Journal of the Torrey Botanical Society 140(4), 2013, pp. 453–464

Multiple factors influence the vegetation composition of Southeast U.S. wetlands restored in the Wetlands Reserve Program 1,2 Diane De Steven 3 USDA Forest Service, Southern Research Station, Center for Bottomland Hardwoods Research, Stoneville, MS 38776 Joel M. Gramling Department of Biology, The Citadel, Charleston, SC 29409

DE STEVEN, D. (USDA Forest Service, Southern Research Station, Center for Bottomland Hardwoods Research, Stoneville, MS 38776) AND J. M. G RAMLING (Department of Biology, The Citadel, Charleston, SC 29409). Multiple factors influence the vegetation composition of Southeast U.S. wetlands restored in the Wetlands Reserve Program. J. Torr. Bot. Soc. 140: 453–464. 2013.—Degradation of wetlands on agricultural lands contributes to the loss of local or regional vegetation diversity. The U.S. Department of Agriculture’s Wetlands Reserve Program (WRP) funds the restoration of degraded wetlands on private ‘working lands’, but these WRP projects have not been studied in the Southeast United States. Wetland hydrogeomorphic type influences hydrodynamics and thus the vegetation of restored sites, but species composition may also be affected by prior land-condition and restoration methods. We examined the variation in restored wetland vegetation of 61 WRP sites (representing 52 projects) across the Southeast region. Field surveys identified the common species at each site, and species composition was analyzed in relation to hydrogeomorphic type and specific restoration methods that were linked to pre-restoration habitat status. At least 380 plant species were recorded across all sites. Site floristic composition generally reflected variation in wetness conditions and vegetation structure. Wetlands restored by ‘non-intensive’ methods overlapped in species composition irrespective of hydrogeomorphic type, as a consequence of successional dynamics related to natural hydrologic variation. More distinctive species composition occurred in wetlands restored by ‘intensive’ methods designed to compensate for intense agricultural land-use before restoration. In the Southeast U.S., WRP wetlands are supporting a variety of plant assemblages influenced by hydro- geomorphic settings, site land-use history, and differing restoration approaches. Key words: Conservation Effects Assessment Project (CEAP), hydrogeomorphic (HGM) classification, vegetation diversity, wetland restoration, Wetlands Reserve Program (WRP).

Wetland degradation caused by drainage and agricultural activity can result in signifi- cant loss of vegetation diversity and other 1 The study was funded through the USDA ecological benefits such as wildlife habitat and Natural Resources Conservation Service (Conserva- tion Effects Assessment Project), the USDA Forest water-quality improvement. In the United Service Southern Research Station, and a Forest States, a mechanism for restoring degraded Service Joint Venture Agreement with The Citadel wetlands on private agricultural lands is the Graduate College. Wetlands Reserve Program (WRP), one of 2 The authors are grateful for outstanding support from CEAP-Wetlands Science Coordinator Diane several conservation programs administered Eckles and from NRCS State and Field Office staffs, by the U.S. Department of Agriculture especially WRP coordinators Glenn Sandifer (South (USDA). Agricultural uses on such ‘working Carolina), Mike Oliver (), and Keith lands’ may include row-cropping, pasture, and Wooster (). We also thank the landowners for allowing access to their WRP sites. Steven rangeland grazing. The WRP offers private Hughes provided technical assistance, with addi- landowners a cost-share for wetland restora- tional help from Kristie Burr, Cody Clark, Anne tion treatments and also pays for acquiring Cubeta, and Kristi Wharton. Content and statistical long-term conservation easements on the reviews were supplied by Beverly Collins, Danny Gustafson, and Tom Dell. Use of trade or firm restored sites. Easement tracts can include names in this publication is for information and some extent of upland buffer land, and the does not imply endorsement by the U.S. Depart- landowners retain rights to limited uses such ment of Agriculture of any product or service. as hunting and fishing. By 2008 there were 3 Author for correspondence, E-mail: ddesteven . @fs.fed.us 10,000 WRP easements on nearly 2 million Received for publication January 7, 2013, and in acres nationwide, mostly in farming-intensive revised form September 2, 2013. regions, at a total cost of . $2 billion since the

453 454 JOURNAL OF THE TORREY BOTANICAL SOCIETY [V OL . 140 program’s inception in 1992 (NRCS 2009). In The initial analyses found few differences light of the large expenditures for the WRP among wetland HGM types or prior habitats and other USDA conservation programs, the in the generalized attributes that indicated Natural Resources Conservation Service natural wetland vegetation (e.g., high per- (NRCS) developed the Conservation Effects centages of wetland species and native Assessment Project (CEAP) as a multi-faceted species) (De Steven and Gramling 2012). effort to evaluate the ecological benefits of However, this does not imply a lack of these programs (Mausbach and Dedrick 2004, floristic variety among the restored wetlands. Duriancik et al. 2008). The underlying species composition could The general objective for WRP restorations vary more widely in relation to multiple is to recover wetland hydrologic function, factors, including how specific restoration native wetland vegetation, and desired wildlife methods were used in relation to HGM type habitat in relation to programmatic and and prior habitat. Nearly all depressions and landowner goals (cf. De Steven and Gramling flats were restored ‘non-intensively’, i.e., by 2012). The outcomes of the WRP for vegeta- installing ditch plugs or simple water-control tion composition and wildlife habitat have structures to block artificial drainage and by been studied in the farming-intensive Southern relying mainly on passive revegetation. The Mississippi Alluvial Valley region (e.g., Twedt methods for riverine wetlands differed with 2002 et al., Heard et al. 2005, King et al. 2006, prior-habitat condition. Restoration of timber- Faulkner et al. 2011). However, in the mixed harvested floodplains was also ‘non-intensive’, forested-agricultural Southeast region (Pied- using simple methods of breaching roads or mont and Coastal Plain), a CEAP review dikes to promote river flows and allowing found that even the basic features of WRP forests to regenerate naturally. In contrast, projects were largely unknown (De Steven and prior-agriculture floodplain and headwater Lowrance 2011). The Southeast has a great tracts were treated with a more ‘intensive’ variety of wetlands and associated plant dual-practice approach, usually without recon- communities, yet there was little information nection to major river flows. The approach on which wetland hydrogeomorphic (HGM) consisted of installing semi-enclosed water- types were being restored on program lands. fowl-habitat impoundments on part of the This is relevant for assessing conservation project tract, plus widespread planting of tree benefits because HGM types represent mor- seedlings across the remaining area. Impound- phological forms (e.g., depressional, flat, ments are managed with periodic water draw- riverine) with different hydrodynamics, which downs to encourage seasonal growth of food leads to differences in some ecological traits for waterfowl, a system called moist-soil and functions (Brinson 1993, Brinson and management (Strader and Stinson 2005). Tree Reinhardt 1998). Consequently, a CEAP seedlings are planted in systematic arrays of survey of over 100 WRP restoration projects multiple species, typically bottomland oaks was conducted in 2010 across several South- and other floodplain trees. east states (De Steven and Gramling 2012). This mix of non-intensive and intensive The project plans were found to encompass restoration treatments could likely amplify four wetland types (depression, flat, riverine the floristic variation among restored wetlands headwater, riverine floodplain) and various beyond any influence of wetland type or other restoration methods (see below). Aerial imag- factors. In this paper, we describe the species ery indicated that ‘prior-habitat’ conditions composition of Southeast WRP wetlands in also differed among the project wetlands greater detail. We specifically examine whether before restoration, such that some had been the interactions between wetland type and the in active agriculture whereas others already distinctive restoration approaches linked to appeared to be semi-naturally vegetated (in- prior-habitat condition (i.e., moist-soil man- cluding bottomlands that had been timber- agement, intensive tree-planting) have pro- harvested). Regardless of wetland type or duced variability in species composition that prior condition, field data indicated that most would not be apparent from considering restorations were successful in providing wetland type alone. The findings provide a functional wetland habitats dominated by general picture of the extent of vegetation native wetland vegetation (De Steven and diversity among WRP program wetlands Gramling 2012). across the Southeast region. 2013] DE STEVEN AND GRAMLING: RESTORED WETLANDS 455

Table 1. Features of 52 WRP projects classed by wetland HGM type. Percent wetland area is estimated as the % of easement-tract area (ha) with mapped hydric soils. Project age is years since restoration. ‘Prior- agriculture’ sites were in row-crops or pasture before restoration.

Easement area % wetland area Project age HGM type n # prior-agriculture (mean ha [range]) (mean 6 SE) (range of years) Depression 14 8 26.7 [4–94] 64 6 5 2–11 Wet flat { 11 5 135.3 [5–779] 64 6 8 2–11 Riverine headwater 13 6 33.4 [6–167] 83 6 3 3–8 Riverine floodplain 14 7 149.3 [4–325] 85 6 5 3–10 { includes 3 very large (81–779 ha) ‘Carolina bays’.

Materials and Methods. SITES AND SURVEYS . indicator categories (Reed 1997) were assigned Field data had been collected on a stratified- to each species and grouped into three classes random subsample of 53 WRP projects chosen based on ACOE 1987 protocols: true ‘wetland’ from the 109 projects in the CEAP survey species (OBL and FACW categories), ‘facul- (details in De Steven and Gramling 2012). For tative’ species often occurring in wetlands this paper we excluded one outlier project with (FAC + and FAC categories), and ‘upland’ recent management disturbance, leaving 52 (non-wetland) species (FAC-, FACU, UPL projects for analysis (Table 1). The projects categories). Wetland and facultative species ranged across three states (, collectively are ‘hydrophytic’ and indicate Georgia, and ‘Coastal-Plain’ Mississippi) and ‘wetland vegetation’ when they comprise $ multiple sub-regions (Piedmont, Hilly Coastal 50 % of all species (ACOE Environmental Plain, Coastal Flats, MS Loess Uplands). Laboratory 1987). Species were also classed Project age (i.e., time since restoration was by growth habit and life history (e.g., herbace- implemented) averaged six years. Four wetland ous forb/graminoid, annual/perennial, woody), HGM types were represented: depressions, flats, native status (USDA 2012), and regional status riverine headwater areas on low-order (1 st –3 rd ) as a potentially ‘invasive’ exotic (derived from streams, and mainstream floodplains on higher- www.invasive.org/south/ and www.invasiveplan- order rivers. Project (tract) sizes varied widely, tatlas.org). partly as a natural consequence of geomorphic As an indicator of hydrology, water depths difference among HGM types. The project were measured at each location and assigned wetlands also had contrasting ‘prior-habitat’ qualitative ranks of none, shallow ( , 0.5 m), condition, with about half having been in or deep ( . 0.5 m) [ranked respectively as 0, 1, active agriculture (cropping or grazing) and the 2]. Sampling occurred in late summer, when rest already appearing more naturally vegetat- the absence or presence of water can indicate ed before restoration (Table 1). The ‘prior- relative seasonal hydroperiods from shorter to vegetated’ projects included depressions, flats, longer duration. Other hydrology indicators and headwaters altered by past ditching and (e.g., water lines) were also noted to aid in drainage, as well as forested floodplains interpreting wetness conditions (see De Steven disturbed by past timber harvest and the and Gramling 2012). associated logging roads. References for plant identification included Field data were collected in July–August Radford et al. (1968), Godfrey and Wooten 2010 with methods adapted from ‘‘routine (1981), Godfrey (1988), and Weakley (2010). wetland determination’’ procedures (ACOE For consistency, species nomenclature in this Environmental Laboratory 1987). Each proj- paper follows the PLANTS database (USDA ect was assessed with spot-surveys at 1–4 2012, at http://plants.usda.gov). distributed sample locations; number of loca- tions was determined by tract size and cover DATA ANALYSIS . The partial linkages among types (details in De Steven and Gramling HGM type, prior-habitat status, and restora- 2012). At each location, we traversed a broad tion approach (non-intensive vs. intensive) area and recorded all dominant plant species were used to classify wetlands for vegetation in each of four strata (tree, sapling/shrub, analysis. Nine riverine projects that used the herb, woody vine), where ‘dominants’ were ‘dual-practice’ approach were split into sub- species representing roughly 20 % or more of areas for the two contrasting intensive treat- stratum cover by visual estimate. Wetland ments (water-managed vs. tree-planted), giving 456 JOURNAL OF THE TORREY BOTANICAL SOCIETY [V OL . 140

Differences in the descriptive variables among groups were tested by analysis of variance (ANOVA) in SYSTAT H (SPSS, Inc.), using Tukey’s tests for group-mean comparisons. Diagnostics did not indicate any substantive departures from model assumptions, but the arcsine-square root transformation was ap- plied generally to proportion data to improve homogeneity of variances. Site floristic composition was analyzed with non-metric multi-dimensional scaling (NMS), multiple-response permutation procedures (MRPP), and indicator species analysis (ISA) in PC-ORD TM (McCune and Mefford 2011). The species data were counts per site, averaged FIG . 1. Distribution of wetland HGM types, over sample locations and strata. To reduce pre-restoration habitat conditions (vegetated or agriculture) and ‘intensive’ (tree-planted or im- data noise and improve analytical stability, we pounded) vs. ‘non-intensive’ restoration approaches merged some highly similar species to generic for 61 WRP project sites. level (mainly in speciose taxa such as Cyperus and Juncus ) to obtain 318 taxa from 384 a total of 61 distinct ‘sites’ for analysis (Fig. 1). species, and then omitted very infrequent taxa # % These 61 sites were classed into six a priori site- found in only 1–3 sites ( 5 of sites; cf. groups: depressions (DEPR), flats (FLAT), McCune & Grace 2002). A 3-dimensional riverine headwaters (RIV-H), prior-forested NMS ordination was identified as optimal 5 floodplains (RIV-F), managed moist-soil im- (final stress 15.3), with site-scores rotated to poundments (MSM), and tree-planted head- principal axes for display in the 2 main water and floodplain areas (HF-PT). Sites in dimensions. Possible explanatory variables the four HGM-defined groups had all been were tested for Pearson correlation with the restored non-intensively (Fig. 1). The MSM ordination scores. These variables included group included the water-managed areas of wetness score, FAQWet4 index, percent non- prior-cropped floodplains and headwaters, native species, project age, and number of plus two prior-cropped flats that had been vertical strata present (from 1–3 for herb, herb modified into managed impoundments. The + sapling, herb + sapling + tree) as an indicator tree-planted sites (HF-PT) were all associated of overall vegetation structure. Significant with prior-cropped floodplains and headwa- (P , 0.05), non-redundant variables were ters. Project age averaged 5–7 years in each displayed as vectors in the ordination graph. group. MRPP tested whether the six site-groups Data were compiled over sample locations accounted for more compositional variation at each of the 61 sites (1–4 locations per site). than if sites were classed only into four HGM Variables characterizing each site included a types; lower (more negative) values of the five-level wetness score (0, 0.5, 1, 1.5, 2) based T-statistic indicate stronger group separation on the average water-depth ranks, and the (McCune and Grace 2002). MRPP was also relative proportions of native species, hydro- used to test whether composition differed phytic species (FAC or wetter), and wetland generally with prior-habitat status (agriculture species (FACW and OBL only). An index of vs. vegetated). The ISA was conducted on the ‘floristic quality’ (FAQWet4 of Ervin et al. same data matrix, where species indicator 2006) was also calculated for each site. values (I.V.) are calculated as the product of FAQWet4 uses species wetland-indicator relative abundance and relative frequency in a scores ranging from 5 (OBL) to 25 (UPL) given group compared to all groups (McCune combined with frequency-weighted percent and Grace 2002). Indicator species for site- nativity; although lacking a fixed range, an groups were tabulated as those species having index . 0 indicates hydrophytic vegetation an I.V. significant at P , 0.10, which detected and higher values (10–20) suggest a prepon- taxa that contrasted more than one group derance of native and ‘true’ wetland species. jointly from other groups. 2013] DE STEVEN AND GRAMLING: RESTORED WETLANDS 457

ENERAL EGETATION AND ABITAT 6 Results. G V H TRAITS . The intensively restored MSM and HF-PT sites differed in general vegetation { ** ** ** attributes from each other and from the non- intensively restored groups (Table 2). Both had more distinctive indicator taxa than most other groups except forested floodplains (the

D least altered habitats). MSM sites were the D D wettest, with the highest proportion of hydro- 0.1 ** 0.5 0.3 * 1.1 2.3 5.33.0 ** ** 0.06 1.3 ** 2.1 n.s.

0.10. Site traits are means phytic and wetland species and the highest 6 6 6 6 6 6 6 6 6 6

, FAQWet4 values. These moist-soil managed 2.1 P sites were typically inundated and had mostly herbaceous species, including many wetland annuals. In contrast, HF-PT sites were com- D paratively the driest, with relatively fewer 0.30.2 1.3 0.8 2.2 6.0 2.65.73.6 80.1 48.8 47.7 0.10.0 0.06 0.8 2.6 1.1 93.9 hydrophytic and wetland species and lower 6 6 6 6 6 6 6 6 6 6 FAQWet4 values. These tree-planted headwa- 0.4 0.3 ters and floodplains had a midstory stratum of the young planted trees plus a variety of other woody and herbaceous species. D D The MSM and HF-PT sites were all prior- 1.1 14.1 0.4 2.14.03.9 92.6 61.3 32.8 0.6 2.0 98.5 0.10.2 0.7 3.0 2.3 1.6 6 6 6 6 6 6 6 6 6 6 agriculture by definition. Excepting prior-for- ested floodplains (RIV-F), sites in the other three HGM-defined groups had a mix of prior- oist-soil managed), DEPR (depression), FLAT (flat), RIV-H (riverine om all remaining group means in the row. habitat and wetness conditions, yet the four groups were similar in most general attributes 1.6 9.2 0.3 1.6 0.10.2 0.4 2.4 1.2 7.2 3.56.77.9 84.9 53.7 45.7 0.5 2.4 1.7 92.0 except for differences in vegetation structure 6 6 6 6 6 6 6 6 6 6

er or floodplain). Indicator taxa are from ISA at (Table 2). RIV-H sites averaged more potential- ly ‘invasive’ non-natives, but the mean number of such species per site was low in all groups. Mean richness of all dominant species (not shown) averaged 23 per site with no substan- 1.4 8.8 0.2 0.6 3.46.45.9 89.2 54.6 39.4 0.20.2 0.2 2.2 1.5 3.7 0.4 0.9 1.9 96.2 tive between-group differences (cf. De Steven 6 6 6 6 6 6 6 6 6 6 and Gramling 2012). Likewise, all six groups 0.10, or n.s. (not significant). $ $ 0.7 1.8 3.6 averaged 90 % native species and 80 % , 86.2 63.5 54.8

P hydrophytic species (Table 2). { FLORISTIC COMPOSITION PATTERNS . MRPP D D D D D D

0.05, tests confirmed that species composition was 2.8 0.8 0.0 0.08 0.4 1.1 9.9 0.1 5.1 3.4 0.4 0.9 2.3 95.0

, differentiated more strongly by the site-groups 6 6 6 6 6 6 6 6 6 6 P (n 5 6, T 5 216.3, P , 0.0000001) than by HGM type alone ( n 5 4, T 5 23.2, P 5 0.005).

0.01, * The NMS ordination (Fig. 2) showed that the

, prior-agriculture moist-soil managed sites P ) 100 57(MSM) 40 and tree-planted 22 sites (HF-PT) 0 differed 100 – n floristically from each other and from other of

) 98.0 groups. A significant overall difference based % ) 81.5 % % 5, 55; ** on prior-habitat condition ( n 5 2, T 5 214.0, ) 90.3 ) 20.2 5

), 93.8 % P 0.0000001) was largely attributable to the ) 12 14 10 9 7 9 – % { % n distinctive composition of those two groups (Fig. 2). There was much greater floristic

Descriptor or trait MSM DEPR FLATsimilarity RIV-H among RIV-F the four HF-PT site-groups ANOVA Sig. that had been restored by non-intensive methods. for each trait, D superscript indicates group means that differ from each other and fr { Table 2. Habitat and vegetation traits of 61 WRP sites in six site-groups: MSM (m Site species composition was significantly ‘Invasive’ non-natives (number) 0.08 FAQWet4 Index 13.2 Vegetation strata (1–3)Hydrophytic species ( Wetland species ( 1.0 GROUP DESCRIPTORS Sites in group ( Prior-agriculture sites ( headwater), RIV-F (riverine forested floodplain), HF-PT (tree-planted headwat Indicator taxa (number per group)SITE TRAITS 12 2 3 7 21 12 – Wetness score (0–2) 1.2 SE. ANOVA d.f. Herbaceous species ( Annual species ( Native species ( Non-native species (number)correlated 1.0 with site wetness (Fig. 2, vector 458 JOURNAL OF THE TORREY BOTANICAL SOCIETY [V OL . 140

facultative Dichanthelium acuminatum ) and/or various woody species including Taxodium ascendens and Liquidambar styraciflua . Most FLAT sites had a variety of herbaceous and woody taxa, particularly trees such as Acer rubrum, Nyssa biflora , T. ascendens , Diospyros virginiana, and L. styraciflua. RIV-H sites (nearly all prior-vegetated) shared many taxa with the timber-harvested RIV-F sites, which were dominated by bottomland trees (e.g., Quercus laurifolia , T. distichum , N. biflora , L. styraciflua ) and native woody vines (e.g., Ampelopsis arborea, Campsis radicans, Smilax spp., Vitis rotundifolia ). With the driest conditions (cf. Table 2), HF-PT sites were FIG . 2. NMS ordination of floristic composition distinctly characterized by certain old-field in 61 WRP sites. Legend shows symbols for six successional herbs and shrubs (e.g., Solidago groups (cf. Table 2); filled and asterisk symbols 5 ‘prior-vegetated’ sites, unfilled symbols 5 ‘prior- spp., Symphyotrichum dumosum , Andropogon agriculture’ sites. Vector arrows show significantly virginicus, Baccharis halimifolia , Lonicera ja- correlated variables, with values increasing from the ponica , Rubus spp.) in addition to the planted 0,0 centroid and vector lengths proportional to joint bottomland trees ( Quercus spp., Fraxinus correlation strength. pennsylvanica , T. distichum ) and other volun- teer trees (e.g., Platanus occidentalis , L. shown), but not with project age. A major styraciflua ). That composition contrasted the pattern of floristic variation reflected a struc- tree-planted sites from the wetter MSM tural gradient (number of vegetation strata) impoundments that were co-located on the from simple herbaceous communities to multi- same tracts. layered forests (Fig. 2, left to right). Floristic Of the 384 total species recorded as quality (FAQWet4) represented a secondary dominants, 31 (8 %) were non-native. Most compositional gradient; this partly reflected of these were naturalized old-field grasses and the prevalence of herbaceous wetland species forbs (e.g., Bromus japonicus, Cynodon dacty- in MSM and some wetter DEPR sites, versus lon, Festuca pratensis, Paspalum spp., Ipomoea greater frequency of facultative and upland spp., Verbena brasiliensis ), thus non-natives species in HF-PT and some drier DEPR sites. occurred somewhat more frequently in prior- With a range of wetness and prior-habitat agriculture sites (63 % of sites) than in prior- conditions, DEPR and FLAT sites varied in vegetated sites (46 % of sites) (chi-square test, structure from open-emergent to shrubby or d.f. 51, n.s.). Ten non-natives of various forested, and they had substantial floristic growth habits were species listed as regionally overlap with each other and with headwater invasive, though not necessarily in wetlands sites (RIV-H) and prior-forested floodplains (Table 3). The two most common were (RIV-F). The wooded RIV-H and RIV-F sites Lonicera japonica (Japanese honeysuckle) and were very similar in species composition. Ligustrum sinense (Chinese privet). L. japonica Some common taxa that accounted for was a dominant in 13 sites and seen in 3 other floristic differences among groups are listed sites. L. sinense was a dominant in 6 sites and in Appendix 1 (see USDA 2012 for nomen- seen in 6 other sites ( Note : the total of 12 for L. clature authorities). Typical plants in MSM sinense corrects an erroneous value of 15 in De sites were semi-aquatic and wetland forbs and Steven and Gramling 2012). Other invasives sedges/rushes such as quadrivalvis , such as Alternanthera philoxeroides (alligator- Polygonum hydropiperoides , Ludwigia spp., weed), Microstegium vimineum (Japanese stilt- Cyperus spp., and Eleocharis spp., with Salix grass), and Triadica sebifera (Chinese tallow) nigra as a woody colonizer. Many common were recorded sporadically in 1–3 sites each. MSM-site taxa also occurred in DEPR sites, but depressions were distinguished by several Discussion. The plant data revealed broad perennial grasses (the semi-aquatics Panicum floristic patterns that reflected the diverse hemitomon and Leersia hexandra, or the species assemblages of Southeastern U.S. 2013] DE STEVEN AND GRAMLING: RESTORED WETLANDS 459

Table 3. Potentially ‘invasive’ exotics recorded in WRP sites. Data for each species are number of sites of occurrence, by wetland group (codes in Table 2). Shown at bottom are the percentages of sampled sites per group with a non-native present, and with an invasive present.

Species Growth habit MSM DEPR FLAT RIV-H RIV-F HF-PT Alternanthera philoxeroides forb { 1 2 Ligustrum sinense shrub 1 1 2 1 1 Lonicera japonica woody vine 4 2 3 1 3 Lygodium japonicum vine 1 Melia azedarach tree 1 1 Microstegium vimineum grass 3 1 Rosa multiflora shrub 1 Sorghum halepense grass 1 1 Triadica sebifera tree 1 2 Wisteria sinensis woody vine 1 % of sites with an exotic – 50.0 57.1 40.0 88.9 28.6 66.7 % of sites with an ‘invasive’ – 8.3 28.6 30.0 77.8 28.6 55.6 { semi-aquatic species. wetlands. Most restored WRP sites had resembled those of comparable natural wet- general attributes indicative of functional lands, although at a younger successional age. wetland vegetation (see also De Steven and Most of the restored riverine-headwater sites Gramling 2012); however, they had variable were seasonally flooded woody wetlands species composition with over 380 species developing toward forest. Likewise, the tim- recorded as within-site dominants. Composi- ber-harvested floodplains experienced natural tion was partly influenced by substantial river flooding and had well-developed forest variation in site wetness that represented regrowth. Restored depressions and flats had inherent differences in hydrodynamics; these differing vegetation related to a range of differences were evident despite summer wetness conditions that reflected different drought conditions that occurred during the hydroperiods. Depression, flat, and headwater survey year. As hypothesized, floristic compo- groups were not strongly differentiated by sition was also shaped by interactions among prior-habitat status, apart from an unsurpris- wetland HGM type, prior habitat condition ing trend for more frequent presence of non- (reflecting land-use history), and restoration native species in agricultural sites. Possibly the approaches. prior-agriculture wetlands had not been The natural plant communities of South- drained or farmed so intensively that they eastern wetlands are very dynamic, owing to a lacked wetland species before restoration; if large regional pool of wetland species and to so, then remnant wetland plants could have hydrologic variability that is influenced by recolonized from seed banks or refuge areas, HGM settings (see Christensen 2000, and making those sites more similar to prior- review in De Steven and Lowrance 2011). vegetated wetlands that perhaps were aban- The result is that wetland types may differ in doned from use at a much earlier time. vegetation structure but overlap considerably The distinctive vegetation of moist-soil in species composition. Riverine headwater managed areas and tree-planted floodplains and floodplain wetlands typically become and headwaters was linked to use of active forested, since flooding regimes are seasonal restoration practices to compensate for an in relation to surface runoffs, overbank flows, intense agricultural history. Typically, these and microtopographic features that vary with larger tracts had been in continuous row position in the watershed. The natural vegeta- cropping and were isolated hydrologically by tion of depressions and flats varies more small flood-control levees or stream deepening widely; it can range from open-emergent ponds for drainage. Consequently, restoration meth- to forests according to individual differences in ods were similar to those used on marginal annual hydroperiods (from semi-permanent to farmlands in the adjacent region of the temporary) and susceptibility to fire. Southern Mississippi Alluvial Valley (Haynes In the WRP sites restored by low-intensity 2004, King et al. 2006, Faulkner et al. 2011). methods, patterns of species composition In both cases it has been impractical to remove 460 JOURNAL OF THE TORREY BOTANICAL SOCIETY [V OL . 140 the flood-control levees; instead, restoration restored WRP sites; however, that difference is uses dikes, water-control structures, and/or related not only to prior land-use but also to excavated swales to create ponded areas at the use of methods that favor species associated lower elevations, plus widespread reforestation with active vegetation management (e.g., wet- on the higher ground. These practices are also land annuals) or constrained hydrologic recov- driven by specific WRP objectives for creating ery (e.g., old-field herbs). Other wetland species waterfowl habitat and recovering bottomland- that we observed in the WRP sites likely also hardwood forest in these floodplain settings occur in natural wetlands of all types, but are (King et al. 2006). The moist-soil impound- merely unreported in the limited reference ments provide distinctive open-water wetland datasets. Since the WRP wetlands will be habitat, with periodic management that main- maintained within a matrix of private working tains early-successional emergent vegetation lands, some floristic differences from natural having more wetland annuals than would wetlands would be expected, depending on the occur in mature wetlands. Conversely, lacking wetland-management practices used and the reconnection to major riverine flows, the tree- extent of upland buffer habitat on the ease- planted sites tend to be drier than natural ment. Further monitoring would be needed to forested floodplains and have an understory determine if the WRP wetlands can retain their vegetation similar to moist old-fields (see compositional variety over the long term. Battaglia et al. 2002). Despite that hydrologic The study findings also suggest how geo- difference, active tree-planting and natural tree graphic differences in land-use and climate recruitment are addressing the habitat goals to might influence the plant diversity of restored accelerate forest development and develop a wetlands on working lands. In the farming- woody assemblage that resembles the oversto- intensive U.S. Central Plains region, prairie- ry component of floodplain forests (see also pothole and playa wetlands that are restored King et al. 2006). under USDA programs do not seem to adequately recover native plant diversity IMPLICATIONS . Wetland restoration efforts (Smith and Haukos 2002, Aronson and often seek to replicate historic plant commu- Galatowitsch 2008). Their vegetation becomes nities, hence success is evaluated by floristic more homogenized over time and is frequently comparison to natural (‘reference’) wetlands. dominated by undesirable exotics. Such wet- However, relying strictly on compositional lands often were severely damaged by cultiva- criteria and reference types can be problematic tion and are isolated from remnant natural for various reasons (National Research Coun- wetlands that might be sources of plant re- cil 2001, see also De Steven et al. 2010). As we colonization; the semi-arid Plains climate also have shown previously (De Steven and Graml- results in drier wetlands that are more ing 2012), guild-based attributes (e.g., repre- susceptible to colonization by invasive exotics sentation of wetland species, native species, (O’Connell et al. 2013). In contrast, within the and growth forms) can provide important mixed forested-agricultural Southeast U.S. supplemental data for assessing whether WRP region, the WRP enrolls wetlands with a range wetlands have recovered desired functions and of prior conditions and wetland types; the structure in relation to project goals. landscape is richer in nearby source wetlands Replicating specific plant communities is and the humid climate favors wetter hydro- not an explicit goal for WRP projects, so logic regimes. These conditions appear to deviation from ‘reference’ floristics does not promote more successful recovery and a necessarily indicate an unsuccessful restoration. greater variety of wetland plant communities, Of the common species that characterize except where hydrologic restoration has been natural depression, flat, and riverine wetlands inadequate (cf. De Steven and Lowrance 2011, (see Wharton et al. 1982, Sharitz and Gibbons De Steven and Gramling 2012). Non-native 1982, Richardson and Gibbons 1993, Sharitz species are sometimes present but rarely and Mitsch 1993, Rheinhardt and Rheinhardt dominate a site; many are likely to decline 2000, Kirkman et al 2000, De Steven and Toner over time as sites mature (e.g. McLane et al. 2004), 80–90 % occurred in non-intensively 2012). The net result is that most Southeastern restored WRP wetlands of the same type wetlands restored in the Wetlands Reserve (tabulations not shown). About 50 % of those Program are currently supporting a range of characteristic species occurred in the intensively native plant assemblages that are shaped by 2013] DE STEVEN AND GRAMLING: RESTORED WETLANDS 461 multiple factors including hydrogeomorphic conservation practices on wetland ecosystem settings, variable site history, and restoration services in the Mississippi Alluvial Valley. Ecol. Appl. 21: S31–S48. approaches linked to specific programmatic GODFREY, R. K. 1988. Trees, Shrubs, and Woody goals. Vines of Northern and Adjacent Georgia and . University of Georgia Press, Literature Cited Athens, GA. 734 p. GODFREY, R. K. AND J. W. W OOTEN. 1981. Aquatic ACOE [A RMY CORPS OF ENGINEERS] E NVIRONMENTAL and Wetland Plants of the Southeastern United LABORATORY . 1987. Corps of Engineers wetlands States (2 volumes). University of Georgia Press, delineation manual. Technical Report Y-87-1. Athens, GA. 944 p. U.S. Army Engineer Waterways Experiment HAYNES , R. J. 2004. The development of bottomland Station, Vicksburg, MS. 92 p. forest restoration in the Lower Mississippi River ARONSON, M. F. J. AND S. G ALATOWITSCH . 2008. Alluvial Valley. Ecol. Res. 22: 170–182. Long-term vegetation development of restored HEARD , L. P., C. R EWA, R. M ISSO, AND N. M ELVIN. prairie pothole wetlands. Wetlands 28: 883–895. 2005. The role of the Natural Resources Conser- BATTAGLIA, L. L., P. R. M INCHIN , AND D. W. vation Service in bottomland hardwood restora- PRITCHETT. 2002. Sixteen years of old-field tion, pp. 533–539. In L. H. Frederickson, S. L. succession and reestablishment of a bottomland King, and R. M. Kaminski [eds.], Ecology and hardwoods forest in the Lower Mississippi management of bottomland hardwood systems: Alluvial Valley. Wetlands 22: 1–17. the state of our understanding. Gaylord Mem. BRINSON, M. M. 1993. A hydrogeomorphic classifi- Lab. Special Publ. No. 10. Univ. of Missouri- cation for wetlands. Technical Report WRP-DE- Columbia, Puxico, MO. 4. U.S. Army Engineer Waterways Experiment KING, S. L., D. J. T WEDT, AND R. R. W ILSON. 2006. Sta., Vicksburg, MS. The role of the Wetland Reserve Program in BRINSON, M. M. AND R. D. R EINHARDT. 1998. conservation efforts in the Mississippi River Wetland functions and relations to societal Alluvial Valley. Wildlife Soc. B. 34: 914–920. values, pp. 29–48. In M. G. Messina and W. H. KIRKMAN, L. K., P. C. G OEBEL, L. W EST, M. B. Conner [eds.], Southern forested wetlands: ecol- DREW , AND B. J. P ALIK. 2000. Depressional ogy and management. Lewis Publishers, Boca Raton, FL. wetland vegetation types: A question of plant community development. Wetlands 20: 373–385. CHRISTENSEN, N. L. 2000. Vegetation of the south- eastern coastal plain, pp. 397–448. In M. G. MAUSBACH, M. J. AND A. R. D EDRICK . 2004. The Barbour and W. D. Billings [eds.], North length we go: Measuring environmental benefits American terrestrial vegetation. Cambridge Uni- of conservation practices. J. Soil Water Conserv. versity Press, New York, NY. 59: 97A–103A. MCCUNE, B. AND J. B. G RACE. 2002. Analysis of DE STEVEN, D. AND J. M. G RAMLING . 2012. Diverse characteristics of wetlands restored under the Ecological Communities. MjM Software, Glene- Wetlands Reserve Program in the Southeastern den Beach, OR. 300 p. United States. Wetlands 32: 593–604. MCCUNE, B. AND M. J. M EFFORD. 2011. PC-ORD: DE STEVEN, D. AND R. L OWRANCE . 2011. Agricul- Multivariate Analysis of Ecological Data. Ver- tural conservation practices and wetland ecosys- sion 6. MjM Software, Gleneden Beach, OR. tem services in the wetland-rich Piedmont- MCLANE, C. R., L. L. B ATTAGLIA , D. J. G IBSON, AND Coastal Plain region. Ecol. Appl. 21: S3–S17. J. W. G RONINGER. 2012. Succession of exotic and DE STEVEN, D., R. R. S HARITZ , AND C. D. B ARTON. native species assemblages within restored flood- 2010. Ecological outcomes and evaluation of plain forests: A test of the parallel dynamics success in passively restored Southeastern depres- hypothesis. Restor. Ecol. 20: 202–210. sional wetlands. Wetlands 30: 1129–1140. NATIONAL RESEARCH COUNCIL. 2001. Compensating DE STEVEN, D. AND M. M. T ONER. 2004. Vegetation for Wetland Losses Under the Clean Water Act. of Upper Coastal Plain depression wetlands: National Academy Press, Washington, D.C. 322 Environmental templates and wetland dynamics p. within a landscape framework. Wetlands 24: NRCS [N ATURAL RESOURCES CONSERVATION SER- 23–42. VICE]. 2009. Interim Final Benefit-Cost Analysis DURIANCIK, L. F., D. B UCKS, J. P. D OBROWOLSKI, for the Wetlands Reserve Program. U.S. Dept. of T. D REWES , S. D. E CKLES, L. J OLLEY , R. L. Agriculture, NRCS, Washington, DC. 30 p. KELLOGG, D. L UND, J. R. M AKUCH, M. P. O’C ONNELL , J. L., L. A. J OHNSON , B. J. B EAS, L. M. O’N EILL , C. A. R EWA , M. R. W ALBRIDGE , R. P ARRY , SMITH, S. T. M CMURRY, AND D. A. H AUKOS. AND M. A. W ELTZ . 2008. The first five years of 2013. Predicting dispersal-limitation in plants: The Conservation Effects Assessment Project. J. Optimizing planting decisions for isolated wet- Soil Water Conserv. 63: 185A–197A. land restoration in agricultural landscapes. Biol. ERVIN, G. N., B. D. H ERMAN , J. T. B RIED, AND D. C. Conserv. 159: 343–354. HOLLY . 2006. Evaluating non-native species and RADFORD, A. E., H. E. A HLES, AND C. R. B ELL. 1968. wetland indicator status as components of wet- Manual of the Vascular Flora of the Carolinas. lands floristic assessment. Wetlands 26: 1114– University of Press, Chapel Hill, 1129. NC. 1183 p. FAULKNER, S., W. B ARROW JR. , B. K EELAND, S. REED, J R., P. B. 1997. Revision of the National List WALLS, AND D. T ELESCO. 2011. Effects of of Plant Species that Occur in Wetlands. U.S 462 JOURNAL OF THE TORREY BOTANICAL SOCIETY [V OL . 140

Department of Interior, Fish and Wildlife watershed disturbance. Conserv. Biol. 16: 964– Service, Washington, DC. 209 p. 974. RHEINHARDT , M. C. AND R. D. R HEINHARDT. 2000. STRADER, R. W. AND P. H. S TINSON . 2005. Moist- Canopy and woody subcanopy composition of Soil Management Guidelines for the U.S. Fish wet hardwood flats in eastern North Carolina and Wildlife Southeast Region. U.S. Dept. of and southeastern . J. Torrey Bot. Soc. Interior, Fish and Wildlife Service, Jackson, MS. 127: 33–43. 44 p. RICHARDSON, C. J. AND J. W. G IBBONS. 1993. TWEDT, D. J., R. R. W ILSON, J. L. H ENNE-K ERR, AND Pocosins, Carolina bays, and mountain bogs, D. A. G ROSSHUESCH . 2002. Avian response to pp. 257–310. In W. H. Martin, S. G. Boyce, and bottomland hardwood reforestation: The first A. C. Echternacht [eds.], Biodiversity of the 10 years. Restor. Ecol. 10: 645–655. Southeastern United States: lowland terrestrial WEAKLEY , A. S. 2010. Flora of the Southern and communities. Wiley & Sons, New York, NY. Mid-Atlantic States: working draft of March 8, SHARITZ, R. R. AND J. W. G IBBONS. 1982. The ecology of Southeastern shrub bogs (pocosins) 2010. University of North Carolina Herbarium, and Carolina bays: a community profile. Report Chapel Hill, NC. 994 p. FWS/OBS-82/04. U.S. Fish and Wildlife Service, WHARTON , C. H., W. M. K ITCHENS , E. C. P ENDLETON , Washington, D.C. 53 p. AND T. W. S IPE . 1982. The ecology of bottomland SHARITZ, R. R. AND W. J. M ITSCH. 1993. Southern hardwood of the Southeast: a community floodplain forests, pp. 311–372. In W. H. Martin, profile. Report FWS/OBS-81/37. U.S. Fish and S. G. Boyce, and A. C. Echternacht [eds.], Wildlife Service, Washington, DC. 148 p. Biodiversity of the Southeastern United States: USDA [U.S. D EPARTMENT OF AGRICULTURE]. 2012. lowland terrestrial communities. John Wiley & The PLANTS Database. Retrieved August 10, Sons, New York, NY. 2012 from USDA Natural Resources Conserva- SMITH, L. M. AND D. A. H AUKOS. 2002. Floral tion Service, National Plant Data Team, Greens- diversity in relation to playa wetland area and boro, NC. ,http://plants.usda.gov . Appendix 1. Differentiated table of relative Indicator Values ( %) for frequent taxa in six WRP groups; + 5 present but I.V. , 10 %. Groups are MSM 5 moist- 2013] soil managed, DEPR 5 depression, FLAT 5 flat, Riv-H 5 riverine headwater, Riv-F 5 riverine forested floodplain, HF-PT 5 tree-planted headwater or floodplain. Significant indicator taxa are noted by * P # 0.05 or { P , 0.10.

Wetland site-group Wetland indicator Taxon class Growth habit MSM DEPR FLAT Riv-H Riv-F HF-PT ISA Sig. HERBACEOUS TAXA

Ludwigia peploides wetland forb { 50 * WETLANDS RESTORED GRAMLING: AND STEVEN DE Hydrolea quadrivalvis wetland forb { 61 + * Eleocharis (6 spp.) 1 wetland rush 28 + + + * Cyperus (9 spp.) 1 wetland sedge 39 + + + + * Diodea virginiana wetland forb 33 + * Ludwigia (palustris, spathulata ) wetland forb 36 + + + * Ludwigia (8 erect-form spp.) 1 wetland forb 14 10 + + + n.s. Polygonum hydropiperoides wetland forb { 35 13 + + * Panicum hemitomon wetland grass { + 34 + * Leersia hexandra wetland grass { + 25 * Dichanthelium acuminatum facultative grass 16 + + n.s. Woodwardia (virginiana, areolata ) wetland fern + 25 + + * Mikania scandens wetland vine + + 18 10 + n.s. Carex glaucescens wetland sedge + + 27 * Saururus cernuus wetland forb + 44 + * Boehmeria cylindrica wetland forb 19 + 17 n.s. Carex lupulina wetland sedge + 18 + + n.s. Juncus (effusus, coriaceus ) wetland rush 12 + + + 30 * Juncus (11 spp.) 1 wetland rush + 11 + + 15 n.s. Carex frankii wetland sedge + 49 * Andropogon virginicus upland grass + + 44 * Symphyotrichum dumosum facultative forb + 71 * Solidago altissima/gigantea upland forb 85 * WOODY TAXA Taxodium ascendens wetland tree 15 14 n.s. Morella cerifera facultative shrub + 10 + + n.s. Diospyros virginiana facultative tree + 19 16 + + n.s. Acer rubrum (Coastal-Plain) wetland tree + 24 16 29 + * Nyssa biflora wetland tree + 11 11 n.s. Toxicodendron radicans facultative vine + 13 25 15 + * Smilax rotundifolia facultative vine + 10 + 50 * Parthenocissus quinquefolia facultative vine + + 23 14 { Vitis rotundifolia facultative vine + + 13 20 + n.s.

Ampelopsis arborea facultative vine + + + 10 26 + * 463 Campsis radicans facultative vine + + + + 26 + * 464 ORA FTETRE OAIA OIT [V SOCIETY BOTANICAL TORREY THE OF JOURNAL Appendix 1. Continued.

Wetland site-group Wetland indicator Taxon class Growth habit MSM DEPR FLAT Riv-H Riv-F HF-PT ISA Sig. Quercus laurifolia wetland tree + + 39 * Liquidambar styraciflua facultative tree 10 16 13 22 14 n.s. Salix nigra wetland tree 25 + + 16 + + { Taxodium distichum wetland tree + + + 33 + * Fraxinus pennsylvanica wetland tree + + 12 26 * Baccharis halimifolia facultative shrub + + + + 17 n.s. Lonicera japonica # upland vine + + + + 13 n.s. Rubus spp. upland shrub + + + 45 * Quercus phellos wetland tree + + + 13 n.s. Quercus nigra facultative tree + + + 23 { Platanus occidentalis wetland tree + + 48 * Quercus (pagoda, texana ) wetland tree 44 * { semi-aquatic species. 1 includes 1 or more annual species. # non-native, considered ‘invasive’ in some habitats. OL 140.