Canopy Composition and Forest Structure Provide Restoration Targets for Low-Order Riparian Ecosystems Richard D. Rheinhardt,1,2 M. McKenney-Easterling,3 Mark M. Brinson,1 Jennifer Masina-Rubbo,1,4 Robert P. Brooks,3 Dennis F. Whigham,5 David O?Brien,6 Jeremy T. Hite,3 and Brian K. Armstrong3 Abstract Many programs are in place to protect and restore low- order streams and riparian zones. However, information on riparian zone forests is sparse for many biogeographi- cal regions, especially compositional and structural data that would provide useful targets for restoration. This study provides quantitative data on riparian zone compo- sition and forest structure from three physiographic prov- inces of eastern United States. Data from 219 low-order (first- to fourth-order) forested reaches were arranged by three basal area (BA) categories meant to represent successional categories and variations in forest structure. Detrended correspondence analysis (DCA) was used to illustrate differences among successional categories and physiographic provinces. The DCA ordination separated stands into four physiographic subregions, based on the species composition of late-successional stands. Many early to mid-successional stands (<30 m2/ha) were similar in composition to late-successional reference stands (BA ? 30 m2/ha) in the same physiographic subregion. In such sites, natural successional processes would likely be suffi- cient to restore the compositional and structural attributes inherent in late-successional stands if provided long-term protection. Other sites with dissimilar compositions may have been recovering from more intensive types of altera- tions, such as mechanized land clearing. In such sites, restoration to historic compositions could benefit func- tionally by planting oaks (Quercus spp. L.) and other heavy mast species. Key words: composition, ordination, riparian, structure, succession. Introduction Forested riparian zones are important for maintaining the physical, biological, and chemical integrity of riparian eco- systems and for buffering the impact of nonpoint source pollutants transported to them from adjacent uplands (Peterjohn & Correll 1984; Jacobs & Gilliam 1985; Phillips et al. 1993). This is especially true of late-successional riparian forests because such forests include large canopy trees, snags, downed wood, and a three-dimensional struc- ture that includes understory strata of multiple-aged can- opy trees, subcanopy trees, shrubs, and herbaceous plants. Such forests support exceptional invertebrate and verte- brate habitat (Hyatt & Naiman 2001; Wipfli et al. 2007), litter for in-stream biota (Thorp et al. 1985; Wallace et al. 1997), and soils high in organic matter to fuel denitrify- ing microbes (Lowrance et al. 1984; Groffman et al. 1992; Tesoriero et al. 2004). Linear dimension is an important spatial characteristic of streams and riparian forests because they interface directly with upland activities where most nonpoint source pollution originates (Brinson 1993; Alexander et al. 2007). Riparian zones of low-order streams, herein defined as first- to fourth-order streams (sensu Strahler 1952), are important to water quality because they comprise about 90% of a stream network?s total length (Rheinhardt et al. 1999, 2005). Headwater reaches, herein defined as first- to second-order streams, are particularly important because they comprise two-thirds of most stream networks (Leopold et al. 1964; Freeman et al. 2007). Due to their prevalence in the landscape, riparian zones of low-order streams provide enormous potential for buffering land-disturbing activities in uplands (Spruill 2000). Riparian ecosystems of headwater reaches (first- to second-order streams) are hydrologically driven by ground- water discharge and so are generally intermittent to perennial, depending on their hydrogeologic and climate setting (Winter 2007). Although most topographic con- tour maps depict intermittent and perennial streams as blue lines (dashed and solid, respectively, e.g., U.S. 1 Biology Department, East Carolina University, Greenville, NC 27858, U.S.A. 2 Address correspondence to R. D. Rheinhardt, email rheinhardtr@ecu.edu 3 Cooperative Wetlands Center, Pennsylvania State University, University Park, PA 16802, U.S.A. 4 Present address: Hudson River Sloop Clearwater, Inc., Poughkeepsie, NY 12601, U.S.A. 5 Smithsonian Environmental Research Center, Edgewater, MD 21037, U.S.A. 6 Center for Coastal Resources Management, Virginia Institute of Marine Science, Gloucester Point, VA 23062, U.S.A.  2007 Society for Ecological Restoration International doi: 10.1111/j.1526-100X.2007.00333.x JANUARY 2009 Restoration Ecology Vol. 17, No. 1, pp. 51?59 51 Geological Survey [USGS] 1:24,000-scale topographic maps), many intermittent streams are missed by these maps. The third- to fourth-order streams tend to be more perennial in nature and so are usually mapped as solid blue lines, even on lower resolution maps (e.g., 1:100,000 scale). Although some low-order streams in the United States are protected from alteration by federal, state, and local regulations, headwater reaches in many areas are either not regulated or receive minimal regulatory protection. Although many headwater reaches are in poor condition (NRC 2002), there is growing awareness of their potential to maintain and improve water quality. A growing number of government programs are available in the United States for restoring low-order riparian ecosystems, including headwater reaches (NRC 2001; Palmer et al. 2005). Strategies for restoring low-order reaches vary from intensive approaches that include realigning channels, designing floodplains, and planting trees and other vegeta- tion in riparian zones to more extensive approaches that rely principally on natural successional processes. Exam- ples of extensive approaches include acquiring conserva- tion easements to restrict or eliminate timber harvesting or by purchasing land outright. Because extensive approaches are far less expensive (per unit land area) than more intensive approaches, their main attraction is that they could be applied for the same cost over a larger pro- portion of a stream network. Although there are a handful of studies describing the forest composition of floodplain vegetation of low-order perennial streams in eastern North America (Monk 1966; Gemborys & Hodgkins 1971; Glascock & Ware 1979; Parsons & Ware 1982; Rheinhardt et al. 1998), forest com- position reference data are sparse for riparian zones of intermittent streams and their stream channels in this region (Rheinhardt et al. 2000) as are data from succes- sional stands (Phillips 2002). This lack of data may be due to intermittent streams being perceived by some as being less valuable than perennial streams and thus less regu- lated. Further, headwater streams (intermittent and perennial) seldom provide habitat for fisheries, and most have low silvicultural potential when contrasted with bottomland hardwood forests on floodplains of large and intermediate-sized rivers (Hodges 1998; Kellison et al. 1998). However, in regions where cropland or pasture cov- ers a large proportion of the land surface, low-order riparian forests often represent habitat quality and complexity not found elsewhere in the landscape (Brinson & Verhoeven 1999). If protecting and restoring low-order riparian zones that are in early stages of forest succession (extensive restora- tion approach) are to become widespread, there is a need for more information about the composition of developing riparian forests as they succeed toward maturity. We know that riparian forests gain biomass and become more strati- fied with age until they reach maturity and that 90% of the aboveground forest biomass consists of trees and tree- derived detritus (Brinson et al. 2006). However, little is known about changes in canopy composition as biomass accumulates throughout succession. Reference data on differences in composition among sites of various ages (and biomass) could be used to plan lower cost alternatives for riparian restoration that could be applied more widely in the landscape (extensive approaches) and determine where more intensive inter- vention might be required. Therefore, our objective was to (1) obtain reference data on riparian zone forest struc- ture and composition in low-order riparian ecosystems from three physiographic provinces of the eastern United States and (2) use those data as a basis for outlining strate- gies for restoring structure and function to low-order riparian zone forests. To do this, stand basal area (BA) (total cross-sectional area of trees) was used as a surrogate for forest biomass, three-dimensional structure, and suc- cessional status. Study Area and Land Use History All study reaches were located in the Delaware River, Chesapeake Bay, and Albemarle/Pamlico Sound drainage basins (Fig. 1). These basins drain an ecologically div- erse area of eastern North America covering 280,479 km2 across eight states and five physiographic provinces. Sampling was conducted in three of the five physiog- raphic provinces: Coastal Plain, Piedmont, and Ridge & Valley. Land use history has varied widely among the physio- graphic provinces. Agricultural areas in the Coastal Plain have been relatively stable over the past 100?200 years, except for more recent changes related to encroaching urbanization and agricultural intensification. Most of the wetter second- to fourth-order riparian forests have been managed for timber, but at a small scale as wood lots, and have been repeatedly harvested over the past 200 years. In contrast, many drier riparian forests on first- order reaches were ditched long ago to maximize arable land or filled and paved when converting them to urban developments. In the Piedmont, agricultural expansion during the nineteenth century led to widespread soil erosion that carried sediment into streams and floodplains. Many larger floodplains became filled with sediment, but when farming on uplands was abandoned due to loss of fertility from the erosion, forests reclaimed the landscape and widespread erosion ceased. Streams are now downcut- ting through accumulated sediment to historic riverbeds, which has left river floodplains high and dry (Ruhlman & Nutter 1999). Little is known specifically about how lower order streams and channels were affected. How- ever, headward cutting of low-order streams is now com- mon (R. D. Rheinhardt, personal observations). In the Ridge & Valley physiographic province, most forests were cleared for farming by 1800. Farming was eventually abandoned on slopes but has continued in Restoration Targets for Vegetation of Low-Order Riparian Ecosystems 52 Restoration Ecology JANUARY 2009 valleys to the present. Farmland abandoned on slopes reverted to forest, but those forests have been repeatedly cut. As is true for the other provinces, no information is available on the land use history specific to low-order riparian zones. However, many headwater reaches in the Ridge & Valley Province arise in mountainous parts of the landscape now mostly covered by forest. In all three physiographic provinces described above, the major land use changes in the last half-century have been urbanization, suburbanization, and sprawl (Johnson 2001; Torrens 2006). In urban areas, many headwater riparian ecosystems have been converted to underground storm drains or obliterated entirely; those that remain are often highly degraded. The proliferation of impervious surface during urbanization is responsible for major alter- ations to the remaining headwater streams (Paul & Meyer 2001). In many cases, ephemeral overland flows that once fed the upper reaches of intermittent streams during storm events have been amplified by storm flows that originate from impervious surfaces in other watersheds. These urban conversions have caused increased peak flows because impervious surfaces can neither effectively detain rainfall nor contribute to groundwater recharge. Increased peak flows have led to flashier hydrographs, channel inci- sion, headward cutting of channels, and increased sedi- ment loads during storm run-off. Although these alterations often destabilize stream banks, they do not dis- rupt succession beyond the near-channel zone. However, incision and flashy hydrographs often reduce the period of soil saturation in the riparian zone, leading to a change in forest composition to species tolerant of drier conditions. Methods We selected 22 stream networks to represent the range of land cover and topography for the three major physio- graphic provinces of the study region (Wardrop et al. 2005). Within each of these networks, we randomly selected 20 stream reaches for sampling. However, an additional set of 20 random locations was identified for sampling in two of the stream networks. Randomization of sampling locations was conducted using a geographic information system (GIS) algorithm available from the Web page of Environmental Systems Research Institute (ESRI) (Eichenlaub; http://arcscripts.esri.com/details. asp?dbid=10296) and adapted by Brooks et al. (2004). The ESRI algorithm uses Avenue script to place random points along line shapefiles (e.g., USGS?National Hydro- logic Data streams). The points randomly selected marked the center of 100-m-long 3 100-m-wide reaches along which various attributes were measured, including BA of trees (cross-sectional area). Although the target number of sample reaches in each stream network was 20, some networks had fewer due to their small size or to other fac- tors such as difficult or denied access. Data Collection In total, 467 reaches were sampled in three physiographic provinces: eight Coastal Plain stream networks, six Pied- mont stream networks, and eight Ridge & Valley stream networks. The majority of reaches were sampled by a sin- gle two-person field team to help ensure data consistency, but three additional field crews sampled some of the Figure 1. The drainage basins of the Atlantic Slope study area. Sampled stream networks were in three physiographic provinces: Coastal Plain (n ? 8 stream networks), Piedmont (n ? 6 stream networks), and Ridge & Valley (n ? 8 stream networks). Restoration Targets for Vegetation of Low-Order Riparian Ecosystems JANUARY 2009 Restoration Ecology 53 reaches. All crews measured tree BA by species. BA of trees was either calculated from plots (about 10% of reaches) by measuring diameter at breast height (cross- sectional area of stems at 1.5 m above ground) of each tree or using an angle gauge (Bitterlich plotless method). Previ- ous studies have shown that data from stands sampled using fixed plots and plotless methods in eastern North American forests are comparable; plotless methods, however, are more efficient (Grosenbaugh 1952; Shanks 1954; Rice & Penfound 1955; Lindsey et al. 1958; Levy & Walker 1971). Three plots (or Bitterlich points) were placed in each stand to represent the proportion of various cover types present in the riparian zone. If the entire riparian zone was of one cover type, one plot was placed at the center point and one point on each side of the stream, one upstream and one downstream (randomly chosen). All plots and Bitterlich points occurred within a defined sec- tion of reach: Bitterlich points were placed at least 30?35 m apart to prevent overlap but still remained within the 1-ha site being sampled. Data Analysis Relative BA was calculated for each tree species within a stand. Because one study objective was to determine com- positional differences in relation to successional status, total stand BA was partitioned into three classes (roughly equivalent to age, biomass, and three-dimensional struc- ture): BA < 20 m2/ha (early successional), 20  BA < 30 m2/ha (mid successional), and BA  30 m2/ha (late succes- sional). Although BA is positively related with biomass (Brinson et al. 2006), it is not a perfect surrogate for age, successional class, or structure. However, the three catego- ries do provide a rough approximation by which to sort reaches by successional status. In using these successional categories, some dense, early successional stands might have been more mid successional in character and some dense, mid-successional stands may have been more late succes- sional in character, but most stands with BA < 20 m2/ha and BA > 30 m2/ha probably represented early and late- successional stands, respectively. Only low-order (first- to fourth-order) sites with at least two points or plots of the same successional class were used in subsequent data analyses to maintain a reasonable number of Bitterlich tallies for points (or sample area for fixed-area plots). Therefore, compositional data (BA by species) for all stands were represented by two or three points (or fixed-area plots) averaged together. Data on physiographic province, successional class, and latitude were compared with composition data and BA categories using the detrended correspondence analysis (DCA) algo- rithm in PC-ORD (McCune & Mefford 1999). Results Reaches of low-order streams and their riparian forests varied widely in age and degree of human alteration. Of 467 reaches sampled, 13 reaches had no trees. Of the 454 forested or partially forested sites, 225 met our criteria of having at least two of the three points or plots belonging to the same successional class. Six of these sites were fifth order or higher and so were deleted from further analysis. Of the remaining 219 sites, there were 109 Coastal Plain reaches, 48 Piedmont reaches, and 62 Ridge & Valley rea- ches. Ninety-two of these 219 reaches were classified as late successional (BA  30 m2/ha), 45 were mid successional (20  BA < 30 m2/ha), and 82 were early successional (BA < 20 m2/ha). However, only 34 of the 92 reaches were late successional in all three points (or plots). A DCA ordination of the study reaches (Fig. 2), based on species composition, showed a strong relationship with latitude from right to left (cutoff for joint plot arrow based on least squares regression: r2 ? 0.525). Stands could be separated into four distinct physiographic subregions (color coded), defined by the ordination position of late- successional stands: northern Ridge & Valley (located north of Virginia), Piedmont, Coastal Plain south of Dela- ware River estuary, and New Jersey Coastal Plain (located north of Delaware River estuary). Enclosures (solid and dashed) drawn on the ordination diagram delineate sites by these subregions. The enclosures separated Coastal Plain (right) and northern Ridge & Valley stands (left) in the lower half of the ordination. The southern Ridge & Valley sites had no late-successional stands on which to base an enclosure, but the early and late-successional stands, located at the top center of the ordination, sepa- rated from the northern Ridge & Valley stands. In contrast, Piedmont stands, delineated by the dashed enclosure in center of ordination, overlapped both the southern Coastal Plain and the northern Ridge & Valley stands. New Jersey Coastal Plain sites (not delimited by its own enclosure) tended to be more compositionally similar to Piedmont stands than with the other Coastal Plain stands because all its late-successional stands occurred within the enclosure delineated for Piedmont stands. Many early and mid-successional stands occurred out- side the enclosures delineating late-successional stands. Of all early and mid-successional stands, 21% (n ? 12) of southern Coastal plain stands, 50% (n ? 14) of Piedmont stands, and 61% (n ? 14) of northern Ridge & Valley stands occurred outside the enclosures for the subregional type based on late-successional reference stands. Because the enclosures delineate the compositional variation of late-successional stands, the enclosures represent targets for restoration for each of the four identified physio- graphic subregions. Table 1 summarizes species composition based on the mean relative BA for all late-successional stands, arranged by physiographic subregion. In other words, each column represents the mean composition of stands within the three enclosures. There were 109 species in all late-successional stands: 43 species in the southern Coastal Plain stands, 21 species in New Jersey Coastal Plain stands, 30 species in Piedmont stands, and 26 species in northern Ridge & Restoration Targets for Vegetation of Low-Order Riparian Ecosystems 54 Restoration Ecology JANUARY 2009 Valley stands. None of the southern Ridge & Valley stands were late successional, and so it is not certain how closely the early and mid-successional stands might be related com- positionally to Ridge & Valley late-successional stands. Red maple (Acer rubrum L.) and Tulip poplar (Lirio- dendron tulipifera L.) were dominant or codominant canopy species in the riparian zone of many stands in all physiographic subregions. Sweetgum (Liquidambar styra- ciflua L.) codominated southern Coastal Plain stands, Red oak (Quercus rubra L.) codominated Piedmont stands, and Hemlock (Tsuga canadensis (L.) Carr.) codominated northern Ridge & Valley stands. A few species occurred in all physiographic subregions and occasionally codomi- nated stands (Table 1): White oak (Q. alba L.), Green ash (Fraxinus pennsylvanica Marsh.), and Blackgum (Nyssa sylvatica Marsh.). There were three late-successional stands that occurred outside their enclosures (outliers). This meant that these outlier stands were compositionally dissimilar to the other late-successional stands of their subregion. The outliers, marked with a box, belonged to southern Coastal Plain, Piedmont, and northern Ridge & Valley stands (one each). The southern Coastal Plain outlier was dominated (95% BA) by Tulip poplar, which is a common succes- sional but long-lived species in all provinces of the study area. The Piedmont outlier was dominated by Red maple and Loblolly pine (Pinus taeda L.), both of which are com- mon in southern Coastal Plain reaches. The northern Ridge & Valley outlier was dominated (100%) by Red maple, which made it compositionally align more closely with Coastal Plain reaches. Although 13 oak species (Quercus spp. L.) occurred in the study area, most of these species comprised a small pro- portion of BA, except for Northern red oak in the northern Ridge & Valley subregion. Oak species included both upland and wetland species. However, most oak species occurred in the southern Coastal Plain subregion (n ? 9). Discussion Land use in the mid-Atlantic region of eastern North America has varied over time and space, but little is known about land use changes along riparian zones of low-order streams in particular. Although ecological 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 Coastal Plain Ridge & Valley (northern) Piedmont Axis 1 (eigenvalue = 0.52) A x is 2 ( e ig e n v a lu e = 0 .4 4 ) CP: BA>30 m 2 /ha (32) CP: 2030 m 2 /ha (10) NJ: 2030 m 2 /ha (20) PM: 2030 m 2 /ha (30) RVn: 20