WETLANDS, Vol. 24, No. 2, June 2004, pp. 364±374 ᭧ 2004, The Society of Wetland Scientists

COLONIZATION AND ESTABLISHMENT OF RED (ACER RUBRUM)INA SOUTHERN APPALACHIAN WETLAND

Robert J. Warren II1,3, Irene M. Rossell2, and Kevin K. Moorhead2 1 Department of Biology Western Carolina University Cullowhee, North Carolina, USA 28723

2 Environmental Studies Department University of North Carolina at Asheville Asheville, North Carolina, USA 28804

3 Present Address: Institute of Ecology University of Georgia Athens, Georgia, USA 30602 E-mail: [email protected]

Abstract: We characterized the post-disturbance recruitment window for red maple (Acer rubrum)ina southern Appalachian wetland using size-class distributions and forest stand models. The DBH and core age of understory and overstory were measured in 108 plots in forested (closed) and unforested (open) fen and ¯oodplain sites at the Tulula Creek wetland complex (a southern Appalachian wetland in Graham County, North Carolina) in 1994 and 2001 as part of a larger ecological study. In addition, the heights of red maple seedlings were measured in 379 quadrats in an unforested ¯oodplain in 1996 and 2001. We examined the temporal patterning of wetland recruitment using red maple size-class data in order to determine (1) the recruitment window for seedling colonization and (2) temporal recruitment patterning based on the size/age structure of established stands. Diameter and height distributions were compared with power function, negative exponential and quadratic models in order to determine goodness of ®t using the coef®cient of determination (R2). Diameter distributions and stand models showed that recruitment continued (at a diminishing rate) at sites that were last cleared 7, 14, and approximately 30 years earlier and ceased at a fourth site cleared approximately 45 years earlier. While there were minor recruitment ¯uctuations that possibly coincided with water-level changes, the unimodal size class distributions indicated that recruitment did not pulse subsequent to initial canopy disturbance. These results show that red maple readily colonized wetland habitats and that the recruitment window lasts at least twice as long as that reported in terrestrial systems. In addition, size-class distribution and regression analysis indicate that the colonization window is directly impacted by canopy disturbance and only indirectly in¯uenced by water levels.

Key Words: red maple, recruitment window, southern Appalachians, fen, ¯oodplain

INTRODUCTION Rossell 1998) as the in¯ux may threaten the survival Red maple (Acer rubrum L.) is a ubiquitous member of rare and endemic herbaceous helophytes that, in of eastern temperate forests, and it often dominates some cases, exist only in small montane wetlands (80±90% of canopy cover and basal area) lowland and (Murdock 1994, Moorhead and Rossell 1998, Warren headwater wetlands in the eastern United States 2002). (Mitsch and Gosselink 2000), including isolated fens Red maple is a fast-growing tree that thrives in a and in the Southern Appalachian Mountain Re- wide variety of habitats from sun to shade, wet to dry, gion (Gaddy 1981, Schafale and Weakley 1990, and acidic to neutral (Will et al. 1995, Elliott et al. Weakley and Schafale 1994, Warren 2002). A prolif- 1997, Abrams 1998). It can reach sexual maturity in eration of woody species in southern Appalachian wet- 4±10 years and regularly sheds copious amounts of lands has garnered growing attention and concern in winged samaras in early spring not long after ¯ower- recent decades (Dennis 1980, Grafton and Eye 1982, ing (Smith 1972, Cook 1980, Canham and Marks Schafale and Weakley 1990, Richardson and Gibbons 1985, Abrams 1998). The germination requirements 1993, Weakley and Schafale 1994, Moorhead and for red maple seeds are not stringent (Abrams 1998,

364 Warren et al., RED MAPLE COLONIZATION AND ESTABLISHMENT 365

Peterson and Pickett 2000), although there may be conditions. Rather than steady recruitment with peri- some preference for mineral (Clinton and Vose odic recruitment failures, we predicted that recruitment 1996). However, the seeds are sensitive to the ratio of would occur in periodic pulses amongst recurrent fail- red/far red light, which inhibits germination in the ures. shady conditions beneath mature forest canopies We used three established patterns of forest tree age/ (Smith 1972, Bazzaz and Pickett 1980, Cook 1980, size distribution for comparison (e.g., Hett and Loucks Canham and Marks 1985, Abrams 1998, Peterson and 1971, Johnson and Bell 1975, Leak 1975, Lorimer and Pickett 2000). Krug 1983, Parker and Peet 1984, Williams and John- Due to the sunlight requirements for germination, son 1990, Shimano 2000): (1) stands with a power red maple recruitment often corresponds with distur- function (yϭaxb) pattern of survivorship, indicating bance events (Henry and Swan 1974, Armesto and decreasing mortality with increasing tree size; (2) Pickett 1985, Canham and Marks 1985, Rankin and stands with a negative exponential (yϭabx) pattern of Pickett 1989, Peroni 1994). The perturbation of tree survivorship, indicating constant mortality rates inde- canopy or ground creates favorable germination con- pendent of tree size; and (3) stands with a quadratic ditions, which may initiate increased seedling coloni- (yϭaxϩbxϩb2) form of survivorship, indicating de- zation. Gross (1980) proposed the concept of a ``suc- clining populations with low recruitment. cession window,'' which suggested that factors such as seed colonization success controlled the initiation SITE DESCRIPTION and cessation of seedling recruitment. The window for red maple recruitment remains open about seven years The 83-ha research site is located in the Tulula after disturbance in terrestrial habitats (Rankin and Creek ¯oodplain and valley bottom in Graham Coun- Pickett 1989, Peroni 1994), although additional factors ty, North Carolina, USA (elev. 800 m). This large, such as herbivory, seed predation, belowground com- relatively ¯at ¯oodplain is a complex mosaic of - petition with grasses, and conditions appear to in- ed and open wetlands, including small depressions ¯uence the initiation and duration (Van Auken and where Sphagnum spp. accumulate. The site is bordered Bush 1988, 1989, Hardt and Foreman 1989, Rankin by U.S. 129 and an abandoned railroad grade and is and Pickett 1989, McPherson 1993, Wilson 1993, Per- crossed by two power lines. Although the site is oni 1994). known locally as Tulula , it is a large wetland Red maple trees can develop numerous shallow lat- complex that includes areas of red maple , eral instead of a to help avoid anaerobic weakly minerotrophic fen, and open meadow (Moor- stress in wetlands and create distinct red maple head and Rossell 1998, Warren 2002). The North Car- (Tiner 1991, Will et al. 1995, Mitsch and Gos- olina Natural Heritage Program has classi®ed the site selink 2000). Red maple swamps are believed to orig- as a Swamp Forest-Bog Complex (Schafale and Weak- inate as open wetlands or marshes (Mitsch and Gos- ley 1990), de®ned as poorly drained bottomlands with selink 2000), but red maple seedlings and saplings soils of alluvial origin now removed from regular struggle in inundated conditions (Day 1987, Will et al. ¯ooding. Developers cleared large areas of the fen and 1995). This might restrict the recruitment of seedlings ¯oodplain and channelized the creek during the late to periods of drought (Keddy and Reznick 1982, Ked- 1980s (Rossell et al. 1999, Rossell and Wells 1999, dy 1990, Kirkman and Sharitz 1994, Mitsch and Gos- Moorhead et al. 2001). The N.C. Department of Trans- selink 2000, Tyndall 2001), possibly coinciding with portation purchased the wetland for restoration as a vegetation disturbance (Canham and Marks 1985, wetland mitigation bank in 1994. Abrams 1998, Peterson and Pickett 2000). Quantitative vegetation sampling was conducted in This study examined red maple recruitment in a three areas of the Tulula Creek wetland complex in southern Appalachian wetland using size-class data in 1994 and 2001 as part of a larger ecological study. order to determine (1) the recruitment window for The three areas were a cleared fen (open fen), forested seedling colonization and (2) temporal recruitment pat- fen (closed fen), and cleared ¯oodplain (open ¯ood- terning based on the size/age structure of established plain). A forested ¯oodplain (closed ¯oodplain) site tree stands. We hypothesized that inhospitable wetland was added in 2001. As part of the site preparation for soils would amplify red maple recruitment obstacles the failed development, the open ¯oodplain and open and thus narrow the recruitment window to less than fen were last cleared 7 and 14 years before 2001, re- seven years after initiation. We also hypothesized that spectively. According to site records, the closed fen size class distributions of juvenile and mature red ma- and closed ¯oodplain were timbered approximately 30 ple stands in a southern Appalachian wetland would and 45 years before 2001, respectively. The two fen have uneven or bimodal distributions re¯ecting pulses sites and the open ¯oodplain adjoin each other, while of recruitment during favorable (presumably drought) the closed ¯oodplain is approximately 1000 m down- 366 WETLANDS, Volume 24, No. 2, 2004

stream from the other three. The fen sites receive a adult red maple basal area (m2/ha) across all seven data steady input of ground-water ¯ow and are removed sets using the General Linear Models Procedure of from the stream, while the ¯oodplain sites are drier SAS Institute, Inc. (2001). and immediately border the stream (Rossell et al. 1999, Rossell and Wells 1999, Moorhead et al. 2001). Height Distribution The herbaceous layer of the fen sites was dominated by wetland dependent (obligate and facultative Height data were collected along permanent linear wetland), while the ¯oodplain sites had more of an transects in July±August 1996 and April±May 2001. even mix of wetland and terrestrial plants (Warren The 25 north-south transects were separated by 20-m 2002). intervals and were of various lengths (7±54 m) as de- termined by the open ¯oodplain boundaries. Red ma- ple stems were measured in 379, 0.25-m2 quadrats lo- METHODS cated at 2-m intervals along the transects. Student's t- Diameter Distribution test was used to test for a signi®cant difference (p Յ 0.05) in mean height between years using the TTEST A 0.8-ha grid of 100-m2 plots was established in the Procedure of SAS Institute, Inc. (2001). fen (open and closed portions) in 1994, with approx- imately half the plots located in the open fen and the Age-Diameter Regressions other half within the closed fen. Twenty plots were randomly selected in each area of the fen and inven- In summer 2001, radial growth cores were extracted toried in 1994 and 2001 with a series of nested quad- from 17 red maple trees (Ͼ 10 DBH) in closed fen rats. Overstory trees with a diameter at breast height plots and from 18 trees in closed ¯oodplain plots in (DBH) Ͼ 10 cm were inventoried in 10 ϫ 10 m quad- order to examine age-diameter relationships and clar- rats. Saplings with a 2±10 cm DBH were inventoried ify ambiguous diameter distribution data (Johnson and in a 4 ϫ 4 m quadrat nested within each 100 m2 plot. Bell 1975). Two ®eld collectors examined the cores Seedlings with a DBH Ͻ 2 cm were inventoried in a independently and repeated the ring counting process 1 ϫ 1 m quadrat nested within each 100 m2 plot but if age estimations were not within two years. Linear not overlapping with the 4 ϫ 4 m quadrat. The closed regressions of age-diameter relationships were tested ¯oodplain site was inventoried in 2001 using a 20 ϫ for signi®cance (p Յ 0.05) using the General Linear 100 m grid of forty, 100-m2 plots. Twenty of the plots Models Procedure of SAS Institute, Inc. (2001). were randomly sampled using the same protocol as in the fen. Six 20 ϫ 30 plots were established in the open Survivorship Models ¯oodplain in 1994. Vegetation was sampled using the same protocol as the fen, with the exception that four We used tree diameter as an indirect measure of age 1 ϫ 1 m quadrats were randomly placed in each of because diameter measurements are less invasive and the six plots rather than in a grid. time-consuming than extracting cores from trees. In Density, frequency, mean DBH, and basal area were addition, the accuracy of tree cores is limited by the calculated for all red maple trees in the 1994 and 2001 frequent occurrence of heartrot and discrepancies be- open and closed fen and in the 2001 closed ¯oodplain tween ring structure and true age (Lorimer and Krug data sets. There were not enough red maple stems to 1983), and age distributions derived from tree cores construct a meaningful diameter pro®le for the open may not re¯ect the ecological dynamics of the stand. ¯oodplain data sets. Basal area was calculated for For example, older trees may be smaller than younger seedlings (DBH Ͻ 4 cm) using the midpoint, 2 cm ones in uneven-aged stands (Mohler et al. 1978), while DBH, of the size class. Importance value (I.V.) was fecundity and population growth may depend more on calculated for red maple by adding relative density size than age (Ross et al. 1982). Size distributions can ([stem density for red /total stem density for all be useful both in understanding inherent demographic species] ϫ 100) and relative basal area ([basal area for patterns in trees and discerning recent disturbance red maples/total basal area for all species] ϫ 100) and events (Johnson and Bell 1975, Leak 1975, Lorimer dividing by two for a constant potential value of 100 and Krug 1983, Shimano 2000). There are also limi- for each stand (Rhodes 1995). Student's t-tests were tations to using diameter distributions. For example, used to test for signi®cant differences (p Յ 0.05) in they assume that diameter is an accurate estimator of mean DBH between years using the TTEST Procedure age and that forest replacement processes are constant of SAS Institute, Inc. (2001). Linear regressions were rather than episodic, both of which are not always true used to investigate relationships between red maple (Ross et al. 1982, Williams and Johnson 1990). In seedlings (Ͻ 4 cm DBH/ha), water levels (cm), and even-aged stands, for example, competition allows Warren et al., RED MAPLE COLONIZATION AND ESTABLISHMENT 367

Table 1. Basal area (m2/ha) for red maple seedlings (Ͻ2 cm DBH), saplings (2±10 cm DBH), and trees (Ͼ10 cm DBH). Importance values are in parentheses.

Basal Area (m2/ha) Data Set Seedlings Saplings Trees Total 1994 open ¯oodplain 0.003 0 0 0.003 2001 open ¯oodplain 0.006 0.01 (42.5) 0 0.02 1994 open fen 0.002 0.4 (95.8) 0 0.4 2001 open fen 0.003 0.4 (86.2) 0.2 (100) 0.6 1994 closed fen 0.001 0.6 (85.8) 12.5 (98.4) 13.1 2001 closed fen 0.0004 0.7 (82.4) 18.1 (96.5) 18.8 2001 closed ¯oodplain 0.0004 0 23.8 (82.3) 23.8 some trees to achieve greater size than others, causing of 80 cm. River gravel was packed into the space be- size differences that give the impression of age differ- tween pipe and hole to keep soil from entering the ences (Mohler et al. 1978). This can be clari®ed with pipe. The surface around each well was sealed with a sampling of radial growth cores (Johnson and Bell clay to keep rainwater from running down the sides of 1975). the pipe, minimizing false readings. Water-table mea- Distribution interpretations most often are made in surements were collected weekly from each well from relation to theoretical frequency distributions, which June 1994 to November 2001. Precipitation measure- are assumed to represent survivorship in a cohort of ments also were collected weekly from a rain gauge trees. These concepts are particularly strong when ap- located near the fen and open ¯oodplain during the plied to a single, shade-tolerant canopy species such same period. Water-level and precipitation data were as red maple that is not subject to heavy thinning mor- averaged into two discrete periods, June±September tality (Johnson and Bell 1975, Lorimer and Krug 1994 and June±September 2001, for analysis and com- 1983). DBH-class distribution in a forest where trees parison. regenerate continuously should have an inverse-J shape (Shimano 2000). An exponential function best ®ts the inverse-J when the death rate is constant, while RESULTS a power function provides a better ®t when the death Tulula Wetland Complex rate changes with age (Hett 1971, Hett and Loucks 1976, Ross et al. 1982, Shimano 2000). A quadratic Red maple is the dominant canopy tree in the closed function provides the best ®t when the size class forms sites of the Tulula Creek wetland complex (I.V. ϭ 82± a ``u'' shape, either concave up or down. 98) and a dominant component of the sapling-layer of In order to determine the best size-distribution mod- both open and closed portions of the fen (I.V. ϭ 82± el, regression analysis of red maple stem frequency- 96) (Table 1). The total basal area of red maple trees size class relationships were tested for signi®cance (p in the closed fen and ¯oodplain (13±23 m2/ha) was Յ 0.05) using the General Linear Models Procedure signi®cantly greater (t ϭ 2.92, p Ͻ 0.014) than that in of SAS Institute, Inc. (2001). The best model was se- the open fen and ¯oodplain (0.003±0.6 m2/ha). lected from three alternatives: (1) Power Function: log Water levels dropped considerably in the open (density ϩ 1) ϭ b(log(size)) ϩ k; (2) Exponential: log (Ϫ5.5 to Ϫ31.9 cm from the surface) and closed (density ϩ 1) ϭ b(size) ϩ k; and (3) Quadratic: density (Ϫ15.5 to Ϫ55.4 cm) portions of the fen and in the ϭ b(size) ϩ c(size2) ϩ k. The adjusted coef®cient of open (Ϫ46.1 to Ϫ65.8 cm) and closed (Ϫ60.7 to determination (R2) was used to determine goodness of Ϫ72.6) ¯oodplains between the growing seasons ®t. (June±September) of 1994 and 2001. Not surprisingly, the average monthly precipitation also dropped at the Tulula wetland complex between the periods June± Hydrologic Survey September 1994 (18 cm) and June±September 2001 In May 1994, shallow water-table wells were in- (9.5 cm). stalled in the open and closed fens (6), in the adjacent open ¯oodplain (6), and in the closed ¯oodplain (4). Open Floodplain Wells were constructed from polyvinyl chloride (PVC) pipe and installed at a depth of 84 cm. Each well had There were numerous (Ͼ 500/ha; 100% of all stems) slits cut at 2-cm intervals from the bottom to a height seedlings and saplings (Ͻ 4 cm DBH) in the open 368 WETLANDS, Volume 24, No. 2, 2004

Figure 1. Structural pro®le of red maple height classes in the open ¯oodplain at the Tulula wetland complex. Plotted values are projected stems per hectare Ϯ SE.

¯oodplain the same year it last was cleared (1994), The diameter distribution was dominated by the ®rst and the recruitment of red maple stems more than dou- size class (DBH Ͻ 4 cm) in 1994 (94% of the stems) bled by 2001 (Ͼ 1,100/ha; 91% of all stems). The and 2001 (83% of the stems). However, larger stems dearth of size classes Ͼ 4 cm DBH in the open ¯ood- (Ͼ 8 and 12 cm DBH) began appearing in 2001. The plain prevented the construction of a meaningful DBH open fen diameter data had an excellent ®t with the pro®le. A height class pro®le based on 1996 and 2001 power function model in both 1994 and 2001 (Table transect data was employed instead to analyze size- 2). class dynamics (Figure 1). Red maple mean stem height more than doubled (20.4±52.7 cm) between Closed Fen 1996 and 2001, a signi®cant (t ϭ 1.65, p Ͻ 0.001) increase. The negative exponential model had a some- Red maple recruitment in the closed fen was less what better goodness of ®t than the power function than half that in the open fen in 2001 and tapered off model to the height-class data, and both were a much between 1994 (Ͼ 275/ha with a DBH Ͻ 4 cm) and better ®t than the quadratic model. However, the qua- 2001 (Ͻ210/ha with a DBH Ͻ 4 cm). Mean stem DBH dratic model had a much better goodness of ®t to the increased (11.1±12.7 cm) between 1994 and 2001 in 2001 data. Both the negative exponential and power the closed fen, but the change was not signi®cant (t ϭ function models also had signi®cant, if not strong, ®ts 1.64, p Ͻ 0.056). The diameter distribution was dom- with the 2001 data (Table 2). inated by the ®rst size class (DBH Ͻ 4 cm) in 1994 (72% of the stems), but that dominance diminished greatly by 2001 (52% of the stems). There was little Open Fen difference in goodness of ®t between the negative ex- Red maple recruitment in the open fen remained ponential and power function models for both years in high (Ͼ 500 stems/ha with a DBH Ͻ 4 cm) in 2001, the closed fen although both were superior to the qua- fourteen years after it was last cleared, yet this con- dratic model (Table 2). stituted only a slight increase from 1994 (Figure 2). According to growth ring estimates, the ages of the

Table 2. Coef®cient of determination (R2) for linear regressions of survivorship models.

Open Floodplain1 Open Fen2 Closed Fen2 Closed Floodplain2 Model 1994 2001 1994 2001 1994 2001 2001 Negative exponential 0.787*** 0.440*** 0.313* 0.532** 0.979*** 0.885** 0.022NS Power function 0.722*** 0.316** 0.698*** 0.854*** 0.798*** 0.856*** 0.039NS Quadratic 0.201* 0.654*** 0.037NS 0.068NS 0.104NS 0.208NS 0.283* *: p Ͻ 0.05, **: p Ͻ 0.01, ***: p Ͻ 0.001, NS: not signi®cant. 1 Height class distribution. 2 Diameter class distribution. Warren et al., RED MAPLE COLONIZATION AND ESTABLISHMENT 369

Figure 2. Structural pro®le of red maple diameter classes in open fen, closed fen, and closed ¯oodplain at the Tulula wetland complex. Plotted values are projected stems per hectare Ϯ SE. (Note: y-axes differ.)

larger trees in the closed fen range from 17 to 59 years Closed Floodplain (mean age ϭ 34 years). There was a signi®cant but not strong correlation (R2 ϭ 0.347, p Ͻ 0.01) between Red maple recruitment essentially ceased by 2001 red maple age and diameter (Figure 3), and the age- in the closed ¯oodplain, which was last cleared about diameter correlation suggests that most closed fen trees 45 years before 2001. The unimodal diameter distri- were 20 to 40 years old. Both the growth ring and age- bution indicated a declining population with little or diameter estimates are consistent with historical re- no recruitment. There were very few stems (Ͻ 10/ha) cords that indicate the closed fen was last cleared ap- with a DBH Ͻ 4 cm in the closed ¯oodplain, and most proximately 30 years before. stems were between 12 and 32 cm DBH (mean stem 370 WETLANDS, Volume 24, No. 2, 2004

Figure 3. Age-diameter regressions for red maple trees in the closed fen and ¯oodplain in 2001. diameter: 23.6) (Figure 2). The unimodal diameter dis- tribution data might be explained by the dissimilar im- tribution of red maple in the closed ¯oodplain only ®t pact of ¯ooding stress on red maple height growth and the quadratic model, although the goodness of ®t was diameter growth (Kozlowski 1984, Will et al. 1995). not exceptionally strong (Table 2). The recruitment dip was unlikely the result of inhibi- According to growth-ring estimates, the ages of the tion due to shading or increased competition as her- larger trees in the closed ¯oodplain ranged from 26 to baceous coverage was unchanged from 1994, and there 59 years (mean age ϭ 45 years). There was a signi®- were no trees and few shrubs in 2001 (Warren 2002). cant correlation (R2 ϭ 0.504, p Ͻ 0.001) between red This suggests that water-induced ¯uctuations might maple age and diameter (Figure 3), and the age-di- have occurred within the larger context of a canopy- ameter correlation suggests that most closed fen trees driven recruitment window. were 40 to 50 years old. Both the growth ring and age- Water level also appeared to impact recruitment in diameter estimates suggest that the closed ¯oodplain the open fen, which was the wettest Tulula site. While was last cleared approximately 45 years before. recruitment remained high between 1994 and 2001, there were substantial losses between the ®rst and sec- ond DBH classes in each year (Figure 2). This sug- DISCUSSION gests a high level of seedling mortality, which may Because timbering and clearing do not initiate a re- have been caused by inundation stress. This also was cruitment starting point with zero woody seedlings, as re¯ected by the excellent ®t of the diameter distribu- with agricultural abandonment, the red maple seed- tion with the power function model, which assumes lings that appeared in the 1994 open ¯oodplain could high early mortality that decreases with age, particu- be pre-disturbance survivors. This makes it dif®cult to larly in shade-tolerant trees (Hett 1971, Hett and establish with certainty the time lag between canopy Loucks 1976, Ross et al. 1982, Shimano 2000). The disturbance and the initiation of recruitment. However, drop between size classes might alternatively be ex- the presence of Ͼ550/ha seedlings in the open ¯ood- plained by slow seedling growth (which might have plain in 1994 when it last was cleared indicates a fairly retarded growth so that seedling-size stems remained rapid commencement. Once initiated, the recruitment in the ®rst class both years), except that mean stem of red maple stems was vigorous, as seedlings more DBH increased (2.8±3.6 cm) signi®cantly (t ϭ 1.65, p than doubled by 2001. This indicates that post-distur- Ͻ 0.001) in the open fen between 1994 and 2001. bance red maple colonization at Tulula likely was rap- While the open and closed fens share the same hy- id. drologic inputs, the closed fen is drier, presumably due While open ¯oodplain recruitment as measured by to evapotranspiration via the larger trees (Warren DBH distribution more than doubled, recruitment as 2002). The lower water table did not appear to bene®t measured by height class shifted from a growing pop- red maple colonization substantially in 2001, however, ulation (as indicated by the negative exponential and as the recruitment of seedlings was less than half that power function models in 1994) to a declining popu- in the open fen. Furthermore, the reduction in closed lation (as indicated by the quadratic model in 2001). fen water levels between 1994 and 2001 also did not The discrepancy between the diameter and height dis- appear to substantially aid red maple recruitment as Warren et al., RED MAPLE COLONIZATION AND ESTABLISHMENT 371

Figure 4. Red maple seedling water level and seedling basal area regressions across all sites and years at the Tulula wetland complex. the number of seedlings decreased. The 2001 drought maple seedlings generally grow rapidly (Walters and (NOAA 2001) may have temporarily muted seedling Yawney 1990, Abrams 1998), but they can be slowed mortality, however. The diameter distribution implies by soil inundation (Day 1987, Will et al. 1995) How- that there were substantial losses between the ®rst and ever, red maple stem growth in both the open fen (rel- second smallest size classes in the 1994 closed fen, atively wet) and open ¯oodplain (relatively dry) was but the losses were only slight in 2001 (Figure 2). The signi®cant between 1994 and 2001, and stems ap- water level dropped during this period, but canopy peared in 2001 that were more than three times as cover did not, suggesting that the drought reduced in- large as any in 1994. This fast growth would be ex- undation stress upon red maple seedlings. However, pected to deplete rather than in¯ate the smallest size the presence of a substantial tree canopy appeared to class, but Ͼ 500 new stems/ha appeared in 2001 at continue dampening overall recruitment. This again in- each site. In contrast, the slow stem growth between dicates that water-induced ¯uctuations might have oc- 1994 and 2001 in the closed fen would be expected to curred within the larger context of a canopy-driven in¯ate the smallest size class rather than deplete it, but recruitment window. recruitment numbers dropped during this period. The strong ®t of the closed fen diameter distribution While wetland tree recruitment often increases with data with both the negative exponential and power decreased water levels, red maple seedling recruitment function models does not reveal a pattern of age-de- remained steady in the open fen and decreased in the pendent mortality, but it does indicate that recruitment closed fen during the 2001 drought. The 2001 drought was ongoing approximately 30 years after it was last did coincide with a doubling of recruitment (as esti- cleared. However, the drop in recruitment between mated by diameter class) in the open ¯oodplain from 1994 and 2001, and the presence of less than half seed- 1994, but the open ¯oodplain was cleared in 1994, lings as those in the open fen, might indicate that the which might better explain the subsequent burst in recruitment window was closing. stem numbers. The closed ¯oodplain was the driest site The closed ¯oodplain was the most heavily wooded in the Tulula wetland complex, and it had almost no and driest of the Tulula sites, and red maple recruit- recruitment in 2001. Regression analysis indicated ment was essentially nonexistent (Figure 2). This sug- that, across sites and years, there was no relationship 2 gests that the mature red maple canopy (approximately (R ϭϪ0.199, p Ͻ 0.9) between water level and seed- 40±50 years of growth) created too much shade for ling recruitment, while there was a negative relation- 2 seedling germination regardless of inundation level. ship (R ϭ 0.596, p Ͻ 0.03) between basal area and seedling recruitment (Figure 4). This indicates that shading from adult trees had a far greater direct impact Direct Impact of Water Level on Recruitment upon seedling recruitment than water levels at Tulula. Interpretations of seedling recruitment based on size-class data can be confounded by slowed growth Indirect Impact of Water Level on Recruitment if older stems linger in the smallest size classes and Flooding stress may have actually aided red maple give a false impression of continued recruitment. Red seedlings by slowing the growth of adult trees, which 372 WETLANDS, Volume 24, No. 2, 2004

would extend favorable light conditions for germina- enough water tables to place lethal stress upon red ma- tion. The age-diameter regressions suggested that red ple seedlings and inhibit recruitment. As wetland in- maple trees grow faster in the drier conditions of the undation increases, there may be a threshold where closed ¯oodplain than in the closed fen (Figure 3). In water level rather than canopy coverage becomes the addition, the closed fen had a strong contingent of driving mechanism in red maple recruitment, but the small- and intermediate-sized trees, which were nearly data suggest that point was not crossed at Tulula. Con- absent from the closed ¯oodplain. Fast growth of in- versely, the water levels in the Tulula Creek wetland termediate-sized trees could have depleted the smaller complex may have enhanced red maple recruitment by classes (Harcombe 1987), perhaps explaining the un- inhibiting grasses and other herbs that compete with imodal diameter distribution in the closed ¯oodplain. woody seedlings and by slowing the closure of cano- Slowed red maple tree growth due to inundation stress py. This appeared to prolong the recruitment window, could explain a Tulula wetland complex recruitment suggesting that intrinsic red maple dynamics such as window 2±4 times as long as that reported for terres- shade from the mature trees played a larger role in trial systems. Further investigation, including red ma- controlling wetland recruitment than extrinsic factors ple seedling demographics and subcanopy PAR levels, such as ¯ooding. is needed to determine the likelihood of this dynamic. Without intervention or disturbance, succession in many southern Appalachian wetlands may culminate in red maple swamps (see Mitsch and Gosselink Pulse Recruitment 2000). The growing dominance of red maples in east- There were small dips and gaps (lack of stems in a ern forests (Abrams 1998), coupled with the size class) in the height and diameter distributions at lack of ecotypic difference between terrestrial and wet- each Tulula site that may have indicated seedling mor- land genotypes (Will et al 1995), suggests a burgeon- tality or slight recruitment pulsing attributable to water ing seed source for wetland colonization. Wetlands levels. These ¯uctuations occurred within the larger with moderate water levels or long dry periods may juggernaut of post-disturbance red maple colonization, not be able to withstand the onslaught, which was rap- however. While diameter distributions may not be sen- id in the Tulula wetland complex. Since many rare and sitive to small recruitment pulses (Ross et al. 1982) endemic southern Appalachian herbaceous wetland and tree age and size were not strongly correlated at species will not persist in the shade beneath a tree Tulula, there were no gaps in the diameter and height canopy (Warren 2002), important conservation issues class distributions large enough to suggest that recruit- are at stake. Further research is needed to examine ment occurred in episodic pulses equivalent to those inundation thresholds for red maple germination, red caused by the initial canopy clearings (Figures 1 and maple productivity in wet versus dry sites, and wetland 2). Episodic pulses of recruitment would have been recruitment dynamics for other woody species. Future re¯ected by bimodal or polymodal size-class distri- research might also explore multiple wetlands or wet- butions (e.g., Johnson and Bell 1975, Ross et al. 1982, land gradients to determine the inundation threshold Williams and Johnson 1990), which were not recorded where water level takes over and drives recruitment at Tulula. However, it might take more than two years dynamics with canopy coverage secondary. worth of data to elucidate such patterns.

ACKNOWLEDGMENTS CONCLUSIONS The authors gratefully acknowledge the contribu- Red maple appears to have a much larger post-dis- tions of Dan Pittillo, Henry Mainwaring, Tom Martin, turbance recruitment window in a southern Appala- Dan Tinker, Loretta Battaglia, Chris Ulrey, and an chian wetland than that reported in terrestrial habitats anonymous reviewer for manuscript suggestions; Da- (Ͼ 7 years), and post disturbance recruitment appears vid Losure and Andy Kilpatrick for ®eld assistance; to be steady, with small ¯uctuations rather than pulsed. and Western Carolina University Graduate School and The recruitment window for red maple colonization North Carolina Department of Transportation and Cen- was initiated by discrete canopy disturbances at four ter for Transportation and Environment for funding. separate sites, and there was little evidence for sub- sequent water-level induced pulsing. While woody vegetation colonization often coincides with droughts LITERATURE CITED in wetlands, this relationship was not apparent in the Tulula Creek wetland complex. Unlike other wetland Abrams, M. D. 1998. The red maple paradox: What explains the widespread expansion of red maple in eastern forests? BioScience types, such as coastal swamps and northern bogs, some 48:355±366. southern Appalachian wetlands may not have high Armesto, J. J. and S. T. A. Pickett. 1985. Experiments on distur- Warren et al., RED MAPLE COLONIZATION AND ESTABLISHMENT 373

bance in old-®eld communities: impact on species richness Mitsch, W. J. and J. G. Gosselink. 2000. Wetlands, third edition. and abundance. Ecology 66:230±240. Van Nostrand Reinhold, New York, NY, USA. Bazzaz, F. A. and S. T. A. Pickett. 1980. Physiological ecology of Mohler, C. I., P. I. Marks, and D. G. Sprugel. 1978. Stand structure tropical succession: a comparative review. Annual Review of Eco- and allometry of trees during self-thinning of pure stands. Journal logical Systematics 11:287±310. of Ecology 66:599±614. Canham, C. D. and P. L. Marks. 1985. The response of woody plants Moorhead, K. K. and I. M. Rossell. 1998. Southern mountain fens. to disturbance: patterns of establishment and growth. p. 201±216. p. 379±403. In M. G. Messina and W. H. Conner (eds.) Southern In S. T. A. Pickett and P. S. White (eds.) The Ecology of Natural Forested Wetlands. Lewis Publishers, Boca Raton, FL, USA. Disturbances and Patch Dynamics. Academic Press, Orlando, FL, Moorhead, K. K., I. M. Rossell, J. W. Petranka, and C. R. Rossell, USA. Jr. 2001. Tulula wetlands mitigation bank. Ecological Restoration Clinton, B. D. and J. M. Vose. 1996. Effects of Rhododendron max- 19:74±80. imum L. on Acer rubrum L. seedling establishment. Castanea 61: Murdock, N. A. 1994. Rare and endangered plants and animals of 38±45. southern Appalachian wetlands. Water, Air and Soil Pollution 77: Cook, R. 1980. The biology of seeds in the soil. p. 107±26. In O. 385±405. T. Solbrig (ed.) Botanical Monographs Vol. 15: Demography and National Oceanic and Atmospheric Administration. 2001. Southeast Evolution in Plant Populations, University of California Press, Region Climate Summary. National Climatic Data Center, Ashe- Berkeley, CA, USA. ville, NC, USA. (http://www.ncdc.noaa.gov). Day, F. P. Jr. 1987. Effects of ¯ooding and nutrient enrichment on Parker, A. J. and R. K. Peet. 1984. Size and age structure of conifer biomass allocation in Ace rubrum seedlings. American Journal of forests. Ecology 65:1685±1689. Botany 74:1541±1554. Peroni, P. A. 1994. Invasion of red maple (Acer rubrum L.) during Dennis, W. M. 1980. Sarracenia oreophila (Kearney) Wherry in the old ®eld succession in the North Carolina Piedmont: age structure Blue Ridge Province of northeastern Georgia. Castanea 45:101± of red maple in young stands. Bulletin of the Torrey Botan- 103. ical Club 121:357±359. Elliott, K. J., L. R. Boring, W. T. Swank, and B. R. Haines. 1997. Peterson, C. J. and S. T. A. Pickett. 2000. Patch type in¯uences on Successional changes in plant species diversity and composition regeneration in a western Pennsylvania, USA, catastrophic wind- after a southern Appalachian watershed. Forest Ecol- throw. Oikos 90:489±500. ogy and Management 92:67±85. Rankin, W. T. and S. T. A. Pickett. 1989. Time of establishment of Gaddy, L. L. 1981. The bogs of the southwestern mountains of red maple (Acer rubrum) in early old®eld succession. Bulletin of North Carolina. Report to N.C. Natural Heritage Program, Ra- the Torrey Botanical Club 116:182±186. leigh, NC, USA. Rhodes, R. 1995. The use of importance values in ecology. Chin- Grafton, W. N. and O. L. Eye. 1982. Vascular ¯ora of eight selected quapin 3:14. West Virginia wetlands with special references to rare species. p. Richardson, C. J. and J. W. Gibbons. 1993. Pocosins, Carolina bays 107±21. In B. R. McDonald (ed.) Proceedings of the Symposium and mountain bogs. p. 257±310. In W. H. Martin, S. G. Boyce, on Wetlands of the Unglaciated Appalachian Region, Morgan- and A. C. Echternacht (eds.) Biodiversity of the Southeastern town, WV, USA. United States. 1. Lowland Terrestrial Communities. John Wiley Gross, K. L. 1980. Colonization by Verbascum thaspus (Mullein) of & Sons, New York, NY, USA. an old ®eld in : experiments of the effects of vegetation. Ross, M. S., T. L. Sharik, and D. W. Smith. 1982. Age-structure Journal of Ecology 68:919±927. Harcombe, P. A. 1987. Tree life tables. BioScience 37:557±368. relationships of tree species in an Appalachian forest in South- Hardt, R. A. and R. T. T. Forman. 1989. Boundary form effects on west Virginia. Bulletin of the Torrey Botanical Club 109:287±298. wood colonization of reclaimed surface mines. Ecology 70:1252± Rossell, C. R., I. M. Rossell, J. W. Petranka, and K. K. Moorhead. 1260. 1999. Characteristics of a partially disturbed southern Appalachian Henry, J. D. and J. M. A. Swan. 1974. Reconstructing forest history forest-gap bog complex. Virginia Museum of Natural History Spe- from live and dead plant materialÐAn approach to the study of cial Publication Number 7:81±89. forest succession in southwest New Hampshire. Ecology 55:772± Rossell, I. M. and C. L. Wells. 1999. The seed banks of a southern 783. Appalachian fen and an adjacent degraded wetland. Wetlands 19: Hett, J. M. 1971. A dynamic analysis of age in sugar maple seed- 365±371. lings. Ecology 52:1071±1074. SAS Institute, Inc. 2001. Version 8.02 for Windows. SAS Institute, Hett, J. M. and O. L. Loucks. 1971. Sugar maple ( Inc. Cary, NC, USA. Marsh.) seedling mortality. Journal of Ecology 59:507±520. Schafale, M. P. and A. S. Weakley. 1990. Classi®cation of the nat- Johnson, F. L. and D. T. Bell. 1975. Size class structure of three ural communities of North Carolina, third approximation. North streamside forests. American Journal of Botany 62:81±85. Carolina Natural Heritage Program, Division of Parks and Rec- Keddy, P. A. 1990. Water level ¯uctuations and wetland conserva- reation, Raleigh, NC, USA. tion. p. 79±91. In J. Kusler and R. Smardon (eds.) Wetlands of Shimano, K. 2000. A power function for forest structure and regen- the Great Lakes. Proceedings of the Wetlands of the Great Lakes eration pattern of pioneer and climax species in patch mosaic for- Symposium, Niagara Falls, NY, USA. ests. Plant Ecology 146:207±220. Keddy, P. A. and A. A. Reznicek. 1982. The role of seed banks in Smith, H. 1972. Light quality and germination: ecological implica- the persistence of 's coastal plain ¯ora. American Journal tions. p. 219±32. In W. Heydecker (ed.) Seed Ecology. Pennsyl- of Botany 69:13±22. vania State University Press, University Park, PA, USA. Kirkman, L. K. and R. R. Sharitz. 1994. Vegetation disturbance and Tiner, R. W. 1991. The concept of a hydrophyte for wetland iden- maintenance of diversity in intermittently ¯ooded Carolina bays ti®cation. BioScience 41:236±245. in South Carolina. Ecological Applications 4:177±188. Tyndall, R. W. 2001. Vegetation change during a thirteen-year pe- Kozlowski, T. T. 1984. Responses of woody plants to ¯ooding. p. riod (1987±1999) in a Carolina bay on the Delmarva Peninsula of 129±163. In T. T. Kozlowski (ed.) Flooding and Plant Growth. Maryland. Castanea 66:245±252. Academic Press, New York, NY, USA. Van Auken, O. W. and J. K. Bush. 1988. Competition between Leak, W. B. 1975. Age distribution in virgin red spruce and northern Schizachyrium scoparium and Prosopis glandulosa. American hardwoods. Ecology 56:1451±1454. Journal of Botany 75:782±789. Lorimer, C. G. and A. G. Krug. 1983. Diameter distributions in even Van Auken, O. W. and J. K. Bush. 1989. Prosopis glandulosa aged stands of shade tolerant and mid tolerant tree species. Amer- growth: in¯uence of nutrients and simulated grazing of Bouteloua ican Midland Naturalist 109:331±345. curtipendula. Ecology 70:512±516. McPherson, G. R. 1993. Effects of herbivory and herb interference Walters, R. S. and H. W. Yawney. 1990. Acer rubrum L. red maple. on oak establishment in a semi-arid temperate savanna. Journal of p. 60±69. In R. M. Burns and B. H. Honkala (eds.) Silvics of Vegetation Science 4:687±692. : Volume 2: Hardwoods. Agriculture Handbook 374 WETLANDS, Volume 24, No. 2, 2004

654. U.S. Department of Agriculture, Forest Service, Washington, and dry-site Acer rubrum (red maple) populations. American Mid- DC, USA. land Naturalist 134:127±140. Warren, R. J. II. 2002. Impact of Woody Vegetation Succession Williams, C. E. and W. C. Johnson. 1990. Age structure and the Upon Herbaceous Communities in a Southern Appalachian Wet- maintenance of Pinus pungens in pine-oak forests of southwestern land. M.S. Thesis. Western Carolina University, Cullowhee, NC, Virginia. American Midland Naturalist 124:130±141. USA. Wilson, S. D. 1993. Belowground competition in forest and prairie. Weakley, A. S. and M. P. Schafale. 1994. Non-alluvial wetlands of Oikos 68:146±150. the Southern Blue Ridge diversity in a threatened ecosystem. Wa- ter, Air and Soil Pollution 77:359±393. Manuscript received 17 January 2003; revisions received 20 June Will, R. E., J. R. Seiler, P. P. Feret, and W. M. Aust. 1995. Effects 2003, 1 December 2003, and 6 February 2004; accepted 1 March of rhizosphere inundation on the growth and physiology of wet 2004.