Ligustrum Sinense LOUR.)

CAUSES AND CONSEQUENCES OF CHINESE PRIVET (Ligustrum sinense LOUR.)

INVASION IN HYDROLOGICALLY ALTERED FORESTED WETLANDS

Meghan Foard

A thesis presented to the faculty of Arkansas State University in partial fulfillment of the requirements of the degree of

MASTER OF SCIENCE IN ENVIRONMENTAL SCIENCE

Arkansas State University August 2014

Approved by Dr. Travis D. Marsico, Thesis Advisor Dr. Jennifer Bouldin Committee Member Dr. Richard Grippo, Committee Member Dr. Esra Ozdenerol, Committee Member Dr. Paul Sikkel, Committee Member UMI Number: 1563273

INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion.

UMI 1563273 Published by ProQuest LLC (2014). Copyright in the Dissertation held by the Author. Microform Edition © ProQuest LLC. All rights reserved. This work is protected against unauthorized copying under Title 17, United States Code

ProQuest LLC. 789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, MI 48106 - 1346

ii ABSTRACT

Meghan Foard

CAUSES AND CONSEQUENCES OF CHINESE PRIVET (Ligustrum sinense LOUR.)

INVASION IN HYDROLOGICALLY ALTERED FORESTED WETLANDS

What drives invasive species success? My research consists of four studies aiming to answer this question for Ligustrum sinense. The four projects are: (1)

Synthesis of invasion literature within passenger-driver-backseat driver frameworks; (2) hydrochory investigation of water as a dispersal mechanism for invasion; (3) ecohydrology investigation of inundation effects on seed viability of L. sinense; (4) dendrochronology study of the effects of stream channelization and L. sinense invasion on bottomland oak tree growth. Results suggest that L. sinense initially invaded habitats that were hydrologically altered, resulting in drier conditions and a “novel niche.”

Dispersal via hydrochory allowed L. sinense to quickly colonize the novel niches. Once established, L. sinense competed with native oak species contributing to reduced growth rates, an “invasion meltdown.” Control strategies should aim to remove L. sinense and return natural hydrologic regimes, or should consist of human-assisted re-establishment of native species that can thrive in altered conditions.

iii ACKNOWLEDGEMENTS

I would like to thank the Arkansas State University (ASU) Environmental Science

(EVS) Program for providing me with funding, space, and equipment necessary to complete my project. I especially thank the Department Chair of Biological Sciences,

Dr. Thomas Risch, for supporting and helping to guide me throughout my graduate studies at ASU. I thank The Judd Hill Foundation, GK-12, TREE Fund, and The

Arkansas Native Plant Society for additional financial support. Special thanks are extended to Dr. Dorian Burnette of the University of Memphis, and Dr. Dave Stahle and

Dr. Malcolm Cleaveland of The University of Arkansas for aiding me in my dendrochronology studies. My exchanges with them strengthened my passion for and understanding of dendrochronology. I also thank John and Elaine Cobb for allowing me to stay in their guest-house while conducting summer field work.

I would like to thank my committee for their support, insight, guidance, and wisdom. I thank my advisor, Dr. Travis Marsico, for encouraging my interest in L. sinense invasion, dendrochronology, and bottomland hardwood forest ecology. I would also like to thank him for initiating my career in academia, for the many helpful and detailed edits he supplied, for challenging me to think critically, and for teaching me practically everything I know about plants and plant identification. I thank Dr. Jennifer

Bouldin for always being available for advice or assistance when I needed her. I thank

Dr. Esra Ozdenerol for believing in me as a scientist and for encouraging me to be

iv myself and set high goals. I thank Dr. Richard Grippo for going the extra mile when I needed help, and for supporting me and giving me professional guidance. I extend a special thanks to Dr. Paul Sikkel for giving me the encouragement I needed to move forward in stagnant times, for lifting me up when I was down, and for engaging me intellectually. Without his support I would not be the scientist I am today.

I would like to thank the members of the Marsico Lab for helping me with numerous edits on manuscripts, thesis drafts, presentations, and posters. A special thanks is extended to David Burge for always going above and beyond when it came to helping.

His drive, ambition, and insights motivated me in the field and strengthened my understanding of forested wetland ecology. I thank the many technicians and fellow graduate students who helped me collect, assess, and analyze my data including: Hannah

Blair, Jennifer Blanchard, Taylor Mackey, Andrew Neighbors, Hunter Whitehurst, and super special thanks go to Kari Harris and Alexandra Hook for teaching me how to be a mentor, and for making it feel effortless. Last, but certainly not least, I thank Jenn Cobb for assisting with data collections, presentations, papers, and thesis formatting, her time spent helping me is immeasurable and greatly appreciated.

v TABLE OF CONTENTS

LIST OF TABLES ...... ix

LIST OF FIGURES ...... x

LIST OF ABBREVIATIONS ...... xiii

I. INTRODUCTION ...... 1 1.1 Driver, Passenger, and Backseat Driver Models ...... 1 1.2 Bottomland Hardwood Forests ...... 3 1.3 The Invasive Shrub, Ligustrum sinense ...... 4 1.4 The Wolf River, TN ...... 6

II. SYNTHESIS OF L. sinense INVASION WITHIN DRIVER, PASSENGER, AND BACKSEAT DRIVER FRAMEWORKS ...... 8 2.1 Abstract ...... 8 2.2 Introduction...... 9 2.3 The Frameworks and Their Significance in Invasive Species Ecology ...... 10 2.4 Application of Frameworks to L. sinense Invasions ...... 11 2.5 Initial establishment and spread supports the passenger model ...... 15 2.6 After establishment the driver mechanism dominates ...... 19 2.7 Ligustrum sinense Invasion Supports the ‘Novel Niche’ and ‘Invasion Meltdown’ Hypotheses ...... 23 2.8 Discussion and Conclusions ...... 25

III. WATER DISPERSAL OF L. sinense (HYDROCHORY) ...... 28

vi 3.1 Background ...... 28 3.2 Methods ...... 32 3.3 Results ...... 35 3.3 Discussion and Conclusions ...... 44

IV. EFFECT OF LONG-TERM WINTER INUNDATION ON L. sinense GERMINATION AND VIABILITY ...... 48 4.1 Background ...... 48 4.2 Methods ...... 52 4.3 Results ...... 60 4.4 Discussion and Conclusions ...... 65

V. EFFECT OF STREAM FLOW AND L. sinense INVASION ON BOTTOMLAND OAK TREE GROWTH ...... 68 5.1 Introduction...... 68 5.2 Site and Tree Selection...... 71 5.3 Methods for Assessing Tree, Site, and Stand Characteristics ...... 75 5.3.1 Tree Characteristics ...... 75 5.3.2 Site and Stand Characteristics ...... 76 5.4 Determining the Timing of L. sinense Invasion ...... 79 5.5 Dendrochronology Methodology ...... 81 5.5.1 Crossdating and Quality Control ...... 81 5.5.2 Producing Chronologies for Statistical Analyses ...... 82 5.5.3 Identifying the effects of Climate and Flow on Tree Growth ...... 86 5.5.4 Identifying the Effect of L. sinense on Tree Growth ...... 87 5.6 Results of Tree, Site and Stand Characteristics ...... 88 5.6.1 Tree Characteristics ...... 88 5.6.2 Site and Stand Characteristics ...... 94 5.7 Dendrochronology Results...... 104 5.7.1 Quality Control and Chronologies ...... 104

vii 5.7.2 Flow and Climate Signals ...... 108 5.7.3 Signal of L. sinense ...... 114 5.8 Discussion and Conclusions ...... 117 5.8.1 Tree, Site, and Stand Characteristics ...... 117 5.8.2 Effects of Climate and Flow ...... 119 5.8.3 Effects of L. sinense ...... 121

VI. DISCUSSION AND CONCLUSIONS ...... 123

REFERENCES ...... 142

Appendix A: Hydrochory Diversity Characteristics ...... 143

Appendix B: Hydrochory Species Relative Abundance Data ...... 145

Appendix C: Site Characteristics, Presence (1)/absence (0) and similarity data of woody species ...... 156

Appendix D: Species Abbreviations Decoded ...... 158

viii LIST OF TABLES

Table 1. Aspects of Ligustrum sinense invasion placed within Passenger, Driver, and Backseat-Driver frameworks...... 14

Table 2. Monthly diversity and relative abundance of L. sinense seeds dispersed via hydrochory from September 2012-March 2014 ...... 39

Table 3. Tree size, health, and age per site +/- 95% margin of error ...... 90

Table 4. Wetland prevalence indices and percent of occurrence of each wetland indicator status in each site ...... 96

Table 5. Total species richness, and diversity estimates with 95% Tukey confidence intervals ...... 98

Table 6. Average number of pieces, volume (m3), and mass (kg/m3) for invaded and non- invaded plots for each debris type, decay class, woody type, and overall ...... 101

Table 7. Raw Ring Width (RW) and Basal Area Increments (BAI) intercorrelation, sensitivity, and dating information fro each site ...... 105

Table 8. Raw Ring Width (RW) and Basal Area Increments (BAI) intercorrelation, sensitivity, and dating information for each species ...... 106

Table 9. Months of flow with the most significant correlation to site level BAI and RW chronologies ...... 111

Table 10. Months of flow with the most significant correlation to species BAI chronologies ...... 113

ix LIST OF FIGURES

Figure 1. 2011 range and extent of Ligustrum sinense invasion ...... 5

Figure 2. Map of Wolf River, TN with study sites labeled...... 7

Figure 3. Comparison of winter aerial photos of the Wolf River ...... 30

Figure 4. Reflectivity of L. sinense in Shelby Farms along the Wolf River ...... 31

Figure 5. Hydrochory trap in Wolf River tributary ...... 34

Figure 6. Bar graphs of total seeds dispersed via hydrochory per month for all sites, and maximum Wolf River gage height for all collection intervals...... 40

Figure 7. Bar graphs of relative abundance of the top five species dispersed via hydrochory for each site...... 43

Figure 8. Line graph of mean gage height per year before (1936-1963) and after channelization (1964-2010) of the Wolf River, TN...... 50

Figure 9. Line graphs of mean gage height per month before channelization (1936 to 1963) and after channelization (1963-2012) of the Wolf River, TN ...... 51

Figure 10. Ligustrum sinense seeds stained with tetrazolium...... 54

Figure 11. Bag of 100 L. sinense seeds being sewn...... 58

Figure 12. Bag of 100 L. sinense seeds in an inundated forest ...... 59

x Figure 13. Box plot of the viability of L. sinense in DRY and WET treatments for preliminary study ...... 62

Figure 14. Interaction plot between site and treatments for percent germinated in the field inundation study...... 63

Figure 15. Pictures of seed bags from the field inundation study after winter treatments...... 64

Figure 16. Map of dendrochronology sites and trees located along the Wolf River, TN. 74

Figure 17. Master chronologies for each species in raw ring width (RW) and basal area increments (BAI) from 1940 to 2010...... 85

Figure 18. Plot of 95% Confidence intervals for mean tree height at each site...... 89

Figure 19. Plot of 95% Confidence intervals for mean percent canopy missing at each site...... 91

Figure 20. Plot of 95% Confidence intervals for mean tree age at each site...... 93

Figure 21. Plot of 95% Confidence intervals for wetland prevalence indices at each site...... 95

Figure 22. Plot of 95% Confidence intervals for Simpson‟s diversity indices (Ds) at each site...... 99

Figure 23. Pie chart of volume of each type of coarse woody debris (CWD) in 300m2 for NON and INV sites...... 102

Figure 24. Pie chart of volume of coarse woody debris (CWD) for each decay class in 300m2 for NON and INV sites...... 103

xi Figure 25. Master chronologies for each site in raw ring width (RW), Basal Area Increments (BAI) and BAI with flow removed (BAI-FR) from 1940 to 2010...... 107

Figure 26. Wolf River gage height and master chronologies for each site in Basal Area Increments (BAI) from 1940 to 2010 ...... 110

Figure 27. Interaction plots for BAI and BAI-FR showing the BACI effects with control sites grouped together ...... 116

xii LIST OF ABBREVIATIONS

ANOVA = analysis of variance

BACI = before/after, control/impact analysis

BAI = basal area increment chronology

BAI-FR = basal area increment chronology with flow removed

BLHF = bottomland hardwood forest

CNT = non-invaded and non-channelized control site/treatment

CNT/NON = combined chronologies of CNT and NON

CWD = coarse woody debris

DBH = diameter at breast height

INV = invaded and channelized site/treatment

NON = non-invaded and channelized site/treatment

RW = raw ring width chronology

xiii CHAPTER I

INTRODUCTION

1.1 Driver, Passenger, and Backseat Driver Models

According to driver, passenger, and backseat driver models (MacDougall and

The passenger model suggests that the spread of invasive species results from environmental conditions within invaded habitats, such as numerous dispersal vectors or a multitude of disturbances that reduce the native species resources, giving advantages to the invasive species (MacDougall and Turkington 2005; Richardson et al. 2007). This situation could occur if the disturbance regime to which the native species is adapted is altered resulting in newly maladaptive traits in the native range. The driver model, on the other hand, states that the success of the invasive species results from direct competition with native species for shared resources such as sunlight, nutrients, and water

1 (MacDougall and Turkington 2005). The driver model suggests that the traits of the invasive species make it more competitive than the native species in the introduced environment. This could be because competition was greater for the invasive species in its native range. The backseat driver model involves a combination of the passenger and driver models. It states that initial invasions are promoted by altered environmental conditions that lead to the decline in health or performance of native habitats and/or species, and once present invasive species further diminish the fitness of native species

(Bauer 2012).

These models are important for understanding invasive species mechanisms, and understanding species invasions in the context of these models will aid ecosystem restoration efforts. If an invasive species is deemed a passenger then ecosystem recovery will not proceed from removal of the invasive alone. However, if the invasive is the driver responsible for the alteration then the removal should result in a return of native biodiversity and pre-invasion ecosystem functions (Colautti and MacIsaac 2004;

Gurevitch and Padilla 2004; Didham et al. 2005; MacDougall and Turkington 2005). To properly mitigate invasions if the invader is a backseat driver rigorous removal must be accompanied by a return to pre-existing conditions, or removal followed by establishment of species that can thrive in the altered conditions (Bauer 2012). When formulating eradication and prevention techniques it is important to understand the source of invasion because it deals with rudimentary issues and not side effects (Gurevitch and Padilla 2004;

2 1.2 Bottomland Hardwood Forests

Bottomland hardwood forests (BLHFs) perform valuable ecosystem functions.

They sequester nutrients and sediments, filter groundwater, and create habitats for a high diversity of endemic plant and animal species (Johnston et al., 1990, Richardson et al.

2007). Species that thrive in these forested wetlands are often specialists that require a restricted set of physical conditions to survive and reproduce (Shankman 1996). These species are relatively non-competitive and rely on seasonal inundation, or flood pulses, to remain prolific. They have many adaptations that give them an advantage in riparian dynamics. Physical adaptations to anaerobic conditions, that result from frequent, long- term inundation include, shallow root systems, tree buttressing, formation of lenticels, and in some cases aerial roots (Middleton 2002). Many bottomland species reproduce vegetatively as an adaptation to frequent high flow events that damage and break limbs

(Planty-Tabacchi et al. 1996; Naiman and Décamps 1997; Hood and Naiman 2000).

Riparian species also take advantage of flood pulses as a means of seed dispersal

(Schneider and Sharitz 1986, Guo et al. 1998, Capon 2003, Richardson et al. 2007).

Because of their dependency upon seasonal flood pulses, BLHFs are fragile ecosystems.

Unfortunately, by 1980 over 50% of all BLHFs in the United States were destroyed as a result of deforestation linked to agriculture (Gosselink et al. 1990) and indirectly by urbanization (DeFries et al. 2010). Furthermore, the few remaining BLHFs are threatened by invasive species and altered hydrologic regimes (Vitousek et al. 1996,

Hood and Naiman 2000, Middleton 2002, Richardson et al. 2007).

3 1.3 The Invasive Shrub, Ligustrum sinense

Ligustrum sinense Lour. (Chinese privet) is a broad-leaved, semi-evergreen, invasive shrub that dominates hydrologically altered riparian zones throughout the world (Ward

2002, Grove and Clarkson 2005, Pokswinski 2008). As one of the top invaders of the southeastern United States L. sinense occupies land in nearly every county in the region and is expanding its range to the north and west (Fig. 1). Ligustrum sinense has many life history traits that make it perfectly adapted to the dynamics of riparian floodplains.

When grown in hydric conditions L. sinense produces lenticels and aerial roots (Brown and Pezeshki 2000), it can reproduce asexually through broken limbs or runners (Morris et al. 2002), and L. sinense seeds are highly buoyant and are dispersed by water into

BLHFs (Ward 2002, Foard Chapter 3.3).

Figure 1. 2011 range and extent of Ligustrum sinense invasion, showing that almost all of the southeastern States are affected. Counties shaded in white are those with documented L. sinense. Counties shaded in black do not have a record of L. sinense but occur within States where L. sinense has been recorded. Gray regions are those States and counties without L. sinense records. Map generated from The Biota of North America Program (Kartesz 2011)

5 1.4 The Wolf River, TN

Globally, many rivers and streams adjacent to BLHFs were hydrologically altered to control flooding to allow for human settlement. These alterations include damming, rerouting, and/or channelizing waterways (Middleton 2002). In the 1920s The U.S.

Army Corps Of Engineers began a large-scale project to channelize the Wolf River, a major tributary of the Mississippi River in southwestern Tennessee. Within 35 years the lower 35.2 km were channelized. Although channelization ceased in 1964, the ecology of the Wolf River BLHF was permanently altered and the water table lowered substantially (Weins and Roberts 2003). Following channelization, the natural flooding regime changed and the surrounding regions no longer contained the necessary hydrology to remain classified as forested wetlands (Fig. 2) (Weins and Roberts 2003). After the land surrounding the river became drier (i.e., decreased flood frequency and duration), new, competitive, woody species from surrounding upland populations should be expected to occupy the area (Shankman 1996). However, L. sinense was one of the few species to establish, thrive, and reproduce (Houston et al. 2010). Years later, the established bottomland hardwood species began to exhibit severe canopy loss and above average tree mortality (Houston et al. 2010, Foard Chapter 5.6.1). The goal of my study is to identify the causes and consequences of the invasion of L. sinense along the Wolf

River, TN.

Figure 2. Wolf River, Tennessee, with the three study sites labeled. There are two channelized sites – L. sinense-invaded (INV) and non-invaded (NON) – and an unchannelized site with no invasion that serves as a control (CNT).

7 CHAPTER II

SYNTHESIS OF L. sinense INVASION WITHIN DRIVER, PASSENGER, AND

BACKSEAT DRIVER FRAMEWORKS

2.1 Abstract

Current researcher suggests that invasive species are less responsible for extinctions as previously reported. Some authors have postulated that invasions cause extinctions, while others have stated that invasions are a side effect of extinction and habitat loss (Vitousek 1996, Gurevitch and Padilla 2004). It is crucial to understand the ecological role of invasive species when implementing effective ecosystem restorations.

Models developed by MacDougall and Turkington (2005) and Bauer (2012) helped identify the role invasive species can play in reducing biodiversity. My goal is to determine which of three conceptual models: “driver,” “passenger,” or “backseat driver,” best explains the successful invasion of Ligustrum sinense (Chinese privet) My approach combines a synthesis of all publications regarding L. sinense invasion, with empirical data collected in Memphis, Tennessee, USA. This combined approach supports the hypothesis that initial L. sinense invasions are passive and begin with habitat alteration and superior dispersal. Once established, however, L. sinense directly competes with native plants causing further reduction in biodiversity. These data

8 support the backseat driver model (Bauer 2012). Additionally, I use the data from my synthesis to determine the role of “novel niches” and “invasion meltdown” in the invasion of L. sinense, and I found that in early stages of invasion, L. sinense fits into the

“novel niche” hypothesis, and, after years of establishment, it fits into the “invasion meltdown” hypothesis. Finally, I use my results to identify potential hotspots for future invasions, and determine guidelines and practices for mitigating L. sinense invasion.

2.2 Introduction

Reduced biodiversity is among the most recognized and debated effects of invasive species (Ricciardi 2002; MacDougall and Turkington 2005; Larson 2007; Light and Marchetti 2007; Bradley et al. 2009). Some researchers argue that invasive species are a major cause of reduced native species biodiversity (Vitousek et al. 1996; Wilcove et al. 1998; Mooney and Cleland 2001; Light and Marchetti 2007; Greene and Blossey

2011), while others have argued that the success of invasive species is a result of factors in the invaded environment, such as a lack of biodiversity resulting from disturbances or altered disturbance regimes (Ward 2002; Gurevitch and Padilla 2004; Didham et al.

2005; MacDougall and Turkington 2005). Thus, recent work suggests that an examination of species invasions within organized frameworks is crucial for determining the cause(s) of invasion success and loss of biodiversity (Colautti and MacIsaac 2004;

Gurevitch and Padilla 2004; Didham et al. 2005; MacDougall and Turkington 2005;

Richardson et al. 2007; Catford et al. 2009; MacDougall et al. 2009; Catford and Downes

2010; Gurevitch et al. 2011). These authors are motivated by a need to develop

9 conservation and restoration strategies that can target the exact source(s) of an invasion to make the greatest impact.

2.3 The Frameworks and Their Significance in Invasive Species Ecology

According to driver, passenger, and backseat driver models (MacDougall and

Turkington 2005, Bauer 2012), invasive species either “drive” ecological degradation through direct interactions with their native competitors, are “passengers” taking advantage of previously altered systems, or are “backseat drivers” passively taking advantage of ecological alterations, and after invasion they further reduce the fitness of the native species present. The driver model states that the success of invasive species results from direct competition with native species for resources such as sunlight, nutrients, water, and space (MacDougall and Turkington 2005). The driver may also alter the environment in ways that make it less habitable for native species and more habitable for invasive species, in a process dubbed “invasion meltdown” (Simberloff and

Von Holle 1999). The passenger model describes that the establishment of invasive species results from environmental conditions within the invaded habitat, such as disturbances that reduce the native species resources and/or ability to reproduce, thus giving a competitive advantage to the invasive species (MacDougall and Turkington

2005; Richardson et al. 2007). This situation could occur if the disturbance regime to which the native species are adapted becomes altered, creating a “novel niche” which promotes establishment of species that can thrive in the new conditions at the expense of the now-maladapted native species. The backseat driver model involves a combination

10 of the passenger and driver models. It states that environmental conditions that lead to the decline in health of native species promotes invasions, and once present invasive species further diminish the fitness of native species (Bauer 2012).

These models aid ecosystem restoration efforts. When formulating eradication, management, and prevention techniques it is helpful to focus these efforts on the rudimentary causes of invasion, and prevent reoccurrences (Gurevitch and Padilla 2004;

Catford et al. 2009; MacDougall et al. 2009; Catford and Downes 2010). Likewise, it is best to recognize influences promoting initial invasions and monitor habitats at risk for invasions, as a prevention effort (Catford and Downes 2010). If an invasive species is deemed the driver responsible for the alteration then removal should result in a return of native biodiversity and pre-invasion ecosystem functions. However, if the invasive species is a passenger then ecosystem recovery will not result from removal of the invasive species alone (Colautti and MacIsaac 2004; Gurevitch and Padilla 2004; Didham et al. 2005; MacDougall and Turkington 2005). If it is a backseat driver then rigorous removal of the invasive species is the first step in eradication. Then either a return to pre- existing conditions, or human-assisted establishment of species that can thrive in the altered conditions, must be implemented (Bauer 2012).

2.4 Application of Frameworks to L. sinense Invasions

Ligustrum sinense Lour. (Chinese privet) is a broad-leaved, semi-evergreen shrub that is invasive in many temperate riparian habitats throughout the world (Grove and Clarkson

2005). It has become one of the top invaders of the southeastern United States where it

11 occupies land in nearly every county in the region, and is currently expanding its range to north- and westward. It has been shown repeatedly that when L. sinense is present in an ecosystem it is accompanied by a substantial decrease in native biodiversity (Merriam and Feil 2002; Grove and Clarkson 2005; Wilcox and Beck 2007; Brantley 2008;

Pokswinski 2008; Hanula et al. 2009; Klock 2009; Osland et al. 2009; Ulyshen et al.

2010; Greene and Blossey 2011; Hanula and Horn 2011a,b). However, a direct causal link has not been established.

Many studies attempt to determine the principal cause of L. sinense invasion success without considering appropriate theoretical models, thereby missing critical elements to understanding the mechanism(s) of success (Colautti and MacIsaac 2004;

Gurevitch and Padilla 2004; Didham et al. 2005; MacDougall and Turkington 2005).

Some invasive species ecologists acknowledge L. sinense as a driver but do not address the mechanism by which L. sinense established (Greene and Blossey 2011). Although the exact cause of invasion success is unknown, rapid growth, large reproductive output, dispersal by native birds and mammals, tolerance of highly variable environmental conditions, and superior competitive ability have all been implicated (Grove and

Clarkson 2005; Wilcox and Beck 2007; Brantley 2008; Pokswinski 2008; Greene and

Blossey 2011).

My analysis shows that early establishment of L. sinense has been assisted by environmental alteration, making L. sinense a passenger. Yet, after long-term establishment, L. sinense drives the alteration of invaded habitats and reduces native species abundance, diversity, and richness. Therefore, since the invasion does not fit into either category, it should be viewed in the “backseat driver” context so that a more

12 comprehensive plan to address invasive species mitigation can be developed. Thus, when implementing recovery in L. sinense-invaded areas, the invasion as well as the associated loss of regime, habitat, and biodiversity must be considered.

13 Table 1. Multiple aspects of Ligustrum sinense invasion are placed within driver, passenger, and backseat driver frameworks. Of the 14 invasion characteristics, the total number of “Yes” answers in each column shows that Driver and Passenger have important and nearly equal impacts on L. sinense success. Five of the characteristics fit into Driver and Passenger scenarios and are placed within the Backseat Driver category Ligustrum sinense invasion Driver? Passenger? Backseat Driver? characteristics Propagule pressure Yes-once Yes- Yes established horticultural reproduces practices vigorously Occupancy of altered No Yes No habitats (novel niche) Occupancy of riparian areas No Yes No Long distance dispersal No Yes No Multiple dispersal vectors in No Yes No non-native range Asexual reproduction Yes-when Yes Yes superior to natives Mass quantities of fruit Yes-when Yes Yes production superior to natives High germination rates Yes-when Yes Yes superior to natives Competition between Yes No No seedlings Increased native species Yes No No after L. sinense removal Allelopathy Yes No No Decreased herbivory Yes-when Yes Yes superior to natives Phenotypic plasticity Yes-when No No superior to natives Alteration of pre-existing Yes No No habitats

TOTAL Yes 10 9 5

14 2.5 Initial establishment and spread supports the passenger model

Ligustrum sinense can be classified as a passenger when considering its horticultural introduction history. Ligustrum sinense was brought to the U.S. in 1852 as an ornamental hedgerow, and it has been widely planted in both commercial and residential areas (Stromayer et al. 1998). This extensive propagule pressure assisted the colonization, reproduction, and spread of the species (Lockwood et al. 2005; Barney and

Whitlow 2008). Commerce is responsible for the initial introduction, colonization, and dispersal of L. sinense into a variety of habitats, and human activities continue to improve its chances of naturalization and success.

Ligustrum sinense is found more commonly in disturbed habitats as a passenger.

Ligustrum sinense populations increase with proximity to urban disturbances in western

Georgia (Burton et al. 2005). Importance values range from 94.8 in the regeneration layer of an urban forest, to a value of 0 in the regeneration layer of diverse pine forest in a rural setting (Burton et al. 2005). However, this could also be a consequence of propagule pressure, because L. sinense is planted by humans whose presence defines

“urban.” In either case, the data from Burton et al. (2005) support the passenger model.

Importance values represent the population size of, the area occupied by, and the likelihood of encountering a particular species, in this case that species is L. sinense. The importance value is calculated as follows:

Importance = Relative density + Relative frequency + Relative Dominance

15 Ligustrum sinense is more densely populated in the channelized sections of a

Mississippi River tributary in western Tennessee than in less altered nearby regions

(Weins and Roberts 2003, Houston et al. 2010). Channelization along the Wolf River,

Tennessee, altered the natural flooding regime, and surrounding BLHFs no longer possessed the necessary hydrology to support the wetland species present (Weins and

Roberts 2003). In a preliminary survey in 2011 of all woody stems greater than 1.5cm diameters, in random areas of 3,852 m2 for channelized portions and 1,841 m2 for unchannelized sections along the Wolf River, I determined that the importance values of L. sinense were 163.1 where channelization occurred and 0 where the river had not been channelized. Ligustrum sinense density is also greatest in channelized and hurricane- damaged portions of Watoula Creek Riparian Area, in Lee County, Alabama (Pokswinski

2008).

Ward (2002) conducted the most comprehensive historical analysis of L. sinense establishment along the Oconee River, Georgia. He investigated the influence of land use history on L. sinense invasion and, as far as I can determine, is the only researcher to indicate water as a major method of dispersal. Using aerial photography he assessed rates of spread and distribution of L. sinense compared to land use history from 1951 to

1999. He concluded that L. sinense frequently invades areas that have anthropogenic disturbances. The most significant of these disturbances was upstream cotton farming that resulted in increased sediment deposition downstream and disrupted the natural flow of the river, resulting in shallower, braided streams (Ward 2002). Ligustrum sinense invasion along the Oconee River began after microhabitat riparian locations became drier. Ward‟s results indicate that L. sinense responded considerably to habitat alterations

16 resulting in reduced inundation in riparian floodplains. The most probable explanation for this account is that the communities that were present before the hydrological alteration were adapted to wet hydrological conditions. Therefore, when the conditions changed the native communities were unable to remain prolific and new upland species were allowed into the resulting drier riparian habitat; the most successful of these species was L. sinense (Capon 2003; Weins and Roberts 2003; Richardson et al. 2007; Catford et al. 2009; MacDougall et al. 2009).

Furthermore, L. sinense may be considered a passenger in all riparian zones regardless of alterations because riparian dynamics promote novel niches and thus, invasions (Richardson et al. 2007; MacDougall et al. 2009). Frequent fluctuations in hydrology as well as numerous other disturbances make riparian zones naturally susceptible to invasions (Thebaud and Debusshe 1991; Planty-Tabacchi et al 1996;

Naiman and Decamps 1997; Hood and Naiman 2000; Richardson et al. 2007). The more frequent the fluctuations, the more vulnerable the ecosystem (Hood and Naiman 2000).

One can also make inferences about the drive of the invasive species by looking at the other species that remain in the invaded ecosystem after a full invasion takes place

(Colautti and MacIsaac 2004; MacDougall and Turkington 2005; Didham et al. 2005;

Catford and Downes 2010). If functionally similar species remain, then the invasive species is considered a passenger because it indicates a lack of direct competition for shared resources (MacDougall and Turkington 2005). In an ecological analysis of L. sinense in New Zealand, it was observed that L. sinense grows with common lowland shrubs (Grove and Clarkson 2005), which indicates the passenger model. However, it is

17 possible that in New Zealand changes in resources enhanced the successful inhabitance of all shrubs, ultimately reducing the necessity of competition for shared resources.

Ligustrum sinense can also be viewed as a passenger when looking at its methods of reproduction and dispersal. It reproduces vegetatively through runners or though root induction on broken limbs, and reproduces sexually via fruit production (Dirr 1998).

Combined asexual and sexual reproduction is most advantageous for survival of woody plant species in riparian zones (Naiman and Décamps 1997). The frequent inundation of riparian zones results in damaged and broken limbs (Planty-Tabacchi et al. 1996; Naiman and Décamps 1997; Hood and Naiman 2000; Richardson et al. 2007), and L. sinense is highly successful at sprouting ramets from these broken stems and stumps that remain

(Morris et al. 2002). The seeds also have a high germination rate of 29.05% (mean of all the following germination trials under all conditions, range 0.3% to 80%; Burrows and

Kohen 1986; Panetta 2000; Grove and Clarkson 2005; Klock 2009). Seeds also are produced in abundance: A single L. sinense trunk (measured as one individual ramet from the base) produces an average of 2,580 single-seeded fruits per year (Westoby et al.

1983; Klock 2009). If 29.05% germinate then each trunk produces approximately 750 offspring per year. Given that a reproductive plant typically has approximately 4.5 basal trunks, an average adult L. sinense plant can be expected to produce approximately 3,375 germinated offspring in a single year. All these fruits are either dropped by L. sinense at its base or are dispersed through various other methods (Matlack 2002).

Birds, deer, opossums, and waterways are conduits for L. sinense dispersal

(Stromayer et al. 1998; Panetta 2000; Williams et al. 2000; Grove and Clarkson 2005;

Strong et al. 2005; Ward 2002; Brantley 2008; Klock 2009). Ward (2002) suggests that

18 water dispersal is an important aspect to the spread of L. sinense, although birds are currently considered the most important of all dispersal vectors (Matlack 2002; Merriam

2003; Grove and Clarkson 2005; Strong et al. 2005; Brantley 2008; Pokswinski 2008;

Klock 2009). Both birds and water as dispersal vectors allow long distance dispersal that is advantageous for reducing intraspecific competition with parent individuals (Howe and

Miriti 2004). Regardless of the dispersal mechanism, multiple methods of dispersal provide L. sinense with the greatest probability of locating a suitable habitat (Schneider and Sharitz 1986; Richardson et al. 2007). Dispersal that is superior to native species is highly suggestive of the passenger model (MacDougall and Turkington 2005).

2.6 After establishment the driver mechanism dominates

The fruits of L. sinense, which ripen from September to December in the southeastern U.S. (Matlack 2002), are a common food source for birds (Grove and

Clarkson 2005; Strong et al. 2005). After birds eat the fruit of L. sinense they often perch atop native hardwood trees and releases the seeds in excrement. The seeds then germinate at the bases of these trees, which leads to direct competition between seedlings, often resulting in L. sinense as the victor (Miller 1997; Langeland and Burkes

1998; Brown and Pezeshki 2000; Kittell 2001; Merriam and Feil 2002; Wilcox and Beck

2007). Morris et al. (2002) showed that L. sinense seedlings grew taller and faster in situ than seedlings of a functionally similar, and related, native shrub Forestiera ligustrina

(upland swamp privet). Furthermore, when L. sinense seedlings compete against native

19 plants, L. sinense makes up the largest proportion of survivors in the next growing season

(Merriam and Feil 2002; Greene and Blossey 2011).

Pokswinski (2008) provides evidence for the driver mechanism by showing that as the abundance of L. sinense increased, the abundance of the forest shrub, Leucothoe axillaris (Lam.) D. Don. (coastal doghobble), decreased; however, the two species prefer different hydrology thus water availability could be the driving force. Others in the southeastern U.S. noted dissimilar plants, such as vines, growing within dense L. sinense stands (Merriam and Feil 2002; Wilcox and Beck 2007; Klock 2009). The implications of these observations are that L. sinense is a driver and has out-competed similar species and those that remain require a different niche than L. sinense, allowing them to persist after L. sinense invaded (Tilman et al. 1997). However, in these situations it is not evident that similar species were present before the invasion and more information about prior community composition is necessary to conclude L. sinense is a driver.

The driver model also states that when an invasive species is directly causing decreased biodiversity then the removal of the invasive from fully invaded sites should result in increased native biodiversity (MacDougall and Turkington 2005). When a dense canopy of L. sinense was removed in November 1999, increases in native species diversity resulted the following spring (Merriam and Feil 2002). Hanula et al. (2009) also showed an increase in native plant abundance after the removal of large areas of L. sinense. Both of these studies show that native plants and/or seeds are still present and being dispersed into regions invaded by L. sinense. In a large-scale study of Oconee

River riparian zones in Georgia, large areas of L. sinense were removed and the community resilience was monitored over a two-year period (Hanula et al. 2009; Ulyshen

20 et al. 2010; Hanula and Horn 2011a, b). The study showed not only a response by plants, but removal of L. sinense resulted in increased beetle (Coleoptera) species richness and diversity (Ulyshen et al. 2010), increased species richness of native bees (Hymenoptera)

(Hanula and Horn 2011a), and increased native butterfly (Lepidoptera) species richness in areas where L. sinense was mulched (Hanula and Horn 2011b). Additionally, native river cane (Arundinaria gigantea (Walter) Muhl.) increased after the removal of L. sinense from riparian zones in Durham County, North Carolina (Osland et al. 2009).

Thus, much evidence supports the driver model once L. sinense is established.

Greene and Blossey (2011), who concluded that L. sinense is a driver, demonstrated that the presence of L. sinense resulted in decreased native herbaceous and woody plant survival. They showed the drive of L. sinense invasion by conducting a transplant experiment. Four common native riparian seedlings were planted in areas with fully established L. sinense canopy and areas with none. Growth and survival was significantly higher for all species in non-invaded areas (Greene and Blossey 2011).

Thus, evidence shows that after long-term invasion L. sinense drives mortality and limits recruitment of native species and is a driver (Greene and Blossey 2011).

Another key factor that indicates direct competition between established L. sinense and native species is that L. sinense foliage is potentially allelopathic.

Allelopathy has been tested in leaf extract and in leaf-and-root extract (Grove and

Clarkson 2005; Pokswinski 2008, respectively). Grove and Clarkson (2005) germinated radish seeds either with or without L. sinense leaf leachate, and the results showed that

80% of the radishes in the leaf leachate germinated vs. 92% germination in the control.

Pokswinski (2008) discovered similar results when he germinated tomato plants in

21 differing concentration of leaf, root, or leaf-and-root extract of L. sinense. He discovered that with increasing concentrations of L. sinense leachate, there were decreasing sizes of radicle lengths of the tomato, showing that L. sinense leaves and roots had an adverse effect on the success of tomatoes. Both of these studies support that L. sinense directly drives the reduction of other species; however, neither of these studies tests allelopathy on native species that grow in the invaded areas of L. sinense, so impacts of essential allelopathic chemicals in situ remain unknown.

Ligustrum sinense has lower levels of herbivory than a native shrub in the same family (Oleaceae) Forestiera ligustrina (Morris et al 2002). However, to frame herbivory in the model, it must be known whether direct defense, or lack of recognition/preference of native herbivores for L. sinense is responsible. Another

Ligustrum species, L. obtusifolium (border privet), produces a nutritive plant defense against herbivores (Konno et al. 2009). This is likely true for many Ligustrum species including L. sinense (Konno et al. 2009). If herbivores choose to feed on L. sinense less than native shrubs, regardless of the reason, a positive feedback results in which native species are eaten but not the invasive species, supporting the driver model. Ligustrum sinense also has numerous qualities of phenotypic plasticity that make it a superior competitor against the native shrub F. ligustrina (Morris et al. 2002). For example, L. sinense is dissimilar in appearance when grown in upland and bottomland ecosystems

(Klock 2009).

Finally, it is probable that in an invaded terrestrial ecosystem, animal dispersers are less likely to visit a native plant simply because native plants are much less frequently encountered. This can lead to changes in the foraging behavior of dispersers (Traveset

22 and Richardson 2006). White tailed deer, for example, prefer to eat native seeds when available, but were found to have consumed a higher percentage (75%) of L. sinense than was represented in the forest (50%). This is likely because the clonal growth of L. sinense results in fruit and foliage that are low to the ground and abundant, allowing for minimization of deer foraging range (Stromayer et al. 1998). Conversely, the Hermit

Thrush, a bird native to Louisiana, prefers to eat the seeds of a native shrub, Ilex vomitoria (Yaupon holly) over L. sinense despite the greater abundance of L. sinense

(Strong et al. 2005). Ligustrum sinense, however, was the second largest component of their diet suggesting competition for dispersers. Also, competition for pollinators can lead to fewer seeds produced by natives, reducing fecundity (Traveset and Richardson

2006).

2.7 Ligustrum sinense Invasion Supports the ‘Novel Niche’ and ‘Invasion Meltdown’

Hypotheses

The results from the synthesis reveal that early establishment of L. sinense was assisted by environmental alteration, making L. sinense a passenger into “novel niches.”

Yet, after long-term establishment L. sinense drives the alteration of invaded habitats and reduces native species abundance, diversity, and richness, through a process termed invasion meltdown. Therefore, the invasion should be viewed in both contexts so that a more comprehensive plan to address invasive species mitigation can be developed.

The most probable explanation for the extremely successful invasion of L. sinense in urbanized sites is that the species that were present before habitat alteration were

23 adapted to pre-disturbance conditions. Therefore, when the conditions changed as a consequence of urbanization, the native species were unable to remain competitive against the better-adapted invader, L. sinense (Weins and Roberts 2003; Catford et al.

2009; MacDougall et al. 2009). This invasion mechanism supplies evidence for the

“novel niche” hypothesis (MacDougall et al. 2009). For example, many native bottomland species require restricted sets of physical conditions to survive and reproduce

(Shankman 1996; Richardson et al. 2007), and after rivers become altered, new woody species from surrounding populations should begin to occupy the area (Shankman 1996).

However, L. sinense is among the only woody species to thrive and reproduce along altered rivers of the Southeast United States (Ward 2002; Weins and Roberts 2003;

Burton et al. 2005; Pokswinski 2008), and hence supports the “novel niche” hypothesis.

The “invasion meltdown” hypothesis postulates that invasive species drive local ecological change that results in more suitable conditions for the invasive (Simberloff and

Von Holle 1999; Ashton et al. 2005; MacDougall and Turkington 2005). Ligustrum sinense alters its environment by producing dense shade in which only highly shade tolerant species, such as itself, can grow (Grove and Clarkson 2005; Brantley 2008;

Smith et al. 2008; Osland et al. 2009; Greene and Blossey 2011). One additional method of habitat modification results from semi-evergreen foliage that contains less lignin and more nitrogen than typical mixed-species foliage from deciduous hardwood forests

(Mitchell 2009). This results in higher rates of decomposition for L. sinense leaves

(Zhang et al. 2009), which presumably increases soil fertility and growth and reproduction of plants present, in this case predominately L. sinense.

24 Ligustrum sinense is similar to the invasive water-dispersed riparian shrub

Sesbania punicea (Cav.) Benth. (rattlebox) in South Africa (Hoffman and Moran 1991) because it is likely that L. sinense alters its environment is by collecting mass amounts of sediments from floodwaters and raising the soil level around the colonies of densely packed stems, further reducing moisture content of topsoil (Ward 2002). Trapped sediment may result in drier surface conditions to which most riparian native species are not well adapted, and L. sinense can take advantage. In this case a novel niche is created by the invasive species, which remains prolific while keeping out competitors

(MacDougall et al. 2009; Richardson et al. 2007).

A final way L. sinense alters the ecosystem and decreases native plant success is by closing off dispersal passageways that occur naturally as a result of abandoned flow routes and seasonal floods. These passageways are crucial for native seed dispersal by wind and vertebrates (Richardson et al. 2007), thus, native species recruitment is limited by occupancy of these pathways by L. sinense. Additional mechanisms for ecosystem alteration by L. sinense are likely yet to be discovered.

2.8 Discussion and Conclusions

Invasions often accompany habitat alteration, and it is difficult to discern which aspect is most influential to ecosystem processes (Vitousek 1994; Simberloff 2000;

Bradley et al. 2009; Walther et al. 2009; Diez et al. 2012). Moreover, many ecosystems differ from historical conditions due to habitat alterations, introductions of non-native species, pollution, climate change, and other anthropogenic effects. Thus, it is difficult to

25 generalize the responses of ecosystems to invasive species (Vitousek 1994; Vitousek et al. 1996; Simberloff 2000; Mooney and Cleland 2001; Clavero and García-Berthou 2005;

Light and Marchetti 2007; Bradley et al. 2009; Walther et al. 2009; Catford and Downes

2010). Much evidence exists to support L. sinense as a backseat driver; it often invades habitats that are altered, taking advantage of a novel niche (Ward 2002; Pokswinski 2008;

MacDougall et al. 2009). Superior long distance dispersal allows L. sinense to colonize these novel niches (Ward 2002; Howe and Miriti 2004; Richardson et al. 2007). Once established, L. sinense competes with native species, reduces biodiversity, and alters the microenvironment (Stromayer et al. 1998; Grove and Clarkson 2005; Strong et al. 2005;

Brantley 2008; Mitchell 2009; Hanula et al. 2009; Osland et al 2009; Ulyshen et al. 2010;

Greene and Blossey 2011; Hanula and Horn 2011a,b).

Because L. sinense is a passenger and a driver, and therefore a backseat driver, resource managers should consider unifying approaches to resolve the problems associated with invasion and impacts. Restoration plans must address both the invasive species and the alteration of habitat; a synergistic approach (MacDougall and Turkington

2005, Bauer 2012). Therefore, the removal of L. sinense from invaded ecosystems accompanied with reintroduction of natives that can thrive in a drier environment may provide ecosystems the best chance for recovery (Colautti and MacIsaac 2004; Gurevitch and Padilla 2004; MacDougall and Turkington 2005; Didham et al. 2005; Richardson et al. 2007; Catford et al. 2009; MacDougall et al. 2009; Catford and Downes 2010).

Finally, riparian ecosystems that have been altered by human activities are most susceptible to L. sinense invasion (Richardson 2007). Moreover, floods mediate dispersal to these disturbed areas (Chapter 3) and these habitats must be closely monitored

26 following early season floods to recognize and mitigate establishment of new invasive populations. I suggest that historical flood regimes be determined and correlated with areas of altered hydrology, such as channelized rivers. Finally, I hypothesize that the hotspots most likely to be invaded by L. sinense are areas where hydrology has been altered and have recently experienced a seed-dispersing flood.

27 CHAPTER III

WATER DISPERSAL OF L. sinense (HYDROCHORY)

3.1 Background

Hydrochory, or water dispersal, is the most common method of dispersal of

BLHF species (Middleton 2002; Richardson 2007). Although birds, deer, and opossums are all recognized dispersers of L. sinense (Stromayer et al. 1998; Panetta 2000; Williams et al. 2000; Grove and Clarkson 2005; Strong et al. 2005; Brantley 2008; Klock 2009;

Hudson 2013), it is likely water also plays a major role in the dispersal into riparian zones, which is evident by its distribution. Many researchers note large populations of L. sinense in ditches, along riparian zones, in lowlands, and in disturbed wetlands (Brown and Pezeshki 2000; Panetta 2000; Grove and Clarkson 2005; Brantley 2008; Osland et al.

2009). Additional researchers found that L. sinense density increases with proximity to lakes, rivers, and streams (Merriam 2003; Hudson 2013), and that proximity to water is the best predictor of L. sinense presence (Wang and Grant 2012). I hypothesized the buoyant fruits of L. sinense, which are produced in higher quantities in full sun (Brown and Pezeshki 2000), are carried by rainwater, drainage canals, rivers, and other flowing water to lowland sites where they germinate after the water recedes.

28 Ligustrum sinense fruits are round, purplish-black, fleshy drupes usually a few millimeters long and highly buoyant (100% of the L. sinense fruits used in this study were buoyant). According to Grove and Clarkson (2005) the removal of the fleshy coat does not significantly affect the germination rate of L. sinense, which supports the idea that bird digestion is not necessary for germination and suggests that water dispersal could play a much greater role in dispersal than previously thought.

Ward (2002) suggested hydrochory plays a role in L. sinense dispersal, which supports my hypothesis; however, he did not collect any empirical data to support these claims. Thus, this research is important in verifying a previously unrecognized dispersal vector for L. sinense. Historical aerial observations and current remote sensing data of L. sinense in forests adjacent to the Wolf River, Tennessee, show that L. sinense is most prominent in the area immediately adjacent to the pre-channelized path of the Wolf River

(Figs. 3 and 4). The L. sinense invasion in this site has a dense and noticeable transition zone (Fig. 4) (Quarles 2012). Prior to channelization, this invaded location was within the Wolf River flood zone (compare Fig. 3a where open gaps in the flood zone are apparent with Fig. 3b where gaps are no longer present). The adult tree species north of the invasion consist of a greater proportion of upland species which supplies further evidence that the non-invaded area is historically drier, and therefore receives less flooding. Water dispersal of L. sinense fruits into the flood zone would explain why the invasion is much more intense in the southern section, and why there is little invasion in the northern section of forest. Given that L. sinense invades dry habitats it must be limited by dispersal in the northern sites (Dirr 1998).

29 30

Figure 3. Comparison of winter aerial photos of the Wolf River near Memphis, Tennessee. Figure 3a. (left) shows the flood zone immediately surrounding the meandering remnant of the Wolf River in 1954. Visible are lighter-colored large gaps between trees, representing the absence of L. sinense. Figure 3b. (right) shows the current range and extent of L. sinense invasion can be seen to overlap the same area as the historical floods. Because L. sinense is the only evergreen in this location today, the presence is clearly indicated by the dark “popcorn” appearance of the area immediately surrounding the historical Wolf River flood zone. The range of L. sinense in the flood zone was verified by remote sensing techniques (Fig. 4).

Figure 4. Reflectivity of L. sinense in Shelby Farms along the Wolf River. L. sinense is accurately identified through Digital Globe's Worldview-2 platform (Modified from Quarles (2012), with permission). Notice the transition zone is identical to image 3a and 3b, and correlates with the historical flood zone boundary.

3.2 Methods

To investigate hydrochory (water dispersal) of L. sinense, seed/fruit traps were placed in periodically flooded Wolf River tributaries. Hydrochory traps were designed using a modification of those described by Middleton (1995). To make these traps, the bottoms of ten gallon buckets were removed to allow water to flow through them.

Adhered inside the buckets were mesh bags with one side open for water and fine sediment to flow through and all solid debris to be trapped inside the bag. To allow for floatation of buckets, empty plastic bottles with lids were secured to the outside of the buckets. The traps were tied to trees with the open bag facing upstream to catch materials drifting downstream in the water column (Fig 5). The traps captured all water- dispersed seeds and fruits, not only those of L sinense. The contents were removed for analysis by changing out each bag monthly (+/- 2-3 days) from September 2012 to March

2014. However, July and December 2013 were not included because traps were being repaired and improved.

I collected fruits from hydrochory sampling devices along tributaries within the

Wolf River, Tennessee, flood zones in the following three study sites, NON (non- invaded/channelized), INV (invaded/channelized), and CNT (non-channelized and non- invaded) (Fig. 2). I placed three sampling devices at each site, for a total of nine traps for each month. Near the first of each month I changed out seed bags, specific dates of collection intervals are displayed in Table 2. For purposes of ease and simplicity I will refer to each collection interval as the month where most of the collections occurred. The

32 river gage height data represented in this study are the same intervals as the seed collection dates.

Figure 5. Hydrochory trap in Wolf River tributary. These traps move up and down with the water column and trap all seeds that flow through the bucket.

34 Contents of the hydrochory traps were sieved and all seeds of woody species were removed and quantified. I determined the resulting species composition and calculated the relative abundance of all species dispersed by water per month within each location

(INV, NON, and CNT) (Appendix B). Relative abundance is the total amount of seeds/fruits of one species divided by the total amount of all species and is expressed as a percent (Smith and Smith 2001). For each site I also calculated species diversity per month (Appendix A). Diversity was calculated using Simpson‟s diversity index (Smith and Smith 2001). I used Simpson‟s Index of Diversity, over other diversity indices, because I was not able to identify all species and data were not collected in a known area

(Smith and Smith 2001). Simpson‟s index of diversity is calculated as follows:

Σ푛(푛 − 1) D = 1- s N(N − 1)

Where n=number of seeds for each species, N=number of seeds each month (per site).

3.3 Results

There was some variability in the sampling success for each month due to trap deficiencies and extreme flow events. The monthly variability helps explain some of the results. In September and October of 2012 buckets were located in locations of little flow, in response collection buckets were moved to areas with greater flow on 2

November 2012. In January 2013, I lost two buckets from INV and one bucket in CNT became detached (though it still had seeds inside) and another was lost in NON. In May

35 2013, one trap in both NON and INV washed up on the bank, likely because of high flow

(Fig. 6), and very few seeds were collected. Also in May 2013, CNT was inaccessible because of high water levels. Thus, May and June collections from CNT are included together. In Figures 6 and 7 the collections in CNT from May and June are displayed in

May 2013. In July 2013, two CNT traps became unattached from trees and were lost downstream, so they were collected for repairs and released in September with the other repaired traps from INV and NON.

Ligustrum sinense was dispersed via hydrochory in INV for 13 of the 17 collection months, was dispersed in NON for seven months, and was dispersed in CNT for four months (Table 2). Ligustrum sinense was dispersed into all three sites in January and February 2013, and January 2014. Overall, INV in September of 2012 had the greatest relative abundance of L. sinense seeds (90.6%) but not the greatest total number of seeds (n=29) (Table 2). INV in December 2012 had the greatest number of total seeds

(n = 545), and had the second greatest relative abundance (76.5%). INV in November

2012 had the second greatest total number of L. sinense seeds dispersed (n=487), however it had the 6th greatest relative abundance (46.6) after February 2014 (75%, n=63), and October 2012 (66.6%, n=237). Ligustrum sinense made up the greatest proportion of all hydrochory seeds/fruits in the late fall and early winter months (Table

2).

The greatest numbers of total seeds were dispersed in INV (Figure 7). The top four months with the greatest total seeds for all sites were January 2014 (n=2,486),

March 2013 (n=2,261), March 2014 (n=2,058), and April 2013 (n=1,597). The top four months with the greatest maximum gage height were May 2013 (3.80m), January 2013

36 (3.71m), January 2014 (3.25m), and March 2014 (2.98m). Overall, peak flow appears to coincide with peak dispersal months (Fig. 6). January and March 2014 were top dispersal months and top peak flow months. April 2013 had the fourth greatest number of seeds dispersed and had the fifth greatest flow (2.40m). Months with the lowest maximum flows (June and September-November 2013) had no L. sinense dispersal and very little dispersal of other species (Fig. 6, Table 2).

There are also months that had high flow with little dispersal and vice-versa.

Many of these discrepancies between peak flow and peak dispersal can be explained by variability in seed trap reactions to flow (Chapter 3.3, first paragraph), or time of seed production and subsequent shed by trees. February and March 2013 were top dispersal months; however, the peak flow during the collection period was relatively low. May

2013 was the month with the greatest maximum flow; however, there were very few seeds dispersed in May, and there were no L. sinense seeds dispersed. This can possibly be explained because a few buckets washed up on the bank (Chapter 3.2); however, it is also likely that many of the fruits dropped from the trees prior to May 2013. There were many more seeds dispersed in November 2012, despite that peak flows from Sept.-Nov.

2012 were nearly the same. This is probably because seed traps were moved to tributaries with greater flow on November 2nd 2012 and more flow brings more opportunities to collect seeds.

Overall CNT had the greatest mean diversity of seeds at 0.711, INV was second at

0.611, and NON had the lowest diversity at 0.568 (Table 2, Appendix A). March 2013 had the lowest mean diversity (0.139), despite having the second greatest number of seeds dispersed overall. This low diversity is attributed to the huge numbers of P.

37 occidentalis seeds in relation to other species (Fig. 7). Contrarily, January 2013 had the greatest mean diversity (0.854) and had the seventh greatest number of total seeds and the second greatest maximum gage height. There was very low diversity in INV in

September 2012 and February 2014, at the same time NON and CNT had very high diversity values. This can be explained by L. sinense dispersal, which had high relative abundance in INV during these months (Table 2).

38 Table 2. Monthly diversity and relative abundance of L. sinense seeds dispersed via hydrochory from September 2012-March 2014. Relative abundance numbers are bold if L. sinense was present. Diversity values are bold for top four most diverse months for each site and are italicized for the four months with the lowest diversity for each site. Relative Abundance of L. sinense Diversity Month Collection interval dates INV NON CNT INV NON CNT September 2012 9/9-10/3 0.906 0.200 0.000 0.179 0.800 0.848 October 2012 10/3-11/2 0.666 0.000 0.000 0.523 0.325 0.814 November 2012 11/2-12/3 0.466 0.007 0.000 0.688 0.576 0.692 December 2012 12/3-1/6 0.765 0.000 0.007 0.404 0.582 0.808 January 2013 1/6-2/3 0.171 0.091 0.008 0.850 0.896 0.817 February 2013 2/3-3/9 0.283 0.011 0.001 0.834 0.496 0.729 March 2013 3/9-3/31 0.115 0.000 0.008 0.262 0.090 0.064 April 2013 3/31-5/6 0.018 0.000 0.000 0.521 0.114 0.786 May 2013 5/6-6/2 0.000 0.000 0.000 (May/June) 0.750 0.662 0.363 (May/June)

39 June 2013 6/2-7/3 0.000 0.000 - 0.836 0.209 -

July 2013 7/3-8/12 0.005 0.000 - 0.798 0.807 - August 2013 - - - - - September 2013 9/12-9/29 0.000 0.000 0.000 0.854 0.333 0.873 October 2013 9/29-11/2 0.000 0.000 0.000 0.691 0.500 0.733 November 2013 11/2-11/30 0.000 0.000 0.000 0.400 1.000 0.779 December 2013 - - - - - January 2014 12/31-2/5 0.163 0.007 0.001 0.739 0.825 0.774 February 2014 2/5-3/1 0.750 0.009 0.000 0.425 0.718 0.740 March 2014 3/1-4/2 0.077 0.002 0.000 0.637 0.720 0.848

Figure 6. Bar graphs of total seeds dispersed via hydrochory per month for all sites from September 2012 to March 2014, and maximum Wolf River gage height for all collection intervals

40 The five most abundant species dispersed in INV from September 2012 to March

2013 were: Platanus occidentalis L. (American sycamore), n=2,488, L. sinense (Chinese privet) n=1,863, Fraxinus pennsylvanica Marshall (green ash) n=1,136, Fraxinus americana L. (white ash) n=799, and Acer negundo L. (box elder), n=422 (Fig. 7). The top five species dispersed in NON were P. occidentalis, n=2,290, Liquidambar styraciflua L. (sweet gum), n=440, Carpinus caroliniana Water (musclewood), n=329,

Taxodium distichum (L.) Rich (bald cypress), n=298, and F. americana, n=203. The top five species dispersed at CNT were P. occidentalis, n=933, C. caroliniana, n=412, T. distichum, n=366, L. styraciflua, n=316, and Liriodendron tulipifera L. (tuliptree), n=202. Four of the top five species dispersed in CNT and NON were the same. In all three sites the top species dispersed by water was P. occidentalis, which was the only species that was in common among all sites. INV and NON had two species in common,

P. occidentalis, and F. americana.

Among the top five most abundant species dispersed by water per site, Platanus occidentalis had the greatest relative abundance from March-July 2013 and October 2013 in INV; Ligustrum sinense was the most abundant species from Sept. 2012 - February

2013, in October 2013, and in February 2014. F. pennsylvanica had the greatest relative abundance in November 2013, and January and March 2014. F. americana was not the top dispersed species in any months in INV, and A. negundo was the most abundant species in September 2013 (Fig. 7). Within NON, P. occidentalis had the greatest relative abundance from February-June 2013, in September 2013, and from January-

March 2014. Liquidambar styraciflua had the greatest relative abundance in November

2012 and December 2013. Carpinus caroliniana was the top species in September and

41 October of 2012. Taxodium distichum did not have the greatest relative abundance for any month, though it was the second greatest in June 2013 and March 2014. Fraxinus americana had the greatest abundance in July 2013. Within CNT P. occidentalis was the top species dispersed by water from January-May/June 2013 and in January and March

2014. Carpinus caroliniana was most abundant in October 2012 and September and

November 2013. Taxodium distichum was the most abundant species in September 2012 and February 2014. Liquidambar styraciflua was the top species in November 2012 and

October 2013 in CNT, and L. tulipifera was not a top dispersed species for any months.

Figure 7. Bar graphs of relative abundance of the top five species dispersed via hydrochory for each site per month from September 2012 to March 2014

43 3.3 Discussion and Conclusions

This is the first experiment to supply empirical evidence supporting water as a dispersal mechanism for L. sinense. These results also suggest that L. sinense is well adapted to BLHFs because the long-distance dispersal mechanism of hydrochory likely allowed L. sinense to colonize the land adjacent to the channelized river and supports the passenger framework of species invasion (MacDougall and Turkington 2005). Most upland species are dispersed by wind and mammals; therefore, L. sinense had an advantage establishing in the “novel niches” created by channelization along the Wolf

River, Tennessee (Shankman 1996). Major flooding events still occurred after channelization, but were more intense and short lived than historical flood pulses (Weins and Roberts 2003). The multitudes of dispersal mechanisms utilized by L. sinense possibly explain some of its invasion success. Furthermore, hydrochory as a dispersal mechanism for L. sinense helps to explain the pattern of riparian establishment across the eastern United States.

For many months flow was visually correlated with dispersal, for example

January 2014 had high flow and high dispersal rates. These similarities indicate that high flow leads to high dispersal events. However, flow and dispersal were not visually correlated every month. As noted in the results, some months with high flow had very little dispersal (May 2013) and some months with low flow had very abundant dispersal

(March 2013). The differential flow and dispersal is likely a result of fruit production and abscission from the plant body. Most fruits ripen in the fall and fall from the trees from autumn to early spring. The months of high flow and low dispersal are potentially a

44 result of seasonality in fruit readiness. Overall, dispersal levels are likely a result of a combination of these factors, high flow disperses more seeds from late fall to early spring, and after the majority of seeds are dispersed in the spring then dispersal will be low regardless of flow. There were more seeds in INV overall. This is probably because there was greater discharge in the INV tributaries due to channelization in the tributaries there (Figure 5 shows a tributary from INV). Channelization streamlines the moving water and promotes faster flow, thus more discharge (Weins and Roberts 2003).

Although there was little L. sinense dispersal in the two non-invaded sites it was still present. The seed traps were only capturing seeds from a small volume of water and it is likely that many more L. sinense seeds made their way past the traps without capture.

There are many possible reasons why L. sinense disperses to these areas but is not well established. One could be a result of the long term inundation (Chapter 4). Another possibility is that L. sinense invasion is on the cusp and numbers of established species will increase over the next few years. Results from the inundation study in Chapter 4 indicate the former.

There were many endemic species to BLHFs that were dispersed into INV, but made up much greater proportions of seeds in CNT and NON. The assemblage of top species dispersed in INV consists of two species common to both BLHFs and upland hardwood forest (L. sinense and A. negundo). Taxodium distichum, which only grows in

BLHFs naturally, was among the top four species dispersed in CNT and NON. However, in NON and CNT, C. caroliniana, a species common to both upland and BLHFs was highly abundant. It is interesting that the common upland species L. tulipifera was highly abundant in CNT because it shows that water is bringing a great variety of species into

45 BLHFs. Thus, supplying evidence that some upland species also utilize hydrochory in addition to bottomland species. Liriodendron tulipifera were also sparsely established in

INV, NON, and CNT, and were present in the quadrat surveys at INV and CNT.

There are limitations to this study. One is that the construction of the buckets was not perfected until the final three months of collection, and as a consequence many would float away, get lifted onto banks, or break from the trees as a result of the force of stream flow. Another limitation is that collections occurred on one river and thus, overall dispersal patterns apply to only one geographic region. This study should be repeated in multiple rivers and tributaries throughout L. sinense-invaded regions around the United

States to create a more robust spatial analysis. Multiple sample sites would show whether the observed dispersal patterns along the Wolf River, TN also occur in other locations. A third disadvantage to this study is that no other dispersal mechanisms were investigated to make comparisons about what type of dispersal is most common within the different BLHFs along the river. It would be valuable to know the primary dispersal mechanism of L. sinense so one could focus control strategies toward those mechanisms.

For example, if birds are the primary dispersers in BLHFs then new invasions may be more likely to appear under tree canopies and other bird perches. A final limitation is that discharge, or the amount of water moving through the river per time, differed for each of the tributaries, and was unknown. If discharge was measured for all tributaries with seed traps then detailed calculations could be conducted to determine the abundance of L. sinense fruits within the water. From those data one could estimate the number of seeds dispersed into riparian zones with each flooding event. Models could be built to

46 predict where L. sinense might disperse in the wake of climate change and how they might be influenced by propagule pressure.

Further evidence of hydrochory of L. sinense could come from investigating the genetic relatedness of L. sinense within the invaded forests. If they are all clones then hydrochory does not play a significant role in the mechanism of L. sinense invasion.

However, if they are all genetically unique that would support the theory that L. sinense mostly establishes by seed and thus, dispersal is a major aspect of L. sinense invasion.

Further evidence would come from a uncovering levels of genetic relatedness along a continuous channel. If the genotypic pattern showed a gradual change in relatedness as one moves downstream, then water dispersal would be further supported as the primary dispersal mechanism. If L. sinense populations with sporadic or clustered distributions of genotypes then birds may play a much greater role in dispersal in BLHFs.

Although I did not investigate other mechanisms of dispersal, such as animals and wind, this study demonstrates that hydrochory is a major method of L. sinense dispersal.

Future investigations of invasive riparian species should consider hydrochory when formulating hypotheses of invasion, or when formulating eradication or mitigation strategies. One mitigation possibility is that prevention should involve seeking out and removing L. sinense seedlings from riparian areas a few months after major flooding events, which may be the source of new propagules. This would be superior to eradication of well established populations because seedlings are easier to pull from the ground by hand, they would be easier to fell, and easier to treat with herbicides.

Furthermore, seedlings are non-reproductive and removal at that stage would dramatically reduce spread and establishment of new seedlings.

47 CHAPTER IV

EFFECT OF LONG-TERM WINTER INUNDATION ON L. sinense GERMINATION

AND VIABILITY

4.1 Background

Flood pulses, maintained by naturally flowing and unaltered rivers, create unique riparian habitats that are crucial for the establishment and survivorship of BLHF species.

Flood pulses usually occur in the winter months and are characterized by overbank flow where the land becomes completely inundated throughout the winter (Fig. 8). Inundation can affect viability of seeds of bottomland and upland species (Middleton 2002; Greet et al. 2011). The correlation between inundation and germination is generally positive for bottomland species and negative for upland species (Guo et al. 1998; Middleton 2002;

Richardson 2007). Thus, when natural flow patterns cease, present communities suffer as they are not adapted to the new stream flow and flood regimes and are unable to adequately regenerate (Middleton 2002; Capon 2003).

A paired t-test revealed that the channelization of the Wolf River significantly lowered the water table by an average of 2.02m (P<0.001), and resulted in frequent, shorter term floods, often of higher intensity and flow (Fig. 8) (Weins and Roberts 2002).

This channelization caused the natural winter inundation to cease, and also shifted

48 the peak flow from December-May to March-June (Fig. 9). This shift in flow may have caused damage to species present, and may have opened gaps for new invasive establishers, like L. sinense (Hood and Naiman 2000; Richardson 2007). Interestingly, the few places that still have prolonged, low intensity inundation episodes remain un- invaded along the channelized portion of the Wolf River, Tennessee, as is the case with the non-invaded (NON) study site along the channelized portion of the Wolf River

(Foard, personal observation). The objective of this study was to quantify the effects of hydrological alteration on seed germination/viability of L. sinense. Moreover, this research helps explain the community dynamics and hydrological conditions that led to the dominance of L. sinense. I hypothesized that long-term winter inundation significantly reduces L. sinense viability and germination, and thus, channelization supplied dry enough conditions to allow establishment when dispersed into these regions.

49 50

Figure 8. Line graph of mean gage height per year before (1936-1963) and after channelization (1964-2010) of the Wolf River, TN. The red arrow represents the year 1963 when the channelization process ceased. Notice that the mean height before channelization was much greater before channelization. (USACE 2013).

Figure 9. Line graphs of mean gage height per month before channelization (1936 to 1963) and after channelization (1964-2012) of the Wolf River, TN. I conducted a one-way continuous factor ANOVA to determine if monthly gage height differs before and after the channelization (before n=27years, after n=49 years). Each population met the normality assumptions for parametric testing. If there were four or more months of missing flow data I excluded that year from the calculations. (USACE 2012)

4.2 Methods

I conducted two studies to identify the effect of long-term inundation on L. sinense viability/germination: a preliminary and a field study. The preliminary inundation study consisted of two treatments: WET, which represented historical or unaltered hydrological regimes, and DRY, to represent a location with no inundation and to serve as a control. Ten L. sinense seeds were placed into each of 60 cups and cups were evenly divided into the 2 treatments (30 per treatment). The cups were placed in a garage where conditions inside the garage were representative of the natural weather conditions outside, and experienced full shade and temperatures below freezing. The treatment lasted for approximately 10 weeks from 27 January to 5 April 2013. All seeds were tested for viability using a tetrazolium stain: a technique where seeds are soaked in a solution and viability is indicated by the presence of a red color (Howarth et al. 1993)

(Fig. 10). These tests were performed in a laboratory at ASTATE. Tetrazolium is a white/off white salt that reacts with digestive enzymes in the embryo of the seed. If the digestive enzymes are present and the seed is alive, the seed will turn red from the chemical reaction with Tetrazolium. If the digestive enzymes are not present then the seed will not stain and is non-viable (Howarth et al. 1993).

Before each seed was stained it was soaked overnight in water to trigger imbibition, a process by which seeds take up water. The act of imbibing causes viable embryos to begin digesting the endosperm, which activates the enzymes that react with the stain. After imbibition the seeds were removed from fruit and cut in half to expose the embryo. Then they were soaked in a 0.5% solution of 2,3,5 Triphenyl Tetrazolium

52 (Howarth et al.1993). Ligustrum sinense seeds were soaked for 12 hours. After seeds were rinsed the color of each seed was assessed, and total viable seeds in each treatment was quantified using the same visual analyses (Fig. 10). Unfortunately, while assessing seed colors a few of the petri dishes fell to the ground and mixed together; thus, they were not used in the analysis and sample sizes were not equal.

Populations did not meet the assumptions for parametric testing and sample size per treatment was less than 30; thus, a non-parametric Mann-Whitney test was conducted to determine if there was a significant difference between treatments.

Figure 10. Ligustrum sinense seeds stained with tetrazolium. The red seed on the left is considered viable and the seed on the right is not.

54 In addition to the preliminary study, a field inundation study was conducted. The field study consisted of 30 replicates of mesh seed bags that were exposed to the following three treatments: (1) A hydric treatment (WET) replicates historic flood pulses, in which inundation often lasted from December to late May or early June (USACE

2013); (2) a hydrologically altered, mesic, treatment (MID) replicates recent flooding conditions that result from river channelization, in which peak flow occurs in May and lasts for very few days (Fig.s 7 and 8); and (3) a xeric control treatment (DRY), which is typical of non-riparian areas, and was given no flooding regime. Ninety mesh bags of

100 L. sinense seeds each were sewn. Bags were completely enclosed to prevent seeds from escaping (Fig. 11). Bags were tied to trees located in three field locations, and each location represented all three treatments described above (WET, MID, and DRY) (Fig.

12).

Treatments within sites were selected based on observations made upon visiting in January. WET locations consisted of mostly obligatory and facultative wetland species (Taxodium distichum (L.) Rich; Ilex decidua Walter; and Quercus lyrata Walter), trees had large buttresses, and the ground was saturated or completely inundated. MID locations were selected in nearby bottomland hardwood forests where hydrology had been altered (channelizing in the Wolf River and damming in the site at Bayou De View), the species present consisted of greater quantities of facultative species than in the WET locations (Chapter 5.3.2) (Liquidambar styraciflua L.; Aesculus pavia L.; and L. sinense), trees had slight buttressing, and the soil was moist approximately three cm below the surface. The dry sites were located in the highest elevation within the site where there was still forest canopy, and there was a greater mixture of facultative and facultative-

55 upland species (Pinus taeda L. and/or Quercus alba L.). The sites were located along the

Wolf River, Tennessee, Bayou De View, Arkansas, and the St. Francis River, Arkansas.

At each site there were ten bags for each treatment (WET, MID, and DRY), for a total of

30 bags at each site.

Bags were left in the field until natural inundation subsided. Locations were checked 6 and 20 May, 2 and 13 June, 3 and 15 July, and 12 August 2013. Water levels at the Wolf, and Bayou De View Rivers subsided by 13 June 2013. Water levels along the St. Francis River did not drop until 12 August, at which point many of the MID and

DRY seeds germinated; however, the roots of the germinated seeds had become entangled and shriveled and were indistinguishable from one another, so the location was removed from analyses. Bags from the Wolf and Bayou De View Rivers were collected

13 June and were planted in the Arkansas Biosciences (ABI) greenhouse at ASTATE on

14 June. Many of the MID treatment seeds were germinated in the bags in the field by

June 14th (Fig. 15). All germinants and remaining seeds were planted in the ABI greenhouse with a photoperiod of 16 hours of light and 8 hours of darkness. The germinated seeds were set aside while the other seeds were planted, then they were carefully planted and grown with the germinated seeds. The total number of germinated seeds was counted on 12 September 2013. There were approximately 10 seedlings that died in the flats, but were easy to recognize and were counted as germinated in this study.

To analyze the effect of inundation on the germination of L. sinense seeds from the field inundation study I conducted a two-way ANOVA with interaction. The response was germination rate of L. sinense seeds, and the factors were the three treatments (WET, MID, DRY) and the two sites (Wolf and Bayou De View Rivers), and

56 the interaction term tested for an interaction between treatment and site (Fig.14). Not all populations met the assumptions for parametric testing; however, ANOVA is robust to violations of the assumptions (Tomarken and Serlin 1986).

Figure 11. Bag of 100 L. sinense seeds being sewn

Figure 12. Bag of 100 L. sinense seeds in an inundated forest (WET)

59 4.3 Results

For the preliminary study, a non-parametric Mann-Whitney revealed that there was a significant difference between viability of seeds in WET and DRY treatments (Fig.

13). The seeds in the DRY treatment were 70% viable on average and 33% of the seeds from the WET treatment were viable, on average. These percentages are much greater than the results from the field inundation study.

The two-way ANOVA revealed that there was a significant treatment effect

(P<0.001), and there was no significant site effect (P=0.56). There was an interaction effect indicating that seeds germinated differently by treatment at each site (p=0.009).

However, all seeds showed the same qualitative pattern (Fig. 14). At Bayou De View

River the mean germination rates for each treatment were significantly different. For

WET, MID, and DRY the mean germination rates were 4.33%, 33.70%, and 20.17% respectively. At the Wolf River MID treatments had significantly greater germination rates than WET and DRY; however, WET and DRY were not significantly different from one another at alpha=0.05. Germination rate mean values were lowest for the WET treatment with a mean of 1.22%, and in the MID location and DRY locations, mean germination was 42.00% and 12.00%, respectively. Both sites showed that the greatest germination was in the MID treatment, followed by DRY, and lowest germination rates resulted from WET treated seeds.

Tukey confidence intervals, at 95% confidence, were created to determine the difference in germination rate for all pairwise comparisons of treatments within sites. For

Bayou De View River confidence intervals revealed that germination in WET was

60 significantly less than MID treatments by between 41.2% and 17.5%, and was less than

DRY by between 2.6% and 29.1%, on average. MID was significantly greater than DRY by between 1.7 and 25.4%, on average. For the Wolf River confidence intervals revealed that germination in MID was significantly more than WET treatments by between 30.0% and 51.6%, and was more than DRY by between 19.2% and 40.8%, on average. WET was not significantly less than DRY, even though mean germination was greater for DRY seeds than for WET seeds.

Figure 13. Box plot of the viability of L. sinense in DRY and WET treatments for preliminary study.

Figure 14. Interaction plot between site and treatments for percent germinated in the field inundation study. Both sites showed the same qualitative pattern with WET having the lowest germination rate, followed by DRY and MID had the greatest rate. The Wolf River showed a greater increase in germination for the MID treatments than the Bayou De View. Within the treatments there was no significant difference between sites. BD = Bayou De View River, WR = Wolf River.

63 64

Figure 15. Pictures of seed bags from the field inundation study after winter treatments. Many MID treated fruits germinated in the bags while in the field, and can be seen in the middle picture above. WET seed bags were often coated in layers of muddy sediment, while DRY seed bags were dry, with few germinating seeds

4.4 Discussion and Conclusions

These inundation studies supply evidence that L. sinense is a passenger of invasion in hydrologically altered BLHFs (MacDougall and Turkington 2005).

Ligustrum sinense passively takes advantage drier habitats where it germinates at significantly greater rates than the hydric habitats. Thus, in places where natural flood pulses still occur, the likelihood of L. sinense populations establishing is significantly lower than places where flood pulses are ceased or reduced. These results support the

“novel niche” mechanism of invasion (MacDougall et al. 2009). A novel niche was created in riparian forests adjacent to the channelized portion of the river. The new niche was unlike surrounding niches that existed historically, and included well-drained soil with infrequent, short-term, high intensity floods, to which L. sinense is adapted. In its native range of Southeast China L. sinense is common in riparian piedmont areas (Wang and Grant 2012).

There are some limitations to these studies. First, the preliminary study had only one replicate and the field study had only two replicates. More replicates would show whether there was spatial variability, or whether results are consistent across different geographical regions. Second, the field study occurred over the course of only one year and flood pulses are variable from year to year, especially in the wake of climate change

(Junk et al.1989; Seager et al. 2012). Thus, temporal variability is unaccounted for in this study. Furthermore, the winter of 2012-2013 was exceptionally wet in the lower

Mississippi alluvial valley, and pules lasted longer than normal (USACE 2013), possibly exaggerating the effects of flooding on privet germination success. This could explain

65 why there were much higher germination rates in the preliminary study, which took place over a much shorter amount of time (10 weeks vs. 23 weeks). Another possible reason for the differing results between the preliminary and field studies could be a result of the differing methods used to test survivability of seeds. For the preliminary study the seeds were tested for viability using tetrazolium stain. The field study tested germination of the seeds with greenhouse germination. It is possible that viable seeds did not germinate under the greenhouse conditions used in this study or that tetrazolium overestimates actual germination ability, such that not all viable seeds actually germinate in nature. A final limitation of the field study is that conditions in sites were not monitored daily to ensure that they were always matching the expected conditions. Given these limitations, additional research should be conducted. Still, two independent experiments provided qualitatively similar results: long-term inundation reduces viability or germination success of L. sinense. These two studies suggest that a change in flooding away from natural conditions has favored the establishment of L. sinense.

The field study should be replicated in time and space. There should be more than two rivers, and more than one year. I suggest five sites over the course of five years would be sufficient to uncover definite effects of inundation on germination, and to control for the high variability in flood pules from river to river and from year to year.

Moreover, one could also observe the effects of flood variability on the germination of L. sinense and those data would supply insight to the true effects of inundation on L. sinense invasion. To investigate the effects of climate change, one could perform the inundation study in microcosms with varying temperatures.

66 Despite the temporal and spatial limitations, the results still support that L. sinense germination and viability are reduced as a result of long-term inundation. Furthermore, if

L. sinense is shown to have reduced germination rates in all BLHFs with natural flood pulses then some suggested management implications would result. In BLHFs where L. sinense is already established, restoration efforts should begin with removal of L. sinense, and then should aim at returning the floodplain to natural pulsing conditions. Or after removal of L. sinense, efforts could focus on planting upland species that may be dispersal limited but are adapted to the drier conditions. Moreover, any landowners may prevent L. sinense invasion into forested wetlands by maintaining natural flood pulses.

These implications would be strengthened if these results were shown to be robust under

BLHF conditions throughout the southeastern U.S.

67 CHAPTER V

EFFECT OF STREAM FLOW AND L. sinense INVASION ON BOTTOMLAND OAK

TREE GROWTH

5.1 Introduction

The results from the previous studies (Chapters 3 and 4) demonstrated that in early life stages (seed) L. sinense is a passenger in the hydrologically altered riparian forests along the lower Wolf River, Tennessee. Channelization of the river created

“novel niches,” with drier conditions where upland species could establish. Utilizing hydrochory as a dispersal mechanism (Chapter 3), in addition to birds and mammals, I contend that L. sinense colonized the forest before most other upland species could arrive, as most upland species are primarily dispersed by animals and wind only

(Shankman 1996). Likewise, my results support the hypothesis that L. sinense took advantage of the drier conditions that allowed for germination rates higher than in historic conditions (Chapter 4), in turn further supporting the “novel niche” and passenger hypotheses. These studies address the early stages of invasion; however, they do not investigate any of the possible consequences of long-term L. sinense invasion, or explain how L. sinense forms monocultures. Possible consequences of established L. sinense might involve competitive exclusion of other species or natural replacement

68 of species in forest gaps. If competitive exclusion were true then it would supply further support for the driver hypothesis (MacDougall and Turkington 2005). If displacement of species in forest gaps resulted then L. sinense would fit as a backseat driver – initially invading as a passenger, then competitively reducing native biodiversity as a driver

(Bauer 2012).

In the highly invaded BLHFs of the Wolf River, Tennessee, large, standing trees appear to be dying at accelerated rates when compared to trees in non-invaded BLHFs, as is evident by the increased abundance of dead wood littering the forest floor (Chapter

5.6.2). If L. sinense were shown to reduce growth rates of BLHFs then it would be a driver in later stages of invasion – ultimately fitting it into the backseat driver framework

(Bauer 2012).

Dendrochronology, or tree ring dating, was used as a tool to investigate the mechanism of L. sinense invasion. I investigated whether altered stream flow and/or L. sinense invasion is limiting the rate of growth of adult Quercus (oak) trees along the channelized and invaded portions of the Wolf River, Tennessee. Isolation of the variable or variables that most influence tree growth of invaded BLHFs could help explain the cause of tree mortality in these forests.

Dendrochronology is the study of annual tree rings to make inferences about the biotic or abiotic conditions present throughout the life of trees (Stokes and Smiley 1968;

Speer 2008). The annual rings produced by trees are indicative of the favorability of conditions present throughout the seasons. In the spring and early summer, when water is usually abundant and conditions are favorable, trees grow fastest. Wood is made of xylem, which are the water conducting cells known as vessels in hardwood

69 (angiosperms), and tracheids in softwood (gymnosperms) species (Stokes and Smiley

1968; Speer 2008). The spring cells are larger than other xylem cells and are referred to as earlywood. In the late summer and autumn they are much smaller because they are conducting less water; these cells are termed latewood. The combination of earlywood and latewood makes up annual rings, and the size of each ring differs from year to year depending upon the environmental conditions (Stokes and Smiley 1968; Speer 2008). If conditions are highly favorable the tree will grow fast and will be reflected by wider than average annual rings. If conditions are less favorable, then the rings will be thinner and growth slower.

The amount of annual growth of trees depends on various intrinsic and environmental factors. In a simple linear model there are five major predictor variables:

(1) age, as trees age and grow, they show a natural declining growth trend; (2) climate, including precipitation (ppt.), temperature (temp.), and Palmer Drought Severity Index

(PDSI); (3) microclimate disturbances, such as canopy gaps or woody species competition; (4) stand level disturbances, such as proximate logging, construction, or river channelization (5) random variability in annual growth not explained by the other variables (Cook and Kairiukstis 1990).

This study aims to answer the questions: Do oak trees in L. sinense-invaded sites have slower growth rates than trees in reference, or control, sites? If so, what roles do climate, river channelization, and L. sinense invasion play in tree growth and the observed tree mortality? I hypothesized that the woody wetland species that once thrived began to struggle in the less suitable, drier, conditions following channelization, and that

L. sinense invasion caused trees to grow slower, and possibly exacerbated tree mortality

70 in these stands. Using dendrochronology as a tool, I discovered the effects of climate, stream flow, and L. sinense invasion on the growth rates of endemic oak species that dominate BLHF canopies.

5.2 Site and Tree Selection

A before-after/control-impact (BACI) experiment was used to investigate the impact of L. sinense (Smith 2002) on oak trees in three sites along the Wolf River,

Tennessee. One impacted site (INV) was located in an invaded BLHF adjacent to the channelized portion of the river. Two control sites were located in non-invaded locations: (NON) adjacent to the channelized river, and (CNT) adjacent to a portion of the river where there was no channelization (Fig. 16). Originally the CNT site was chosen as a control to look for effects of L. sinense and channelization. However regressions revealed that CNT responded just as much to flow as the other sites, and could not be used as a control for channelization. Thus, CNT was used in conjunction with NON to reveal whether there were BACI impacts of L. sinense on tree growth after invasion (sections 5.53 and 5.54).

Oak trees were cored because they produce consistent annual rings (Stahle 2012 pers. comm.) and are present in all three sites. I cored and analyzed 12-18 Quercus nigra

L. (water oak), 6-8 Q. pagoda Raf. (cherrybark oak), 5 Q. michauxii Nutt. (cow oak), and

2 Q. lyrata Walter (overcup oak) in each of the three locations along the Wolf River,

Tennessee. All trees were greater than 25cm diameter at breast height (DBH) and within

300 meters of the old Wolf River channel to minimize variability in the response to

71 channelization while maximizing flow responses, and to ensure that stand level conditions were similar and consistent (Speer 2001; Copenheaver et al. 2007; Speer

2008). Cores were collected from 25-30 trees per site (INV, NON, CNT) and a total of

83 trees were used in the analyses (Table 3). The chosen oak species are endemic to

BLHFs and likely represent typical BLHF responses (Burke and Chambers 2003).

One GPS coordinate per location (INV, NON, and CNT) was selected using a true randomized number generator from the website http://www.random.org. Coordinates were located and transects were established parallel to the Wolf River. Every living tree along the transect, that was greater than 40cm in diameter, within 300 meters of the old river, and that matched the selected oak species, was sampled until I cored the target number of trees. I selected trees in INV first because there was a scarcity of healthy- looking trees, and I wanted to make sure trees in all sites were physically comparable.

Therefore, when coring in NON and CNT, species and DBH were matched as closely as possible. Following methods from Speer (2008) I extracted two perpendicular cores from each tree (north and east) and stored them in labeled straws. I then dried the cores at room temperature with books on top to prevent bowing. Cores were prepared for analysis using methods suggested by Speer (2010), Stokes and Smiley (1968), and personal consultations with Dr. Dave Stahle of The University of Arkansas, Fayetteville

(Stahle 2012). First the cores were glued into wooden core mounts using water-soluble glue, with vessels facing appropriately. Then they were sanded until the annual rings were smooth and easily visible. I used five series of grits of sandpaper: 50, 120, 220,

320, and 400 followed by a series of three sanding films: 30, 15, and 9 µm and finished with a suede polishing cloth and/or steel scrubber. The core samples were used to

72 determine the rate of growth of each individual (in microns per year). For each cored tree, DBH and tree height were recorded, tree age was calculated, common tree health variables were assessed, and surrounding species and site compositions were surveyed.

Additionally, coarse woody debris (CWD) was quantified at INV and NON.

73 74

Figure 16. Map of dendrochronology sites and trees located along the Wolf River, Tennessee. INV=invaded and channelized or “impact” site. NON=non-invaded and channelized site, and CNT=non-invaded/non-channelized control site, Map made with assistance from David Burge.

5.3 Methods for Assessing Tree, Site, and Stand Characteristics

5.3.1 Tree Characteristics

To infer about physical differences in the size of the trees among sites, I measured diameter at breast height (DBH) and estimated the height of each tree. To investigate tree health characteristics I estimated percent dieback, live crown ratio, and percent canopy missing for each tree. Percent dieback was the percent of dead limbs visible on the tree when compared to the total number of limbs. Live crown ratio was the percent of the trunk with living branches, and was estimated as the number of meters of canopy height divided by the total height of the tree. To predict percent canopy missing I imagined a full circle encompassing the farthest reaching branches of the canopy, then I estimated the percent of that circle that was empty space. Analysis of variance (ANOVA) was conducted in Minitab (version 16, Minitab Inc., State College, PA) comparing the measured tree size and health variables to determine if there were significant differences

among sites. The ANOVA model is Yijk = μ+ αij + βij + (αβ)ij + εijk. The normality assumption was not met for live crown ratio, and the equality of variance assumption was not met for percent dieback, however I still used parametric testing because ANOVA is robust to moderate violations of the assumptions (Tomarken and Serlin 1986), and parametric testing would not allow for testing of interaction.

To calculate tree age for trees that did not contain the first few years of growth, an area known as the pith, I used the dendrochronology program BiaPlt (version 9, Lamont-

Doherty Earth Observatory, Palisades, NY, http://www.ldeo.columbia.edu/tree-ring-

75 laboratory/resources/software). I compared the tree ages among sites with a one-way

ANOVA in Minitab, because age samples met assumptions for parametric testing.

5.3.2 Site and Stand Characteristics

Quadrat and vegetation sampling techniques from Krebs (1999) and U.S.

Environmental Protection Agency (2011) were incorporated to gather species and site composition data. To determine the direction of a quadrat I randomly generated a number 1-360 (90°=East, 180°=South, 270°=West, 360°=North, etc.). I walked 10 meters in that direction away from the cored tree, then randomly generated a direction

(left (1) or right (2)) and turned 90° in that direction, walked another 10 meters and completed a square quadrat with an area of 100 m2. Within each quadrat I identified all woody species and estimated percent cover, average diameter, and mean height of all stems in the quadrat. I used the information from the quadrat measures to calculate similarity, wetland prevalence, and diversity indices (Calculation described in Chapter

3.2) for woody species present in each site. These indices were used to determine the level of comparability of sites.

Sorenson‟s index of similarity (Ss) was calculated to identify resemblance of species composition for all pairwise comparisons of sites (Wolda 1981). Sorenson‟s index of similarity was calculated as:

2퐶 푆 = 퐴 + 퐵

76 Where C= number of species two sites have in common, A = total number of species of site A, and B = total number of species for site B.

Wetland prevalence indices were calculated for woody species presence within each plot to determine the mean wetland prevalence index of each stand. Each species was assigned a wetland indicator status by USDA plant database (USDA 2014).

Obligatory wetland (OBL) species are found in wetlands >99% of the time. Facultative- wetland (FACW) species are present in wetlands 67-99%, facultative (FAC) species are present in wetlands approximately 50% of the time, facultative-upland (FACU) species are present in wetlands 1-33%, and upland species (UPL) are present in wetlands <1% of the time. The wetland indicator classifications are assigned the following values:

OBL=1, FACW=2, FAC=3, FACU=4, and UPL=5, and the values are used to calculate mean wetland prevalence indices for each site (Weins & Roberts 2003). Robust parametric ANOVA was conducted to determine whether there were significant differences among sites (Tomarken and Serlin 1986). Wetland Prevalence Index was calculated as:

퐼 푊퐼 = 푂푛

Where I = Value of the wetland indicator status for each occurrence of the species in the site, and On = Total number of occurrences of each prevalence index in the site.

Measures of diversity included species richness and diversity using the calculation for Simpson‟s diversity index (Ds) described in chapter 3.2 (Smith and Smith 2001;

77 Burton 2005). I verified the eligibility of Simpson‟s Diversity, as opposed to other diversity indices, by calculating the minimum number of samples required to get a full representation of the data. In INV over 98% of the species were represented after 20 samples. In NON 96% of the species were represented after 20 samples, and in CNT

95% of all species were represented after 20 samples. Therefore, my sample size of 30 is adequate for the use of Simpson‟s Diversity Index (Smith and Smith 2002). Simpson‟s index of diversity is calculated as follows:

Σ푛(푛 − 1) D = 1- s N(N − 1)

Where n =number of stems for each species, N = number of stems for each plot.

Richness is a measure of the total number of different species, and Simpson‟s diversity is a measure of the richness and evenness of individual species within each site.

All parametric assumptions were met; thus, a one-way ANOVA was used to test for significant differences in diversity among sites.

Coarse woody debris (CWD) assessments were conducted in both channelized sites (INV and NON) to determine if the amount of fallen trees and broken limbs is greater when L. sinense was present vs. absent (Bragg and Heitzman 2009). All CWD was assessed along three randomly generated transects that were perpendicular to the channelized river, and each transect was 30m long and 10m wide (300m2). I measured and recorded wood type (Hardwood/Softwood, or unknown), debris type (Log, stump, or snag), large end diameter (cm), small end diameter cm), length (m), evidence of cutting

78 (yes or no), and decay class (1: least-3: most), modified from Vose et al. (1999). The average number of pieces of woody debris, volume, and mass were reported per 100m2.

CWD volume (V, m3) was calculated using Smallian‟s formula (Bragg and Heitzman

2009):

(π[D + d]L) V= 8

Where D is equal to the diameter (m) of the large end, d is equal to the diameter (m) of the small end, and L is equal to the length (m) of the woody debris. Specific gravity is based on the Bragg (2013) classification of decay. Mass for the woody debris was calculated by multiplying the volume with the specific gravity.

5.4 Determining the Timing of L. sinense Invasion

To decide which dendrochronology methods were appropriate for this study I determined the timing of L. sinense arrival first; therefore, results of that investigation are included in the methods below. Ligustrum sinense does not produce annual rings because it is an evergreen shrub and has multiple growth spurts throughout the year. Therefore aging L. sinense requires alternative methods to dendrochronology. To determine when

L. sinense became well established I used three methods: (1) visual recognition dating through historical winter aerial photographs, (2) first-hand accounts, and (3) statistical inferences.

79 In the first method I used winter photographs because L. sinense is the only colonial, broad leaved, evergreen species in the BLHFs along the Wolf River, so it was easy to distinguish from the other species when leaves of deciduous trees were absent.

Secondly, two documented first-hand accounts came from long-term residents of the

Memphis region who were frequent visitors of the Wolf River, Tennessee, throughout their lives. Finally, statistical inferences were derived from site level chronologies via non-parametric Superposed Epoch Analyses in Dendro Tools (version 1.0 http://www.djburnette.com/programs/dendrotools/index.html). All methods indicated the same approximate time interval for L. sinense establishment.

It is clear from the old aerial photos that L. sinense was not present in 1954 and was fully established by 1981 (Figs. 3 and 4). Therefore the establishment date is narrowed down to the 27 years in between the images. According to first-hand account from Larry Smith, former Executive Director of the Wolf River Conservancy from 1997-

2003 who has frequently visited the Wolf River bottomlands since 1967, L. sinense was dense and widespread by the early1980s (pers. comm. 4 Dec. 2013). Another first-hand account comes from Mary Winslow Chapman in her 1977 memoir entitled I Remember

Raleigh. She wrote about the Wolf River near INV, stating in the past she “could always see where we were going…even from a distance…but now we were plunging over our heads in underbrush (Chapman 1977).” The only “underbrush” along the Wolf River that would grow so intensely to reduce visibility is L. sinense. I also conducted a non- parametric superposed epoch analysis using the dendrochronology data. This method removes excess noise in tree chronologies to look for any significant dates where an event occurred (Samson and Yeung 1986). The RW and BAI series from INV revealed

80 that in 1975 a significant event occurred (P<0.001, P=0.041, respectively). For NON and

CNT sites there were no significant events in the 1960s, 1970s, or 1980s. Therefore, the superposed epoch analysis suggests that trees in INV were first impacted by L. sinense in the mid 1970s. Therefore, the 1970s were removed from our dendrochronology analyses because trees may show differential response times, and establishment of woody species often has temporal variability. Times before privet invasion were established from 1940-

1969, and times after invasion were determined to be from 1980-2011. I did not conduct any statistical analyses before 1940 because the majority of the trees in CNT were not present until the 1940s (Fig. 20).

5.5 Dendrochronology Methodology

5.5.1 Crossdating and Quality Control

Cores were visually crossdated to gain temporal overviews of annual influences on tree growth in each location (Speer 2008). Next, rate of growth of each tree core was measured to an accuracy of 1μm using Bausch & Lomb StereoZoom 4 Dissecting Scope

(Bausch &Lomb Inc., Rochester, NY), VELMEX unislide measuring device (Velmex

Inc., Bloomfield, NY), and ACU-RITE linear encoder (ACU-RITE Inc., Jamestown, NY) at the University of Arkansas, Fayetteville. The measures were transferred to quality control software, COFECHA (Holmes 1983), which looks for any outliers and confirms manual dating. Three site-level master chronologies were created: INV, NON, and CNT.

COFECHA outputs series intercorrelation values known as “r-bar.” The values range from 0-1 and the values must be greater than 0.4 to be used for data analysis (Speer

81 2008). I visually checked each value and if it was less than 0.4 I re-measured the core, corrected the dating, or if these methods failed I removed the core from the chronology altogether (Speer 2008). Using this approach, I had 14 unreliable cores that had to be removed from the sample before production of chronologies. Another output from

COFECHA is the mean sensitivity value for each core. Mean sensitivity is a measure of the year-to-year variability in ring width and ranges from 0-1, with 1 being the most sensitive and 0 being the most complacent. If a tree has little variation in growth then it is not a good candidate for dendrochronology; likewise, it is not useful if it has too much variability. Preferred mean sensitivity values are around 0.2 for a chronology(Speer

2008).

5.5.2 Producing Chronologies for Statistical Analyses

After ensuring quality, the data were transferred to ARSTAN (Cook 1985), a dendrochronology analysis software program that creates chronologies that are compatible with multiple statistical software applications, including Microsoft Excel and

Minitab. I created three master chronologies for each site, INV, NON, CNT, and four species level master chronologies: Q. lyrata, Q. michauxii, Q. nigra, and Q. pagoda.

ARSTAN supplied four different types of chronologies: raw ring width (RW), standard, residual, and ARSTAN chronologies. The raw chronology (RW) is the true measure of each ring in 1μm. The standard, residual, and ARSTAN chronologies remove excess noise from the chronologies often to amplify climate signals and/or remove age related

82 variability in growth. Because the signature of L. sinense may potentially be removed by the standard, residual, and ARSTAN chronologies, they were not used in the analyses.

Tree growth in the Wolf River BLHFs was viewed in terms of a conceptual linear regression. Each factor in the model has an influence on the overall growth rate, or chronology, of the trees in each site. The basic formula for the model is:

푅푡 = 퐴푡 + 퐶푡 + 퐻푡 + 퐿푡+∈ 푡

Where Rt = Growth rate for chronology at time=t, At = Age-related growth factor, Ct =

Climatological influences (Ct) on growth rate (Ct = ppt., temp., and PDSI), Ht =

Hydrological influence, which is the channelization effect based on river gage height data, Lt = L. sinense-related growth factor, which is the effect of L. sinense, and ∈t =

Variability in annual growth left unexplained (Cook and Kairiukstis 1990).

To remove the age related growth factor (At), which is typically a negative exponential curve, I transformed RW to basal area increments (BAI) (Speer 2008). I used the statistical software program BiaPlt to transform the chronologies. BAI are measures of the total area accumulated by the trees per year. I used the RW and BAI chronologies for climate and flow correlation analyses. I wanted to specifically investigate the effect of channelization and L. sinense on oak tree growth rate; thus, I identified the signatures of the remaining climate predictor (Ct), and if they were strong enough they would be removed from the chronology to isolate effects of flow and L. sinense.

83 To investigate if there were any significant differences among the BAI of species overall, and of species within sites, a one-way ANOVA and a two-way ANOVA were conducted, respectively. I determined that, although some species grew at different rates, all species showed the same general growth trend over time for both RW and BAI chronologies, and I grouped them together by site for analysis (Fig. 17).

Figure 17. Master chronologies for each species in raw ring width (RW, above) and basal area increments (BAI, below) from 1940 to 2010. The chronologies show that all species grew with the same general response patterns over time. Quercus pagoda grew significantly faster than all other species, and Q. lyrata grew significantly slower than all other species, on average. Chronologies were made with KaleidaGraph (4.1.1 Synergy Software Inc., Reading, PA). Chronologies were graphed with curved fits set at 100 fit points.

85 5.5.3 Identifying the effects of Climate and Flow on Tree Growth

To determine if flow (Ht) had a significant influence on tree growth I first needed to remove significant climatic signatures from the chronologies (having already removed age by producing BAI chronologies). The climatic variables ppt., precip., and PDSI were obtained from National Climatic Data Center (NOAA 2012). Using Dendro Tools, correlation function analyses were conducted between monthly climate data versus each site chronology (INV, NON, CNT) to determine which variables and months were significantly correlated with RW and/or BAI.

When strong signals of monthly climate data were detected, monthly data were averaged together and correlation matrices were conducted between strong monthly averages versus each RW and BAI site chronologies. I was most interested in parametric outputs from the correlation matrices because the sample size was n ≥ 54 years, and linear relationships were assumed. If the Pearson‟s correlation coefficient was greater than |± 0.7| and p<0.001 for site level chronologies then the variable was thought to be a significant influence on tree growth and regression models would be built to identify what the influence is, and to determine the strength of the influence. Then the signature would be removed from the site level chronology using the adjusted chronology tool in

Dendro Tools. However, none of the climatic variables met the Pearson‟s correlation criterion listed above, so no climate signals were removed and the RW and BAI chronologies were directly analyzed for flow influence (Table 8).

Wolf River flow (gage height) data were acquired from the United States Army

Corps of Engineers (USACE 2013). Using RW and BAI chronologies, significant flow

86 signatures were identified and analyzed the same way as the climate data described above. There was a strong correlation with BAI and flow (i.e., the correlation coefficient was greater than |± 0.7| and p<0.001 for all three sites, Table 8). Thus, regression models were created to identify the details of the flow effect on tree growth for each site.

An additional analysis was conducted to determine if oak tree species show similar or different responses to flow. Correlation Matrices were created using the above methods.

5.5.4 Identifying the Effect of L. sinense on Tree Growth

To isolate the effects of L. sinense I removed the most significant flow signals

(months of greatest correlation) from the BAI chronologies for each site using the adjusted chronology tool in Dendro Tools. The output chronology consisted partially of negative numbers so I added the constant value of 40 to each year in each chronology to make all growth values positive. The resulting chronology had no age or flow related growth factors and is referred to as BAI-FR.

To determine if trees in all three sites grew significantly different before and after

L. sinense invasion, Minitab was used to conduct a BACI-ANOVA (2-factor mixed effect

ANOVA with interaction: Yijk = μ+ αij + βij + (αβ)ij + εijk), for the BAI, and BAI-FR chronologies. The dependent variable was the tree growth measurements (BAI, and BAI-

FR), the fixed independent variable was before/after L. sinense, and the random independent variable was site (NON, INV, CNT – because within each chronology, annual growth is dependent upon the growth of the surrounding years). The output from

87 the interaction term in the model revealed if there was a BACI effect. I removed the 10 years when L. sinense was establishing and compared 30 years before (1940-1969) vs. 31 years after (1980-2010) because I wanted to account for the variability of tree response time. Furthermore, NON and CNT sites were combined together as the control chronology, and INV was the impacted chronology.

5.6 Results of Tree, Site and Stand Characteristics

5.6.1 Tree Characteristics

ANOVA indicated that there were significant differences among sites for tree sizes (DBH, height), health rankings (percent dieback, live crown ratio, and percent canopy missing), and tree ages. There was no significant difference in tree DBH among sites (P=0.936, F=0.07). Trees in INV were shorter than trees in NON and CNT

(P<0.001, F=12.48) by 5.14 meters on average (Fig. 18, Table 3). There were no significant differences between mean tree heights in CNT versus NON.

Evaluation of tree health indicated no significant differences among sites for percent dieback (P=0.76, F=0.28) and live crown ratio (P=0.955, F=0.05). Trees in INV had greater percent canopy missing than trees in NON and CNT (P<0.001, F=15.57), and trees in NON and CNT were not significantly different (Fig. 19). A Tukey confidence interval, at 95% confidence, shows that trees in INV had a mean of 22.6% more canopy missing than trees in NON and CNT.

Figure 18. Plot of 95% Confidence intervals for mean tree height at each site with mean labeled. Asterisk represents a significant difference. Trees in the INV site were significantly shorter than trees in CNT and NON. There were no significant differences between CNT and NON. Plot generated in Minitab.

89 Table 3. Tree size, health, and age per site +/- 95% margin of error. Tukey confidence intervals were calculated in Minitab. The program BiaPlt was used to estimate tree ages. Variables and locations with significant differences from the other sites are highlighted in bold.

Tree Variable INV NON CNT ALL Mean DBH ± 95% M.E. 65.7±6.4 67.3±6.6 66.1±6.4 66.3±3.4 Mean Height ± 95% M.E. 32.5±2.1 37.6±1.4 37.7±1.6 35.8±1.1 Percent Dieback ± 95% M.E. 9.0±3.3 7.3±3.1 8.1±3.1 - Live Crown Ratio ± 95% M.E. 27.9±5.9 26.8±5.0 27.81±5.9 - Percent Canopy Missing ± 95% M.E. 39.0±8.1 19.3±6.1 15.0±5.2 - Mean Age ± 95% M.E. 64.1 ±3.6 62.5 ±3.9 54.6 ±2.8 60.4 ±2.1

Figure 19. Plot of 95% Confidence intervals for mean percent canopy missing at each site with mean labeled. Asterisk represents a significant difference. Trees in the INV site had greater percent canopy missing than trees in CNT and NON. There were no significant differences between CNT and NON. Plot generated in Minitab.

91 Trees in INV and NON showed no significant difference in age (Fig. 20).

However, trees in CNT were significantly younger than trees in both INV and NON.

Tukey confidence intervals, at 95% confidence, show that trees in CNT are younger than trees in INV and NON by 8.5 years, on average. Despite the difference in mean ages of the trees, the age difference is minimal and not great enough to cause significant differences in growth rate (Pederson et al. 2005). These trees are old enough to show a signal at the scale investigated, and young enough to still be growing rapidly; thus, age related effects are minimal and chronologies are still comparable. Furthermore, age related effect was removed before conducting final analyses.

Figure 20. Plot of 95% Confidence intervals for mean tree age at each site. Trees in the CNT site were significantly younger than trees in INV and NON. There were no significant differences between INV and NON. Plot generated in Minitab.

93 5.6.2 Site and Stand Characteristics

Mean wetland prevalence for all sites was between values for facultative-wetland

(FACW=2) and facultative (FAC=3) species. The mean wetland prevalence for INV

(2.869) was significantly higher than NON (2.632), and CNT (2.643), indicating that the species present in INV have slightly greater tolerances to xeric conditions than the other two sites, on average (Fig. 21). Species in INV are likely to be found in uplands 1.14%-

6.43% more often than species in NON and CNT. The presence of L. sinense in all plots in INV can explain this slight increase in the occurrence of upland species.

Establishment of xeric species after channelization could also explain this increased wetland prevalence, as there was one upland species present in INV (Viburnum rufidulum), and none in the other sites. Facultative (FAC) species were present approximately twice as often as facultative wetland (FACW) species in INV and CNT

(Table 4). In NON, FAC and FACW species were present at about the same frequency.

Facultative upland (FACU) species were present a similar percent of the time for all sites.

Figure 21. Plot of 95% Confidence intervals for wetland prevalence indices at each site with mean labeled. Woody species abundance in the INV site were significantly less indicative of wetlands than trees in NON and CNT. There were no significant differences between NON and CNT. Plot generated in Minitab.

95 Table 4. Wetland prevalence indices and percent of occurrence of each wetland indicator status in each site. Wetland Indicator Status Percent of Occurrences INV NON CNT Obligate (1) 3% 3% 8% Facultative Wetland (2) 25% 41% 30% Facultative (3) 55% 45% 51% Facultative Upland (4) 17% 10% 11% Upland (5) 0.4% 0% 0%

Mean Wetland Prevalence Index (WI) ± 95% M.E 2.869 ± 0.094 2.632 ± 0.079 2.643 ± 0.099

96 Species richness values for INV, NON, and CNT were 47, 50, and 37 respectively

(Table 5). NON had the highest species richness with 13 more species than CNT, and three more species than INV. INV had 10 more species than CNT, which had the lowest species richness.

Simpson‟s diversity index values for INV, NON, and CNT were 0.16, 0.74, and

0.72 respectively. Simpson‟s diversity index was significantly lower in INV than in

NON and CNT. INV had lower diversity than CNT by between 0.48 and 0.64, and lower diversity than NON by between 0.49 and 0.66, on average (Fig. 22).

The similarity index between INV and NON is 0.72, between INV and CNT is

0.69, and between NON and CNT is 0.65. All values range between 0 and 1, and represent the proportion of species in common between two sites (Smith and Smith

2001). Similarity of species among sites ranges from 65% to 72%; therefore, the sites are considered to be similar with respect to woody species composition and are acceptable to use for comparisons in this study. Appendix C shows a list of all species and presence absence data per site and also shows the species in common for the pairwise comparisons of sites.

97 Table 5. Total species richness, and diversity estimates with 95% Tukey confidence intervals.

INV NON CNT Richness 47 49 34 Diversity (n=30) 0.16±0.039 0.74±0.047 0.72±0.063

Figure 22. Plot of 95% Confidence intervals for Simpson‟s diversity indices (Ds) at each site with mean labeled. Woody species diversity in the INV site was significantly less than NON and CNT. There were no significant differences between NON and CNT. Plot generated in Minitab.

99 Overall there were greater numbers of pieces, volumes, and masses of CWD in

INV than in NON. The total number of CWD pieces sampled was 184 with 141 pieces in

INV and 43 in NON. There were 15.67 pieces of CWD per 100m2 in INV, and 4.78 pieces of CWD per 100m2 in NON, on average (Table 6). Mean CWD volume was greater in INV than NON by 6.08 m3/100m2. There was greater mass of CWD in INV then NON by 2,431kg/100m2.

Logs (fallen branches or tree trunks) were the most common type of CWD found in both locations, followed by stumps (short trunkless bases of trees), and then snags

(standing dead trees) (Fig. 23). In INV there were 13.44 logs/100m2, which was greater than NON by 8.77 logs, on average. There was a mean of 1.33 stumps per 100m2 in

NON which was greater than INV by 1.22 stumps. There were no snags in NON, however there was a mean of 0.89 snags/100m2 in INV.

The decay classes were more evenly distributed than the type of CWD, especially in INV where all three decay classes constituted approximately the same percentage of

CWD overall (Fig. 24). Decay class 3 (highest decomposition; most wood not intact) was most common in INV while decay class 2 (wood intact with some bark loss) was most common in NON. Decay class 1 (wood intact with little bark loss) had the fewest number of pieces in both sites; however, it accounted for the largest portion of mass and volume in INV and NON. Decay class 1 was 3 times greater in INV than NON.

100 Table 6. Mean number of pieces, volume (m3), and mass (kg/m3) for invaded and non-invaded plots for each debris type, decay class, woody type, and overall.

Mean CWD Pieces Mean Volume Mean Mass (per 100m2) (m3 per 100m2) (kg per 100m2) Debris Type Invaded Non-Invaded Invaded Non-Invaded Invaded Non-Invaded Log 13.44 4.67 7.15 1.81 2907.04 807.33 Stump 1.33 0.11 0.19 0.01 77.21 5.93 Snag 0.89 0.00 0.57 0.00 260.23 0.00

Decay Class 1 3.22 1.56 3.03 1.13 1514.11 565.73 Decay Class 2 5.22 1.67 2.66 0.38 1064.58 152.75 Decay Class 3 7.22 1.56 2.22 0.32 665.33 94.67

Total CWD 15.67 4.78 7.91 1.83 3244.33 813.33

1 10

2 10

Figure 23. Pie chart of volume of each type of coarse woody debris (CWD) in 300m2 for NON and INV sites.

3 10

Figure 24. Pie chart of volume of coarse woody debris (CWD) for each decay class in 300m2 for NON and INV sites

5.7 Dendrochronology Results

5.7.1 Quality Control and Chronologies

COFECHA output revealed that each site level RW and BAI chronology has a series intercorrelation value greater than 0.45 and they have mean sensitivity values between 0.201 and 0.235. Therefore, they are considered to be accurately dated and appropriate for dendrochronological analysis (Table 7).

Species level chronologies were also checked for quality control using

COFECHA (Table 8). Mean sensitivity values were all within the acceptable range and values were between 2.10-2.31. Series intercorrelation values were greater than 4 for all species except Q. michauxii which was close at 0.359. The lack of correlation of Q. michauxii is likely because of the differential site level responses (Fig. 25), or could possibly indicate that Q. michauxii is more variable in response to climatic signals which were not removed.

All chronologies (RW, BAI, and BAI-FR) were analyzed in ARSTAN to produce master chronologies for each site (Fig. 25), and each species (Fig. 17). The master chronologies were used in all statistical analyses to follow.

104 Table 7. Raw Ring Width (RW) and Basal Area Increments (BAI) intercorrelation, sensitivity, and dating information for each site (INV=invaded/channelized, NON=non- invaded/channelized, CNT=non-invaded/non-channelized control, as output by COFECHA

Site INV NON CNT ALL # Dated Cores 60 47 55 162 # Dated Trees 30 25 28 83 RW Series Inter-correlation 0.456 0.499 0.491 0.439 RW Mean Sensitivity 0.235 0.207 0.201 0.217 BAI Series Inter-correlation 0.448 0.503 0.488 0.430 BAI Mean Sensitivity 0.234 0.206 0.203 0.216

105 Table 8. Raw Ring Width (RW) and Basal Area Increments (BAI) intercorrelation, sensitivity, and dating information for each species, as is output by COFECHA

Site Quercus lyrata Quercus michauxii Quercus nigra Quercus pagoda # Dated Cores 11 24 86 41 # Dated Trees 6 12 45 21 RW Series Inter-correlation 0.418 0.359 0.486 0.433 RW Mean Sensitivity 0.231 0.219 0.210 0.221 BAI Series Inter-correlation 0.419 0.371 0.489 0.438 BAI Mean Sensitivity 0.230 0.220 0.211 0.221

6 10

Figure 25. Master chronologies for each site in raw ring width (RW), Basal Area Increments (BAI) and BAI with flow removed (BAI-FR) from 1940 to 2010. Chronologies made with KaleidaGraph (4.1.1 Synergy Software Inc., Reading, PA). Chronologies are graphed with curved fits set at 100 fit points.

107 5.7.2 Flow and Climate Signals

I tested the RW, BAI, plus the standard, residual, and ARSTAN chronologies against the climatic variables to identify any significant climate signals. I tested standard, residual, and ARSTAN chronologies to ensure that I did not miss an important signal.

There was no significant effect of the climatic variables precipitation, temperature, and palmer drought severity index (PDSI) on tree growth rate for the three chronologies for any month or group of months. Thus, climatic variables did not produce a signature within the tree chronologies and did not need to be removed for analysis of flow and L. sinense. These finding also support those found by Copenheaver et al. (2007), that trees in BLHFs are not sensitive to climatic variables.

Pearson correlation coefficients revealed that BAI chronologies for all sites (INV,

NON, and CNT) were significantly correlated with flow (Table 9). BAI chronologies, which remove the age related growth factor, were more strongly correlated with flow than RW chronologies. INV-BAI was most strongly correlated with July flow, NON-

BAI was most strongly correlated with the mean of July and August flow, and CNT-BAI was most strongly correlated with August flow. For the RW chronologies, only CNT-

RW was correlated with July flow, there were no significant correlations between flow and INV-RW and NON-RW chronologies (Table 9).

Contrary to my hypothesis, all chronologies revealed negative relationships with flow, meaning that as flow increases tree growth decreased. Thus, all trees grew faster after channelization (Fig. 26). Regression equations reveal that for every one meter drop in river gage height BAI increased by 12.37cm2 in INV, by 13.65 cm2 in NON, and by

108 25.23 cm2 in CNT, on average. After channelization, the Wolf River gage height dropped an average 2.019 meters compared to prior levels. Thus, based on the regression equations, tree growth in INV increased an average of 24.98 cm2/year, NON increased an average of 27.56 cm2/year, and CNT growth increased an average of 50.94 cm2/year after channelization.

109

Figure 26. Wolf River gage height and master chronologies for each site in Basal Area Increments (BAI) from 1940 to 2010, shows the strong inverse relationship between flow and tree growth at all sites.

110 Table 9. Months of flow with the most significant correlation to site level BAI and RW chronologies, plus the resulting Spearman and Pearson Correlation coefficients and regression results for BAI versus flow of highly correlated months. All values with significance are in bold text. Highest correlated Pearson’s Pearson’s P- Spearman’s Spearman’s n = # of Years R2 Month/s Correlation Value Correlation P-Value used in Reg’n INV-BAI July Flow -0.749 < 0.001 -0.737 < 0.001 56.1 57 NON-BAI July-Aug Mean Flow -0.744 < 0.001 -0.59 < 0.001 55.4 57 CNT-BAI August Flow -0.835 < 0.001 -0.599 < 0.001 76.6 54 INV-RW July-Aug Mean Flow -0.309 0.019 -0.291 0.026 - - NON-RW Oct Flow -0.193 0.145 -0.174 0.192 - - CNT-RW July Flow -0.720 < 0.001 -0.690 < 0.001 - -

1 11

Despite showing the same overall trend, Q. pagoda grew significantly faster than the other species while Q. lyrata grew significantly slower than the other species, on average (p<0.001) (Fig. 17). Using BAI chronologies (to remove age as a factor) Q. pagoda grew faster than Q. nigra by 1.9 cm2 to 12.8 cm2, faster than Q. michauxii by 1.1 cm2 to 12.3 cm2, and faster than Q. lyrata by 15.1cm2 and 26.3 cm2, per year, on average. Quercus lyrata grew slower than Q. nigra by between 15.1cm2 and 26.3cm2, and slower than Q. michauxii by between 15.1cm2 and 26.3cm2, per year, on average.

Each species‟ BAI chronology was also strongly correlated with flow (Table 10).

Correlation matrices revealed that Q. michauxii was the most responsive to flow, followed by Q. lyrata, Q. nigra, and finally Q. pagoda. For every one meter drop in river gage height in July, annual BAI growth increase in Q. lyrata by 12.17cm2, and increased in Q. michauxii 18.90cm2, on average. For every one meter drop in July/August mean river gage height, annual BAI growth increased in Q. pagoda by 18.21cm2, and annual

BAI growth of Q. nigra increased by 12.89cm2, on average.

Quercus michauxii had the greatest increase in growth after channelization, followed by Q. pagoda, Q. nigra, and Q. lyrata. Given that the river dropped an average of 2.02 meters after channelization annual growth of Q. michauxii increased by 38.16 cm2, Q. pagoda increased by 36.77 cm2, Q. nigra increased by 22.48 cm2, and Q. lyrata increased by 24.58 cm2, on average after channelization.

112 Table 10. Months of flow with the most significant correlation to species BAI chronologies, plus the resulting Spearman and Pearson Correlation coefficients Species Month/s with Pearson’s Pearson’s P- Spearman’s Spearman’s P- n = # of Years R2 greatest Flow Signal Correlation Value Correlation Value used in Reg’n QULY July -0.844 < 0.001 -0.727 < 0.001 71.2 57 QUMI July -0.859 < 0.001 -0.677 < 0.001 73.8 57 QUNI July & August -0.835 < 0.001 -0.714 < 0.001 65.4 58 QUPA July & August -0.809 < 0.001 -0.674 < 0.001 64.7 58

3 11

5.7.2 Signal of L. sinense

Two-factor mixed effects ANOVA, with CNT and NON combined (CNT/NON) as control sites revealed that there was a significant BACI effect (interaction) for BAI, and BAI-FR chronologies in response to L. sinense (BAI; p<0.001 F=23.24, and BAI-FR; p=0.03, F=4.8) (Fig. 27). There were no significant site effects or before/after effects for

BAI and BAI-FR chronologies (P>0.1).

For BAI chronologies all sites had an increase in growth after L. sinense invasion; however, the impacted site, INV, had less of an increase and CNT/NON had the greatest increase in growth. INV mean growth rate was significantly less than the control sites,

NON/CNT, after L. sinense invasion, but there was no significant difference in mean growth rate before L. sinense invasion (Fig. 27).

When flow effects were removed from the chronologies (BAI-FR), the control sites, CNT/NON, increased in growth after L. sinense invasion, and the impacted INV site showed a reduction in growth (Fig. 27). The BAI-FR chronologies revealed that trees in all sites grew similarly before L. sinense invasion; however, trees in INV grew slower than the other sites after invasion, which indicates that trees there are responding to L. sinense invasion (Fig. 27). ANOVA was conducted because it is robust to the assumptions, despite that one assumption, equality of variance, was not met (Tomarken and Serlin 1986).

The results reveal that after accounting for age, climate, and flow, trees in the L. sinense invaded habitat grew significantly slower than trees in non-invaded habitats after invasion. Given that the majority of all other contributing factors were constant (size,

114 similarity, wetland prevalence, similarity, and richness) this slowing of growth can be attributed L. sinense invasion.

115

Figure 27 a and b. Interaction plots for BAI (A) and BAI-FR (B) showing the BACI effects after L. sinense invasion. Notice that for BAI (A) chronologies the increase in growth after invasion is less for INV than for CNT/NON. Also notice that after flow is removed from the chronology (B)the trees in the invaded site grew slower after invasion than they did prior to invasion.

116 5.8 Discussion and Conclusions

5.8.1 Tree, Site, and Stand Characteristics

In general tree, site, and stand characteristics were similar among all sites. The similarity is mostly a result of the selection methods, as none of the damaged trees were cored in INV (Chapter 5.2). Because only “healthy-looking” the differences that were present seem more significant and can be explained by the invasion of L. sinense. First, trees in INV were significantly shorter (by a mean of 5m) and had greater percent canopy missing (by a mean of 23%) than trees in the other two sites. The shorter trees in INV are likely a result from the lack of competition with other canopy trees because the canopy trees in INV are competing with the shrub, L. sinense more than other canopy trees, because large standing trees in INV are sparsely distributed. The greater percent canopy missing in INV may also explain the shortness of trees.

The tree health data indicate that trees in INV are self-thinning. Trees in INV had low percent dieback (mean = 9%), which indicates that there were few visible dead limbs.

However, percent canopy missing (mean = 39%) was greater in INV than the other two sites. This suggests that trees in INV are concentrating resources toward growing the remaining living canopy branches. The results indicate that dead branches mostly disconnect and drop. This “self-thinning” postulation is further supported by the CWD data.

Trees in INV lost mass amounts of limbs (represented as dead logs), as indicated by the three times more pieces of CWD in INV than in NON, and 99% of the pieces in

117 INV were logs. Moreover, the amount of CWD from each decay class was evenly distributed in INV, which indicates that trees have been losing limbs at a steady rate, and that rate is comparable to the decomposition rate. Thus, the data further support that trees in INV are responding to L. sinense by self-thinning, which may eventually lead to tree mortality.

The increased richness in BLHFs along the channelized portion of the river is likely a response to disturbance and urbanization (Armesto 1985). Furthermore, the combination of disturbance and increased species richness is indicative of low forest productivity; likewise the lack of disturbance and low species richness of CNT could suggest that there is greater forest productivity in CNT (Kondoh 2001). The dendrochronology data also support the hypothesis that forests in the non-channelized, rural location (CNT) produce more biomass than forests in channelized urban locations

(INV and NON) (Fig. 25).

Reduced diversity in INV is undoubtedly a result of the densely packed L. sinense stems – as diversity is a measure of richness and evenness. In CNT and NON the woody species were much more evenly distributed, and a different species was encountered approximately 73% of the time, while in INV a different species was encountered 16% of the time on average. These lowered diversity values support the findings of many other

L. sinense investigations (Merriam and Feil 2002; Grove and Clarkson 2005; Pokswinski

2008; Hanula et al. 2009; Klock 2009; Osland et al. 2009; Ulyshen et al. 2010; Greene and Blossey 2011; Hanula and Horn 2011a,b).

118 5.8.2 Effects of Climate and Flow

To my knowledge, this is the most comprehensive analysis of effects of climate, flow and woody invasion on BLHF tree growth that has ever been accomplished.

In support of the findings of Copenheaver et al. (2007), trees in the BLHFs along the

Wolf River did not show significant responses to climatic variables including: precipitation, temperature, and PDSI. They did, however, show a significant response to flow, which was expected (Copenheaver et al. 2007). These finding support that species in BLHFs are more receptive to flow and microclimate characteristics than to climatic variables.

The responses of BLHF oak trees to flow and channelization are similar to those observed by Weins and Roberts (2003), who looked at the effect of river headcutting

(another result of channelization) on the rates of willow oak (Quercus phellos) growth along the Wolf River, Tennessee. They cored trees in an impacted, or “headcut,” site located about 24.1 km upstream of INV, where the water table was observed to have dropped. Their reference site (control) was located directly across the river from CNT.

They observed that trees in the reference reach did not increase in growth as much as trees in the headcut portion of the river. Contrarily, my data show that trees in non- impacted areas (CNT) also respond significantly to downstream flow. In fact, trees in

CNT had a much greater relationship to flow than did trees in other sites (Table 9).

One explanation for the differing results between my results and those of Weins and Roberts (2003) could result from the differing analyses conducted. Weins and

Roberts (2003) looked at decadal growth of BAI from 1940 to 2000, and observed that all

119 trees increased in growth after channelization. However, trees in the control site did not continue to increase in growth as much as trees in the headcut portion for the ten years between 1990-1999. Compared to my study, they did not perform any regression analyses with climate or flow, and they were limited in the amount of time analyzed (i.e.,

I had 10 more years of data). Another possible reason could be that different species were cored, and Q. phellos might respond differently than the oaks cored in my study.

An interesting observation about tree growth response to flow is that trees in CNT showed the greatest relationship to flow, yet the trees in INV and NON are located directly adjacent to the channelized river. This is possibly due to changes in the water table. Trees in BLHFs get the majority of their water from the water table (Naiman and

Décamps 1997; Copenheaver 2007), which is correlated with stream gage height (Weins and Roberts 2003). Furthermore, as noted by Weins and Roberts, the water table dropped significantly in channelized and headcut portions of the river after channelization, but did not drop in the headcut reach. Thus, trees in CNT are likely still relying on the water table, contributing to the strong correlation and greater increase in growth. Trees in INV and NON may no longer be able to access water from the water table and have less correlation to flow.

Trees in all sites showed a negative correlation to flow, and thus a positive growth trend as a result to channelization. This is likely due to slow growth of BLHF trees from anoxic conditions, thus supplying more oxygen to the roots and allowing the trees to grow more rapidly (Shankman 1996). These results supplies further evidence that endemic bottomland trees rely on flood pulses to reduce interspecific competition. The adaptations to anaerobic conditions allow BLHF species to thrive where upland species

120 cannot (Middleton 2002). These results also support that BLHF species are vulnerable in altered hydrologic conditions because they are specialists, adapted to flood pulsing

(Richardson et al. 2007, Middleton 2002).

5.8.3 Effects of L. sinense

These dendrochronology data supply evidence that L. sinense is responsible for reduced rate of oak tree growth in BLHFs of the Wolf River, Tennessee. After a thorough literature search, only one other study sought to determine if an invasive shrub slows canopy tree growth (Hartman and McCarthy 2007). Similar to what Hartman and

McCarthy (2007) observed with the invasive shrub Lonicera maackii (Rupr.) Maxim.

(amur honeysuckle), I discovered that L. sinense is likely responsible for reduction in growth of native trees. There are, however, differences between the methods of the two studies including matching tree species to site in my study vs. no site matching in their study, differences in sample size , differences in number of sampling locations , differences in the handling of climatic variables, and investigation of flow effects on growth rate. All of these differences in the study approach could result in different outcomes, but our two studies show the same qualitative results – the invasive shrubs not only prevent growth of other shrubs and herbaceous species, but they are also detrimental to established canopy trees (Hartman and McCarthy 2007).

If I could expand on this research I would core more trees and more tree species, and have multiple site replicates in more than one state. Questions I would address include: “Do oak tree responses to invasion show an edge effect? Does proximity to

121 streams affect tree responses? Do tree species other than oaks respond differently to channelization and invasion?” Another way to elaborate on this study would be to investigate the differential responses of trees to climate and flow variables over time. I would also like to answer questions about variability of tree growth in response to climate, L. sinense, and altered hydrology over time. In my study, I investigated correlations between the full chronologies vs. climate and flow variables; however, I did not investigate partial chronologies. For example, one could look at the effects of variables in the years before channelization, and compare the correlations to responses after channelization. I hypothesize that trees in INV and NON would show stronger correlations with flow before channelization, and that they began responding more to precipitation or PDSI more than flow, after channelization, as a response to the lowered water table.

122 CHAPTER VI

DISCUSSION AND CONCLUSIONS

To my knowledge, I am the first researcher to supply empirical evidence that 1)

Ligustrum sinense seeds are dispersed by water (hydrochory) in riverine wetlands, 2) long-term inundation of L. sinense fruits reduces their viability and germination, and 3)

Ligustrum sinense slows oak tree growth rates in hydrologically altered riparian forests.

In conclusion, L. sinense invasion is placed in the “backseat driver” category of invasive mechanisms. Ligustrum sinense initially invades drier habitats adjacent to a channelized river which caused drier conditions – a “novel niche” (MacDougall et al. 2009). The novel niche no longer supports the conditions that restricted establishment of upland competitors and that allowed the adapted BLHF woody plant species to regenerate

(Middleton 2002). The long distance dispersal mechanism of hydrochory allowed L. sinense to quickly colonize these “novel niches” (Richardson et al. 2007). Once established, L. sinense then competed with native oak species and reduced tree growth, possibly leading to an early death of the native oaks. Thus, eradication strategies should aim at rigorous removal of the invasive species and a return to pre-existing (un-altered) conditions, or should consist of human-assisted establishment of native early succession species that can thrive in the altered conditions.

123 Bottomland hardwood forests with altered flood pulses are highly vulnerable to L. sinense invasion, and once established L. sinense dramatically changes forest productivity and function. My research supplies evidence that the invasive shrub, L. sinense, uses hydrochory as a dispersal mechanism in addition to the previously documented dispersal mechanism of birds. Therefore L. sinense utilizes multiple dispersal mechanisms, while taking advantage of well-drained soil, to establish and dominate the understory of BLHFs in hydrologically altered riverine wetlands. Moreover, L. sinense poses threats to longevity of native BLHFs. In support of my findings, that L. sinense reduces tree growth rate, Hartman and McCarthy (2007) found that the invasive shrub Lonicera maackii reduces tree growth rate in Southwest Ohio bottomland forests. Boyce (2009), who conducted an all-inclusive synthesis of invasive shrubs in the United States, found data to further support that invasive shrubs are major threats to forests. He found that most of the known invasive shrub species prevented forest regeneration and altered forest functions, and he found that over 100 species of shrubs are established as invasive and the characteristics that make invasive shrubs successful in forests are asexual growth, prolific seed production, and evergreen or semi-evergreen foliage (Boyce 2009).

Ligustrum sinense possesses all of those traits (Chapter 1.3).

Water dispersal is the most common method of dispersal for BLHF species

(Richardson et al. 2007). However, dispersal alone does not ensure establishment and longevity. This is made evident by the conflicting diversity values between dispersed and established species in the BLHFs of the Wolf River. The diversity values for established species were significantly different among sites with INV being much lower than the other two sites (INV= 0.16, NON=0.74, and CNT= 0.72). The low diversity values of

124 established woody species in INV can be explained by the mass amount of clustered L. sinense stems. Naturally, this should translate to mass amount of L. sinense seed production and lower seed diversity values in INV. However, diversity of water- dispersed seeds was comparable among all sites (INV= 0.611, NON=0.568, and CNT=

0.711). These conflicting results supply evidence that great diversities of seeds are still dispersed into the L. sinense invaded forests but are not able to establish or thrive. Some possible reasons for this could be that L. sinense shades out the forest floor and prevents germination of seeds (Grove and Clarkson 2005; Brantley 2008; Smith et al. 2008; Boyce

2009; Osland et al. 2009; Greene and Blossey 2011). This could also be caused by reduction of established native populations as a result of competitive exclusion (Merriam and Feil 2002; Greene and Blossey 2011).

Forest integrity, or the measure of a forest‟s ability to support diverse communities and produce timber (Seymour and Hunter 1999), is greatly reduced in L. sinense invaded forests (Houston et al. 2010). Many researchers have shown that regeneration of native woody species is nearly non-existent in invaded forests (Grove and

Clarkson 2005; Brantley 2008; Smith et al. 2008; Boyce 2009; Osland et al. 2009;

Greene and Blossey 2011). Thus, when the native tree dies, these invaded forests are destined to become L. sinense monocultures that support very little biodiversity, regenerate only L. sinense seedlings, and thus have almost no ecological value (Merriam and Feil 2002; Greene and Blossey 2011). Monocultures of L. sinense would result in little-to-no timber production, and thus have little economic value (Houston 2011).

Furthermore, these invasions create major recreational risks, where there are greater amounts of fallen limbs compared to non-invaded BLHFs (Table 6). People who bike

125 and hike these forests are at greater risk of being hit by falling limbs, and these forests should be avoided in heavy wind and rain.

Unfortunately, the high cost of L. sinense removal discourages hope for future eradication. It is estimated that L. sinense removal costs between $370-$420 per hectare to remove, and that approximately140,550 hectares of the Southeast United States are invaded (Johnson 2010). Therefore, it would cost between $52.1 and 59.0 million dollars to eradicate L. sinense in the Southeast United States alone. Furthermore, the public may be resistant to eradication based on conflicting agendas. In recreational bike trails privet forms long bush tunnels that bicyclists enjoy riding through (NEABC 2014). Still, other citizens may find the dense understory to be comforting and protecting from the elements. Some citizens may also have concerns over wildlife such as birds that eat the fruits in the winter (Reichard et al. 2005). Furthermore, it is likely that L. sinense would return to these regions because of the large amount of local landscape introductions, known as propagule pressure. Propagule pressure of L. sinense, which is greater in urban areas (Burton et al. 2005), plays a major role in the invisibility of urban forests (Colautti et al. 2006). Furthermore, if privet could be eradicated from its current geographic extent, urban forests would be at greater risk for a return of invasion than rural forests because there are more nearby propagules, or seed sources (Colautti et al. 2006).

There are currently no federal or state laws restricting the sale and planting of L. sinense in the United States, even though it is included on many states‟ noxious weed lists (USDA 2014). Connecticut and New Hampshire have laws prohibiting the planting, sale, and distribution of L. sinense‟s larger leaved, deciduous cousin, border privet (L. obtusifolium Siebold & Zucc.), but none for L. sinense (CGA 2004; NHDA 2012). Major

126 difficulties in making new policies involve lack of public education, lack of scientific communication, and little public and private activism. Thus, emphasis must be given to each of these categories if new policies are to be made.

One important place to begin the mission of preventing invasive species spread and success is with public awareness and education. If people were aware of the extreme negative consequences of invasive species on native biodiversity then they might be less likely to plant exotic species, especially those known to be invasive and/or noxious

(Randall 1996). One simple solution would involve attaching an informative tag to every privet plant that is sold in the United States. The tag could include information about the effects of the invasion as well as suggestions of native plant lists that would supply alternatives. Furthermore, educating the public about invasive plants should be a priority in k-12 and college level schools, at city, county, state, and national parks, at nature centers, museums, and other community events. Additionally, these actions can help inform the public on how people can change their lifestyles to reduce the negative consequences of invasive species.

Another major obstacle in formulating new invasive species policies is that there is a lack of communication among scientists and among scientists and politicians

(Reichard 1997). There are numerous organizations designed to inform scientists and share information, such as the Invasive Plant Atlas of New England, the Invasive Plant

Atlas of the Mid-South, CitiSci, the Global Invasive Species Information Network, and the Florida Exotic Plant Pest Council (Reichard 1997; Pyke et al. 2008; Simpson et al.

2009; Langeland et al. 2011). However, many of these organizations focus on different species and ecosystems, do not consider L. sinense and/or BLHFs, and do not

127 communicate with each other or organizations that formulate legislation.

Communication among these groups, citizens, landscapers, and lawmakers is imperative to suggest appropriate legislation.

Citizens can also play an influential role when it comes to invasive L. sinense mitigation. Citizens should stop introducing L. sinense to novel regions. Landscapers should shift their focus toward planting native species that are also hardy (Dirr 1998).

There are many native alternatives for L. sinense such as Viburnum prunifolium L.

(blackhaw), Osmanthus aamericana (L.) A.Gray (devilwood), and Prunus caroliniana

(Mill.) Aiton (Carolina laurelcherry) (Burrell 2006). Not only should people stop introducing L. sinense, but they should also remove it from their property. People should do their part to prevent L. sinense on their land because it may lower property value, and removal will reduce establishment of new populations. Finally, citizens should volunteer for local community eradication events that often occur in urban parks (Langeland et al.

2011).

128 REFERENCES

Arkansas Multi-Agency Wetland Planning Team (MAWPT), 2001. Wetlands in Arkansas. Accessed 07/22/2011 at URL http://www.mawpt.org/wetlands/.

Armesto, J.J., Pickett, S.T.A., 1985. Experiments on disturbance in old-field plant communities: impact on species richness and abundance. Ecology 66, 230-240.

Ashton, I.W., Hyatt L.A., Howe K.M., Gurevitch J., Lerdau M.T., 2005. Invasive species accelerate decomposition and litter nitrogen loss in a mixed deciduous forest. Ecological Applications 1, 1263–1272.

Barney, J.N., Whitlow, T.H., 2008. A unifying framework for biological invasions: the state factor model. Biological Invasions 10, 259–272.

Bauer, 2012. Invasive species; „backseat drivers of eco-system change.‟ Biological Invasions 14, 1-10.

Boyce, R.L., 2009. Invasive shrubs and forest tree regeneration. Journal of sustainable forestry, 28(1-2), 152-217.

Bradley, B., Wilcove, D., Oppenheimer, M., 2009. Climate change increases risk of plant invasion in the eastern United States. Biological Invasions 12, 1855-1872.

Bragg, D.C., 2013. Composition, Biomass, and Overstory Spatial Patterns in a Mature Pine-Hardwood Stand in Southeastern Arkansas. Castanea 78(1), 37-55.

Bragg, D.C., Heitzman, E., 2009. Composition, structure, and dynamics of a mature, unmanaged, pine-dominated old-field stand in southeastern Arkansas. Southeastern Naturalist 8(3), 445-470.

129 Brantley, E.F., 2008. Influence of Chinese Privet (Ligustrum sinense Lour.) on Riparian Forests of the Southern Piedmont: Net Primary Productivity, Carbon Sequestration, and Native Plant Regeneration. Ph.D. Dissertation, Auburn University.

Brown, C.E., Pezeshki, S.R., 2000. A study on waterlogging as a potential tool to control Ligustrum sinense populations in western Tennessee. Wetlands 20, 429–437.

Burke, M.K., Chambers, J.L., 2003. Root dynamics in BLHFs of the Southeastern United States Coastal Plain. Plant and Soil Science 250(1), 141-153.

Burnette, D.J., 2013 Dendro-Tools, Dendrochronology Data Analysis Software. Accessed 12/13/2013 at URL http://www.djburnette.com/programs/dendrotools/index.html.

Burrell, C.C., 2006. Native alternatives to invasive plants. No. 185, Brooklyn Botanic Garden.

Burrows, F.J., Kohen, J., 1986. Inhibition of germination in privet. Plant Protection Quarterly 1, 107-108.

Burton, M.L., Samuelson, L.J., Pan, S., 2005. Riparian woody plant diversity and forest structure along an urban-rural gradient. Urban Ecosystems 8, 93–106.

Capon, 2003. Plant Community Responses to Wetting and Drying in a Large Arid Floodplain. River Research and Applications 19, 509–520.

Capon, S., Brock, M., 2006. Flooding, soil seed bank dynamics and vegetation resilience of a hydrologically variable desert floodplain. Freshwater Biology 51, 206–223.

Catford, J.A., Downes, B.J., 2010. Using multi-scale species distribution data to infer drivers of biological invasion in riparian wetlands. Diversity Distributions 16, 20- 32.

Catford, J.A., Jansson, R., Nilsson, C., 2009. Reducing redundancy in invasion ecology by integrating hypotheses into a single theoretical framework. Diversity Distributions 15, 22–40.

130 CGA (Connecticut General Assembly), 2004. Public Act No. 04-203. Accessed 05/29/2014 at URL http://www.cga.ct.gov/2004/act/Pa/2004PA-00203-R00SB- 00547-PA.htm.

Chapman, M.W., 1977. I remember Raleigh. Riverside Press.

Clavero, M., García-Berthou, E., 2005. Invasive species are a leading cause of animal extinctions. Trends in Ecology and Evolution 20, 110.

Colautti, R.I., MacIsaac, H.J., 2004, A neutral terminology to define „invasive‟ species. Diversity Distributions 10, 135-141.

Colautti, R.I., Grigorovich, I.A., & MacIsaac, H.J. 2006. Propagule pressure: a null model for biological invasions. Biological Invasions 8(5), 1023-1037.

Cook, E.R., 1985. A Time Series Approach to Tree-Ring Standardization. PhD Dissertation. University of Arizona, Tucson, AZ, USA.

Cook, E.R., Kairiukstis, L.A., 1990. Methods of Denrochronology, Applications in the Environmental Sciences. Kluwer Academic Publishers. Norwell, MA.

Copenheaver, C.A., Yancey, M.W., Pantaleoni, E., Emrick, V.R., 2007. Dendroclimatic Analysis of a Bottomland Hardwood Forest: Floodplain vs. Terrace Responses. The Journal of the Torrey Botanical Society 134(4), 505-511.

DeFries, R.S., Rudel, T., Uriarte, M., Hansen, M., 2010. Deforestation driven by urban population growth and agricultural trade in the twenty-first century. Nature Geoscience 3, 178-181.

Diez, J.M., D'Antonio, C.M., Dukes, J.S., Grosholz, E.D., Olden, J.D., Sorte, C.J., Miller, L.P., 2012. Will extreme climatic events facilitate biological invasions? Frontiers in Ecology and the Environment 10(5), 249-257.

Didham, R.K., Tylianakis, J.M., Hutchison, M.A., Ewers, R.M., Gemmel, N.J., 2005. Are invasive species the drivers of ecological change? Trends in Ecology and Evolution 20, 470–474.

131 Dirr, M.A., 1998. Manual of woody landscape plants. The American Phytopathological Society (APS). St. Paul, MN.

Drummond, M.A., Loveland, T.R., 2010. Land-use pressure and a transition to forest- cover loss in the eastern United States. Bioscience 60(4), 286-298

Gosselink, J.G., Lee, L.C., Muir, T.A., 1990. Ecological processes and cumulative impacts: illustrated by bottomland hardwood wetland ecosystems. Lewis Publishers, Chelsea, Michigan, USA.

Greene, B.T., Blossey, B., 2011. Lost in the weeds: Ligustrum sinense reduces native plant growth and survival. Biological Invasions 14(1), 139-150.

Grove, E., Clarkson, B.D. 2005. An ecological study of Chinese privet (Ligustrum sinense Lour.) in the Waikato region. Centre for Biodiversity and Ecology Research Department of Biological Sciences: University of Waikato Contract Report No. 41.

Guo, Y., Shelton, M., Lockhart, B., 1998. Effects of flood duration and season on germination of black, cherrybark, northern red, and water oak acorns. New Forests 15, 69–76.

Gurevitch, J., Fox, G.A., Wardle, G.M., Inderjit, Taub, D., 2011. Emergent insights from the synthesis of conceptual frameworks for biological invasions. Ecology Letters 14, 407–418.

Gurevitch, J., Padilla, D.K., 2004. Are invasive species a major cause of extinctions? Trends in Ecology and Evolution 19, 470–474.

Hanula, J.L., Horn, S., 2011a. Removing an invasive shrub (Chinese privet) increases native bee diversity and abundance in riparian forests of the southeastern United States. Insect Conservation and Diversity 4, 275-283.

Hanula, J.L., Horn, S., 2011b. Removing an exotic shrub from riparian forests increases butterfly abundance and diversity. Insect Conservation and Diversity 262:674- 680.

132 Hanula, J.L., Horn, S., Taylor, J.W., 2009. Chinese privet (Ligustrum sinense) removal and its effect on native plant communities of riparian forests. Invasive Plant Science Management 2, 292–300.

Hartman, K.M., McCarthy, B.C., 2007. A dendro-ecological study of forest overstory productivity following the invasion of the non-indigenous shrub Lonicera maackii. Applied Vegetation Science 10, 3-14.

Hoffmann, J.H., Moran, V.C., 1991. Biological control of Sesbania punicea (Fabaceae) in South Africa. Agriculture, Ecosystems, & Environment 37, 137–157.

Holmes, R., 1983. Computer assisted quality control in tree-ring dating and measurement. Tree-Ring Bulletin 44, 69-75.

Hood, W.G., Naiman, R.J., 2000. Vulnerability of riparian zones to invasion by exotic woody plants. Plant Ecology 148, 105–114.

Houston, A.E., Goudie, C.A., Simco, B.A., Kennedy, M.L., 2010. An Assessment of Forested Habitat at Shelby Farms Park, Memphis, Shelby County, Tennessee. Prepared for: Tennessee Department of Agriculture Division of Forestry Ecological Research Center Department of Biological Sciences.

Howe, H.F., Miriti, M.H., 2004. When seed dispersal matters. Biological Science 54, 651–660.

Hudson, J.R., 2013. Effects of removing Chinese privet (Ligustrum sinense) on plant communities, pollinator communities, and tree growth in riparian forests five years after removal with mechanisms of reinvasion. Doctoral dissertation, University of Georgia.

Johnson, G.R., 1997. Tree preservation during construction: a guide to estimating costs. Minnesota Extension service, University of Minnesota.

Johnson, J., 2010. Aerial Privet Eradication. Georgia Forestry Commission, Athens, GA.

133 Johnston, C.A., Detenbeck, N.E., and Niemi G.J., 1990. The Cumulative Effect of Wetlands on Stream Water Quality and Quantity: A Landscape Approach. Biogeochemistry 10(2), 105-141.

Jordan, Bowers, L., 1973, Treering Dating of the Bald Cypress (Taxodium distichum L. Rich.) in the Lower Mississippi Valley.MS thesis, Arkansas State University. Jonesboro.

Junk, W.J., Bayley, P.B., & Sparks, R.E., 1989. The flood pulse concept in river- floodplain systems. Canadian Special Publication of Fisheries and Aquatic Sciences 106(1), 110-127.

Kartesz, J.T., 2011. The Biota of North America Program (BONAP) North American Plant Atlas Chapel Hill, N.C., Accessed 11/09/2011 at URL http://www.bonap.org/MapSwitchboard.html.

Kittell, M.M., 2001. Relationships among Invasive Chinese Privet, Plant Diversity, and Small Mammal Captures in Southeastern Deciduous Forests. MS Thesis, Clemson University.

Klock, M., 2009. Using Demographic Models to Manage Chinese Privet (Ligustrum sinense Lour.). MS Thesis, Louisiana State University.

Kondoh, M., 2001. Unifying the relationships of species richness to productivity and disturbance. Proceedings of the Royal Society of London. Series B. Biological Sciences 268(1464), 269-271.

Konno, K., Chikara, Hirayma, C., Shinbo, H., Nakamura, M., 2009. Glycine addition improves feeding performance of non-specialist herbivores on the privet, Ligustrum obtusifolium: In vivo evidence for the physiological impacts of anti- nutritive plant defense with iridoid and insect adaptation with glycine. Applied Entomology and Zoology 44, 595–601.

Krebs, C.J., 1999. Ecological methodology. 2nd. ed. Benjamin/Cummings, Menlo Park, CA, US.

134 Langeland, K.A., Burkes, K.C., 1998. Identification and Biology of Non-Native Plants in Florida‟s Natural Areas. University of Florida Press, Gainesville.

Langeland, K.A., Ferrell, J.A., Sellers, B., MacDonald, G.E., Stocker, R.K., 2011. Integrated Management of Nonnative Plants in Natural Areas of Florida. University of Florida, IFAS Extension.

Larson, B.M.H., 2007. An alien approach to invasive species: objectivity and society in invasion biology. Biological Invasions 9, 947–956.

Light, T., Marchetti, M.P., 2007. Distinguishing between invasions and habitat changes as drivers of diversity loss among California‟s freshwater fishes. Conservation Biology 21, 434–446.

Lockwood, J.L., Cassey, P., Blackburn, T., 2005. The role of propagule pressure in explaining species invasions. Trends in Ecology and Evolution 20, 223–228.

MacDougall, A.S., Gilbert, B., Levine, J.M., 2009. Plant invasions and the niche. Journal of Ecology 97, 609–615.

MacDougall, A.S., Turkington, R., 2005. Are invasive species the drivers or passengers of change in degraded ecosystems? Ecology 86, 42–55.

Matlack, G.R., 2002. Exotic plant species in Mississippi, USA: critical issues in management and research. Natural Areas Journal 22, 241-247.

McEuen, A.B., Curran, L.M., 2004. Seed dispersal and recruitment limitation across spatial scales in temperate forest fragments. Ecology 85, 507–518.

Merriam, R.W., 2003. The abundance, distribution and edge associations of six nonindigenous, harmful plants across North Carolina. Journal of the Torrey Botanical Society 130, 283-291.

Merriam, R.W., Feil, E., 2002. The potential impact of an introduced shrub on native plant diversity and forest regeneration. Biological Invasions 4, 369–373.

135 Middleton, B., 1995. Sampling Devices for the Measurement of Seed Rain and Hydrochory in Rivers. Bulletin of the Torrey Botanical Club 122, 152-155.

Middleton, B., 2002. Flood Pulsing in Wetlands: Restoring the Natural Hydrological Balance. Wiley and Son. New York.

Miller, J.H., 1997. Exotic invasive plants in southeastern forests. In: Britton KO, Proceedings of exotic pests of eastern forests. Tennessee Exotic Pest Plant Council, Nashville, TN 97–106.

Mitchell, J.D., 2009. Biogeochemical Influence of Chinese Privet in Riparian Forests in West Georgia and the Influences of Oyster Harvesting in Apalachicola Bay, FL. Ph.D. Dissertation, Auburn University.

Mooney, H.A., Cleland, E.E., 2001. The evolutionary impact of invasive species. Proceeding of the National Academy of Science 98, 5446–5451.

Morris, L.L., Walck, J.L., Hidayati, S.N., 2002. Growth and reproduction of the invasive Ligustrum sinense and native Forestiera ligustrina (Oleaceae): implications for the invasion and persistence of a nonnative shrub. International Journal of Plant Sciences 163, 1001–1010.

Naiman, R.J., Décamps, H., 1997. The ecology of interfaces -- riparian zones. Annual Review of Ecology, Evolution, and Systematics 28, 621-658.

Naiman, R.J., 2010. Riparian Zones. River Ecosystem Ecology: A Global Perspective 182.

NEABC (Northeast Arkansas Bicycle Coalition), 2014. Personal communication.

NHDA (New Hampshire Department of Agriculture), 2012. Prohibited Invasive Plant Species Rules, Agr 3800. Accessed 05/29/2014 at URL http://www.agriculture.nh.gov/publications-forms/documents/prohibited-invasive- species.pdf

136 NOAA, 2012. National Oceanic Atmospheric Administration, National Climate Data Center Accessed 05/12/2013 at URL http://www.ncdc.noaa.gov, TN Zone 4.

Osland, M.J., Pahl, J.W., Richardson, C.J., 2009. Native bamboo [Arundinaria gigantea (Walter) Muhl., Poaceae] establishment and growth after the removal of an invasive nonnative shrub (Ligustrum sinense Lour., Oleaceae): implications for restoration. Castanea 74, 247–258.

Panetta, F.D., 2000, Fates of fruits and seeds of Ligustrum lucidum W.T.Ait. and L. sinense Lour. maintained under natural rainfall or irrigation. Australian Journal of Botany 48, 701-705.

Pederson, N., Cook, E.R., Myvonwyn-Hopton, H., Jacoby, G., 2005. Does age matter? Evidence of vigorously growing, ancient oaks in the eastern U.S. 14-16. in Bennett, K.P., (technical coordinator), Moving Toward Sustainable Forestry: Lessons from Old Growth Forests. University of New Hampshire Cooperative Extension Natural Resource Network Report.

Planty-Tabacchi, A.M., Tabacci, E., Naiman, J., Deferrari, C., Décamps, H., 1996. Invasibility of species-rich communities in riparian zones. Conservation Biology 10, 598-607.

Poff, N.L., Allan, J.D., Bain, M.B, Karr, K.L., Prestegaard, K.L., Richter, B.D., Stromberg, J.C., 1997. The natural flow regime: A paradigm for river conservation and restoration. Biological Science 47, 769-784.

Pokswinski, S., 2008. Invasive Characteristics of Chinese Privet (Ligustrum sinense Lour.) in a Bay Swamp in the Fall Line Hills of East-Central Alabama. MS Thesis, Auburn University.

Pyke, C.R., Thomas, R., Porter, R.D., Hellmann, J.J., Dukes, J.S., Lodge, D.M., Chavarria, G., 2008. Current practices and future opportunities for policy on climate change and invasive species. Conservation Biology 22(3), 585-592.

Quarles, R., 2012. A Method for Determining the Appropriate Spatial Resolution to Detect Ligustrum sinense in Shelby Farms Park using WorldView-2 multi- spectral satellite imagery. MS Thesis, University of Memphis.

137 Randall, J.M., Marinelli, Hanson, and Lindstrom, 1996. Invasive Plants; Weeds of the Global Garden.” Brooklyn Botanic Gardens.

Reichard, S.H., Schmitz, D.C., Simberloff, D., Morrison, D., Lehtonen, P.P., Windle, P.N., Mezitt, R.W., 2005. The tragedy of the commons revisited: invasive species. Frontiers in Ecology and the Environment 3, 103-109.

Reichard, S., Hayden, C., Hamilton, W., 1997. Predicting invasions of woody plants introduced into North America. Conservation Biology 11(1), 193-203.

Ricciardi, A., 2004. Assessing species invasions as a cause of extinction. Trends in Ecology and Evolution 19, 619.

Richardson, D.M., Holmes, P.M., Esler, K.J., Galatowitsch, S.M., Stromberg, J.C., Kirkman, S.P., Pyšek, P., Hobbs, R.J., 2007. Riparian vegetation: degradation, alien plant invasions, and restoration prospects. Diversity Distributions 13, 126- 139.

Samson, J.C., Yeung, K.L., 1986. Some generalizations on the method of superposed epoch analysis. Planetary and Space Science 34(11), 1133-1142.

Schneider, R.L., Sharitz, R.R., 1986. Seed bank dynamics in a southeastern riverine swamp. American Journal of Botany 73, 1022-1030.

Seager, R., Naik, N., Vogel, L., 2012. Does Global Warming Cause Intensified Interannual Hydroclimate Variability? Journal of Climate 25, 9.

Seymour, R.S., Hunter, M.L. Jr., 1999. Principles of ecological forestry. In: Hunter ML Jr (ed) Maintaining bio-diversity in forested ecosystems. Cambridge University Press, Cambridge, pp 22–61.

SFPC, 2010. Shelby Farms Park Conservancy Master Plan. Accessed 03/12/2012 at URL http://www.shelbyfarmspark.org/.

Shankman, D., 1996. Stream channelization and changing vegetation patterns in the U. S. Coastal Plain. Geographical Review 86, 216-232.

138 Sher, A.A., Marshall, D.L., Gilbert, S.A., 2000. Competition between native Populus deltoids and invasive Tamarix ramosissima and the implications for reestablishing flooding disturbance. Conservation Biology 14, 1744-1754.

Shibata, M., Nakashizuka, T., 1995. Seed and seedling demography of four co-cccurring Carpinus species in a temperate deciduous forest. Ecology 76, 1099-1108.

Simberloff, D., 2000. Global climate change and introduced species in United States forests. Science of the Total Environment 262, 253-261.

Simberloff, D., Von Holle, B., 1999. Positive interactions of nonindigenous species: invasional meltdown? Biological Invasions 1, 21–32.

Simpson, A., Jarnevich, C., Madsen, J., Westbrooks, R., Fournier, C., Mehrhoff, L., Sellers, E., 2009. Invasive species information networks: collaboration at multiple scales for prevention, early detection, and rapid response to invasive alien species. Biodiversity 10(2-3), 5-13.

Smith, E., 2002. BACI design. Encyclopedia of Environmetrics. Edited by El-Shaarawi and Piegorsch. Wiley & Sons, Chichester. 141-148.

Smith, K.E, Runion, G.B., Prior, S.A., Price, A.J., Rogers, H., Torbert, H.A., 2008. Chinese Privet (Ligustrum sinense) in an elevated CO2 Environment. Botanical Resource Journal 1, 43–48.

Smith, R.L., Smith, T.M., 2001. Ecology and Field Biology. (6th ed.) Addison Wesley Longman Inc., NY Speer (2008) Fundamentals of tree ring research. The University of Arizona Press, Tucson.

Speer, J.H., 2008. Fundamentals of Tree Ring Research. The University of Arizona Press, Tucson.

Stahle, D., 1996. Tree rings and ancient forest relics. Arnoldia 1-10.

Stahle, D., 2012. Distinguished Professor, Tree-Ring Laboratory, Department of Geosciences, University of Arkansas, Fayetteville, AR. Personal communication.

139 Stanturf, Conner, Gardiner, Schweitzer, Ezell, 2004. Recognizing and Overcoming Difficult Site Conditions for Afforestation of Bottomland Hardwoods. Ecological Restoratoin 22, 183-193.

Stokes, M.A., Smiley, T.L., 1968, An Introduction to Tree-Ring Dating. University of Chicago Press, Chicago, Illinois, USA.

Stromayer, K.A.K., Warren, R.J., Johnson, A.S., Hale, P.E., Rogers, C.L., Tucker, C.L 1998. Chinese privet and the feeding ecology of white-tailed deer: the role of an exotic plant. Journal of Wildlife Management 62, 1321–1329.

Strong, C., Brown, D., Stouffer, P., 2005. Frugivory by wintering hermit thrush in Louisiana. Southeastern Naturalist 4, 627–638.

Thebaud, C., Debussche, M., 1991. Rapid invasion of Fraxinus ornus L. along the Herault River system in southern France: the importance of seed dispersal by water. Journal of Biogeography 18, 7-12.

Tilman, D., Knops, J., Wedin, D., Reich, P., Ritchie, M, Siemann, E., 1997. The influence of functional diversity and composition on ecosystem processes. Science 277, 1300–1302.

Tomarken, A.J., Serlin, R.C., 1986. Comparison of ANOVA alternatives under variance heterogeneity and specific noncentrality structures. Psychological Bulletin 99(1), 90.

Traveset, A., Richardson, D.M., 2006. Biological invasions as disruptors of plant reproductive mutualisms. Trend in Ecology an Evolution 21, 208–216.

U.S. Environmental Protection Agency, 2011. National Wetland Condition Assessment: Field Operations Manual. EPA-843-R-10-001. U.S. Environmental Protection Agency, Washington, DC.

Ulyshen, M.D., Horn, S., Hanula, J., 2010. Response of beetles (Coleoptera) at three heights to the experimental removal of an invasive shrub, Chinese privet (Ligustrum sinense), from floodplain forests. Biological Invasions 12, 1573–1579.

140 USACE, 2013. United States Army Corps of Engineers Wolf River Gage WF111. Accessed 03/13/2013 at URL http:// www.rivergages.com.

USDA, 2014. The plants database. National plant data center, Greensboro, NC 27401- 4901 USA.

Veraart, A., Romaní, A., Tornés, E., Sabater, S., 2008. Algal responses to nutrient enrichment in forested oligotrophic stream. Journal of Phycology 44, 564–572.

Vitousek, P.M., 1994. Beyond global warming: ecology and global change. Ecology 75, 1861-1876.

Vitousek, P.M., D‟Antonio, C.M., Loope, L., Westbrooks, R., 1996. Biological changes as global environmental change: our mobile society is redistributing the species on the earth at a pace that challenges ecosystems, threatens human health and strains economies. American Scientist 84, 468-478.

Vose, J.M., Swank, W.T., Clinton, B.D., Knoepp, J.D., Swift, L.W., 1999. Using stand replacement fires to restore southern Appalachian pine–hardwood ecosystems: effects on mass, carbon, and nutrient pools. Forest Ecology and Management 114(2), 215-226.

Walther, G.R., Roques, A., Hulme, P.E., Sykes, M.T., Pyšek, P., Kühn, I., Zobel, M., Bacher, S., Botta-Dukát, Z., Bugmann, H., Czúcz, B., Dauber, J., Hickler, T., Jarošík, V., Kenis, M., Klotz, S., Minchin, D., Moora, M., Nentwig, W., Ott, J., Panov, V.E., Reineking, B., Robinet, C., Semenchenko, V., Solarz, W., Thuiller W., Vilà, M., Vohland, K., Settele, J., 2009. Alien species in a warmer world: risks and opportunities. Trends in Ecology and Evolution 24, 686–693.

Wang, H.H., Grant, W.E., 2012. Determinants of Chinese and European Privet (Ligustrum sinense and Ligustrum vulgare) Invasion and Likelihood of Further Invasion in Southern US Forestlands. Invasive Plant Science and Management, 5(4), 454-463.

Ward, R.W., 2002. Extent and dispersal rates of Chinese privet (Ligustrum sinense) invasion on the upper Oconee River floodplain, North Georgia. Southeastern Geographer 1, 29–48.

141 Weins, K., Roberts, T.H., 2003. Effects of headcutting on the bottomland hardwood wetlands adjacent to the Wolf River, Tennessee, ERDC TN-WRP-HS-CP-2.1, U.S. Army Engineer Research and Development Center, Vicksburg, MS.

Westoby, M., Dalby, J., Adams-Acton, L., 1983. Fruit production by two species of privet Ligustrum sinense Lour. and L. lucidum W.T.Ait. in Sydney. Australian Weeds 2, 127-129.

Wilcove, D.S., Rothstein, D., Dubow, J., Phillips, A., Losos, E., 1998. Quantifying threats to imperiled species in the United States. BioScience 48, 607–615.

Wilcox, J., Beck, C.W., 2007. Effects of Ligustrum sinense Lour. (Chinese privet) on abundance and diversity of songbirds and native plants in a southeastern nature preserve. Southeastern Naturalist 6, 535-550.

Williams, P., Karl, B., Bannister, P., Lee, W., 2000. Small mammals as potential seed dispersers in New Zealand. Australian Ecology 25, 523-532.

Wolda, H., 1981. Similarity indices, sample size and diversity. Oecologia 50(3), 296-302.

Young, PJ., Keeland, B.D., Sharitz, R.R., 1995. Growth response of baldcypress [Taxodium distichum (L.) Rich.] to an altered hydrologic regime. American Midland Naturalist 133, 206-212.

Zhang Y, Sun J, Hanula JL (2009) Biology and life history of Argopistes tsekooni (Coleoptera: Chrysomelidae) in China, a promising biological control agent of Chinese privet. Annals of the Entomological Society of America 102, 508–516.

142 Appendix A

Hydrochory Diversity Characteristics

Hydrochory diversity data for all sites including total number of seeds, Simpson‟s diversity, and species richness

Site Year Month Number of seeds Simpson Diversity index Richness INV 2012 September 32 0.179 3 INV 2012 October 356 0.523 11 INV 2012 November 1045 0.688 13 INV 2012 December 712 0.404 18 INV 2013 January 164 0.850 12 INV 2013 February 466 0.834 13 INV 2013 March 1383 0.262 10 INV 2013 April 1080 0.521 19 INV 2013 May 65 0.750 8 INV 2013 June 51 0.836 8 INV 2013 July 623 0.798 13 INV 2013 September 19 0.854 8 INV 2013 October 11 0.691 4 INV 2013 November 5 0.400 2 INV 2014 January 502 0.739 20 INV 2014 February 84 0.425 7 INV 2014 March 1020 0.637 14 NON 2012 September 5 0.800 3 NON 2012 October 16 0.325 2 NON 2012 November 138 0.576 5 NON 2012 December 18 0.582 3 NON 2013 January 286 0.896 15 NON 2013 February 277 0.496 8 NON 2013 March 754 0.090 10 NON 2013 April 509 0.114 9 NON 2013 May 17 0.662 5 NON 2013 June 214 0.209 7 NON 2013 July 43 0.807 9 NON 2013 September 6 0.333 2 NON 2013 October 4 0.500 2 NON 2013 November 2 1.000 1 NON 2014 January 1046 0.825 17 NON 2014 February 111 0.718 10 NON 2014 March 645 0.720 13

143 CNT 2012 September 12 0.848 6 CNT 2012 October 23 0.814 8 CNT 2012 November 39 0.692 6 CNT 2012 December 137 0.808 13 CNT 2013 January 398 0.817 17 CNT 2013 February 662 0.729 13 CNT 2013 March 124 0.064 4 CNT 2013 April 8 0.786 4 CNT 2013 May/June 25 0.363 5 CNT 2013 September 49 0.873 9 CNT 2013 October 6 0.733 3 CNT 2013 November 183 0.779 11 CNT 2014 January 918 0.846 17 CNT 2014 February 47 0.774 6 CNT 2014 March 393 0.740 13

144 Appendix B

Hydrochory Species Relative Abundance Data

Hydrochory relative abundance data for all species in all sites and months, species abbreviations decoded in Appendix D

# Seeds Relative Site Year Month Species per month abundance INV 2012 September ACNE 1 0.031 INV 2012 September LISI 29 0.906 INV 2012 September VISP 2 0.063 INV 2012 October ACNE 15 0.042 INV 2012 October UNK5 1 0.003 INV 2012 October BESC 4 0.011 INV 2012 October BROV 1 0.003 INV 2012 October FRAM 52 0.146 INV 2012 October FRPE 38 0.107 INV 2012 October LISI 237 0.666 INV 2012 October PLOC 4 0.011 INV 2012 October QUSP 1 0.003 INV 2012 October TADI 1 0.003 INV 2012 October VISP 2 0.006 INV 2012 November ACNE 285 0.273 INV 2012 November BROV 18 0.017 INV 2012 November CACA 2 0.002 INV 2012 November FRAM 137 0.131 INV 2012 November FRPE 48 0.046 INV 2012 November FRPR 7 0.007 INV 2012 November FRSP 31 0.030 INV 2012 November LISI 487 0.466 INV 2012 November LITU 2 0.002 INV 2012 November NYSY 4 0.004 INV 2012 November PLOC 2 0.002 INV 2012 November QUSP 10 0.010 INV 2012 November VISP 12 0.011 INV 2012 December ACNE 67 0.094 INV 2012 December UNK5 1 0.001 INV 2012 December BESC 1 0.001 INV 2012 December BROV 14 0.020 INV 2012 December BROV 14 0.020 INV 2012 December CACA 9 0.013

145 INV 2012 December FRAM 9 0.013 INV 2012 December FRPE 4 0.006 INV 2012 December FRSP 2 0.003 INV 2012 December LISI 545 0.765 INV 2012 December LIST 3 0.004 INV 2012 December LITU 2 0.003 INV 2012 December NYSY 1 0.001 INV 2012 December PINUS 2 0.003 INV 2012 December PLOC 20 0.028 INV 2012 December QUSP 7 0.010 INV 2012 December UNK3 3 0.004 INV 2012 December VISP 8 0.011 INV 2013 January ACNE 4 0.024 INV 2013 January BESC 4 0.024 INV 2013 January BROV 11 0.067 INV 2013 January CASP 1 0.006 INV 2013 January FRAM 34 0.207 INV 2013 January FRPE 30 0.183 INV 2013 January FRSP 27 0.165 INV 2013 January LISI 28 0.171 INV 2013 January PINUS 1 0.006 INV 2013 January PLOC 2 0.012 INV 2013 January QUSP 21 0.128 INV 2013 January UNK3 1 0.006 INV 2013 February ACNE 8 0.017 INV 2013 February BROV 34 0.073 INV 2013 February FRAM 44 0.094 INV 2013 February FRPE 37 0.079 INV 2013 February FRSP 37 0.079 INV 2013 February LISI 132 0.283 INV 2013 February LITU 1 0.002 INV 2013 February NYSY 21 0.045 INV 2013 February PLOC 106 0.227 INV 2013 February QUSP 37 0.079 INV 2013 February ROPS 2 0.004 INV 2013 February UNK4 3 0.006 INV 2013 February VISP 4 0.009 INV 2013 March BESC 1 0.001 INV 2013 March BROV 10 0.007 INV 2013 March CACA 1 0.001 INV 2013 March FRAM 13 0.009

146 INV 2013 March FRPE 3 0.002 INV 2013 March FRSP 4 0.003 INV 2013 March LISI 159 0.115 INV 2013 March PINUS 1 0.001 INV 2013 March PLOC 1177 0.851 INV 2013 March QUSP 14 0.010 INV 2013 April ACRU 1 0.001 INV 2013 April ACSA 6 0.006 INV 2013 April BESC 2 0.002 INV 2013 April BROV 38 0.035 INV 2013 April CACA 2 0.002 INV 2013 April CASP 1 0.001 INV 2013 April FRAM 109 0.101 INV 2013 April FRPE 54 0.050 INV 2013 April FRSP 13 0.012 INV 2013 April ILDE 1 0.001 INV 2013 April LISI 19 0.018 INV 2013 April LITU 2 0.002 INV 2013 April NYAQ 5 0.005 INV 2013 April PINUS 2 0.002 INV 2013 April PLOC 734 0.680 INV 2013 April QUSP 58 0.054 INV 2013 April TADI 13 0.012 INV 2013 April ULSP 5 0.005 INV 2013 April VISP 15 0.014 INV 2013 May BROV 11 0.169 INV 2013 May CASP 1 0.015 INV 2013 May FRAM 7 0.108 INV 2013 May FRPE 11 0.169 INV 2013 May PLOC 28 0.431 INV 2013 May QUSP 5 0.077 INV 2013 May VISP 2 0.031 INV 2013 June BROV 8 0.157 INV 2013 June FRAM 6 0.118 INV 2013 June FRPE 9 0.176 INV 2013 June FRPR 1 0.020 INV 2013 June LIST 1 0.020 INV 2013 June PLOC 15 0.294 INV 2013 June QUSP 5 0.098 INV 2013 June VISP 6 0.118 INV 2013 July ACSP 1 0.002

147 INV 2013 July BROV 76 0.122 INV 2013 July CACA 4 0.006 INV 2013 July CASP 5 0.008 INV 2013 July FRAM 139 0.223 INV 2013 July FRPE 100 0.161 INV 2013 July FRSP 53 0.085 INV 2013 July LISI 3 0.005 INV 2013 July LIST 5 0.008 INV 2013 July PLOC 200 0.321 INV 2013 July QUSP 30 0.048 INV 2013 July TADI 4 0.006 INV 2013 July VISP 3 0.005 INV 2013 September ACNE 6 0.316 INV 2013 September ASTR 3 0.158 INV 2013 September BROV 1 0.053 INV 2013 September CACA 2 0.105 INV 2013 September FRSP 1 0.053 INV 2013 September PLOC 4 0.211 INV 2013 September QUSP 1 0.053 INV 2013 September UNK4 1 0.053 INV 2013 October FRAM 1 0.091 INV 2013 October NYAQ 2 0.182 INV 2013 October PLOC 2 0.182 INV 2013 October QUSP 6 0.545 INV 2013 November FRAM 1 0.200 INV 2013 November FRPE 4 0.800 INV 2014 January ACNE 18 0.036 INV 2014 January UNK5 6 0.012 INV 2014 January BESC 1 0.002 INV 2014 January BROV 9 0.018 INV 2014 January CACA 6 0.012 INV 2014 January CASP 2 0.004 INV 2014 January UNK1 1 0.002 INV 2014 January FRAM 88 0.175 INV 2014 January FRPE 223 0.444 INV 2014 January FRPR 5 0.010 INV 2014 January ILDE 2 0.004 INV 2014 January LISI 82 0.163 INV 2014 January NYSY 9 0.018 INV 2014 January PINUS 1 0.002 INV 2014 January PLOC 36 0.072

148 INV 2014 January QUSP 5 0.010 INV 2014 January UNK2 1 0.002 INV 2014 January TADI 2 0.004 INV 2014 January ULSP 1 0.002 INV 2014 January VISP 4 0.008 INV 2014 February CACA 2 0.024 INV 2014 February FRAM 2 0.024 INV 2014 February FRPE 6 0.071 INV 2014 February LISI 63 0.750 INV 2014 February PLOC 9 0.107 INV 2014 February QUSP 1 0.012 INV 2014 February TADI 1 0.012 INV 2014 March ACNE 18 0.018 INV 2014 March BROV 1 0.001 INV 2014 March CACA 2 0.002 INV 2014 March CASP 4 0.004 INV 2014 March FRAM 157 0.154 INV 2014 March FRPE 569 0.558 INV 2014 March FRPR 2 0.002 INV 2014 March LISI 79 0.077 INV 2014 March LITU 2 0.002 INV 2014 March NYSY 3 0.003 INV 2014 March PINUS 2 0.002 INV 2014 March PLOC 149 0.146 INV 2014 March QUSP 23 0.023 INV 2014 March VISP 9 0.009 NON 2012 September CACA 2 0.400 NON 2012 September LISI 1 0.200 NON 2012 September QUSP 2 0.400 NON 2012 October CACA 3 0.188 NON 2012 October QUSP 13 0.813 NON 2012 November CACA 45 0.326 NON 2012 November FRAM 1 0.007 NON 2012 November LISI 1 0.007 NON 2012 November LIST 77 0.558 NON 2012 November QUSP 14 0.101 NON 2012 December CACA 8 0.444 NON 2012 December LIST 9 0.500 NON 2012 December PLOC 1 0.056 NON 2013 January ACNE 12 0.042 NON 2013 January BESC 1 0.004

149 NON 2013 January BROV 17 0.059 NON 2013 January CACA 24 0.084 NON 2013 January CASP 5 0.017 NON 2013 January FRAM 24 0.084 NON 2013 January FRPE 21 0.073 NON 2013 January FRSP 32 0.112 NON 2013 January LISI 26 0.91 NON 2013 January LIST 45 0.157 NON 2013 January NYAQ 2 0.007 NON 2013 January PINUS 3 0.010 NON 2013 January PLOC 51 0.178 NON 2013 January QUSP 20 0.070 NON 2013 January VISP 3 0.010 NON 2013 February CACA 46 0.166 NON 2013 February CASP 1 0.004 NON 2013 February LISI 3 0.011 NON 2013 February LIST 21 0.076 NON 2013 February NYSY 1 0.004 NON 2013 February PLOC 190 0.686 NON 2013 February QUSP 13 0.047 NON 2013 February TADI 2 0.007 NON 2013 March ACRU 1 0.001 NON 2013 March BROV 1 0.001 NON 2013 March CACA 6 0.008 NON 2013 March FAGR 1 0.001 NON 2013 March FRPE 1 0.001 NON 2013 March LIST 2 0.003 NON 2013 March LITU 2 0.003 NON 2013 March PLOC 719 0.954 NON 2013 March QUSP 9 0.012 NON 2013 March TADI 12 0.016 NON 2013 April ACSA 1 0.002 NON 2013 April BROV 1 0.002 NON 2013 April FAGR 2 0.004 NON 2013 April FRPE 7 0.014 NON 2013 April LIST 10 0.020 NON 2013 April PLOC 479 0.941 NON 2013 April QUSP 1 0.002 NON 2013 April TADI 7 0.014 NON 2013 April VISP 1 0.002 NON 2013 May ACSP 1 0.059

150 NON 2013 May ILDE 1 0.059 NON 2013 May PLOC 9 0.529 NON 2013 May QUSP 5 0.294 NON 2013 May VISP 1 0.059 NON 2013 June CASP 2 0.009 NON 2013 June LIST 2 0.009 NON 2013 June PLOC 190 0.888 NON 2013 June QUSP 8 0.037 NON 2013 June CASP 1 0.005 NON 2013 June TADI 10 0.047 NON 2013 June VISP 1 0.005 NON 2013 July BROV 1 0.023 NON 2013 July CASP 3 0.070 NON 2013 July FRAM 3 0.070 NON 2013 July FRPE 10 0.233 NON 2013 July FRSP 6 0.140 NON 2013 July PLOC 1 0.023 NON 2013 July QUSP 15 0.349 NON 2013 July TADI 3 0.070 NON 2013 July VISP 1 0.023 NON 2013 September PLOC 5 0.833 NON 2013 September QUSP 1 0.167 NON 2013 October LIST 3 0.750 NON 2013 October QUSP 1 0.250 NON 2013 November QUSP 2 1.000 NON 2014 January ACNE 34 0.032 NON 2014 January BROV 2 0.002 NON 2014 January CACA 153 0.143 NON 2014 January CASP 1 0.001 NON 2014 January FRAM 67 0.064 NON 2014 January FRPE 14 0.013 NON 2014 January ILDE 6 0.006 NON 2014 January LISI 8 0.008 NON 2014 January LIST 219 0.209 NON 2014 January LITU 29 0.028 NON 2014 January NYAQ 16 0.015 NON 2014 January NYSY 5 0.005 NON 2014 January PINUS 2 0.002 NON 2014 January PLOC 314 0.300 NON 2014 January QUSP 68 0.065 NON 2014 January TADI 105 0.100

151 NON 2014 January VISP 3 0.003 NON 2014 February ACNE 2 0.018 NON 2014 February CACA 12 0.108 NON 2014 February FRAM 6 0.054 NON 2014 February ILDE 1 0.009 NON 2014 February LISI 1 0.009 NON 2014 February LIST 22 0.198 NON 2014 February NYAQ 2 0.018 NON 2014 February PLOC 53 0.477 NON 2014 February QUSP 5 0.045 NON 2014 February TADI 7 0.063 NON 2014 March ACNE 13 0.020 NON 2014 March BESC 1 0.002 NON 2014 March CACA 30 0.047 NON 2014 March CASP 1 0.002 NON 2014 March FRAM 102 0.158 NON 2014 March FRPE 5 0.008 NON 2014 March LISI 1 0.002 NON 2014 March LIST 30 0.047 NON 2014 March NYAQ 9 0.014 NON 2014 March NYSY 1 0.002 NON 2014 March PLOC 278 0.431 NON 2014 March QUSP 22 0.034 NON 2014 March TADI 152 0.236 CNT 2012 September QUSP 1 0.083 CNT 2012 September CASP 1 0.083 CNT 2012 September FRPE 4 0.333 CNT 2012 September NYAQ 3 0.250 CNT 2012 September TADI 1 0.083 CNT 2012 September VISP 2 0.167 CNT 2012 October CACA 3 0.130 CNT 2012 October FRPE 1 0.043 CNT 2012 October ILDE 4 0.174 CNT 2012 October NYAQ 1 0.043 CNT 2012 October NYSY 2 0.087 CNT 2012 October QUSP 9 0.391 CNT 2012 October TADI 1 0.043 CNT 2012 October VISP 2 0.087 CNT 2012 November ACNE 3 0.077 CNT 2012 November CACA 12 0.308 CNT 2012 November ILDE 1 0.026

152 CNT 2012 November LIST 18 0.462 CNT 2012 November LITU 4 0.103 CNT 2012 November QUSP 1 0.026 CNT 2012 December ACNE 37 0.270 CNT 2012 December BROV 15 0.109 CNT 2012 December CACA 2 0.015 CNT 2012 December CASP 2 0.015 CNT 2012 December ILDE 1 0.007 CNT 2012 December LISI 1 0.007 CNT 2012 December LIST 1 0.007 CNT 2012 December LITU 37 0.270 CNT 2012 December NYAQ 3 0.022 CNT 2012 December NYSY 1 0.007 CNT 2012 December PLOC 25 0.182 CNT 2012 December QUSP 10 0.073 CNT 2012 December TADI 2 0.015 CNT 2013 January ACNE 65 0.163 CNT 2013 January BESC 3 0.008 CNT 2013 January BROV 5 0.013 CNT 2013 January CACA 99 0.249 CNT 2013 January CASP 13 0.033 CNT 2013 January LISI 3 0.008 CNT 2013 January FAGR 6 0.015 CNT 2013 January FRAM 3 0.008 CNT 2013 January FRPE 2 0.005 CNT 2013 January FRSP 2 0.005 CNT 2013 January ILDE 1 0.003 CNT 2013 January LITU 62 0.156 CNT 2013 January NYAQ 1 0.003 CNT 2013 January PLOC 22 0.055 CNT 2013 January QUSP 7 0.018 CNT 2013 January TADI 103 0.259 CNT 2013 January VISP 1 0.003 CNT 2013 February ACNE 16 0.028 CNT 2013 February UNK5 13 0.023 CNT 2013 February CACA 33 0.058 CNT 2013 February CASP 2 0.003 CNT 2013 February FAGR 4 0.007 CNT 2013 February FRSP 8 0.014 CNT 2013 February ILDE 12 0.021 CNT 2013 February LISI 1 0.002

153 CNT 2013 February LIST 151 0.264 CNT 2013 February LITU 10 0.017 CNT 2013 February NYAQ 2 0.003 CNT 2013 February PLOC 288 0.503 CNT 2013 February QUSP 32 0.056 CNT 2013 March BROV 2 0.016 CNT 2013 March CASP 1 0.008 CNT 2013 March LISI 1 0.008 CNT 2013 March PLOC 120 0.968 CNT 2013 April FRAM 1 0.125 CNT 2013 April FRPE 1 0.125 CNT 2013 April PLOC 3 0.375 CNT 2013 April TADI 3 0.375 CNT 2013 May/June ACSP 2 0.087 CNT 2013 May/June BROV 1 0.043 CNT 2013 May/June CASP 1 0.043 CNT 2013 May/June FRAM 1 0.043 CNT 2013 May/June PLOC 20 0.870 CNT 2013 September ACSP 4 0.082 CNT 2013 September BROV 4 0.082 CNT 2013 September CACA 6 0.122 CNT 2013 September FRAM 3 0.061 CNT 2013 September FRPE 12 0.245 CNT 2013 September LIST 1 0.020 CNT 2013 September PLOC 5 0.102 CNT 2013 September QUSP 8 0.163 CNT 2013 September TADI 6 0.122 CNT 2013 October NYAQ 1 0.167 CNT 2013 October FRPE 3 0.500 CNT 2013 October LIST 2 0.333 CNT 2013 November ACNE 10 0.055 CNT 2013 November BROV 1 0.005 CNT 2013 November CACA 72 0.393 CNT 2013 November FAGR 2 0.011 CNT 2013 November FRPE 15 0.082 CNT 2013 November ILDE 4 0.022 CNT 2013 November LIST 10 0.055 CNT 2013 November LITU 28 0.153 CNT 2013 November PLOC 3 0.016 CNT 2013 November QUSP 5 0.027 CNT 2013 November TADI 33 0.180

154 CNT 2014 January ACNE 53 0.058 CNT 2014 January BROV 6 0.007 CNT 2014 January CACA 143 0.156 CNT 2014 January FAGR 2 0.002 CNT 2014 January FRAM 63 0.069 CNT 2014 January FRPE 49 0.053 CNT 2014 January FRSP 2 0.002 CNT 2014 January LISI 1 0.001 CNT 2014 January LIST 117 0.127 CNT 2014 January LITU 30 0.033 CNT 2014 January NYAQ 2 0.002 CNT 2014 January NYSY 35 0.038 CNT 2014 January PLOC 272 0.296 CNT 2014 January QUSP 38 0.041 CNT 2014 January TADI 97 0.106 CNT 2014 January VISP 7 0.008 CNT 2014 January ULSP 1 0.001 CNT 2014 February ACNE 2 0.043 CNT 2014 February FAGR 1 0.021 CNT 2014 February LIST 7 0.149 CNT 2014 February LITU 9 0.191 CNT 2014 February PLOC 12 0.255 CNT 2014 February TADI 16 0.340 CNT 2014 March ACNE 12 0.031 CNT 2014 March BROV 3 0.008 CNT 2014 March CACA 42 0.107 CNT 2014 March FAGR 1 0.003 CNT 2014 March FRAM 23 0.059 CNT 2014 March FRPE 4 0.010 CNT 2014 March LIST 9 0.023 CNT 2014 March LITU 22 0.056 CNT 2014 March NYAQ 1 0.003 CNT 2014 March NYSY 2 0.005 CNT 2014 March PLOC 163 0.415 CNT 2014 March QUSP 7 0.018 CNT 2014 March TADI 104 0.265

155 Appendix C

Site Characteristics

Presence (1), absence (0), and similarity data of woody species (species abbreviations decoded in Appendix D)

Species INV NON CNT NON & NON & INV & INV CNT CNT shared shared shared ACNE 1 1 1 1 1 1 ACRU 0 1 1 0 1 0 ACSA 1 1 0 1 0 0 ARSP 1 0 0 0 0 0 ASTR 1 1 0 1 0 0 BESC 1 1 0 1 0 0 BICA 1 1 0 1 0 0 CAAL 1 1 1 1 1 1 CACA 1 1 1 1 1 1 CAIL 0 1 0 0 0 0 CALA 1 1 1 1 1 1 CAOV 1 1 0 1 0 0 CARA 1 1 1 1 1 1 CASP 1 0 0 0 0 0 CELA 1 1 0 1 0 0 CEOC 0 1 0 0 0 0 COCA 1 0 0 0 0 0 CRFL 0 1 0 0 0 0 CRMA 0 1 0 0 0 0 DIVI 0 1 0 0 0 0 EUSP 1 0 0 0 0 0 FAGR 0 0 1 0 0 0 FOAC 0 1 0 0 0 0 FRPE 1 1 1 1 1 1 FRSP 1 1 0 1 0 0 ILDE 1 1 1 1 1 1 ILOP 1 0 1 0 0 1 LISI 1 1 1 1 1 1 LIST 1 1 1 1 1 1 LITU 1 0 1 0 0 1 LOJA 1 1 1 1 1 1 MAGR 1 0 0 0 0 0 MOAL 1 1 0 1 0 0 MORU 1 1 0 1 0 0

156 MOSP 1 1 1 1 1 1 NADO 1 0 0 0 0 0 NYAQ 0 1 1 0 1 0 NYSY 1 1 1 1 1 1 OSVA 0 0 1 0 0 0 PAQU 1 1 1 1 1 1 PLOC 1 1 1 1 1 1 POHE 0 1 0 0 0 0 PRSE 0 1 0 0 0 0 QULY 1 1 1 1 1 1 QUMI 1 1 1 1 1 1 QUNI 1 1 1 1 1 1 QUPA 1 1 1 1 1 1 QUPH 0 1 1 0 1 0 ROPS 1 0 1 0 0 0 RUSP 0 1 0 0 0 1 SAAL 1 1 0 1 0 0 SACA 1 1 0 1 0 0 STAM 0 1 0 0 0 0 TADI 1 0 1 0 0 1 TORA 1 1 1 1 1 1 ULAL 0 0 1 0 0 0 ULAM 1 1 1 1 1 1 ULRU 1 0 0 0 0 0 ULSP 1 1 1 1 1 1 VICI 1 1 1 1 1 1 VIRO 1 1 1 1 1 1 VIRU 1 0 0 0 0 0 VISP 1 1 1 1 1 1 WISP 0 1 0 0 0 0 TOTAL 47 49 34 35 27 28

157 Appendix D

Species Abbreviations

Species Abbreviation Species Name Species MAGR Magnolia grandiflora Abbreviation Species Name MORU Morus rubra ACNE Acer negundo MOSP Morus spp. ACRU Acer rubrum NADO Nandina domestica ACSA Acer Saccharinum NYAQ Nyssa aquatica ARSP Aralia spinosa NYSY Nyssa sylvatica ACSP Acer spp. OSVA Ostrya virginiana ASTR Asimina triloba PAQU Pinus spp. BESC Berchemia scandens PLOC Platanus occidentalis BICA Bignonia capreolata POHE Populus heterophylla BROV Brunnichia ovata PRSE Prunus serotina CAAL Carya alba QULY Quercus lyrata CACA Carpinus caroliniana QUMI Quercus michauxii CAIL Carya illinoinensis QUNI Quercus nigra CALA Carya laciniosa QUPA Quercus pagoda CAOV Carya ovata QUPH Quercus phellos CARA Campsis radicans QUSP Quercus spp. CASP Carya spp. ROPS Robinia pseudoacacia CELA Celtis laevigata RUSP Rubus spp. Cephalanthus SAAL Sassafras albidum CEOC occidentalis SACA Sambucus canadensis COCA Cocculus caroliniana STAM Styrax aamericana CRFL Crataegus flava TADI Taxodium distichum CRMA Crataegus marshallii Toxicodendron DIVI Diospyros virginiana TORA radicans EUSP Euonymus spp. ULAL Ulmus alata FAGR Fagus grandifolia ULAM Ulmus americana FOAC Fraxinus americana ULRU Ulmus rubra FRPE Fraxinus pennsylvanica ULSP Ulmus spp. FRPR Fraxinus profunda VICI Vitis cineria FRSP Fraxinus spp. VIRO Vitis rotundifolia ILDE Ilex decidua VIRU Viburnum rufidulum ILOP Ilex opaca VISP Vitis spp. LISI Ligustrum sinense WISP Wisteria spp. LIST Liquidambar styraciflua LITU Liriodendron tulipifera LOJA Lonicera japonica

158