• adaptive management;
  • alternative community state;
  • coastal plain;
  • depression wetland;
  • ecological threshold;
  • fire;
  • hardwood removal;
  • longleaf pine ecosystem;
  • restoration;
  • seed bank


  1. Top of page
  2. Summary
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

1. The restoration of disturbance-maintained ecosystems may require management to overcome ecological thresholds and re-establish feedbacks that perpetuate an alternative community. We use hardwood-dominated depression wetlands (locally known as oak domes) embedded in the fire-maintained longleaf pine–wiregrass Pinus palustrisAristida stricta ecosystem as an example where concepts developed from alternative state theory are applied to practical restoration.

2. As extant communities were not available as reference sites, we based our restoration objectives on knowledge of vegetation dynamics, land-use history and historical data. We quantified a hardwood encroachment pattern beginning with the establishment of central nuclei during fire-free periods. Expansion of this core of hardwoods is positively reinforced by the accumulation of fuels that impede the spread of fire.

3. In order to examine the feasibility of re-establishing herbaceous communities, we selected 10 depression wetlands in 2000 and randomly assigned a hardwood removal treatment to half of them. During the transition period of fine fuel accumulation, we adapted the management regime as necessary for control of hardwood re-sprouts and to promote the development of a fire-maintained community.

4. After 5 years, hardwood removal communities had shifted toward herbaceous dominance, characterized by multi-layered, species-rich, native, wetland-specific ground flora. The rapid recovery of herbaceous species was probably possible because of initial seedling recruitment from a persistent wetland soil seed bank. This immediate recruitment of herbaceous vegetation produced fine fuels, allowing for the reintroduction of frequent prescribed fire and, thus, the re-establishment of the herbaceous community-fire feedback mechanism necessary to maintain the community state.

5.Synthesis and applications. Our findings confirm that it is possible to re-establish a rare alternative community state in a fire-maintained ecosystem. Establishment of a desired transition trajectory required decoupling ecological feedbacks that inhibit reintroduction of fire while facilitating positive feedbacks to promote fire. Our approach incorporating ecological thresholds and biotic legacies, such as a persistent seed bank, can serve as a model to inform restoration strategies for other disturbance-maintained ecosystems.


  1. Top of page
  2. Summary
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Restoration of alternative community states

Ecological literature is rich with conceptual models and discussions of ecosystems in which multiple community states are possible (e.g. Holling 1973; Beisner et al. 2003; Mayer & Rietkerk 2004; Suding et al. 2004; Hobbs 2007). Alternative state dynamics, including feedback cycles and thresholds, may be particularly applicable to restoration and management of disturbance-maintained ecosystems (Peterson 2002a; Groffman et al. 2006; Nowacki & Abrams 2008). In these ecosystems, restoration goals often target community states that are rare or extirpated due to elimination of natural disturbances (Groffman et al. 2006). The objective of restoration is therefore to restore and maintain landscape heterogeneity and γ-diversity (Holling 1973; Nowacki & Abrams 2008). Yet, re-establishment of alternative communities can be complicated when the species in the existing community respond differently to disturbance (Beisner et al. 2003; Suding et al. 2004; Groffman et al. 2006). In such cases, it is important to couple the reintroduction of the disturbance regime with adaptive management to direct the community through a transition period and across an alternative state threshold (Bennett et al. 2005; Groffman et al. 2006; Nowacki & Abrams 2008).

We developed concepts from alternative state theory and applied them to a practical restoration of depression wetlands embedded in the fire-maintained longleaf pine Pinus palustris P. Mill–wiregrass Aristida stricta Michx ecosystem. Previous discussion of ecological thresholds, feedback cycles and alternative states in the longleaf pine ecosystem has been centred in upland communities (Peterson 2002a,b; Bennett et al. 2005; Groffman et al. 2006). Similarly, we propose that hardwood dominance represents one community state in depression wetlands. Although remnant herbaceous vegetation was not present, historical evidence from property maps and aerial photography chronosequences suggested that complete oak dominance of small depression wetlands is a recent phenomenon, and many depressions would have been herbaceous communities within the last century (Martin 2006).

Fire in the longleaf pine ecosystem and embedded depression wetlands

Fire is understood to be an integral process across the longleaf pine–wiregrass ecosystem. Regular, low-intensity burns with return intervals of 2–4 years create canopies of widely spaced longleaf pine, which allow sufficient light to support a highly diverse herb-dominated ground flora (Lemon 1949; Walker & Peet 1983; Bridges & Orzell 1989; Clewell 1989; Peet & Allard 1993; Maliakal et al. 2000; Kirkman et al. 2004a). Periods of extended fire suppression cause a shift in community structure from an herbaceous ground flora to one dominated by shrubs and hardwood sprouts, and with sufficient fire-free periods, species such as oaks Quercus spp. reach a fire-resistant stage (Lemon 1949; Guerin 1993; Glitzenstein et al. 1995; Maliakal et al. 2000).

The length of fire suppression necessary to encourage hardwood encroachment may vary depending on local site conditions (Glitzenstein et al. 1995; Hinman & Brewer 2007). Guerin (1993) demonstrated that oak sprouts taller than 2 m are capable of surviving low-intensity burns, which represents an important threshold. Once established, fire-impeding hardwood species further exclude burns (Platt et al. 1991; Kane, Varner & Hiers 2008). As the hardwood canopy develops, the increased shade (Walker & Peet 1983; Clewell 1989; Kirkman et al. 2004a) and increased litter layer (Hiers et al. 2007) associated with hardwood dominance nearly eliminates the rich ground flora. This further shifts the available fuels from flammable ground flora species to fire-impeding hardwood leaf litter (Peterson 2002a; Groffman et al. 2006). In turn, fire-sensitive hardwood patches transition across an ecological threshold and into an alternative state. After hardwoods reach this fire-suppressing state, re-introduction of fire alone does not restore previous ecosystem dynamics (Harrington & Edwards 1999; Brockway & Outcalt 2000; Peterson 2002a; Groffman et al. 2006).

Study objectives

Depression wetlands may have been isolated from burning, even within longleaf pine forests that are managed with prescribed fire. In particular, land management practices, such as the traditional use of winter burns, coincided with periods during which seasonally flooded depression wetlands were most likely to be inundated, and therefore would not burn (Lemon 1949; Clewell 1989; Goolsby 2006). Similarly, depression wetlands near roads and field edges were often excluded from fire simply due to their proximity to adjacent fire-excluded areas. Consequently, seasonal wetlands, and particularly small depressions, might have served as fire refugia for species such as live oak Quercus virginiana P. Mill., water oak Q. nigra L., and laurel oak Q. laurifolia Michx. that tolerate periodically saturated soil conditions (Menges et al. 1993; Jacqmain et al. 1999) but are fire intolerant at early stages. In turn, the increased evapotranspiration rates of hardwood species may have resulted in drier conditions and decreased duration of inundation, further promoting woody establishment and accumulation of less flammable fuels (Sun et al. 2001; W. Hicks, personal communication). The accumulation of fire-inhibiting litter of species including Q. virginiana, Q. nigra and Q. laurifolia (Kane et al. 2008) creates a fire shadow and disconnects the fire corridor between uplands and wetland interiors (Kirkman et al. 1998).

Small, herbaceous depression wetlands support a unique suite of wetland and ecotone species and contribute to regional or γ-diversity (Sutter & Kral 1994; Kirkman et al. 1998; Semlitsch & Bodie 1998). Mesic longleaf pine communities, including depression wetlands and wetland–upland ecotones, harbour the highest plant diversity within the longleaf pine ecosystem (Sutter & Kral 1994; Kirkman et al. 1998, 2004b). Often, studies of the regional depression wetland systems in the south-eastern USA do not include the small depressions (1 ha or less) that may be more similar hydrologically to a wetland–upland ecotone or mesic savanna (Kirkman et al. 2000; Sharitz 2003; DeSteven & Toner 2004). Small depression wetlands with short, seasonal hydroperiods are sometimes recognized as ‘oak domes’ or ‘oak depression hammocks’ because of the conspicuous island-like dominance of hardwoods (NatureServe 2007). Historically, a regime of regular fire probably maintained a persistent herbaceous community state in some small depression wetlands. However, today this community state may be extremely rare due to widespread hardwood encroachment during fire-free periods.

Our central hypothesis was that a fire-maintained, species-rich herbaceous wetland ground flora could be restored in these depression wetlands using hardwood removal and adaptive management to re-instate fire. To understand the driving mechanisms during the hardwood–herbaceous transition, we quantified the pattern of historic hardwood encroachment. Finally, we gathered evidence to determine whether a persistent soil seed bank contributed native wetland species to the re-vegetation process.

Material and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Site description

The hardwood-dominated depression wetlands selected for study were located on Ichauway, a privately owned property of the Joseph W. Jones Ecological Research Center, encompassing 11 300 ha of Baker County, on the Lower Coastal Plain and Flatwoods (LCPF) Province of south-west Georgia, USA (McNab & Avers 1994). The climate is humid sub-tropical with annual average rainfall of 1310 mm evenly distributed throughout the year and average daily temperature ranges (low–high) 21–34 °C in summer and 5–17 °C in winter. Currently, 7500 ha of 70- to 90-year-old second growth longleaf pine-dominated forest spans the property interspersed with depression wetlands. The presence of species such as wiregrass that do not re-establish following soil perturbation indicates that although the forest is second growth, it has not experienced major anthropogenic alteration (Drew et al. 1998).

Throughout most of the 20th century, Ichauway land managers used annual, cool season-prescribed burning to maintain longleaf pine habitat for northern bobwhite quail Colinus virginianus. In 1992, Ichauway was designated an ecological research centre and prescribed fire management shifted to a 2-year burn rotation. Burns are often applied during the growing season to top-kill hardwood sprouts and to encourage reproduction in many of the fire-dependent ground flora species, including wiregrass (Goolsby 2006). While the prescribed fire techniques used do not actively discourage fire from spreading from uplands through ecotones and into depression wetlands, the current community of hardwood species and associated fire-impeding leaf litter serves to exclude fire from small depressions.

We selected 10 small, oak-dominated depression wetlands for the study. These sites were chosen based on: (1) similar size (approximately 1 ha and smaller) and circular shape; (2) evidence of seasonal inundation, particularly high water marks on trees; (3) soil profile traits with redoximorphic features indicative of hydric soils; (4) the presence of wiregrass in the adjacent ecotone, indicating potential connectivity with undisturbed and fire-maintained ground flora; and (5) a complete, closed canopy of oak and other hardwoods with minimal ground flora. All study sites were initially dominated by Q. virginiana, Q. laurifolia and Q. nigra. The very sparse ground flora included woody vine species such as trumpet creeper Campsis radicans (L.) Seem. ex Bureau, muscadine Vitis rotundifolia Michx. and grape vine Vitis aestivalis Michx. Characteristic soils include aquic paleudults, grossarenic paleaquults, aquic arenic paleudults and typic and arenic paleudults (Goebel et al. 1997). Typically, these sites are inundated for a few weeks in the late winter to early spring prior to the increased rate of evapotranspiration following deciduous bud break, but hydroperiod varies considerably with local precipitation events.

Pattern of hardwood encroachment

We hypothesized that initial oak establishment was related to a fire shadow (fire-free area) associated with the most mesic conditions of the depression wetlands. Therefore, we expected to find the oldest trees in the centre of the depression wetlands. Due to the difficulty of extracting accurate cores from large live oaks, we first tested whether diameter at breast height, d.b.h. (1·37 m) was an appropriate field surrogate for age. We obtained cross-sections of 25 boles each of Q. virginiana and Q. laurifolia from similar small depression wetlands. The boles spanned a range of size classes. A cross-section cut from each tree was sanded and annual growth rings were counted. We used linear regression to establish the relationship between age and d.b.h. of the oaks. We then used the age–size relationship to examine the spatial pattern of hardwood establishment. As part of a larger long-term study, all woody stems larger than 2·5 cm d.b.h. were recorded in 20 × 20 m sequential permanent plots in transects across each depression wetland (Fig. 1). In the transects, plots were divided into two zones: (1) interior wetland plots and (2) ecotone plots (upland–wetland transitional) plots designated by the presence of wiregrass, an upland species. To examine the hardwood encroachment pattern, we calculated a mean d.b.h. from the three largest (presumably oldest) individuals of both oak species in the interior and ecotone zones in each of our study depression wetlands (Fig. 1). We then tested our hypothesis that stems in the centre of the depression wetlands were larger, and, therefore, older using PROC GLM in sas with the individual depression wetlands as a block factor (sas on-line documentation:


Figure 1.  Sampling design using 100 m2 ground flora sub-plots within 20 × 20 m permanent plots in a transect from two upland anchor plots.

Download figure to PowerPoint

Description of hardwood removal treatment

Our restoration treatment included mechanical and chemical hardwood control. In the summer of 2000, we randomly assigned five study depression wetlands to a complete canopy removal treatment, and five were maintained as un-manipulated control sites. Prior to the canopy harvest, mean (±SE) stem density across all our sites was 29·85 ± 3·59 stems per 400 m2 with a mean basal area of (±SE) of 15·42 ± 1·32 m2 ha−1. At the hardwood removal sites, we used an industrial mower to remove all small saplings up to approximately 10 cm d.b.h.. Trees up to 43 cm d.b.h. were removed with a feller-buncher, and stumps were treated with Pathway (5·4% picloram, 20·9% 2,4 d-Amine; Dow AgroSciences, Indianapolis, IN, USA) to prevent re-sprouting. Due to equipment limitations, larger trees were girdled and sprayed with Pathway herbicide. All large hardwoods that showed signs of incomplete mortality in 2003 were re-girdled and sprayed with Arsenal 11 (27·6% imazapyr, isopropylamine salt; Basf Corporation, Research Triangle Park, NC, USA). All sites (hardwood removal and un-manipulated) were included in larger prescribed fire management units on 2-year burn rotations that were burned three times during the 2000–2005 period. These fires did not penetrate un-manipulated sites due to abundant fire-resistant oak litter and a lack of groundcover fuels to carry fire. In 2002 and 2004, all hardwood re-sprouts approaching the 2-m fire-tolerant stage were spot treated with mowing and herbicide.

Vegetation response to hardwood removal

Ground flora vegetation was sampled in 100-m2 sub-plots within permanent 20-m-wide transects designed for the long-term study (Fig. 1). Due to variability in depression wetland diameter, each site contained between three and seven sub-plots. The presence of each ground flora species was recorded in 2000 prior to hardwood removal, and again 5 years post-hardwood removal in 2005. We included all woody species less than 2·5 cm d.b.h.

In order to examine the effects of the hardwood removal on community composition, we first examined species presence–absence data at the 100-m2 plot level in a non-metric multidimensional scaling (NMS) analysis using a Jaccard dissimilarity index (McCune & Grace 2002). All recorded species were included in the analyses. NMS depicts objects in space such that distances represent the dissimilarity while minimizing the discrepancy between the rank-order dissimilarity calculations and ordination distances (Kruskal 1964). The optimal dimensionality of the NMS solution was determined by examining a plot of the sas badness-of-fit criterion as a function of NMS dimension. Beyond graphical illustration, NMS is also a useful tool for data reduction and the dimension coordinates can be used as response variables for hypothesis testing (Quinn & Keough 2002).

Non-metric multidimensional scaling dimension coordinates were then used as multivariate response variables in a general linear mixed models analysis (Wright 1998; Schabenberger & Pierce 2002). A Shapiro–Wilk’s test was used to confirm that dimension coordinates met the normal distribution assumption of mixed models analysis. Mixed models allow for testing of fixed effects as well as covariant components (Littell et al. 1996), which in our case were the random effects generated by our hierarchical sampling design (nested effects of zones, depressions and plots) and repeated measures in time. In the multivariate analysis of variance, we used a multivariate structure for the residuals and a no-intercept model with dummy variable coding for the multivariate dimension coordinates (Wright 1998). We used information criteria statistics to select the appropriate covariance structures for the repeated measures and random effects. Multivariate contrasts were constructed to test for significant main effects within the interactions between treatments and zone. The contrasts identify changes in plot community composition over time using the NMS dimension coordinates as the response variable. The Kenward–Roger adjustment for the denominator degrees of freedom (Kenward & Roger 1997) was used to improve the test statistic P-values (see also Kirkman et al. 2007). To identify species that increased in frequency, we used Pearson correlations between year and species presence–absence, and report all species with a correlation value exceeding 0·30. Analyses were performed using the sas System 9·0 (sas on-line documentation:

Soil seed bank

To determine the composition and potential contribution of the soil seed bank, we evaluated the germination of seeds from soil samples collected in each of the un-manipulated depression wetlands (Poiani & Johnson 1988). In March 2005, five soil cores (10 cm in diameter, 6 cm deep) were collected and combined into a single sample at 20-m intervals immediately adjacent to the depression wetland transects. Soil samples were stored at 4 °C for 2 months. Soil was then sieved to remove large rhizomes and debris, spread over potting soil mix in tubs with holes drilled in the bottom to allow drainage, and placed randomly on greenhouse benches. Moist soil conditions were maintained by watering daily with reverse osmosis filtered (RO) water. From May 2005 to September 2006, each emergent seedling was identified, recorded and discarded. Unknown seedlings were removed, potted and grown until identification was possible. To determine the presence of targeted wetland-specific perennial species, each species was assigned a wetland indicator status based on the USDA PLANTS data base (USDA, NRCS 2006) and consolidated into three categories (upland, facultative and wetland) as in DeSteven et al. (2006).

Species nomenclature follows Wunderlin & Hansen (2003) with the exception of Dicanthelium, which follows Freckmann & Lelong (1993).


  1. Top of page
  2. Summary
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Hardwood establishment pattern

Mean diameter of the largest trees of Q. laurifolia and Q. virginiana present in the wetland differed by zone (interior vs. ecotone) (F = 10·02, P = 0·003, d.f. = 1 Q. laurifolia; F = 4·71, = 0·04, d.f. = 1 Q. virginiana). The interior zone harboured larger diameter trees of both species than that of the ecotone. Mean (±SE) d.b.h. for Q. laurifolia was 14·9 (±2·9) cm in the ecotone zone and 26·6 (±2·2) cm in the interior zone. Likewise, d.b.h. values for Q. virginiana were 28·6 (±4·7) cm in the ecotone zone and 55·2 (±5·7) cm in the interior zone. For both oak species, d.b.h. and age were positively correlated (r2 = 0·758, < 0·001 Q. laurifolia, r2 = 0·742, < 0·001 Q. virginiana); thus, presumably, differences in size represent relative differences in dates of establishment.

Changes in vegetation composition following restoration

Shifts in species composition between 2000 and 2005 were striking in the interior zone of hardwood removal plots with the development of a dense herbaceous ground cover (Fig. 2). The NMS solution clearly shows a greater dissimilarity in the vegetation in the interior zones of hardwood removal sites over time. Un-manipulated plots were clustered, and plots in the ecotone zone were the most similar (Fig. 3). A two-dimensional NMS solution was selected after 12 iterations (badness-of-fit = 0·151). This solution explained 76% of the variance, with 52·4% explained by dimension one and 23·7% by dimension two.


Figure 2.  Changes in an example site 5 years post-treatment. Large oaks in the background of 2005 picture were too large for mechanical removal and were girdled.

Download figure to PowerPoint


Figure 3.  Non-metric multidimensional scaling two-dimensional solution for species presence–absence in 100-m2 ground flora plots. Lines connect plots in 2000 and 2005. Dotted lines connect hardwood removal and solid lines connect un-manipulated control plots. Open circles are in the ecotone zone of the depression, closed are in depression interiors.

Download figure to PowerPoint

After 5 years, species composition and dominant guilds of species differed as a result of the hardwood removal (Table 1). In the mixed models analysis, information criteria confirmed selection of an unstructured error covariance matrix, and a no diagonal first-order factor analytic structure for the random effects. The selected model provided a better fit than the null model (likelihood ratio test: χ2 = 72·83, < 0·001, d.f. = 8). Multivariate contrasts performed on the mixed models results indicated that depression wetland interiors differed, but no changes were detected in the ecotones (Table 2). Pearson correlations of species presence–absence changes from 2000 to 2005 support a community shift toward herbaceous, wetland-specific dominance (Table 3). In 2005, wetland forbs and graminoids such as Panicum, Dichanthelium, Carex, Saccharum, Ludwigia and Eleocharis were more abundant in hardwood removal plots. Likewise, woody species were less abundant. (We define abundance here as frequency, or the number of plots where the species was present.) Of the 26 species with an increased presence over time, approximately 58% were obligate and 11·5% were facultative wetland species.

Table 1.   Mixed models analysis of community composition measured using non-metric multidimensional scaling (NMS) ordination of species presence–absence
EffectNum DFDen DFFP
  1. Treatments include canopy harvest and control. Zones refer to interior depression of ecotone areas of sites. Year refers to measures taken in 2000 and 2005. *Significance at α = 0·05.

NMS dimension (dimension)2 8·242·91  0·1106
Dimension × treatment2 8·2415·290·0017*
Dimension × zone211·181·06<0·0001*
Dimension × treatment × zone211·120·280·0002*
Dimension × year247·475·11<0·0001*
Dimension × treatment × year247·462·49<0·0001*
Dimension × zone × year247·49·910·0003*
Dimension × treatment × zone × year247·423·55<0·0001*
Table 2.   Multivariate contrasts examining community composition dissimilarity
ContrastNum DFDen DFFP
  1. NMS dimension coordinates for each vegetation plot are the response variables. Effects of treatment (un-manipulated and hardwood harvest) and zone (ecotone and interior) are contrasted. *Significance at α = 0·05.

Oak treatment × zone for un-manipulated232·888·79<0·0001*
Oak treatment × zone for harvest236·221·08<0·0001*
Oak treatment × zone for interior242·237·17<0·0001*
Oak treatment × zone for ecotone252·90·79  0·4582
Table 3.   Species appearing on more plots in 2005 than in 2000
SpeciesPearson CorrelationDetected seed bankGrowth formWetland statusDuration
  1. Species shown had Pearson correlations of presence–absence changes exceeding 0·30.

  2. *Although Eleocharis spp. in the standing vegetation were not flowering, they are, thus, identified only to genus.

  3. Eleocharis spp. in the region are wetland obligates.

Panicum anceps0·69027YesHerbFacultativePerennial
Oldenlandia boscii0·31944YesHerbObligatePerennial
Eupatorium compositifolium0·56309YesHerbUplandPerennial
Mecardonia acuminata0·53452YesHerbObligatePerennial
Panicum verrucosum0·53452YesHerbObligateAnnual
Polypremum procumbens0·50709YesHerbUplandA/P
Erechtites hieracifolius0·47673NoHerbUplandAnnual
Cyperus haspan0·44721YesHerbObligatePerennial
Dichanthelium wrightianum0·44721YesHerbUplandPerennial
Acalypha gracilens0·44544YesHerbUplandAnnual
Carex festucacea0·41703YesHerbObligatePerennial
Phytolacca americana0·41703YesHerbUplandPerennial
Diodia virginiana0·38983YesHerbObligateA/P
Iva microcephala0·38592YesHerbObligateAnnual
Ludwigia linearis0·38592YesHerbObligatePerennial
Polygonum hydropiperoides0·38592YesHerbObligatePerennial
Dichanthelium acuminatum0·38139YesHerbFacultativePerennial
Carex verrucosa0·35355NoHerbObligatePerennial
Eleocharis spp.0·35355NA*HerbObligate†A/P
Eupatorium capillifolium0·35355YesHerbUplandPerennial
Saccharum giganteum0·35355NoHerbObligatePerennial
Dichanthelium erectifolium0·34783YesHerbObligatePerennial
Rubus cuneifolius0·33806YesWoodyUplandPerennial
Ludwigia glandulosa0·31944YesHerbObligatePerennial
Pluchea camphorata0·31944YesHerbObligateA/P
Scleria ciliata0·31944YesHerbFacultativePerennial

Seed bank contributions and standing vegetation

Seedling emergence from soil samples totalled 64 species and an additional three genera not separated to species level. Mean (±SE) seedling density was 1046 (±103) per m2 and mean (±SE) species richness was 34·6 (±3·0) per wetland. Very small emergents that did not survive until identification were not included, which represent less than 7% of unidentified individual germinants. The seed bank was dominated by herbaceous species (65 of 67 taxa), the majority of which were perennial. Over 70% of the species were classified as obligate or facultative wetland species, including many of the most common such as Oldenlandia boscii (DC.) Chapman, Juncus elliottii Chapman, Cyperus pseudovegetus Steud. and Mecardonia acuminata (Walter) Small. Nearly half (28) of the seed bank species were found in the standing vegetation only in the post-treatment hardwood removal plots. Many species that correlated with increased presence over time in field sites were also present in the seed bank (Table 3).


  1. Top of page
  2. Summary
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

This study demonstrates that the reconnection of feedback cycles between the vegetation and disturbance dynamics was a necessary initial goal to cross an alternative state threshold (Suding et al. 2004; Groffman et al. 2006; Nowacki & Abrams 2008). In this case, removal of the dense hardwood canopy was necessary to encourage herbaceous fuels as a means to overcome the feedback cycles reinforcing the hardwood community state (Fig. 4). The initial lack of fuels during the transition period subsequent to canopy removal necessitated a period of adaptive management to prevent hardwood re-sprouts from re-establishing dominance (Provencher et al. 2001). During this transition, the additional management reinforced the feedback between fine fuels, regular fire and herbaceous dominance.


Figure 4.  Conceptual model of community states. Each state is maintained by feedback loops within the landscape scale disturbance process (fire).

Download figure to PowerPoint

Although the un-manipulated depression wetlands in our study were included in portions of the landscape managed with frequent prescribed fire, these hardwood communities remained as a fire shadow and consequently changed little over time. The hardwood encroachment pattern we observed is consistent with our conjecture that fire-sensitive species such as oaks become established in an initial central fire shadow, and expand with an increasing radius of fire-suppressing hardwood litter. As hardwood species tolerant of a range of moisture conditions gain dominance, they tend to homogenize formerly distinct and diverse vegetation across soil moisture gradients. Furthermore, it was difficult if not impossible to locate any small wetlands in the herbaceous-dominated state as appropriate reference sites. It is likely that only high-intensity catastrophic fires during extreme drought would have pushed the hardwood community back across the state transition threshold and returned herbaceous dominance. However, given regional fragmentation and current management of longleaf pine sites with frequent, low-intensity prescribed fire, such stand replacing fires are extremely unlikely (Platt & Schwartz 1990). Therefore, intervention (i.e. hardwood removal) is required to re-establish herbaceous dominance.

Following hardwood removal treatments, a multi-layered ground cover flora containing many desirable wetland obligate and facultative grasses and sedges developed. The seed bank probably contributed many perennial, wetland-specific species to the re-vegetation process. The important role of the seed bank has been described for other south-eastern coastal plain depression wetlands, such as Carolina bays (Kirkman & Sharitz 1994; Singer 2001; Mulhouse et al. 2005; DeSteven et al. 2006) and more mesic longleaf sites (Cohen et al. 2004). Yet, our finding of a viable, species-rich seed bank at sites dominated by hardwoods for 50–60 years is unique. Although validation of the longevity of the seed bank is outside the scope of this paper, the persistence of these species in the soil is supported by the absence of many of them from the standing vegetation of both the depression wetland interiors and adjacent ecotones at un-manipulated sites during the 5 years of the study. Attrition over years of dormancy may explain why seedling density is somewhat lower than that reported in Carolina bays (Kirkman & Sharitz 1994; Singer 2001). At the same time, these depression wetland seed banks appear to be more species rich and dense than short-term seed banks in upland longleaf sites (Cohen et al. 2004). The distinctness of species composition of the uplands, illustrated by the NMS ordination plot, indicates that adjacent communities are unlikely to be a dominant source of wetland propagules. Additional studies are needed to determine the relative roles of the seed bank and seed dispersal in the initial vegetation, and continued monitoring of the sites may indicate that dispersal plays an increasing role in changes in the vegetation community over time (Laughlin et al. 2008).

Although the study wetlands are transitioning to a desired community condition, we recognize that the present species composition is not necessarily an endpoint. Rather, restoration and management is an iterative process, requiring continued monitoring. Based on the current herbaceous community, future considerations for these sites could also include experimental reintroductions of species thought to be appropriate for the habitat but absent from the present community. In particular, rare and threatened species associated with the types of communities that have developed could also be identified and prioritized for reintroduction. The recruitment of some upland ruderal species including dog fennel Eupatorium spp., fireweed Erechtites hieraciifolia (L.) Raf. ex DC, and pokeweed Phytolacca americana L. in the hardwood removal sites is probably attributable to several dry years following hardwood removal. These less desirable species are likely to be eliminated once inundated conditions are present.

This fire-maintained, species-rich and regionally rare herbaceous community state may be a desirable restoration goal for some depression wetlands currently dominated by a few generalist hardwood species. We predict that progressive increases in herbaceous vegetation will permit future fires to burn from adjacent upland areas and maintain the herbaceous community state. Once the positive feedback with fine fuels is re-established, the regular prescribed fire already applied in the landscape should maintain the herbaceous communities. However, sites will be monitored for hardwood sprouts that escape burns, particularly at the centre of the depression wetlands. Regular monitoring will allow any additional management to occur before hardwoods reach fire-resistant stages (Guerin 1993).

Future restoration goals may increasingly emphasize the rarity or near elimination of alternative community states, in many cases within the context of restoring and maintaining landscape heterogeneity and γ-diversity (Holling 1973; Choi 2004, 2007; Asbjornsen et al. 2005; Hobbs 2007; Kirkman et al. 2007; Nowacki & Abrams 2008). Our findings confirm that it is possible to re-establish a fire-maintained herbaceous community in depression wetlands dominated by more generalist oak species. Such management may be particularly successful in cases where ecological legacies, such as persistent seed banks, are present that facilitate the resilience of the community. Our approach incorporating community dynamics and ecological thresholds may serve as a model to inform restoration strategies for component communities within the longleaf pine ecosystem, as well as other disturbance-maintained ecosystems.


  1. Top of page
  2. Summary
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We acknowledge the intellectual contribution of Dr Barry Moser, deceased, for the statistical analyses of this study. We thank Analie Barnett, the Nature Conservancy, for further statistical support. Many others provided valuable assistance to this project, including Melanie Kaeser, Kim Coffey, Heather Summer, Jimmy Atkinson, Mike Conner, Liz Cox, Michelle Creech and numerous field assistants. R.J. Mitchell, P.C. Goebel, R.J. Naiman, M. Hunter and several anonymous reviewers provided helpful comments on drafts of this manuscript. Funding was provided by the Joseph W. Jones Ecological Research Centre, the Robert W. Woodruff Foundation and the University of Georgia Graduate School.


  1. Top of page
  2. Summary
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Asbjornsen, H., Brudvig, L.A., Mabry, C.M., Evans, C.W. & Karnitz, H.M. (2005) Defining reference information for restoring ecologically rare tallgrass oak savannas in the Midwestern United States. Journal of Forestry, 103, 345350.
  • Beisner, B.E., Haydon, D.T. & Cuddington, K. (2003) Alternative stable states in ecology. Frontiers in Ecology and the Environment, 1, 376382.
  • Bennett, E.M., Cumming, G.S. & Peterson, G.D. (2005) A systems model approach to determining resilience surrogates for case studies. Ecosystems, 8, 114.
  • Bridges, E.L. & Orzell, S.L. (1989) Longleaf pine communities of the west Gulf Coastal Plain. Natural Areas Journal, 9, 246263.
  • Brockway, D.G. & Outcalt, K.W. (2000) Restoring longleaf pine wiregrass ecosystems: hexazione application enhances effects of prescribed fire. Forest Ecology and Management, 137, 121138.
  • Choi, Y.D. (2004) Theories for ecological restoration in changing environments: toward ‘futuristic’ restoration. Ecological Research, 19, 7581.
  • Choi, Y.D. (2007) Restoration ecology to the future: a call for a new paradigm. Restoration Ecology, 15, 351353.
  • Clewell, A.F. (1989) Natural history of wiregrass (Aristida stricta Michx., Gramineae). Natural Areas Journal, 9, 223233.
  • Cohen, S., Braham, R. & Sanchez, F. (2004) Seed bank viability in disturbed longleaf pine sites. Restoration Ecology, 12, 503515.
  • DeSteven, D. & Toner, M.M. (2004) Vegetation of upper coastal plain depression wetlands: environmental templates and wetland dynamics within a landscape framework. Wetlands, 24, 2342.
  • DeSteven, D., Shartiz, R.R., Singer, J.H. & Barton, C.D. (2006) Testing a passive revegetation approach for restoring coastal plain depression wetlands. Restoration Ecology, 14, 452460.
  • Drew, M.B., Kirkman, L.K. & Gholson, A.K., Jr (1998) The vascular flora of Ichauway, Baker County, Georgia: a remnant longleaf pine/wiregrass ecosystem. Castanea, 63, 124.
  • Freckmann, R.W. & Lelong, M.C. (1993) Dichanthelium. The Flora of North America (ed. Flora of North America Editorial Committee), pp. 406450. Oxford University Press, New York.
  • Glitzenstein, J.S., Platt, W.J. & Streng, D.R. (1995) Effects of fire regime and habitat on tree dynamics in north Florida longleaf pine savannas. Ecological Monographs, 65, 441476.
  • Goebel, P.C., Palik, B.J., Kirkman, L.K. & West, L. (1997) Field Guide: Landscape Ecosystem Types of Ichauway. Technical Report No. 97-1, Joseph W. Jones Ecological Research Center at Ichauway, Newton, GA.
  • Goolsby, D. (2006) Patterns of landuse and succession in a longleaf pine forest. MS Thesis, University of Georgia, Athens, GA, USA.
  • Groffman, P.M., Baron, J.S., Blett, T., Gold, A.J., Goodman, I., Gunderson, L.H., Levinson, B.M., Palmer, M.A., Paerl, H.W., Peterson, G.D., Poff, N.L., Rejeski, D.W., Reynolds, J.F., Turner., M.G., Weathers, K.C. & Weins, J. (2006) Ecological thresholds: the key to successful environmental management or an important concept with no practical application? Ecosystems, 9, 113.
  • Guerin, D.N. (1993) Oak dome clonal structure and fire ecology in a Florida longleaf pine dominated community. Bulletin of the Torrey Botanical Club, 120, 107114.
  • Harrington, T.B. & Edwards, M.B. (1999) Understory vegetation, resources availability, and litterfall responses to pine thinning and woody vegetation control in longleaf pine plantations. Canadian Journal of Forest Research, 29, 10551064.
  • Hiers, J.K., O’Brien, J.J., Will, R.E. & Mitchell, R.J. (2007) Forest floor depth mediates understory vigor in xeric Pinus palustris ecosystems. Ecological Applications, 17, 806814.
  • Hinman, S.E. & Brewer, J.S. (2007) Responses of two frequently-burned wet pine savannas to an extended period without fire. Journal of the Torrey Botanical Society, 134, 512526.
  • Hobbs, R.J. (2007) Setting effective and realistic restoration goals: key directions for research. Restoration Ecology, 15, 354357.
  • Holling, C.S. (1973) Resilience and stability in ecological systems. Annual Review of Ecology and Systematics, 4, 125.
  • Jacqmain, E.I., Jones, R.H. & Mitchell, R.J. (1999) Influences of frequent cool-season burning across a soil moisture gradient on oak community structure in longleaf pine ecosystems. American Midland Naturalist, 141, 85100.
  • Kane, J.M., Varner, J.M. & Hiers, J.K. (2008) The burning characteristics of southeastern oaks: discriminating fire facillitators from fire impeders. Forest Ecology and Management, 256, 20392045.
  • Kenward, M.G. & Roger, J.H. (1997) Small sample inference for fixed effects from restricted maximum likelihood. Biometrics, 53, 983997.
  • Kirkman, L.K. & Sharitz, R.R. (1994) Vegetation disturbance and maintenance of diversity in intermittently flooded Carolina bays in South Carolina. Ecological Applications, 4, 177188.
  • Kirkman, L.K., Drew, M.B., West, L.T. & Blood, E.R. (1998) Ecotone characterization between upland longleaf pine/wiregrass stands and seasonally-ponded isolated wetlands. Wetlands, 18, 346364.
  • Kirkman, L.K., Goebel, P.C., West, L., Drew, M.B. & Palik, B.J. (2000) Depression wetland ecosystem development. Wetlands, 20, 373385.
  • Kirkman, L.K., Coffey, K.L., Mitchell, R.J. & Moser, E.B. (2004a) Ground cover recovery patterns and life-history traits: implications for restoration obstacles and opportunities in a species-rich savanna. Journal of Ecology, 92, 409421.
  • Kirkman, L.K., Goebel, P.C., Palik, B.J. & West, L.T. (2004b) Predicting plant species diversity in a long leaf pine landscape. Ecoscience, 11, 8093.
  • Kirkman, L.K., Mitchell, R.J., Kaeser, M.J., Pecot, S.D. & Coffey, K.L. (2007) The perpetual forest: utilizing undesirable species as a bridge for restoration. Journal of Applied Ecology, 44, 604614.
  • Kruskal, J.B. (1964) Nonmentric multidimensional scaling: a numerical method. Psychometrika, 29, 115129.
  • Laughlin, D.C., Bakker, J.D., Daniels, M.D., Moore, M.M., Casey, C.A. & Springer, J.D. (2008) Restoring plant species diversity and community composition in a ponderosa pine-bunchgrass ecosystem. Plant Ecology, 197, 139151.
  • Lemon, P.C. (1949) Successional responses of herbs in the longleaf-slash pine forest after fire. Ecology, 30, 135145.
  • Littell, R.C., Milliken, G.A., Stroup, W.W. & Wolfinger, R.D. (1996) sas System for Mixed Models. SAS Institute, Cary, NC.
  • Maliakal, S.K., Menges, E.S. & Denslow, J.S. (2000) Community composition and regeneration of Lake Wales Ridge wiregrass flatwoods in relation to time-since-fire. Journal of the Torrey Botanical Society, 127, 125138.
  • Martin, K.L. (2006) Restoration of Fire Dependent Wetlands. MS Thesis, University of Georgia, Athens, GA, USA.
  • Mayer, A.L. & Rietkerk, M. (2004) The dynamic regime concept for ecosystem management and restoration. BioScience, 54, 10131020.
  • McCune, B. & Grace, J.B. (2002) Analysis of Ecological Communities. MJM Software Design, Gleneden Beach, OR.
  • McNab, W.H. & Avers, P.E. (1994) Ecological Subregions of the United States: Section Descriptions. Administrative Publication WO-WSA-5. United States Department of Agriculture, Forest Service, Washington, DC.
  • Menges, E.S., Abrahamson, W.G., Givens, K.T., Gallo, N.P. & Layne, J.N. (1993) Twenty years of vegetation change in 5 long-unburned Florida plant communities. Journal of Vegetation Science, 4, 375386.
  • Mulhouse, J.M., De Steven, D., Lide, R.F. & Shartiz, R.R. (2005) Effects of dominant species on vegetation change in Carolina bay wetlands following a multi-year drought. Journal of the Torrey Botanical Society, 132, 411420.
  • NatureServe (2007) NatureServe Explorer: An Online Encyclopedia of Life [web application] Version 6·2. NatureServe, Arlington, VA. Available at: (accessed 17 December 2007).
  • Nowacki, G.J. & Abrams, M.D. (2008) The demise of fire and “mesophication” of forests in the eastern United States. BioScience, 58, 123138.
  • Peet, R.K. & Allard, D.J. (1993) Longleaf pine-dominated vegetation of the southern Atlantic and eastern Gulf Coast region, USA. Proceedings of the Tall Timbers Fire Ecology Conference, No. 18 (ed. S.Hermann), pp. 4582. Tall Timbers Research Station, Talahassee, FL.
  • Peterson, G.D. (2002a) Forest dynamics in the southeastern US: managing multiple stable states. Resilience and the Behaviour of Large Scale Systems (eds L.H.Gunperson & L.Pritchard), pp. 227246. Island Press, Washington, DC, USA.
  • Peterson, G.D. (2002b) Estimating resilience across landscapes. Conservation Ecology, 6, 17.
  • Platt, W.J. & Schwartz, M.W. (1990) Temperate hardwood forests. Ecosystems of Florida (eds R.L.Myers & J.J.Ewel), pp. 194229. University of Central Florida Press, Orlando, FL.
  • Platt, W.J., Glitzenstein, J.S. & Streng, D.R. (1991) Evaluating pyrogenicity and its effects on vegetation in longleaf pine savannas. Proceedings of the Tall Timbers Fire Ecology Conference, No. 17 (ed. S.Hermann), pp. 143161. Tall Timbers Research Station, Talahassee, FL.
  • Poiani, K.A. & Johnson, W.C. (1988) Evaluation of the emergence method in estimating seed bank composition of prairie wetlands. Aquatic Botany, 32, 9197.
  • Provencher, L., Herring, B.J., Gordon, D.R., Rodgers, H.L., Galley, K.E.M., Tanner, G.W., Hardesty, J.L. & Breannan, L.A. (2001) Effects of hardwood reduction techniques on longleaf pine sandhill vegetation in northwest Florida. Restoration Ecology, 9, 1327.
  • Quinn, G.P. & Keough, M.J. (2002) Experimental Design and Analysis for Biologists. Cambridge University Press, New York.
  • Schabenberger, O. & Pierce, F.J. (2002) Contemporary Statistical Models for the Plant and Soil Sciences. CRC Press, Boca Raton, FL.
  • Semlitsch, R.D. & Bodie, J.R. (1998) Are small, isolated wetlands expendable? Conservation Biology, 12, 11291133.
  • Sharitz, R.R. (2003) Carolina bay wetlands: unique habitats of the southeastern United States. Wetlands, 23, 550562.
  • Singer, J.H. (2001) Effects of overstory removal and fire on wetland vegetation and recruitment from the seed bank in a hyrologically restored Carolina bay wetland. MS Thesis, University of Georgia, Athens, GA, USA.
  • Suding, K.N., Gross, K.L. & Houseman, G.R. (2004) Alternative states and feedbacks in restoration ecology. Trends in Ecology and Evolution, 19, 4653.
  • Sun, G., McNulty, S.G., Shepard, J.P., Amatya, D.M., Riekerk, H., Comerford, N.B., Skaggs, W. & Swift, L. (2001) Effects of timber management on the hydrology of wetland forests in the southern United States. Forest Ecology and Management, 143, 227236.
  • Sutter, R.D. & Kral, R. (1994) The ecology, status, and conservation of two non-alluvial wetland communities in the south Atlantic and eastern Gulf coastal plain, USA. Biological Conservation, 68, 235243.
  • USDA, NRCS. (2006) The PLANTS Database (, 18 October 2006). National Plant Data Centre, Baton Rouge, LA.
  • Walker, J. & Peet, R.K. (1983) Composition and species diversity of pine-wiregrass savannas of the Green Swamp, North Carolina. Vegetatio, 55, 163179.
  • Wright, S.P. (1998) Multivariate Analysis Using the MIXED Procedure. Paper 229 in Proceedings of the Twenty-third Annual sas Users Group International Conference. SAS Institute, Cary, NC (
  • Wunderlin, R.P. & Hansen, B.F. (2003) Guide to the Vascular Plants of Florida. University Press of Florida, Gainesville, FL.