Ecosystem dynamics in Central Appalachian riparian forests affected by hemlock woolly adelgid

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The State University

By

Katherine Lee Martin

Graduate Program in Environment and Natural Resources

The Ohio State University

2012

Dissertation Committee:

Dr. P. Charles Goebel, Advisor

Dr. John Cardina

Dr. Peter S. Curtis

Dr. Ronald L. Hendrick

Copyrighted by

Katherine Lee Martin

2012

Abstract

In an era of rapid ecological change due to loss of biodiversity, climate change,

and altered nutrient dynamics, predicting ecosystem dynamics and maintaining

ecosystem services provided by complex natural ecosystems is an increasing concern.

Novel disturbances, including those caused by invasive insects, provide an unfortunate

opportunity to test the applicability of ecological theories to practical problems. One such case study is the loss of eastern hemlock (Tsuga canadensis, hemlock) from much of eastern North America due to the invasive pest hemlock woolly adelgid (Adelges tsugae; HWA).

Foundation species such as hemlock are abundant and define the ecosystem processes of a community through a small number of strong interactions. As these interactions are lost, ecosystems are predicted to transition rapidly and develop distinctly different energy exchanges, nutrient cycles, and compositions that define a new community state. Much of the understanding of such transitions is conceptual. To advance the understanding of the function of foundation species and transitions to alternate community states when a foundation species is lost, my dissertation examines vegetation community composition and ecosystem function in 38 hemlock-dominated riparian forests across the central Appalachians. On the unglaciated Allegheny Plateau in

Ohio, uninvaded forests provide baseline data on hemlock as a foundation species.

iii

Across West Virginia and Virginia, sites impacted by HWA for 9-32 years were selected

as a chronosequence to compare changes in compostion and function.

Hemlock forests exhibit low species richness, and thus have low resiliency. In

uninvaded forests of Ohio, hemlock dominates the vegetation, although other species are structured by environmental gradients. Structural equation modeling indicates hemlock has a negative influence on vegetation species richness, light availability and productivity. Thus, a likely future HWA arrival will result in a complete reorganization of these ecosystems, but impacts will differ across environmental gradients. Data from sites impacted by HWA 9-32 years in West Virginia and Virginia indicate all hemlock forests will likely be impacted. Although mortality is initially slowed at higher elevations

and on steeper slopes with northerly aspects, eventually, the duration of HWA invasion is

the most important driver of mortality and ecosystem change. As decline progress,

hemlock remains dominant in sites impacted for decades, although compositions are

shifting and diverging across overstory hemlock decline classes. Some species, including

the native evergreen shrub rhododendron (Rhododendron maximum) and other evergreen

species including red spruce (Picea rubens), may be particularly influential during

community reorganization. Environmental gradients, including elevation and soil

characteristics, are also important ecologial drivers. Among overstory hemlock decline categories, resource availability and nutrient cycling are accelerating, but this varies with environmental context.

This research supports broad patterns in compositional and functional shifts found in other regions, but also highlights the complexity of the loss of hemlock as a foundation

iv species. As a model system, hemlock provides and example where resilience and atlternate state theories apply, but require some expansion for complex systems dominated by long-lived species. This indicates restoration and management will be more efficient when adapted based on the environmental context and component species.

v

For everyone that supported me on this journey, for the resilience of the Appalachians,

and for eastern hemlock.

vi

Acknowledgments

This dissertation research was supported by funding from an Ohio Agricultural Research

and Development Center (OARDC) SEEDS graduate grant as well as The Charles E.

Thorne Memorial Associateship, the OARDC Director’s Associateship, the College of

Food, Agriculture & Environmental Science Fellowship, and National Science

Foundation GK-12 Graduate Fellowship (Grant 0638669). Data collection assistance was provided by Tom Macy, Lara Kobelt, Cody Clifton, Stephen Rist, and Jack Martin. In particular, Tom, Lara, and Jack maintained positive attitudes during unexpected

Appalachian surprises. Advice and assistance with research permits was provided by

John Perez, Dr. Thomas Schuler, Randy Beinlich, Merrick Smith, Wanda San Jule, Kent

Karriker, Terry Slater, Edward Haverlack, Stephen Tanguay, David Glass and Sam

Cowell. Research permits were provided by the United States Department of the Interior

National Park Service (Gauley River National Recreation Area and New River Gorge

National River), the United States Department of Agriculture National Forest Service

(Monongahela National Forest, Fernow Experimental Forest, Washington and Jefferson

National Forest), the West Virginia Department of Natural Resources (Carnifex Ferry

State Battleground Park), and the Nature Conservancy Virginia Chapter (Warm Springs

Preserve). I am very grateful to the consistent support and encouragement of my advisor,

Dr. Charles Goebel who helped me develop and design all phases of this project and

vii

supported my development as an independent scientist. Drafts of my dissertation were

improved by the generous consideration and helpful suggestions of my committee

members, Drs. John Cardina, Peter Curtis, and Ron Hendrick. In the School of

Environment & Natural Resources, Mary Cappocia, Amy Schmidt, Pat Patterson,

Anthony Utz, and Bev Winner were wonderfully helpful. Rocky & Belynda Smiley were

great personal and professional supporters and friends, particularly with MWGL-SER.

This journey was supported by my dear family and friends, particularly my parents Jackie

and Bunn Martin, and numerous wonderful people including, but not limited to: Julia

Barton, Ray Ball, Sarah Becker, Sabrina Grossman, Jessica Fouke, Krista Jacobsen,

Laura Kearns, Kay Kirkman, Jessica Rutledge, Shannan Reichenberg, and Rachel

Schultz.

viii

Vita

June 1997 ...... Hayfield Secondary School

May 2001 ...... B.S. Biology, minor Computer Science,

College of William & Mary

December 2006 ...... M.S. Ecology, University of Georgia

June 2007 to present ...... Graduate Fellow, School of Environment &

Natural Resources, The Ohio State

University

Publications

Martin, K.L., D.M. Hix , and P.C. Goebel. 2011. Coupling of vegetation layers and

environmental influences in a mature, second growth Central Hardwood forest

ecosystem. Forest Ecology and Management 261: 720-729.

Martin, K.L. and P.C. Goebel. 2011. Preparing for hemlock woolly adelgid in Ohio:

communities associated with hemlock-dominated ravines of Ohio’s Unglaciated

Allegheny Plateau. Proceedings, 17th Central Hardwood Forest Conference; 2010

April 5-7; Lexington, KY. USDA Forest Service General Technical Report.

Refereed paper.

Martin, K.L. and P.C. Goebel. 2010. Impacts of hemlock decline on successional

pathways and ecosystem function at multiple spatial scales in forests of the central ix

Appalachians, USA. Pp. 147-152, In: J.C. Azevedo, M. Feliciano, J. Castro and

M.A. Pinto (Editors), Proceedings of the IUFRO Landscape Ecology International

Conference, Sept. 21-27. Bragança, Portugal. Non-referred paper.

Martin, K.L. and L.K. Kirkman. 2009. Management of ecological thresholds to re-

establish disturbance-maintained herbaceous wetlands of the Southeastern USA.

Journal of Applied Ecology 46: 906-914.

Fields of Study

Major Field: Environment and Natural Resources

x

Table of Contents

Abstract ...... iii

Acknowledgments...... vii

Vita ...... ix

List of Tables ...... xiv

List of Figures ...... xvi

Chapter 1: Introduction ...... 1

Chapter 2: Resilience, alternate states, and biodiversity of eastern forests during rapid environmental change ...... 13

2.1 Abstract ...... 13

2.2 Introduction ...... 14

2.3 Resilience and alternate states ...... 16

2.4 The role of diversity in resilience ...... 17

2.5 Resilience in forest ecosystems ...... 18

2.6 Alternate state transitions from oak to maple: mesophication ...... 19

2.7 Reshaping resilience across eastern deciduous forests: EAB ...... 24

xi

2.8 Foundation species ecosystems exhibit low resilience: eastern hemlock ...... 26

2.9 Implications ...... 30

2.10 Acknowledgements ...... 31

2.11 References ...... 32

Chapter 3: The foundation species influence of Tsuga canadensis (eastern hemlock) on biodiversity and ecosystem function on the Unglaciated Allegheny Plateau ...... 47

3.1 Abstract ...... 47

3.2 Introduction ...... 48

3.3 Methods ...... 52

3.4 Results ...... 60

3.5 Discussion ...... 65

3.6 Acknowledgements ...... 70

3.7 References ...... 70

Chapter 4: Decline in riparian Tsuga canadensis forests of the central Appalachian across an Adelges tsugae invasion chronosequence ...... 90

4.1 Abstract ...... 90

4.2 Introduction ...... 91

4.3 Materials and Methods ...... 94

4.4 Results ...... 99

xii

4.5 Discussion ...... 100

4.6 Conclusion ...... 104

4.7 Acknowledgements ...... 105

4.8 Literature Cited ...... 105

Chapter 5: Removal of eastern hemlock by an invasive pest causes community divergence across environmental gradients ...... 119

5.1 Abstract ...... 119

5.2 Introduction ...... 120

5.3 Materials and Methods ...... 125

5.4 Results ...... 133

5.5 Discussion ...... 138

5.6 Acknowledgements ...... 146

5.7 References ...... 146

Chapter 6: Conclusion...... 173

References ...... 179

xiii

List of Tables

Table 3.1 Mean overstory species relative basal area (±SE) in transects at 78

10, 30 and 50 meters from small streams

Table 3.2 Environmental and functional metrics (± SE) in eastern hemlock 79

ravines are linked at distances of 10, 30 and 50 meters from

headwater streams

Table 3.3 Sapling species frequency (stem counts ± SE) in transects at 10, 80

30 and 50 meters from small streams

Table 3.4 Mean stem counts of seedling species (±SE) at three distances 81

from headwater streams

Table 3.5 Mean percent cover (± SE) of ground-flora taxa at 10, 30 and 50 82

m from headwater streams

Table. 4.1 Study sites across the Central Appalachians 112

Table 4.2 Overstory species relative basal area (± SE) at three distances 113

from the stream across 30 streams in the central Appalachians

impacted by Adelges tsugae

Table 4.3 Relative basal area (± SE) of sapling (2.5-10 cm dbh) species in 114

xiv

30 sites impacted by Adelges tsugae in the central Appalachians

Table 5.1 Study sites across the Central Appalachians with mean 156

environmental conditions across transects

Table 5.2 Mean overstory relative basal area (% ±SE) of common species 157

(contributing > 1% mean relative basal area in overstory or sapling

layers) in 30 hemlock-dominated riparian sites in the central

Appalachians

Table 5.3 Mean sapling relative basal area ( % ±SE) of common species 158

(contributing > 1% mean relative basal area in overstory or sapling

layers) in 30 hemlock dominated riparian sites in the central

Appalachians

Table 5.4 Mean seedling counts (density ±SE) of common species 159

(contributing > 1% mean relative basal area in overstory or sapling

layers) in the seedling layer of 30 hemlock dominated riparian sites

in the central Appalachians

xv

List of Figures

Figure 1.1 Conceptual models of alternate state dynamics 3

Figure 1.2 Conceptual model of hypothesized alternate state changes caused by 4

HWA

Figure 1.3 USDA Forest Service Forest Health Protection Hemlock Woolly 5

Adelgid (HWA) Map

Figure 1.4 HWA appears as small white tuffs on the underside of needles 7

Figure 1.5 Hemlock snags in Bath County, Virginia where HWA was detected 8

in 1993

Figure 2.1 Generalized model of alternate state dynamics 43

Figure 2.2 The transition from oak to mesic maple forests 44

Figure 2.3 Resilience is decreasing in eastern deciduous forests due to the loss of 45

ash

Figure 2.4 Hemlock forests are low resilience and the loss of hemlock results in 46

a state change

Figure 3.1 Species richness (S) (± SE) of vegetation 83

Figure 3.2 Shannon diversity (H) (± SE) of vegetation 84

Figure 3.3 Non-metric multidimensional scaling of the overstory with PCA axes 85

xvi

of environmental variables

Figure 3.4 Non-metric multidimensional scaling of the overstory with PCA axes 86

of environmental variables

Figure 3.5 Non-metric multidimensional scaling of the overstory and functional 87

metrics as PCA axes

Figure 3.6 Structural Equation model (SEM) of diversity metrics in Ohio 88

hemlock forests

Figure 3.7 Structural equation model (SEM) of the influence of hemlock 89

dominance on canopy openness and leaf litter biomass

Figure 4.1 Location of study counties across West Virginia and Virginia with 115

the initial year of HWA detection

Figure. 4.2 Health of individual T. canadensis trees by size (dbh) across 30 sites 116

in the central Appalachians

Figure 4.3 Structural equation model (SEM) of factors influencing Tsuga 117

candensis decline in the overstory and sapling layers of thirty sites across

West Virginia and Virginia invaded 10-32 years

Figure 4.4 Structural equation model (SEM) of factors influencing health in the 118

overstory and sapling layers in sites invaded within the last ten years in

West Virginia

Figure 5.1 Generalized conceptual model of alternate state shifts 159

Figure 5.2 Conceptual model of alternate state shifts where changes that surpass 160

threshold result in community divergence, the development of multiple

xvii

alternate community states

Figure 5.3 Map of study site locations across the central Appalachians in 161

Virginia and West Virginia with year hemlock woolly adelgid was

detected in the county

Figure 5.4 Non metric multidimensional scaling (NMDS) analysis of the 162

overstory community composition

Figure 5.5 Non metric multidimensional scaling (NMDS) analysis of the 163

overstory community composition

Figure 5.6 Comparisons of multivariate variance in overstory community 164

composition compared by average overstory hemlock decline class (1-5)

Figure 5.7 Tukey’s honestly significant diference (HSD) comparison of variance 165

among overstory hemlock decline classes

Figure 5.8 Non metric multidimensional scaling NMDS of the sapling 166

community composition

Figure 5.9 Non metric multidimensional scaling NMDS of sapling community 167

composition

Figure 5.10 Non metric multidimensional scaling (NMDS) analysis of the 168

seedling community composition

Figure 5.11 Non metric multidimensional scaling (NMDS) analysis of the 169

seedling community composition

Figure 5.12 Structural equation model (SEM) of the influence of hemlock 170

decline on soil texture on soil cation exchange capacity and soil carbon to

xviii

nitrogen ratio

Figure 5.13 Structural equation model (SEM) illustrating the influence of 171

overstory community compostion on canopy openness in the growing

season

Figure 5.14 Structural equation model (SEM) of the influence of overstory 172

hemlock decline on relative decomposition rate constant k and leaf litter

carbon to nitrogen ratio

xix

Chapter 1: Introduction

Forest ecosystems of eastern North America have always been in continual

compositional fluctuation in response to broad changes in climate, including glacial

advancement and retreat (Davis 1969, Delcourt and Delcourt 1998, Foster et al. 2002).

On smaller spatial and temporal scales, ecosystem composition and function are regularly

reshaped by natural disturbances, including fire, windstorms, flooding, and mortality.

While change has been a consistent property of ecosystems, over the past two centuries,

the rate of community composition and change has accelerated (Foster et al. 2002). In

part this is due to novel and compounded disturbances, from climate change to invasive species, altered natural disturbance regimes, anthropogenic nutrient additons, acid deposition, and altered herbivore populations (Likens et al. 1996, Vitousek et al. 1997,

Paine et al. 1998, Foster et al. 2002, Cote et al. 2004, Groffman et al. 2006, Lovett et al.

2006, Nowacki and Abrams 2008, Ehrenfeld 2010).

The amount of change an ecosystem can absorb while maintaining functional

processes is known as ecological resilience (Holling 1973, Gunderson 2000). Ecological

resilience differs from engineering resilience, which is defined as the time for a system to

recover from change (Besiner et al. 2003, Groffman et al. 2006). Instead, ecological

resilience places emphasis on thresholds, beyond which new energy and nutrient cycles

develop, creating an alternate community state (Besiner et al. 2003, Groffman et al.

1

2006). In other words, ecological resilience with its inclusion of multiple alternate community states focuses on adaptation and change, as well as feedbacks between ecological communities and ecosystem processes (Mayer and Rietkerk 2004, Thrush et al. 2009). Ecological resilience and alternate state dynamics have been explored most extensively using mathematical approaches and small scale experiments, but are increasingly recognized to have important application in the management of some complex natural systems (Suding et al. 2004, Groffman et al. 2006, Suding and Hobbs

2009, Dodds et al. 2010, Daily et al. 2012).

The current loss of eastern hemlock (Tsuga canadensis, hemlock) due to an invasive pest insect, the hemlock woolly adelgid (Adelges tsugae, HWA) provides an unfortunate opportunity to explore the applicability of ecological resilience and alternate dynamics in complex natural systems. Eastern hemlock is a foundation species, one that dominates both composition and is an important driver of ecosystem processes (Bruno et al. 2003, Ellison et al. 2005). As evergreen forests within a largely deciduous forest region, hemlock communities contribute to landscape or beta and gamma diversity, with unique suites of aquatic macroinvertebrates, salamanders, cold water fish, and birds

(Snyder et al. 2002, Tingley et al. 2002, Ross et al. 2003). At the same time, vegetation in hemlock forests is species poor when compared to eastern mixed deciduous forests.

As ecological resiliency is thought to be a property of biodiversity (Peterson et al.

1998), hemlock forests have low resiliency. In particular, resiliency is thought to be tied to both functional diversity, which groups species based on attributes such as growth form, life history traits, and traits such as nitrogen fixation or shade tolerance, and response diversity, which groups species with similar reactions to environmental change

2

(Elmqvist et al. 2003). Ecosystems dominated by foundation species are thought to be

structured by a few strong interactions (Ellison et al. 2005), making them low in both

functional and response diversity. The loss of hemlock should therefore result in a

reorganization of energy pathways and nutrient dynamics as composition changes,

leading to the development of alternate community states. Alternate state theory generally

predicts shifts from one state to another (Beisner et al. 2003), however, in some cases

community states diverge and develop along multiple pathways (Fig 1.1, Houseman et al.

2008, Thrush et al. 2010).

The overall objective of my dissertation is to examine the impacts of hemlock loss

on ecosystem structure and function in riparian forests across the central Appalachians.

My central hypothesis is that mortality

caused by HWA will result in several unique

trajectories and community states, dependent

upon the initial composition (Fig. 1.2).

Using a chronosequence approach including

uninvaded sites as well as sites impacted for

decades, I examine how these impacts Figure 1.1 Conceptual models of alternate progress over time. Insights from my study state dynamics. Communities are often will contribute to a more complete picture of pictured moving from one state (A) to the complex compositional and functional another (B) when a threshold between impacts of HWA. I hope it will also provide basins of attraction is crossed. In other

3

information that can help guide the focus of future restoration of HWA-invaded riparian forests of the central Appalachians.

Eastern hemlock as a foundation species

Within the eastern deciduous forest region, hemlock communities create unique habitat characterized by a damp, shady microclimate where seasonal extremes in temperature and hydrology are muted (Ellison et al. 2005, Ford and Vose 2007).

Compared to surrounding deciduous forests, hemlock communities are characterized by slowly decaying, acidic litter, lower light availability, and transpiration rates distributed

Figure 1.2 Conceptual model of hypothesized alternate state changes caused by HWA.

While hemlock forests are similar throughout the range, HWA causes communities to

diverge.

4

throughout the year but with high rates in the spring (Canham et al. 1994, Ellison et al.

2005, Ford and Vose 2007). Hemlock is also an important structural component,

providing regeneration sites protected from deer browse (Krueger and Peterson 2006) and

slowly decaying large wood in streams (Wallace et al. 2001, Morris et al. 2007). In the

Northeast, hemlock occurs throughout the landscape, while south of approximately the

state of Pennsylvania, hemlock is largely restricted to cove and riparian forests, perhaps

due to cold air drainage and naturally higher humidity (Oosting and Hess 1956, Adams and Loucks 1971, Runkle and Whitney 1987).

Across much of the central and southern Appalachians, hemlock dominance increased in the early twentieth century as a former foundation species, American

Figure 1.3 USDA Forest Service HWA distribution map for 2011

5

chestnut (Castanea dentata) was functionally removed by chestnut blight (Cryphonectria parasitica) (Day and Monk 1974, Elliott et al. 1999, Vandermast and VanLear 2002,

Ellison et al. 2005, Elliott and Swank 2008). Changes in ecosystem function following the decline of American chestnut, which composed 40-50% of the basal area of eastern forests (Keever 1953, Anagonostakis 1987, Elliott and Swank 2008), are largely unknown, which underscores the vanishing opportunity to quantify the influence of hemlock while some forests remain un-invaded and hemlock remains in others (Young et al. 2002).

Introduced in 1951 on nursery stock in Richmond, Virginia, HWA has since spread from northern Georgia to southern Maine, with scattered occurrences west of the

Appalachian Mountains (Fig. 1.3). Evans and Gregorie (2006) estimate the mean rate of spread at 12 km·yr-1 but this is accelerated in warmer climates in the southern part of the

range. Both spread and mortality are slowed by cold winter temperatures, and Dukes et

al. (2009) estimate HWA is impacted by average winter temperatures of -5°C and cannot

survive minimum temperatures below -28.8°C, which may be increasingly less common

due to climate change. Already, the southern portions of the hemlock range, particularly

at lower elevations, are warmer than this threshold (Dukes et al. 2009). A small, aphid

like insect (Fig. 1.4), HWA feeds on the parenchyma cells of xylem rays, causing the loss

of needles and buds, leading to the death of branches and eventually, total tree mortality

(McClure 1991, Young et al. 1995). There is no evidence of resistance and almost

complete mortality occurs over four to fifteen years (McClure 1991). Individual

hemlocks can be treated with horticultural oils or systemic insecticides (Onken and

Reardon 2005), but this is costly and labor intensive. On a landscape scale, biological

6

Figure 1.4 HWA appears as small white tuffs on the underside of needles.

control efforts have focused on the release of predatory beetles from the native HWA

range in Asia, but such experiments have yet to yield great success (Onken and Reardon

2008). While biological control research continues, many forests are already

experiencing high levels of hemlock mortality (Ellison et al. 2005).

Impacts of HWA

Although alternate state theory predicts rapid shifts to alternate states, most examples come from shallow lakes and grassland ecosystems, both dominated by short lived species that rapidly respond to change (Mayer and Rietkerk 2004). In cases such as forests, dominated by long-lived species, the theory may require some expansion

(Groffman et al. 2006). The literature suggests when alternate state transitions occur in

7

Figure 1.5 Hemlock snags in Bath County, Virginia. HWA was detected in 1993.

complex ecosystems of long-lived species such as forests, functional changes are rapid,

while composition and structure change over decades (Fig. 1.5).

In the case of HWA, changes in productivity, nutrient dynamics, and water

budgets are apparent nearly as soon as HWA arrives and even before major decline

(Jenkins et al. 1999, Cobb et al. 2006, Stadler et al. 2006, Daley et al. 2007, Ford and

Vose 2007, Nuckolls et al. 2008, Orwig et al. 2008). Functional processes are likely to

continue to fluctuate as composition changes.

Some of the changes in composition following hemlock decline are apparent from

historical records and detailed studies in New England and the southern Appalachians.

Approximately 5,500 ybp, hemlock went through a major population crash due to

climatic shifts (Heard and Valente 2009), and pollen records indicate increases in birch

8

(Betula), maple (Acer) and oak (Quercus). This broad pattern seems to hold for the current decline of hemlock, although evidence suggests that community response differs and thus, hemlock forests throughout the range will likely reorganize along different pathways. In New England, sweet birch (Betula lenta) is increasing (Orwig and Foster

1998, Orwig et al. 2008), but these changes vary with environmental context (Small et al.

2005). Southern Appalachian forests may be more diverse, and also exhibit differences by environmental context (Dhungel et al. 2010, Spaulding and Rieske 2010, Krapfl et al.

2011, Ford et al. 2012). Further, central and southern Appalachian sites often have a significant shrub layer, mainly the native evergreen shrub rhododendron (Rhododendron maximum). Rhododendron is known to be an aggressive competitor, altering gap dynamics, light availability, nutrient cycling, and hyrology (Clinton et al. 1994,

Kominoski et al. 2007, Ball et al. 2008, Ford and Vose 2007, Beckage et al. 2008,

Wurzberger and Hendrick 2008, Ford et al. 2012). Following hemlock decline in North

Carolina, Ford et al. (2012) found rhododendron increased its growth rate 2.6 times over a five year study period.

Focus of this dissertation

To test for evidence of community divergence, indicating transitions to multiple alternate community states following hemlock loss, and clarify some patterns in the compositional and functional changes as hemlock declines, my dissertation examines sites across the central Appalachians. This region is an important area for study, as it is not well-represented in the literature and includes the natural hemlock populations closest the site of HWA introduction. Further, the central Appalachians are a transition zone

9

between northern and southern floras (Braun 1950), which may further emphasize local and regional differences in HWA response. To explore the influence of hemlock as a foundation species in the central Appalachians, including un-invaded Ohio, my dissertation includes four main chapters. The theoretical framework of my approach is outlined in Chapter Two, which provides a review of the literature on ecological resilience and alternate states. After summarizing the background of this theoretical framework, I explore how it might apply in complex natural ecosystems experiencing novel changes. What is known and suggestions for further study are outlined for the loss of hemlock, as well as two additional cases: the mesophication of oak forests due to changing fire regimes and the invasion of emerald ash borer (Agrilus planipennis) in mixed deciduous forests.

The remaining chapters focus on data collected from 38 hemlock dominated riparian forests across the central Appalachians: in Ohio, where HWA is not currently present, in West Virginia, where HWA was detected 1993-2002, and in Virginia where

HWA was detected 1979-1993. Sites are located on land managed by state and federal agencies and in one case, the Nature Conservancy. Physiography and soil characteristics vary as sites are spread throughout the Appalachian Plateau, Allegheny Mountains, and

Ridge and Valley Phyiographic provinces (Fralish 2003). Sites were located from the species range map complied by Prasaad and Iverson (1999) and the advice of local managers, and selected as representative of relatively undisturbed hemlock dominated riparian forests. Hemlock dominance along the stream was assessed visually by the presence of live trees and snags. At each site, a series of three transects at 10, 30, and 50 meters from the bank-full channel of a small headwater stream were used to collect

10 vegetation and ecosystem function data. Transects consisted of a series of five 100-m2 plots.

Chapter Three is an investigation of the influence of hemlock on forest community compositon and function in un-invaded forests on the unglaciated Allegheny

Plateau of southeastern Ohio. The objective was to quantify the influence of hemlock as a foundation species prior to HWA arrival. The chapter examines how hemlock structures biodiversity and community composition, and key aspects of ecosystem function, including light availability (using hemispherical photography during the growing season and deciduous leaf-off period), leaf litter biomass and chemistry, and relative decomposition rates.

In Chapter Four, we move over to West Virginia and Virginia and examine hemlock dominance and mortality in sites impacted by HWA for 9-32 years. The literature suggests that hemlock mortality may initially vary by environmental context, but is eventually complete (Orwig et al. 2008). The objective is to determine whether any hemlock remains healthy in the overstory or sapling layers, whether hemlock declines by environmental context, and whether invasive plant species are taking advantage of hemlock canopy openings.

Chapter Five takes a more in-depth view of the impact of HWA on ecosystem structure and function across thirty riparian forests in West Virginia and Virginia. The central hypothesis for the chapter is that as a foundation species, hemlock is the most important driver of forest dynamics across wide environment gradients, in terms of soil, physiography, and elevation. Therefore, as it declines, we expect to find evidence of change in community composition, including increased variance in communities

11 impacted by HWA the longest. This would support the hypothesis that hemlock communities are transitioning to multiple alternate community states. As composition changes, we also expected changes in resource availability, productivity, and nutrient cycling as hemlock declines.

12

Chapter 2: Resilience, alternate states, and biodiversity of eastern forests during rapid environmental change

2.1 Abstract

In an era of rapid ecological change from loss of biodiversity, climate change, and

altered nutrient dynamics, the resilience of complex ecosystems is an increasing concern.

Ecological resilience is defined as the capacity of a system to absorb change while

maintaining a specific structure or composition and function. Concepts of ecological

resilience are largely based on theory and small-scale experiments, and thus are not well

understood or tested in complex natural systems. At the same time, forest management

plans are increasingly disturbance-based, which includes an underlying assumption of

ecosystem resilience, but metrics of this assumed resilience are unclear. Our objective is

to initiate a more critical examination of current and future resilience in eastern forest

ecosystems in the context of continued ecological change. We suggest that experimental

studies designed to identify important thresholds and non-linear dynamics are required to

develop a more quantitative understanding of resilience in complex forest ecosystems.

Such investigations will expand and advance resilience theory, but are also directly

applicable to management challenges. Examples of rapidly changing forests that may

provide model systems to explore forest resilience include the ongoing shift from oak

(Quercus) to maple (Acer) dominance as a result of shifting fire regimes, the loss of ash

(Fraxinus) species from diverse hardwood forests due to an exotic pest insect the

Emerald Ash Borer (Agrilus planipennis), and the loss of a foundation species, eastern hemlock (Tsuga canadensis) due to hemlock woolly adelgid (Adelges tsugae).

13

2.2 Introduction

In the context of rapid changes in biodiversity and climate, ecological resilience is

increasingly thought to be an important component of complex natural ecosystems

(Millar et al 2007), but it is not well understood or tested (Dodds et al. 2010). A greater

mechanistic and quantitative understanding of forest ecosystem resilience is necessary to make accurate predictions of future dynamics and achieve management goals, particularly as acid and nitrogen deposition (Likens et al. 1996, Vitousek et al. 1997), a

continually expanding list of exotic pests and pathogens (Lovett et al. 2006), invasive

species (Ehrenfeld 2010) and overpopulation of herbivores such as white-tailed deer

(Cote et al. 2004) continue to shift forest composition and ecosystem processes,

particularly as the climate changes, creating compounded disturbances (Paine et al.

1998).

In ecological theory, resilience is defined as the amount of change an ecosystem can

absorb, where key functional processes are maintained (Holling 1973, Gunderson 2000,

Beisner er al. 2003, Groffman et al. 2006). Disturbances, both natural and anthropogenic,

that exceed resilience thresholds result in the development of alternate energy and

nutrient cycles, characterizing a distinctly different alternate community state (Groffman

et al. 2006, Besiner et al. 2003). Ecological resilience differs slightly from engineering

resilience, which is defined as the amount of time required for a system to recover

(Besiner et al. 2003, Groffman et al. 2006). Perhaps the key differentiation in

perspectives is the ecological focus on alternate state dynamics which emphasizes a

community’s ability to adapt to change (Thrush et al. 2009). Identification and

acceptance of alternate state dynamics has been debated in ecological literature for

14

decades and may not be applicable in all ecosystems (Schröder et al. 2005, Suding and

Hobbs 2009). Further, alternate states may not always be stable (Schröder et al. 2005,

Thrush et al. 2009, Dodds et al. 2010), and perhaps it is most illustrative to consider

alternate states to be defined by feedbacks between the community and ecosystem

processes (Mayer and Rietkerk 2004). Much of the literature on resilience is either theoretical or occurs in small-scale experiments and threshold identification is debated from a statistical perspective (Daily et al. 2012). Yet, an increasing number of examples of threshold dynamics the resilience of complex natural systems are reported in the literature; in some cases changes are so obvious they can be identified without non-linear statistics (Dodds et al. 2010). Often, these examples are identified because thresholds have been exceeded and systems do not respond to management as expected (Suding et al. 2004, Groffman et al. 2006, Suding and Hobbs 2009).

We apply current understanding of alternate state dynamics, foundation species, and

biodiversity-ecosystem function to examine the resilience of forest ecosystems of the

eastern United States. We outline what is understood and what remains unclear using

examples where ecosystems are experiencing both rapid and gradual transitions to

alternate states. We also highlight an example where the ecosystem remains resilient, but

may have less capacity to absorb additional future changes. We consider the examples

chosen to be model systems due to their prominent forest management concerns and

because of the general understanding of the important ecosystem processes associated

with these ecosystems. In each of our examples, ecosystem services may be lost at the

landscape level as these forests move toward landscape traps with feedbacks that are

15

difficult to manage (Lindenmayer 2011). Our objective is to add to the discussion of

resilience, particularly as it may be applicable in forest ecosystems.

2.3 Resilience and alternate states

Conceptually, resilience and alternate state theory are represented by a ball

(community) moving across a surface with hills (thresholds) and valleys (states or basins

of attraction) (Beisner et al. 2003; Fig. 1). Deeper, wider valleys represent greater

resilience, where changes may move the ball, but it usually settles back into place.

Movement between alternate states results when the ball escapes the valley. This can be

the result of a change large enough for the ball to cross the threshold and leave the valley, or if the valley changes shape, and the ball moves across the lowered threshold. A change that results in movement of the ball is referred to as a change in the state or fast variable, as these changes are characterized as occurring rapidly (Suding and Hobbs

2009). Examples of alternate states have been reported primarily from grassland and lake ecosystems. A classic example of this kind of state change is the shift of shallow lakes from a clear to a turbid state (Scheffer et al. 1993). Changes to the shape of the surface can be referred to as shifts in slow variables, parameters, or controlling variables (Beisner et al. 2003). For example, a recent study by Hilt et al. (2011) identified water retention time as an important parameter determining shifts in between clear and turbid alternate states in shallow lakes and rivers. Slow and fast variables are sometimes referred to as external and internal, and Mayer and Rietkerk (2004) provide additional examples.

Although conceptual models generally indicate movement from one community state to another, in some cases shifts may instead result in community dispersion, depending on the disturbance type and whether multiple disturbances interact (Thrush et al. 2010).

16

For example, when Sasaki and Laurenroth (2011) removed the dominant C4 grass from a short grass steppe in Colorado to test stability, the resulting communities were heterogeneous and less stable, possibly indicating differential responses depending on environmental context, even on local scales. In an experiment designed to examine patterns of community response to disturbance, Houseman et al. (2008) applied factorial disturbance treatments to low productivity grasslands in Michigan. The authors found community dispersion decreased following a biomass removal disturbance, continuous fertilization increased dispersion, and the combined treatment resulted in three distinct alternate community states.

In many cases, alternate state changes are considered undesirable and some states do not provide desired ecosystem services, but are reinforced by internal feedbacks that complicate management intervention. In a review of alternate states in restoration ecology, Suding et al. (2004) outline examples from grasslands and coral reefs where changes in system dynamics challenge restoration treatments designed based on more linear succession models. Advancement of theory and its practical application in management requires greater understanding the mechanisms and feedbacks that maintain or restore resilience of alternate states.

2.4 The role of diversity in resilience

Resilience is thought to be a property of biodiversity, particularly functional diversity

(Peterson et al. 1998). While functional diversity typically measures species groups based on attributes such as growth form, life history traits, and traits such as nitrogen fixation or shade tolerance, response diversity groups species with similar reactions to environmental change (Elmqvist et al. 2003). Functional and response diversity create

17

redundancy that maintains ecological feedbacks in energy and nutrient exchanges during disturbances. For example, in a grassland community, Walker et al. (1999) found that under stable conditions, the community is characterized by a few dominant species from disparate functional groups and minor species more similar to the dominant species.

Selective grazing pressure on the dominant species results in an increase in minor species with functional overlap, providing resilience within the community.

If diversity is the key driver of resilience, it is likely that ecosystems structured by one or a few foundation species (Ellison et al 2005) ,are low resilience systems.

Foundation species are common or abundant species that define ecosystem exchanges of

nutrients and energy as well as microclimate, thus providing habitat for other species

(Bruno et al. 2003, Ellison et al. 2005). The low diversity within these systems indicates

these communities are structured by few, strong interactions with little functional

redundancy (Peterson et al. 1998). The loss of foundation species therefore destabilizes

community structure as well as energy and nutrient dynamics, causing a rapid re-

organization to an alternate state (Ellison et al. 2005). In the example from the short

grass steppe of Colorado, Sasaki and Laurenroth (2011) demonstrate that when one

dominant, or perhaps foundation, species was removed, the resulting communities were

more diverse but less stable. As stability is thus dependent on one species, the community

exhibits low resilience to disturbances that remove this species.

2.5 Resilience in forest ecosystems

A greater depth of understanding ecosystem resilience to disturbance is increasingly

recognized as a critical component for accurate predictive models and restoration and

management plans for ecosystem services (Scheffer et al. 2001, Mayer and Rietkerk

18

2004, Groffman et al. 2006, Suding and Hobbs 2009). While forest ecosystems provide

many ecosystem services, there is limited understanding of resilience in these complex

systems dominated by long-lived species. Many of the forested landscapes of eastern

North America are second-growth ecosystems that developed following a history of

widespread logging and land conversion for agriculture and urbanization in the early

twentieth century. Although, or perhaps because they have been in consistent flux over

the last two centuries (Foster et al. 2002, Nowacki and Abrams 2008), these forests are

presumed to be resilient, a fact emphasized by management plans designed to maintain

resilience, often incorporating natural disturbance (Attiwill 1994, Drever et al. 2006).

For example, while Millar et al. (2007) underscore the importance of maintaining and

increasing resilience in forests facing uncertain climate futures, they do not provide detail

about current levels of resilience or differentiate resilience between forests. Eastern

forests provide important examples to examine current and changing ecosystem resilience

from both ecological and societal perspectives. These changes in eastern forests occur across a heavily populated region where ecosystem services including carbon capture,

flood storage, nutrient retention, and water and air quality are critical to society

(Constanza et al. 1997).

2.6 Alternate state transitions from oak to maple: mesophication

Much of the research and management focus across the Central Hardwoods Forest

Region of the eastern U.S. is devoted to an ongoing compositional shift as oak-hickory

forests (Quercus- Carya) are replaced by other hardwoods, particularly maple (Acer;

Fralish and McArdle 2009, Fei et al. 2011). Species including red maple (Acer rubrum)

increasingly dominate the forest regeneration layers and gradually, canopies shift in

19

composition (Hutchinson et al. 2008). Drury and Runkle (2006) found maple dominated

regeneration layers in both older oak-dominated and younger forests where maple already

dominated the canopy, supporting the oak bottleneck as suggested by Sutherland et al.

(2000). This transition is considered undesirable in many cases, as mature oak-hickory

forests are valued for forest products, natural heritage, and wildlife habitat, including

game species such as white-tailed deer and turkey (Pierce et al. 2006).

Approaches to restore oak-hickory ecosystems typically use prescribed fire,

mechanical canopy thinning, herbicides, or a combination to encourage oak recruitment.

Yet, the success of many oak restoration efforts is limited. In some cases, maple and

mesic species continued to dominate the regeneration layers after multiple prescribed

fires (Hutchinson et al. 2005, Green et al. 2010) or combinations of prescribed fire and

mechanical thinning (Albrecht and McCarthy 2006). Waldrop et al. (2007) suggest that

goals to re-create open oak forests require continued management and treatment over

time. In a review outlining studies from across the Central Hardwood Forest, McEwan et

al. (2011) demonstrate that there are no clear patterns of restoration success in multiple

treatment combinations. A more comprehensive understanding of the oak-maple

transition is required to improve restoration success.

Nowacki and Abrams (2008) described the oak-hickory transition as part of a broader

process they define as mesophication, emphasizing that frequent fire is thought to be the

main driver perpetuating communities including shade intolerant oak and hickory species. Mesophication suggests that over decades of fire suppression, more shade

tolerant, mesic species such as maples established and altered fuel and moisture

dynamics, further suppressing fire and reinforcing their dominance (Nowacki and

20

Abrams 2008). The failure of restoration efforts including prescribed fire, mechanical

thinning, and combinations of both indicates feedbacks reinforce in the maple-dominated state creates challenges for restoration treatments.

Conceptually, the transition between oak-hickory and maple can be represented as

two shallow, broad valleys with a gradual hill (threshold) in between, each maintained

through by an interaction between the controlling (slow) and state (fast) variables (Fig. 2,

see also Nowacki and Abrams (2008)). In this model, the oak-hickory state is maintained

by a feedback between fire and oak litter. During extended periods without fire, the

establishment of mesic species shifts the microclimate and fuel conditions. Maples and

other mesic species create fire-resistant conditions by increasing humidity and soil

moisture (Nowacki and Abrams 2008). As the transition progresses, the maple dominated

state becomes resilient to fire, complicating restoration treatments. Such transitions may

be common across fire-dependent ecosystems that exhibit alternate state transitions

(Groffman et al. 2006, Martin and Kirkman 2009).

While the transition between these states has been outlined by Nowacki and Abrams

(2008), many details of the process are not understood, particularly at smaller scales.

Forested systems may require an expansion of alternate states, where shifts may not be as

rapid for long lived tree species when compared to species such as aquatic macrophytes,

rocky intertidal organisms, and grasses, upon which much of theory is based. For

example, in forest ecosystems, transitions in function as species and their influence

decline may precede more obvious shifts in composition. Investigations in the oak-maple

transition have documented some components of the shifts in ecosystem function

(Alexander and Arthur 2010, Fox et al. 2010). The influence of specific tree species,

21

including comparisons of oak and maple, on nutrient cycling and pH has been established

(Finzi et al. 1998, Finzi et al. 1998b, Lovett et al. 2004). Thus, the shift in dominance

from oak to maple should result in a significant shift in nutrient cycles, which is

supported by studies examining this transition. In a study comparing nutrient dynamics of

leaf litter packs composed entirely of oak and litter containing approximately half oak

and half deciduous species, Piatek et al. (2010) found slower decomposition and greater

retention of carbon, nitrogen, phosphorus and calcium in the oak litter. The increased

nutrient cycling rate is consistent with accelerated nitrogen cycling rates under red maple

when compared to both chestnut and scarlet oak identified by Alexander and Arthur

(2010). In addition to altering the temporal scale of nutrient cycling, the authors

identified spatial shifts in nutrients and hydrology by identifying differences in

throughfall and stem flow under red maple compared with chestnut oak and scarlet oak.

While shifts in processes such as nutrient cycling are documented, and lack of

response to restoration treatments are apparent, the current literature does not outline

details of the resilience of the oak and maple. This is an important gap in our

understanding of how resilience is structured by feedbacks between a community and its

environment, and limits implementation of effective restoration treatments. In the

resilience literature, Thrush et al (2010) point to a lack of understanding of the context

dependency of threshold responses, and in the case of oak there is some indication of a

broad differentiation between xeric and mesic sites. For example, oaks may have more of

a competitive advantage on more xeric south facing slopes (Goebel and Hix 1996), while

more mesic species are more competitive on north and east slopes (Iverson et al. 1997).

Conceptually, Nowacki and Abrams (2008) indicate lower thresholds and shallower

22

basins for mesic uplands compared to xeric uplands. In addition, McEwan et al. (2011)

present a compelling argument for a wider examination of compounding factors,

including climate, land use change, loss of American chestnut (Castanea dentata) and herbivory.

Current restoration and management treatments are not designed to specifically identify thresholds between oak and maple community states. Without new approaches, it is unlikely we will be able to answer vital questions such as: 1) What are the feedbacks that maintain oak and maple dominance and where and how does the shift occur? 2) How do feedbacks that maintain oak dominance vary with environmental context, is there a critical xeric/mesic division? 3) How important is the canopy composition in determining regeneration, thus determining the level of manipulation necessary? 4) What role does fire behavior play and how do fuel characteristics regulate this behavior? 5) What are the

relative strengths of the influence of light availability, nutrients, and seed source in

determining sapling composition? These are just some of the remaining questions that

could be addressed to increase the efficiency and success of management. This is

particularly critical as mesophication progresses and an increasing portion of the

landscape transitions to maple, particularly in the Central Hardwoods region (Nowacki

and Abrams 2008, Fei et al. 2011). Widespread transition to the maple-dominated state

results in homogenization or loss of landscape (gamma) diversity (Amatangelo et al.

2011), which in turn decreases resilience across eastern forests. While Matthews et al.

(2011) predict red maple to be fairly resilient to climate change, large areas of maple forest are increasingly susceptible to Asian longhorned beetle (ALB), an exotic pest that

23 feeds preferentially on maples and has already escaped to natural forests in the northeast

(Dodds and Orwig 2011).

2.7 Reshaping resilience across eastern deciduous forests: EAB

Species-specific mortality events such as ash (Fraxinus) mortality due to Emerald

Ash Borer (Agrilus planipennis; EAB) provide an additional opportunity to examine resilience and response diversity. Unlike oak-hickory forests, ash rarely defines a community state and the loss of ash will have more subtle impacts on forest structure and function. Although mono-dominant black ash stands occur in forested wetlands of the northern Great Lakes and Canada (Poland and McCullough 2006), we limit our current discussion to more diverse forests where ash occurs as a component species. These forests should exhibit some functional and response redundancy and thus, resilience.

Across eastern forests, functional overlap between ash and other deciduous species may provide similar resiliency.

Native to Asia, EAB is a phloem-feeding beetle that feeds on all species of the

Fraxinus genus. The EAB-impacted area currently includes 15 states and two Canadian provinces and has already resulted in the death of tens of millions of ash trees in

Michigan alone (Kashian and Wittier 2011). The first signs of ash mortality from EAB were discovered in the greater Detroit, Michigan area in 2002. While the broader ecosystem consequences of EAB remain the subject of investigation (Gandhi and Herms

2010, Kashian and Witter 2011), EAB continues to spread. In the Central Hardwood

Forest Region, Sutherland et al. (2000) place green (F. pensylvanica) and white (F. americana) ash in a functional group characterized fast-growing species with the ability to persist in lower canopy positions, also including black cherry (Prunus serotina),

24

flowering dogwood (Cornus florida), hackberry (Celtis occidendalis), boxelder (Acer

negrundo), and mulberry (Morus alba). There is also some evidence that ash expanded in

response to elm decline due to Dutch elm disease (Barnes 1976), and thus may have

provided response diversity in the past. The future of ash in eastern forests remains

uncertain. In areas of Michigan with high ash mortality, Kashian and Witter (2011) found high ash regeneration levels, thus the continued presence of EAB in these forests will determine whether ash can recover or be restored in the future.

We represent the simplification of diversity from the loss of ash, as a change in the basin shape, making it narrower, with shorter sides (Fig. 3). Thus, while the state (fast) variable does not change, there are shifts in parameters (also referred to as the slow or controlling variables, Beisner et al. 2003). Although the loss of ash will not result in a change in ecosystem state immediately, a reduction in functional and response diversity will make these ecosystems more vulnerable to further perturbation (Paine et al. 1998,

Peterson et al. 1998, Walker et al. 1999, Scheffer et al. 2001).

This understanding of resiliency is conceptual and merits quantification. Important questions remain regarding how we measure resiliency across diverse forests. In the case of ash, uncertainties include: 1) What are the quantified functional contributions of ash in eastern forests that remain uninvaded and how does this differ between ash species? 2)

What species provide functional overlap with ash and for which function? It is likely that multiple species provide redundancy for different functions, such as habitat structure, productivity and soil nutrient feedbacks (Finzi et al. 1998). Understanding both the role of ash and the impact of its loss on diversity will provide a model for species-specific mortality in other diverse forests.

25

2.8 Foundation species ecosystems exhibit low resilience: eastern hemlock

Ecosystems structured by foundation species are low in diversity and are susceptible

to rapid shifts to alternate states (Ellison et al. 2005), indicating low resiliency. While

alternate state theory generally predicts ecosystems moving from one state to another

specific state (Beisner et al. 2003), increasing evidence indicates this does not hold for

ecosystems dominated by foundation species that occur across wide ecological gradients.

Instead, as foundation species are removed, ecosystems develop along multiple pathways

(Houseman et al. (2008), which are context dependent (Thrush et al. 2010). Some of the

resulting states may be considered degraded as they may not provide the desired level of

ecosystem services in terms of habitat structure and function.

The decline of eastern hemlock (Tsuga candensis) due to hemlock woolly adelgid

(Adelges tsugae; HWA) is an often cited case study examining the impact of an invasive

pest (Lovett et al. 2006, Ehrenfeld 2011). Given the current rates of hemlock mortality

and lack of effective landscape scale HWA control, hemlock will likely be eliminated

from much of its range within the next few decades (Ford et al. 2012). Ironically,

hemlock expanded on the landscape, particularly in riparian forests, following the decline

of a former foundation species, American chestnut (Castanea dentata) that was

functionally eliminated by chestnut blight (Cryphonetria parasitica) in the early

twentieth century (Ellison et al. 2005).

In alternate state terms, foundation species ecosystems such as hemlock forests

should have a narrow basin of attraction without steep sides, illustrating that a small

perturbation will push the system across state thresholds (Fig. 4). While alternate state

26 shifts are thought to be triggered by either a shift in state or (fast) variables, such as a large population changes, or by parameter shifts (controlling or slow variables) caused by a change in the environment (Besiner et al. 2003, Suding and Hobbs 2009), state and parameter variables are coupled in cases such as hemlock. In such ecosystems, both the state and parameter values, including environmental conditions such as microclimate, nutrient exchanges, and hydrology are controlled by feedbacks with the foundation species.

As foundation species are lost, the rapid shifts to an alternate state are probably first apparent in functional processes. Even at initial stages of decline and mortality, HWA alters productivity, nutrient dynamics, and water budgets. In North Carolina, Nuckolls et al. (2008) found rapid declines in productivity following invasion. As hemlock productivity declines, patterns of water uptake and evapotranspiration become more seasonal (Daley et al. 2007, Ford and Vose 2007). At initial infestation densities, HWA also alters throughfall and forest floor chemistry (Stadler et al. 2006), resulting in accelerated rates of decomposition (Cobb et al. 2006) and nitrogen cycling (Jenkins et al.

1999, Orwig et al. 2008). Overall, these changes in ecosystem function are likely part of the transition phase and long-term nutrient and energy cycling will be restructured in ways that are not yet clear. We believe these processes will continue to fluctuate as compositional changes occur more gradually (Krapfl et al. 2011). Furthermore, understanding the scope of these changes is complicated by the variation in environmental context throughout the range of eastern hemlock, which occurs from the

Northeast and Lake States south to the southern Appalachians of Georgia and Alabama.

27

While forests throughout the range represent one hemlock community state, evidence

indicates mortality is causing divergence at local and region scales. As a foundation species, hemlock exerts biotic control that probably dampens the influence of abiotic factors that occur throughout its range (Ellison et al. 2005). As forest communities reassemble through much of the eastern U.S., the biotic control exerted by hemlock will be replaced by a greater abiotic control. Redeveloping forests across the hemlock range will be influenced by the variability in physiography, soil, habitat fragmentation, seed sources, and the current forest community (Young et al. 2002, Small et al. 2005, Ford et al. 2012). Comparisons between forests in New England and the southern Appalachians support this increased importance of abiotic factors in hemlock forests and indicate that

predictions from one part of the hemlock range are not applicable elsewhere. In New

England, hemlock is being replaced largely by black birch (Betula lenta) (Orwig and

Foster 1998, Kizlinski et al. 2002). Yet, variation is evident at regional or local scales, as

Small et al. (2005) found different species trajectories for xeric ridges and mesic ravines

in Connecticut. The future composition and structure of ravine and riparian forests is not

as clear for the central and southern portions of the hemlock’s range (Ellison et al. 2005).

Perhaps most importantly, post-hemlock forest dynamics will be determined in part based

on the presence of an additional foundation species in the central and southern

Appalachians, the evergreen shrub Rhododendron maximum. While Ford et al. (2011) found a mix of Acer, Betula, Fagus, and Quercus in the Nantahala Mountains of North

Carolina, Krapfl et al. (2011) found no differences in community composition after 5-10 years of HWA presence in the Smoky Mountains. Similarly, communities across central

Appalachian sites impacted by HWA for decades continue to be dominated by hemlock,

28

although in severe decline (Chapter 3). The resulting impact on ecosystem processes also

remains a subject of inquiry and likely depends on the abundance of Rhododendron. In a

four-year study in North Carolina, Knoepp et al. (2011) identified different nutrient cycling between hardwood and hemlock stands, but did not find significant changes either within or between HWA-infested and girdled hemlock stands, possibly due in part to Rhododendron.

Rhododendron is known to alter gap dynamics (Clinton et al. 1994, Beckage et al.

2000, Beckage and Clark 2003, Beckage et al 2008), nutrient cycling (Kominoski et al.

2007, Ball et al. 2008, Wurtzberger and Hendrick 2009; Knoepp et al. 2011), and forest hydrology (Ford and Vose 2007). In areas of hemlock mortality throughout the central and southern Appalachians, alternate communities driven by a Rhododendron foundation are developing (Ford et al. 2012). These shrub-dominated communities differ from hemlock riparian forests in significant ways, resulting in a loss of ecosystem services that will persist without management intervention.

The divergent communities that develop following hemlock mortality will present complex management challenges that will vary regionally. In particular, loss of ecosystem services may be particularly significant in the central and southern

Appalachians (Ford et al. 2012). Management to address landscape traps created by communities such as Rhododeondron shrub forests will require continued expansion of the understanding of foundation species and alternate states. In particular, there is a gap in understanding of the following: 1. How similar is the hemlock community state across its geographic range in terms of species composition and function? 2. What is the extent of the divergence in community development following hemlock mortality at local and

29

regional scales? 3. What is the timescale for community redevelopment and how does

this vary across hemlock forests? 4. How does the presence of Rhododendron shape the composition, function, and ecosystem services of forests in the central and southern

Appalachians? 5. How will functional processes respond on short and long term scales and how is this influenced by the rate of development in the vegetation? As increasing numbers of species-specific pests and pathogens are introduced in forest ecosystems, hemlock may serve as an important model for other foundation species structured ecosystems.

2.9 Implications

Eastern forests are losing diversity and this regional homogenization lowers

resilienceat the landscape scale, limiting their ability to respond to increasingly

compounded environmental change (Scheffer et al. 2001). What remains unclear is how

well our understanding of resilience will aid in predicting and managing future forest

composition and function, and thus, the ecosystem services available to society.

Understanding resilience in eastern forests requires both observational and experimental

studies specifically designed to examine feedbacks, thresholds, and the relative strengths

of environmental factors in complex systems. The example forest ecosystems we outline

may provide models for resilience studies in complex ecosystems with long-lived species.

Oak-hickory forests are gradually but steadily transitioning to maple dominance, with

ramifications for ecosystem function and services. Management to restore oak-hickory

forests must take into account the resilience of the maple-dominated state altered fuel

conditions and thus, the complex response to fire. Additional perturbations, including

climate change, the loss of American chestnut, and increased herbivory probably also

30

contribute to the balance between states (McEwan et al. 2011). As efforts to shift the balance back to oak have yielded mixed results, specifically addressing interactions that

structure resilience will likely be the most effective and translatable across landscape positions (Thrush et al. 2009). The motivation to prevent landscape homogenization by mesophication may be more urgent in light of the susceptibility of maple to invasion by

Asian longhorned beetle (Dodds and Orwig 2011). Even more diverse mixed hardwood

forests are losing resilience to future perturbations, as EAB eliminates ash from forests

that have already experienced a decline in elm and elimination of American chestnut. The

decline of functional and response diversity is not confined to mixed hardwood forests.

As eastern hemlock forests rapidly transform to alternate states due to HWA,

understanding the variation of novel states at local and regional levels is crucial. This

understanding will advance alternate state theory, while at the same time allowing

managers to adapt strategies appropriate for different environmental contexts and land

management goals.

2.10 Acknowledgements

Funding for this research was provided by the Ohio Agricultural Research and

Development Center from a SEEDS Interdisciplinary and a Graduate SEEDS grant, the

Thorne Memorial Fellowship, and the Directors Associateship. Support was also

provided to the first author from an Ohio State University College of Food, Agriculture,

and Environmental Science Environmental Fellowship and NSF GK-12 Grant 0638669.

Earlier drafts of this manuscript were improved by valuable comments of Kay Kirkman,

Brian Palik, and Peter Curtis.

31

2.11 References

Albrecht, M.A. and B.C. McCarthy. 2006. Effects of prescribed fire and thinning on tree

recruitment patterns in Central Hardwood forests. Forest Ecology and

Management 226:88-103.

Alexander, H.D. and M.A. Arthur. 2010. Implications of a predicted shift from upland

oaks to red maple on forest hydrology and nutrient availability. Canadian Journal

of Forest Research 40:716-726.

Attiwill, P.M. 1994. The disturbance of forests ecosystems- the ecological basis for

conservative management. Forest Ecology and Management 63:247-300.

Ball, B.A., M.D. Hunter, J, S. Kominoski, C.S. Swan and M.A.Bradford. 2008.

Consequences of non-random species loss for decomposition dynamics:

experimental evidence for additive and non-additive effects. Journal of Ecology

96:303-313.

Barnes, B.V. 1976. Succession in deciduous swamp communities of southeastern

Michigan formerly dominated by American elm. Canadian Journal of Forest

Research 54: 19-24.

Beckage, B. and J.S. Clark, J.S. 2003. Seedling survival of three forest tree species: the

role of spatial heterogeneity. Ecology 84:1849-1861.

Beckage, B., J.S. Clark, B.D. Clinton, and B.L. Haines. 2000. A long-term study of tree

seedling recruitment in southern Appalachian forests: the effects of canopy gaps

and shrub understories. Canadian. Journal of Forest Research. 30:1617-1631.

32

Beckage, B., B.D. Kloeppel, J. A. Yeakley, S.F. Taylor, and D.C. Coleman. 2008.

Differential effects of understory and overstory gaps on tree regeneration. Journal

of the Torrey Botanical Society 135:1-11.

Beisner, B.E., D.T. Haydon, and K. Cuddington. 2003. Alternative stable states in

ecology. Frontiers in Ecology and the Environment 1: 376-382.

Bruno, J.,J. Stachowicz, and P.A. Townsend. 2003. Inclusion of facilitation into

theoretical ecology. 2003. Trends in Ecology and Evolution 18:119-125.

Chapin, F.S., E. S. Zavaleta, V.T. Eviner, R. L. Naylor, P.M. Vitousek, H. L. Reynolds,

D.U. Hooper, S. Lavorel, O.E. Sala, S.E. Hobbie, M.C. Mack, and S. Diaz. 2000.

Consequences of changing biodiversity. Nature 405:234-242.

Clinton, B.D., L.R. Boring, and W.T. Swank. 1994. Regeneration patterns in canopy gaps

of mixed-oak forests of the southern Appalachians: influences of topographic

position and evergreen understory. American Midland Naturalist 132:308-319.

Cobb, R.C., D.A. Orwig, and S. Currie. 2006. Decomposition of green foliage in eastern

hemlock forests of southern New England impacted by hemlock woolly adelgid

infestations. Canadian Journal of Forest Research 36:1331-1341.

Constanza, R., R. d’Arge, R. de Groot, S. Farber, M. Grasso, B. Hannon, K. Limburg, S.

Naeem, R. V. O’Neill, J. Paruelo, R. G. Raskin, P. Sutton, and M. van den Belt.

1997. The value of the world’s ecosystem services and natural capital. Nature

387:253-260.

Cote, S.D. T. P. Rooney, J. Tremblay, C. Dussault, and D.M. Waller. 2004. Ecological

impacts of deer overabundance. Annual Review of Ecology, Evolution and

Systematics 35:113-147.

33

Daley, M.J., N. G. Phillips, C. Pettijohn, and J.L. Hadley. 2007. Water use by eastern

hemlock (Tsuga canadensis) and black birch (Betula lenta): implications of

effects of the hemlock woolly adelgid. Canadian Journal of Forest Research

37:2031-2040.

Dodds, K.J. and D.A. Orwig. 2011. An invasive urban forest pest invades natural

environments- Asian longhorned beetle in northeastern US hardwood forests.

Canadian Journal Forest Research 41:1729-1742.

Drever, C.R., G. Peterson, C. Messier, Y. Bergeron, and M. Flannigan. 2006. Can forest

management based on natural disturbances maintain ecological resilience?

Canadian Journal of Forest Research 36:2285-2299.

Ehrenfeld, J.G. 2011. Ecosystem consequences of biological invasions. Annual Review

of Ecology, Evolution, and Systematics 41:59-80.

Ellison, A. M., M. S. Banks, B. D. Clinton, E. A. Colburn, K. Elliott, C. R. Ford, D. R.

Foster, B. D. Kloeppel, J. D. Knoepp, G. M. Lovett, J. Mohan, D. A. Orwig, and

N. L. Rodenhouse. 2005. Loss of foundation species: consequences for the

structure and dynamics of forested ecosystems. Frontiers in Ecology and the

Environment 3:479-486.

Elmqvist, T., C. Folke, M. Nystrom, G. Peterson, J. Bengtsson, B. Walker, and J.

Norberg. 2003. Response diversity, ecosystem change, and resilience. Frontiers in

Ecology and the Environment 1:488-494.

Fei, S., N. Kong, K.C. Steiner, W.K. Moser, and E.B. Steiner. 2011. Change in oak

abundance in the eastern United States from 1980 to 2008. Forest Ecology and

Management 262:1370-1377.

34

Finzi, A.C., N. Van Breemen, and C. D. Canham. 1998. Canopy tree soil interactions

within temperate forests: species effects on soil carbon and nitrogen. Ecological

Applications 8:440-446.

Folke, C., S. Carpenter, B. Walker, M. Scheffer, T. Elmqvist, L. Gunerson, and C.S.

Holling. 2004. Regime shifts, resilience, and biodiversity in ecosystem

management. Annual Review of Ecology, Evolution, and Systematics 35:557-

581.

Ford, C.R. and J.M. Vose. 2007. Tsuga canadensis (L.) Carr. mortality will impact

hydrologic processes in southern Appalachian forest ecosystems. Ecological

Applications 17:1156-1167.

Ford, C.R., K.J. Elliott, B.D. Clinton, B.D. Kloeppel., J.M. Vose. 2012. Forest dynamics

following eastern hemlock mortality in the southern Appalachians. Oikos

121:523-536.

Foster, D.R., S. Clayden, D. A. Orwig, and B. Hall. 2002. Oak, chestnut and fire: climatic

and cultural controls of long-term forest dynamics in New England, USA. Journal

of Biogeography 29:1359-1379.

Fox, V.L., C. P. Buehler, C.M. Byers, S. E. Drake. 2010. Forest composition, leaf litter,

and songbird communities in oak-vs. maple-dominated forests in the eastern

eastern United States. Forest Ecology and Management 259:2426-2432.

Fralish, J.S. and T.G. McArdle. 2009. Forest dynamics across three centry-length

disturbance regimes in the Illinois Ozark Hills. American Midland Naturalist

162:418-449.

35

Franklin, J.F., and K.N. Johnson. 2011. Societal challenges in understanding and

responding to regime shifts in forest landscapes. Proceedings of the National

Academy of Sciences of the United States of America 108:16863-16864.

Gandhi, K.J.K. and D.A. Herms. 2010. Direct and indirect effects of alien insect

herbivores on ecological processes and interactions in forests of eastern North

America. Biological Invasions 12:389-405.

Gilliam, F.S., M.B. Adams, B.M. Yurish. 1996. Ecosystem nutrient responses to chronic

nitrogen inputs at Fernow Experimental Forest, West Virginia. Canadian Journal

of Forest Research 16:196-205.

Green, S.R., M.A. Arthur, B. A. Blakenship. 2010. Oak and red maple seedling survival

and growth following periodic prescribed fire on xeric ridgetops on the

Cumberland Plateau. Forest Ecology and Management 259:2256-2266.

Groffman, P.M, J.S. Baron, T. Blett, A.J. Gold, I. Goodman, L.H. Gunderson, B. M.

Levinson, M.A. Palmer, H. W. Paerl, G.D. Peterson, N. LeRoy Poff, D. W.

Rejeski, J. F. Reynolds, M. G. Turner, K.C. Weathers, and J. Wiens. 2006.

Ecological thresholds: the key to successful environmental management or an

important concept with no practical application? Ecosystems 9:1-13.

Gunderson, L.H. 2000. Ecological resilience- in theory and application. Annual Review

of Ecology and Systematics 31:425-439.

Hilt, S. J. Kohler, H. Kozerski, E.H. van Nes, M. Scheffer. 2010. Abrupt regime shifts in

space and time along rivers and connected lake systems. Oikos 120:766-775.

Holling, C.S. 1973. Resilience and stability of ecological systems. Annual Review of

Ecology and Systematics 4:1-23.

36

Hooper, D.U., F.S. Chapin, J.J. Ewel, A. Hector, P. Inchausti, S. Lavorel, J.H. Lawton,

D.M. Lodge, M. Loreau, S. Naeem, B. Schmid, H. Setala, A.J. Symstad, J.

Vendermeer, and D.A. Wardel. 2005. Effects of biodiversity on ecosystem

functioning: a consensus of current knowledge. Ecological Monographs 75:3-35.

Houseman, G.R., G.G. Mittelbach, H.L. Reynolds, and K.L. Gross. 2008. Perturbations

alter community convergence, divergence, and formation of multiple community

states. Ecology 89:2172-2180.

Huston, M.A. 1997. Hidden treatments in ecological experiments: reevaluating the

ecosystem function of biodiversity. Oecologia 110:449–460

Hutchinson, T.F., R. J. Boerner, S. Sutherland, E. K. Sutherland, M. Ortt, L. R. Iverson.

2005. Prescribed fire effects on the herbaceous layer of mixed-oak forests.

Canadian Journal of Forest Research 35:877-890.

Iverson, L.R., M.E. Dale, C.T. Scott, A. Prasad. 1997. A GIS-derived integrated moisture

index to predict forest composition and productivity of Ohio forests (USA).

Landscape Ecology 12:331-348.

Jenkins, J.C., J.D. Aber, J.D. and C.D. Canham. 1999. Hemlock woolly adelgid impacts

on community structure and N cycling rates in eastern hemlock forests. Canadian

Journal of Forest Research 29:630-645.

Kashian, D.M. and J.A. Wittier. 2011. Assessing the potential for ash canopy tree

replacement via current regeneration following emerald ash borer-caused

mortality on southeastern Michigan landscapes. Forest Ecology and Management

261:480-488.

37

Kizlinski, M.L, D.A. Orwig, R.C. Cobb and D.R. Foster. 2002. Direct and indirect

ecosystem consequences of an invasive past on forests dominated by eastern

hemlock. Journal of Biography 29:1489-1503.

Knoepp, J.D., J.M. Vose, B. D. Clinton, N. D. Hunter.. 2011. Hemlock infestation and

mortality: impacts on nutrient pools and cycling in Appalachian forests. Soil

Science Society of America Journal 75:1935-1945.

Kominoski, J.S., C.M. Pringle, B.A. Ball, M.A. Bradford, D.C. Coleman, D.B. Hall, and

M.D. Hunter. 2007. Nonadditive effects of leaf litter species diversity on

breakdown dynamics in a detritus-based stream. Ecology 88:1167-1176.

Likens, G.E., C.T. Driscoll, and D. C Buso. 1996. Long-term effects of acid rain:

response and recovery of a forest ecosystem. Science 272:244-246.

Lindenmayer, D.B., R.J. Hobbs, G.E. Likens, C.J. Krebs, S.C. Banks. 2011. Newly

discovered landscape traps produce regime shifts in wet forests. Proceedings of

the National Academy of Sciences of the United States of America 108:15887-

15891.

Lovett, G.M., K.C. Weathers, M.A. Arthur and J.C. Schultz. 2004. Nitrogen cycling in a

northern hardwood forest: do species matter? Biogeochemistry 67:289-308.

Lovett, G.M., C.D. Canham, M.A. Arthur, K.C. Weathers, and R.D. Fitzhugh. 2006.

Forest ecosystem responses to exotic pests and pathogens in eastern North

America. BioScience 56:395- 405.

Martin, K.L. and L.K. Kirkman. 2009. Management of ecological thresholds to re-

establish disturbance-maintained herbaceous wetlands of the south-eastern USA.

Journal of Applied Ecology 46:906-914.

38

Martin, K.L., D.M. Hix, and P.C. Goebel. 2011. Coupling of vegetation layers and

environmental influences in a mature, second-growth Central Hardwood forest

landscape. Forest Ecology Management 261:720-729.

Mayer, A.L. and M. Rietkerk, 2004. The dynamic regime concept for ecosystem

management and restoration. BioScience 54:1013- 1020.

McEwan, R.W. J.M. Dyer and N. Pederson. 2011. Multiple interacting ecosystem

drivers: toward an encompassing hypothesis of oak forest dynamics across eastern

North America. Ecography 34:244-256.

Minchin, P.R. 1987. An evaluation of the relative robustness of techniques for ecological

ordination. Vegetatio 69:89-107.

Nowacki , G.J. and M.D. Abrams. 2008. The demise of fire and “mesophication” of

forests in the eastern United States. BioScience 58:123-138.

Nuckolls, A, N. Wurzburger, C.R. Ford, R. Hendrick, J. M. Vose, and B. Kloeppel. 2008.

Hemlock Declines Rapidly with Hemlock Woolly Adelgid Infestation: Impacts on

the Carbon Cycle of Southern Appalachian Forests. Ecosystems 12:179-190.

Orwig, D. A. and D. R. Foster. 1998. Forest response to the introduced hemlock woolly

adelgid in southern New England, USA.. Journal of the Torrey Botanical Society

125:60–73.

Orwig, D.A, D.R. Foster, and D.L. Mausel. 2002. Landscape patterns of hemlock decline

in New England due to the introduced hemlock woolly adelgid. Journal of

Biogeography. 29: 1475-1487.

39

Orwig, D.A., R.C. Cobb, A.W. D’Amato, M.L. Kizlinski, and D.R. Foster. 2008. Multi-

year ecosystem response to hemlock woolly adelgid infestation in southern New

England forests. Canadian Journal of Forest Research 38:834-843.

Paine, R.T., M.J. Tegner, and E.A. Johnson. 1998. Compounded perturbations yield

ecological surprises. Ecosystems 1:534-545.

Peterson, G., C.R. Allen, and C.S. Holling. 1998. Ecological resilience, biodiversity, and

scale. Ecosystems 1:6-18.

Piatek, K.B., P. Munasinghe, W.T. Peterjohn, M.B. Adams, and J.R. Cumming. 2010. A

Decrease in oak litter mass changes nutrient dynamics in the litter layer of a

Central Hardwood Forest. Northern Journal of Applied Forestry:97-104.

Pierce, A.R., G. Parker, and K. Rabenold. 2006. Forest succession in an oak-hickory

dominated stand during a 40-year period at the Ross Biological Reserve, Indiana.

Natural Areas Journal 26:351-359.

Poland , T.M.and D.G. McCullough, D.G. Emerald ash borer: invasion of the urban

forest and the threat to North America’s ash resource. Journal of Forestry 104:

118-124.

Sasaki, T. and Lauenroth, W.K. 2011. Dominant species, rather than diversity, regulates

temporal stability of plant communities. Oecologia 166:761-768.

Scheffer, M., S. Carpenter, J.A. Foley, C. Folke, and B. Walker. 2001. Catastrophic shifts

in ecosystems. Nature 413:591-596.

Scheffer, M., S.H. Hosper, M-L. Meijer, B. Moss, E. Jeppensen. 1993. Alternative

equilibia in shallow lakes. Trends in Ecology Evolution 8:275-279.

40

Schröder, A., L. Persson, and A.M. De Roos. 2005. Direct experimental evidence for

alternative stable states: a review. Oikos 110:3-19.

Small, M.J, C.J. Small, and G.D. Dreyer. 2005. Changes in a hemlock-dominated forest

following woolly adelgid infestation in southern New England. Journal of the

Torrey Botanical Society 132:458-470.

Snyder C.D., J.A. Young and D. Smith. 2002. Influence of eastern hemlock (Tsuga

canadensis) forests on aquatic invertebrate assemblages in headwater streams.

Canadian Journal of Fisheries and Aquatic Science 59:262–75.

Spaulding, H.L. and L.K. Rieske. 2010. The aftermath of an invasion: Structure and

composition of Central Appalachian hemlock forests following establishment of

hemlock woolly adelgdi, Adelges tsugae. Biological Invasions 12:3135- 3143.

Stadler, B., T. Müller and D.A. Orwig. 2006. The ecology of energy and nutrient fluxes

in hemlock forests invaded by hemlock woolly adelgid. Ecology 87:1792-1804.

Suding, K.N. and R. J. Hobbs. 2009. Threshold models in restoration and conservation: a

developing framework. Trends in Ecology and Evolution 24:271-279.

Suding, K.N., K.L. Gross, and G.R. Houseman. 2004. Alternate states and positive

feedbacks in restoration ecology. –Trends in Ecology and Evolution 19:46-53.

Sutherland, E.K., B.J. Hale, and D.M. Hix, 2000. Defining species guilds in the Central

Hardwood Forest, USA. Plant Ecology:1-19.

Thrush, S.F., J.E. Hewitt, P.K. Dayton, G. Coco, A.M. Lohrer, A. Norkko, J. Norkko, and

M. Chiantore. 2010. Forecasting the limits of resilience: integrating emphirical

research with theory. Proceedings of the Royal Society B. 276:3209-3217.

41

Tingley, M.W., D.A. Orwig, R. Field and G. Motzkin. 2002. Avian response to removal

of a forest dominant: consequences of hemlock woolly adelgid infestations.

Journal of Biogeography 29:1505-1516.

Vitousek , P.M., J.D. Aber, R.W. Howarth, G.E. Likens, P.A. Matson, D.W. Schindler,

W. H. Schlesinger, and D. G. Tilman. 1997. Human alteration of the global

nitrogen cycles: sources and consequences. Ecological Applications 7:737-750.

Waldrop, T.A. D.A. Yaussy, R.J. Phillips, T.A. Huchinson, L. Brudnak, and R.J.

Boerner. 2007. Fuel reduction treatments affect stand structure of hardwood

forests in western North Carolina and southern Ohio, USA. Forest Ecology and

Management 255:3117-3129.

Walker, B., A. Kinzig, and J. Langridge. 1999. Plant attribute diversity, resilience, and

ecosystem function: the nature and significance of dominant and minor species.

Ecosystem 2:95-113.

Wurzburger, N. and R.L. Hendrick. 2009. Plant litter chemistry and mycorrhizal roots

promote a nitrogen feedback in a temperate forest. Journal of Ecology 97: 528-

536.

Young, R.F, K.S. Shields, and G.P.Berlyn. 1995. Hemlock woolly adelgid (Homoptera:

Adelgidae): stylet bundle insertion and feeding sites. Annals of the Entomological

Society of America 88:827- 835.

42

Figure 2. 1 Generalized model of alternate state dynamics. State changes, also referred to as fast changes, are caused by a movement of the ball (state) out of a valley and across a threshold. Parameter or slow changes result from a change in the shape of the surface.

Resilience is determined by the shape of the valleys, which represents feedbacks and

functions that maintain a community state.

43

Figure 2.2 The transition from oak to mesic maple forests. While this state transition and the general feedbacks are understood, the details of the threshold are not well documented.

44

Figure 2. 3 Resilience is decreasing in eastern deciduous forests due to the loss of ash.

While the state variable remains within the valley, the valley shrinks and thresholds are lowered. This represents a change in the slow variables or parameters.

45

Figure 2. 4 Hemlock forests are low resilience and the loss of hemlock results in a state change. The resulting state and parameters are determined by the local and regional environment and in some cases, the presence of species including Rhododendron.

46

Chapter 3: The foundation species influence of Tsuga canadensis (eastern hemlock) on biodiversity and ecosystem function on the Unglaciated Allegheny Plateau

3.1 Abstract

What is the role of foundation species in community processes and what are the

relative strengths of their direct and indirect influences on diversity, productivity, and resource availability? Ecosystems structured by foundation species, abundant species that define the ecosystem processes of a community, are thought to depend on a small number of strong interactions. Yet, how this translates to ecosystem composition and function is not well understood. Opportunities to quantify the influence of Tsuga

canadensis on microclimate, ecosystem composition, and functional processes are limited

as an increasing portion of its range is impacted by an invasive pest insect, Adelges

tsugae. Prior to invasion, we tested the foundational role of Tsuga canadensis in forests

on the Unglaciated Allegheny Plateau. We explored eight Tsuga canadensis dominated

ravine forests on the Unglaciated Allegheny Plateau of Ohio, USA. In transects at 10, 30

and 50 meters from headwater streams, we measured all vegetation strata in a series of

five 100-m2 plots and recorded the physiographic context. Light availability was quantified using hemispherical photography during the growing season and the deciduous

leaf-off period. We also measured soil properties and leaf litter biomass and chemistry as

metrics of nutrient cycling and productivity, and determined the relative decomposition

rate using cellulose paper.

47

Comparisons across transects indicated a high degree of similarity. Species

richness was low, with slight increases moving upslope from the streams. Productivity

(leaf litter biomass), light availability (canopy openness) in the growing season and

deciduous leaf-off period, and nutrient cycling (decomposition and leaf litter chemistry)

were also similar across transects. Non-metric multidimensional scaling analyses

indicated differences in species composition near the stream compared to 30 and 50

meters away. Structural equation modeling (SEM) indicated Tsuga canadensis

dominance has a strong negative influence on vegetation species richness, as well as light

availability and productivity.

As a foundation species, Tsuga canadensis dampens environmental gradients and

overwhelms local differences in species composition. However, ecosystem response will

likely differ near streams and upslope when Tsuga canadensis dominance declines and is

removed by HWA. Whether other foundation species, particularly conifers, structure

ecosystems through similar mechanisms merits further investigation.

Keywords: Alternate states; Adelges tsugae; Central Hardwood Forest; Hemlock woolly

adelgid; resilience; Structural equation modeling

3.2 Introduction

Ecological literature frequently refers to foundation species, common or abundant species that are thought to define the microclimate and processes of a community (Ellison et al. 2005), but examples of the mechanisms and strengths of foundation species influence are uncommon. Unlike keystone species, which are often predators that exert

48

top-down control, foundation species are primary producers and are likely to be trees in

forested ecosystems (Ellison et al. 2005). A recent review by Ellison et al. (2005)

indicates that foundation species are varied and include Castanea dentata and Quercus

species that dominate diverse deciduous forests, but also evergreen conifers including

Tsuga canadensis and Pinus albicaulis that are frequently monodominant. From these

examples, it is unclear whether foundation species are competitive dominants that

exclude other species, facilitate them by creating locally stable conditions, or more likely,

a more complex relationship (Bruno et al. 2003). For example, in harsh environments

such as desert ecosystems, nurse plants facilitate other species by creating small patches

of increased soil moisture and nutrients, and lower soil temperature; however, they also

cause shading (Bruno et al. 2003). Some species categorized as foundational may be

responsible for the sampling effect in biodiversity-ecosystem function theory, where one

dominant species is responsible for increased productivity in diverse communities

(Huston 1997; Bruno et al. 2003), but this may not always be the case. Rather, in some

cases, foundation species may result in lower diversity, but increased local stability

(Sasaki & Laurenroth 2011). In an era of rapid ecological change, perhaps the most

important prediction of foundation species theory indicates they define a specific community state structured by a small number of strong interactions; therefore, the loss

of foundation species destabilizes energy and nutrient dynamics, and the community will

rapidly shift to an alternate state with distinct properties (Ellison et al. 2005). The

widespread mortality of Tsuga canadensis due to Adelges tsugae (hemlock woolly

adelgid, HWA) provides a model system to explore fundamental questions of the role of

foundation species in community interactions and processes.

49

Tsuga canadensis communities are unique evergreen ecosystems within the largely deciduous forest region of eastern North America. Characterized by a damp, shady microclimate and slowly decomposing, acidic litter (Ellison et al. 2005), transpiration in Tsuga canadensis forests is distributed more evenly throughout the year, with higher rates in the spring than in deciduous hardwood forests (Ford &Vose 2007).

Tsuga canadensis stands support unique suites of species, including salamanders, cold water fish (Snyder et al. 2002), birds (Tingley et al. 2002), and aquatic macroinvertebrates (Ross et al. 2003), contributing to landscape or beta and gamma diversity. At the same time, Tsuga canadensis communities exhibit low plant species richness or alpha diversity (Ellison et al. 2005). For the plants species that have adapted to the low light and nutrient availability, Tsuga canadensis may provide valuable regeneration sites with a level of protection from deer browse (Krueger & Peterson

2006). Beyond the terrestrial system, streams in Tsuga canadensis dominated riparian areas tend to have more large wood jams, and because of the slow decay rate of Tsuga canadensis, many logs remain in the large wood pool much longer than other associated hardwoods, with the exception being the now functionally extinct Castanea dentata

(Wallace et al. 2001; Morris et al. 2007).

While the full impact of Tsuga canadensis mortality due to HWA is not yet clear, evidence supports a rapid transition to alternate states, as functional changes are evident almost as soon as HWA arrives. From New England to the southern Appalachian

Mountains, HWA invasion alters ecosystem energy exchanges, nutrient cycling, and hydrology beginning at initial stages of decline (Jenkins et al. 1999; Cobb et al. 2006;

Stadler et al. 2006; Daley et al. 2007; Ford &Vose 2007; Orwig et al. 2008; Nuckolls et

50

al. 2008). These processes likely continue to fluctuate as Tsuga canadensis declines, dies, and a new vegetation community develops. Eventually, a new community state characterized by distinct feedbacks between composition and functional pathways should develop. A more complete understanding of the influence of Tsuga canadensis prior to decline will enhance knowledge of this transition and allow more accurate projections of future forest development. In particular, in areas outside of the invasion front where forests remain healthy, there is a window of opportunity to test and expand the largely conceptual understanding of the role of Tsuga canadensis as a foundation species.

The loss of a foundation species is not a new phenomenon for the forests of eastern

North America, as these forests experienced a significant foundation species loss at the

turn of the twentieth century with the introduction of the pathogen Cryphonectria

parasitica. Prior to introduction of Cryphonectria parasitica, Castanea dentata

represented approximately 40-50% of the basal area of eastern forests, which served as an

important species for wildlife habitat and was valued as a timber species (Keever 1953:

Anagnostakis 1987). As Castanea dentata declined, many upland areas transitioned to

Quercus-Carya forest (Keever 1953), while Tsuga canadensis dominance increased in

riparian areas (Day &Monk 1974; Elliott et al. 1999; Elliott & Swank 2008). Castanea dentata is a slow decaying species that likely had a significant impact on ecosystem dynamics, and remains an important component of biogeochemical processes for decades beyond its lifespan (Wallace et al. 2001; Rhoades 2006). Unfortunately, there are very few baseline data on the structure and function of Castanea dentata dominated forests prior to the blight (Anagnostakis 1987), and its prevalence in riparian areas, now dominated by Tsuga canadensis, has only been recognized more recently (Vandermast &

51

Van Lear 2002). As the changes in ecosystem function following the loss of Castanea dentata are not well understood (Keever 1953; Elliott & Swank 2008), details of the functional role Castanea dentata played across eastern deciduous forests and the transitions that occurred following its loss will largely remain speculative. This gap in knowledge underscores the importance of quantifying the ecosystem dynamics associated with Tsuga canadensis dominated forests while some remain intact (Yorks et al. 2002).

We investigated the influence of Tsuga canadensis on the forest community

composition and function in healthy Tsuga canadensis forests on the unglaciated

Allegheny Plateau of southeastern Ohio. Our objective was to develop a more quantitative understanding of Tsuga canadensis as a foundation species and establish

baselines of ecosystem function in the Tsuga canadensis dominated community state

prior to HWA arrival. Our central hypothesis was that Tsuga canadensis dampens the influence of physiographic gradients and couples community composition and ecosystem function, including light availability, leaf litter biomass and chemistry, and decomposition rates across local environmental gradients within Tsuga canadensis

dominated ravines.

3.3 Methods

Study Sites

Study sites were located within the Unglaciated Allegheny Plateau physiographic

province of southeastern Ohio (Brockman 1998). All sites were selected within ravine

riparian areas dominated by second-growth forests with little evidence of recent human

disturbance and a dominance of Tsuga canadensis (approximately 50 percent or greater

of the total overstory basal area). The Unglaciated Allegheny Plateau physiographic

52

province is characterized by sandstone bedrock that forms deep valleys and cliffs, and the

majority of Tsuga canadensis stands within Ohio are found in this part of the Ironton and

Shawnee-Mississippian Plateau physiographic subsections (Black & Mack 1976;

Brockman 1998; Prasad et al. 2007). Three of our Tsuga canadensis dominated ravines were located in Lake Katharine State Nature Preserve, an 817-ha preserve in Jackson

County. The remaining five ravines were located at in Hocking County at Sheick Hollow

State Nature Preserve and Hocking State Forest. Sheick Hollow is a 61-ha preserve adjacent to the 3,924-ha Hocking State Forest. Hocking and Jackson Counties have continental climates with cold, snowy winters with an average temperature of 0°C and warm, humid summers averaging 21.6°C, with an average precipitation of 1020 mm. evenly distributed throughout the year (Kerr 1985; Lemaster and Gilmore 1989).

Hocking County is characterized by Mississippian and Pennsylvanian sedimentary bedrock. In the Mississippian system, the Logan formation of sandstone, shale and conglomerate, overlays Cuyahoga Shale and Blackhand Sandstone (Lemaster & Gilmore

1989). The bedrock of Jackson County is similarly porous, formed by Sharon conglomerate and Pottsville sandstone (Beatley 1959; Runkle & Whitney 1987). Soils in the upland and slope portions of Lake Katharine State Nature Preserve are well-drained

Hapludults, mainly the Clymer silt loam formed from sandstone residuum and Rigley sandy loam formed by colluvium at the base of slopes. Orville fluvaquents occur in the floodplains of some small coves (Kerr 1985; Runkle & Whitney 1987). Lemaster and

Gilmore (1989) describe the predominant soils of Hocking State Forest as part of the

Cedarfalls-rock outcrop complex with 40-70 percent slopes and steep, well-drained soil.

53

Sampling design

At each study site, we established three transects parallel to the stream at

distances of 10, 30 and 50 meters from the bank-full channel. The side of the ravine

chosen for sampling was determined based on logistical considerations due to the

location of cliffs, evidence of disturbance, and uniformity of landform. In each transect, we used a series of five 100-m2 circular plots to characterize local differences in

environmental conditions and community composition and function. Within each

circular plot, we recorded basic physiographical data including slope percent using a

clinometer, slope shape (covex, concave, or linear), slope position (summit, hillslope, flood plain, structural bench) and aspect. We calculated metrics of heat load and direct incident radiation for each plot using latitude, slope percent, and aspect in Equation 1 from McCune & Keon (2002). At three randomly chosen sites, we measured the microclimate using HOBO© Pro v2 temperature and humidity data loggers (Bourne,

MA, USA) installed in the center plot of each transect. To ensure the security of the

loggers, they were attached to the base (approx. 10 cm above the forest floor) of the tree

closest to the center of the plot on the steam side of the trunk. Temperature and humidity

were recorded every 15 minutes throughout the growing season (June- November).

Soil samples were collected from near the center of a random plot in each

transect. During collection, the organic matter was moved aside, and three 5-cm diameter

cores were taken to a depth of five cm were collected. Samples were dried at 40°C and

sieved through a two mm screen prior to analysis. Samples were then analyzed for texture

using the hydrometer method (American Society for Testing and Materials 1985), and pH

and CEC determined using an Orion pH meter (ThermoFisher Scientific, Waltham, MA,

54

USA) and ammonium acetate extraction, respectively. Total carbon and nitrogen was

determined using an Elementar Vario-Max CN analyzer, while major elements

(phosphorus, potassium, calcium and magnesium) were determined using digestion with

HClO4/HNO3 followed by inductively coupled plasma (ICP) emission spectrometry using a Teledyne Leeman Labs Prodicgy Dual view ICP. All soils analyses were conducted by

the Ohio Agricultural Research and Development Center (OARDC) Service Testing and

Research (STAR) lab.

Vegetation community composition was measured in the circular plots within the

three transects. All species and dbh of the woody vegetation (all stems >2.5 cm dbh)

were recorded. Stems between 2.5 cm and 10 cm dbh were classified as saplings, while

those > 10 cm were classified as overstory. At the center of each plot, we established a 4-

m2 sub-plot and recorded all ground flora and seedlings. Ground-flora species (herbs and

woody species < 1.37 m tall that have no overstory potential) were identified and assigned an estimated percent cover on a seven-point scale (1: <1 %, 2: 1-5%, 3: 5-15%,

4: 15-25%, 5: 25-50%, 6: 50-75%, 7: > 75%). Bare ground without live vegetation was also assigned a percent cover value. Seedlings (tree and shrub species < 2.5 cm dbh) were identified, counted, and categorized as large (>1.37 m tall) or small (<1.37 m tall).

Vegetation data from the plots and subplots in each transect were combined for data analyses; environmental metrics were averaged across plots of each transect.

Functional metrics

We examined metrics of light, productivity, and nutrient cycling to examine the role of Tsuga canadensis in ecosystem function. Estimates of light availability were

55

determined by hemispherical photographs taken 1.2-m above the ground in the center of

each 100-m2 plot using a Nikon COOLPIX E8400 (Nikon Corporation, Tokyo, Japan )

equipped with a fisheye lens (Nikon LC-ER2) and mounted on Self-leveling O-Mount

(Regent Instruments, Canada) during the growing season (June – September) and

deciduous leaf-off periods (December – March). Canopy openness was determined by

analyzing photographs of each plot using standard protocols and WinSCanopy digital

image processing software (Regents Instruments, Canada). This methodology excluded

obstructions in photos created by cliffs or boulders. Canopy openness was averaged

across plots to provide a single value per transect.

As a metric of productivity and nutrient cycling, leaf litter fall was collected using plastic litter traps located just outside the central 4-m2 subplot of a randomly selected

100-m2 plot within each transect (24 traps total, 3 per site). Leaf litter traps were

constructed from an approximately 35 L (0.035 m3, 1 US bushel) plastic laundry basket lined with nylon mesh and secured to the ground using a metal stake to prevent shifting.

Litter was collected during fall leaf-off (September- December), dried at 40°C to constant mass, sorted by species and weighed for biomass estimates. Species in the leaf litter were identified to the finest taxonomic resolution, which was generally to species, but Quercus

rubra and Q. coccinea were grouped. Litter carbon and nitrogen content were determined by combustion at the OARDC STAR Laboratory using an Elementar Vario-Max CN analyzer (International Standard ISO 10694 1995; AOAC Official Methods of Analysis

2002).

To establish a standard metric of nutrient cycling, relative decomposition rates

were determined using cellulose paper (Kizlinski et al. 2002). Four replicates made of

56

three grams of cellulose paper wrapped in 1-mm2 nylon mesh were placed at ground level

in each transect, secured to the litter traps. One of the four packs was randomly collected

at nine weeks and all remaining packs were collected at the end of the growing season

(28 weeks). Upon collection, all packs were dried at 40°C until constant mass and

weighed. Mass loss was averaged and decomposition rates were compared using the

decay constant k, calculated from the decay equation: ln (initial mass- end mass)= -kt

where t= number of days).

Statistical Analyses

We first identified local-scale differences in environmental metrics, community

composition and function by comparing metrics among transects. Statistical analyses were performed using R version 2.10.0 (online documentation available at http://www.r-

project.org/). Prior to analyses, we ensured that each variable fit the model assumptions,

using a Shapiro-Wilk test for normality and Levene’s test for homogeneity of variance.

Skewness and kurtosis were examined with the D’Agostino and Bonett-Seier tests

respectively. In cases where assumptions were not met, data were transformed or a non-

parametric test was substituted. Significant results were examined in detail with Tukey’s

honestly significant difference.

Environmental metrics, including soil and microclimate characteristics were

analyzed using analysis of variance (ANOVA) for a split plot design with site as the whole plot factor and transect as the within subject factor. Soil characteristics included pH, cation exchange capacity (CEC), carbon to nitrogen ratio, and major elements, including phosphorus, potassium, calcium, and magnesium. As soils were all quite sandy

57

(with >60% sand particle size), percent sand was used as an indicator of texture. Each characteristic was compared by distance to the stream (10, 30, and 50 m). For the microclimate, we compared maximum temperature, minimum temperature, and minimum relative humidity by distance to the stream.

Several analyses were used to examine differences in species composition and diversity. To test for differences in species composition among transects, we compared the vegetation using multiple response permutation procedure (MRPP) with the Bray-

Curtis dissimilarity index. Diversity in vegetation layers was examined by both species and functional groups. Functional groups for woody species were defined by Sutherland et al. (2000), based on characteristics including germination requirements and growth rates (Table 1). However, due to its foundational role and evergreen habit, we defined

Tsuga canadensis as its own functional group. Overstory analyses were based on relative basal area by species. As there were few individuals, frequency (stem counts) were used for analyses of the sapling and seedling layers. For the ground flora, species abundance was calculated by adding the cover represented by the midpoint of the cover class on each of the five subplots to create one cover value per transect. We tested for differences in species richness, diversity, and functional metrics by distance from the stream (10, 30, and 50 m) using a split- plot ANOVA. Differences in diversity were examined using the response variables of species richness (S) and Shannon’s diversity (H’) by distance to the stream. These metrics were calculated for all vegetation layers (overstory, sapling and ground flora) and leaf litter. Finally, functional metrics, including light availability

(canopy openness), leaf litter biomass, and leaf litter quality (carbon to nitrogen ratio),

58

and relative decomposition constant (k) were also compared by distance from the stream

using plot- plot ANOVA.

We used multivariate techniques to examine the environmental gradients

structuring Tsuga canadensis ravine communities in Ohio. Vegetation layers were

subjected to non-metric multidimensional scaling analysis (NMDS) to compare species

occurrence across environmental gradients. For the purposes of analysis, species appearing on fewer than five percent of subplots (6 subplots out of a total of 120) were excluded. NMDS is an unconstrained ordination method regarded as highly robust for community data (Minchin 1987). We used the metaMDS procedure with a Bray-Curtis

distance matrix using the vegan package for community ecology (Dixon & Palmer 2003,

http://finzi.psych.upenn.edu/R/library/vegan/html/metaMDS.html). The metaMDS procedure automatically transforms data with large ranges using square root and

Wisconsin double standardization. In addition, the metaMDS procedure uses an iterative approach with random starts to insure against entrapment by local minima while converging on a solution that minimizes stress. The final solution is rotated so that the first axis explains the greatest variance. To examine associations with specific factors,

environmental vectors determined by principal components analysis were overlaid on the

community NMDS diagram using the envfit procedure, also part of the vegan package

(Dixon & Palmer 2003, http://finzi.psych.upenn.edu/R/library/vegan/html/envfit.html).

We repeated this procedure using PCA vectors of functional metrics to explore patterns of community and function. We allowed NMDS iterations to continue until the convergence on a global solution. Categorical variables including slope shape and slope position were coded as dummy variables.

59

Finally, to quantify the strength of the relationships between Tsuga canadensis, environmental gradients and vegetation composition and function, we used structural equation modeling (SEM) in AMOS version19 (SPSS Inc, Chicago, IL). SEM uses maximum likelihood to solve path equations simultaneously and allows closer inspection of indirect effects, interactions, and reciprocal relationships between variables (Grace

2006, 2008). For the purposes of SEM, we used the NMDS axis scores to quantify the vegetation community composition of each layer; in this way we reduced the dimensionality of the species composition data into a single response variable for the

SEM (Grace 2006, 2008). We used patterns in the data from previous analyses to test and build candidate SEM models. Models were considered an acceptable fit when the chi square fit statistic indicated the model was not significantly different than the data (Χ2 p

> 0.05), and model pathways were significant (p < 0.05). Grace (2008) suggests 5-20 observations per model parameter, thus we constructed multiple smaller models for parts of the system as relating all variables was not feasible with our dataset.

3.4 Results

Microclimate and soil characteristics

On a local scale moving from the stream to 50 meters upslope, the Tsuga canadensis ravines in our study exhibit similar microclimates and soil characteristics. We found no differences in maximum (Table 2, F= 0.12, df= 2, p= 0.89) or minimum (F=

0.49, df= 2, p= 0.64) temperature. The maximum relative humidity reached 100% during periods of rain, and the minimum humidity was similar at all slope positions (Table 2, F=

0.29, df= 2, p= 0.76). Soils were sandy across the ravines (Table 2, F= 2.47, df= 2, p=

60

0.12). Soil cation exchange capacity was slightly higher at 30-m than either 10-m or 50-m

from the stream (Table 2, F= 3.69, df=2, p= 0.05), and soil pH was slightly lower at 30-m than in either 10-m or 50-m from the stream (Table 2, non-normal distribution, Kruskal-

Wallis Χ2= 9.72, df= 2, p = 0.01).

Species composition and diversity

Throughout all vegetation layers, Tsuga canadensis was the dominant species.

Tsuga canadensis composed approximately half of the overstory basal area at all

distances from the stream, with only two other species, Lirodendron tulipifera and Fagus

grandifolia exceeding ten percent of the basal area (Table 3). Saplings were sparse and

the majority of those encountered were Tsuga canadensis (Table 4). Small Acer rubrum

seedlings were abundant, but there were few additional seedlings (Table 4). The ground

flora was likewise sparse, with bare ground exceeding 80% (Table 5). MRPP indicated

similarity in the species composition of the overstory (A= 0.03, n=8, p= 0.09) and the

sapling layer (A= -0.03, n=8, p= 0.85), regardless of distance from the stream.

Species richness and diversity were low in all vegetation layers across transects.

In the overstory, species richness (±SE) was only 4.75 ± 0.37 adjacent to the stream, with

a slight increase moving upslope (Fig. 3.1; F= 3.64, df=2, p= 0.05); however, diversity

exhibited an insignificant decrease moving upslope (Fig. 3.2; F= 1.24, df= 2, p= 0.32.

The sapling layer had mean species richness between 1.62 ± 0.18 and 2.88 ± 0.67 across

the ravines (Fig.3.1; F= 1.68 df=2, p= 0.22) and diversity was also similar (Figure 1b;

Kruskal-Wallis Χ2= 1.04, df= 3, p= 0.79). In the ground flora, we found similar mean

species richness (Fig 3.1; F=0.90, df=2, p= 0.43) and diversity (Fig. 3.2; F=0.32 df=2, p=

61

0.73). Functional richness was likewise low, but there was a slight increase in the

overstory functional richness moving upslope (Fig. 3.1; F= 9.04, df=2, p= 0.02). The

sapling and seedling layers contained so few species we did not analyze the functional

groups. The similarities in species richness across transects were reflected in the leaf litter

(Fig. 3.1) as there was an insignificant increase in leaf litter richness across transects (Fig

3.1; F= 1.26, p= 0.30).

Functional metrics

Ecosystem function was similar across Tsuga canadensis dominated ravines irrespective of distance from the stream (Table 1). All variables met statistical assumptions with the exception of leaf litter carbon to nitrogen ratio, which was analyzed with a non-parametric Kruskal-Wallis test. Light availability, as measured by canopy openness, did not differ across the transects (Table 2) in either the growing season (F=

1.28, df=2, p= 0.30) or deciduous leaf-off period (F=0.05, df=2, p= 0.95). The additional functional metrics (Table 2) were also similar across transects including leaf litter biomass (F= 0.58, df=2, p= 0.56), leaf carbon to nitrogen ratio (F= 2.50, df=2, p= 0.11) and decomposition constant k (F= 2.00, df=2, p= 0.18).

Gradient analyses

Ordination analyses indicated patterns of species occurrence along environmental gradients associated with the Tsuga canadensis dominated ravines. In our analysis of the overstory, NMDS axis one was associated with slope shape, slope position, heat load, and the 30-m transect (Fig. 3.3). The second axis was more strongly associated with the 10-m

62

and 50-m transects, slope percent, and incident radiation. The ordination diagram places

Tsuga canadensis near the origin of the environmental gradients, suggesting a ubiquitous

distribution across transects. Lower slope positions are characterized by species such as

Betula lenta, Fagus grandifolia, and Liriodendron tulipifera. Moving upslope, species

including Acer saccharum, Acer rubrum, and Quercus alba become more abundant, with

Q. rubra and Q. prinus on summits.

As species occurred infrequently in the sapling layer, and a number of sample

transects did not have any saplings that were not Tsuga canadensis, we completed an

NMDS with all species (including rare) to examine possible environmental gradients

(Fig. 3.4). Species including Oxydendrum arboreum, Magnolia tripetala, Nyssa sylvatica,

Acer rubrum, and Juglans nigra were associated with convex slopes, associated with the

first axis. As in the overstory layer, Tsuga canadensis’s ubiquity was apparent from its

position near the origin. Ulmus rubra, Carpinus caroliniana, and Acer saccharum

grouped together on summits and convex slopes. The short environmental vectors

indicate sapling distribution may be driven by a factor not included in the analysis, such

as the canopy composition, or there were too few saplings to indicate a clear pattern.

An NMDS of the overstory composition with PCA vectors of functional metrics

identified patterns in functional metrics (Fig. 3.5). In the analysis, leaf litter biomass,

growing season canopy openness, and soil carbon to nitrogen ratio were significantly

associated with the first axis (p< 0.05). The second axis was associated with the

decomposition rate (k) (p<0.05), and was weakly associated with soil pH (p= 0.13).

Species including Liriodendron tulipifera and Fagus grandifolia were associated with

higher soil pH and greater percent sand in the soil, Betula lenta was associated with

63

higher decomposition rate (k), which grouped opposite higher leaf carbon to nitrogen

ratio, which was associated with Quercus prinus and Q. rubra along the first axis.

Quercus alba and Acer rubrum were associated with increased growing season canopy

openness, while Nyssa sylvatica was associated with higher leaf litter biomass and higher

soil carbon to nitrogen ratio.

Models of Tsuga canadensis influence

To explore the influence of Tsuga canadensis on the vegetation community, we

constructed a model of Tsuga canadensis dominance and metrics of diversity in the

sapling layer, ground flora, leaf litter, and overstory functional groups (Fig. 3.6, Χ2= 10.1, df= 6, p= 0.12). Our model indicates Tsuga canadensis dominance (as expressed by relative basal area) as an influential factor, negatively associated with species richness throughout the vegetation community. Tsuga canadensis dominance was most negatively influential on sapling layer richness (-0.54) and leaf litter richness (-0.51), and explained nearly half of the variance for both (R2= 0.51, 0.49 for saplings and leaf litter

respectively). Tsuga canadensis was also quite influential in regulating ground flora

richness (-0.46, R2= 0.44) and overstory functional richness (-0.41, R2=0.39).

Taking a more functional approach, we constructed a second model of the

influence of Tsuga canadensis dominance on metrics of 1) light availability, measured by

canopy openness, and 2) productivity, measured by leaf litter biomass (X2 = 3.1, df= 2, p

= 0.21, Fig. 3.7). Our model again identified Tsuga canadensis as negatively associated with growing season canopy openness (-0.41, R2= 0.39) and leaf litter biomass (-0.36).

We found an indirect influence of Tsuga canadensis dominance on sapling community

64

composition (-0.23), structured through a direct influence of growing season canopy

openness (0.56, total sapling community R2= 0.53). Most of the variance (R2= 0.76) in

leaf litter biomass was influence by a negative association with Tsuga canadensis

dominance (-0.36) and a positive association with sapling community composition (0.67).

We were unable to fit models of Tsuga canadensis dominance and other functional

metrics.

3.5 Discussion

Tsuga canadensis is often cited as a model foundation species (Ellison et al. 2005;

Lovett et al. 2006), yet the strength of its influence on ecosystem processes is not well

understood on a quantitative level. While we found Tsuga canadensis functions as a

foundation by suppressing plant diversity, resource availability, and productivity, this

seems unlikely to be the case for every foundation species. Our measures of

microclimate, community composition, productivity, nutrient cycling, and light

availability across ravine forests indicate Tsuga canadensis as the major driver of ecosystem composition and function, with both direct and indirect influence. Tsuga canadensis has a direct, negative influence on richness of canopy functional group richness, the sapling and ground flora layers, and leaf litter. It also has a strong, negative influence on light availability and leaf litter biomass, an index of productivity. By suppressing light availability, Tsuga canadensis indirectly determines the sapling layer composition, which further suppresses overall ecosystem productivity. At the same time,

Tsuga canadensis plays a facultative role by supporting unique habitat for other taxa

(Snyder et al. 2002; Tingley et al. 2002; Ross et al. 2003). On the landscape, the most

65 important role of foundation species may be in defining a unique community that in turn enhances heterogeneity and diversity at larger scales.

Throughout the unglaciated Allegheny Plateau of Ohio, Tsuga canadensis creates a unique community state. It occurs in specialized habitats, dominating both the overstory and sapling layers, regardless of physiography and across riparian-upland transitions. Few other species occur in any of the vegetation layers of these steep riparian ravines, due in part to Tsuga canadensis dominance. The seedling layer was species poor and dominated by Acer rubrum; however, Acer rubrum seedling survival is likely low based on the scarcity of saplings. Key aspects of ecosystem function are also similar across Tsuga canadensis dominated ravines in the unglaciated Allegheny Plateau, and the microclimate of Tsuga canadensis ravines did not vary and across our riparian-upland transects. Gradient analyses indicated Tsuga canadensis was consistently associated with functional metrics, and our measures of relative decomposition rate and leaf litter chemistry indicate little difference in nutrient cycling. Similarity in nutrient cycling was reflected in the soil, as there was little difference in soil chemistry. All soil samples had a high sand content and low pH, which is probably due to the underlying sandstone bedrock as well as the influence of specific species, including Tsuga canadensis. We were not able to fit a model to quantify Tsuga canadensis influence on soil properties, or vice-versa, but this could be attributed to our small samples size and the high degree of similarity between sample sites.

The loss of the Tsuga canadensis foundation due to the arrival of HWA would be a major disturbance to ecosystems that occur along headwater areas of the Ohio River watershed as ecosystems shift to alternate states (Ellison et al. 2005). Tsuga canadensis

66 mortality would likely occur quickly, as the minimum winter temperatures in this region of Ohio are above the -5°C minimum average temperature and -28°C threshold minimum temperature thought to limit HWA population growth and Tsuga canadensis mortality

(Dukes et al. 2009). Should Tsuga canadensis mortality occur, which is likely, the terrestrial system may leach nitrate to streams (Jenkins et al. 1999), which is a cause for water quality concerns. Evidence indicates functional responses occur rapidly following

HWA arrival, and forest productivity, hydrology, and nutrient dynamics are all significantly altered (Jenkins et al. 1999; Cobb et al. 2006; Daley et al. 2007; Ford &

Vose 2007, Nuckolls et al. 2008; Orwig et al. 2008). Forest functional processes will ultimately be driven by the forest composition, which shifts more gradually (Ford et al.

2012; Krapfl et al. 2011). Even during the transition, there is evidence that forest composition can buffer some of the shifts in nutrient and energy cycling. For example, during decline in Tsuga canadensis productivity in North Carolina, co-occurring hardwoods initially increased growth by 40% (Nuckolls et al. 2008) and over a five-year period, the growth rate of the evergreen shrub Rhododendron maximum increased 2.6 times (Ford et al. 2012). Although the shrub layer of the central and southern

Appalachians is a major factor in post Tsuga canadensis forest processes elsewhere, shrubs do not occur in Tsuga canadensis forests in Ohio. Thus, Ohio forests will likely respond differently than areas of the central and southern Appalachians. At the same time, given differences in environmental context, climate, and current abundance of

Betula, New England Tsuga canadensis forests may not provide a model for Ohio either.

Dominated by interactions with Tsuga canadensis, there is little indication of resiliency in Ohio Tsuga canadensis ravine forest ecosystems. The sparse sapling layer,

67

which is mostly Tsuga canadensis, indicates seed dispersal would be important to forest

regeneration. Pollen records from a Tsuga canadensis population crash approximately

5,500 years ago provide some insight into the future forest (Allison et al. 1986; Heard &

Valente 2009). Following the historic pattern, studies indicate that Tsuga canadensis will

likely be replaced by Betula lenta in the New England, where Tsuga canadensis occurs

across a variety of landscape positions (Orwig & Foster 1998; Kizlinski et al. 2002). In

ravines at the Connecticut Arboretum, however, Small et al. (2005) found that forests

shifted toward a mixed hardwood overstory with a more developed understory. Sites in

the southern Appalachians indicate that historic Tsuga canadensis declines resulted in a

transition to Betula, followed by Acer and Quercus (Allison et al 1986; Heard & Valente

2009). These large-scale patterns from pollen records provide a general model, but do not account for current forest fragmentation, introduced species, or altered climate and nutrient dynamics. Furthermore, transitions to Quercus and or Acer may be complicated by a legacy of fire suppression (Nowacki & Abrams 2008) and other emerging threats such as Anoplophora glabripennis (Asian long-horned beetle), an invasive pest that prefers Acer species.

Although Tsuga canadensis forests represent a unique community state, loss of

Tsuga canadensis will like result in divergence at regional and local scales. The future

composition and structure of Tsuga canadensis forests is not as clear for the central and

southern portions of the Tsuga canadensis range (Ellison et al. 2005). Divergence across

the Tsuga canadensis range is indicated by the importance of Betula. Unlike the New

England, we observed Betula lenta sporadically, generally at lower slope positions.

Betula lenta was rarely found in the current understory, although it is known as an

68

opportunistic species well adapted to disturbances from canopy gaps to Tsuga canadensis

mortality (Orwig and Foster 1998; Catovsky and Bazzaz 2000). The distribution of

Betula lenta also differs on a local scale and perhaps more importantly, our data indicate

some local differences in community composition which are currently overwhelmed by

the influence of Tsuga canadensis. In near the stream, species such as Betula lenta and

Liriodendron tulipifera are more likely to occur with Tsuga canadensis. These fast

growing, deciduous species produce higher quality, faster-decaying leaf litter than Tsuga

canadensis and would alter the cycling of nutrients and energy across riparian and

headwater streams if they were to replace Tsuga canadensis (Kominoski et al. 2007; Ball

et al. 2008). At upper slope positions, Quercus species and Carya species are more

common. If they were able to expand, species composition and function may not shift

quite as much, which may indicate more resilience to the loss of Tsuga canadensis at

certain landscape positions. Future studies may indicate whether areas immediately

adjacent to streams should be prioritized for management and restoration.

Understanding Tsuga canadensis is important from an ecological perspective to

advance theories of foundation species in a more quantified and comparable way. At the

same time, it has immediate practical application. In Ohio, significant recreation and

tourism is centered in Tsuga canadensis dominated ravines. Work to further understand

Tsuga canadensis ecosystem ecology and continue refining successional predictions is

critical to develop possible management planning before HWA becomes more

problematic in Ohio, and the Central Hardwood region following several contained

introductions. Management based on studies from the New England or southern

Appalachians may not be as applicable, given the expected divergence across the Tsuga

69 canadensis range. For example, Betula lenta is currently an occasional species in Ohio, and thus may not increase as much in HWA impacted forests as it does in New England.

At the same time, unlike the southern Appalachians, Ohio hemlock forests to not have a significant shrub layer and Rhododendron maximum does not occur. Instead, it is more likely mixed deciduous species structured by ecological gradients would develop in Ohio following widespread hemlock mortality.

3.6 Acknowledgements

Salaries and funding for this study was provided in part by the state of Ohio through funds to the Ohio Agricultural Research and Development Center (OARDC).

Additionally funding was provided to the lead author via an OARDC SEEDS graduate grant. We are grateful for access to research sites, granted by the Ohio Division of

Natural Areas and Preserves and the Ohio Division of Forestry. R. Beinlich at Lake

Katharine and D. Glass at Hocking State Forest logistical advice. S. Rist, C. Clifton, K.

Davidson-Bennett and J. Martin provided valuable field assistance. Finally, we thank anonymous reviewers for their comments on an earlier version of the manuscript.

3.7 References

Allison, T.D., R.E. Moeller, and M.B. Davis, M.B. 1986. Pollen in laminated sediments

provides evidence for a mid-Holocene forest pathogen outbreak. Ecology

67:1101-1105.

American Society for Testing and Materials. 1985. Standard test method for particle-size

analysis of soils. Pages 117-127 in: Annual Book of ASTM Standards, American

Society for Testing Materials. West Conshohocken, PA, USA.

70

Anagnostakis, S.L. 1987. Chestnut blight: the classical problem of an introduced

pathogen. Mycologia 79:23-37.

AOAC Official Methods of Analysis. 2002. Method 990.03. Protein (crude) in Animal

Feed Combusion Method (Dumas method). JAOAC 72:770.

Ball, B.A., M.D. Hunter, J, S. Kominoski, C.S. Swan and M.A.Bradford. 2008.

Consequences of non-random species loss for decomposition dynamics:

experimental evidence for additive and non-additive effects. Journal of Ecology

96:303-313.

Beatley, J.C. 1959. The primeval forests of a periglacial area in the Allegheny Plateau

(Vinton and Jackson Counties). Bulletin of the Ohio Biological Survey, (New

Series) 1:1-182.

Black, R.A. and R.N. Mack. 1976. Tsuga canadensis in Ohio: Synecological and

phytogeographical relationships. Plant Ecology 32:11-19.

Brockman, C.S. 1998. Physiographic regions of Ohio [map]. Division of Geological

Survey, State of Ohio.

Bruno, J.,J. Stachowicz, and P.A. Townsend. 2003. Inclusion of facilitation into

theoretical ecology. 2003. Trends in Ecology and Evolution 18:119-125.

Catovsky, S. and F.A. Bazzaz. 2000. The role of resource interactions and seedling

regeneration in maintaining a positive feedback in hemlock stands. Journal of

Ecology 88:100-112.

Cobb, R.C., D.A. Orwig, and S. Currie. 2006. Decomposition of green foliage in eastern

hemlock forests of southern New England impacted by hemlock woolly adelgid

infestations. Canadian Journal of Forest Research 36:1331-1341.

71

Daley, M.J., N. G. Phillips, C. Pettijohn, and J.L. Hadley. 2007. Water use by eastern

hemlock (Tsuga canadensis) and black birch (Betula lenta): implications of

effects of the hemlock woolly adelgid. Canadian Journal of Forest Research

37:2031-2040.

Day, F.P. and C.D. Monk. 1974. Vegetation pattern on a Southern Appalachian

watershed. Ecology 55:1064-1074.

Dixon, P. and M.W. Palmer. 2003. Vegan, a package of R functions for community

ecology. Journal of Vegetation Science 14: 927-930.

Dukes, J.S., J. Pontius, D. Orwig, J.R. Garnas, V.L. Rodgers, N. Brazee, B. Cooke, K.A.

Theoharides, E.E. Stange, R. Harrington, J. Ehrenfeld, J. Gurevitch, M. Lerdau,

K. Stinson. R. Wick, and M. Ayres. 2009. Responses of insect pests, pathogens,

and invasive plant species to climate change in the forests of northeastern North

America: What can we predict? Canadian Journal of Forest Research 39:231-248.

Elliott, K.J, J.M. Vose, W.T. Swank, and P.V. Bolstad. 1999. Long-term patterns in

vegetation-site relationships in southern Appalachian forests. Journal of the

Torrey Botanical Society 126:320–334

Elliott, K.J. and W.T. Swank. 2008. Long-term changes in forest composition and

diversity following early logging (1912-1923) and the decline of the American

chestnut (Castanea dentata). Plant Ecology 197:155-172.

Ellison, A. M., M. S. Banks, B. D. Clinton, E. A. Colburn, K. Elliott, C. R. Ford, D. R.

Foster, B. D. Kloeppel, J. D. Knoepp, G. M. Lovett, J. Mohan, D. A. Orwig, and

N. L. Rodenhouse. 2005. Loss of foundation species: consequences for the

72

structure and dynamics of forested ecosystems. Frontiers in Ecology and the

Environment 3:479-486.

Ford, C.R. and J.M. Vose. 2007. Tsuga canadensis (L.) Carr. mortality will impact

hydrologic processes in southern Appalachian forest ecosystems. Ecological

Applications 17:1156-1167.

Ford, C.R., K.J. Elliott, B.D. Clinton, B.D. Kloeppel., J.M. Vose. 2012. Forest dynamics

following eastern hemlock mortality in the southern Appalachians. Oikos

121:523-536.

Grace, J. B. 2006. Structural equation modeling and natural systems. Cambridge

University Press. Cambridge, UK.

Grace, J.B. 2008. Structural equation modeling for observational studies. Journal of

Wildlife Management 72:14-22.

Heard, M.J. and M.J. Valente, M.J. 2009. Fossil pollen records forecast response of

forests to hemlock woolly adelgid invasion. Ecography 32:881887.

Huston, M.A. 1997. Hidden treatments in ecological experiments: reevaluating the

ecosystem function of biodiversity. Oecologia 110:449–460

International Standard, ISO 10694. 1995. Soil quality- determination of organic and total

carbon after dry combustion (elementary analysis). International Organization for

Standardization. Geneva, Switzerland.

Jenkins, J.C., J.D. Aber, J.D. and C.D. Canham. 1999. Hemlock woolly adelgid impacts

on community structure and N cycling rates in eastern hemlock forests. Canadian

Journal of Forest Research 29:630-645.

73

Keever, C. 1953. Present composition of some stands of former oak-chestnut forest in the

Southern Blue Ridge Mountains. Ecology 34:44–54

Kerr, J.W. 1983. Soil survey of Jackson County, Ohio. USDA Soil Conservation

Service. Washington, DC

Kizlinski, M.L, D.A. Orwig, R.C. Cobb and D.R. Foster. 2002. Direct and indirect

ecosystem consequences of an invasive past on forests dominated by eastern

hemlock. Journal of Biography 29:1489-1503.

Kominoski, J.S., C.M. Pringle, B.A. Ball, M.A. Bradford, D.C. Coleman, D.B. Hall, and

M.D. Hunter. 2007. Nonadditive effects of leaf litter species diversity on

breakdown dynamics in a detritus-based stream. Ecology 88:1167-1176.

Krapfl, K.J., E.J. Holzmueller, and M.A. Jenkins. 2011. Early impacts of hemlock woolly

adelgid in Tsuga canadensis forest communities of the southern Appalachian

Mountains. Journal of the Torrey Botanical Society 138:93-106.

Krueger, L.M. and C.J. Peterson. 2006. Effects of white-tailed deer on Tsuga canadensis

regeneration: evidence of microsites as refugia from browsing. American Midland

Naturalist 156:353-362.

Lemaster, D.D. and G.M. Gilmore.1989. Soil survey of Hocking County, Ohio. USDA

Soil Conservation Service. Washington DC

Lovett, G.M., C.D. Canham, M.A. Arthur, K.C. Weathers, and R.D. Fitzhugh. 2006.

Forest ecosystem responses to exotic pests and pathogens in eastern North

America. BioScience 56:395- 405.

McCune, B. and D. Keon. 2002. Equations for potential annual direct incident radiation

and heat load. Journal of Vegetation Science 13:603-606.

74

Minchin, P.R. 1987. An evaluation of the relative robustness of techniques for ecological

ordination. Vegetatio 69:89-107.

Morris, A.L. and P.C. Goebel. 2007. Geomorphic and riparian forest influences on

characteristics of large wood and large wood jams in old-growth and second-

growth forests in northern Michigan, SA. Earth Surface Processes and Landforms

32:1131-1153.

Nowacki , G.J. and M.D. Abrams. 2008. The demise of fire and “mesophication” of

forests in the eastern United States. BioScience 58:123-138.

Nuckolls, A, N. Wurzburger, C.R. Ford, R. Hendrick, J. M. Vose, and B. Kloeppel. 2008.

Hemlock Declines Rapidly with Hemlock Woolly Adelgid Infestation: Impacts on

the Carbon Cycle of Southern Appalachian Forests. Ecosystems 12:179-190.

Orwig, D. A. and D. R. Foster. 1998. Forest response to the introduced hemlock woolly

adelgid in southern New England, USA.. Journal of the Torrey Botanical Society

125:60–73.

Orwig, D.A, D.R. Foster, and D.L. Mausel. 2002. Landscape patterns of hemlock decline

in New England due to the introduced hemlock woolly adelgid. Journal of

Biogeography. 29: 1475-1487.

Orwig, D.A., R.C. Cobb, A.W. D’Amato, M.L. Kizlinski, and D.R. Foster. 2008. Multi-

year ecosystem response to hemlock woolly adelgid infestation in southern New

England forests. Canadian Journal of Forest Research 38:834-843.

Prasad, A. M. L.R. Iverson, S. Matthews and M. Peters. 2007. A Climate Change Atlas

for 134 Forest Tree Species of the Eastern United States

75

[database].http://www.nrs.fs.fed.us/atlas/tree. USDA Forest Service Northern

Research

Rhoades, C.C. 2007. The influence of American chestnut (Castanea dentata) on nitrogen

availability, organic matter and chemistry of silty and sandy loam soils.

Pedobiologia 50:553-562.

Ross, R.M., R.M. Bennett, C.D. Snyder, J.A. Young, D.R. Smith, and D.P. Lemarie.

2003. Influence of eastern hemlock (Tsuga Canadensis L.) on fish community

structure and function in headwater streams of the Delaware River basin.

Ecology of Freshwater Fish 12:60-65.

Runkle, J.R. and G.G. Whitney. 1987. Vegetation-site relationships in Lake Katharine

State Nature Preserve Ohio: A northern outlier of the mixed mesophytic forest.

Ohio Journal of Science 87:36-40.

Sasaki, T. and Lauenroth, W.K. 2011. Dominant species, rather than diversity, regulates

temporal stability of plant communities. Oecologia 166:761-768.

Small, M.J, C.J. Small, and G.D. Dreyer. 2005. Changes in a hemlock-dominated forest

following woolly adelgid infestation in southern New England. Journal of the

Torrey Botanical Society 132:458-470.

Snyder C.D., J.A. Young and D. Smith. 2002. Influence of eastern hemlock (Tsuga

canadensis) forests on aquatic invertebrate assemblages in headwater streams.

Canadian Journal of Fisheries and Aquatic Science 59:262–75.

Stadler, B., T. Müller and D.A. Orwig. 2006. The ecology of energy and nutrient fluxes

in hemlock forests invaded by hemlock woolly adelgid. Ecology 87:1792-1804.

76

Sutherland, E.K., B.J. Hale, and D.M. Hix, 2000. Defining species guilds in the Central

Hardwood Forest, USA. Plant Ecology:1-19.

Tingley, M.W., D.A. Orwig, R. Field and G. Motzkin. 2002. Avian response to removal

of a forest dominant: consequences of hemlock woolly adelgid infestations.

Journal of Biogeography 29:1505-1516.

United States Department of Agriculture, Natural Resources Conservation Service. 2012.

The Plants Database url: http://plants.usda.gov. [National Plant Data Team]

Greensboro, NC, USA.

Vandermast, D.B. and D.H. Van Lear. 2002. Riparian vegetation in the southern

Appalachian mountains (USA) following chestnut blight. Forest Ecology &

Management 155: 97-106.

Wallace, J.B., J.R. Webster, S.L.. Eggert, J.L. Meyer, and E.R. Siler.. 2001. Large

woody debris in a headwater stream: long-term legacies of forest disturbance.

International Review of Hydrobiology. 86:501-513.

Yorks, T.E. D.J. Leopold, and D.J. Raynal. 2002. Effects of Tsuga canadensis mortality

on soil water chemistry and understory vegetation: possible consequences of an

invasive insect herbivore. Canadian Journal of Forest Research 33:1525-1537.

77

Table 3.1 Mean overstory species relative basal area (±SE) in transects at 10, 30 and 50

meters from small streams. FG refers to the functional group identity, adapted from

Sutherland et al. (2000). Tsuga canadensis was removed from group 4 given its own group (10) as a foundation species and an evergreen. Species from group 9 were not found in our study. Symbols are used in figures.

Overstory 10 m 30 m 50 m Symbol Species FG

Acer rubrum 4.0% ± 0.0% 6.1% ± 0.2% 10.4% ± 0.0% 8 ACRU Acer saccharum 0.3% ± 2.8% 1.1% ± 6.1% 0.7% ± 6.9% 8 ACSA Betula lenta 8.8% ± 2.3% 11.5% ± 3.9% 1.3% ± 1.7% 1 BELE Carya CACO cordiformis 1.0% ± 5.7% 0.0% ± 1.9% 0.0% ± 2.6% 6 Carya glabra 0.0% ± 0.0% 1.0% ± 1.3% 0.7% ± 3.3% 6 CAGL Carya ovata 0.0% ± 0.0% 1.5% ± 0.7% 0.0% ± 0.1% 6 CAOV Carya ± CATO tomentosa 0.0% 0.0% 0.0% ± 0.0% 1.9% ± 2.8% 6 Celtis ± CEOC occidentalis 0.1% 0.6% 0.0% ± 0.3% 0.0% ± 0.9% 3 Cornus florida 0.0% ± 0.0% 0.0% ± 0.0% 0.1% ± 0.7% 3 COFL Fagus FAGR grandifolia 13.0% ± 4.0% 5.6% ± 3.8% 8.5% ± 3.1% 4 Fraxinus FRAM americana 0.0% ± 0.0% 1.3% ± 1.3% 0.0% ± 0.0% 3 Liriodendron LITU tulipifera 12.8% ± 3.9% 6.8% ± 4.3% 3.1% ± 4.6% 1 Nyssa sylvatica 0.0% ± 0.0% 0.0% ± 0.0% 1.2% ± 0.1% 5 NYSY Oxydendrum OXAR arborum 0.6% ± 0.1% 0.7% ± 0.0% 1.8% ± 0.0% 4 Pinus PIVI virginiana 0.0% ± 0.0% 0.0% ± 0.0% 5.3% ± 1.9% 2 Prunus serotina 0.0% ± 0.0% 3.8% ± 0.5% 0.1% ± 0.0% 3 PRSE Quercus alba 2.3% ± 0.0% 6.2% ± 0.7% 10.3% ± 0.0% 7 QUAL Quercus prinus 0.0% ± 0.0% 1.3% ± 0.7% 4.0% ± 0.7% 7 QUPR Quercus rubra 7.2% ± 1.0% 2.8% ± 0.0% 6.8% ± 0.0% 6 QURU Qurcus QUCO coccinea 2.3% ± 3.0% 4.8% ± 4.4% 2.6% ± 0.8% 6 Tsuga TSCA canadensis 47.5% ± 0.3% 45.5% ± 1.1% 41.0% ± 0.5% 10 Ulmus rubrum 0.0% ± 1.6% 0.2% ± 2.1% 0.0% ± 2.5% 8 ULRU

78

Table 3.2 Environmental and functional metrics (± SE) in eastern hemlock ravines are linked at distances of 10, 30 and 50 meters from headwater streams. Soil characteristics include: CEC: cation exchange capacity; P: soil phosphorus, K: soil potassium, Ca: soil calcium. Microclimate metrics of temperature (T) and relative humidity (RH) were recorded throughout the growing season (June- November) by data loggers at three of the eight sites. Light is estimated by Openness GS, WI: canopy openness in the deciduous growing season, winter. Leaf litter was collected during deciduous leaf fall (Sept-

December). Decay constant was determined by mass loss of cellulose paper.

10m 30m 50m Soil Percent sand (%) 72.00 ± 6.72 64.00 ± 6.01 64.00 ± 8.13 pH 4.51 ± 0.30 3.75 ± 0.10 4.24 ± 0.14 CEC 11.30 ± 1.91 15.97 ± 1.24 10.96 ± 1.16 Soil C:N 16.09 ± 1.36 17.51 ± 1.35 17.48 ± 1.86 P (μg/g) 3.42 ± 0.38 5.28 ± 1.43 4.78 ± 1.13 13.7 K (μg/g) 39.44 ± 6.18 34.44 ± 5.51 48.7 ± 6 17.7 20.6 Ca (μg/g) 62.59 ± 21.86 56.33 ± 8 55.96 ± 3 19.2 Mg (μg/g) 12.92 ± 2.69 28.61 ± 3 14.78 ± 2.78 Microclimate T min (°C) -0.96 ± 0.09 -1.24 ± 0.20 -0.98 ± 0.29 T max (°C) 29.30 ± 1.26 30.15 ± 0.71 30.05 ± 1.06 RH min (%) 38.93 ± 14.19 31.38 ± 2.32 35.38 ± 1.85 Light availability Openness GS 3.38 ± 0.49 3.66 ± 0.50 4.31 ± 0.45 Openness WI 21.53 ± 2.77 21.95 ± 1.53 22.39 ± 2.75 Leaf litter Leaf mass (g) 24.92 ± 2.74 31.13 ± 6.45 33.29 ± 6.07 Leaf C:N 40.45 ± 3.08 46.80 ± 3.16 48.99 ± 1.87 Relative decomposition Decay constant k 0.03 ± 0.00 0.02 ± 0.04 0.03 ± 0.04

79

Table 3.3 Sapling species frequency (stem counts ± SE) in transects at 10, 30 and 50 meters from small streams. Symbols are used in figures.

10m 30m 50m Symbol Acer rubrum 0.00 ± 0.00 0.50 ± 0.50 1.50 ± 0.93 ACRU Acer saccharum 0.00 ± 0.00 0.00 ± 0.00 0.25 ± 0.25 ACSA Betula lenta 0.00 ± 0.00 0.63 ± 0.63 0.00 ± 0.00 BELE Carpinus caroliniana 0.00 ± 0.00 0.00 ± 0.00 0.25 ± 0.25 CACA Fagus grandifolia 0.38 ± 0.18 0.25 ± 0.16 1.00 ± 0.63 FAGR Hamamelis virginiana 0.00 ± 0.00 0.00 ± 0.00 0.13 ± 0.13 HAVI Juglans nigra 0.00 ± 0.00 0.13 ± 0.13 0.00 ± 0.00 JUNI Magnolia tripetala 0.00 ± 0.00 0.38 ± 0.38 0.88 ± 0.88 MATR Nyssa sylvatica 0.00 ± 0.00 0.63 ± 0.38 0.25 ± 0.16 NYSY Oxydendrum arboreum 0.00 ± 0.00 0.25 ± 0.16 0.13 ± 0.13 OXAR Qurcus coccinea 0.00 ± 0.00 0.00 ± 0.00 0.25 ± 0.25 QUCO Tsuga canadensis 10.25 ± 1.25 17.63 ± 1.97 14.00 ± 2.76 TSCA Ulmus rubrum 0.00 ± 0.00 0.00 ± 0.00 0.38 ± 0.38 ULRU Vitis aestivalis 0.25 ± 0.16 0.13 ± 0.13 0.00 ± 0.00 VIAE Total Stems 10.88 ± 1.27 20.50 ± 1.69 19.00 ± 2.79

80

Table 3.4 Mean stem counts of seedling species (±SE) at three distances from headwater streams.

10 m 30 m 50 m Symbol Acer rubrum 82.00 ± 18.90 97.88 ± 15.58 130.00 ± 23.76 ACRU Acer saccharum 0.00 ± 0.00 0.00 ± 0.00 0.13 ± 0.13 ACSA Betula lenta 0.13 ± 0.13 0.25 ± 0.16 0.00 ± 0.00 BELE Carpinus caroliniana 0.00 ± 0.00 0.00 ± 0.00 3.00 ± 2.73 CACA Carya spp 0.00 ± 0.00 0.25 ± 0.16 0.13 ± 0.13 CARYA Cornus florida 0.00 ± 0.00 0.00 ± 0.00 0.25 ± 0.25 COFL Fagus grandifolia 0.50 ± 0.27 0.25 ± 0.16 0.38 ± 0.18 FAGR Liriodendron tulipifera 0.88 ± 0.88 0.00 ± 0.00 0.13 ± 0.13 LITU Nyssa sylvatica 0.00 ± 0.00 0.63 ± 0.63 3.13 ± 3.13 NYSY Ostrya virginia 0.00 ± 0.00 0.00 ± 0.00 4.63 ± 3.20 OSVI Oxydendrum arboreum 0.13 ± 0.13 0.00 ± 0.00 0.50 ± 0.50 OXAR Pinus virginiana 0.00 ± 0.00 0.00 ± 0.00 0.13 ± 0.13 PIVI Prunus serotina 0.00 ± 0.00 0.00 ± 0.00 0.25 ± 0.16 PRSE Quercus prinus 0.00 ± 0.00 0.00 ± 0.00 0.25 ± 0.16 QUPR Quercus rubrum 0.13 ± 0.13 0.63 ± 0.26 0.13 ± 0.13 QURU Tsuga canadensis 0.13 ± 0.13 0.75 ± 0.62 4.13 ± 3.84 TSCA

81

Table 3.5 Mean percent cover (± SE) of ground-flora taxa at 10, 30 and 50 m from headwater streams. In the 50 meter transect, bare ground exceeded 75% at all sites, which is reflected in a SE of 0%.

10 m 30 m 50 m Symbol Adiantum pedatum 0.00% ± 0.00% 0.00% ± 0.00% 0.01% ± 0.01% ADPE Arisaema triphyllum 0.15% ± 0.09% 0.03% ± 0.02% 0.00% ± 0.00% ARAT Actaea racemosa 0.03% ± 0.03% 0.01% ± 0.01% 0.03% ± 0.03% CIRA Cypripedium acaule 0.00% ± 0.00% 0.00% ± 0.00% 0.03% ± 0.03% CYAC Thelypteris cordata 0.99% ± 0.39% 0.36% ± 0.24% 0.13% ± 0.06% DRIM Erythronium americanum 0.01% ± 0.01% 0.00% ± 0.00% 0.00% ± 0.00% ERAM Geranium ± ± ± maculatum 0.00% 0.00% 0.00% 0.00% 0.03% 0.03% GEMA Iris spp 0.00% ± 0.00% 0.00% ± 0.00% 0.01% ± 0.01% IRIS Lindera benzoin 0.04% ± 0.03% 0.00% ± 0.00% 0.09% ± 0.07% LIBE Maianthemum canadense 0.19% ± 0.09% 0.00% ± 0.00% 0.01% ± 0.01% MACA Medeola virginiana 0.30% ± 0.24% 0.19% ± 0.08% 0.03% ± 0.02% MEVI Mitchella repens 0.06% ± 0.03% 0.13% ± 0.05% 0.09% ± 0.04% MIRE Osmunda ± ± ± claytoniana 1.96% 1.86% 0.29% 0.27% 0.00% 0.00% OSCL Parthenocissus quinquefolia 0.03% ± 0.03% 0.03% ± 0.03% 0.05% ± 0.04% PAQU Polystichum acrosticohides 0.18% ± 0.10% 0.08% ± 0.05% 0.14% ± 0.12% POAC Polygonatum ± ± ± biflorum 0.15% 0.09% 0.10% 0.04% 0.14% 0.06% POBI Podophyllum ± ± ± peltatum 0.08% 0.03% 0.00% 0.00% 0.01% 0.01% POPE Rubus alleghenensis 0.01% ± 0.01% 0.03% ± 0.02% 0.09% ± 0.09% RUAL Smilax spp 0.09% ± 0.03% 0.11% ± 0.04% 0.26% ± 0.10% SMILAX Maianthemum racemosum 0.05% ± 0.04% 0.03% ± 0.02% 0.00% ± 0.00% SMRA Tiarella cordifolia 0.14% ± 0.09% 0.00% ± 0.00% 0.00% ± 0.00% TICO Toxicodendron radicans 0.01% ± 0.01% 0.00% ± 0.00% 0.05% ± 0.03% TORA Trillium erectum 0.04% ± 0.03% 0.03% ± 0.03% 0.00% ± 0.00% TRER Trillium grandiflora 0.04% ± 0.03% 0.00% ± 0.00% 0.00% ± 0.00% TRGR Vaccinium spp 0.00% ± 0.00% 0.00% ± 0.00% 0.01% ± 0.01% VACC Viburnum ± ± ± acerifolium 0.00% 0.00% 0.01% 0.01% 0.08% 0.05% VIAC Viburnum dentatum 0.00% ± 0.00% 0.00% ± 0.00% 0.01% ± 0.01% VIDE Viola spp 0.08% ± 0.08% 0.05% ± 0.03% 0.09% ± 0.05% VIOLA Unknown Fern 0.00% ± 0.00% 0.00% ± 0.00% 0.01% ± 0.01% FERN Bare ground 81.19% ± 3.15% 84.69% ± 2.19% 87.50% ± 0.00% BARE

82

Figure 3.1 Species richness (S) (± SE) of vegetation. FG: functional groups. Letters indicate differences by distance at P ≤ 0.05.

83

Figure 3. 2 Shannon diversity (H) (± SE) of vegetation. FG: functional groups. Letters indicate differences by distance at P ≤ 0.05.

84

Figure 3. 3 Non-metric multidimensional scaling of the overstory with PCA axes of environmental variables. Continuous variables are shown as vectors, categorical variables as points. Species abbreviations can be found in Table 1. Abbreviations for environmental variables: Transects are set at three distances from headwater streams

10M, 30M, and 50M; slope shape was categorized as CC: concave, CV: convex, LI: linear; slope position includes BE: structural bench, HS: hill slope, RT: ridge top; slope refers to slope percent; Rad: incident radiation, Heat: heat load; Sand: percent soil particles characterized as sand.

85

Figure 3. 4 Non-metric multidimensional scaling of saplings with PCA axes of environmental variables. Continuous variables are shown as vectors, categorical vectors as points. Species abbreviations can be found in Table 1. Abbreviations for environmental variables: Transects are set at three distances from headwater streams

10M, 30M, 50M; slope shape was categorized as CC: concave, CV: convex, LI: linear; slope position includes BE: structural bench, HS: hill slope, RT: ridge top; slope refers to slope percent; Rad: incident radiation, Heat: heat load; Sand: percent soil particles characterized as sand.

86

Figure 3. 5 Non-metric multidimensional scaling of the overstory and functional metrics as PCA axes. Species symbols can be found in Table 1. Soil variables include sand

(percent sand), soil carbon:nitrogen (soilCN), soil pH (pH), and cation exchange capacity

(CEC). Leaf litter was measured in terms of mass (Leafmass) and carbon:nitrogen

(leafCN). Relative decomposition rate measured with cellulose paper mass loss is represented by k, the decomposition constant. Metrics of canopy openness determined through analysis of hemispherical photography include gap fraction during the growing

season (GSOPEN) and deciduous leaf-off period (WOPEN).

87

Figure 3. 6 Structural Equation model (SEM) of diversity metrics in Ohio hemlock forests. Hemlock dominance is defined by relative basal area, functional richness refers to functional groups of overstory species modified by Sutherland et al. (2000).

88

Figure 3. 7 Structural equation model (SEM) of the influence of hemlock dominance on canopy openness and leaf litter biomass. Hemlock dominance refers to relative basal area.

Growing season canopy openness was determined using hemispherical photography.

Sapling community composition represents the first axis of a non-metric multidimensional scaling analysis of the sapling layer.

89

Chapter 4: Decline in riparian Tsuga canadensis forests of the central Appalachian across an Adelges tsugae invasion chronosequence

4.1 Abstract

Forests of eastern North American are losing Tsuga candensis (eastern hemlock)

ecosystems throughout an expanding portion of its range due to the invasive pest insect

Adelges tsugae (hemlock woolly adelgid: HWA). Tsuga canadensis represents a small portion of the landscape, particularly in the central and southern Appalachians where it is largely restricted to cove and riparian areas; however, it is a foundation species that defines a unique forest ecosystem. Consequently, loss of T. canadensis will result in a re- organization of ecosystem structure and function as alternate communities develop. A greater understanding this transition will advance ecological theory, but is also directly applicable to management and restoration planning. While county-level patterns of detection are readily available, less is known about the process of decline and compositional shifts. We identified riparian T. canadensis forests along thirty headwater streams across West Virginia and Virginia, representing an invasion chrononosequence of nine to thirty-two years. Sites encompassed a range in elevation, slope, and aspect,

intended to identify patterns of T. canadensis dominance and decline. At each site, we

sampled the overstory and sapling vegetation and recorded the crown health of each T.

canadensis in transects at 10, 30, and 50 meters from the bank-full stream channel.

Although in severe decline, T. canadensis continued to dominate both the overstory and

90

sapling layers across riparian-upland transects across the central Appalachians. Structural equation modeling (SEM) indicated decline was moderated by higher elevations and landscape positions that received less incident radiation due to aspect and slope, but duration of Adelges tsugae invasion was the most influential factor in the decline of

crown health for the overstory. SEM also identified decline in the overstory as the most

influential factor in the decline of the sapling layer. Future forest function depends on

composition, which will likely vary depending on the presence of Rhododendron

maximum, an aggressive native shrub, and woody invasives including Ailanthus

altissima. Most importantly, functional processes that respond rapidly to T. canadensis

decline are likely to experience continuing fluctuation as the composition shifts more

gradually.

Key Words: Adelges tsugae; Central Appalachians; Eastern Hemlock; Foundation

Species; Hemlock Woolly Adelgid; Invasive Species; Structural Equation Modeling;

Tsuga canadensis

4.2 Introduction

Tsuga canadensis (L.) Carriere, (eastern hemlock) is an evergreen canopy tree

that functions as a foundation species across eastern North America (Ellison et al. 2005).

An invasive pest insect is causing widespread mortality across an expanding portion of

the T. canadensis range from the southern Appalachians to the Northeast. Adelges tsugae

Annand, (hemlock woolly adelgid, HWA) was introduced to a nursery in Richmond,

Virginia in 1951. Preferring fresh needles, HWA feeds on the parenchyma cells of xylem

91

rays, causing the loss of needles and buds, leading to the death of branches, and

eventually, tree mortality (McClure 1991a). Currently, horticultural oils and systemic

insecticides are the most reliable method for controlling HWA (Onken and Reardon

2005) but cost and labor requirements make this an unrealistic control strategy on a

landscape scale. A more promising option may be the use of predatory beetles as

biological control agents (Onken and Reardon 2008). While research on biocontrol

agents continues, many forests are already experiencing high levels of T. canadensis

mortality, altering forest dynamics (Ellison et al. 2005), and HWA continues to spread at

a mean rate of spread of 12 km·yr-1 across the native range of T. canadensis (Evans and

Gregorie 2006).

South of approximately the state of Pennsylvania, T. canadensis is largely

restricted to riparian and cove forests of the Appalachian Mountains and foothills (Ellison et al. 2005). While it does not occupy a large portion of the landscape, it is a unique habitat that moderates temperature and hydrology along headwater streams (Ford and

Vose 2007), and contributes to landscape diversity. Tsuga canadensis became increasing dominant in riparian forests following the elimination of Castanea dentata (Marsh.)

Borkh. (American chestnut) by Cryphonectria parasitica (Murrill) Barr (chestnut blight),

(Vandermast and Van Lear 2002). Thus, riparian forests of the central and southern

Appalachians have historically been dominated by a foundation species (either C. dentata or T. canadensis) (Day and Monk 1974, Elliott et al. 1999, Elliott and Swank 2008). The transiton from foundation species dominance to a more diverse deciduous forest following the loss of T. canadensis is a significant ecological change with many questions that merit further study. Understanding the transition process will illustrate the

92

timeframe of changes in both compostion and function, while providing a point to

compare future dynamics.

Although county-level infestations are recorded and available from the United

States Department of Agriculture Forest Service Forest Health Protection Agency, the rate and pattern of decline and mortality is less understood. In New England, HWA populations experience high winter mortality which may slow T. canadensis decline

(Skinner et al. 2003, Orwig et al. 2012); however, the effects of HWA may be accelerated in more southern, warmer portions of the T. canadensis range (Evans and Gregorie 2007,

Ford and Vose 2007). Dukes et al. (2009) estimate HWA is impacted by average winter temperatures of -5°C and cannot survive minimum temperatures below -28.8°C, which may be increasingly less common. Already, the southern portions of the T. canadensis range, particularly at lower elevations, are warmer than this threshold (Dukes et al. 2009).

Throughout areas impacted by HWA, T. canadensis complete mortality is estimated to occur between four to fifteen years of an invasion (McClure 1991a), but this varies with environmental context. Latitude, an estimate of the time of invasion, and minimum winter temperature seems to slow mortality and HWA impacts in the

Northeast, including central Massachusetts (Orwig et al. 2012). Farther south along the

Connecticut River Valley, initial variation in mortality across physiological gradients was ultimately overwhelmed by the duration of HWA invasion, and many sites experienced high levels (50-100%) of mortality within five years (Orwig and Foster 1998, Orwig et al.

2002, Orwig et al. 2008). In the Delaware Water Gap National Recreation Area of New

Jersey and Pennsylvania, the majority of stems remained healthy or in slight decline over nine years following initial detections (Eschtruth et al 2006). However, farther south in

93

North Carolina, HWA resulted in 50% mortality in six years, with complete mortality

expected in eight years (Ford et al. 2012). Mortality at Great Smoky Mountain National

Park over five-six years was 11% for overstory and 34% for sapling T. canadensis across

multiple forest types (Krapfl et al. 2011). In both the Northeast and the southern

Appalachians, saplings are thought to experience higher mortality than overstory trees

(Orwig and Foster 1998, Krapfl et al. 2011)

These studies suggest that mortality varies across the range of T. canadensis but is

eventually complete. Mortality also seems to eliminate saplings first, making future

recruitment of T. canadensis less likely. To determine whether similar trends are

occurring in riparian forests of the central Appalachians, we investigated the health of T.

canadensis stands across Virginia and West Virginia. This region includes the natural

populations of T. canadensis closest to the site of the original introduction, as well as

more recent invasions. In sites impacted for 9-32 years, we quantified the plant

community composition and determined: 1) Do any trees remain healthy in either the

overstory or sapling layer? 2) Does health or mortality vary by environmental context,

including elevation, slope or by T. canadensis dominance? 3) What is the prevalence of

invasive plant species?

4.3 Materials and Methods

STUDY SITES

We selected study sites along 30 headwater streams across the central Appalachians in

West Virginia and Virginia to encompass the variation in the physiography of T. canadensis forests (Table 1). As mapped by Dukes et al. (2009), sites in West Virginia

94

may currently experience minimum winter temperatures below -28.8°C, thought to be

lethal to HWA, but this is not expected to continue under current climate change

predictions. Tsuga canadensis stands in Virginia are already outside this minimum

temperature range. Sites were located on federal lands managed by the United States

Department of the Interior National Park Service and United States Department of

Agriculture Forest Service, the West Virginia Division of Natural Resources, and The

Nature Conservancy. Physiographic characteristics of the sites vary, with elevations from

363-1057 m and slopes from nearly level to 90%.

In West Virginia, ten sites occurred on the Appalachian Plateau in Nicholas,

Fayette, Greenbriar and Pocahontas counties. Within the counties, sites were located in

Carnifex Ferry State Battleground Park, the Gauley River National Recreation Area, New

River Gorge National River and Monongahela National Forest. The Appalachian Plateau is a region of highly dissected topography of deep and narrow valleys and ridges with dendritic drainage, with Ultisol and Inceptisol soils underlain by sandstone (Fralish

2003). The climate of the region is characterized by cold, snow winters but without persistent snow cover, particularly in the valleys. Precipitation is evenly distributed throughout the year, but is heavier on winward, west-facing slopes, and averages between

1041-1479 mm. Temperatures also vary with elevation, but average daily temperature is between -5- 1 °C in the winter, with minimums in the range of -10- -5°C. Summer temperatures average approximately 17- 21°C (Gorman and Espy 1975, Carpenter 1992,

Flegel 1998, Flegel 2007). We located five additional sites in the Allegheny Mountains in

Tucker County at the United States Department of Agriculture Forest Service Fernow

Experimental Forest and within the Monongahela National Forest. Tucker County

95 receives approximately 1450 mm precipitation, correlated with increasing elevation, and with more falling in June and less in January (Gilliam et al. 1996).

In Virginia, we located fifteen streams in the Ridge and Valley physiographic province in Alleghany, Augusta, Bath, Botetourt and Rockingham counties in the George

Washington National Forest and The Nature Conservancy’s Warm Springs Preserve. The

Ridge and Valley province is characterized by fold mountains, a series of resistant sandstone ridges separated by level shale and limestone valleys. Soils in the Ridge and

Valley Province are Ultisols and Inceptisols (Fralish 2003). Winter temperatures average approximately 0-2.5°C, with minimums averaging in the range of -6- 4° C. Summer temperatures average 21- 23°C. Precipitation in the Ridge and Valley increases along an elevation gradient and averages approximately 914- 1270 mm, with approximately half occurring during the growing season (Cook and Slabaugh 2006, Wolf and Thomas 2006).

Sites were selected as representative of relatively undisturbed T. canadensis dominated riparian forests, located from the species range map compiled by Prasad and

Iverson (1999) and recommendations of local managers. Tsuga canadensis dominance along the stream was assessed by the presence of live trees and snags. Counties where sites were located have been impacted by HWA for 9-32 years (Fig. 4.1). As we did not have detailed historical documentation for many areas, we may have excluded former T. canadensis dominated streams where forest succession is more advanced.

VEGETATION SAMPLING

At each site, we established three transects parallel to the stream at distances of

10, 30 and 50 m from the bank-full channel. The side of the stream valley or ravine chosen for sampling was determined based on logistical considerations due to the

96

location of cliffs, evidence of disturbance, and uniformity of landform. In each transect, we used a series of five 100-m2 circular plots. Within each circular plot, we recorded

basic physiographical data including slope percent (using a clinometer), slope shape,

slope position and aspect. A Garmin eTrex GPS (Garmin Ltd, Kansas, USA) unit was

used to record the latitude, longitude, and altitude of each plot. We calculated metrics of

direct incident radiation for each plot using latitude, slope percent and aspect in Equation

1 from McCune & Keon (2002). All species and dbh (diameter-at-breast-height, 1.37 m)

of the woody vegetation (all stems >2.5 cm dbh) were recorded. Stems between 2.5 cm

and 10 cm dbh were classified as saplings, while those > 10.0 cm dbh were classified as

overstory. Tsuga canadensis stems were assigned a canopy health class based on the

estimated foliage remaining 1: >75%; 2: 51-75%; 3: 26-50%; 4: 1-25%; 5: dead, as

outlined by Orwig and Foster (1998). As many plots had a high density of shrubs with

stems < 2.5 cm dbh, overall shrub cover was also estimated for each plot. Data from the

five plots were combined (added), creating one value for each transect. Environmental

metrics were averaged across each transect.

STATISTICAL ANALYSES

We investigated local differences in the response to T. canadensis mortality near the stream and moving upslope, as well as differences across our study region. Aside from location, there is some evidence that saplings decline faster than overstory individuals (Orwig and Foster 1998, Krapfl et al. 2011), so we used linear regression to test for correlation between size (dbh) and health decline score. To examine the potential

effect of distance from the stream, we used analysis of variance (ANOVA) to test for

differences in health decline between transects at 10, 30, and 50 meters from the stream

97

for the overstory and also for the sapling layer. We also used ANOVA to determine

differences in T. canadensis distribution relative to the stream, both before and after

HWA. Tsuga canadensis before HWA was estimated by relative basal area including

both live stems and standing snags. This estimate is conservative, as it did not include

any downed wood. We repeated the analysis of T. canadensis relative basal area by

distance to the stream with only live trees. For our ANOVA based analyses, in cases

where statistical assumptions were violated (normality, homoscedasity), data was

transformed or the non-parametric Krustal-Wallis test was used. To examine the change

in the composition of the overstory vegetation across all transects, we compared the estimated overstory vegetation community (including T. canadensis snags) and the community with only live stems using Multiple Response Permutation Procedure

(MRPP) using the Bray-Curtis distance metric.

To examine differences on a regional level and quantify the strength of the relationships between the decline in T. canadensis health and physiographic gradients, we

used structural equation modeling (SEM). SEM uses maximum likelihood to solve path

equations simultaneously and allows closer inspection of indirect effects, interactions,

and reciprocal relationships between variables (Grace 2008). We constructed one model

with the entire data set (30 sites) and a second model to examine any possible differences for the influence of factors on sites invaded in the last ten years, which included the thirteen sites from Tucker, Nicholas, and Fayette counties of West Virginia.

Statistical analyses were calculated in R version 2.10.0 (online documentation available at http://www.r-project.org/) with the exception of the structural equation

98

modeling, for which we used AMOS version 19 (SPSS Inc., Chicago, IL). Nomenclature

follows the United States Department of Agriculture PLANTS Database.

4.4 Results

In the central Appalachians where HWA has been present for decades, T.

canadensis are dead or in severe decline, but still dominates the overstory across all

transects from the stream (Table 2). Following T. canadensis, the most common species

were Betula lenta L. (sweet birch), B. alleghaniensis Britton (yellow birch) and

Liriodendron tulipifera L (tulip-poplar). Two evergreen species, Picea rubens Sarg. (red

spruce) and Pinus strobus L. (white pine) occurred occasionally with T. canadensis but

were not very dominant in terms of basal area, and the remaining overstory species are

deciduous. The sapling layer was dominated by the evergreen shrub, Rhododendron

maximum L., and T. canadensis (Table 3). We found one invasive tree, Ailanthus

altissima (Mill.) Swingle at high sapling densities at two sites, Hone Quarry Run and

Kephart Run, but few other woody invasives.

HWA has impacted T. canadensis throughout our study region regardless of local

gradients. Linear regression did not indicate any correlation between tree size (dbh) and

health decline status (Fig. 4.2). We did not detect a difference in health by distance to the

stream in the overstory (Kruskal-Wallis Χ2= 4.12, df= 2, P= 0.127) or sapling layer

(Kruskal-Wallis Χ2 = 0.5784, df = 2, P = 0.749). Our comparison of estimated T.

canadensis dominance prior to HWA (live stems and snags) did not differ by distance to

the stream (square root transformation was used to achieve normality, F= 2.33, df= 2, P=

0.103). Live T. canadensis was also distributed across the transects (Kruskal-Wallis Χ2 =

99

0.329, df= 2, P= 0.848). MRPP indicated that vegetation composition before HWA

(estimated) and current overstory composition are not significantly different across the 30 sites in West Virginia and Virginia (A= =0.001, P=0.697).

On a regional level, SEM indicated some differences in factors influencing mortality. When all 30 sites were examined, we were able to fit a model that explained

43% of the variance in T. canadensis decline in the overstory and 47% for the sapling layer (Fig. 4.3, Χ2= 14.73, P = 0.142, df= 10). Years impacted by HWA was the strongest driver of overstory T. canadensis decline (0.53), but the relative basal area of T. canadensis was also influential (0.30). There was a weak negative relationship between

T. canadensis decline and altitude (-0.16), indicating trees at higher altitudes, where

HWA may be impacted by cold temperatures, had healthier canopies. Increasing incident radiation resulted in a slightly increased score of overstory decline. (0.18). The decline of the sapling layer was directly influenced only by the overstory (0.47). When only the subset of more recently invaded sties (9-10 years) were examined we fit a model that explained 62% of the sapling layer decline, but only 10% of the variance in the overstory decline (Fig. 4.4, Χ2=0.227, P =0.893, df=2). In this model, incident radiation and altitude had negative influences on overstory decline (-0.58, -0.31 respectively). The sapling layer decline was influenced directly by overstory decline (0.48) and altitude

(0.34).

4.5 Discussion

Tsuga candensis forests are experiencing a major reorganization of structure and

function as they shift to new community states (Ellison et al. 2005), but many details of

100

this transition are not entirely clear. Our objective in this study was to examine T. canadensis decline in the central Appalachians, where HWA was originally introduced.

Information regarding the status of T. canadensis in the central Appalachians was not readily available, representing a gap in understanding of long-term patterns of mortality and forest succession. In addition, the central part of the T. canadensis range may differ from the Northeast and southern Appalachians. Although in decline, T. canadensis remains a dominant part of Central Appalachian forests impacted by HWA for decades, both in the overstory and sapling layer. Remaining T. canadensis are in severe decline, rarely maintaining even 25% crown foliage, with no indication of resistant trees.

Complete T. canadensis mortality is therefore inevitable in many forests.

The impacts of HWA may initially vary by landscape position and local climate, but this is eventually overwhelmed by the duration of HWA impact. In Massachusetts,

Orwig et al. (2012) found HWA density was controlled by winter climate, elevation and

latitude. In our Central Appalachian sites, SEM indicated decline was slowed by

increasing elevation and decreased incident radiation, presumably at sites with lower

temperatures, more northerly aspects, and steeper slopes. The decline in the sapling layer

was driven by the decline in the overstory rather than physiographic drivers. In the

subset of our sites where HWA has been present approximately nine to ten years, our

SEM did not explain much of the variance in overstory decline, and thus, unmeasured

factors such as soil quality might be influential. There is evidence that fertilization

experiments designed to increase tree resilience to HWA have instead increased HWA

populations (MClure 1991b); thus, more productive sites may experience more rapid

decline. At the same time, some variation in mortality could be attributed to pattern of

101

HWA invasion, even within individual counties. As a unique late-successional forest

type, T. canadensis is not very contiguous even on local scales, which likely slows

dispersal. For analyses, we used the year HWA was first identified at the county level as

our index of invasion duration. However, within counties, particularly in Virginia, our study sites likely represent stands impacted later during the invasion process. Our site selection was based on a visual approximation of T. canadensis dominance in riparian

areas, and in areas first impacted by HWA, much of the T. canadensis has probably died

and fallen, and forest succession progressed. Association between decline and

environmental factors are likely irrelevant as HWA impacts progress, particularly in this

portion of the range where winter temperatures exceed thresholds thought to affect HWA

(Dukes et al. 2009). In Great Smoky Mountains National Park, Krapfl et al. (2011) did

not find any pattern of T. canadensis mortality and forest type or physiography.

Although the loss of a foundation species is characterized by rapid change in

community state (Ellison et al. 2005), the transition in composition following T.

canadensis mortality occurs over decades as HWA causes decline, then mortality. We

found remaining live, but stressed trees in sites where HWA has been present for 30

years. Sites where HWA arrived 9-10 years ago had trees with healthy crowns, with >

75% foliage. This seems inconsistent with estimates of complete mortality in eight years

projected for a site in the southern Appalachians (Ford et al. 2012), or perhaps even the

rates of 50-100% mortality estimated for the Northeast (Orwig and Foster 1998, Orwig et

al. 2008). Rather, mortality in the central Appalachians might be more similar to the

gradual decline over nine years observed at the Delaware Water Gap National Recreation

Area in Pennsylvania and New Jersey (Eschtruth et al. 2006). It is also possible that

102

mortality is initially rapid, then proceeds more slowly. This contributes to a slow turnover in community composition. Consistent with our study, Krapfl et al. (2011) did not detect compositional shifts five to six years following initial HWA infestation in the southern

Appalachians. As a slow decaying species, T. canadensis will likely leave a lasting legacy in the form of snags and downed wood, much in the same way that Castanea dentata logs influenced functional processes for decades following decline (Wallace et al.

2001, Rhodes et al. 2007). Species that will likely replace T. canadensis such as Betula lenta, Betula alleghaniensis, and in some cases Rhododendron maximum will not provide similar legacies in structure and biogeochemistry. Yet, even at initial stages of HWA invasion, forest function begins to reorganize.

While composition changes gradually, rapid shifts in productivity, hydrology and evapotranspiration, and nutrient dynamics have been noted at sites in both the Northeast and the southern Appalachians (Jenkins et al. 1999, Cobb et al. 2006, Daley et al. 2007,

Ford and Vose 2007, Nuckolls et al. 2008, Orwig et al. 2008). Forest functional processes such as productivity and nutrient cycling will continue to fluctuate as communities redevelop from T. canadensis mortality. Ultimately, functional processes will depend on community composition, which will likely vary. Species such as Betula spp. and

Liriodendron tulipifera produce faster decaying leaf litter which would accelerate nutrient cycles across terrestrial and stream ecosystems (Kominoski et al. 2007, Ball et al.

2008). Invasive species such as Ailanthus altissima will likely play a significant role in the future of at least some T. canadensis forests. While we did not find A. altissima at many sites, at two sites it was fairly dominant throughout the sites, with few other species in the sapling layer. On the other hand, tree recruitment may be limited by the evergreen

103

shrub, Rhododendron maximum. A dominant in the sapling layer across many of our sites, R. maximum may be the new foundation of many Appalachian T. canadensis

riparian forests, as it is known to dominate gap dynamics (Clinton et al. 1994, Beckage et

al. 2000, Beckage and Clark 2003, Beckage et al 2008), nutrient cycling (Kominoski et

al. 2007, Ball et al. 2008, Wurtzberger and Hendrick 2008), and forest hydrology (Ford

and Vose 2007). While Siderhurst et al. (2010) suggest the evergreen cover provided by

R. maximum may moderate impacts to stream temperature, over time it will not provide

the same riparian and in-stream habitat structure in the form of downed wood.

4.6 Conclusion

Eastern forests are losing a unique foundation species ecosystem as T. canadensis

is eliminated from an expanding portion of its range by HWA. While functional change

seems to occur rapidly, these processes will continue to fluctuate as compositional

change occurs more gradually. While research to control HWA continues, T. canadensis

across physiological gradients and size classes are impacted and will likely die. This is a

significant ecological shift in the riparian and cove forests of the Appalachian Mountains

and foothills, the implications of which are still not fully understood. Invasive species

such as Ailanthus altissima and the native evergreen shrub Rhododendron maximum may

be particularly influential as Appalachian riparian forests re-organize.

104

4.7 Acknowledgements

Salaries and financial support for this research were provided by the Ohio Agricultural

Research and Development Center (OARDC) SEEDS Program, The Ohio State

University College of Food, Agriculture and Environmental Science, and The Ohio State

University. We thank T. Macy, L. Kobelt, C. Clifton, and J. Martin for field assistance.

We are also grateful for the advice and assistance with research permits provided by J.

Perez, T. Schuler, M. Smith, W. San Jule, K. Karriker, T. Slater, E. Haverlack, S.

Tanguay, and S. Cowell.

4.8 Literature Cited

Ball, B.A., M.D. Hunter, J, S. Kominoski, C.S. Swan and M.A.Bradford. 2008.

Consequences of non-random species loss for decomposition dynamics:

experimental evidence for additive and non-additive effects. Journal of Ecology

96:303-313.

Beckage, B. and J.S. Clark, J.S. 2003. Seedling survival of three forest tree species: the

role of spatial heterogeneity. Ecology 84:1849-1861.

Beckage, B., J.S. Clark, B.D. Clinton, and B.L. Haines. 2000. A long-term study of tree

seedling recruitment in southern Appalachian forests: the effects of canopy gaps

and shrub understories. Canadian. Journal of Forest Research. 30:1617-1631.

Beckage, B., B.D. Kloeppel, J. A. Yeakley, S.F. Taylor, and D.C. Coleman. 2008.

Differential effects of understory and overstory gaps on tree regeneration. Journal

of the Torrey Botanical Society 135:1-11.

105

Carpenter, S.G. 1992. Soil Survey of Nicholas County, West Virginia. United States

Department of Agriculture Soil Conservation Service. Washington, DC.

Clinton, B.D., L.R. Boring, and W.T. Swank. 1994. Regeneration patterns in canopy gaps

of mixed-oak forests of the southern Appalachians: influences of topographic

position and evergreen understory. American Midland Naturalist 132:308-319.

Cobb, R.C., D.A. Orwig, and S. Currie. 2006. Decomposition of green foliage in eastern

hemlock forests of southern New England impacted by hemlock woolly adelgid

infestations. Canadian Journal of Forest Research 36:1331-1341.

Cook, M E. M. and J. D. Slabaugh. 2006. Soil Survey of Alleghany County, Virginia.

United States Department of Agriculture Natural Resources Conservation Service.

Washington, DC.

Daley, M.J., N. G. Phillips, C. Pettijohn, and J.L. Hadley. 2007. Water use by eastern

hemlock (Tsuga canadensis) and black birch (Betula lenta): implications of

effects of the hemlock woolly adelgid. Canadian Journal of Forest Research

37:2031-2040.

Dukes, J.S., J. Pontius, D. Orwig, J.R. Garnas, V.L. Rodgers, N. Brazee, B. Cooke, K.A.

Theoharides, E.E. Stange, R. Harrington, J. Ehrenfeld, J. Gurevitch, M. Lerdau,

K. Stinson. R. Wick, and M. Ayres. 2009. Responses of insect pests, pathogens,

and invasive plant species to climate change in the forests of northeastern North

America: What can we predict? Canadian Journal of Forest Research 39:231-248.

Elliott, K.J, J.M. Vose, W.T. Swank, and P.V. Bolstad. 1999. Long-term patterns in

vegetation-site relationships in southern Appalachian forests. Journal of the

Torrey Botanical Society 126:320–334

106

Elliott, K.J. and W.T. Swank. 2008. Long-term changes in forest composition and

diversity following early logging (1912-1923) and the decline of the American

chestnut (Castanea dentata). Plant Ecology 197:155-172.

Ellison, A. M., M. S. Banks, B. D. Clinton, E. A. Colburn, K. Elliott, C. R. Ford, D. R.

Foster, B. D. Kloeppel, J. D. Knoepp, G. M. Lovett, J. Mohan, D. A. Orwig, and

N. L. Rodenhouse. 2005. Loss of foundation species: consequences for the

structure and dynamics of forested ecosystems. Frontiers in Ecology and the

Environment 3:479-486.

Eschtruth, A. K., N. L. Cleavitt, J. J. Battles, R. A. Evans, and T. J. Fahey. 2006.

Vegetation dynamics in declining eastern hemlock stands: 9 years of forest

response to hemlock woolly adelgid infestation. Canadian Journal of Forest

Research 36:1435-1450.

Evans, A.M and T.G. Gregoire. 2007. A geographically variable model of hemlock

woolly adelgid spread. Biological Invasions 9:369-382.

Flegel, D.G. 1998. Soil Survey of Pocahontas County, West Virginia. United States

Department of Agriculture Natural Resources Conservation Service. Washington,

DC.

Ford, C.R. and J.M. Vose. 2007. Tsuga canadensis (L.) Carr. mortality will impact

hydrologic processes in southern Appalachian forest ecosystems. Ecological

Applications 17:1156-1167.

Ford, C.R., K.J. Elliott, B.D. Clinton, B.D. Kloeppel., J.M. Vose. 2012. Forest dynamics

following eastern hemlock mortality in the southern Appalachians. Oikos

121:523-536.

107

Fralish, J.S. 2003. The Central Hardwood Forest: Its boundaries and physiographic

provinces, pp 1-20 in J.W. Van Sambeek, J.O. Dawson, F. Ponder, Jr., E.F.

Loewenstein, and J.S. Fralish [eds]. Proceedings, 13th Central Hardwood Forest

Conference; 2002 April 1-3; Urbana, IL. Gen. Tech. Rep. NC-234. St. Paul, MN:

U.S. Department of Agriculture, Forest Service, North Central Research Station.

Fralish, J.S. and T.G. McArdle. 2009. Forest dynamics across three centry-length

disturbance regimes in the Illinois Ozark Hills. American Midland Naturalist

162:418-449.

Gilliam, F.S., M.B. Adams, B.M. Yurish. 1996. Ecosystem nutrient responses to chronic

nitrogen inputs at Fernow Experimental Forest, West Virginia. Canadian Journal

of Forest Research 16:196-205.

Gorman, J.L. and L.E. Espy. 1975. Soil Survey of Fayette and Raleigh Counties, West

Virginia. United States Department of Agriculture Soil Conservation Service.

Washington, DC.

Grace, J.B. 2008. Structural equation modeling for observational studies. Journal of

Wildlife Management 72:14-22.

Jenkins, J.C., J.D. Aber, J.D. and C.D. Canham. 1999. Hemlock woolly adelgid impacts

on community structure and N cycling rates in eastern hemlock forests. Canadian

Journal of Forest Research 29:630-645.

Kizlinski, M.L, D.A. Orwig, R.C. Cobb and D.R. Foster. 2002. Direct and indirect

ecosystem consequences of an invasive past on forests dominated by eastern

hemlock. Journal of Biography 29:1489-1503.

108

Kominoski, J.S., C.M. Pringle, B.A. Ball, M.A. Bradford, D.C. Coleman, D.B. Hall, and

M.D. Hunter. 2007. Nonadditive effects of leaf litter species diversity on

breakdown dynamics in a detritus-based stream. Ecology 88:1167-1176.

Krapfl, K.J., E.J. Holzmueller, and M.A. Jenkins. 2011. Early impacts of hemlock woolly

adelgid in Tsuga canadensis forest communities of the southern Appalachian

Mountains. Journal of the Torrey Botanical Society 138:93-106.

McClure, M.S. 1991a. Role of wind, birds, deer, and humans in the dispersal of hemlock

woolly adelgid (Homoptera, Adelgidae). Environmental Entomology 19:36-43.

McClure, M.S. 1991b. Nitrogen fertilization of hemlock increases susceptibility to

hemlock woolly adelgid. Journal of Arboriculture 17: 227-230.

McCune, B. and D. Keon. 2002. Equations for potential annual direct incident radiation

and heat load. Journal of Vegetation Science 13:603-606.

Nuckolls, A, N. Wurzburger, C.R. Ford, R. Hendrick, J. M. Vose, and B. Kloeppel. 2008.

Hemlock Declines Rapidly with Hemlock Woolly Adelgid Infestation: Impacts on

the Carbon Cycle of Southern Appalachian Forests. Ecosystems 12:179-190.

Oksanen, J., F. G. Blanchet, R. Kindt, P. Legendre, R.B. O’Hara, G.L. Simpson, P.

Solymos, M.H. H. Stevens, and H. Wagner. 2010. Vegan: A community ecology

package. R package version 1.17-4. http://CRAN. R-project.org/package=vegan.

Onken, B. and R. Reardon, eds. 2005. Proceedings, Third Symposium on Hemlock

Woolly Adelgid in the Eastern United States. Feb 1-3, 2005. United States

Department of Agriculture, Forest Service, Northern Research Station,

Morgantown, WV, USA. 367 p.

109

Onken, B. and R. Reardon, eds. 2008. Proceedings, Fourth Symposium on Hemlock

Woolly Adelgid in the Eastern United States. Feb 12-14, 2005. United States

Department of Agriculture, Forest Service, Northern Research Station,

Morgantown, WV, USA.

Orwig, D. A. and D. R. Foster. 1998. Forest response to the introduced hemlock woolly

adelgid in southern New England, USA.. Journal of the Torrey Botanical Society

125:60–73.

Orwig, D.A, D.R. Foster, and D.L. Mausel. 2002. Landscape patterns of hemlock decline

in New England due to the introduced hemlock woolly adelgid. Journal of

Biogeography. 29: 1475-1487.

Orwig, D.A., R.C. Cobb, A.W. D’Amato, M.L. Kizlinski, and D.R. Foster. 2008. Multi-

year ecosystem response to hemlock woolly adelgid infestation in southern New

England forests. Canadian Journal of Forest Research 38:834-843.

Orwig, D.W. J.R. Thompson, N.A. Povak. M. Manner. D. Niebyl, and D.R. Foster. 2012.

A foundation tree at the precipice: Tsuga canadensis health after the arrival of

Adelges tsugae in central New England. Ecosphere 3:1-16.

Prasad, A. M. L.R. Iverson, S. Matthews and M. Peters. 2007. A Climate Change Atlas

for 134 Forest Tree Species of the Eastern United States

[database].http://www.nrs.fs.fed.us/atlas/tree. USDA Forest Service Northern

Research

Rhoades, C.C. 2007. The influence of American chestnut (Castanea dentata) on nitrogen

availability, organic matter and chemistry of silty and sandy loam soils.

Pedobiologia 50:553-562.

110

Siderhurst, L.A., H.P. Griscom, M. Hudy, and Z.J. Bortolot. 2010. Changes in light levels

and stream temperatures with the loss of eastern hemlock (Tsuga canadensis) at a

southern Appalachian stream: implications for brook trout. Forest Ecology and

Management. 260:1677-1688.

Skinner, M., B. L. Parker, S. Gouli, and T. Ashikaga. 2003. Regional responses of

hemlock woolly adelgid (Homoptera: Adelgidae) to low temperatures.

Environmental Entomology 32:523–528.

Vandermast, D.B. and D.H. Van Lear. 2002. Riparian vegetation in the southern

Appalachian mountains (USA) following chestnut blight. Forest Ecology &

Management 155: 97-106.

Wallace, J.B., J.R. Webster, S.L.. Eggert, J.L. Meyer, and E.R. Siler.. 2001. Large

woody debris in a headwater stream: long-term legacies of forest disturbance.

International Review of Hydrobiology. 86:501-513.

Wolf, B.L. and J.R. Thomas. 2006. Soil Survey of Bath County, Virginia. United States

Department of Agriculture Natural Resources Conservation Service. Washington,

DC.

Wurzburger, N. and R.L. Hendrick. 2009. Plant litter chemistry and mycorrhizal roots

promote a nitrogen feedback in a temperate forest. Journal of Ecology 97: 528-

536.

111

Table. 4.1. Study sites across the Central Appalachians

Year Elev Year Elev Site Name Invaded County (m) Lat Long Site Name Invaded County (m) Lat Long Back Creek Alleghany, Tributary 1 1993 Bath, VA 741 38.115 79.474 Jerrys Run 1991 VA 648 37.472 80.113 Back Creek Rockingham, Tributary 2 1993 Bath, VA 619 38.103 79.450 Kephart Run 1991 VA 597 38.325 79.035 Barenshee Nicholas, Left Prong Run 2002 WV 699 38.168 79.173 Wilson Creek 1993 Bath, VA 582 37.564 79.475 Greenbriar, Bear Run 1998 WV 1042 38.111 80.220 Lindy Run 2001 Tucker, VA 950 39.055 79.312 Greenbriar, Little Laurel Big Run 1998 WV 1008 38.123 80.233 Run 1993 Bath, VA 1057 38.390 79.051 Bradley Little Mare Pond 1991 Augusta, VA 605 38.165 79.173 Mountain 1993 Bath, VA 718 38.004 79.461 Carnifex Ferry Nicholas, Stream 1 2002 WV 457 38.124 80.556 Mare Run 1993 Bath, VA 345 38.011 79.463 Carnifex Ferry Nicholas, Masons Nicholas,

Stream 2 2002 WV 494 38.124 80.556 Branch 2002 WV 363 38.132 80.592 112 Cranberry River Pochahontas, Pounding Mill Alleghany, Tributary 1993 WV 1051 38.123 80.170 Run 1991 VA 583 37.485 79.550 Botetourt, Elklick Run 2001 Tucker, WV 600 39.043 70.392 Roaring Run 1979 VA 512 37.424 79.542 Elklick Run Simpsons Alleghany, Tributary 2001 Tucker, WV 603 39.042 79.394 Creek 1991 VA 551 37.503 79.384 Skidamore Rockingham, Engine Run 2001 Tucker, WV 973 39.060 79.291 Fork 1991 VA 758 38.323 79.094 Williams Hedricks River Pochahontas, Creek 2002 Fayette, WV 510 38.092 80.563 Tributary 1993 WV 869 38.162 80.133 Hone Rockingham, Quarry Run 1991 VA 609 38.281 79.082 Wilson Creek 1993 Bath, VA 561 37.565 79.472 Horseshoe Run 2001 Tucker, WV 571 39.103 79.364 Wolf Creek 2002 Fayette WV 552 38.031 81.052

112

Table 4.2. Overstory species relative basal area (± SE) at three distances from the stream across

30 streams in the central Appalachians impacted by Adelges tsugae. EST Tsuga candensis includes live stems and snags, as an estimate of pre-HWA T. canadensis basal area. T. canadesis live individuals with some remaining canopy foliage. Other hardwoods includes 22 incidental species contributing <1% basal area at any distance.

10 m 30 m 50 m

EST Tsuga canadensis 50.03% ± 4.73% 39.86% ± 4.14% 34.00% ± 4.70%

Tsuga canadensis 37.35% ± 4.90% 31.85% ± 4.17% 28.79% ± 5.01%

Betula lenta 15.11% ± 3.68% 8.90% ± 2.71% 4.59% ± 1.40%

Betula alleghaniensis 6.29% ± 2.34% 7.00% ± 2.56% 3.80% ± 1.66%

Liriodendron tulipifera 5.90% ± 1.55% 9.16% ± 2.52% 6.69% ± 2.02%

Picea rubens 4.75% ± 2.52% 4.47% ± 1.89% 8.30% ± 4.04%

Quercus prinus 4.13% ± 2.36% 6.74% ± 2.33% 9.41% ± 2.85%

Quercus alba 4.03% ± 2.14% 5.56% ± 2.37% 5.36% ± 2.21%

Pinus strobes 3.88% ± 1.98% 3.58% ± 2.36% 4.84% ± 2.81%

Acer rubrum 3.28% ± 1.18% 5.94% ± 1.58% 7.40% ± 1.76%

Quercus rubra 2.83% ± 1.58% 3.33% ± 1.30% 6.83% ± 2.41%

Nyssa sylvatica 2.34% ± 1.50% 2.21% ± 1.16% 1.61% ± 0.96%

Acer saccharum 2.29% ± 1.11% 0.32% ± 0.19% 0.40% ± 0.21%

Tillia americana 1.50% ± 1.01% 1.05% ± 0.64% 0.20% ± 0.20%

Fagus grandifolia 1.36% ± 0.71% 1.38% ± 0.80% 1.04% ± 0.70%

Quercus coccinea 0.91% ± 0.53% 0.97% ± 0.41% 2.42% ± 1.27%

Quercus velutina 0.00% ± 0.00% 2.16% ± 0.97% 2.51% ± 1.12%

Other Hardwoods 4.04% ± 3.48% 5.36% ± 4.39% 5.82% ± 5.20%

113

Table 4.3. Relative basal area (± SE) of sapling (2.5-10 cm dbh) species in 30 sites impacted by

Adelges tsugae in the central Appalachians. Alanthis altissima is an invasive tree found at high levels at a few sites.

10 m 30 m 50 m

Rhododendron maximum 42.19% ± 6.20% 29.45% ± 5.50% 24.10% ± 5.54%

Tsuga canadensis 18.19% ± 2.95% 34.68% ± 4.82% 25.90% ± 4.29%

Acer saccharum 5.34% ± 3.18% 1.32% ± 0.85% 2.09% ± 1.31%

Fagus grandifolia 5.08% ± 2.90% 2.41% ± 1.10% 1.74% ± 0.62%

Hamamelis virginiana 4.18% ± 1.90% 2.92% ± 1.56% 5.90% ± 2.51%

Acer pensylvanicum 3.49% ± 2.47% 2.91% ± 1.64% 1.84% ± 1.31%

Aralia spinosa 3.47% ± 1.96% 2.36% ± 1.40% 2.94% ± 1.74%

Pincea rubens 3.47% ± 1.96% 2.36% ± 1.40% 2.94% ± 1.74%

Betula lenta 2.43% ± 0.94% 2.47% ± 1.51% 0.83% ± 0.42%

Betula alleghaniensis 2.29% ± 1.25% 1.58% ± 0.77% 2.68% ± 1.28%

Acer rubrum 2.18% ± 1.01% 5.48% ± 1.47% 8.76% ± 2.41%

Kalmia latifolia 1.80% ± 0.58% 2.53% ± 1.37% 4.41% ± 2.57%

Ailanthis altissima* 0.73% ± 0.73% 1.76% ± 1.69% 3.81% ± 2.64%

Carpinus caroliniana 0.67% ± 0.34% 1.95% ± 0.88% 1.00% ± 0.68%

Nyssa sylvatica 0.63% ± 0.50% 2.36% ± 1.43% 5.07% ± 2.35%

Vitus aestivalis 0.33% ± 0.29% 2.08% ± 1.73% 1.58% ± 1.16%

Amelanchier arborea 0.28% ± 0.14% 1.53% ± 1.26% 0.34% ± 0.24%

Other 2.40% ± 2.28% 2.21% ± 2.10% 7.00% ± 5.36%

114

Figure 4.1. Location of study counties across West Virginia and Virginia with the initial year of

HWA detection. A full list of sites can be found in Table 1.

115

Figure. 4.2. Health of individual T. canadensis trees by size (dbh) across 30 sites in the central

Appalachians. Regressions were not significant. Shapes refer to the distance from headwater streams: ■10, ●30, ∆50 meters. Health categories follow Orwig & Foster (1998): 1: > 75% canopy foliage remaining, 2: 51-75% canopy foliage, 3: 26-50% canopy foliage, 4: 1-25% canopy foliage, 5: dead.

116

Figure 4.3. Structural equation model (SEM) of factors influencing Tsuga candensis decline in the overstory and sapling layers of thirty sites across West Virginia and Virginia invaded 10-32 years. The overstory decline was the only direct factor influencing sapling decline.

117

Figure 4.4. Structural equation model (SEM) of factors influencing health in the overstory and sapling layers in sites invaded within the last ten years in West Virginia.

118

Chapter 5: Removal of eastern hemlock by an invasive pest causes

community divergence across environmental gradients

5.1 Abstract

Forests of eastern North America are losing a foundation species as the invasive pest

hemlock woolly adelgid (Adelges tsugae; HWA) impacts an increasing portion of the native

range of eastern hemlock (Tsuga canadensis; hemlock). We examined vegetation community composition and ecosystem function in thirty hemlock-dominated riparian forests across the central Appalachians impacted by HWA for 9-32 years. Transects at 10, 30 and 50-m from headwater streams were used to compare local and regional differences in impacts to composition and function. We found increases in species richness in the overstory and seedling layers, but not the sapling layer. Multivariate analysis of variance indicated a change in composition of the overstory and seedling layers among overstory hemlock decline categories.

Further, increases in the multivariate variance of the overstory communities indicate they are diverging, and gradient analyses indicate associations between some species and environmental gradients, including elevation and distance from the stream. Our analyses indicated fewer changes in the sapling layer, which is likely due to an aggressive native evergreen shrub, rhododendron (Rhododendron maximum), which occurred in high densities at some, but not all, of our sites. Structural equation modeling (SEM) indicated hemlock decline influences resource availability and nutrient cycling. Soil cation exchange capacity and carbon to nitrogen ratio increased with hemlock decline, but this varied with soil texture. Likewise, canopy openness

119 increased during the growing season as well as the deciduous dormant season, but at higher elevations this change is modified by the presence of red spruce (Picea rubens). As hemlock declines, nutrient availability and cycling rates are increasing, as relative decomposition rates increased and leaf litter carbon to nitrogen rates decreased as overstory hemlock declined. Our study agrees with some of the compositional and functional shifts found in other regions, but also highlights the complexity of the loss of hemlock as a foundation species. Differences in community reorganization will likely become more apparent as composition changes over the coming decades.

Keywords: Adelges tusgae, alternate states, hemlock woolly adelgid, resilience, Tsuga canadensis

5.2 Introduction

Novel disturbances, such as those caused by invasive species, are impacting ecosystems at an accelerating rate, resulting in altered species composition and functions that cause management challenges (Gandhi and Herms 2009). The impacts of invasive species on ecosystem processes are increasingly apparent and some general patterns have emerged, but with caveats. Ehernfeld (2011) suggests invasives generally increase pools and fluxes of nutrients and energy, but such responses vary by environmental context. Perhaps in particular, introduced pests and pathogens are causing rapid shifts across forest ecosystems, in many cases by removing specific species, which results in cascading effects throughout energy and nutrient processes (Lovett et al. 2006). Changes such as large areas of tree mortality, further invasion by exotic plant species, and loss of ecosystem services such as carbon storage, biodiversity, wildlife

120 habitat, water quality and recreation opportunities, create management challenges (Gandhi and

Herms 2009). However, relevant ecological theories to understand such rapid transitions and address management concerns are generally not well tested in complex ecosystems such as forests. Ecologists have made progress in the areas of ecological resilience, alternate states, and the role of biodiversity in ecosystem function using theoretical models, shallow lakes, and grassland ecosystems. The applicability of these concepts to forests, which are dominated by long lived species, is uncertain. Further, these changes cannot be separated from the compounded disturbances (Paine et al. 1998) affecting forest composition and function, including climate change, nitrogen deposition and acidification, altered natural disturbance regimes, overpopulation of herbivores, and land use changes (Likens et al. 1996, Vitousek et al.

1997, Nowacki and Abrams 2008, McEwan et al. 2011).

Across the eastern United States, many forests are experiencing widespread change in composition and function as eastern hemlock (Tsuga canadensis, hemlock) is eliminated by an invasive pest insect, hemlock woolly adelgid (Adelges tsugae, HWA). Intact hemlock forests are unique evergreen ecosystems within the largely deciduous forest region of eastern North

America, characterized by a damp, shady microclimate and slowly decomposing acidic litter.

Hemlock communities are more widely distributed on the landscape in the northern portion of their range, while south of approximately the state of Pennsylvania, hemlock generally dominates riparian and cove forests of the Appalachian Mountains and foothills (Ellison et al.

2005). Compared to adjacent deciduous forests, transpiration in hemlock forests is more evenly distributed throughout the year, with higher rates in the spring that moderate seasonal extremes

(Ford and Vose 2007). Hemlock is also an important structural component, providing regeneration sites protected from deer browse (Krueger and Peterson 2006) and slowly decaying

121

large wood in adjacent streams (Morris et al. 2007). Unique suites of salamanders, cold water

fish, aquatic macroinvertebrates, and birds inhabit hemlock forests, contributing to landscape

diversity (Snyder et al. 2002, Tingley et al. 2002, Ross et al. 2003).

Hemlock is characterized as a foundation species, one that exerts a dominant influence on

ecosystem processes and creates a specific habitat for other species (Bruno et al. 2003, Ellison et.

al 2005). Communities dominated by foundation species are thought to be structured by a few

strong interactions, and therefore, the loss of the foundation should result in a rapid shift to an

alternate community state with distinctly different functional processes (Ellison et al. 2005).

Generally, alternate state theory predicts ecosystems moving from one state to another. However,

increasing evidence indicates alternate state transitions can be more complex (Fig 5.1, Fig 5.2,

Houseman et al. 2008, Thrush et al. 2010). When a foundation species is removed across a range

that includes wide ecological gradients, the resulting communities may diverge, resulting in

multiple alternate states with distinct ecosystem composition, structure, and function (Houseman

et al. 2008, Thrush et al. 2010). Unfortunately, the ongoing invasion of HWA presents an

opportunity to examine the rate and pattern of changes in the forest community composition,

resource availability, and nutrient cycling.

An aphid-like insect, HWA was accidentally introduced to a nursery in Richmond,

Virginia in 1951. Spread primarily by wind and animals at an average rate of 12 km·yr-1, rates of spread and mortality are accelerated in the warmer climates of the southern portion of the range

(Evans and Gregorie 2006), and forests from southern Maine to Georgia have been impacted.

HWA feeds on the xylem ray parenchyma cells, which causes a progressive loss of needles and buds to branches, leading to eventual tree mortality (McClure 1991). Near complete mortality is estimated to occur between four and fifteen years of invasion (McClure 1991). In New England,

122

HWA experience higher winter mortality which seems to slow hemlock decline (Skinner et al.

2003, Orwig et al. 2012), particularly when compared to the southern Appalachian Mountains

(Ford et al. 2012). Dukes et al. (2009) estimate an average temperature below -5° C or minimum temperatures below -28.8° C to be the lethal threshold, but project that such temperatures are likely to be increasingly less common due to climate change. Likewise, while elevation and physiographic position may initially slow the impact, duration of invasion is ultimately the most important driver of mortality (Orwig et al. 2008, Orwig et al. 2012, Chapter 4). There is no evidence of resistance, and while horticultural oils and systemic insecticides are effective in treating individual trees (Onken and Reardon 2005), it is costly and labor intensive.

Experimental releases of predatory beetles from the HWA native range in Asia that might provide a landscape scale biological control are ongoing (Onken and Reardon 2005), but many forests are already experiencing high levels of hemlock mortality (Ellison et al. 2005).

Once HWA arrives, shifts in functional processes are rapidly apparent, and hemlock productivity declines almost immediately (Nuckolls et al. 2008). Nutrient cycling rates also increase even before mortality occurs, as throughfall chemistry changes and both nitrogen cycling and decomposition rates accelerate (Jenkins et al. 1999, Cobb et al. 2006, Stadler et al.

2006, Orwig et al. 2008). While functional processes shift rapidly, as predicted by foundation species theory, compositional shifts seem to occur over decades as mortality progresses. For example, in the Smoky Mountains of Tennessee, Krapfl et al. (2011) did not detect any compositional shifts following six years of HWA infestation. Even as mortality proceeds, hemlock is a slow decaying species and thus, snags may remain for years or decades. The length of the transition in both structural and functional processes is not very clear, but it seems likely

123 that functional processes will fluctuate while during vegetation regeneration, and ultimately be driven by the communities that develop.

As hemlock declines, there is some evidence that communities are diverging at local and regional scales, influenced by environmental gradients. Comparisons between forests in New

England and the southern Appalachians already indicate some differences. In New England, hemlock is largely being replaced by sweet birch (Betula lenta) (Orwig and Foster 1998,

Kizlinski et al. 2002), but Small et al. (2005) found different species composition on xeric ridges compared to mesic ravines in Connecticut. In North Carolina, Ford et al. (2012) found a mix of maple (Acer), birch (Betula), beech (Fagus grandifolia) and oak (Quercus). Further, rhododendron (Rhododendron maximum), a native evergreen shrub, will be very influential in the central and southern Appalachians. Rhododendron is known to document gap dynamics, nutrient cycling, and hydrology (Clinton et al. 1994, Kominoski et al. 2007, Ball et al. 2008,

Ford and Vose 2007, Beckage et al. 2008, Wurzberger and Hendrick 2008).

To investigate whether communities are diverging and to further clarify patterns in compositional and functional changes as hemlock declines, we examined forests across the central Appalachian region, representing an HWA invasion chronosequence. This region includes forests near the site of the original introduction, where HWA was likely introduced in

1979, as well as forests were invaded within the last decade. Our central hypothesis was that as a foundation species, hemlock was the most important driver of forest dynamics. Therefore, we expected to see 1. Evidence of a change in community composition, including increased divergence in communities impacted by HWA the longest; 2. Changes in resource availability, productivity, and nutrient cycling as hemlock declines.

124

5.3 Materials and Methods

Study sites

We selected study sites along 30 headwater streams across the central Appalachians in

West Virginia and Virginia (Table 1). Sites were selected as representative of relatively undisturbed (aside from HWA) hemlock dominated riparian forests, located from the species range map compiled by Prasad and Iverson (1999) and recommendations of local managers.

Hemlock dominance was assessed by the presence of live hemlock trees and snags. Counties where sites were located have been impacted by HWA for 9-32 years (Fig. 5.3). As we did not have detailed historical documentation for many areas, we may have excluded former hemlock dominated riparian forests where succession is more advanced.

Sites were located on lands managed by the United States Department of the Interior

National Park Service, the United States Department of Agriculture Forest Service, the West

Virginia Division of Natural Resources, and Virginia Chapter of The Nature Conservancy. The region included the variation in hemlock forest physiographic and environmental contexts, with elevations from 363-1057 m, slopes from nearly level to 90%, and all aspects represented. All sites were dominated by hemlock, which foundation species theory suggests should be more influential in ecosystem processes than the local environment.

In West Virginia, ten sites occurred on the Appalachian Plateau in Carnifex Ferry State

Battleground Park (Nicholas County), the Gauley River National Recreation Area (Nicholas

County), New River Gorge National River (Fayette County) and Monongahela National Forest

(Nicholas, Greenbrier, and Pocahontas counties). The Appalachian Plateau is a region of highly dissected topography of deep and narrow valleys and ridges with dendritic drainage, with Ultisol and Inceptisol soils underlain by sandstone (Fralish 2003). The climate of the region is

125

characterized by cold, snowy winters, but without persistent snow cover, particularly in the

valleys. Precipitation is evenly distributed throughout the year, but is heavier on windward, west-facing slopes, and averages between 1041-1479 mm. Temperatures also vary with elevation, but average daily temperature is between -5- 1°C in the winter, with minimums in the

range of -10- -5°C. Summer temperatures average approximately 17- 21°C (Gorman and Espy

1975; Carpenter 1992, Flegel 1998; Flegel 2007). We located five additional sites in the

Allegheny Mountains at the United States Department of Agriculture Forest Service Fernow

Experimental Forest and within the Monongahela National Forest (Tucker County). This section

of the Allegheny Mountains receives approximately 1450 mm precipitation, correlated with

increasing elevation, and with more falling in June and less in January (Gilliam et al. 1996).

In Virginia, we located fifteen streams in the Ridge and Valley physiographic province in

the George Washington National Forest and The Nature Conservancy’s Warm Springs Preserve

(including Alleghany, Augusta, Bath, Botetourt, and Rockingham counties). The Ridge and

Valley province is characterized by fold mountains, a series of resistant sandstone ridges

separated by level shale and limestone valleys with Ultisols and Inceptisols (Fralish 2003).

Winter temperatures average approximately 0-2.5°C, with minimums averaging in the range of -

6- 4°C. Summer temperatures average 21- 23°C. Approximately half of the annual precipitation

occurs during the growing season, with an average range of 914-1270 mm which increases along

an elevation gradient (Cook and Slabaugh 2006; Wolf and Thomas 2006).

Environmental measurements

At each site, we established three transects parallel to the stream at distances of 10, 30

and 50 m from the bank-full channel. The side of the stream valley or ravine chosen for sampling

126

was determined haphazardly, in some cases based the location of cliffs. In each transect, we used

a series of five 100-m2 circular plots where we recorded basic physiographical data: slope percent (using a clinometer), slope shape, slope position and aspect. A Garmin eTrex GPS

(Garmin Ltd, Kansas, USA) unit was used to record the latitude, longitude, and elevation of each plot. Aspect, along with slope percent and latitude were later used to calculate metrics of direct incident radiation for each plot using Equation 1 from McCune & Keon (2002).

Soil samples were collected from the center of each transect at the 23 sites where it was permitted (permission to collect soil samples not granted at sites in the Monongahela National

Forest (seven sites); thus, those sites are not included in any analyses that include soil). At each site where soil was sampled, three 5-cm diameter cores were taken to a depth of five cm . At sites with many boulders, soil was taken to the deepest possible depth and additional cores were sometimes added to ensure adequate sample for analysis. Soil samples were kept on ice and transported in coolers to the laboratory, where they were dried at 40°C and sieved through a two mm screen prior to analysis. The hydrometer method was used to determine soil particle size fractionation (American Society for Testing and Materials 1985), and pH and CEC determined using an Orion pH meter (ThermoFisher Scientific, Waltham, MA, USA) and ammonium acetate extraction, respectively. Total carbon and nitrogen was determined using an Elementar Vario-

Max CN analyzer, while major elements (phosphorus, potassium, calcium and magnesium) were determined using digestion with HClO4/HNO3 followed by inductively coupled plasma (ICP)

emission spectrometry using a Teledyne Leeman Labs Prodicgy Dual view ICP (AOAC Official

Methods of Analysis 2002). All soils analyses were conducted by the Ohio Agricultural

Research and Development Center (OARDC) Service Testing and Research (STAR) lab.

127

Vegetation community composition

In each plot, we recorded all woody species stems >2.5 cm dbh and measured the dbh

(diameter-at-breast-height, 1.37 m), and classified stems 2.5- 10 cm dbh as saplings, and while

those > 10.0 cm dbh were considered overstory. Each hemlock individual was assigned a

canopy decline class based on the estimated foliage remaining, as outlined by Orwig and Foster

(1998) 1: >75%; 2: 51-75%; 3: 26-50%; 4: 1-25%; 5: dead. Seedlings (woody trees and shrubs

<2.5 cm dbh) were identified and counted in central 4-m2 subplots. As many plots had a high density of shrubs with stems < 2.5 cm dbh, overall shrub cover was estimated for each plot. Data from the five plots were combined (added), creating one value for each transect. Environmental metrics were averaged across each transect. Nomenclature follows the USDA PLANTS Database

(USDA and NRCS 2012).

Metrics of ecosystem function

As metrics of ecosystem function along our chronosequence of HWA invasion, we collected data relevant to resource availability and nutrient cycling. Specifically, we measured light availability, leaf litter biomass and chemistry, and the relative decomposition rate. Light availability was estimated in the growing season (June- September) and dormant season

(December- March) by analysis of hemispherical photos taken 1.2-m above the ground in the center of each 100-m2 plot using a Nikon COOLPIX E8400 (Nikon Corporation, Tokyo, Japan )

equipped with a fisheye lens (Nikon LC-ER2) and mounted on Self-leveling O-Mount (Regent

Instruments, Canada) . Canopy openness and leaf-area-index (LAI) was determined using

WinSCanopy digital image processing software (Regents Instruments, Canada), which allows

exclusions of obstructions in photos, such as those created by cliffs or boulders. Canopy

128

openness is a metric of the amount of open sky captured in a hemispherical photograph,

accounting for the transformation caused by the equilinear projection of the fisheye lens (Regent

Instruments 2007).

Leaf litter fall was collected as an index of productivity (determined by biomass) and

nutrient availability (carbon to nitrogen ratio). In September prior to leaf fall, one leaf litter trap constructed from a 50 X 40 X 24 cm rectangular plastic laundry basked lined with nylon mesh

was secured to the ground using metal stakes near the center of each transect. Baskets were

removed in November once most leaves had fallen, but prior to winter storms that make some

sites inaccessible. In some cases, baskets were found tipped sideways, destroyed, or missing, due to wildlife activity. To ensure enough litter material for chemical analysis, litter from tipped baskets was included in chemical and species richness measures, but were not included in analyses of biomass. In some cases where the basket was missing or destroyed, an area of fresh litter equivalent to the area of the basket was collected from the basket location for chemical analysis and to measure species richness, but was not included in biomass estimates. In all cases, it is possible some litter, particularly conifer needles, was lost. Litter was dried at 40°C to constant mass.

Relative decomposition rate was determined as a standard index of nutrient cycling

(Kizlinski et al. 2002). Three replicates made of approximately three grams of cellulose paper enclosed in 1-mm2 nylon mesh and identified with a metal tag were weighed in the lab to the

nearest 0.01 gram, then placed in the field at ground level in the center of each transect in June,

secured by wire and a metal stake. Packs were collected in November, dried at 4° C until

constant mass and weighed. Mass loss of all intact packs was averaged and decomposition rates

129

between transects was compared using the decay constant k, calculated from: ln (initial mass-

end mass)= -kt where t= number of days.

Statistical Analyses

As hemlock declines, we expected to find a shift in community diversity and composition

and an increase in community variance, indicating multiple pathways of community

reorganization. Hemlock communities are low diversity, which our previous work suggests is

due in part to the direct influence of hemlock dominance (Chapter 3). Therefore, species richness

and diversity should increase as hemlock declines. As transects were nested within sites, changes

in richness and diversity in the overstory, sapling, and seedling layers were examined using

linear effects mixed models calculated using the lme procedure from the R package nlme

(Pinherio et al. 2012) with distance from the stream (10, 30, and 50m) as a nested factor within site, a random factor. We compared models using only the decline, decline and distance, and the two main effects with an interaction and selected the model with the smallest log likelihood value.

To examine patterns of community composition in the overstory and sapling layers, we used non-metric multidimensional scaling analysis (NMDS). NMDS is an unconstrained ordination method regarded as highly robust for community data (Minchin 1987). We used the metaMDS procedure with a Bray-Curtis distance matrix using the vegan package for community ecology (Oksanen et al. 2003). Species that appeared on less than five percent of transects (4 of

90) were excluded to focus on general patterns (total of 13 overstory, 31 sapling and 17 seedling species). In many cases, species found infrequently in the sapling layer were common overstory species. The metaMDS procedure automatically transforms data with large ranges using square root and Wisconsin double standardization. The metaMDS procedure uses an iterative approach

130

with random starts to insure against entrapment by local minima while converging on a solution

that minimizes stress. The procedure also rotates the final solution so that the first axis explains the greatest variance. In the NMDS diagrams, each site was categorized by distance and average

level of hemlock decline at the site. To visualize possible shifts, we added confidence ellipses

around the centroid of each decline category. We fit environmental principal component vectors

to the NMDS diagrams using the envfit procedure in the vegan package to explore correlation

with environmental gradients.

To further examine whether the overstory, sapling, and seedling communities were

different beyond the gradient analysis approach, we used a permutational multivariate analysis of

variance using the adonis procedure in the R package vegan. A multivariate equivalent of

analysis-of-variance, adonis can examine differences between groups where the response

variable is a distance matrix, such as a matrix based on community composition. To examine

within and between site factors, we tested the effect of decline, distance, and the interaction of

decline and distance on Bray-Curtis comparisons of canopy composition. To identify changes in

community variance (dispersion) as hemlock declines, we tested multivariate homogeneity of

group variance using the betadisper function with a Bray-Curtis distance matrix in the R package

vegan. Betadisper is a multivariate analogue of Lavene’s test for homogeneity of variance that

examines distances between points and group centroids, and has been used to determine whether

disturbances increase community variance (Houseman et al. 2008).

As community composition begins to diverge, we expected functional processes would

become more variable across the range of hemlock sites sampled in the central Appalachians. As

species are known to influence soil chemistry (Finzi et al. 1998, Lovett et al. 2004), we used

betadisper as a multivariate test for homogeneity of variance based on Euclidean distances in the

131

soil cation composition (P, K, Ca, Fe, Na, Mg, Mn, S, Al, B, Cu, and Zn) among average

overstory decline categories. For other functional measures where we had a single response

variable, we instead used Levene’s test to compare soil metrics (pH, cation exchange capacity,

carbon to nitrogen ratio), light availability (LAI and canopy openness in the growing and

dormant seasons), leaf litter (biomass and carbon to nitrogen ratio), and decomposition constant

k.

Beyond the possible divergence in functional processes, we were also interested in

quantifying the influence of hemlock decline on these processes. The influence of hemlock

decline is likely complex, particularly in terms of functional processes that likely vary by

environmental factors and community composition. Therefore, we examined the functional

response of hemlock decline using structural equation modeling (SEM). A multi-equational

approach, SEM uses maximum likelihood to solve path equations simultaneously and can

highlight indirect effects, interactions, and reciprocal relationships between variables (Grace

2006, Grace 2008). In the SEM, community composition of the vegetation layers was

represented by the scores of the first axis of the NMDS, reducing the dimensionality to a single

variable (Grace 2006, 2008). Models were constructed based on previous analyses and preliminary explorations of the data, including scatter plots of variables. Candidate models were

considered an acceptable fit when: 1. The chi-square fit statistic indicated the model was not a

significant departure from the data (Χ2 p > 0.05), and 2. Model pathways were significant (p <

0.05).

132

5.4 Results

Patterns of community composition and dispersion

Plant species richness in hemlock riparian forests is lower when compared to surrounding

deciduous forest, with few species aside from hemlock contributing to total basal area (Table 2).

Across all transects, the most common overstory species following hemlock were sweet birch,

tulip-poplar (Lirodendron tulipifera), chestnut oak, yellow birch (Betula alleghaniensis), red maple (Acer rubrum), red spruce (Picea rubens), and white oak (Quercus alba). In the sapling layer, rhododendron (Rhododendron maximum) was the most dominant followed by hemlock

(Table 3). Wild grape (Vitis aestivalis) and red maple were the only other species contributing more than five percent of the relative basal area. Red maple, rhododendron, sweet birch, red spruce, mountain laurel (Kalmia latifolia), and red oaks (Quercus rubra, Q. coccinea, Q. velutina) were most common in the seedling layer (Table 4).

As hemlock declines, linear effects mixed models indicate an increase in species richness.

In the overstory, a model with overstory hemlock decline as the sole predictor variable was the best fit and indicated a significant difference (t= 2.15, df= 28, p= 0.04). For overstory diversity, the log liklihood estimates indicated a model with canopy decline, distance from the stream, and an interaction between canopy decline and distance from the stream as most likely, but there were no significant differences. In the sapling layer, there were no changes in species richness

(best estimated by canopy decline alone) or diversity (best estimated by canopy decline, distance, and an interaction between them). In the seedling layer, a significant increase in richness (t=

3.02, df= 28, p= 0.005) was best estimated by a canopy decline, but there was not a change in diversity (the best fit included decline, distance, and an interaction).

133

Across a chronosequence of HWA impact in the central Appalachians, our analyses indicate shifts in both composition and dispersion. In the overstory, NMDS analyses (final stress value 21.83) visually illustrate differences in group averages, identified by confidence ellipses, and dispersion, identified by increasing scatter by average decline category (Fig 5.4). When we fit environmental vectors to the NMDS diagram, the first axis was associated most strongly with latitude and elevation, but also with duration of HWA infestation, hemlock decline, and undulating slope shape. The second axis was mostly associated with slope (Fig 5.5). Hemlock was near the origin, indicating some ubiquity across sites, but somewhat associated with latitude and elevation, and opposite decline and duration on HWA invasion. A number of species were particularly associated with the duration of HWA invasion and hemlock decline including oaks

(Quercus prinus, Q. rubra), blackgum (Nyssa sylvatica), and hornbeam (Carpinus caroliniana), sourwood (Oxydendrum arborum). Sugar maple (Acer saccaharum), tulip-poplar, and sassafras

(Sassafras albidum) were associated with steeper slopes as well as hemlock decline. Sweet birch was associated with steeper slopes, and yellow birch was found with increased elevations. A typical shade tolerant component of hemlock forests, American beech (Fagus grandifolia) occurred near streams. Of the other evergreen species, red spruce was found at higher elevations, and white pine (Pinus strobus) was associated with the 50-m transect. Permutational multivariate analysis of variance identified a significant effect of hemlock decline on overstory composition (F= 3.59, df= 1, p= 0.003), but distance was not a significant factor, and there was not a significant decline by distance interaction. Analysis of the multivariate homogeneity of variance indicated significant difference between average site decline categories (Fig 5.6, F=

10.90, df= 4, p <0.001), and Tukey’s honestly significant difference illustrate significant difference in variance as decline progresses (Fig 5.7).

134

Patterns of change were not quite as clear for the sapling layer NMDS (final stress

19.42), with the largest confidence ellipse and perhaps greatest variation in category 1,

representing sites where overstory canopies were >75% intact (Fig 5.8). When we fit

environmental vectors, the first axis was weakly associated with latitude and the second axis was

weakly associated with the overstory decline (Fig 5.9). Hemlock saplings were again near the

origin, but were associated with overstory hemlock decline, particularly at the 30-m transect, and

sugar maple was the species most closely associated with length of HWA invasion.

Rhododendron was most associated with the 10-m transect (near the stream), and a number of species were associated with the 50-m transect including white pine, mountain laurel, blackgum, sourwood, witch hazel (Hamamelis virginiana), wild grape. Yellow birch was associated with increasing elevation, as was sweet birch. Permutational multivariate analysis of variance identified a difference in sapling composition by average overstory decline (F= 2.2, df= 1, p=

0.048), which differed by distance (F= 2.56, df= 2, p= 0.006), but no significant interaction (F=

1.53, df= 2, p= 0.127). In agreement with the visual dispersion identified in the NMDS, analysis of multivariate homogeneity of variance was not different among average decline classes (F=

1.99, df= 4, p= 0.10).

In contrast to the sapling layer, seedling communities seem to be more strongly

associated with environmental gradients. In the NMDS (Fig. 5.10, Fig. 5.11, final stress 24.54),

the first axis was associated with overstory hemlock decline, duration of HWA invasion, and

latitude, longitude, and elevation, and the second axis was associated with incident radiation.

Slope shape and position were also influential, with both axes associated with terraces and the

hillslope. Rhododendron was associated with hillslopes, but was not near other species. Several

species were associated with overstory hemlock decline and duration of HWA invasion,

135

including white pine, striped maple (Acer pennsylvanica), hornbeam, hickories (Carya spp), and

the invasive tree of heaven (Ailanthus altissima). Hemlock, yellow birch, and red spruce were

associated with higher elevations. Certain species, including sweet birch, tulip-poplar, and red

oaks were located near the origin, indicating more ubiquitous distribution. Permutational multivariate analysis of variance identified a difference in seedling composition by average overstory decline (F= 5.13, df= 1, p= 0.003), but no difference by distance (F= 1.18, df= 2, p=

0.273), and no significant interaction (F= 0.073, df= 2, p= 0.761). Analysis of multivariate homogeneity of variance was not different among average decline classes (F= 0.419, df= 4, p=

0.795).

Influence of hemlock decline on functional processes

We did not find clear evidence of increasing variation in functional ecosystem metrics with increasing hemlock decline. A multivariate homogeneity of variance test for soil cation composition (P, K, Ca, Fe, Na, Mg, Mn, S, Al, B, Cu, and Zn) based on Euclidean distance did not differ by average overstory decline class. Levene’s test for homogeneity of variance did not indicate any significant differences in variance by overstory decline class in the soil (pH, cation

exchange capacity, carbon to nitrogen ratio, percent organic matter), light availability during the

growing season (LAI, canopy openness), leaf litter (carbon to nitrogen ratio, biomass),

decomposition constant k. During the dormant season, variance in canopy openness differed by

average overstory hemlock decline (F= 2.99, df= 4, p= 0.02). Variance was higher with

increasing decline, but the most variance occurred in categories 5 (mostly dead) and category 3

(25-50% hemlock canopy remaining).

136

Using prior analyses and preliminary examinations of functional metrics, we fit four structural equation models relating the impacts of HWA to ecosystem function. To understand the relative strengths of environmental conditions and hemlock influence on soil properties, we fit an SEM of soil cation exchange capacity and soil carbon to nitrogen ratio using data from only the 23 sites where we collected soil (Fig. 5.12). The model explained almost half of the variance in cation exchange capacity (R2= 0.46), which was directly influenced by the soil texture (clay particle size fractionation -0.46; sand particle size fractionation -0.26) and the community composition of the overstory (-.20). Indirectly, cation exchange capacity was influenced by the years of HWA invasion (-0.09) and hemlock dominance (measured by relative basal area, 0.15), through their influence on overstory composition. Thus, total effects of HWA invasion (-0.09) on cation exchange capacity were outweighed by the total effects of hemlock dominance (0.15). Hemlock dominance was more influential (-0.75) on overstory community composition than the total effect of years of HWA invasion (0.44), which had a direct effect

(0.20), as well as an indirect effect (0.02), as hemlock dominance was negatively influenced by years of HWA invasion (-0.32). The model explained much less of the variation in soil carbon to nitrogen ratio (R2= 0.11), but like cation exchange capacity it was influenced by a combination of soil texture (clay particle size fractionation, 0.27) and the overstory community composition

(0.34). For soil carbon to nitrogen ratio, the total effects of hemlock dominance (-0.18) outweighed the total effects of years of HWA (0.11).

Canopy openness during the dormant season (Fig 5.13,. Χ2= 6.20, df= 3, p= 0.10) can be explained (R2= 0.17) by direct effects of the canopy composition (0.52), years of HWA invasion

(0.46), and relative basal area of red spruce (0.16). Indirectly, canopy openness during the dormant season was influence by years of HWA invasion (0.24), elevation (-0.19), and red

137

spruce basal area (-0.16). Total effects indicate years of HWA invasion was most influential factor (0.70), followed by canopy composition (0.52), elevation (-0.19), and a small effect of red spruce relative basal area (0.001).

The decline of hemlock was also an influential factor for metrics of nutrient cycling and availability, the relative decomposition rate k (R2= 0.27) and leaf litter carbon to nitrogen ratio

(R2= 0.23) (Fig. 5.14). Hemlock decline had a slightly stronger influence on decomposition rate

(0.30) than leaf litter carbon to nitrogen ratio (-.20) and was influenced by years of hemlock invasion (0.54). Thus, years of hemlock invasion had indirect effects on decomposition rate k

(0.16) and leaf litter carbon to nitrogen ratio (-0.14). As this model was fit on all sites, we were not able to explore the influence of soil factors which were only available for a subset of study sites.

5.5 Discussion

Results from thirty sites widely distributed across the central Appalachians support

findings from studies across other areas experiencing hemlock decline, but also highlight the

complexities in the impacts of this novel disturbance on ecosystem composition and function.

As hemlock declines, we expected to find changes in the composition of the vegetation layers, in

part due to divergence across central Appalachian sites that vary in physiographic context,

elevation, soil type, and potentially land use history. We expected species richness to increase as

communities shift from hemlock dominance, which was true in the overstory, although diversity

indices did not differ. In the overstory, community composition also exhibited the expected

change and divergence. This change is likely due to a combination of hemlock decline and

environmental context, supported by the NMDS of overstory species and environmental vectors.

A number of overstory deciduous species were associated with hemlock decline, while others

138 were more closely associated with environmental factors. In New England, birch is expected to increase following hemlock decline (Orwig et al. 2008). Across the central Appalachian sites, sweet birch was more associated with steeper slopes than hemlock decline. Perhaps similar ecologically, yellow birch was more closely associated with increased elevations. This suggests that birch may increase following hemlock decline, but probably not across all landscape positions. In the southern Appalachians, Ford et al. (2012) suggest hemlock will be replaced with a mix of maple, birch, beech, and oak. This is also likely in the central Appalachians, but our analyses suggest the distribution of these species vary across physiographic and soil gradients. We found birch and beech were more commonly near the stream, while red maple, oaks, and hickories were more common upslope. In the overstory, tulip-poplar and particularly blackgum were associated with hemlock decline and duration of HWA invasion, which Ford and

Vose (2007) also suggest could increase over time in southern Appalachian riparian forests.

Patterns of change due to HWA were not as clear in the sapling layer. Community composition of the sapling layer seems to be changing with hemlock decline, and differences between transects indicate these shifts differ on local scales. While composition differed with hemlock overstory decline, we did not find differences in pecies richness or diversity. Unlike the overstory, we did not find an increase in variance in the sapling layer, which was somewhat contrary to our expectation that communities are diverging as hemlock declines. Sapling layers were probably quite different in hemlock forests throughout the central Appalachians prior to

HWA invasion. Healthy hemlock forests often have a very sparse sapling layer (Chapter 3), which may indicate it is driven by factors such as tree fall gaps. Further, in the central

Appalachians, some hemlock forests have a dense layer of ericaceous shrubs, particularly rhododendron, while shrubs are absent from other areas. In the sapling NMDS, rhododendron

139 was associated with transects nearest to the stream, while the other major contributor to shrub layers, mountain laurel, was associated with the upslope transect. In the sapling layer NMDS, species were only weakly associated with environmental gradients and overstory hemlock decline, indicating other factors such as shrub abundance and overstory composition may be influential. Additionally, soil characteristics may be particularly important, as both rhododendron and mountain laurel are associated with acidic soils but we were unable to collect soil from a quarter of our study sites.

As overstory hemlock declines, the seedling layer composition is changing, most likely due to increases in light and nutrients. NMDS suggest many of the seedling species are associated with duration of HWA invasion and overstory hemlock decline, as well as the 50-m transect. On the other hand, rhododendron separated from the other species and is associated with the 10-m transect. Hemlock, yellow birch, red spruce, and beech are more common at higher elevations and in the 30-m transect. While we did not find an increase in variance in the composition of the seedling layer, like the sapling layer this may be due to a high variance even within intact hemlock forests. Many species were associated with environmental gradients and differed across the transects, which supports different community compositions at local scales.

Similar species seem to occur in hemlock forest communities throughout its range, but there is not a great deal of information comparing community composition at different scales.

Exploring community divergence has not been a primary objective of the majority of HWA studies, thus how these communities may be diverging at local and regional scales as impacts of

HWA progresses is not clear. In a study of eight uninvaded hemlock forests on the Unglaciated

Plateau of Ohio, we found sweet birch, beech, and tulip-poplar more commonly near streams, red and sugar maple upslope, and oaks (white oak, Northern red oak, chestnut oak) more abundant

140

50 meters from streams (Chapter 3). In eight hemlock stands in the Connecticut River Valley,

Orwig et al (2008) found red oak, sweet birch, and red maple, but none of these were significant contributors to the overall basal area of sites, which were 69-89% hemlock. Previously, at a different set of sites (except one) in southern Connecticut, Orwig and Foster (1998) also found black oak (Quercus velutina) and chestnut oak, which indicates some differences in community composition. Compositional differences that result in divergence between communities in southern Connecticut is further supported by Small et al. (2005) who followed changes in forests at the Connecticut Arboretum that lost 70% of their hemlock basal area over twenty years of

HWA infestation. On xeric ridges, black oaks (grouped Q. rubra, Q. velutina, and Q. coccinea) increased while distinct mixed canopies developed in mesic ravines. Also in their study, the relative basal area rankings of species remained relatively similar, but many become more dominant, particularly beech.

Evidence from the Appalachian Mountains indicates hemlock communities may be compositionally different from New England forests, and this may vary with environmental context. As in New England, hemlock is often associated with birch, oaks, and maples.

However, central and southern Appalachian hemlock forests may be more diverse, and hemlocks dominate a number of associations that have been characterized as different forest types (Krapfl et al. 2011, Spaulding and Rieske 2010). Some of these hemlock communities are unique associations, such as the hemlock-silverbell (Halesia tertaptera) forests of the Great Smoky

Mountains (Dhungel et al. 2010). Across forest types, in addition to the oaks, sweet birch, and beech found at sites in New England ,yellow birch, sugar maple, beech, tulip-poplar, basswood

(Tilia americana), sassafras and blackgum are often common in the overstory as well as the sapling layer (Krapfl et al. 2011, Spaulding and Rieske 2010, Ford et al. 2012). Also unlike New

141

England, central and southern Appalachian hemlock forests often have a significant shrub layer, particularly rhododendron. The shrub layer and overstory and sapling composition are likely structured along environmental gradients. At Great Smoky Mountains National Park, Krapfl et al. (2011) examined plots distributed across six forest ecogroups containing hemlock, and found yellow birch, red maple, basswood, and chestnut oak were the most common overstory species following hemlock (or in once case exceeding hemlock), but their distribution varied by ecogroup.

In addition to shifting compositions of native species, forests experiencing hemlock decline may provide habitat for invasive plant species. For example, Small et al. (2005) found the first invasive species in a 50-year record at the Connecticut arboretum following the arrival of HWA. Likewise, Eschtruth et al. (2006) found invasive plant species on 35% of their plots after nine years of hemlock decline at the Delaware Water Gap National Recreation Area in

Pennsylvania and New Jersey. In this assessment of woody species, we found one invasive, tree of heaven, in the sapling layer, although it occurred on less than 5% of plots in the overstory. In the NMDS, tree of heaven was not closely associated with any environmental vectors. It is most likely dependent on proximity to seed sources, and a few sites with very abundant tree of heaven saplings were closer to roads, towns, and campgrounds, which are factors worthy of closer examination. In the ground flora, Japanese stiltgrass (Microstegium vimineum) was present, but mainly at the sites with abundant tree of heaven. There were no invasives present at the majority of our sites, and further investigation of the factors that contribute to the presence or absence of invasive recruitment as hemlock declines presents an opportunity for future studies.

In non-invaded hemlock forests, hemlock is influential in determining resource availability, productivity, and nutrient cycling (Ellison et al. 2005). In a study of uninvaded

142 hemlock forest on the Unglaciated Allegheny Plateau of Ohio, we found hemlock dominance suppresses light availability, measured by growing season canopy openness, and productivity, measured by leaf litter biomass (Chapter 3). As this influence wanes, the reorganization of functional processes will depend on varied community compositions that develop following hemlock mortality.

Across our gradient of overstory hemlock decline, we found changes in functional processes, which were dependent on environmental context. Our data support the hypothesis that hemlock decline will result in an increase in resource availability. At the subset of sites where we were able to collect soil samples, we found changes in cation exchange capacity and soil carbon to nitrogen ratio, determined by a combination of factors, including overstory composition and soil texture. Our study is consistent with the lower soil carbon to nitrogen ratio

Orwig et al. (2008) found in invaded sites when compared to uninvaded forests, but they also identified clear increases in soil pH and decreases in soil organic matter, which were not apparent in our data. Consistent with other studies, we found evidence that hemlock decline results in increased nutrient cycling rates (measured by relative decomposition constant) and decreased leaf litter carbon to nitrogen ratio (Jenkins et al. 1999, Kizlinski et al. 2002, Orwig et al. 2008, Cobb 2009, Knoepp et al. 2011). Although nutrient cycling may be increasing, productivity, measured by leaf litter biomass, did not exhibit any clear patterns in our study. This is likely because of a complex interaction of factors influencing productivity and leaf fall as well as the divergence we found in overstory composition. In some sites, deciduous species productivity may be increasing (Nuckolls et al. 2008), but in others, rhododendron may limit tree recruitment and productivity. In addition, we found very little rhododendron leaf litter in our traps, and this metric of productivity likely varies with shrub cover. We could not fit a structural

143

equation model explaining this relationship, which is likely because complex compounding factors make it non-linear. For instance, productivity may initially decline, then increase as new

individuals exploit the increased availability of light and nutrients as hemlock dies.

Across all our sites, as hemlock declines and the canopy composition changes, there was an increase in canopy openness, and thus light availability, during the the dormant season, which is expected as the forest transitions to a more deciduous composition. This change was modified by the presence of the evergreen red spruce at higher elevations. It is also likely that

rhododendron is influential in determining light availability during both the growing and dormant seasons. In the southern Appalachians, Ford et al. (2012) found increases in light transmittance, particularly during the dormant season, in HWA invaded sites when compared with a girdling treatment. By measuring light at one and five meters above the ground, Ford et al.

(2012) also identified that evergreen shrubs (mainly rhododendron) play a significant role in intercepting light. We were unable to quantify the strength of rhododendrons’ influence on light availability across our sites, likely due to the distribution of rhododendron. While some transects had a dense layer of rhododendron, it was completely absent from others. In the sapling and seedling layers, rhododendron was associated with transects closest to the stream, but this does not necessarily explain its distribution on a regional scale.

The majority of studies examining changes following HWA invasion have been in forests without a significant component of evergreen canopy species. However, evergreen species are commonly associated with hemlock in some parts of its range, particularly at northern latitudes and higher elevations (Evans et al. 2011). Potential functional redundancy between hemlock and

other evergreen canopy species has not been explored extensively, most likely because the

spread of HWA is slowed by cold winter temperatures and thus, it is not yet present across areas

144

containing significant portions of spruce (Picea) and fir (Abies). Pines are occasional species in

many studies, with white pine abundant in some areas (Evans et al. 2011) and, its role in forests

impacted by HWA is not very clear. In our study, red spruce was a significant component at six sites that occurred at elevations 869-1051 meters, and was a significant influence on functional processes, particularly light availability during the dormant season. Thus, in some forests changes in light levels and perhaps additional functional processes may be modified by red spruce.

A more complete understanding of feedbacks between community composition, functional process and ecosystem resilience and change is an important frontier in ecology, particularly in complex natural systems (Hooper et al.2005). Compounded disturbances of invasive pests and pathogens, climate change, altered nutrient dynamics and loss of biodiversity are happening at an accelerated rate (Likens et al. 1996, Vitusek et al. 1997, Paine et al. 1998,

Lovett et al. 2006, Gandhi and Herms 2009, Ehrenfeld 2010). Hemlock provides a model foundation species system that encompasses a wide geographic area, and some consequences of its loss are well studied in parts of its range. Some responses such as an increase in nutrient cycling rates may be generalizable, which concurs with the patterns of ecosystem response to invasive species outlined by Ehrenfeld (2010). At the same time, our data suggest changes will be complex and vary by environmental context and the presence of some key species including evergreen canopy trees and evergreen ericaceous shrubs in some parts of the range.

Understanding this complexity should contribute to the success of restoration and management planning. For example, although an HWA extension is available for the forest vegetation simulator (FVS), predictions are not likely very accurate for areas with a dense shrub layer, as it does not account for this (Spaulding and Rieske 2010). The entire suite of ecological tools, from

145 simulation models to mechanistic manipulations, and field observation data are necessary to account for the complexity in the response of eastern forests to the loss of hemlock.

5.6 Acknowledgements

Salaries and financial support for this research were provided by the Ohio Agricultural Research and Development Center (OARDC) SEEDS Program, The Ohio State University College of

Food, Agriculture and Environmental Science, and The Ohio State University. We thank T.

Macy, L. Kobelt, C. Clifton, and J. Martin for field assistance. We are also grateful for the advice and assistance with research permits provided by J. Perez, T. Schuler, M. Smith, W. San Jule, K.

Karriker, T. Slater, E. Haverlack, S. Tanguay, and S. Cowell.

5.7 References

American Society for Testing and Materials. 1985. Standard test method for particle-size

analysis of soils. Pages 117-127 in: Annual Book of ASTM Standards, American Society

for Testing Materials. West Conshohocken, PA, USA.

AOAC Official Methods of Analysis. 2002. Method 990.03. Protein (crude) in Animal Feed

Combusion Method (Dumas method). JAOAC 72:770.

Ball, B.A., M.D. Hunter, J, S. Kominoski, C.S. Swan and M.A.Bradford. 2008. Consequences of

non-random species loss for decomposition dynamics: experimental evidence for additive

and non-additive effects. Journal of Ecology 96:303-313.

Beckage, B., B.D. Kloeppel, J. A. Yeakley, S.F. Taylor, and D.C. Coleman. 2008. Differential

effects of understory and overstory gaps on tree regeneration. Journal of the Torrey

Botanical Society 135:1-11.

146

Bruno, J.,J. Stachowicz, and P.A. Townsend. 2003. Inclusion of facilitation into theoretical

ecology. 2003. Trends in Ecology and Evolution 18:119-125.

Carpenter, S.G. 1992. Soil Survey of Nicholas County, West Virginia. United States Department

of Agriculture Soil Conservation Service. Washington, DC..

Clinton, B.D., L.R. Boring, and W.T. Swank. 1994. Regeneration patterns in canopy gaps of

mixed-oak forests of the southern Appalachians: influences of topographic position and

evergreen understory. American Midland Naturalist 132:308-319.

Cobb, R.C., D.A. Orwig, and S. Currie. 2006. Decomposition of green foliage in eastern

hemlock forests of southern New England impacted by hemlock woolly adelgid

infestations. Canadian Journal of Forest Research 36:1331-1341.

Cobb, R.C. 2009. Species shift drives decomposition rates following invasion by hemlock

woolly adelgid. Oikos 119:1291-1298.

Cook, M E. M. and J. D. Slabaugh. 2006. Soil Survey of Alleghany County, Virginia. United

States Department of Agriculture Natural Resources Conservation Service. Washington,

DC.

Dhungel, E.K., J.W. Groninger, and E.J. Holzmueller. 2010. Tree species and environment

associations with hemlock-silverbell stands treated for hemlock woolly adelgid in Great

Smoky Mountains National Park. Journal of the Torrey Botanical Society 137:401-409.

Dukes, J.S., J. Pontius, D. Orwig, J.R. Garnas, V.L. Rodgers, N. Brazee, B. Cooke, K.A.

Theoharides, E.E. Stange, R. Harrington, J. Ehrenfeld, J. Gurevitch, M. Lerdau, K.

Stinson. R. Wick, and M. Ayres. 2009. Responses of insect pests, pathogens, and

invasive plant species to climate change in the forests of northeastern North America:

What can we predict? Canadian Journal of Forest Research 39:231-248.

147

Ehrenfeld, J.G. 2011. Ecosystem consequences of biological invasions. Annual Review of

Ecology, Evolution, and Systematics 41:59-80.

Ellison, A. M., M. S. Banks, B. D. Clinton, E. A. Colburn, K. Elliott, C. R. Ford, D. R. Foster, B.

D. Kloeppel, J. D. Knoepp, G. M. Lovett, J. Mohan, D. A. Orwig, and N. L. Rodenhouse.

2005. Loss of foundation species: consequences for the structure and dynamics of

forested ecosystems. Frontiers in Ecology and the Environment 3:479-486.

Eschtruth, A. K., N. L. Cleavitt, J. J. Battles, R. A. Evans, and T. J. Fahey. 2006. Vegetation

dynamics in declining eastern hemlock stands: 9 years of forest response to hemlock

woolly adelgid infestation. Canadian Journal of Forest Research 36:1435-1450.

Evans, A.M and T.G. Gregoire. 2007. A geographically variable model of hemlock woolly

adelgid spread. Biological Invasions 9:369-382.

Evans. D.M., W.M. Aust, A.C. Dolloff, B.S. Templeton, J.A. Peterson. 2011. Eastern hemlock

decline in riparian areas from Maine to Alabama. Northern Journal of Applied Forestry

28: 97-104.

Finzi, A.C., N. Van Breemen, and C. D. Canham. 1998. Canopy tree soil interactions within

temperate forests: species effects on soil carbon and nitrogen. Ecological Applications

8:440-446.

Flegel, D.G. 1998. Soil Survey of Pocahontas County, West Virginia. United States Department

of Agriculture Natural Resources Conservation Service. Washington, DC.

Flegel, D.G. 2007. Soil Survey of Greenbrier County, West Virginia. United States Department

of Agriculture Natural Resources Conservation Service. Washington, DC.

148

Ford, C.R. and J.M. Vose. 2007. Tsuga canadensis (L.) Carr. mortality will impact hydrologic

processes in southern Appalachian forest ecosystems. Ecological Applications 17:1156-

1167.

Ford, C.R., K.J. Elliott, B.D. Clinton, B.D. Kloeppel., J.M. Vose. 2012. Forest dynamics

following eastern hemlock mortality in the southern Appalachians. Oikos 121:523-536.

Fralish, J.S. 2003. The Central Hardwood Forest: Its boundaries and physiographic provinces, pp

1-20 in J.W. Van Sambeek, J.O. Dawson, F. Ponder, Jr., E.F. Loewenstein, and J.S.

Fralish [eds]. Proceedings, 13th Central Hardwood Forest Conference; 2002 April 1-3;

Urbana, IL. Gen. Tech. Rep. NC-234. St. Paul, MN: U.S. Department of Agriculture,

Forest Service, North Central Research Station.

Gandhi, K.J.K. and D.A. Herms. 2010. Direct and indirect effects of alien insect herbivores on

ecological processes and interactions in forests of eastern North America. Biological

Invasions 12:389-405.

Gilliam, F.S., M.B. Adams, B.M. Yurish. 1996. Ecosystem nutrient responses to chronic

nitrogen inputs at Fernow Experimental Forest, West Virginia. Canadian Journal of

Forest Research 16:196-205.

Gorman, J.L. and L.E. Espy. 1975. Soil Survey of Fayette and Raleigh Counties, West Virginia.

United States Department of Agriculture Soil Conservation Service. Washington, DC.

Grace, J. B. 2006. Structural equation modeling and natural systems. Cambridge University

Press. Cambridge, UK.

Grace, J.B. 2008. Structural equation modeling for observational studies. Journal of Wildlife

Management 72: 14-22.

149

Hooper, D.U., F.S. Chapin, J.J. Ewel, A. Hector, P. Inchausti, S. Lavorel, J.H. Lawton, D.M.

Lodge, M. Loreau, S. Naeem, B. Schmid, H. Setala, A.J. Symstad, J. Vendermeer, and

D.A. Wardel. 2005. Effects of biodiversity on ecosystem functioning: a consensus of

current knowledge. Ecological Monographs 75:3-35.

Houseman, G.R., G.G. Mittelbach, H.L. Reynolds, and K.L. Gross. 2008. Perturbations alter

community convergence, divergence, and formation of multiple community states.

Ecology 89:2172-2180.

Jenkins, J.C., J.D. Aber, J.D. and C.D. Canham. 1999. Hemlock woolly adelgid impacts on

community structure and N cycling rates in eastern hemlock forests. Canadian Journal of

Forest Research 29:630-645.

Kizlinski, M.L, D.A. Orwig, R.C. Cobb and D.R. Foster. 2002. Direct and indirect ecosystem

consequences of an invasive past on forests dominated by eastern hemlock. Journal of

Biography 29: 1489-1503.

Knoepp, J.D. , J.M. Vose, B. D. Clinton, N. D. Hunter.. 2011. Hemlock infestation and mortality:

impacts on nutrient pools and cycling in Appalachian forests. Soil Science Society of

America Journal 75:1935-1945.

Kominoski, J.S., C.M. Pringle, B.A. Ball, M.A. Bradford, D.C. Coleman, D.B. Hall and M.D.

Hunter. 2007. Nonadditive effects of leaf litter species diversity on breakdown dynamics

in a detritus-based system. Ecology 88:1167-1176.

Krapfl, K.J., E.J. Holzmueller, and M.A. Jenkins. 2011. Early impacts of hemlock woolly

adelgid in Tsuga canadensis forest communities of the southern Appalachian Mountains.

Journal of the Torrey Botanical Society 138: 93-106.

150

Krueger, L.M. and C.J. Peterson. 2006. Effects of white-tailed deer on Tsuga canadensis

regeneration: evidence of microsites as refugia from browsing. American Midland

Naturalist 156:353-362.

Likens, G.E., C.T. Driscoll, and D. C Buso. 1996. Long-term effects of acid rain: response and

recovery of a forest ecosystem. Science 272:244-246.

Lovett, G.M., K.C. Weathers, M.A. Arthur and J.C. Schultz. 2004. Nitrogen cycling in a

northern hardwood forest: do species matter? Biogeochemistry 67:289-308.

Lovett, G.M., C.D. Canham, M.A. Arthur, K.C. Weathers, and R.D. Fitzhugh. 2006. Forest

ecosystem responses to exotic pests and pathogens in eastern North America. BioScience

56:395- 405.

McClure, M.S. 1991. Role of wind, birds, deer, and humans in the dispersal of hemlock woolly

adelgid (Homoptera, Adelgidae). Environmental Entomology 19:36-43.

McCune, B. and D. Keon. 2002. Equations for potential annual direct incident radiation and heat

load. Journal of Vegetation Science 13: 603-606.

McEwan, R.W. J.M. Dyer and N. Pederson. 2011. Multiple interacting ecosystem drivers:

toward an encompassing hypothesis of oak forest dynamics across eastern North

America. Ecography 34:244-256.

Minchin, P.R. 1987. An evaluation of the relative robustness of techniques for ecological

ordination. Vegetatio 69:89-107.

Morris, A.L. and P.C. Goebel. 2007. Geomorphic and riparian forest influences on

characteristics of large wood and large wood jams in old-growth and second-growth

forests in northern Michigan, SA. Earth Surface Processes and Landforms 32:1131-1153.

151

Nowacki , G.J. and M.D. Abrams. 2008. The demise of fire and “mesophication” of forests in

the eastern United States. BioScience 58: 123-138.

Nuckolls, A, N. Wurzburger, C.R. Ford, R. Hendrick, J. M. Vose, and B. Kloeppel. 2008.

Hemlock Declines Rapidly with Hemlock Woolly Adelgid Infestation: Impacts on the

Carbon Cycle of Southern Appalachian Forests. Ecosystems 12: 179-190.

Oksanen, J., F. G. Blanchet, R. Kindt, P. Legendre, R.B. O’Hara, G.L. Simpson, P. Solymos,

M.H. H. Stevens, and H. Wagner. 2010. Vegan: A community ecology package. R

package version 1.17-4. http://CRAN. R-project.org/package=vegan.

Onken, B. and R. Reardon, eds. 2005. Proceedings, Fourth Symposium on Hemlock Woolly

Adelgid in the Eastern United States. Feb 12-14, 2005. United States Department of

Agriculture, Forest Service, Northern Research Station, Morgantown, WV, USA.

Orwig, D. A. and D. R. Foster. 1998. Forest response to the introduced hemlock woolly adelgid

in southern New England, USA. Journal of the Torrey Botanical Society 125:60–73.

Orwig, D.A., R.C. Cobb, A.W. D’Amato, M.L. Kizlinski, and D.R. Foster. 2008. Multi-year

ecosystem response to hemlock woolly adelgid infestation in southern New England

forests. Canadian Journal of Forest Research 38:834-843.

Orwig, D.W. J.R. Thompson, N.A. Povak. M. Manner. D. Niebyl, and D.R. Foster. 2012. A

foundation tree at the precipice: Tsuga canadensis health after the arrival of Adelges

tsugae in central New England. Ecosphere 3:1-16.

Paine, R.T., M.J. Tegner, and E.A. Johnson. 1998. Compounded perturbations yield ecological

surprises. Ecosystems 1:534-545.

Pinherio, J., D. Bates, S. DebRoy, D. Sarkar and the R core team. 2009. NLME: Linear and

nonlinear mixed effects models. R package version 3.1-96.

152

Prasad, A. M. L.R. Iverson, S. Matthews and M. Peters. 2007. A Climate Change Atlas for 134

Forest Tree Species of the Eastern United States

[database].http://www.nrs.fs.fed.us/atlas/tree. USDA Forest Service Northern Research

Regent Instruments Inc. 2007. WinSCanopy for Canopy Analysis. Regent Instruments, Canada.

Ross, R.M., R.M. Bennett, C.D. Snyder, J.A. Young, D.R. Smith, and D.P. Lemarie. 2003.

Influence of eastern hemlock (Tsuga Canadensis L.) on fish community structure and

function in headwater streams of the Delaware River basin. Ecology of Freshwater Fish

12:60-65.

Skinner, M., B. L. Parker, S. Gouli, and T. Ashikaga. 2003. Regional responses of hemlock

woolly adelgid (Homoptera: Adelgidae) to low temperatures. Environmental

Entomology. 32:523–528.

Small, M.J, C.J. Small, and G.D. Dreyer. 2005. Changes in a hemlock-dominated forest

following woolly adelgid infestation in southern New England. Journal of the Torrey

Botanical Society 132:458-470.

Snyder C.D., J.A. Young and D. Smith. 2002. Influence of eastern hemlock (Tsuga canadensis)

forests on aquatic invertebrate assemblages in headwater streams. Canadian Journal of

Fisheries and Aquatic Science 59:262–75.

Spaulding, H.L. and L.K. Rieske. 2010. The aftermath of an invasion: Structure and composition

of Central Appalachian hemlock forests following establishment of hemlock woolly

adelgdi, Adelges tsugae. Biological Invasions 12:3135- 3143.

Stadler, B., T. Müller and D.A. Orwig. 2006. The ecology of energy and nutrient fluxes in

hemlock forests invaded by hemlock woolly adelgid. Ecology 87:1792-1804.

153

Thrush, S.F., J.E. Hewitt, P.K. Dayton, G. Coco, A.M. Lohrer, A. Norkko, J. Norkko, and M.

Chiantore. 2010. Forecasting the limits of resilience: integrating emphirical research with

theory. Proceedings of the Royal Society B. 276:3209-3217.

Tingley, M.W., D.A. Orwig, R. Field and G. Motzkin. 2002. Avian response to removal of a

forest dominant: consequences of hemlock woolly adelgid infestations. Journal of

Biogeography 29:1505-1516.

USDA and NRCS. 2012. The PLANTS Database (http://plants.usda.gov, 18 April 2012).

National Plant Data Team. Greensboro, NC USA.

USDA Forest Service Forest Health Protection. 2012. Northeastern Area State and Private

Forestry hemlock woolly adelgid infefstations by state and county. Available:

http://na.fs.fed.us/fhp/hwa/infestations/hwa_infestations11.pdf

Vitousek , P.M., J.D. Aber, R.W. Howarth, G.E. Likens, P.A. Matson, D.W. Schindler, W. H.

Schlesinger, and D. G. Tilman. 1997. Human alteration of the global nitrogen cycles:

sources and consequences. Ecological Applications 7:737-750.

Wolf, B.L. and J.R. Thomas. 2006. Soil Survey of Bath County, Virginia. United States

Department of Agriculture Natural Resources Conservation Service. Washington, DC.

Wurzburger, N. and R.L. Hendrick. 2009. Plant litter chemistry and mycorrhizal roots promote a

nitrogen feedback in a temperate forest. Journal of Ecology 97: 528-536.

154

Table 5.1. Study sites across the Central Appalachians with mean environmental conditions across transects. Year invaded is the year hemlock woolly adelgid was reported in the county

(USDA Forest Service Forest Health Protection Agency 2012). Hemlock decline refers to mean overstory hemlock decline, where each tree was assigned a canopy health score based on canopy fullness 1: >75%; 2: 51-75%; 3: 26-50%; 4: 1-25%; 5- dead (Orwig and Foster 1998). Sites were categorized by the rounded average of all individuals.

Year Lat Long Hemlock Invaded County State Site Name Alt (m) (N) (W) Slope Asp decline 1979 Botetourt VA Roaring Run 512 37.424 79.542 31 E 5 1991 Alleghany VA Jerrys Run 648 37.472 80.113 57 N 4 1991 Alleghany VA Pounding Mill 583 37.485 79.550 24 S 4 1991 Alleghany VA Simpsons Cr 551 37.503 79.384 39 N 5 1991 Augusta VA Bradley Pond 605 38.165 79.173 56 NE 4 1991 Rockingham VA Hone Quarry 609 38.281 79.082 4 SW 5 1991 Rockingham VA Kephart Run 597 38.325 79.035 7 S 5 1991 Rockingham VA Skidamore Fork 758 38.323 79.094 28 W 3 1993 Bath VA Back Creek Trib 1 741 38.115 79.474 27 W 5 1993 Bath VA Back Creek Trib 2 619 38.103 79.450 50 W 5 1993 Bath VA Left Prong Wilson Cr 582 37.564 79.475 29 N 4 1993 Bath VA Little Laurel Run 1057 38.390 79.051 34 NE 4 1993 Bath VA Little Mare Mt 718 38.004 79.461 29 S 4 1993 Bath VA Mare Run 345 38.011 79.463 50 S 5 1993 Bath VA Wilson Creek 561 37.565 79.472 30 NE 4 1993 Pochahontas WV Cranberry River Trib 1051 38.123 80.170 23 E 4 1993 Pochahontas WV Williams River Trib 869 38.162 80.133 34 S 3 1998 Greenbriar WV Bear Run 1042 38.111 80.220 29 E 2 1998 Greenbriar WV Big Run 1008 38.123 80.233 32 N 3 2001 Tucker WV Elklick Run 600 39.043 70.392 56 N 3 2001 Tucker WV Elklick Run Trib 603 39.042 79.394 63 NE 3 2001 Tucker WV Engine Run 973 39.060 79.291 3 SW 1 2001 Tucker WV Horseshoe Run 571 39.103 79.364 60 W 1 2001 Tucker WV Lindy Run 950 39.055 79.312 7 N 1 2002 Fayette WV Hedricks Cr 510 38.092 80.563 45 SE 2 2002 Fayette WV Wolf Creek 552 38.031 81.052 14 N 3 2002 Nicholas WV Barenshee Run 699 38.168 79.173 39 N 3 2002 Nicholas WV Carnifex Ferry 1 457 38.124 80.556 47 SW 1 2002 Nicholas WV Carnifex Ferry 2 494 38.124 80.556 40 S 1 2002 Nicholas WV Masons Branch 363 38.132 80.592 43 SE 3

155

Table 5.2. Mean overstory relative basal area (% ±SE) of common species (contributing > 1% mean relative basal area in overstory or sapling layers) in 30 hemlock-dominated riparian sites in the central Appalachians. Symbols are used in figures. Incidental overstory species (<1% relative basal area in overstory and/or saplings) included in analyses: Carya glabra (CAGL), Carya tomentosa (CATO), Fraxinus americana (FRAM), Magnolia acuminata (MAAC), Oxydendrum arboreum (OXAR),Oystra virginiana (ORVI), Prunus serotina (PRSE), and Sassafras albidum

(SAAL).

Symbol 10 m 30 m 50 m Acer pensylvanicum ACPE 0.14 ± 0.14 0.00 ± 0.00 0.00 ± 0.00 Acer rubrum ACRU 3.28 ± 1.18 5.94 ± 1.58 7.40 ± 1.76 Acer saccharum ACSA 2.29 ± 1.11 0.32 ± 0.19 0.40 ± 0.21 Ailanthis altissima AIAL 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 Amelanchier arborea AMAR 0.00 ± 0.00 0.04 ± 0.04 0.04 ± 0.04 Aralia spinosa ARSP 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 Betula alleghaniensis BEAL 6.29 ± 2.34 7.00 ± 2.56 3.80 ± 1.66 Betula lenta BELE 15.11 ± 3.68 8.90 ± 2.71 4.59 ± 1.40 Carpinus caroliniana CACA 0.18 ± 0.16 0.49 ± 0.31 0.41 ± 0.36 Fagus grandifolia FAGR 1.36 ± 0.71 1.38 ± 0.80 1.04 ± 0.70 Hamamelis virginiana HAVI 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 Kalmia latifolia KALA 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 Liriodendron tulipifera LITU 5.90 ± 1.55 9.16 ± 2.52 6.69 ± 2.02 Nyssa sylvatica NYSY 2.34 ± 1.50 2.21 ± 1.16 1.61 ± 0.96 Picea rubens PIRU 4.75 ± 2.52 4.47 ± 1.89 8.30 ± 4.04 Pinus strobus PIST 3.88 ± 1.98 3.58 ± 2.36 4.84 ± 2.81 Quercus alba QUAL 4.03 ± 2.14 5.56 ± 2.37 5.36 ± 2.21 Quercus coccinea QUCO 0.91 ± 0.53 0.97 ± 0.41 2.42 ± 1.27 Quercus prinus QUPR 4.13 ± 2.36 6.74 ± 2.33 9.41 ± 2.85 Quercus rubra QURU 2.83 ± 1.58 3.33 ± 1.30 6.83 ± 2.41 Quercus velutina QUVE 0.00 ± 0.00 2.16 ± 0.97 2.51 ± 1.12 Rhododendron maximum RHMA 0.44 ± 0.22 0.15 ± 0.11 1.72 ± 1.69 Tillia americana TIAM 1.50 ± 1.01 1.05 ± 0.64 0.20 ± 0.20 Tsuga canadensis TSCA 37.35 ± 4.90 31.85 ± 4.17 28.79 ± 5.01 Vitus aestivalis VIAE 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 Species Richness 5.5 ± 0.39 6.3 ± 0.38 6.5 ± 0.47 Shannon Diversity (H') 1.17 ± 0.24 1.33 ± 0.06 1.28 ± 0.08

156

Table 5.3. Mean sapling relative basal area ( % ±SE) of common species (contributing > 1% mean relative basal area in overstory or sapling layers) in 30 hemlock dominated riparian sites in the central Appalachians. Incidental sapling species (<1% relative basal area in overstory and/or saplings) included in analyses: Oxydendrum arboreum (OXAR), Prunus allegheniensis (PRAL), and Sassafras albidum (SAAL).

Symbol 10 m 30 m 50 m Acer pensylvanicum ACPE 3.49 ± 2.47 2.91 ± 1.64 1.84 ± 1.31 Acer rubrum ACRU 2.18 ± 1.01 5.48 ± 1.47 8.76 ± 2.41 Acer saccharum ACSA 5.34 ± 3.18 1.32 ± 0.85 2.09 ± 1.31 Ailanthis altissima AIAL 0.73 ± 0.73 1.76 ± 1.69 3.81 ± 2.64 Amelanchier arborea AMAR 0.28 ± 0.14 1.53 ± 1.26 0.34 ± 0.24 Aralia spinosa ARSP 3.47 ± 1.96 2.36 ± 1.40 2.94 ± 1.74 Betula alleghaniensis BEAL 2.29 ± 1.25 1.58 ± 0.77 2.68 ± 1.28 Betula lenta BELE 2.43 ± 0.94 2.47 ± 1.51 0.83 ± 0.42 Carpinus caroliniana CACA 0.67 ± 0.34 1.95 ± 0.88 1.00 ± 0.68 Fagus grandifolia FAGR 5.08 ± 2.90 2.41 ± 1.10 1.74 ± 0.62 Hamamelis virginiana HAVI 4.18 ± 1.90 2.92 ± 1.56 5.90 ± 2.51 Kalmia latifolia KALA 1.80 ± 0.58 2.53 ± 1.37 4.41 ± 2.57 Liriodendron tulipifera LITU 0.40 ± 0.40 0.00 ± 0.00 0.42 ± 0.38 Nyssa sylvatica NYSY 0.63 ± 0.50 2.36 ± 1.43 5.07 ± 2.35 Picea rubens PIRU 3.47 ± 1.96 2.36 ± 1.40 2.94 ± 1.74 Pinus strobus PIST 0.00 ± 0.00 0.26 ± 0.26 1.51 ± 0.97 Quercus alba QUAL 0.05 ± 0.05 0.00 ± 0.00 0.30 ± 0.23 Quercus coccinea QUCO 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 Quercus prinus QUPR 0.00 ± 0.00 0.14 ± 0.14 0.33 ± 0.26 Quercus rubra QURU 0.01 ± 0.01 0.14 ± 0.14 0.00 ± 0.00 Quercus velutina QUVE 0.00 ± 0.00 0.00 ± 0.00 0.07 ± 0.07 Rhododendron maximum RHMA 42.19 ± 6.20 29.45 ± 5.50 24.10 ± 5.54 Tillia americana TIAM 0.90 ± 0.74 0.00 ± 0.00 0.00 ± 0.00 Tsuga canadensis TSCA 18.19 ± 2.95 34.68 ± 4.82 25.90 ± 4.29 Vitus aestivalis VIAE 0.33 ± 0.29 2.08 ± 1.73 1.58 ± 1.16 Species Richness 5.20 ± 0.50 4.67 ± 0.40 5.67 ± 0.56 Shannon Diversity (H') 0.99 ± 0.09 1.05 ± 0.08 1.16 ± 0.09

157

Table 5.4. Mean seedling counts (density ±SE) of common species (contributing > 1% mean relative basal area in overstory or sapling layers) in the seedling layer of 30 hemlock dominated riparian sites in the central Appalachians. Red oaks (QURED: Quercus rubra, Q. cocinea, Q. velutina) were not separated in the seedling layer. Incidental seedling species (<1% relative basal area in overstory and/or saplings) included in analyses: Carya species (CASPP), Ulmus rubra

(ULRU), Fraxinus americana (FRAM), Magnolia acuminata (MAAC), Oxydendrum arboreum

(OXAR), Prunus serotina (PRSE), Rhododendon calendulaceum (RHCAL) and Sassafras albidum (SAAL), Vaccinium species (VASPP).

Symbol 10m 30m 50m Acer pensylvanicum ACPE 0.97 ± 0.40 0.97 ± 0.40 0.30 ± 0.19 Acer rubrum ACRU 40.77 ± 10.74 40.77 ± 10.74 41.63 ± 10.05 Acer saccharum ACSA 0.70 ± 0.24 0.70 ± 0.24 0.60 ± 0.34 Ailanthis altissima AIAL 0.10 ± 0.06 0.10 ± 0.06 1.30 ± 0.93 Amelanchier arborea AMAR 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 Aralia spinosa ARSP 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 Betula alleghaniensis BEAL 0.77 ± 0.44 0.77 ± 0.44 0.37 ± 0.23 Betula lenta BELE 15.47 ± 4.97 15.47 ± 4.97 3.60 ± 1.60 Carpinus caroliniana CACA 0.23 ± 0.20 0.23 ± 0.20 1.37 ± 0.81 Fagus grandifolia FAGR 0.33 ± 0.23 0.33 ± 0.23 0.57 ± 0.47 Hamamelis virginiana HAVI 1.70 ± 0.83 1.70 ± 0.83 3.13 ± 1.41 Kalmia latifolia KALA 4.03 ± 2.63 4.03 ± 2.63 9.27 ± 4.58 Liriodendron tulipifera LITU 5.80 ± 1.78 5.80 ± 1.78 2.10 ± 0.65 Nyssa sylvatica NYSY 1.03 ± 0.47 1.03 ± 0.47 0.70 ± 0.28 Picea rubens PIRU 5.27 ± 3.33 5.27 ± 3.33 9.13 ± 5.64 Pinus strobus PIST 0.20 ± 0.07 0.20 ± 0.07 0.40 ± 0.18 Quercus alba QUAL 4.00 ± 3.63 4.00 ± 3.63 0.37 ± 0.20 Quercus prinus QUPR 2.43 ± 1.25 2.43 ± 1.25 7.40 ± 3.81 Quercus (red) QURED 5.33 ± 2.30 5.33 ± 2.30 3.43 ± 1.12 Rhododendron maximum RHMA 24.73 ± 4.83 24.73 ± 4.83 13.20 ± 4.04 Tillia americana TIAM 0.43 ± 0.34 0.43 ± 0.34 0.10 ± 0.07 Tsuga canadensis TSCA 2.53 ± 1.04 2.53 ± 1.04 0.73 ± 0.37 Vitus aestivalis VIAE 0.00 ± 0.00 0.00 ± 0.00 0.17 ± 0.12 Species Richness 8.47 ± 0.70 8.47 0.70 7.10 ± 0.54 Shannon Diversity (H') 1.26 ± 0.09 1.12 0.08 1.13 ± 0.09

158

Figure 5.1. Generalized conceptual model of alternate state shifts. Communities move between basins of attraction separated by a threshold.

159

Figure 5.2. Conceptual model of alternate state shifts where changes that surpass threshold result in community divergence, the development of multiple alternate community states. Modified from Houseman et al (2008).

160

Figure 5.3. Map of study site locations across the central Appalachians in Virginia and West

Virginia with year hemlock woolly adelgid was detected in the county: 1979: Boutetourt, VA;

1991: Alleghany, Augusta, and Rockingham, VA; 1993: Bath, VA and Pochahontas,WV; 1998:

Greenbriar, WV; 2001: Tucker, WV; 2002: Fayette and Nicholas, WV. A full list of sites can be found in Table 1.

161

Figure 5.4. Non metric multidimensional scaling (NMDS) analysis of the overstory community composition. Points are site mean overstory hemlock decline category (Orwig and Foster 1998): green circles- 1 (>75% foliage remaining); yellow diamond- 2 (50-75% foliage remaining); orange triangle- 3 (25-50% foliage remaining); red triangle- 4 (1-25% foliage remaining);black squares- 5 (dead). Ellipses represent 95% confidence intervals around group centroid of each overstory decline class.

162

Figure 5.5. Non metric multidimensional scaling (NMDS) analysis of the overstory community

composition. Points represent species, codes can be found in Table 2. Vectors represent

continuous environmental gradients: Long- longitude; Lat- latitude; Elev- elevation. Categorical environmental variables are represented by Xs: Slope shape: CC (concave), CV (convex), LI

(linear); Slope position: FP (flood plain), HS (hillslope), SB (structural bench), RT (ridge top);

Distance from stream: 10m, 30m, 50m.

163

Figure 5.6. Comparisons of multivariate variance in overstory community composition compared by average overstory hemlock decline class (1-5). Red points represent group centroids, lines connect centroid to each transect score based on bray-curtis distance. Axes represent the first two dimensions of the three dimensional analysis of multivariate variance.

164

Figure 5.7.Tukey’s honestly significant diference (HSD) comparison of variance among overstory hemlock decline classes. Confidence intervals that do not overlap with the center line

(0) indicate a significant difference.

165

Figure 5.8. NMDS of the sapling community composition. Points represent sites coded by mean overstory hemlock decline category (Orwig and Foster 1998): green circles- 1 (>75% foliage remaining); yellow diamond- 2 (50-75% foliage remining); orange triangle- 3 (25-50% foliage remaining); red triangle- 4 (1-25% foliage remaining);black squares- 5 (dead). Ellipses are 95% confidence intervals around group centroid of each overstory decline class.

166

Figure 5.9. NMDS of sapling community composition. Points represent species, codes can be found in Table 5.5. Vectors represent continuous environmental gradients: Long- longitude; Lat- latitude; Alt- elevation. Categorical environmental variables are represented by Xs: Slope shape:

CC (concave), CV (convex), LI (linear); Slope position: FP (flood plain), HS (hillslope), SB

(structural bench), RT (ridge top); Distance from stream: 10m, 30m, 50m.

167

Figure 5.10. Non metric multidimensional scaling (NMDS) analysis of the seedling community composition. Points represent sites coded by mean overstory hemlock decline category ( the rounded mean of all individuals across the site): green circles- 1 (>75% foliage remaining); yellow diamond- 2 (50-75% foliage remining); orange triangle- 3 (25-50% foliage remaining); red triangle- 4 (1-25% foliage remaining);black squares- 5 (dead). Categories developed by

Orwig and Foster (1998). Ellipses represent 95% confidence intervals around group centroid of each overstory decline class.

168

Figure 5.11. Non metric multidimensional scaling (NMDS) analysis of the seedling community

composition. Points represent species, codes can be found in Table 4. Vectors represent

continuous environmental gradients: Long- longitude; Lat- latitude; Elev- elevation. Categorical environmental variables are represented by Xs: Slope shape: CC (concave), CV (convex), LI

(linear); Slope position: FP (flood plain), HS (hillslope), SB (structural bench), RT (ridge top);

Distance from stream: 10m, 30m, 50m.

169

Figure 5.12. Structural equation model (SEM) of the influence of hemlock decline and soil texture on soil cation exchange capacity and soil carbon to nitrogen ratio. Model fit:

Χ 2 17.0, df= 11, p= 0.11

170

Figure 5.13. Structural equation model (SEM) illustrating the influence of overstory community composition on canopy openness in the dormant season. Model fit: Model fit:

Χ2 6.2, df= 3, p= 0.10

171

Figure 5.14. Structural equation model (SEM) of the influence of overstory hemlock decline on relative decomposition rate constant k and leaf litter carbon to nitrogen ratio.

Model fit: Χ2 10.3, df= 6, p= 0.11.

172

Chapter 6: Conclusion

Although ecosystem composition and function continually fluctuate and

reorganize, the rate of ecosystem change has accelerated due to anthropogenic

disturbances including loss of biodiversity, invasion by exotic species, and altered

nutrient and energy cycles (Chapin et al. 2000, Hooper et al. 2005). While progress has

been made to understand the implications of these changes, particularly through

theoretical and small-scale experiments, much remains unclear (Hooper et al. 2005). As

increasingly novel and compounded disturbances reshape complex natural systems, in

some cases the theoretical frameworks of resilience and alternate state dynamics may be

particularly helpful to understand and predict ecosystem responses (Paine et al. 1998,

Beisner et al. 2003, Suding et al. 2004, Thrush et al. 2009). This necessitates testing the

applicability of these concepts on broader scales and in ecosystems dominated by long-

lived individuals (Groffman et al. 2006). The focus of this dissertation, the loss of

eastern hemlock (Tsuga canadensis) to hemlock woolly adelgid (Adelges tsugae)

provides an important case study as a part of this discussion.

A central hypothesis of this research is that forests defined by hemlock as a

foundation are low resilience, because compositional and functional processes are driven largely by hemlock, with little redundancy. This is supported by data from the

173 unglaciated Allegheny Plateau of Ohio where hemlock remains uninvaded by HWA

(Chapter 3). When compared to surrounding deciduous forest, hemlock communities exhibit low plant species richness and diversity. Although hemlock is dominant throughout ravine ecosystems along headwater streams, gradient analyses suggested component species are structured by environmental context. Sweet birch (Betula lenta),

American beech (Fagus grandifolia), and tulip-poplar (Liriodendron tulipifera) occur more commonly closer to the stream, with sugar maple (Acer saccharum), red maple

(Acer rubrum), and white oak more (Quercus alba) more common moving upslope, and

Northern red oak (Quercus rubra) and chestnut oak (Quercus prinus) most common 50- m from headwater sreams. Structural equation modeling identifies hemlock as an important driver of biodiversity in the sapling layer, ground flora, leaf litter, as well as overstory functional diversity. In functional processes, structural equation models indicate hemlock as a strong determinant of resource availability, determined by light availability (canopy openness) in the growing season and leaf litter biomass. If HWA arrives in Ohio, as is likely, functional processes will likely respond rapidly (Ellison et al.

2005, Nuckolls et al. 2008, Orwig et al. 2008), and will likely fluctuate as composition shifts. Differences in species composition across ecological gradients will likely become more apparent as communities reorganize during hemlock decline and mortality.

Although other researchers have observed that hemlock forests begin to change as soon as HWA arrives (Jenkins et al. 1999, Stadler et al. 2006, Nuckolls et al. 2008), data from West Virginia and Virginia indicate compositional changes occur over decades

(Chapter 4). Across the sites in Virginia and West Virginia impacted by HWA for 9-32

174

years, hemlock remained a dominant species in both the overstory and sapling layers.

Examining sites impacted within the last decade, the impacts of HWA seem to be slowed at higher elevations and at cooler, shadier landscape positions (northerly aspects, steeper slopes). However, when the full HWA invasion choronosequence is included, duration of

HWA invasion is the most influential factor in hemlock decline, and saplings are further impacted by the decline in the overstory. This finding is in agreement with studies from the Connecticut River Valley that initial differences in HWA impacts are eventually overwhelmed by the duration of invasion (Orwig et al. 2008, Orwig et al. 2012).

Community compositional and functional shifts are further explored in Chapter 5.

The central hypothesis of this chapter is that as hemlock is lost from central Appalachian

ecosystems, communities that occur across elevation, physiographic, soil, and probably

land use history gradients are re-developing along multiple pathways. This would

indicate a divergence, ultimately resulting in multiple new community states. In the

overstory and seedling layers, there species richness increases among overstory hemlock decline categories (1: >75%; 2: 51-75%; 3: 26-50%; 4: 1-25%; 5: dead (Orwig and Foster

1998)), but diversity does not. Sapling layer richness and diversity did not differ by

overstory hemlock decline.

As overstory hemlock declines, the composition of the vegetation layers is

changing. Gradient analyses suggest some species are more closely associated with

environmental factors, while others are driven by the duration of HWA invasion and

overstory hemlock decline. In the overstory, oaks (Quercus prinus, Q. rubra), blackgum

(Nyssa sylvatica), and sourwood (Oxydendrum arborum) were associated with hemlock

175

decline. Sweet birch was more closely associated with steep slopes and yellow birch

(Betula allghenensis) and red spruce (Picea rubens) were found at higher elevations.

Analyses indicate overstory community compostion is changing among overstory

hemlock decline classes, and the variance in composition is increasing, indicating

divergence.

Dynamics in the sapling layer are not as straightforward, most likely due to

rhododendron (Rhododendron maximum) which was present at some, but all sites.

Factors driving the distribution of rhododendron are not entirely clear, although it was

associated with the transect closest to the stream and away from other species in the

sapling layer. Soil factors, which we were unable to measure at a quarter of the study sites, may explain much of the variation in rhododendron presence. At the same time, the influence of rhododendron may explain compositional differences in the sapling layer both among overstory hemlock decline classes as well as by distance from the stream.

Variance in the species composition did not increase in the sapling layer, but that is likely because it already varies. Many species were infrequent in the sapling layer, with 31 of

50 occurring on less that 5% of plots (4 of 90).

Like the sapling layer, the seedling layer is likely quite variable across central

Appalachain hemlock forests. Although sapling composition differed among overstory hemlock decline classes, compositional variance did not increase. Gradient analyses indicate environmental measures including elevation, incident radiation, and slope shape and position may be more influential in the seedling distribution than in the sapling layer.

Again, rhododendron was located away from other species in the ordination diagram.

176

Several species were associated with overstory hemlock decline and duration of HWA invasion, while sweet birch, tulip poplar, and red oaks (Quercus rubra, Q. coccinea, Q. velutina) exhibit fairly ubiquitous distribution.

As hemlock declines, data from the central Appalachians indicate an increase in nutrient cycling rates and resource availability, but this varies with environmental context. Structural equation models identified an increase in soil cation exchange capacity and a decrease in soil carbon to nitrogen ratio with hemlock decline, but both properties varied with soil texture. Likewise, light availability (canopy openness) increased during both the growing and dormant season, but this was modified at higher elevations by the presence of red spruce. Unlike other functional metrics, hemlock decline seemed to accelerate decomposition rates and decrease leaf litter carbon to nitrogen ration across environmental gradients.

Findings from this dissertation support broad patterns identified with hemlock decline in historic pollen records from a hemlock population crash approximately 5500 ybp (Allison et al. 1986, Heard and Valente 2009) and several other well studied regions, including the Connecticut River Valley in New England (Orwig and Foster 1998, Orwig et al. 2008, Orwig et al. 2012), the Delaware Water Gap National Recreation Area in

Pennsylvania and New Jersey (Eschtruth et al. 2006) and the Coweeta Hyrdologic

Laboratory in the Nantahala Mountains of North Carolina (Ford and Vose 2007,

Nuckolls et al. 2008 , Ford et al. 2012). At the same time, this research highlights the complexity of the loss of hemlock as a foundation species across a wide native habitat that encompasses the southern Appalachians to the Northeast and Lake States. Although

177

evidence from this research supports the transition of hemlock communities to multiple

alternate states, questions remain for further study. While foundation species theory

indicates communities structured by hemlock should be similar across the range, this has

not been a primary objective of study and thus, remains uncertain. Some of the

differences throughtout the range are suggested by Evans et al. (2011), and central and southern Appalachian hemlock forests may be more diverse (Orwig and Foster 1998,

Orwig et al. 2008, Small et al. 2005, Spaulding and Rieske 2010, Krapfl et al. 2011).

Further, it is not clear whether there are there generalizable patterns in alternate state

development and which environmental gradients and/or species are most influential. In

the central and southern Appalachians, in some cases rhododendron may be a new

foundation species of a novel shrub ecosystem. In other areas with an evergreen component such as red spruce, the impacts of the loss of hemlock on functional processes may be modified by some functional overlap, but this is unclear.

Hemlock ecosystems are fairly well studied throughout their range, and the

impacts of HWA are well documented in certain areas; thus, it serves and important case

study of foundation species and forests resilience. As outlined in Chapter 2, comparisons

with hemlock will also advance our understanding of the role of resilience in the ongoing

shift from oak to maple dominance and the loss of ash (Fraxinus) due to the invasive

emerald ash borer (Agrilus planipennis, EAB). Such understanding expands ecological

theory, but is also necessary for predictive models and management to maintain

ecosystem services including carbon capture, nutrient and flood retention, and biodiversity.

178

References

Adams, M.S. and O.L. Loucks. 1971. Summer air temperatures as a factor affecting net

photosynthesis and distribution of eastern hemlock (Tsuga canadensisL. (Carriere)) in

southwestern Wisconsin. American Midland Naturalist 85:1-10

Albrecht, M.A. and B.C. McCarthy. 2006. Effects of prescribed fire and thinning on tree

recruitment patterns in Central Hardwood forests. Forest Ecology and Management

226:88-103.

Alexander, H.D. and M.A. Arthur. 2010. Implications of a predicted shift from upland oaks

to red maple on forest hydrology and nutrient availability. Canadian Journal of Forest

Research 40:716-726.

Allison, T.D., R.E. Moeller, and M.B. Davis, M.B. 1986. Pollen in laminated sediments

provides evidence for a mid-Holocene forest pathogen outbreak. Ecology 67:1101-

1105.

Amatangelo, K.L., M.R. Fulton, D.A. Rogers, and D.M. Waller. 2011. Converging forest

community composition along an edaphic gradient threatens landscape-level

diversity. Diversity and Distributions17: 201-213.

179

American Society for Testing and Materials. 1985. Standard test method for particle-size

analysis of soils. Pages 117-127 in: Annual Book of ASTM Standards, American

Society for Testing Materials. West Conshohocken, PA, USA.

Anagnostakis, S.L. 1987. Chestnut blight: the classical problem of an introduced pathogen.

Mycologia 79:23-37.

AOAC Official Methods of Analysis. 2002. Method 990.03. Protein (crude) in Animal Feed

Combusion Method (Dumas method). JAOAC 72:770.

Attiwill, P.M. 1994. The disturbance of forests ecosystems- the ecological basis for

conservative management. Forest Ecology and Management 63:247-300.

Ball, B.A., M.D. Hunter, J, S. Kominoski, C.S. Swan and M.A.Bradford. 2008.

Consequences of non-random species loss for decomposition dynamics: experimental

evidence for additive and non-additive effects. Journal of Ecology 96:303-313.

Barnes, B.V. 1976. Succession in deciduous swamp communities of southeastern Michigan

formerly dominated by American elm. Canadian Journal of Forest Research 54: 19-

24.

Beatley, J.C. 1959. The primeval forests of a periglacial area in the Allegheny Plateau

(Vinton and Jackson Counties). Bulletin of the Ohio Biological Survey, (New Series)

1:1-182.

Beckage, B. and J.S. Clark, J.S. 2003. Seedling survival of three forest tree species: the role

of spatial heterogeneity. Ecology 84:1849-1861.

180

Beckage, B., J.S. Clark, B.D. Clinton, and B.L. Haines. 2000. A long-term study of tree

seedling recruitment in southern Appalachian forests: the effects of canopy gaps and

shrub understories. Canadian. Journal of Forest Research. 30:1617-1631.

Beckage, B., B.D. Kloeppel, J. A. Yeakley, S.F. Taylor, and D.C. Coleman. 2008.

Differential effects of understory and overstory gaps on tree regeneration. Journal of

the Torrey Botanical Society 135:1-11.

Beisner, B.E., D.T. Haydon, and K. Cuddington. 2003. Alternative stable states in ecology.

Frontiers in Ecology and the Environment 1: 376-382.

Black, R.A. and R.N. Mack. 1976. Tsuga canadensis in Ohio: Synecological and

phytogeographical relationships. Plant Ecology 32:11-19.

Braun, E.L. 1950. Deciduous forests of eastern North America. MacMillian, New York,

USA.

Brockman, C.S. 1998. Physiographic regions of Ohio [map]. Division of Geological Survey,

State of Ohio.

Bruno, J.,J. Stachowicz, and P.A. Townsend. 2003. Inclusion of facilitation into theoretical

ecology. 2003. Trends in Ecology and Evolution 18:119-125.

Canham, C.D., A.C. Finzi, S.W. Pacala, and D.H. Burbank. 1994. Causes and consequences

of resource heterogeneity in forests: interspecific variation in light transmission by

canopy trees. Canadian Journal of Forest Research 24:337-349.

Carpenter, S.G. 1992. Soil Survey of Nicholas County, West Virginia. United States

Department of Agriculture Soil Conservation Service. Washington, DC.

181

Catovsky, S. and F.A. Bazzaz. 2000. The role of resource interactions and seedling

regeneration in maintaining a positive feedback in hemlock stands. Journal of

Ecology 88:100-112.

Chapin, F.S., E. S. Zavaleta, V.T. Eviner, R. L. Naylor, P.M. Vitousek, H. L. Reynolds, D.U.

Hooper, S. Lavorel, O.E. Sala, S.E. Hobbie, M.C. Mack, and S. Diaz. 2000.

Consequences of changing biodiversity. Nature 405:234-242.

Clinton, B.D., L.R. Boring, and W.T. Swank. 1994. Regeneration patterns in canopy gaps of

mixed-oak forests of the southern Appalachians: influences of topographic position

and evergreen understory. American Midland Naturalist 132:308-319.

Cobb, R.C., D.A. Orwig, and S. Currie. 2006. Decomposition of green foliage in eastern

hemlock forests of southern New England impacted by hemlock woolly adelgid

infestations. Canadian Journal of Forest Research 36:1331-1341.

Cobb, R.C. 2009. Species shift drives decomposition rates following invasion by hemlock

woolly adelgid. Oikos 119:1291-1298.

Constanza, R., R. d’Arge, R. de Groot, S. Farber, M. Grasso, B. Hannon, K. Limburg, S.

Naeem, R. V. O’Neill, J. Paruelo, R. G. Raskin, P. Sutton, and M. van den Belt. 1997.

The value of the world’s ecosystem services and natural capital. Nature 387:253-260.

Cook, M E. M. and J. D. Slabaugh. 2006. Soil Survey of Alleghany County, Virginia. United

States Department of Agriculture Natural Resources Conservation Service.

Washington, DC.

182

Cote, S.D. T. P. Rooney, J. Tremblay, C. Dussault, and D.M. Waller. 2004. Ecological

impacts of deer overabundance. Annual Review of Ecology, Evolution and

Systematics 35:113-147.

Daley, M.J., N. G. Phillips, C. Pettijohn, and J.L. Hadley. 2007. Water use by eastern

hemlock (Tsuga canadensis) and black birch (Betula lenta): implications of effects of

the hemlock woolly adelgid. Canadian Journal of Forest Research 37:2031-2040.

Davis, M.B. 1969. Clilmatic changes in southern Connecticut recorded by pollen deposition

at Rogers Lake. Ecology 50:409-422.

Day, F.P. and C.D. Monk. 1974. Vegetation pattern on a Southern Appalachian watershed.

Ecology 55:1064-1074.

Delcourt, P.A. and H.R. Delcourt. 1998. Paleoecological insights on conservation of

biodiversity: a focus on species, ecosystems, and landscapes. Ecological Applications

8:921-934.

Dhungel, E.K., J.W. Groninger, and E.J. Holzmueller. 2010. Tree species and environment

associations with hemlock-silverbell stands treated for hemlock woolly adelgid in

Great Smoky Mountains National Park. Journal of the Torrey Botanical Society

137:401-409.

Dixon, P. and M.W. Palmer. 2003. Vegan, a package of R functions for community ecology.

Journal of Vegetation Science 14: 927-930.

Dodds, K.J. and D.A. Orwig. 2011. An invasive urban forest pest invades natural

environments- Asian longhorned beetle in northeastern US hardwood forests.

Canadian Journal Forest Research 41:1729-1742.

183

Drever, C.R., G. Peterson, C. Messier, Y. Bergeron, and M. Flannigan. 2006. Can forest

management based on natural disturbances maintain ecological resilience? Canadian

Journal of Forest Research 36:2285-2299.

Dukes, J.S., J. Pontius, D. Orwig, J.R. Garnas, V.L. Rodgers, N. Brazee, B. Cooke, K.A.

Theoharides, E.E. Stange, R. Harrington, J. Ehrenfeld, J. Gurevitch, M. Lerdau, K.

Stinson. R. Wick, and M. Ayres. 2009. Responses of insect pests, pathogens, and

invasive plant species to climate change in the forests of northeastern North America:

What can we predict? Canadian Journal of Forest Research 39:231-248.

Ehrenfeld, J.G. 2011. Ecosystem consequences of biological invasions. Annual Review of

Ecology, Evolution, and Systematics 41:59-80.

Elliott, K.J, J.M. Vose, W.T. Swank, and P.V. Bolstad. 1999. Long-term patterns in

vegetation-site relationships in southern Appalachian forests. Journal of the Torrey

Botanical Society 126:320–334

Elliott, K.J. and W.T. Swank. 2008. Long-term changes in forest composition and diversity

following early logging (1912-1923) and the decline of the American chestnut

(Castanea dentata). Plant Ecology 197:155-172.

Ellison, A. M., M. S. Banks, B. D. Clinton, E. A. Colburn, K. Elliott, C. R. Ford, D. R.

Foster, B. D. Kloeppel, J. D. Knoepp, G. M. Lovett, J. Mohan, D. A. Orwig, and N.

L. Rodenhouse. 2005. Loss of foundation species: consequences for the structure and

dynamics of forested ecosystems. Frontiers in Ecology and the Environment 3:479-

486.

184

Elmqvist, T., C. Folke, M. Nystrom, G. Peterson, J. Bengtsson, B. Walker, and J. Norberg.

2003. Response diversity, ecosystem change, and resilience. Frontiers in Ecology and

the Environment 1:488-494.

Eschtruth, A. K., N. L. Cleavitt, J. J. Battles, R. A. Evans, and T. J. Fahey. 2006. Vegetation

dynamics in declining eastern hemlock stands: 9 years of forest response to hemlock

woolly adelgid infestation. Canadian Journal of Forest Research 36:1435-1450.

Evans, A.M and T.G. Gregoire. 2007. A geographically variable model of hemlock woolly

adelgid spread. Biological Invasions 9:369-382.

Evans. D.M., W.M. Aust, A.C. Dolloff, B.S. Templeton, J.A. Peterson. 2011. Eastern

hemlock decline in riparian areas from Maine to Alabama. Northern Journal of

Applied Forestry 28: 97-104.

Fei, S., N. Kong, K.C. Steiner, W.K. Moser, and E.B. Steiner. 2011. Change in oak

abundance in the eastern United States from 1980 to 2008. Forest Ecology and

Management 262:1370-1377.

Finzi, A.C., N. Van Breemen, and C. D. Canham. 1998. Canopy tree soil interactions within

temperate forests: species effects on soil carbon and nitrogen. Ecological

Applications 8:440-446.

Flegel, D.G. 1998. Soil Survey of Pocahontas County, West Virginia. United States

Department of Agriculture Natural Resources Conservation Service. Washington,

DC.

185

Folke, C., S. Carpenter, B. Walker, M. Scheffer, T. Elmqvist, L. Gunerson, and C.S. Holling.

2004. Regime shifts, resilience, and biodiversity in ecosystem management. Annual

Review of Ecology, Evolution, and Systematics 35:557-581.

Ford, C.R. and J.M. Vose. 2007. Tsuga canadensis (L.) Carr. mortality will impact

hydrologic processes in southern Appalachian forest ecosystems. Ecological

Applications 17:1156-1167.

Ford, C.R., K.J. Elliott, B.D. Clinton, B.D. Kloeppel., J.M. Vose. 2012. Forest dynamics

following eastern hemlock mortality in the southern Appalachians. Oikos 121:523-

536.

Foster, D.R., S. Clayden, D. A. Orwig, and B. Hall. 2002. Oak, chestnut and fire: climatic

and cultural controls of long-term forest dynamics in New England, USA. Journal of

Biogeography 29:1359-1379.

Fox, V.L., C. P. Buehler, C.M. Byers, S. E. Drake. 2010. Forest composition, leaf litter, and

songbird communities in oak-vs. maple-dominated forests in the eastern eastern

United States. Forest Ecology and Management 259:2426-2432.

Fralish, J.S. 2003. The Central Hardwood Forest: Its boundaries and physiographic

provinces, pp 1-20 in J.W. Van Sambeek, J.O. Dawson, F. Ponder, Jr., E.F.

Loewenstein, and J.S. Fralish [eds]. Proceedings, 13th Central Hardwood Forest

Conference; 2002 April 1-3; Urbana, IL. Gen. Tech. Rep. NC-234. St. Paul, MN:

U.S. Department of Agriculture, Forest Service, North Central Research Station.

186

Franklin, J.F., and K.N. Johnson. 2011. Societal challenges in understanding and responding

to regime shifts in forest landscapes. Proceedings of the National Academy of

Sciences of the United States of America 108:16863-16864.

Gandhi, K.J.K. and D.A. Herms. 2010. Direct and indirect effects of alien insect herbivores

on ecological processes and interactions in forests of eastern North America.

Biological Invasions 12:389-405.

Gilliam, F.S., M.B. Adams, B.M. Yurish. 1996. Ecosystem nutrient responses to chronic

nitrogen inputs at Fernow Experimental Forest, West Virginia. Canadian Journal of

Forest Research 16:196-205.

Goebel, P.C., and D.M. Hix, D.M. 1996. Development of mixed-oak forests in southeastern

Ohio: A comparison of second-growth and old-growth forests. Forest Ecology and

Management 84:1-21.

Gorman, J.L. and L.E. Espy. 1975. Soil Survey of Fayette and Raleigh Counties, West

Virginia. United States Department of Agriculture Soil Conservation Service.

Washington, DC.

Grace, J. B. 2006. Structural equation modeling and natural systems. Cambridge University

Press. Cambridge, UK.

Grace, J.B. 2008. Structural equation modeling for observational studies. Journal of Wildlife

Management 72:14-22.

Green, S.R., M.A. Arthur, B. A. Blakenship. 2010. Oak and red maple seedling survival and

growth following periodic prescribed fire on xeric ridgetops on the Cumberland

Plateau. Forest Ecology and Management 259:2256-2266.

187

Groffman, P.M, J.S. Baron, T. Blett, A.J. Gold, I. Goodman, L.H. Gunderson, B. M.

Levinson, M.A. Palmer, H. W. Paerl, G.D. Peterson, N. LeRoy Poff, D. W. Rejeski, J.

F. Reynolds, M. G. Turner, K.C. Weathers, and J. Wiens. 2006. Ecological

thresholds: the key to successful environmental management or an important concept

with no practical application? Ecosystems 9:1-13.

Gunderson, L.H. 2000. Ecological resilience- in theory and application. Annual Review of

Ecology and Systematics 31:425-439.

Heard, M.J. and M.J. Valente, M.J. 2009. Fossil pollen records forecast response of forests to

hemlock woolly adelgid invasion. Ecography 32:881887.

Hilt, S. J. Kohler, H. Kozerski, E.H. van Nes, M. Scheffer. 2010. Abrupt regime shifts in

space and time along rivers and connected lake systems. Oikos 120:766-775.

Holling, C.S. 1973. Resilience and stability of ecological systems. Annual Review of

Ecology and Systematics 4:1-23.

Hooper, D.U., F.S. Chapin, J.J. Ewel, A. Hector, P. Inchausti, S. Lavorel, J.H. Lawton, D.M.

Lodge, M. Loreau, S. Naeem, B. Schmid, H. Setala, A.J. Symstad, J. Vendermeer,

and D.A. Wardel. 2005. Effects of biodiversity on ecosystem functioning: a

consensus of current knowledge. Ecological Monographs 75:3-35.

Houseman, G.R., G.G. Mittelbach, H.L. Reynolds, and K.L. Gross. 2008. Perturbations alter

community convergence, divergence, and formation of multiple community states.

Ecology 89:2172-2180.

Huston, M.A. 1997. Hidden treatments in ecological experiments: reevaluating the ecosystem

function of biodiversity. Oecologia 110:449–460

188

Hutchinson, T.F., R. J. Boerner, S. Sutherland, E. K. Sutherland, M. Ortt, L. R. Iverson.

2005. Prescribed fire effects on the herbaceous layer of mixed-oak forests. Canadian

Journal of Forest Research 35:877-890.

International Standard, ISO 10694. 1995. Soil quality- determination of organic and total

carbon after dry combustion (elementary analysis). International Organization for

Standardization. Geneva, Switzerland.

Iverson, L.R., M.E. Dale, C.T. Scott, A. Prasad. 1997. A GIS-derived integrated moisture

index to predict forest composition and productivity of Ohio forests (USA).

Landscape Ecology 12:331-348.

Jenkins, J.C., J.D. Aber, J.D. and C.D. Canham. 1999. Hemlock woolly adelgid impacts on

community structure and N cycling rates in eastern hemlock forests. Canadian Journal

of Forest Research 29:630-645.

Kashian, D.M. and J.A. Wittier. 2011. Assessing the potential for ash canopy tree

replacement via current regeneration following emerald ash borer-caused mortality on

southeastern Michigan landscapes. Forest Ecology and Management 261:480-488.

Keever, C. 1953. Present composition of some stands of former oak-chestnut forest in the

Southern Blue Ridge Mountains. Ecology 34:44–54

Kerr, J.W. 1983. Soil survey of Jackson County, Ohio. USDA Soil Conservation Service.

Washington, DC

Kizlinski, M.L, D.A. Orwig, R.C. Cobb and D.R. Foster. 2002. Direct and indirect ecosystem

consequences of an invasive past on forests dominated by eastern hemlock. Journal of

Biography 29:1489-1503.

189

Knoepp, J.D., J.M. Vose, B. D. Clinton, N. D. Hunter.. 2011. Hemlock infestation and

mortality: impacts on nutrient pools and cycling in Appalachian forests. Soil Science

Society of America Journal 75:1935-1945.

Kominoski, J.S., C.M. Pringle, B.A. Ball, M.A. Bradford, D.C. Coleman, D.B. Hall, and

M.D. Hunter. 2007. Nonadditive effects of leaf litter species diversity on breakdown

dynamics in a detritus-based stream. Ecology 88:1167-1176.

Krapfl, K.J., E.J. Holzmueller, and M.A. Jenkins. 2011. Early impacts of hemlock woolly

adelgid in Tsuga canadensis forest communities of the southern Appalachian

Mountains. Journal of the Torrey Botanical Society 138:93-106.

Krueger, L.M. and C.J. Peterson. 2006. Effects of white-tailed deer on Tsuga canadensis

regeneration: evidence of microsites as refugia from browsing. American Midland

Naturalist 156:353-362.

Lemaster, D.D. and G.M. Gilmore.1989. Soil survey of Hocking County, Ohio. USDA Soil

Conservation Service. Washington DC

Likens, G.E., C.T. Driscoll, and D. C Buso. 1996. Long-term effects of acid rain: response

and recovery of a forest ecosystem. Science 272:244-246.

Lindenmayer, D.B., R.J. Hobbs, G.E. Likens, C.J. Krebs, S.C. Banks. 2011. Newly

discovered landscape traps produce regime shifts in wet forests. Proceedings of the

National Academy of Sciences of the United States of America 108:15887-15891.

Lovett, G.M., K.C. Weathers, M.A. Arthur and J.C. Schultz. 2004. Nitrogen cycling in a

northern hardwood forest: do species matter? Biogeochemistry 67:289-308.

190

Lovett, G.M., C.D. Canham, M.A. Arthur, K.C. Weathers, and R.D. Fitzhugh. 2006. Forest

ecosystem responses to exotic pests and pathogens in eastern North America.

BioScience 56:395- 405.

Martin, K.L. and L.K. Kirkman. 2009. Management of ecological thresholds to re-establish

disturbance-maintained herbaceous wetlands of the south-eastern USA. Journal of

Applied Ecology 46:906-914.

Martin, K.L., D.M. Hix, and P.C. Goebel. 2011. Coupling of vegetation layers and

environmental influences in a mature, second-growth Central Hardwood forest

landscape. Forest Ecology Management 261:720-729.

Mayer, A.L. and M. Rietkerk, 2004. The dynamic regime concept for ecosystem

management and restoration. BioScience 54:1013- 1020.

McClure, M.S. 1991a. Role of wind, birds, deer, and humans in the dispersal of hemlock

woolly adelgid (Homoptera, Adelgidae). Environmental Entomology 19:36-43.

McClure, M.S. 1991b. Nitrogen fertilization of hemlock increases susceptibility to hemlock

woolly adelgid. Journal of Arboriculture 17: 227-230.

McCune, B. and D. Keon. 2002. Equations for potential annual direct incident radiation and

heat load. Journal of Vegetation Science 13:603-606.

McEwan, R.W. J.M. Dyer and N. Pederson. 2011. Multiple interacting ecosystem drivers:

toward an encompassing hypothesis of oak forest dynamics across eastern North

America. Ecography 34:244-256.

Minchin, P.R. 1987. An evaluation of the relative robustness of techniques for ecological

ordination. Vegetatio 69:89-107.

191

Morris, A.L. and P.C. Goebel. 2007. Geomorphic and riparian forest influences on

characteristics of large wood and large wood jams in old-growth and second-growth

forests in northern Michigan, SA. Earth Surface Processes and Landforms 32:1131-

1153.

Nowacki , G.J. and M.D. Abrams. 2008. The demise of fire and “mesophication” of forests

in the eastern United States. BioScience 58:123-138.

Nuckolls, A, N. Wurzburger, C.R. Ford, R. Hendrick, J. M. Vose, and B. Kloeppel. 2008.

Hemlock Declines Rapidly with Hemlock Woolly Adelgid Infestation: Impacts on the

Carbon Cycle of Southern Appalachian Forests. Ecosystems 12:179-190.

Oksanen, J., F. G. Blanchet, R. Kindt, P. Legendre, R.B. O’Hara, G.L. Simpson, P. Solymos,

M.H. H. Stevens, and H. Wagner. 2010. Vegan: A community ecology package. R

package version 1.17-4. http://CRAN. R-project.org/package=vegan.

Onken, B. and R. Reardon, eds. 2005. Proceedings, Third Symposium on Hemlock Woolly

Adelgid in the Eastern United States. Feb 1-3, 2005. United States Department of

Agriculture, Forest Service, Northern Research Station, Morgantown, WV, USA. 367

p.

Onken, B. and R. Reardon, eds. 2008. Proceedings, Fourth Symposium on Hemlock Woolly

Adelgid in the Eastern United States. Feb 12-14, 2005. United States Department of

Agriculture, Forest Service, Northern Research Station, Morgantown, WV, USA.

Oosting, H.J. and D.W. Hess. 1956. Microclimate and a relic stand of Tsuga canadensisin the

lower piedmont of North Carolina. Ecology 37:28-39.

192

Orwig, D. A. and D. R. Foster. 1998. Forest response to the introduced hemlock woolly

adelgid in southern New England, USA.. Journal of the Torrey Botanical Society

125:60–73.

Orwig, D.A, D.R. Foster, and D.L. Mausel. 2002. Landscape patterns of hemlock decline in

New England due to the introduced hemlock woolly adelgid. Journal of

Biogeography. 29: 1475-1487.

Orwig, D.A., R.C. Cobb, A.W. D’Amato, M.L. Kizlinski, and D.R. Foster. 2008. Multi-year

ecosystem response to hemlock woolly adelgid infestation in southern New England

forests. Canadian Journal of Forest Research 38:834-843.

Orwig, D.W. J.R. Thompson, N.A. Povak. M. Manner. D. Niebyl, and D.R. Foster. 2012. A

foundation tree at the precipice: Tsuga canadensis health after the arrival of Adelges

tsugae in central New England. Ecosphere 3:1-16.

Paine, R.T., M.J. Tegner, and E.A. Johnson. 1998. Compounded perturbations yield

ecological surprises. Ecosystems 1:534-545.

Peterson, G., C.R. Allen, and C.S. Holling. 1998. Ecological resilience, biodiversity, and

scale. Ecosystems 1:6-18.

Piatek, K.B., P. Munasinghe, W.T. Peterjohn, M.B. Adams, and J.R. Cumming. 2010. A

Decrease in oak litter mass changes nutrient dynamics in the litter layer of a Central

Hardwood Forest. Northern Journal of Applied Forestry:97-104.

Pierce, A.R., G. Parker, and K. Rabenold. 2006. Forest succession in an oak-hickory

dominated stand during a 40-year period at the Ross Biological Reserve, Indiana.

Natural Areas Journal 26:351-359.

193

Pinherio, J., D. Bates, S. DebRoy, D. Sarkar and the R core team. 2009. NLME: Linear and

nonlinear mixed effects models. R package version 3.1-96.

Poland , T.M.and D.G. McCullough, D.G. Emerald ash borer: invasion of the urban forest

and the threat to North America’s ash resource. Journal of Forestry 104: 118-124.

Prasad, A. M. L.R. Iverson, S. Matthews and M. Peters. 2007. A Climate Change Atlas for

134 Forest Tree Species of the Eastern United States

[database].http://www.nrs.fs.fed.us/atlas/tree. USDA Forest Service Northern

Research

Regent Instruments Inc. 2007. WinSCanopy for Canopy Analysis. Regent Instruments,

Canada.

Rhoades, C.C. 2007. The influence of American chestnut (Castanea dentata) on nitrogen

availability, organic matter and chemistry of silty and sandy loam soils. Pedobiologia

50:553-562.

Ross, R.M., R.M. Bennett, C.D. Snyder, J.A. Young, D.R. Smith, and D.P. Lemarie. 2003.

Influence of eastern hemlock (Tsuga Canadensis L.) on fish community structure and

function in headwater streams of the Delaware River basin. Ecology of Freshwater

Fish 12:60-65.

Runkle, J.R. and G.G. Whitney. 1987. Vegetation-site relationships in Lake Katharine State

Nature Preserve Ohio: A northern outlier of the mixed mesophytic forest. Ohio

Journal of Science 87:36-40.

Sasaki, T. and Lauenroth, W.K. 2011. Dominant species, rather than diversity, regulates

temporal stability of plant communities. Oecologia 166:761-768.

194

Scheffer, M., S. Carpenter, J.A. Foley, C. Folke, and B. Walker. 2001. Catastrophic shifts in

ecosystems. Nature 413:591-596.

Scheffer, M., S.H. Hosper, M-L. Meijer, B. Moss, E. Jeppensen. 1993. Alternative equilibia

in shallow lakes. Trends in Ecology Evolution 8:275-279.

Schröder, A., L. Persson, and A.M. De Roos. 2005. Direct experimental evidence for

alternative stable states: a review. Oikos 110:3-19.

Siderhurst, L.A., H.P. Griscom, M. Hudy, and Z.J. Bortolot. 2010. Changes in light levels

and stream temperatures with the loss of eastern hemlock (Tsuga canadensis) at a

southern Appalachian stream: implications for brook trout. Forest Ecology and

Management. 260:1677-1688.

Skinner, M., B. L. Parker, S. Gouli, and T. Ashikaga. 2003. Regional responses of hemlock

woolly adelgid (Homoptera: Adelgidae) to low temperatures. Environmental

Entomology 32:523–528.

Small, M.J, C.J. Small, and G.D. Dreyer. 2005. Changes in a hemlock-dominated forest

following woolly adelgid infestation in southern New England. Journal of the Torrey

Botanical Society 132:458-470.

Snyder C.D., J.A. Young and D. Smith. 2002. Influence of eastern hemlock (Tsuga

canadensis) forests on aquatic invertebrate assemblages in headwater streams.

Canadian Journal of Fisheries and Aquatic Science 59:262–75.

Spaulding, H.L. and L.K. Rieske. 2010. The aftermath of an invasion: Structure and

composition of Central Appalachian hemlock forests following establishment of

hemlock woolly adelgdi, Adelges tsugae. Biological Invasions 12:3135- 3143.

195

Stadler, B., T. Müller and D.A. Orwig. 2006. The ecology of energy and nutrient fluxes in

hemlock forests invaded by hemlock woolly adelgid. Ecology 87:1792-1804.

Suding, K.N. and R. J. Hobbs. 2009. Threshold models in restoration and conservation: a

developing framework. Trends in Ecology and Evolution 24:271-279.

Suding, K.N., K.L. Gross, and G.R. Houseman. 2004. Alternate states and positive feedbacks

in restoration ecology. –Trends in Ecology and Evolution 19:46-53.

Sutherland, E.K., B.J. Hale, and D.M. Hix, 2000. Defining species guilds in the Central

Hardwood Forest, USA. Plant Ecology:1-19.

Thrush, S.F., J.E. Hewitt, P.K. Dayton, G. Coco, A.M. Lohrer, A. Norkko, J. Norkko, and M.

Chiantore. 2010. Forecasting the limits of resilience: integrating emphirical research

with theory. Proceedings of the Royal Society B. 276:3209-3217.

Tingley, M.W., D.A. Orwig, R. Field and G. Motzkin. 2002. Avian response to removal of a

forest dominant: consequences of hemlock woolly adelgid infestations. Journal of

Biogeography 29:1505-1516.

United States Department of Agriculture, Natural Resources Conservation Service. 2012.

The Plants Database url: http://plants.usda.gov. [National Plant Data Team]

Greensboro, NC, USA.

USDA Forest Service Forest Health Protection. 2012. Northeastern Area State and Private

Forestry hemlock woolly adelgid infefstations by state and county. Available:

http://na.fs.fed.us/fhp/hwa/infestations/hwa_infestations11.pdf

Vandermast, D.B. and D.H. Van Lear. 2002. Riparian vegetation in the southern Appalachian

mountains (USA) following chestnut blight. Forest Ecology & Management 155: 97-

196

106.

Vitousek , P.M., J.D. Aber, R.W. Howarth, G.E. Likens, P.A. Matson, D.W. Schindler, W.

H. Schlesinger, and D. G. Tilman. 1997. Human alteration of the global nitrogen

cycles: sources and consequences. Ecological Applications 7:737-750.

Waldrop, T.A. D.A. Yaussy, R.J. Phillips, T.A. Huchinson, L. Brudnak, and R.J. Boerner.

2007. Fuel reduction treatments affect stand structure of hardwood forests in western

North Carolina and southern Ohio, USA. Forest Ecology and Management 255:3117-

3129.

Walker, B., A. Kinzig, and J. Langridge. 1999. Plant attribute diversity, resilience, and

ecosystem function: the nature and significance of dominant and minor species.

Ecosystem 2:95-113.

Wallace, J.B., J.R. Webster, S.L.. Eggert, J.L. Meyer, and E.R. Siler.. 2001. Large woody

debris in a headwater stream: long-term legacies of forest disturbance. International

Review of Hydrobiology. 86:501-513.

Wolf, B.L. and J.R. Thomas. 2006. Soil Survey of Bath County, Virginia. United States

Department of Agriculture Natural Resources Conservation Service. Washington,

DC.

Wurtzburger, N. and R.L. Hendrick. 2006. Rhododendron thickets alter N cycling and soil

extracellular enzyme activities in southern Appalachian hardwood forests.

Pedobiologia 50:563-576.

Wurzburger, N. and R.L. Hendrick. 2009. Plant litter chemistry and mycorrhizal roots

promote a nitrogen feedback in a temperate forest. Journal of Ecology 97: 528-536.

197

Yorks, T.E. D.J. Leopold, and D.J. Raynal. 2002. Effects of Tsuga canadensis mortality on

soil water chemistry and understory vegetation: possible consequences of an invasive

insect herbivore. Canadian Journal of Forest Research 33:1525-1537.

Young, R.F, K.S. Shields, and G.P.Berlyn. 1995. Hemlock woolly adelgid (Homoptera:

Adelgidae): stylet bundle insertion and feeding sites. Annals of the Entomological

Society of America 88:827- 835.

198