these seeds could have been deposited in the fields many years ago. These deposited seeds would have had no niche for establishment in cultivated fields.

herbaceous species

Unlike the woody species in this study, seeds of herbaceous species are available for restoration in the seed banks of farmland including a large number of species (species richness = 173–207 species) with generally high densities (Table 1). Densities of herbaceous seeds in the seed bank was not related to the length of time of farming, but generally either increased or decreased with farming. A study of the seed banks of farmed prairie potholes also found that wetland species were maintained in seed banks despite farming (e.g. A. plantago-aquatica, Elatine triandra Schkuhr, G. neglecta and P. pensylvanicum), but that the density of some species decreased over time, especially if farming continued for more than 20 years (e.g. G. neglecta; Wienhold & van der Valk 1989). European studies of seed longevity in agricultural soils suggest a very rapid rate of seed density loss in ploughed soils (30–72·4% loss per year (Roberts & Dawkins 1967; Froud-Williams, Chancellor & Drennan 1983). In contrast, this study suggests that seed loss of the herbaceous species of baldcypress swamps may be negligible over 50 years of cultivation, although admittedly the methodology applied here did not provide direct comparison of seed loss under specific types of cultivation. Herbaceous species may be maintained either because they are long-lived or because they are maintained as ‘weeds’ during cultivation. This study does not answer that question, nor does it shed any light on the effects of the type of tillage, fertilization and chemical applications on species persistence in the seed banks of agricultural fields (Kremer 1993; Mayor & Dessaint 1998). European grasslands that are the least agriculturally improved (i.e. less cultivated, fertilized and time under cultivation) are the most readily restored (Bekker et al. 1997).

Certain herbaceous species have wind-borne disseminules such as Ricciocarpos natans (Middleton 1999), and are found throughout the spore-banks of agricultural fields (Table 1). Nonetheless, flood pulsing is likely to increase the distribution of R. natans because the species is present at much higher densities at elevations with winter flooding/summer drawdown (Conrad 1997).

Some common herbaceous species of baldcypress swamps are absent in the seed banks of these fields, notably those that are primarily dispersed by vegetative organs (see Middleton 1995, 1999), including Hottonia inflata, Limnobium spongia (Bosc) L.C. Rich. ex Steud., L. minor, L. trisulca and Wolffiia columbiana. In farmed prairie potholes, lemnids were missing both from the propagule and seed banks, except for Lemna minor (Wienhold & van der Valk 1989). Perhaps because of this, Lemna minor established in farmed fields soon after hydrologic restoration, but not S. polyrhiza (Galatowitsch & van der Valk 1996a; S.M. Galatowitsch, pers. comm.). In this study, S. polyrhiza was present in the seed banks of farmed sites suggesting that it would readily recolonize fields after hydrologic restoration. Alternatively, birds can disseminate entire plants of lemnids on their feathers, and this may aid in their eventual reestablishment in restoration sites (Middleton 2002). However, apart from certain of the very small lemnids, most species that disperse primarily via vegetative organs may require active reintroduction in restoration sites, unless clever hydrologic re-engineering can promote their dispersal from forested floodplain fragments to restoration sites.

It is also worth noting that exotic herbaceous species were maintained in the seed banks of both farmed and intact baldcypress forests, including P. compressa, R. sardous and S. oleraceus. Additionally, seeds of A. artemisiifolia, A. retroflexus and S. faberi were also found; these were noted as characteristic of the weed seed bank in farmed fields by Zhang et al. (1998). Xanthium strumarium was common in the seed banks of farmed fields in this study, and also in recently restored prairie potholes (Galatowitsch & van der Valk 1995).

water delivery in wetland restoration

This study demonstrates that baldcypress swamps diverse in woody species are unlikely to regenerate naturally on abandoned farmed floodplains in the south-eastern United States. This is because seed banks are depleted following farming. Natural restoration is also likely to require hydrologic re-engineering that incorporates seasonally driven pulses of flood water to abandoned fields delivering seeds of dominant species.

It is not possible that flood pulsed hydrology can be delivered to all abandoned farm fields on floodplains where the restoration of nature conservation areas is desirable. The historical and political changes that were put into place for development across these landscapes would have to be reversed, and this is usually not feasible after people have moved onto the floodplains (McCorvie & Lant 1993). In such cases, restoration goals need to be adjusted to recognize the limitations set by these constraints.

Intact baldcypress swamps also rely on flood pulsed hydrology to maintain dispersal to the short-lived seed banks of woody species. The majority of nature conservation areas on floodplains have been cut off from flood pulsing from channels because of re-engineering for development, i.e. channel straightening and deepening. These engineering works result in the downcutting of rivers (i.e. a deepening of the channel bed) and this limits the pulsing of water from the channel onto otherwise undisturbed floodplains (Dister et al. 1990; Middleton 1999). This problem has broad implications because 77% of the rivers in the northern part of the world have been re-engineered. In nature conservation areas, woody and exotic species associated with drier environments are replacing those of wetter environments in otherwise undisturbed forests adjacent to downcut channels (Weller 1995; Shear, Lent & Fraver 1996).

Along the Cache River, some re-engineering has addressed water delivery problems to nature areas and restoration sites on the floodplain. For example, the channel is downcut downstream from Heron Pond, as well as portions of upper Cache, which lie above the Post Creek Cutoff (Sengupta 1995). Downcutting is the incision of a stream channel due to erosion, and the process can be accelerated as a result of the re-engineering of channels and floodplains. Stream beds along altered rivers may drop to elevations much below their original levels, and because of this, little or no flood pulsed water emanates from the channel onto the floodplain (Dister et al. 1990). Managers have fortified the natural embankment that separates Heron Pond from the Cache River so that water does not flow as quickly out of the swamp into the downcut channel (IDNR 2003).

Other reengineering strategies along the Cache River have dealt with the lack of downstream water flow to the Lower Cache River. The Diehl Dam was constructed downstream of Buttonland Swamp in the 1980s and has increased water depths in this swamp (Muir et al. 1995; Middleton 2002). The Diehl Dam and Big Creek projects both have attempted to reduce sedimentation in Buttonland Swamp (Muir et al. 1995; USDA 2003). After the construction of the Diehl Dam, Buttonland Swamp was permanently impounded so that regeneration in this swamp was restricted to old field sites near the margins of the swamp near the highest elevations of winter flooding (Middleton 2000). Managers lowered the Diehl Dam in November 2002 in an attempt to decrease tree damage from long-term flooding (Winkeler 2003). While dam construction is a common restoration technique in North America, it restricts flow and flood pulsing, reduces tree survivorship and production (Middleton 1999, 2002), and maintains subimpoundments with distinctive flora (Jansson, Nilsson & Renöfält 2000).

Ultimately, the successful restoration of abandoned farm fields probably depends on the restoration of flood pulsing across the landscape. In the case of the Cache River, Illinois, alternatives might be considered to supply flood pulsing to nature conservation areas. The reconnection of the river may become more possible in the future with the acquisition of the Grassy Slough Preserve, which lies between the dissected waterway (TNC 2003). Flood pulse restoration has been attempted in only a few places thus far, including the Kissimmee River, Florida (Toth et al. 2002), and along some western rivers in the USA (Scott & Auble 2002; Stromberg & Chew 2002). The success of this approach is not yet demonstrated in resupplying depleted seed banks of farmed fields with floodplain forest species.

synthesis and applications

The short-lived seeds of woody species of seasonal baldcypress swamps depend on frequent episodes of flood pulsing to replenish seed banks in natural swamps. Seed banks of farmed fields harbour almost no woody species that are available for restoration, and this fact is undoubtedly related to the landscape setting of floodplains altered for agriculture. Because dominant species probably regulate function, e.g. production, carbon and nutrient cycling, the composition of species that establish in restoration sites may be important. For the Cache River watershed in Illinois, the regional hydrology has been dried to facilitate development. Similar hydrologic modifications, designed to benefit development and navigation, affect as much as 77% of the rivers in the northern hemisphere (Dynesius & Nilsson 1994). If replanting were designed to restore many functional types on sites, it might be possible to maintain biodiversity on these sites to a greater extent than at present. However, only very few tree species are planted at restoration sites, and this has failed to re-establish anything more than plantations in what were once diverse swamp forests of the south-eastern United States. Where feasible, the re-establishment and maintenance of diverse floodplain wetlands would be best served by addressing the critical hydrologic problems hindering the supply of seeds to seed banks. Our inattention to the flood pulsed hydrology, which is often required in the life history of plant species, is hampering our ability both to restore and maintain nature conservation areas. The knowledge necessary to maintain diverse wetlands on floodplains spans the disciplines of both population and landscape ecology, and illustrates the fact that the intermingling of ideas from both fields aids problem solving in restoration ecology and ecosystem management.

Acknowledgements

This study was funded with a grant from the Office of Research Development and Administration, Southern Illinois University and the United States Department of the Interior via the Illinois Water Resources Center, Urbana-Champaign, Illinois (IL-WRC-95–220). Ernie Lewis of Southern Illinois University and Rassa Dale of the USGS National Wetlands Research Center provided statistical support. At least 30 private farmers allowed me access to their farmed fields and provided information regarding the history of these sites. Thanks to Carol Wienhold, Susan Galatowitsch and anonymous referees who commented on earlier drafts of this manuscript. Many students helped with field and glasshouse work, including Mark Basinger, Erin Conley, Holly Harris, Charlie Giedeman, Scott Keykendall, Eduardo Sanchez, Edmond Schott, Allison Strauss, John Rivera, Jonathan Taylor and John Wilker.

Ancillary