Plant Ecology

, Volume 197, Issue 2, pp 155–172

Long-term changes in forest composition and diversity following early logging (1919–1923) and the decline of American chestnut (Castanea dentata)

Authors

    • Forest Service, US Department of AgricultureCoweeta Hydrologic Laboratory, Southern Research Station
  • Wayne T. Swank
    • Forest Service, US Department of AgricultureCoweeta Hydrologic Laboratory, Southern Research Station
Article

DOI: 10.1007/s11258-007-9352-3

Cite this article as:
Elliott, K.J. & Swank, W.T. Plant Ecol (2008) 197: 155. doi:10.1007/s11258-007-9352-3

Abstract

Chestnut blight fungus (Endothia parasitica [Murr.] P.J. And. & H.W. And.)) is a classic example of an invasive species, which severely damaged populations of its host, Castanea dentata, and had widespread and long-term impacts on eastern North American forests. Concurrently, forests were further disturbed by lumbering, which was common across the region from the mid 1800s to the early 1900s. In 1926, local infestations of chestnut blight were reported in the Coweeta Basin, Southern Appalachian Mountains of North Carolina. We used permanent plot inventories of the Basin (first sampled in 1934 and twice afterward in 1969–72 and 1988–93) to describe the distribution of species along a complex environmental gradient. Specifically, we asked: How does vegetation change over approximately 60 years following logging and the demise of C. dentata? Does the association between vegetation and environment determine the pattern of species distributions through time? Which species replaced C. dentata across this complex environmental gradient? We used nonmetric multidimensional scaling ordination and multiresponse permutation procedure for the analyses of the inventory periods. In 1934, C. dentata was the most important species in the Coweeta Basin. It was present in 98% of the plots and contributed 22% of the total density and 36% of the total basal area. Diversity increased significantly over time and was attributed to an increase in evenness of species distribution. The canopy dominant, C. dentata, was replaced by more than one species across the environmental gradient. Importance values of Quercus prinus, Acer rubrum, Cornus florida, Tsuga canadensis, and Oxydendrum arboreum increased by 2–5% across the basin following the decline of C. dentata. Tsuga canadensis increased in abundance and distribution, especially near streams across elevations. Liriodendron tulipifera replaced C. dentata in moist coves, which have low terrain shape and high organic matter content. In contrast, Q. prinus and A. rubrum were ubiquitous, much like C. dentata before the chestnut blight becoming dominant or co-dominant species across all environmental conditions.

Keywords

Quercus prinusQuercus rubraTsuga canadensisDisturbanceDisease pandemicLong-term data

Introduction

In forested ecosystems throughout the world, tree species are disappearing due to introductions and subsequent outbreaks of exotic pests and pathogens, irruptions of native pests, logging, and deliberate removals of single species (Ellison et al. 2005). The loss of tree species from North American forests is resulting in fundamental shifts in population and community dynamics of associated species and immediate and long-term changes in ecosystem dynamics (Ellison et al. 2005). Chestnut blight fungus (Cryphonectria parasitica (Murr.) Barr (= Endothia parasitica [Murr.] P.J. And. & H.W. And.)), a classic example of an invasive species, had widespread and long-term influences on eastern North American forests (Day and Monk 1974; Shugart and West 1977; Ellison et al. 2005). This fungus disease devastated the former oak-chestnut forest association (Braun 1950) by eliminating the dominant or co-dominant Castanea dentata.

Castanea dentata was a dominant species in the upland forest communities of much of eastern North America, until it was virtually eliminated by chestnut blight at the beginning of the 20th century. Chestnut blight, caused by the bark-inhabiting canker fungus, was introduced from Asia in the late 19th century (Milgroom et al. 1996). The blight was first discovered in New York in 1904, spread rapidly (∼37 km yr−1) across the range of chestnut, and within 50 years had converted this stately tree to a rarely flowering understory shrub across ∼3.6 million ha of chestnut forest (Anagnostakis 1987). Areas were further disturbed by lumbering and fire, which frequently followed the blight (Brose et al. 2001).

Reconstructing forest dynamics following chestnut blight is challenging because it requires information on forest structure and composition pre- and post-blight. For most historical invasions, such as the chestnut blight, good baseline data on the distribution and abundance of the species assemblages are rare, and little to no information on basic ecosystem processes is available. Although various studies provided an indication of early regeneration for a variety of sites ranging from New England to Tennessee and North Carolina (Korstian and Stickel 1927; Braun 1940; Keever 1953; Nelson 1955; Woods and Shanks 1957, 1959; Good 1968), most of these studies were completed within the first 20 years after the blight infected a region. Several studies were conducted in other regions of the United States (Korstian and Stickel 1927; Aughanbaugh 1935; Good 1968; Mackey and Sivec 1973) and in the Southern Appalachians (Keever 1953; Nelson 1955; Woods and Shanks 1959; Shugart and West 1977; McCormick and Platt 1980) to examine the replacement of chestnut following the chestnut blight in the 1920s. To our knowledge, records of pre-blight vegetation are available only for the Connecticut plots of Korstian and Stickel (1927). The characteristics of pre-blight vegetation could be reconstructed for permanent plots sampled in Virginia in 1939 (McCormick and Platt 1980) and for plots sampled in Pennsylvania in 1934 (Aughanbaugh 1935). Most studies have been confined to investigations of forests after the elimination of chestnut by reconstruction of the pre-blight forest using existing chestnut stumps as a reference (Woods and Shanks 1959; Good 1968; Karban 1978; Van Lear et al. 2002; Myers et al. 2004) or public land records that recorded trees at boundaries or corners (McEwan et al. 2005).

In 1926, local infestations of chestnut blight were reported in Western North Carolina and North Georgia (Gravatt and Marshall 1926; Keever 1953); and it was first noted in the Coweeta Basin. Soon after establishment of Coweeta Hydrologic Laboratory as an experimental forest administered by USDA Forest Service, vegetation surveys were initiated along permanent transects across the Basin. One vegetation survey occurred in 1934–35 and included measurements of standing, infected and dying C. dentata. This survey provides a basis for reconstructing the forest structure, composition, and diversity prior to the chestnut blight in the Southern Appalachians. Data from later inventories on these same permanent plots afford a unique opportunity to assess what species have replaced C. dentata along an environmental gradient. Our study area, a 1,626 ha basin, is larger and more environmentally varied than those examined previously, and the period covered by our observations (ca. 60 years) is longer than those in other studies.

The biological significance of invasive species can have effects on five levels: (1) individuals (mortality and growth), (2) genetic (including hybridization), (3) population abundance and dynamics, (4) community effects (diversity, trophic structure), and (5) ecosystem processes (primary productivity, nutrient cycling) (Parker et al. 1999). Our goal is to address how chestnut blight fungus impacted individuals, population abundance, and community diversity and composition. We also offer brief comments on how chestnut blight impacted genetic diversity and ecosystem processes. Our objectives were to: (1) examine the change in vegetation over approximately 60 years in a southern mixed deciduous forest following logging and the demise of C. dentata, (2) investigate the association between the vegetation and the abiotic and physical site variables that determine the pattern of species distribution through time, and (3) determine which species replaced C. dentata across this complex environmental gradient.

Methods

Study area

Coweeta Hydrologic Laboratory is a research site of the USDA Forest Service, Southern Research Station. It is located in the Nantahala Mountain Range of Western North Carolina, USA, within the Blue Ridge Physiographic Province, near the southern end of the Appalachian Mountain chain (latitude 35°03′ N, longitude 83°25′ W). The 2,185 ha laboratory consists of two adjacent, east-facing, bowl-shaped basins; Coweeta Basin and Dryman Fork Basin. The Coweeta Basin encompasses 1,626 ha and is drained by Ball Creek and Shope Fork, two fourth-order streams. These streams join to form Coweeta Creek, which flows 7 km east to the Little Tennessee River. The vegetation survey did not include Drymans Fork, a 559 ha basin, since it was not added to the laboratory until 1946. Elevations range from 675 to 1,592 m. Slopes are steep ranging from 30 to over 100%. Soils are deep sandy loams and are underlain by folded schist and gneiss. Two soil orders are found within Coweeta, immature Inceptisols and older developed Ultisols (Thomas 1996). The relief has a major influence on hydrologic, climatic, and vegetation characteristics. Streams flow throughout the year, fed by approximately 2,000 mm of precipitation per year, most of which is rain. Mean annual temperature is 12.6°C and ranges from an average of 11.7°C in winter to 21.6°C in summer. Frequent rain, more than 130 storms distributed throughout the year, sustains high evapotranspiration rates and a humid climate (Swift et al. 1988). Vegetation is southern mixed deciduous forest. The principal overstory species are of the genera Quercus, Carya, and Liriodendron. There are scattered groups of Pinus rigida and Tsuga canadensis. Evergreen shrubs (Rhododendron and Kalmia) combine with Cornus, Robinia, Acer, and Betula to form a dense understory cover (Day et al. 1988).

Disturbance history

The Basin was part of the Cherokee Indian Nation until the Nation was moved in 1837. The Cherokee practiced semiannual burning to improve forage for livestock (Douglass and Hoover 1988). In 1842, Charles Dryman built the first house on the present administrative area. Between 1842 and 1900, European settlers continued the practice of light semiannual burning and grazing. The Forest Service purchased the Basin in 1918, but rights to timber over 38 cm (15 inches) at the stump were reserved for the J.A. Porter Logging Company. Logging began in 1919, and selective but heavy cutting continued until 1923. By 1923, when the Forest Service took over administration of the Coweeta Basin, 8 million board feet of timber had been removed (Douglas and Hoover 1988).

Chestnut blight was first noted in Coweeta in the early 1920s. In the decades between 1970 and 1990, two large-scale, natural disturbances have been documented. Dogwood anthracnose, caused by Discula destructiva Redlin, average incidence of infection of Cornus florida was 87% by 1990 in the Coweeta Basin (Chellemi et al. 1992). A major drought was recorded from 1985–88 in the Coweeta Basin (Swift et al. 1989). Clinton et al. (1993) showed that there was an increase in canopy gap formation due to hardwood tree mortality and significant mortality of P. rigida occurred due to a combination of drought and subsequent attack by the southern pine beetle (Dendroctonus frontalis Zimmerman) (Smith 1991).

Vegetation surveys

Permanent plots were used to study changes in forest vegetation patterns over a 60-year-period at Coweeta; 987 permanent plots were inventoried during 1934–1935 (hereafter 1934), 403 of the original permanent plots were re-inventoried during 1969–1973 (hereafter 1970s), and 297 permanent plots were re-inventoried during 1988–1993 (hereafter 1990s). Only data from the 281 permanent plots that were re-inventoried during all three periods (1934, 1970s, and 1990s) were used in our analyses. These 281 plots were located in reference watersheds, which have been without human disturbance since 1923 (Douglass and Hoover 1988). Permanent 0.08 ha (20 m × 40 m) plots were established along 13 parallel, approximately North–South transects (330°) spanning the Coweeta Basin at 200 m intervals. Chestnut stakes were used to establish plot corners in 1934 and most of these stakes were still standing in the 1970s inventory. Subsequently, plot corners have been referenced with pvc pipe. Trees > 1.37 m height and ≥ 2.5 cm dbh (diameter at 1.37 m height) were tallied by species in 2.5 cm diameter classes in each plot. The 1934 survey included diameter classes for live but blight infected C. dentata (no notes were taken as to the condition of the standing C. dentata trees). Diameter class of cut stumps was recorded for trees harvested during the 1919–1923 logging operations. Cut stumps were not identified to species on the 1934 survey sheets. For all species and inventories, we calculated frequency, density (number of stems per hectare), relative density (number of stems of a given species as a proportion of total number of stems of all species), basal area (m2/ha), relative basal area (basal area of a given species as a proportion of total basal area of all species), and importance value (IV; {relative density + relative basal area} ÷ 2). We also calculated the density and basal area of cut stumps that were measured in the 1934 inventory. Plant species nomenclature follows Kirkman et al. (2007).

Site variables

Site variables used in our analyses included percent slope, elevation, modified azimuth, terrain shape, soil depth, depth of A-horizon, soil clay content, soil organic matter content, mean temperature during the growing season, growing season precipitation, and potential solar radiation. Values of site variables were determined by direct measurements or calculated by digital geographical information system (GIS) mapping methods (Elliott et al. 1999). During the 1970s survey, percent slope, aspect, elevation, and slope position (ridge, upper slope, middle slope, lower slope, or cove) were recorded for each plot. Terrain shape, mean temperature, and potential solar radiation were derived using a GIS (ARC/INFOtm).1 Soils data for individual plots were obtained from a first-order soil survey completed in 1985 by the Natural Resources Conservation Service (Thomas 1996). This soil survey map was overlain onto the permanent vegetation plots map using a GIS (ARC/INFOtm). Maximum values for percent soil organic matter content, percent clay content, soil water content, soil depth, and depth of A-horizon were derived from soil series descriptions and used in our analyses.

Previous research at Coweeta has shown that the above site variables modify the local environment. Plot soil moisture is a function of precipitation (input), terrain shape (water drainage), and soil characteristics (water holding capacity) (Yeakley 1993). Availability of soil nutrients is a complex function of soil characteristics including organic matter content, clay content, soil depth, A-horizon depth, terrain shape (affecting downslope movement of nutrients with water drainage), and soil temperature and moisture (Knoepp and Swank 1997a, b). Temperature is a function of elevation, terrain shape (e.g., ridge, middle slope, or cove bottom), aspect (e.g., south-facing slope or north-facing slope), and percent slope. Potential solar radiation, adjusted for solar declination (latitude), at any plot is influenced by the slope inclination and aspect, shading by adjacent hills, and cloud cover (Swift and Knoerr 1973).

Digital terrain shape was determined by taking a scaled 9-celled zero-sum difference, centric minus the weighted average of perimeter cells (Burrough 1986; McNab 1989), using a GIS (ARC/INFOtm). This yielded an unitless value for each ground position. Calculated values for terrain shape were rescaled linearly, where the range −2 to +2 was rescaled from 0 (highest concavity–deep coves) to 10,000 (highest convexity–narrow, well-defined ridges).

For each plot, mean temperature during the growing season was calculated by means of a temperature prediction model (Bolstad et al. 1998b). Temperatures decreased with elevation at mean rates of 7°C km−1 (maximum temperature) and 3°C km−1 (minimum temperature); growing season temperature range was 16.8–22.0°C. Mean annual precipitation for each plot was estimated on the basis of long-term data and isohyetal maps that provide weighting factors (Swift 1968); mean annual precipitation range was 177–255 cm. Solar radiation on mountain slopes was estimated using an algorithm for calculating potential radiation (Swift and Knoerr 1973; Swift 1976). This algorithm estimates solar radiation within 10% of measured radiation values when slope, aspect, and a variable to adjust for topography are included; potential solar radiation range was 42–131%.

Data analyses

We evaluated species diversity using species richness (S), Shannon-Wiener’s index of diversity (H′), and species evenness (E). Shannon-Wiener’s index incorporates both species richness and the evenness of species abundance (Magurran 2004). H′ and E were calculated at the plot level and averaged for each survey period. We used repeated measures ANOVA (PROC GLM, SAS 1999) to determine significant differences in density, basal area, and diversity (S, H′, and E) among inventory periods.

To examine the relationships among tree species composition and environmental variables, we used separate nonmetric multidimensional scaling (NMS) (Clarke 1993; McCune and Mefford 1999; McCune and Grace 2002) ordination analyses for each inventory period (1934, 1970s, and 1990s). In addition, vegetation data were combined for all three inventory periods into one NMS ordination analysis to examine how sample plots moved across ordination space through time. We used tree species basal area as the measure of abundance in the NMS analyses. With NMS, we explored the vegetation–environment relationships among tree species abundance (basal area) and site characteristics. NMS was conducted using the Sorensen distance measure with 60 runs of real data and 50 runs of randomized data (with a maximum of 200 iterations for each run). A three-dimensional solution was chosen for the final iterative ordination, and the best ending point in the preliminary analysis was used as the starting point in the final run. We report the final stress of the ordination and coefficients of determination (R2) for each ordination axis calculated as a proportion of the variation explained in the reduced matrix relative to that in the original matrix. Ecological community data sets often have final stress values between 10 and 20, and values in the lower half of this range indicate reliable solutions (McCune and Grace 2002).

A secondary matrix of site environmental variables (percent slope, elevation, terrain shape index, modified azimuth, soil organic matter content, soil clay content, soil depth, soil A-horizon depth, air temperature, solar radiation, annual precipitation, and basal area of cut stumps) was used to help interpret the ordination results. We chose to include basal area of cut stumps in our NMS analysis as an indicator of harvesting disturbance level. In the resulting ordination graph, plot points indicate sites and line-vectors indicate environmental variables. The length of each line-vector represents the rate of change in the weighted average as inferred from the biplot, showing how much the species distributions and sites differ along that environmental variable. The most important variables are those with the longest line-vector. Monte Carlo permutations were applied to NMS analyses to determine if the strength of species sorting along the environmental variable was greater than would be expected by chance (P < 0.020) (ter Braak 1986; Crowley 1992). We chose to graph only the environmental variables with an R≥ 0.15.

We used the “proportion of max” feature in PC-ORD and rotated the diagrams using terrain shape in the secondary matrix to create better uniformity among the NMS graphical displays of the three inventory periods. To compare tree species composition among the inventory periods (1934–1935, 1969–1972, and 1988–1993), we used a multiresponse permutation procedure (MRPP) (McCune and Mefford 1999; McCune and Grace 2002). MRPP is a nonparametric multivariate test of differences between a priori groups (Mielke 1991), which provides a T-test statistic and a P-value. A stronger separation between groups is reflected by a more negative T-statistic. We compared Sørensen distance and Euclidean distance (with and without ranking). Since these distances showed similar results and Sørensen distances are less prone to exaggerate the influence of outliers (McCune and Grace 2002), we are reporting on Sørensen distance without ranking. When comparisons across inventory periods were significant, we applied multiple pair-wise comparisons using Sørensen distance. We used PC-ORD version 4 (McCune and Mefford 1999) for the NMS ordination analyses and MRPP group separation tests.

Results

In the 1934 survey, C. dentata was the most important species in the Coweeta Basin. It was present in 98% of the plots and contributed 22% of the total density and 36% of the total basal area (Table 1). The species that ranked highest in importance value in 1934 were C. dentata Q. prinus > A. rubrum > C. florida > Carya spp. (Table 1). In the 1970s survey, species were ranked in the order Q. prinus > A. rubrum > C. florida > Q. coccinea > Carya spp., and C. dentata was ranked eighth. By the 1990s, species were ranked in the order A. rubrum > Q. prinus > C. florida > Q. coccinea > Carya spp., and C. dentata was ranked twelfth. Although C. dentata was present in 61% of the plots, its basal area was less than 1.0% of the total (Table 1). Tsuga canadensis occurred in only 16% of the plots in 1934; by the 1990s inventory it occurred in 50% of the plots and its basal area had increased by more than 400% (Table 1).
Table 1

Percent frequency (n = 281), density (stems ha1), basal area (BA; m2 ha1), and importance value (IV; (relative density + relative basal area) ÷ 2) for 1934–1935, 1969–1972, and 1988–1993 tree inventories of the Coweeta Basin

Species

1934–1935

1969–1972

1988–1993

Freq

Density

BA

IV

Freq

Density

BA

IV

Freq

Density

BA

IV

Castanea dentata

98

686

10.08

29.0

80

152

0.39

4.8

61

66

0.16

2.6

Quercus prinus

89

246

3.43

10.1

89

181

5.48

15.0

89

118

6.32

14.6

Acer rubrum

95

339

1.50

8.1

100

299

2.69

12.9

99

256

4.13

16.1

Quercus coccinea

73

172

2.61

7.4

68

82

3.31

8.3

59

47

2.59

5.9

Carya spp.

88

272

1.60

7.2

85

141

1.77

7.0

80

79

1.74

5.7

Cornus florida

77

234

0.63

4.9

82

284

0.93

9.3

75

192

0.79

8.3

Nyssa sylvatica

91

170

1.03

4.6

88

100

1.12

4.7

82

91

1.29

5.4

Liriodendron tulipifera

47

171

0.69

4.0

54

65

1.51

4.5

54

41

2.08

4.9

Hamamelis virginiana

42

157

0.33

3.1

39

70

0.17

2.2

37

55

0.15

2.3

Quercus velutina

68

58

1.01

2.7

62

30

1.39

3.4

40

14

0.99

2.1

Quercus rubra

57

50

0.88

2.4

62

32

1.85

4.3

55

24

2.43

4.8

Robinia pseudoacacia

82

72

0.57

2.2

64

26

0.70

2.0

58

19

0.68

1.8

Oxydendrum arboreum

79

80

0.45

2.1

79

82

1.08

4.2

81

73

1.50

5.1

Quercus alba

42

54

0.62

2.0

40

28

0.83

2.3

32

15

0.84

1.9

Castanea pumila

44

99

0.22

2.0

23

16

0.03

0.5

16

8

0.02

0.3

Pinus rigida

26

26

0.81

1.9

23

23

0.92

2.3

19

10

0.51

1.2

Betula lenta

32

57

0.36

1.6

46

53

0.70

2.7

48

39

0.97

3.0

Tsuga canadensis

16

13

0.25

0.7

42

66

0.71

3.1

50

71

1.17

4.5

Species nomenclature follows Kirkman et al. (2007). Density and basal area were calculated from trees >1.37 m height and ≥2.5 cm dbh (diameter at 1.37 m height). Only species with an IV of ≥2.0 in any sample period were included in the table; minor species with an IV < 2.0 were Fraxinus americana, Betula alleghaniensis, Magnolia fraseri, Sassafras albidum, Diospyros virginiana, Magnolia acuminata, Prunus serotina, Aesculus octandra, Acer saccharum, Fagus grandifolia, Quercus falcata, Alnus serrulata, Ilex opaca, Pinus strobus, Acer pensylvanicum, Amelanchier arborea, and Tilia americana

Tree density was significantly higher in 1934 than in the other two inventory periods (Table 2). Average basal area of the forest at the time of the first survey was only 28.0 m2 ha−1 (Table 2). By the 1990s, basal area had increased significantly, to 30.1 m2 ha−1. In 1934, cut stumps averaged 7.0 m2 ha−1 and basal area removed from individual plots ranged from 0 to 52 m2 ha−1. Only two plots had no cut stumps, 40% of the plots had ≥5.0 m2 ha−1, and 24% of the plots had ≥10.0 m2 ha−1 of basal area removed. H′basal area was significantly higher in the two later inventories than in 1934, but species richness did not significantly decrease until the 1990s (Table 2).
Table 2

Average density, basal area, species richness (S), Shannon’s Index of diversity based on density (H′ density) and basal area (H′ basal area) for inventory periods in 1934–1935, 1969–1972, and 1988–1993 for the Coweeta Basin

 

1934–1935

1969–1972

1988–1993

Density (stems ha−1)

3121 a (89.4)

1886 b (58.1)

1372 c (40.0)

    Cut stumps (stems ha−1)

209 (8.4)

  

Basal area (m2 ha−1)

27.98 a (0.65)

26.96 a (0.55)

30.69 b (0.68)

    Cut stumps (m2 ha−1)

6.92 (0.61)

  

Richness (# species plot−1)

14.0 a (0.16)

14.0 a (0.17)

13.4 b (0.17)

H′ density

2.068 a (0.015)

2.119 b (0.015)

2.093 ab (0.017)

E density

0.784

0.803

0.807

H′ basal area

1.756 a (0.019)

1.852 b (0.018)

1.843 b (0.017)

E basal area

0.666

0.702

0.710

Average density and basal area of cut stumps in 1934–1935 inventory

Within rows, values followed by different letters are significantly different (P < 0.05) based on repeated measures ANOVA (PROC GLM, SAS 1999). Standard errors are in parentheses. Cut stumps were not identified to species in the 1934 inventory

In 1934, size class distribution of species indicated that the forest was experiencing high recruitment of small stems (2.5–7.5 cm dbh) (Fig. 1). Castanea dentata had a high number of small stems, and it had more stems in all size classes than any other co-occurring species (Fig. 1). After 1934, the number of large size-class stems decreased for C. dentata, with <1.0 stems ha−1 for size classes ≥15.0 cm dbh (Fig. 1), and the number of small stems (those <15 cm dbh) had decreased by 75% in the 1970s and 90% in the 1990s. Acer rubrum had many stems in the small size-classes (<15.0 cm dbh) in all survey periods, but an increasing number of large stems in the 1970s and 1990s (Fig. 1). Liriodendron tulipifera had fewer small size-class stems through time and an increasing number of larger stems by the 1970s (Fig. 1). Tsuga canadensis had more stems in all size-classes in the 1970s and 1990s than in 1934 (Fig. 1), and this demonstrates the recruitment and growth of this species after the demise of C. dentata.
https://static-content.springer.com/image/art%3A10.1007%2Fs11258-007-9352-3/MediaObjects/11258_2007_9352_Fig1_HTML.gif
Fig. 1

Diameter size class distribution of select tree species for three inventory periods in the Coweeta Basin, Western North Carolina: (a) 1934–1935 survey; (b) 1969–1972 survey; and (c) 1988–1993 survey. Each stacked-bar graph has been split and presented on two different scales

The NMS ordination for the 1934 data (Fig. 2a) converged on three axes for the final solution (McCune and Mefford 1999). The final stress for the three-dimensional solution was 15.06. The proportion of variance explained was 19% for Axis 1, 37% for Axis 2, and 30% for Axis 3 (cumulative R= 86.5% for the first three ordination axes) (Table 3). When the secondary matrix of site environmental variables was examined, Axis 2 was positively correlated with terrain shape and annual temperature and negatively correlated with basal area of cut stumps, soil organic matter, and A-horizon depth (Table 3, R ≥ 0.30). Axis 3 was positively correlated with elevation and annual precipitation and negatively correlated with solar radiation (Fig. 2a, Table 3). In the NMS ordination diagram, tree species that were most positively related to Axis 1 were Q. prinus and Q. coccinea (Table 4, tau ≥ 0.30); and species most negatively related were C. dentata and Carya spp. (Fig. 2a, Table 4). Species most positively related to Axis 2 were Q. coccinea, Castanea pumila, P. rigida, and Q. prinus; and species most negatively related were B. lenta, L. tulipifera, Hamamelis virginiana, Fraxinus americana, Q. rubra, A. pensylvanicum, Tsuga canadensis, Cornus florida, A. rubrum, and Tiliaamericana (Fig. 2a, Table 4). Species most positively related to Axis 3 were C. dentata, Q. prinus, and Q. rubra; and species most negatively related were Q. alba and Q. velutina (Fig. 2a, Table 4).
https://static-content.springer.com/image/art%3A10.1007%2Fs11258-007-9352-3/MediaObjects/11258_2007_9352_Fig2_HTML.gif
Fig. 2

Coweeta Basin, Western North Carolina for inventory periods: (a) 1934–1935, (b) 1969–1972, and (c) 1988–1993; nonmetric multidimensional scaling ordination diagram of species scores and environmental variables. Species codes: ACEPEN = Acer pensylvanicum; ACERUB = Acer rubrum; ACESAC = Acer saccharum; AESOCT = Aesculus octandra; AMEARB = Amelanchier arborea; BETALL = Betula alleghaniensis; BETLEN = Betula lenta; CARYA = Carya spp.; CASDEN = Castanea dentata; CASPUM = Castanea pumila; DIOVIR = Diospyros virginiana; FRAAME = Fraxinus americana; FAGGRA = Fagus grandifolia; HAMVIR = Hamamelis virginiana; LIRTUL = Liriodendrontulipifera; MAGACU = Magnolia acuminata; MAGFRA = Magnolia fraseri; NYSSYL = Nyssa sylvatica; OXYARB = Oxydendrum arboreum; PINRIG = Pinus rigida; PRUSER = Prunus serotina; QUEALB = Quercus alba; QUECOC = Quercus coccinea; QUEPRI = Quercus prinus; QUERUB = Quercus rubra; QUEVEL = Quercus velutina; ROBPSE = Robinia pseudoacacia; SASALB = Sassafras albidum; TILAME = Tilia americana; TSUCAN = Tsuga canadensis. Environmental codes: SHAPE = Terrain shape index; OM = soil organic matter content; BA cut = basal area of cut stumps; ELEV = elevation; TEMP = mean annual temperature; and ADEP = soil A-horizon depth. Only environmental variables with an r2 ≥ 0.15 were included in the ordination diagram. The two axes that explained the most variation are shown for each inventory period. Axes are displayed using “proportion of maximum” feature in PC-ORD and rotated using terrain shape in the secondary matrix to create better uniformity among the NMS graphical displays of the three inventory periods

Table 3

Nonmetric multidimensional scaling (NMS) correlations for the 12 site variables with the first three ordination axes of vegetation data from inventories in 1934–1935, 1969–1972, and 1988–1993 in the Coweeta Basin

 

Axis 1

Axis 2

Axis 3

r

tau

r

tau

r

tau

Inventory 1934–1935

Slope

0.126

0.071

0.038

0.016

0.244

0.143

Elevation

0.030

0.008

−0.073

0.024

0.481

0.299

Terrain shape

0.282

0.184

0.470

0.358

−0.016

−0.040

Modified azimuth

0.156

0.099

0.291

0.207

−0.229

−0.171

Soil organic matter

0.046

0.048

0.438

−0.321

0.025

0.067

Soil clay content

−0.084

−0.067

0.268

0.174

−0.231

−0.174

Soil depth

0.030

0.052

−0.200

−0.182

0.022

0.029

Soil A-horizon depth

0.033

0.025

0.409

−0.318

0.033

0.049

Temperature

0.117

0.082

0.301

0.209

−0.263

−0.170

Solar radiation

0.123

0.084

0.289

0.184

0.310

−0.206

Annual precipitation

0.077

0.059

−0.218

−0.103

0.322

0.185

Basal area of cut stumps

−0.107

−0.079

0.470

−0.326

−0.027

−0.045

Coefficients of determination

0.191

0.376

0.297

Cummulative % variance explained

19.1

56.7

86.5

Final stress of 3-dimensional solution

15.063

Inventory 1969–1972

Slope

−0.236

−0.145

0.197

0.119

−0.015

−0.014

Elevation

0.377

−0.272

0.375

0.201

0.143

0.001

Terrain shape

−0.180

−0.095

0.070

0.033

0.460

−0.358

Modified azimuth

0.135

0.091

−0.144

−0.100

0.335

−0.225

Soil organic matter

0.175

0.039

0.146

0.153

0.423

0.312

Soil clay content

0.161

0.139

0.311

−0.253

−0.248

−0.151

Soil depth

0.110

−0.265

0.181

0.162

0.236

0.198

Soil A-horizon depth

0.183

0.093

0.126

0.115

0.383

0.295

Temperature

0.236

0.155

−0.118

−0.060

0.359

−0.234

Solar radiation

0.068

0.035

0.327

−0.214

−0.285

−0.179

Annual precipitation

−0.114

−0.096

0.398

0.240

0.241

0.104

Basal area of cut stumps

0.095

0.041

−0.134

−0.099

0.477

0.299

Coefficients of determination

0.176

0.140

0.502

Cummulative % variance explained

17.6

31.6

81.8

Final stress of 3-dimensional solution

15.494

Inventory 1988–1993

Slope

−0.217

−0.140

−0.013

−0.008

−0.081

−0.045

Elevation

0.441

0.269

−0.024

−0.034

0.040

−0.09

Terrain shape

0.012

0.020

0.373

−0.287

0.375

−0.299

Modified azimuth

0.247

0.160

−0.252

−0.173

−0.186

−0.112

Soil organic matter

−0.163

−0.174

0.236

0.193

0.449

0.250

Soil clay content

0.374

0.293

−0.095

−0.060

−0.177

−0.072

Soil depth

−0.122

−0.111

0.120

0.108

0.272

0.176

Soil A-horizon depth

−0.134

−0.125

0.191

0.171

0.422

0.277

Temperature

0.297

0.180

−0.251

−0.176

−0.169

−0.091

Solar radiation

0.284

0.179

−0.107

−0.070

−0.249

−0.161

Annual precipitation

0.339

−0.216

0.013

0.013

0.253

0.094

Basal area of cut stumps

−0.089

−0.02

0.511

0.330

0.334

0.181

Coefficients of determination

0.121

0.323

0.374

Cummulative % variance explained

12.1

44.4

81.8

Final stress of 3-dimensional solution

15.923

All inventories combined

Slope

−0.118

−0.075

−0.204

−0.135

−0.047

−0.029

Elevation

−0.035

−0.100

0.425

−0.252

−0.012

−0.065

Terrain shape

0.337

−0.238

0.114

0.091

0.375

−0.266

Modified azimuth

−0.150

−0.087

0.282

0.182

−0.210

−0.129

Soil organic matter

0.354

0.207

−0.141

−0.155

0.260

0.172

Soil clay content

−0.124

−0.052

0.317

0.239

−0.080

−0.025

Soil depth

0.198

0.139

−0.077

−0.088

0.127

0.100

Soil A-horizon depth

0.328

0.213

−0.105

−0.101

0.241

0.184

Temperature

−0.149

−0.089

0.386

0.240

−0.216

−0.130

Solar radiation

−0.142

−0.082

0.261

0.165

−0.143

−0.083

Annual precipitation

0.143

0.044

0.304

−0.188

0.046

−0.005

Basal area of cut stumps

0.301

0.191

−0.153

−0.073

0.410

0.235

Coefficients of determination

0.340

0.163

0.279

Cummulative % variance explained

34.0

50.3

78.3

Final stress of 3-dimensional solution

16.564

Pearson’s parametric (r) and Kendall’s non-parametric (tau) correlations with ordination axes, n = 281. Environmental variables with r ≥ 0.30 are highlighted in bold. Monte Carlo Test for stress in real data was P < 0.02 for all axes shown

Table 4

Nonmetric multidimensional scaling (NMS) correlations for the tree species variables with the first three ordination axes for inventories in 1934–1935, 1969–1972, and 1988–1993

Species

1934–1935

1969–1972

1988–1993

Axis 1

Axis 2

Axis 3

Axis 1

Axis 2

Axis 3

Axis 1

Axis 2

Axis 3

Acer pensylvanicum

−0.050

0.399

0.165

−0.026

0.049

0.295

−0.114

0.300

0.192

Acer rubrum

−0.135

0.051

0.032

−0.027

0.145

0.302

−0.232

0.170

0.075

Acer saccharum

−0.106

−0.173

0.129

−0.111

0.041

0.164

−0.158

0.235

0.194

Aesculus octandra

0.120

−0.269

0.051

0.036

−0.036

0.276

−0.031

0.258

0.186

Amelanchier arborea

0.081

0.068

0.068

0.030

0.129

−0.243

0.061

0.379

−0.083

Betula alleghaniensis

0.016

−0.247

0.109

0.047

0.196

0.290

−0.094

0.081

0.266

Betula lenta

0.003

0.493

0.009

0.207

0.087

0.487

−0.133

0.346

0.445

Carya spp.

−0.249

−0.177

−0.022

−0.103

0.392

0.204

0.154

0.340

−0.062

Castanea dentata

0.551

−0.133

0.604

−0.200

−0.104

−0.140

0.027

−0.113

0.310

Castanea pumila

0.139

0.552

−0.194

0.034

−0.039

0.350

0.119

0.315

−0.191

Cornus florida

−0.157

−0.269

−0.176

0.011

0.365

0.159

0.230

0.239

−0.094

Fraxinus americana

0.013

0.428

−0.045

0.053

−0.069

0.268

0.001

0.191

0.153

Hamamelis virginiana

0.088

−0.292

−0.021

0.143

0.183

0.189

−0.059

0.009

0.204

Liriodendron tulipifera

−0.142

0.534

−0.101

0.128

−0.202

0.517

0.056

0.497

0.402

Magnolia acuminata

−0.135

−0.171

0.134

−0.026

0.004

0.235

−0.066

0.221

0.162

Magnolia fraseri

−0.014

−0.266

0.033

0.142

0.125

0.194

0.074

0.042

0.189

Nyssa sylvatica

0.164

0.245

−0.086

−0.022

0.112

−0.244

−0.054

−0.263

0.310

Oxydendrum arboreum

−0.002

0.120

−0.090

0.069

−0.024

−0.221

0.132

−0.149

−0.225

Pinus rigida

0.118

0.447

−0.291

0.170

−0.064

0.480

0.265

0.376

−0.153

Prunus serotina

−0.071

−0.240

0.155

−0.059

0.020

0.153

−0.179

0.223

0.080

Quercus alba

−0.090

0.064

0.362

0.368

0.346

−0.095

0.425

−0.021

0.085

Quercus coccinea

0.243

0.647

−0.238

0.156

−0.147

0.650

0.330

0.515

0.315

Quercus prinus

0.258

0.297

0.329

0.638

0.289

−0.236

0.477

−0.098

0.503

Quercus rubra

−0.178

0.438

0.285

−0.219

0.056

0.386

−0.208

0.462

0.169

Quercus velutina

−0.033

0.189

−0.284

0.045

0.457

−0.056

0.258

0.015

−0.263

Tilia americana

0.077

0.314

−0.046

0.060

−0.019

0.308

−0.122

0.266

0.319

Tsuga canadensis

0.063

0.373

−0.131

0.304

−0.000

0.377

−0.025

0.223

0.383

Species not included in table with rank correlations less than 0.20 for any axis and sample period were Fagus grandifolia, Quercus falcata, Robiniapseudoacacia, and Sassafras albidum

Kendall’s non-parametric (tau) correlations with ordination axes, n = 281. Species variables with tau ≥ 0.30 are highlighted in bold

The NMS ordination for the 1970s data (Fig. 2b) converged on three axes for the final solution (McCune and Mefford 1999). The final stress for the three-dimensional solution was 15.49. The proportion of variance explained was 17% for Axis 1, 14% for Axis 2, and 50% for Axis 3 (cumulative R= 81.8% for the first three ordination axes) (Table 3). When the secondary matrix was examined, Axis 1 was negatively correlated with elevation (Table 3, R ≥ 0.30). Axis 2 was positively correlated with annual precipitation and elevation and negatively correlated with solar radiation and soil clay content (Fig. 2b, Table 3). Axis 3 was positively correlated with basal area of cut stumps, soil organic matter, and A-horizon depth and negatively correlated with terrain shape, temperature, and modified azimuth (Fig. 2b, Table 3). In the NMS ordination diagram, tree species that were most positively related to Axis 1 were T. canadensis, B. lenta, and Q. alba (Table 4, tau ≥ 0.30); and species most negatively related were Q. prinus and Q. rubra (Fig. 2b, Table 4). The species most positively related to Axis 2 was Q. prinus; and species most negatively related were Q. velutina, Cornus florida, and Q. alba. The species most positively related to Axis 3 were L.tulipifera, B. lenta, Q. rubra, Tilia americana, Fraxinus americana, and Acer rubrum; and the species most negatively related were Q. coccinea, Pinus rigida, Castanea pumila, and Q. prinus (Fig. 2b, Table 4).

The NMS ordination for the 1990s data (Fig. 2c) converged on three axes for the final solution (McCune and Mefford 1999). The final stress for the three-dimensional solution was 15.92. The proportion of variance explained was 12% for Axis 1, 32% for Axis 2, and 37% for Axis 3 (cumulative R= 81.8% for the first three ordination axes) (Table 3). When the secondary matrix was examined, Axis 1 was positively correlated with soil clay content and negatively correlated with elevation and precipitation (Table 3, R ≥ 0.30). Axis 2 was positively correlated with basal area of cut stumps and negatively correlated with terrain shape. Axis 3 was positively correlated with soil organic matter, A-horizon depth, basal area of cut stumps, and negatively correlated with terrain shape (Fig. 2c, Table 3). In the NMS ordination diagram, the tree species most positively related to Axis 1 were Q. alba, P. rigida, and Q. coccinea; and the species most negatively related to that axis was Q. prinus (Fig. 2c, Table 4). The species most positively related to Axis 2 were Carya spp., L. tulipifera, and Q. rubra; and the species most negatively related were Amelanchier arborea, P. rigida, Q. coccinea (Fig. 2c, Table 4). Species that were most positively related to Axis 3 were B. lenta, Tsuga canadensis, Tilia americana, L. tulipifera, and B. alleghaniensis; and species most negatively related were Q. prinus, Q. coccinea, Oxydendrum aboreum, and Nyssa sylvatica (Fig. 2c, Table 4).

In the combined NMS analysis (all three inventory periods) (Fig. 3), the final stress for the three-dimensional solution was 16.56. The proportion of variance explained was 34% for Axis 1, 16% for Axis 2, and 28% for Axis 3 (cumulative R= 78.3% for the first three ordination axes) (Table 3). In the combined NMS analysis, plots moved across ordination space from the upper left coordinate in the ordination diagram in 1934 (Fig. 3) toward the lower right coordinate for the 1970s and 1990s inventory periods (Fig. 3). A clear separation emerged between 1934 plots and those from the 1970s and 1990s, while the 1970s and 1990s plots overlapped. The site variables most positively correlated to Axis 1 were soil organic matter, A-horizon depth, and basal area of cut stumps and the variable most negatively correlated was terrain shape (Table 3, R ≥ 0.30). Axis 2 was positively correlated with temperature and soil clay content; and negatively correlated with elevation and annual precipitation. Axis 3 was positively correlated with basal area of cut stumps; and negatively correlated with terrain shape (Table 3, R ≥ 0.30).
https://static-content.springer.com/image/art%3A10.1007%2Fs11258-007-9352-3/MediaObjects/11258_2007_9352_Fig3_HTML.gif
Fig. 3

Coweeta Basin, western North Carolina; nonmetric multidimensional scaling ordination diagram of plot scores using combined vegetation data for inventory periods 1934–1935, 1969–1972, and 1988–1993. Axis 1 and Axis 3 are displayed because they explain more of the variation in the NMS results than Axis 2 (see Table 4)

Tree species composition differed significantly among the three survey periods (MRPP; T = −167.57, A = 0.077, P < 0.0001). In addition, average distance increased from 1934 to the 1970s indicating greater within-group heterogeneity through time. Based on Sørensen distance measures, average within-group distance was 0.592 for 1934, 0.652 for the 1970s, and 0.644 for the 1990s. Multiple pair-wise comparisons showed significant differences between 1934 and the 1970s (T = −151.16, A = 0.077, P < 0.0001) and the 1990s (T = −184.51, A = 0.094, P < 0.0001); and a significant, although smaller difference between the 1970s and the 1990s (T = −10.20, A = 0.050, P < 0.0001).

Discussion

The loss of native species due to introductions of exotic pests and pathogens can alter long-term community and ecosystem dynamics (Ellison et al. 2005). The effects of large scale logging and the chestnut blight pandemic that co-occurred across eastern North America in the early 20th century are difficult to separate; however, it is important to examine the combined influences of these disturbance events on forest dynamics. In this study, we used long-term permanent plot data and multivariate analyses to evaluate the changes in species-environment relationships as affected by the loss of a canopy dominant, Castanea dentata, across a complex spatial and temporal gradient. In a previous study, Elliott et al. (1999) described seven vegetation community types that occur across elevation and topographic positions within the Coweeta Basin. Transitions between the different community types were considered gradual (Day and Monk 1974; Elliott et al. 1999) and thought to reflect gradients in species distribution (sensu Whittaker 1956) rather than distinct communities (Bolstad et al. 1998a; Elliott et al. 1999). We hypothesized that tree species distribution patterns overtime are attributed to the combined effects of environment and major disturbances of harvesting and the chestnut blight pandemic.

Harvesting of large trees from the late 1800s to the early 1900s was common across the southern Appalachian region (Brose et al. 2001). Castaneadentata was harvested preferentially because it yielded valuable timber and was a good source of tannic acid (Frothingham 1924; Steer 1948). Douglass and Hoover (1988) estimated that more than 8 million board feet of timber were removed from the Coweeta Basin from 1919 to 1923. On the basis of cut stump data from the 1934 survey, we estimated that an average of 20% of the basal area was removed from the Basin. Thus, in 1934, the Coweeta Basin forest was in a recently disturbed condition and was recovering from logging. Many small stems were present at that time, and this suggests that the forest was then in the initiation stage of stand development (Oliver and Larson 1996). Average basal area in the Basin at the time of the first survey was nearly as low as that in a relatively young (17-year-old clearcut) forested watershed in the Coweeta Basin (Elliott et al. 1997). In 1923, the US Forest Service took over administration of the recently logged area and no further logging was done except in the experimental watersheds (not included in these data). By the time of the 1934 survey, the forest had only 10 years to recover from logging before the chestnut blight induced mortality of virtually all remaining chestnut trees. Unfortunately, no species-specific information is available about the trees removed during the 1919–1923 harvesting period as cut stumps were not identified to species on the 1934 survey sheets. Since C. dentata and Quercus species were the preferred commercial trees (Ashe 1911), we assumed that these species were selectively cut from the forest. From 1934 to the 1970s, density decreased by 40% and basal area by 3.6% as the forest matured and natural thinning proceeded during the stem exclusion stage of stand development (Oliver and Larson 1996). Between the 1970s and 1990s, density decreased by another 27% and basal area increased by 14%. Much of this increased basal area growth was attributed to two species, Q. prinus and A. rubrum.

In 1934, C. dentata was the dominant or co-dominant species (high frequency, density and basal area) across the environmental gradient; however, it was the most abundant at higher elevation (i.e., based NMS ordination for 1934). Many small C. dentata stems were present in 1934 indicating successful recruitment following the earlier harvest and it also had more large stems than other species at that time. Based on the diameter distribution that remained after the harvesting between 1919 and 1923, we propose that C. dentata would have continued its dominant position in the forest in the absence of further disturbance. However, the chestnut blight pandemic dramatically altered this successional trajectory. By the 1970s, only sprouting stems of C. dentata remained in the forest due to chestnut blight induced mortality. Another less dramatic, yet significant, decline in C. dentata density occurred between the 1970s and 1990s. Castanea dentata sprouts remain a major component of some forests through a cycle of sprouting, blight infection, and dieback, which has persisted for decades (Stephenson 1986; Paillet 1988; Stephensen et al. 1991; Griffin 1989, 1992; Parker et al. 1993; Schwadron 1995).

In our study, the decline in density between the 1970s and 1990s suggests that C. dentata sprouting ability has decreased. Parker et al. (1993) attributed an observed decline in the number of sprouting C. dentata stems in southwestern Virginia to the interaction of continuing chestnut blight infection and drought between 1983 and 1988. The same phenomena are likely responsible for the decline in C. dentata sprouts in the Coweeta Basin. As the sprouting stems of C. dentata continued to decline from the 1970s to 1990s, co-occurring species became more abundant, replacing the formerly dominant overstory tree.

The NMS ordination results allowed us to interpret where species were located across the environmental gradient. Some species replaced C. dentata in specific environmental conditions, while others responded to the loss of C. dentata across environmental conditions. Tsuga canadensis increased in abundance (density and basal area) and distribution (frequency) between 1934 and the 1970s. Its increase was most notable near streams (low terrain) at low- to mid-elevations. Since T. canadensis is a shade tolerant species (Burns and Honkala 1990), it may have been present as saplings <2.5 cm dbh at the time of the 1934 inventory and not recorded, or it may have recruited into new areas by seed dispersal. However, observations at the time of the 1970s inventory (W.T. Swank, personal observation) point to recruitment by seed dispersal as the explanation of the increase in frequency and abundance of T. canadensis.

Early successional, disturbance adapted species, such as L. tulipifera, and R. pseudoacacia (Phillips and Shure 1990; Elliott et al. 1998), had high density and high numbers of small stems in the 1934 survey and L. tulipifera was associated with a high level of cutting (i.e., BA cut in the NMS ordination). They both declined in numbers in the later surveys. However, Liriodendron tulipifera increased in size and basal area over time, whereas R. pseudoacacia declined overall by the 1990s. Liriodendron tulipifera occurred primarily in moist coves, which are characterized by low terrain shape and high organic matter content, at low- to mid-elevations (i.e., high association with axes 2 and 3 in the NMS ordination). In contrast, Q. prinus and A. rubrum were ubiquitous (high frequency and abundance), much like C. dentata before the chestnut blight, becoming dominant species across all environmental conditions.

Quercus species responded differently overtime to canopy gaps creation. Between 1934 and 1970s, all of the Quercus species (prinus, coccinea, velutina, rubra, and alba) increased in importance value and basal area primarily due to the canopy openings created by mortality of C. dentata. Quercus species most likely filled in gaps where they were already present, since their frequency did not increase between the first and second inventories. Quercus rubra increased in basal area in coves and mesic slope positions (low terrain shape), with high soil organic matter content. Between the 1970s an 1990s, Q. prinus and Q. rubra continued to increase in basal area and remained about the same in importance value, however, Q. rubra declined in frequency. In contrast, Q.coccinea and Q. velutina decreased in all estimates of abundance (i.e., frequency, basal area, and importance value). The species-specific responses to the drought that occurred from 1985 to 1988 may partially explain differences among Quercus in long-term changes. For example, Clinton et al. (1993) found that 83% of the canopy gaps in the Coweeta Basin were formed by Quercus during this drought and the most frequent gap-forming species were Q. coccinea (44%), Q. rubra (16%), Q. velutina (11%), Carya spp. (11%), and Q. prinus (7%). In our study, Q. coccinea was found on upper slopes and dry ridges (high terrain shape) where drought effects would be most extreme.

We attributed the significant increase in diversity (H′ basal area) from 1934 through the 1990s to the increase in evenness (E basal area) of species distribution, since species richness decreased and average basal area increased. The canopy dominant, C. dentata was replaced by more than one species across the environmental gradient. Quercus prinus, A. rubrum, C. florida, T. canadensis, and Oxydendrum arboreum increased by 2–5% in importance value across the basin following the decline of C. dentata, and many other species increased by 1.0%.

We found that average distance in the NMS ordination was significantly (i.e., results from MRPP separation test) greater for plots in the 1970s and 1990s than for plots in 1934, which suggests that species diversity was greater after C. dentata decline across the landscape (Palmer 1993). It is not surprising that species diversity increased after the virtual loss of C. dentata as this species was overwhelmingly dominant in 1934. Thus, the increase in diversity is attributable to a change in evenness of species distribution rather than an increase in species richness. The harvesting in the early 1900s coupled with the decline of C. dentata enhanced the general landscape pattern of diversity (i.e., NMS ordination, plots in 1934 vs. the 1970s and 1990s).

In addition to chestnut blight, other diseases and insects have been significant factors in the dynamics of vegetation in the Southern Appalachians. Studies have shown that the regional climate is extremely conducive to disease development and the incidence of disease and mortality in the Coweeta Basin is high (Chellemi et al. 1992; Britton 1993). For example, we recorded a decline in C. florida between the 1970s and 1990s. This decline of C. florida was strongly influenced by dogwood anthracnose caused by Disculadestructiva Redlin, a serious disease in Southern Appalachian forests since about 1985 (Anderson 1991; Redlin 1991). Mortality of R. pseudoacacia was attributed to the locust borer (Megacyllene robiniae), which commonly infests young R. pseudoacacia in the Southern Appalachians (Boring and Swank 1984). However, none of these examples was as widespread and catastrophic as the functional loss of C. dentata due to the chestnut blight pandemic.

The invasion of the chestnut blight fungus did have a biologically significant impact on the Eastern deciduous forests as a whole (sensu Parker et al. 1999). The mortality and growth of the individual species, C. dentata, were affected dramatically as is shown by our data and that of others. Population abundance and dynamics of the species have shifted; our study shows that C. dentata abundance and distribution have been reduced substantially. The growth form has changed from a dominant tree to a small sprouting sapling, and the number of sapling sprouts has declined over time. The structure and diversity of forests have been altered through replacement of C. dentata by species with varying physiological and morphological characteristics. Tsuga canadensis, previously unimportant in the canopy, has emerged and become increasingly important, particularly near stream channels at low- to high-elevations. Liriodendron tulipifera has recruited into moist coves at low-elevations and co-occurs with the ubiquitous Q. prinus and A. rubrum. Quercus prinus and A. rubrum, co-dominant species before the blight, tended to fill in the available space across all environmental conditions. The phenotypic and genotypic characters of C. dentata have undoubtedly changed. For decades, C. dentata has succumbed to the fungus before it could flower or produce seed; thus, much of the potential genetic composition of the species has been lost or stagnant. In addition, the preponderance of the research on C. dentata has been directed to the development of hybrids for the reintroduction of the species (Forest et al. 1990; Hebard 2004; Jacobs 2007), and development and reintroduction of hybrids would obviously alter the genetic structure of the species. Ecosystem processes have likely changed with the demise of C. dentata and subsequent replacement of that species by others with different growth rates, litter qualities, and decomposition and nutrient cycling rates (Smock and MacGregor 1988; Rhoades 2007).

Footnotes
1

The use of trade or firm names in this publication is for reader information and does not imply endorsement by the US Department of Agriculture of any product or service.

 

Acknowledgements

This research was primarily supported by USDA Forest Service and partially supported through grants from the National Science Foundation, Division of Environmental Biology to the University of Georgia for the LTER program. Special acknowledgement is made to the Lake City Forest Ranger School, (now Lake City Community College), Florida who provided students during the summer months of the years 1969–1972. As part of their on-the-job training, the students re-inventoried permanent plots established in 1934. Ranger School faculty members Tom Centner and Tom Hirt provided field leadership and Walt Knudsen and Don Lewis provided institutional support. We also acknowledge the valuable contributions of James Buchanan and Mary Lou Rollins, Forest Service staff, who provided field and logistical assistance throughout the conduct of this cooperative project. We thank Alan White, Jack Webster, Ryan McEwan, and James Vose for providing valuable comments on a previous version of the manuscript.

Copyright information

© Springer Science+Business Media B.V. 2007