Twentieth Century Changes in the Climate Response of Yellow Pines in Great Smoky Mountains National Park, Tennessee, U.S.A

University of Tennessee, Knoxville TRACE: Tennessee Research and Creative Exchange

Masters Theses Graduate School

8-2009

Twentieth Century Changes in the Climate Response of Yellow Pines in Great Smoky Mountains National Park, Tennessee, U.S.A.

Christine Patricia Biermann University of Tennessee - Knoxville

Follow this and additional works at: https://trace.tennessee.edu/utk_gradthes

Part of the Arts and Humanities Commons

Recommended Citation Biermann, Christine Patricia, "Twentieth Century Changes in the Climate Response of Yellow Pines in Great Smoky Mountains National Park, Tennessee, U.S.A.. " Master's Thesis, University of Tennessee, 2009. https://trace.tennessee.edu/utk_gradthes/18

This Thesis is brought to you for free and open access by the Graduate School at TRACE: Tennessee Research and Creative Exchange. It has been accepted for inclusion in Masters Theses by an authorized administrator of TRACE: Tennessee Research and Creative Exchange. For more information, please contact [email protected]. To the Graduate Council:

I am submitting herewith a thesis written by Christine Patricia Biermann entitled "Twentieth Century Changes in the Climate Response of Yellow Pines in Great Smoky Mountains National Park, Tennessee, U.S.A.." I have examined the final electronic copy of this thesis for form and content and recommend that it be accepted in partial fulfillment of the equirr ements for the degree of Master of Science, with a major in Geography.

Henri D. Grissino-Mayer, Major Professor

We have read this thesis and recommend its acceptance:

Sally Horn, Carol Harden

Accepted for the Council: Carolyn R. Hodges

Vice Provost and Dean of the Graduate School

(Original signatures are on file with official studentecor r ds.) To the Graduate Council:

I am submitting herewith a thesis written by Christine Patricia Biermann entitled “Twentieth Century Changes in the Climate Response of Yellow Pines in Great Smoky Mountains National Park, Tennessee, U.S.A.” I have examined the final electronic copy of this thesis for form and content and recommend that it be accepted in partial fulfillment of the requirements for the degree of Master of Science with a major in Geography.

Henri D. Grissino-Mayer, Major Professor

We have read this thesis and recommend its acceptance:

Sally Horn

Carol Harden

Accepted for the Council:

Carolyn R. Hodges Vice Provost and Dean of the Graduate School

(Original signatures are on file with official student records.)

TWENTIETH CENTURY CHANGES IN THE CLIMATE RESPONSE OF YELLOW PINES IN GREAT SMOKY MOUNTAINS NATIONAL PARK, TENNESSEE, U.S.A.

A Thesis Presented

for the Master of Science Degree

The University of Tennessee, Knoxville

Christine Patricia Biermann

ACKNOWLEDGEMENTS

This thesis would not have been possible without the help and support of my advisor, committee members, colleagues, friends, and family. I want to thank my advisor Dr. Henri D. Grissino-Mayer for his guidance and enthusiasm. His passion for dendrochronology is contagious. I also thank my committee members Drs. Sally Horn and Carol Harden for their support, suggestions, and willingness to help. Working in the Laboratory of Tree-Ring Science has been a pleasure, and I am grateful to my fellow graduate students for inspiring me, motivating me, helping me, and sometimes distracting me. Many thanks to Lisa LaForest, John Sakulich, Mark Spond, Grant Harley, Ian Feathers, Saskia van de Gevel, Nancy Li, and Monica Rother. I also want to thank Runy Muñoz and Brad Yarger for assistance with processing and measuring samples. My field work would not have been possible without the help of my advisor and fellow students. Thank you to James Baginski, Henri D. Grissino-Mayer, John Sakulich, Monica Rother, Matthew Kookogey, Nancy Li, Brad Yarger, Josh Brown, Ian Feathers, Mark Spond, and Philip White. Also, many thanks to Lisa LaForest and Jessica Slayton for developing and sharing their yellow pine chronology at Gold Mine Trail. Finally, I want to thank my family and friends for their love and support. Thank you to my parents, for instilling in me a love of learning. Thank you to my sisters, for giving me footsteps to follow in and big shoes to fill. Thank you to James Baginski, for encouraging me and helping me with every step of this thesis, from field work to lab work to cartography. I could not have done it without you, and I wouldn’t have wanted to, either.

I am grateful for the fellowship support from the National Science Foundation. This material is based upon work supported under a National Science Foundation Graduate Student Fellowship. Any opinions, findings, conclusions or recommendations expressed in this publication are those of the author and do not necessarily reflect the views of the National Science Foundation.

iii

ABSTRACT

Previous dendroclimatological research has shown that tree growth is primarily a function of temperature and precipitation. At mid-latitude temperate forest sites, trees have been found to be mainly moisture-sensitive rather than temperature-sensitive.

Researchers at the 2007 North American Dendroecological Fieldweek were surprised to find a winter temperature signal in a shortleaf pine (Pinus echinata) chronology from

Great Smoky Mountains National Park, Tennessee. Building on this finding, I evaluated yellow pine climate-tree growth relationships at five sites on the western end of Great

Smoky Mountains National Park using correlation, response function, moving correlation, and wavelet analyses. Winter mean minimum temperatures influenced yellow pine growth at all five sites, but spring precipitation and growing season moisture conditions also affected growth. Growth was positively associated with Atlantic Ocean sea surface temperature anomalies (SSTAs) and North Atlantic Ocean (NAO) index values, suggesting that positive phases of the both the Atlantic Multidecadal Oscillation

(AMO) and the NAO lead to above average annual tree growth. Pacific Ocean climate variability did not have a strong influence on yellow pine growth in Great Smoky

Mountains National Park.

Climate-growth relationships were temporally unstable at all of the five sites. In the mid-20th century, the response to growing season precipitation and moisture conditions weakened. Simultaneously, the response to winter and fall mean minimum temperatures strengthened. The shift may have been caused by an AMO phase change, age-dependent climate responses, changes in phenology, decreased drought frequency, data quality, or atmospheric pollution. Because the relationship with temperature

iv strengthened since 1950, yellow pines in Great Smoky Mountains National Park do not show evidence of divergence between temperature and tree growth. Still, this network of chronologies is not ideal for climate reconstruction because the climate-growth relationships were unstable over time. In the future, climate analysis of other chronologies from the southeastern U.S. will be necessary to determine whether the mid-

20th century shift in climate response at Great Smoky Mountains National Park occurred throughout the region.

TABLE OF CONTENTS

1. INTRODUCTION TO DENDROCLIMATOLOGY AND DIVERGENCE...... 1 1.1 Background...... 1 1.2 The Divergence Problem ...... 2 1.2.1 Overview...... 2 1.2.2 Possible Causes...... 4 1.2.3 Implications...... 6 1.3 Purpose of Research...... 8 1.4 Organization of Thesis...... 10

2. LITERATURE REVIEW OF TEMPERATURE RECONSTRUCTIONS AND THE DIVERGENCE PROBLEM ...... 12 2.1 Tree-Ring Based Temperature Reconstructions ...... 12 2.1.1 Mann et al. (1998)...... 13 2.1.2 McIntyre and McKitrick (2005)...... 15 2.1.3 Moberg et al. (2005) ...... 16 2.1.4 Esper and Frank (2004)...... 18 2.2 Issues of Consistency and Divergence...... 20 2.2.1 Briffa et al. (1998) ...... 20 2.2.2 Barber et al. (2000)...... 21 2.2.3 Driscoll et al. (2005)...... 22 2.2.4 Carrer and Urbinati (2006)...... 23 2.2.5 Frank et al. (2007)...... 24 2.2.6 Oberhuber et al. (2008)...... 26 2.2.7 D’Arrigo et al. (2008)...... 28

3. SITE DESCRIPTION OF GREAT SMOKY MOUNTAINS NATIONAL PARK...... 30 3.1 General Setting...... 30 3.1.1 Geology...... 30 3.1.2 Soils ...... 32 3.1.3 Climate...... 33 3.1.3.1 North Atlantic Oscillation...... 34 3.1.3.2 Atlantic Multidecadal Oscillation...... 35 3.1.3.3 El Niño-Southern Oscillation...... 36 3.1.3.4 Pacific Decadal Oscillation...... 36 3.1.4 Vegetation...... 37 3.1.5 Land Use History ...... 39 3.2 Microsite Characteristics ...... 41 3.2.1 Gold Mine Trail ...... 41 3.2.2 Cooper Road Trail...... 43 3.2.3 Shaw Grave Gap ...... 44 3.2.4 Maynard Creek...... 44 3.2.5 Gregory Ridge Trail...... 45

4. METHODS ...... 46 4.1 Field Methods ...... 46 4.2 Laboratory Methods...... 48 4.2.1 Processing and Preparation...... 48 4.2.2 Crossdating ...... 49 4.2.3 Standardization ...... 50 4.3 Climate Analysis...... 52 4.3.1 Climate Data ...... 52 4.3.2 DENDROCLIM2002...... 54 4.3.3 Correlation Analysis ...... 54 4.3.4 Response Function Analysis...... 55 4.3.5 Moving Correlation Analysis...... 56 4.3.6 Wavelet Analysis ...... 57 4.3.7 Composite Chronology ...... 58

5. RESULTS OF CLIMATE ANALYSIS ...... 59 5.1 Chronology Construction...... 59 5.2 Gold Mine Trail ...... 61 5.2.1 Temperature, Precipitation, and PDSI ...... 61 5.2.2 Moving Correlation Analysis...... 66 5.2.3 Ocean-Atmospheric Teleconnections ...... 70 5.2.4 Wavelet Analysis ...... 72 5.3 Cooper Road Trail...... 72 5.3.1 Temperature, Precipitation, and PDSI ...... 72 5.3.2 Moving Correlation Analysis...... 75 5.3.3 Ocean-Atmospheric Teleconnections ...... 79 5.3.4 Wavelet Analysis ...... 81 5.4 Shaw Grave Gap ...... 81 5.4.1 Temperature, Precipitation, and PDSI ...... 81 5.4.2 Moving Correlation Analysis...... 84 5.4.3 Ocean-Atmospheric Teleconnections ...... 87 5.4.4 Wavelet Analysis ...... 90 5.5 Maynard Creek...... 90 5.5.1 Temperature, Precipitation, and PDSI ...... 90 5.5.2 Moving Correlation Analysis...... 93 5.5.3 Ocean-Atmospheric Teleconnections ...... 96 5.5.4 Wavelet Analysis ...... 96 5.6 Gregory Ridge Trail...... 100 5.6.1 Temperature, Precipitation, and PDSI ...... 100 5.6.2 Moving Correlation Analysis...... 102 5.6.3 Ocean-Atmospheric Teleconnections ...... 106 5.6.4 Wavelet Analysis ...... 106 5.7 Composite Chronology ...... 109 5.7.1 Temperature, Precipitation, and PDSI ...... 109 5.7.2 Moving Correlation Analysis...... 109 5.7.3 Ocean-Atmospheric Teleconnections ...... 111

vii

5.7.4 Wavelet Analysis ...... 116

6. DISCUSSION OF YELLOW PINE CLIMATE-GROWTH RELATIONSHIPS...... 118 6.1 Growth Response to Climate ...... 118 6.1.1 Response to Temperature...... 118 6.1.2 Responses to Precipitation and PDSI...... 121 6.1.3 Differences in Climate Response Between Sites...... 122 6.1.4 Relationships to Ocean-Atmospheric Teleconnections ...... 125 6.2 Temporal Consistency of Climate-Growth Relationships ...... 128 6.3 Evidence of Divergence...... 133 6.4 Potential for Reconstruction ...... 134

7. CONCLUSIONS AND FUTURE RESEARCH ...... 136 7.1 Major Conclusions...... 137 7.2 Future Research ...... 141

REFERENCES ...... 144 APPENDICES ...... 156 APPENDIX A1. Gold Mine Trail COFECHA Summary Statistics...... 157 APPENDIX A2. Cooper Road Trail COFECHA Summary Statistics ...... 163 APPENDIX A3. Shaw Grave Gap COFECHA Summary Statistics...... 165 APPENDIX A4. Maynard Creek Trail COFECHA Summary Statistics ...... 168 APPENDIX A5. Gregory Ridge Trail COFECHA Summary Statistics...... 169 VITA...... 171

viii

LIST OF TABLES

Table 3.1 Location of study sites ...... 42 Table 4.1 List of chronologies ...... 47 Table 5.1 Summary statistics of chronologies ...... 60 Table 5.2 Summary data of correlation analysis...... 63

LIST OF FIGURES

Figure 2.1 Hockey stick reconstruction by Mann et al. (1998) ...... 14 Figure 3.1 Map of the five sites in Great Smoky Mountains National Park...... 42 Figure 4.1 Sampling at Gregory Ridge Trail ...... 47 Figure 4.2 Standardization of raw ring-width data ...... 51 Figure 5.1 Standard chronologies at each site ...... 62 Figure 5.2 Correlation analysis at Gold Mine Trail...... 65 Figure 5.3 Response function analysis at Gold Mine Trail ...... 65 Figure 5.4 Moving analysis of temperature at Gold Mine Trail...... 67 Figure 5.5 Moving analysis of precipitation at Gold Mine Trail...... 68 Figure 5.6 Moving analysis of PDSI at Gold Mine Trail ...... 69 Figure 5.7 Ocean-atmospheric teleconnections at Gold Mine Trail...... 71 Figure 5.8 Wavelet analysis at Gold Mine Trail...... 73 Figure 5.9 Correlation analysis at Cooper Road Trail ...... 74 Figure 5.10 Response function analysis at Cooper Road Trail...... 74 Figure 5.11 Moving analysis of temperature at Cooper Road Trail ...... 76 Figure 5.12 Moving analysis of precipitation at Cooper Road Trail ...... 77 Figure 5.13 Moving analysis of PDSI at Cooper Road Trail...... 78 Figure 5.14 Ocean-atmospheric teleconnections at Cooper Road Trail ...... 80 Figure 5.15 Wavelet analysis at Cooper Road Trail...... 82 Figure 5.16 Correlation analysis at Shaw Grave Gap...... 83 Figure 5.17 Response function analysis at Shaw Grave Gap ...... 83 Figure 5.18 Moving analysis of temperature at Shaw Grave Gap...... 85 Figure 5.19 Moving analysis of precipitation at Shaw Grave Gap...... 86 Figure 5.20 Moving analysis of PDSI at Shaw Grave Gap ...... 88 Figure 5.21 Ocean-atmospheric teleconnections at Shaw Grave Gap...... 89 Figure 5.22 Wavelet analysis at Shaw Grave Gap...... 91 Figure 5.23 Correlation analysis at Maynard Creek ...... 92 Figure 5.24 Response function analysis at Maynard Creek...... 92 Figure 5.25 Moving analysis of temperature at Maynard Creek ...... 94 Figure 5.26 Moving analysis of precipitation at Maynard Creek ...... 95 Figure 5.27 Moving analysis of PDSI at Maynard Creek...... 97 Figure 5.28 Ocean-atmospheric teleconnections at Maynard Creek ...... 98 Figure 5.29 Wavelet analysis at Maynard Creek...... 99 Figure 5.30 Correlation analysis at Gregory Ridge Trail ...... 101 Figure 5.31 Response function analysis at Gregory Ridge Trail...... 101 Figure 5.32 Moving analysis of temperature at Gregory Ridge Trail...... 103 Figure 5.33 Moving analysis of precipitation at Gregory Ridge Trail ...... 104 Figure 5.34 Moving analysis of PDSI at Gregory Ridge Trail...... 105 Figure 5.35 Ocean-atmospheric teleconnections at Gregory Ridge Trail...... 107 Figure 5.36 Wavelet analysis at Gregory Ridge Trail ...... 108 Figure 5.37 Correlation analysis of composite chronology...... 110 Figure 5.38 Response function analysis of composite chronology...... 110 Figure 5.39 Moving analysis of temperature in the composite chronology ...... 112 Figure 5.40 Moving analysis of precipitation in the composite chronology ...... 113

Figure 5.41 Moving analysis of PDSI in the composite chronology...... 114 Figure 5.42 Ocean-atmospheric teleconnections in the composite chronology ...... 115 Figure 5.43 Wavelet analysis of composite chronology...... 117

CHAPTER 1

INTRODUCTION TO DENDROCLIMATOLOGY AND DIVERGENCE

1.1 Background

Climate change, whether caused by natural or anthropogenic factors, has been a major focus within the scientific community in recent years. To determine whether contemporary climate change is anomalous or within the historical range of variability, we must first understand how climate functioned in the past. At best, high-quality instrumental climate records are available for the past century. Long-term warming or cooling trends (i.e., century timescale) and low-frequency climate oscillations, therefore, are difficult to investigate using only instrumental data.

Proxy data sources such as tree rings and lake sediments allow us to reconstruct climate over centennial to millennial time scales. Tree rings are a particularly good source of proxy data because tree growth is largely a function of climate variables such as temperature and precipitation (Fritts 1976). For example, in the high latitudes and at high elevations, tree growth is most often limited by temperature (LaMarche and Stockton

1974; Buntgen et al. 2005; Frank et al. 2005; Frank et al. 2007). A warmer than average growing season generally means more tree growth, and thus, a wider than average annual ring. Warm growing seasons have also been shown to lead to increased density of the latewood in the annual ring in boreal and subalpine forests (Fritts 1976; Jacoby et al.

1988; Briffa et al. 1992; D’Arrigo et al. 1992; Park and Telewski 1993). In lower latitudes and elevations, precipitation is generally the most significant climate variable for tree growth (Fritts 1976; Cook and Jacoby 1979; Cook et al. 1999).

The technique of crossdating allows for patterns of wide and narrow rings (or high and low latewood densities) to be matched within and between stands of trees (Fritts

1976; Stokes and Smiley 1996). Tree-ring records, or chronologies, can thus be extended beyond the lifespan of an individual tree through overlapping sequences of tree rings.

Because annual tree growth is associated with climate, past temperature and precipitation can be reconstructed using regression analysis based on standardized ring widths or density measurements (Fritts 1976; Briffa et al. 1990).

Tree-ring based temperature reconstructions indicate that global climate has been dynamic throughout the past millennium. In the Northern Hemisphere, the Medieval

Warm Period occurred between A.D. 1000 and 1200, followed by a cooling event from

A.D. 1600 to 1800, termed the Little Ice Age (Briffa et al. 1990; Mann et al. 1998;

Moberg et al. 2005; Esper and Frank 2004). Climate change itself is not unique to the

20th and 21st centuries, but not all climate reconstructions show the same trends. For example, a temperature reconstruction by Moberg et al. (2005) showed mid-20th century temperatures similar to temperatures that occurred between A.D. 1000 and 1100, the height of the Medieval Warm Period. This finding contradicted Mann et al. (1998), as well as the 2001 Intergovernmental Panel on Climate Change (IPCC) report, both of which stated that 20th century warming was unprecedented and outside the historical range of variability.

1.2 The Divergence Problem 1.2.1 Overview Disagreement between temperature reconstructions and a lack of clarity regarding statistical techniques have spurred criticism of dendroclimatology (McIntyre and

McKitrick 2005a, 2005b). Within the dendroclimatological community, however, the recent discovery of a “divergence problem” is even more problematic (D’Arrigo et al.

2008). First identified by Briffa et al. (1998a), divergence between climate and tree growth has since been found in several other dendroclimatological studies (Briffa et al.

1998b; Barber et al. 2000; Driscoll et al. 2005; Carrer and Urbinati 2006; Oberhuber et al. 2008). The divergence problem is defined as a weakening of the temperature-tree growth relationship in recent decades, particularly within sites at high northern latitudes

(D’Arrigo et al. 2008). In other words, divergence occurs between the instrumental temperature records and the predicted temperatures derived from tree-ring chronologies.

Many stands that previously responded to temperature are no longer temperature- sensitive. At some sites, the weakened temperature signal is coupled with greater drought sensitivity (Driscoll et al. 2005; Barber et al. 2000), but this is not true in all cases.

Briffa et al. (1998a) first noted divergence between temperature and tree growth in a dataset of ring-width and maximum latewood density series from over 300 sites in northern North America and northern Europe. Ring-width series are standardized measurements of the width of annual tree rings, while maximum latewood density refers to the density of the wood put on in the late growing season. They compared correlations between instrumental temperature records and tree growth between 1881–1960 and

1881–1980. When including the most recent data (1961–1980), correlations significantly decreased. Briffa et al. (1998a) identified divergence in both maximum latewood density and ring-width datasets, but the density-based reconstructions were overall better estimates of recent temperatures. Oberhuber et al. (2008) also found evidence of

3 divergence in stone pine (Pinus cembra L.) chronologies from the Austrian Alps. The relationship between tree growth and the historically most significant climate variables

(June and July temperature) began to diminish around 1980. This finding suggested that divergence is also problem at high-elevation, temperature-sensitive sites outside the high latitudes.

White spruce (Picea glauca (Moench) Voss) ring-width chronologies from

Alaska also exhibited changing sensitivities over the last half-century (Driscoll et al.

2005). However, different stands of trees and even different trees within the same stand exhibited unique responses to the warmer, longer growing seasons. Prior to 1950, all of the trees in the study had similar responses to climate, but after 1950, correlations between separate stands and correlations between growth and climate both deteriorated.

This indicates that dendrochronologists can no longer assume a coherent climate response across an entire study site (Wilmking et al. 2005).

1.2.2 Possible Causes

Several theories have been proposed to explain why temperature-sensitive trees have recently stopped responding to increasing temperatures. Some examples of divergence at high latitudes have been attributed to temperature-induced drought stress

(Barber et al. 2000; Davi et al. 2003; Driscoll et al. 2005; D’Arrigo et al. 2004). As the climate warms and the growing season lengthens, soil moisture is depleted and drought, rather than temperature, may begin to limit tree growth. Evidence of this is seen in strengthened correlations between tree growth and growing season precipitation over the past few decades, coupled with weakened correlations between growth and temperature

(D’Arrigo et al. 2004; Driscoll et al. 2005). Also, earlier snowmelt may reduce late growing season soil moisture, thus leading to a stronger response to precipitation

(D’Arrigo et al. 2008). Critics of this hypothesis, however, argue that drought-sensitivity is an inadequate, site-specific explanation for such a broad-scale hemispheric phenomenon (Briffa et al. 1998a; D’Arrigo et al. 2008).

Other potential causes of divergence include differing responses to mean, maximum, and minimum annual/seasonal temperatures, global dimming, increased competition due to more favorable growing conditions, carbon dioxide fertilization, and statistical end effects (including standardization, detrending, and sampling bias) (Briffa et al. 1998a; D’Arrigo et al. 2008). When reconstructing temperature, dendroclimatologists must choose the season and temperature variables (minimum, maximum, mean) that are most strongly and consistently related to tree growth. Selecting maximum summer temperature, for example, may produce a divergent trend, while mean annual temperature may show no divergence problem.

Global dimming is another possible explanation for the divergence problem

(D’Arrigo et al. 2008). Since approximately 1960, the amount of solar radiation reaching the Earth’s surface has decreased due to atmospheric pollution (Abakumova et al. 1996).

Therefore, despite the increase in temperature, less solar radiation may reduce tree growth rates. Global dimming may essentially be masking the temperature signal, particularly in high northern latitudes (D’Arrigo et al. 2008). An advantage of this hypothesis is that it provides a global-scale phenomenon to explain widespread divergence. Still other dendroclimatologists point to increased competition in more favorable growing conditions (Briffa et al. 1998a) and statistical end effects (D’Arrigo et

5 al. 2008) as explanations for divergence. However, these hypotheses also fail to provide a global explanation for the sudden loss of temperature sensitivity.

1.2.3 Implications

The divergence problem calls into question the validity of tree-ring based temperature reconstructions. If climate-tree growth relationships change over time, it is inappropriate to assume that the factors driving tree growth today also controlled tree growth a millennium, a century, or even a decade ago. Dendroclimatologists have attempted to resolve this issue by calibrating their climate reconstructions without the most recent instrumental data (1970–present) (D’Arrigo et al. 2008). Less than a century’s worth of temperature data, however, may not be enough to recognize long- term, multidecadal, or centennial climate oscillations. Also, omitting the most recent instrumental data may be criticized as statistically impure (i.e., “data censoring”).

Several recent studies have tested the consistency and stationarity of climate-tree growth relationships over time to determine if growth responses are constantly changing, or if responses were stable prior to 1970 (D’Arrigo et al. 2004; Carrer and Urbinati 2006;

Oberhuber et al. 2008; Frank et al. 2007). Carrer and Urbinati (2006) found that

European larch (Larix decidua Mill.) responses to the most significant variables exhibited the greatest variability over time, as compared to the responses to less significant variables. Conversely, Oberhuber et al. (2008) concluded that the variables with the strongest effects maintained their strength over time, while growth responses to less

“powerful” variables were more inconsistent. These findings indicate that stationarity and variability of climate responses may be both species- and site-specific. Also, it is

6 essential to test tree-ring chronologies rigorously for consistent climate responses before attempting to reconstruct past climate or predict responses to future climate scenarios.

Perhaps most importantly, the divergence problem challenges the very principle that dendroclimatology is based upon: uniformitarianism (D’Arrigo et al. 2008). The principle of uniformitarianism states that Earth system processes acting today also acted similarly in the past. However, the divergence problem reveals that even if tree growth is driven by precipitation today, it may have actually been driven by temperature (or other climate factors) in the past. This also undermines our ability to predict how forests will respond to future climate change. Warmer temperatures may or may not mean increased tree growth or carbon sequestration (Driscoll et al. 2005). The climate-tree growth relationship is both nonlinear and nonmonotonic (D’Arrigo et al. 2008)—equal increases in temperature do not always yield equal increases in growth, and a one-degree increase in temperature may increase growth, but a second one-degree increase may actually decrease growth. Because of the divergence problem, dendroclimatologists have become more cautious with their reconstructions, and statistical tests of consistency and stationarity of the climate-growth response are necessary before a reconstruction is considered reliable (Biondi and Waikul 2004).

No consensus has been reached among dendroclimatologists on the cause(s) or implications of the divergence problem. It is currently unclear if it is limited to boreal forests at high northern latitudes or high elevations, or if divergence between temperature and tree-growth also occurs in temperate, mid-latitude mixed forests. This issue has largely been unexplored because temperature-sensitive stands are less common in mid- latitude forests (Fritts 1976). A study of shortleaf pine (Pinus echinata Mill.) in northern

Georgia showed decreased sensitivity to both precipitation and summer temperature beginning around 1963, but it is unclear whether this weakening of the climate-tree growth relationship is related to the divergence problem (Grissino-Mayer and Butler

1993). Even at high northern latitudes, where the problem is most prevalent, divergence is not found at all sites. Moreover, the sites where divergence has been identified are heterogenous. Some sites exhibit recent drought sensitivity (Barber et al. 2000), while other sites do not respond at all to moisture stress (Driscoll et al. 2005). While a consensus is far from being reached, most researchers believe that this problem is in some way anthropogenic, resulting from increased human impacts on tree growth

(D’Arrigo et al. 2008). More research is necessary, however, before any generalizations can be made regarding causes of this phenomenon.

1.3 Purpose of Research

The purpose of this thesis is to evaluate the climate-tree growth relationship for yellow pines, primarily shortleaf pine (Pinus echinata Mill.) and pitch pine (Pinus rigida

Mill.), in the western portion of the Great Smoky Mountains National Park, Tennessee.

The climate response of shortleaf pine in Great Smoky Mountains National Park is particularly unique. Researchers at the 2007 North American Dendroecological

Fieldweek discovered a winter temperature signal in a shortleaf pine chronology from

Gold Mine Trail in Great Smoky Mountains National Park (Grissino-Mayer et al. 2007).

This relationship is extraordinary because most trees at mid-latitudes and mid-elevations are more drought-sensitive than temperature-sensitive (Fritts 1976). Also, the western part of the park is the driest, making the strong shortleaf pine temperature response even

8 more startling. In this thesis, I further examine the relationships between climate and yellow pine growth in Great Smoky Mountains National Park, using five yellow pine tree-ring chronologies from the area.

Specific research objectives and hypotheses include:

1. Identify the climate variables to which yellow pines in Great Smoky

Mountains National Park respond. This is the first step in determining if

chronologies from the region may be used for climate reconstruction.

H0: No significant relationship exists between climate and annual tree

growth.

HA: Certain climate variables (temperature, precipitation) are significantly

related to annual tree growth.

2. Evaluate the temporal consistency of the climate-tree growth relationship.

If the relationship is inconsistent, a chronology should not be used for

climate reconstruction.

H0: The climate-tree growth relationship does not change over time. The

same climate variables remain significant throughout the 20th century.

HA: The climate response changes over time. The relative importance of

certain climate variables shifts during the 20th century.

3. Determine if these chronologies show evidence of divergence in the past

half-century. If so, this will indicate that divergence is indeed a problem

in mid-latitude mid-elevation forests.

H0: These chronologies show no evidence of divergence. The climate-tree

growth relationship (and particularly the temperature-tree growth

relationship) is stable during the past half-century.

HA: These chronologies suggest evidence of divergence. The

temperature-tree growth relationship weakens during the past half-century.

4. Determine if the climate responses are strong and consistent enough to

reliably isolate and reconstruct past climate.

H0: No significant relationship exists between climate and tree growth.

Therefore, past climate cannot be reconstructed from these chronologies.

HA: A significant, stable relationship exists between climate and tree

growth. This network of chronologies may be useful in reconstructing

past climate patterns.

While these objectives are undoubtedly important on a local and regional scale, they also address global issues of anthropogenic climate change and natural climate variability. Most importantly, this research confronts the divergence problem and assesses the ability of mid-latitude tree-ring chronologies to reconstruct past temperature.

1.4 Organization of Thesis

This thesis is divided into seven chapters. Chapter One serves as an introduction to tree ring-based climate reconstructions and the divergence problem. The research objectives and hypotheses are also introduced in Chapter One. Chapter Two is a review of the recent literature on temperature reconstructions, the divergence problem, and the

10 fidelity of tree rings in reconstructing past climate. The literature review will help to situate this thesis within a larger context.

In Chapter Three, I describe the area in which this study was performed, including the geology, soils, vegetation, climate, and land use history. I also discuss important differences in site characteristics between the five chronology study sites. Chapter Four outlines the field, laboratory, and statistical methods performed in this study. The results of this research and the discussion of results are included in Chapters Five and Six.

Chapter Seven concludes the thesis and highlights potential opportunities and directions for future dendroclimatology research in the southeastern U.S.

CHAPTER 2

LITERATURE REVIEW OF TEMPERATURE RECONSTRUCTIONS AND THE DIVERGENCE PROBLEM

2.1 Tree-Ring Based Temperature Reconstructions

In the past decade, several temperature reconstructions based on tree-ring data and other proxy data sources (e.g., ice cores, coral, lake sediments) have received considerable attention from both the scientific community and policy makers. The Mann et al. (1998) reconstruction, for example, was included in the Third Assessment Report of the Intergovernmental Panel on Climate Change (IPCC) in 2001. In response, McIntyre and McKitrick (2005a) highlighted numerous problems with the famed “hockey stick” reconstruction and criticized the idea of unprecedented warming in the 20th century.

Several other tree-ring based temperature reconstructions have since been published, and the dendroclimatological community has faced criticism from both within and beyond the discipline. The main criticism is that different temperature reconstructions (both multiproxy and solely tree-ring based) show different temperature patterns (Mann et al. 1998;

Jones et al. 1998; Briffa 2000; Esper et al. 2002; Moberg et al. 2005). Esper and Frank

(2004) addressed the uncertainty regarding reconstructions, finding that different data standardization techniques often yield significantly different temperature patterns.

Overall, these studies demonstrate the need for more research on temperature-tree growth relationships, as well as the role that such research plays in the broader climate change debate.

2.1.1 The Hockey Stick Reconstruction of Mann et al. (1998)

In 1998, Michael Mann, Raymond Bradley, and Malcolm Hughes published a

Northern hemisphere temperature reconstruction known as the “hockey stick” because of the shape of the temperature curve that showed rapid, unprecedented warming in the 20th century (Mann et al. 1998) (Figure 2.1). This famous reconstruction extended back six centuries and was based on a vast multiproxy network that included ice core, ice melt, coral, and historical records, but was dominated by tree-ring data. Global mean annual temperatures were reconstructed using principal components analysis (PCA) and regression analysis. The proxy dataset was calibrated with monthly instrumental grid- point data from 1902 to 1980. Despite spatial gaps in the data, Mann et al. (1998) asserted that the instrumental data were indeed representative of actual Northern

Hemisphere mean annual temperatures. The results of PCA indicated that the first five eigenvectors explained 93% of the global mean temperature variability in the calibration period. However, this percentage, and thus the skill of the reconstruction, diminished as the record extended back in time. Overall, the reconstruction showed cold periods in the mid-1600s and 1800s, and warm periods in the mid-1500s, late 1700s, and 1900s. Mean annual temperatures in the mid-to-late 20th century were warmer than any other period in the past six centuries. According to Mann et al. (1998), 20th century warming was unprecedented and anomalous.

However, a recent study by Frank et al. (2007) described an offset between reconstructed and actual temperatures prior to 1900, in which reconstructed values were significantly cooler than actual recorded temperatures. This suggests that temperature

Figure 2.1 The “hockey stick” temperature reconstruction by Mann et al. (1998), as published in the Third Assessment Report of the Intergovernmental Panel on Climate Change (2001).

14 reconstructions may underestimate past temperature, and thus overestimate recent warming.

2.1.2 Response to Hockey Stick Reconstruction by McIntyre and McKitrick (2005)

In 2005, Stephen McIntyre and Ross McKitrick responded critically to the hockey stick reconstruction of Mann et al. (1998) by noting several problems with the reconstruction, beginning with an unusual data transformation that was omitted from the methods reported by Mann et al. (1998). This transformation almost always produces a hockey stick shaped curve, regardless of the data used. McIntyre and McKitrick (2005a) accused Mann et al. (1998) of “mining the data for hockey stick patterns” to show unprecedented warming in the 20th century. They also noted low proxy data sample depth in the 15th and 16th centuries, as fewer tree-ring chronologies and other proxies were part of the reconstruction as it extended back in time. The 15th century temperature reconstruction was based primarily on bristlecone pine (Pinus longaeva D.K. Bailey,

Pinus aristata Engelm.) and foxtail pine (Pinus balfouriana Grev. & Balf.) chronologies.

Other species that did not show a hockey stick growth pattern were weighted less in the reconstruction. McIntyre and McKitrick (2005a) also criticized the use of the Reduction of Error (RE) statistic because it cannot be adequately tested for significance. McIntyre and McKitrick (2005a) tested various other verification statistics and found them to be statistically insignificant. Other potential issues with the hockey stick reconstruction of

Mann et al. (1998) include data selection and screening (i.e., using drought-sensitive tree- ring chronologies to reconstruct temperature), lack of recent years (1980–1998) in the calibration or analysis, and poor spatial coverage (i.e., reconstructing global temperature

15 from proxy records concentrated in the mid- and high latitudes of the Northern

Hemisphere.)

McIntyre and McKitrick have continued to challenge the hockey stick reconstruction and anthropogenic climate change in general in both academic papers and

McIntyre’s Climate Audit blog (McIntyre 2005; McIntyre and McKitrick 2005b). Their criticisms underscore the need for a thorough record of methods and perhaps a more stringent peer review process. The ability to replicate an experiment is at the crux of dendroclimatology and science in general, but given the methods outlined in Mann et al.

(1998), the hockey stick reconstruction could not be replicated. Overall, the scientific debate surrounding the hockey stick reconstruction reveals the importance of dendroclimatological research to both policymaking and the representation of climate change in the media.

2.1.3 Northern Hemisphere Temperature Reconstruction by Moberg et al. (2005)

Moberg et al. (2005) reconstructed Northern Hemisphere temperatures over the past 2000 years using both high-resolution tree-ring data and low-resolution proxies from lake and ocean sediments. This reconstruction included far fewer datasets (seven high resolution tree-ring datasets, 11 low-resolution series) than the famed hockey stick reconstruction of Mann et al. (1998), but the data selection was better justified. Moberg et al. (2005) only used temperature-sensitive tree-ring series from high elevation and high northern latitude sites, while Mann et al. (1998) used more data but were less selective.

Because of the mix of high- and low-resolution proxies weighted at different timescales, low-frequency, multicentennial variability was retained (Moberg et al. 2005).

Conversely, Mann et al. (1998) relied primarily on tree-ring data that had been aggressively standardized, thus removing, or at least minimizing, any low frequency trends in the data. Using a wavelet transform technique, Moberg et al. (2005) also showed changes in dominant frequency modes over time. This technique allowed proxies to be weighted individually at different time scales, to preserve both low- and high-frequency trends.

Moberg et al. (2005) suggested that recent warming is not anomalous. Although the reconstruction showed a relatively warm 20th century, the warmest period was actually A.D. 1000–1100, during the height of the Medieval Warm Period (MWP). This contradicted Mann et al. (1998), as well as the 2001 Intergovernmental Panel on Climate

Change (IPCC) report, both of which stated that 20th century warming was unprecedented. According to Moberg et al. (2005), the climate in the 20th century was still within the historical range of variability. While the highest temperatures in the 2000- year reconstruction were during the Medieval Warm Period, the lowest temperatures occurred around A.D. 1600, during a cold period referred to as the Little Ice Age.

Overall, the Moberg et al. (2005) reconstruction showed that there was likely more century- and millennial-scale natural climate variability than previously thought. Also, this reconstruction is one of few temperature reconstructions that incorporated wavelet analysis, a technique that I will use to analyze dominant modes of frequency in tree-ring data from Great Smoky Mountains National Park, Tennessee.

2.1.4 Analysis of Temperature Reconstructions by Esper and Frank (2004)

Esper and Frank (2004) responded to the discrepancies between various tree-ring based temperature reconstructions produced in the late-1990s and early-2000s (Mann et al. 1998; Jones et al. 1998; Briffa 2000; Esper et al. 2002.) As seen by comparing Mann et al. (1998) and Moberg et al. (2005), temperature reconstructions often differ significantly from one another. Esper and Frank (2004) noted that some reconstructions

(Briffa 2000; Esper et al. 2002; Moberg et al. 2005) showed well-established centennial to multicentennial climatic periods such as the Medieval Warm Period and Little Ice Age, while other reconstructions (Mann et al. 1998; Jones et al. 1998) showed no evidence of these low-frequency trends.

The goal of Esper and Frank (2004) was to determine why tree-ring based temperature reconstructions differ so significantly. Their four hypotheses included (1) insufficient data, (2) differences in spatial coverage of data, (3) different detrending techniques, and (4) different seasons represented by the proxy data. They first examined issues with the raw tree-ring data that went into the various temperature reconstructions.

Many reconstructions share tree-ring chronologies; for example, both Mann et al. (1998) and Esper et al. (2002) use the same dataset from the Polar Urals. Esper and Frank

(2004) examined the two reconstructions with and without this shared dataset to determine the impact of this data overlap. They concluded that the data used in the reconstructions were sufficient, and that even without the series from the Polar Urals, the same climatic trends were apparent. Next, Esper and Frank (2004) assessed the role of spatial coverage of data. They found that while spatial coverage may have slightly

18 influenced the amount of variability in a record, the difference was statistically insignificant.

The effects of different standardization and detrending techniques were then considered. Esper and Frank (2004) noted that long-term, low-frequency trends are the most difficult to preserve when a series is detrended. The various temperature reconstructions differ most significantly in low-frequency, multicentennial trends, implying that these differences may be caused by detrending. Also, age growth trends are commonly removed from tree-ring series when a chronology is developed. However,

Esper and Frank (2004) suggested that removing supposed growth trends could remove long-term cooling signals as well. They concluded that detrending has led to the major differences between tree-ring based temperature reconstructions.

Finally, annual versus warm season temperature discrepancies were assessed.

Esper and Frank (2004) found that annual temperature and warm season temperature data correlated significantly at 0.94, which suggests that choice of target season does not cause vast differences in the reconstructions. Also, they noted that seasonality would be more likely to cause greater variation at annual to interannual time scales, rather than the centennial to multicentennial scales. Therefore, Esper and Frank (2004) concluded that different types of detrending most likely cause discrepancies between temperature reconstructions. This is exemplified in the Moberg et al. (2005) reconstruction; significant low-frequency variability is maintained because proxy data were detrended or standardized on individual time scales. Overall, Esper and Frank (2004) substantiated the claim that temperature reconstructions are sensitive to standardization.

Dendrochronologists are now increasingly cognizant of the effects of different standardization trials on reconstructions.

2.2 Issues of Consistency and Divergence

2.2.1 Identification of Weakened Temperature Sensitivity by Briffa et al. (1998)

The divergence problem was first identified by Keith Briffa et al. (1998a).

Compositing both tree-ring density series and tree ring-width series from over 300 sites across the northern America and Eurasia, Briffa et al. (1998a) compared correlations between temperature and tree growth from 1881–1960 and 1881–1980. When the most recent data were included (1961–1980), correlations significantly decreased for both ring- widths and maximum latewood density. This weakened sensitivity was apparent at both annual and decadal timescales, but was not yet termed “divergence.”

Briffa et al. (1998a) hypothesized that the offset between temperature and tree- growth may have began as early as the 1930s. Although the northern hemisphere experienced extremely warm temperatures in the 1980s, tree growth did not respond positively. Rather, hemispheric tree growth consistently decreased from approximately

1940 to the mid-1990s, despite the overall warming trend. Briffa et al. (1998a) argued that the divergence between temperature and tree-growth occurred gradually, as correlations between temperature and tree-growth were at their lowest in the most recent years on record (late 1980s and 1990s).

This seminal work introduced the issue of temporal instability to dendroclimatology and tree-ring based temperature reconstructions. Also, it indirectly challenged the assumption that the factors limiting tree growth today also limited tree

20 growth in the past and will continue to limit tree growth in the future. The research of

Briffa et al. (1998a; 1998b) opened a new line of research in dendroclimatology that evaluates the temporal sensitivity of climate-growth relationships.

2.2.2 Evidence of Divergence in White Spruce in Alaska by Barber et al. (2000)

In a study of white spruce (Picea glauca (Moench) Voss) in east central Alaska,

Barber et al. (2000) found that radial growth of white spruce decreased from about 1970 to the mid-1990s, despite longer growing seasons. They claimed that the recent decrease in radial growth was caused by temperature-induced drought stress. As the climate warmed and the growing season lengthened, drought—rather than temperature—began to limit tree growth.

Barber et al. (2000) also showed that a favorable climate existed from 1915 to

1965 for white spruce in Alaska, with roughly decadal cycles of approximately five warmer, drier years followed by five cooler, moister years. Around 1970, the climate of east central Alaska diverged from this cycle. Significant, prolonged warming has occurred over the past 30 to 40 years without any marked increase in precipitation. This long period of unfavorable climate led to decreased white spruce growth, and likely decreased carbon sequestration in white spruce forests.

The findings of Barber et al. (2000) indicated that trees at high northern latitudes have faced temperature-induced drought stress over the past few decades. For species that have historically responded positively to warmer temperatures, this drought stress could weaken the temperature-tree growth relationship, thus leading to divergence

21 between instrumental temperature records and the temperature values predicted from tree-ring chronologies.

2.2.3 Further Analysis of White Spruce Climate-Growth Relationships by Driscoll et al. (2005)

Driscoll et al. (2005) followed up on the research of Barber et al. (2000) in southwestern Alaska. This study developed four new white spruce chronologies and tested the growth response to 20th century climate. Two of the four chronologies responded positively to warming over the last half-century, while the remaining two chronologies included individual trees that responded positively and trees that responded negatively to the increased temperatures. Only one sub-population within the four sites showed evidence of divergence.

Driscoll et al. (2005) maintained that the divergent response among one sub- population was due to temperature-induced drought stress. This hypothesis supported the results of Barber et al. (2000). Also, this shows that dendrochronologists should not assume a coherent climate response across an entire study site without testing climate- tree growth relationships. Negative and positive responders were then compared using principal components analysis. Both positive and negative-responding chronologies corresponded until about 1950, at which point the relationship began to weaken.

Although multiple hypotheses exist to explain the complexities of divergence, Driscoll et al. (2005) suggested that temperature-induced drought stress was the most important factor in this study. Overall, this research indicated that climate-growth relationships are highly complex and involve extensive interplay between climate variables. Also,

Driscoll et al. (2005) supported the claim that recent climate change has altered climatic stresses and has led to non-uniform climate responses within and between stands of trees of the same species.

2.2.4 Evidence of Divergence in the European Alps by Carrer and Urbinati (2006)

Since Briffa et al. (1998) identified the divergence problem, much of the research on divergence has taken place in two regions: northwestern Canada and Alaska, and the

European Alps. Using a regional European larch (Larix decidua Mill.) chronology from the eastern Italian Alps, Carrer and Urbinati (2006) analyzed the consistency and stationarity of the climate-growth relationship over the past two centuries (1800–1999).

They first created a regional chronology by compositing ring-width chronologies from 17 sites. Then, using moving correlations, response function analysis, and principal components analysis, they analyzed the climate-tree growth relationship over time.

Spring and early summer temperature were the most significant variables affecting tree growth throughout the entire period (1800–1999).

Carrer and Urbinati (2006) then split the record into two periods: 1800–1896 and

1897–1993. They found significant differences between the climate responses in the two periods, indicating that climate sensitivity of European larch is nonmonotonic. In particular, responses to the most significant variables exhibited the least consistency over time. For example, June temperature was the most significant climate variable, but the strength of its correlation with tree growth also exhibited the greatest variability over time.

These findings underscore the importance of testing tree-ring chronologies rigorously for consistent climate responses before attempting to reconstruct past climate or predict responses to future climate scenarios. However, it is also important to recognize that stationarity and variability of climate responses may be both species- and site-specific. Carrer and Urbinati (2006) found significant temporal variability in the climate-growth relationship for European larch in the eastern Italian Alps, but these results cannot responsibly be extrapolated to other geographic areas or other tree species.

Also, this was a landmark study because it explicitly suggested a departure from the uniformitarianism principle as applied to dendroclimatology. Still, further research is needed to determine if inconsistencies in a climate response are typical or anomalous.

2.2.5 Identification of 19th Century Divergence by Frank et al. (2007)

Frank et al. (2007) noted an early divergence different from that identified by

Briffa et al. (1998). Using regional chronologies from the European Alps and large-scale hemispheric reconstructions, Frank et al. (2007) identified an offset between pre-20th century “warmer” instrumental temperature records and “colder” tree-ring based reconstructed temperatures. In this work, they assessed four potential explanations for this offset: (1) tree-ring detrending procedures, (2) biological persistence, (3) instability or uncertainty in the climate-tree growth relationship, and (3) instrumental data quality, quantity, and homogeneity. This divergence between temperature reconstructions and instrumental data in the 19th century is particularly problematic because it suggests that late-20th century divergence is not unprecedented, and it challenges the hypothesis that recent divergence is a product of human influence and anthropogenic warming. Also, the

19th century offset calls into question the processes of climate calibration and reconstruction and decreases overall confidence in tree-ring based temperature reconstructions.

Frank et al. (2007) eliminated detrending procedures as a cause of the 19th century divergence, as similar offsets occur in numerous reconstructions that use different datasets and different standardization techniques. Biological persistence also seems an unlikely explanation, as it is generally a larger problem in ring-width chronologies than maximum latewood density chronologies, yet the offset between reconstructed and instrumental values is seen using both proxy methods.

Frank et al. (2007) concluded that the early divergence was likely related to climate-growth relationships and instrumental data quality, quantity, and homogeneity.

To best isolate the temperature signal in tree rings, Frank et al. (2007) maintained that reconstructions should rely solely on chronologies developed at sites near the thermal treeline. Temperature sensitivity, however, was recently identified in a shortleaf pine

(Pinus echinata Mill.) chronology from Great Smoky Mountains National Park,

Tennessee (Grissino-Mayer et al. 2007). More research is needed on temperature-tree growth relationships below the thermal treeline to determine if mid-elevation, mid- latitude sites may be used for temperature reconstruction. Frank et al. (2007) also admitted that the specific climatic and non-climatic variables that influence tree growth vary over time. It remains unclear, however, the extent to which these influences vary temporally, and if particular tree species exhibit more consistent growth responses than others.

Data quality, quantity, and homogeneity may have also contributed to the offset between reconstructed and instrumental temperature values. Frank et al. (2007) favored this hypothesis and supported it with data on number and spatial extent of climate stations in the Alps over time. Instrumental climate records were less homogenous and more unreliable in the 18th century than the 19th century. Also, stations in the European Alps were moved frequently in the 18th century. While they may not have been moved great spatial distances, any elevational change could have affected temperature records and trends. Frank et al. maintained that warm biases likely existed in early thermometer shelters, suggesting that the “warmer” instrumental records may be a product of a warm bias. Although Frank et al. (2007) endorsed the data issues hypothesis, they acknowledged that many concerns remain, particularly involving uncertainty and instability in climate-growth relationships.

2.2.6 Further Analysis of Climate-Growth Relationships in the European Alps by

Oberhuber et al. (2008)

This study in the Austrian Alps tested climate-tree growth relationships in five

Stone pine (Pinus cembra L.) chronologies. This was the first study of its kind using chronologies from this tree species. The goals of the study were to (1) determine what monthly climate variables influenced Stone pine growth and (2) assess the stability of climate response over time (from 1866 to 1999). The five chronologies contained multiple cores from over 150 trees located near the timberline. Oberhuber et al. (2008) used the residual chronologies from ARSTAN and evaluated climate-tree growth relationships from 1866 to 1999. They performed bootstrapped moving response

26 function analysis with the dendroclimatology statistical program DENDROCLIM2002

(Biondi and Waikul 2004) to analyze response to climate variables across moving periods of a fixed length (moving intervals).

Oberhuber et al. (2008) found current-year July temperature to be the most significant predictor of tree growth, followed closely by June temperature. These two most significant variables also showed the greatest stability of response over time.

Numerous other monthly climate variables (October and November temperature, August precipitation) were significant during certain times, but did not remain significant throughout the entire study period. These results are somewhat encouraging to dendroclimatology as a whole—the variables that had the strongest effect on tree growth tended to maintain their strength over time. However, in a similar study also in the

European Alps, Carrer and Urbinati (2006) found the opposite relationship; the most significant variables also had the most inconsistent relationships with tree growth. These two findings suggest that levels of consistency in climate-tree growth relationships are species-specific.

Like Carrer and Urbinati (2006), Oberhuber et al. (2008) found evidence of divergence between tree growth and summer temperature in recent decades. The climate- growth relationship was fairly stable from 1866 to about 1980, but from 1980 to 1999, the strong relationship between temperature and Stone pine growth diminished.

Oberhuber et al. (2008) suggested causal agents such as threshold effects and increased drought stress (also suggested in studies from the boreal forests of Alaska and western

Canada). They also mentioned the potential role of short-term extremes that are not reflected in monthly climate variables (e.g., ice storms, droughts, extreme heat events.)

Overall, this study provided yet another example of divergence between temperature and tree growth in recent years. It also supported the hypothesis that divergence is a recent phenomenon, and that climate-tree growth relationships were stable prior to the mid-19th century.

2.2.7 Review of the Divergence Problem by D’Arrigo et al. (2008)

In 2008, D’Arrigo et al. published the first review paper on the divergence problem in dendroclimatology. They defined the divergence problem as the tendency for the temperature-tree growth relationship to weaken or disappear at high northern latitude sites within the past 30 to 40 years. This definition limited divergence to the high latitudes, but more research is necessary to determine whether or not this weakening is seem at mid-latitude, mid-elevation sites. Such sites are presumed to be rare, however, as tree growth in temperate zones is generally a function of precipitation rather than temperature (Fritts 1976).

D’Arrigo et al. (2008) discussed several potential causes of this marked loss of temperature sensitivity: temperature-induced drought stress, change in seasonality and timing of snowmelt, differing responses to mean, maximum, and minimum annual temperatures, global dimming, and end effects (including standardization, detrending, and sampling bias.) Each hypothesis was assessed, but no conclusion was reached.

D’Arrigo et al. (2008) found global dimming and temperature-induced drought stress the most likely explanations for the divergence problem. If global dimming (a decrease in solar radiation due to atmospheric pollution) did indeed cause divergence, then it is

28 unlikely that divergence would be an issue at mid-latitude sites, because global dimming is strongest near the poles.

D’Arrigo et al. (2008) encouraged research on divergence in varied locations using many different tree species. The publication of this review article also signified the magnitude of the divergence problem in dendroclimatology. Since its discovery in 1998

(Briffa et al.), the divergence problem has been identified at numerous sites and has spurred a new line of research on the temporal stability of climate-tree growth relationships.

CHAPTER 3

SITE DESCRIPTION OF GREAT SMOKY MOUNTAINS NATIONAL PARK

3.1 General Setting

Officially established in 1934, Great Smoky Mountains National Park spans over

210 hectares in the Southern Appalachian mountain range (Figure 3.1) (Clark 2001; NPS

2008). In 1926, President Calvin Coolidge signed a bill that established the park upon purchase of the land, but it was not until 1934 that the federal government officially acquired the land and turned over management to the National Park Service (Campbell

1960). The park straddles the state boundary between Tennessee and North Carolina, with about half of its area in each state. With over nine million visitors per year, Great

Smoky Mountains National Park is the most visited national park in the United States and has been designated a UNESCO International Biosphere Reserve (1976) and World

Heritage Site (1983). It is highly acclaimed for both its biodiversity and the recreation opportunities provided by its 3,400 km of streams and 16 peaks over 1,800 m in height

(NPS 2008).

3.1.1 Geology

Elevations in Great Smoky Mountains National Park range from approximately

265 m at the mouth of Abrams Creek in the western part of the park, to 2,025 m at the peak of Clingmans Dome, the highest point in Tennessee (NPS 2008). This topographic diversity has facilitated the development of diverse habitats and species assemblages.

The Great Smoky Mountains are part of the larger Appalachian Mountain range that

30 extends from northern Georgia to southeastern Canada (Clark 2001). The United States

Geological Survey divides the U.S. Appalachian highlands into seven physiographic provinces: Piedmont, Blue Ridge, Valley and Ridge, St. Lawrence Valley, Appalachian

Plateaus, Adirondack, and New England. Great Smoky Mountains National Park is located in the southern section of the Blue Ridge province, bounded on the east by the

Piedmont province and on the west by the Valley and Ridge province (Clark 2001; U.S.

Geological Survey 2004).

The Appalachian Mountains formed approximately 250 to 330 million years ago

(mya) during the Alleghany orogeny (Clark 2001; Christopherson 2006). The present- day continents of North America and Africa collided and much of the east coast of North

America was uplifted, creating a northeast-southwest oriented mountain chain and series of folds and faults (Clark 2001). At the time of orogeny, the Appalachians were much taller than they are today. Over the last 100 million years, they have been gradually worn down by weathering and erosion, and these processes continue to shape the Appalachians today (Davis 2000).

Most of the rocks in Great Smoky Mountains National Park are Proterozoic sedimentary rocks, formed from sediments deposited in a shallow sea along the North

American continental margin before the North American and African continents collided

(Clark 2001; NPS 2008). Most of the rocks in the park are between 450 and 800 million years old (NPS 2008). On the far western side of the park, Precambrian aged sandstone, siltstone, and shale are most common. Quartzite (metamorphosed sandstone), schist

(metamorphosed siltstone) and slate (metamorphosed shale) are also present (King et al.

1968; Moore 1988). Limestone and dolomite are found in isolated areas, such as on the valley floor of Cades Cove (King et al. 1968; NPS 2008).

3.1.2 Soils

The soils of the southern Appalachians are primarily loams formed from sandstone, gneiss, schist, shale, or granitic parent material (Bennett 1921). Because of the varied topography, soils range from thin on steep slopes to thick and well developed on valley bottoms. The soils of Great Smoky Mountains National Park are mainly

Ultisols and Inceptisols (Bennett 1921; King et al. 1968; Bryant and Reed 1970).

Ultisols are extremely weathered forest soils, often with high moisture content but relatively low fertility due to the leaching of minerals and nutrients. Inceptisols are younger soils that often do not exhibit distinct soil horizons. Ultisols are more likely to be found on valley bottoms, while inceptisols occur on slopes and rocky surfaces

(Christopherson 2006). In terms of soil temperature, low elevation slopes and valleys in the southern Appalachians are mesic, while upper slopes and mountain peaks are frigid.

Mesic soils have a mean annual temperature between 8 °C and 15 °C, while frigid soils have a mean annual temperature below 8 °C. Within Great Smoky Mountains National

Park, the mesic-frigid boundary occurs between 1280 m and 1402 m, depending on the aspect of the slope (Lester et al. 2007).

On the western side of the park, soils are mainly formed from weathered metasedimentary rocks, such as quartzite and phyllite. These soils are well drained with low water storage capacity. Soils on mountainsides and steep slopes have particularly low water storage capacity and are often rocky and thin (Lester et al. 2007). In terms of

32 moisture, the western side of the park is relatively dry. South- and west-facing slopes are generally xeric (dry), while north and east-facing slopes are mesic (moist) (NPS 2008).

3.1.3 Climate

The climate of Great Smoky Mountains National Park is broadly classified as humid subtropical by the Köppen system of climate classification (Christopherson 2006).

The park is located in NOAA’s East Tennessee and Southern Mountains of North

Carolina climate divisions. However, all of the sites in this study are located in

Tennessee. Temperature and precipitation in the park vary with elevation. Precipitation is particularly variable throughout the park because of the orographic effect of the topography (Davis 2000). Average annual precipitation varies from 137 cm at low elevations to 216 cm at Clingmans Dome (NPS 2008). Mean annual temperature is 13.9

°C, but temperatures between high and low elevations vary significantly (NCDC 2008).

Daily high temperatures often differ by elevation up to 7.5 °C, while daily lows vary less, between 1.5 °C and 3.5 °C (Gaffin et al. 2002). Precipitation occurs year round, although the driest period is September to November. The climate of the region is also linked to ocean-atmospheric teleconnections, including the North Atlantic Oscillation (NAO), the

Atlantic Multidecadal Oscillation (AMO), El Niño-Southern Oscillation (ENSO), and the

Pacific Decadal Oscillation (PDO) (Grissino-Mayer et al. 2007).

3.1.3.1 North Atlantic Oscillation

The North Atlantic Oscillation (NAO) is a 1.7–7.5 year climate teleconnection measured by winter sea level pressure (SLP) differentials between the subpolar Icelandic

Low and the subtropical Azores High (Walker 1924; Lamb and Peppler 1987). The cycle results in a winter temperature seesaw between Europe and eastern Canada. Positive phases of the NAO (an increased pressure differential between subpolar low and subtropical high) bring stronger and more frequent storms across the Atlantic. This causes warm, wet winters in Europe, cold, dry conditions in Northern Canada, and mild, wet conditions in the eastern U.S., including the southern Appalachians (van Loon and

Rogers 1978; Hurrell and van Loon 1997). Conversely, negative phases of the NAO (a reduced pressure differential) are marked by extreme cold in Northern Europe, mild winters in Northern Canada, and cold conditions in the eastern U.S. (van Loon and

Rogers 1978; Hurrell 1996). Thus, positive phases of the NAO generally lead to mild, wet winters in the southern Appalachians, while negative phases bring colder temperatures to the region.

Relationships between NAO and tree growth have been found at several sites in the eastern U.S. (Cook et al. 1998; Grissino-Mayer et al. 2007). Using tree-ring chronologies from the northeastern U.S. and Europe, Cook et al. (1998) reconstructed pressure differentials back to 1700. For this reconstruction, Cook et al. (1998) tested over 66 chronologies from the eastern U.S., but ultimately only used six from the

Northeast and four from Europe. Surprisingly, Grissino-Mayer et al. (2007) found an

NAO signal in a shortleaf pine chronology from Great Smoky Mountains National Park,

Tennessee. Because NAO is a driver of winter climate for the region, it can also affect tree growth, particularly at sites that are highly sensitive to winter temperature.

3.1.3.2 Atlantic Multidecadal Oscillation

The Atlantic Multidecadal Oscillation (AMO) is a 65–80 year cycle of changes in sea surface temperature anomalies (SSTAs) in the North Atlantic Ocean (Kerr 2000;

Gray et al. 2004). The oscillation is composed of warm and cool phases, each lasting approximately 20–40 years, with a brief organizational period between phases. Warm phases of the AMO generally coincide with increased precipitation over Florida and the

Pacific Northwest and decreased precipitation over the rest of the continental U.S. Cool phases of the AMO correspond with increased drought frequency in Florida and the

Pacific Northwest and increased precipitation elsewhere in the U.S. (Enfield et al. 2001).

Although instrumental measurements only exist for about the past century, sea surface temperature anomalies (SSTAs) have been reconstructed back to the 16th century using tree-ring chronologies from locations bordering the North Atlantic (Gray et al. 2004).

The AMO exerts its greatest influence on the continental U.S. during the summer months (Kerr 2000; Enfield et al. 2001). In the case of Great Smoky Mountains National

Park, drought frequency and severity slightly increase during warm phases and slightly decrease during cool phases (Enfield et al. 2001). Thus, SSTAs in the North Atlantic are most likely to affect trees that are drought sensitive and particularly responsive to summer conditions.

3.1.3.3 El Niño-Southern Oscillation

The El Niño-Southern Oscillation (ENSO) is a 3–10 year cycle of shifting pressure and SST in the Pacific Ocean (Philander 1983). It exerts the greatest influence on global climate of all oscillations yet identified (Christopherson 2006). ENSO consists

35 of El Niño and La Niña events. During an El Niño event, abnormal warming of the sea surface occurs in the central and eastern Pacific and a low-pressure cell forms. High pressure develops in the western Pacific and causes the wind and water currents to weaken or even shift directions. Conversely, abnormal cooling off the coast of Peru marks La Niña events (Philander 1983). El Niño conditions are generally stronger than

La Niña events, and thus have a greater effect on temperature and precipitation globally

(Christopherson 2006).

In the U.S., ENSO exerts its greatest influence in the West. The climate of Great

Smoky Mountains National Park is only marginally affected by this oscillation

(Ropelewski and Halpert 1986). In eastern Tennessee, winter and early spring precipitation tends to be below average during strong El Niño episodes. El Niño episodes also tend to coincide with above average November–December temperatures and slightly below average January to March temperatures (National Weather Service 2003).

3.1.3.4 Pacific Decadal Oscillation

The Pacific Decadal Oscillation (PDO) is a 20–30 year pattern of climate variability in the North Pacific, consisting of positive (or warm) phases and negative (or cool) phases (Mantua et al. 1997; Mantua and Hare 2002). During positive phases, waters in the eastern Pacific Ocean (off the coast of North America) are anomalously warm, while waters in the western and central North Pacific are anomalously cool. The opposite SST pattern exists during negative phases. The effects of PDO on North

American climate are similar to the effects of ENSO, but less extreme (Latif and Barnett

1994; MacDonald and Case 2005). Positive phases generally lead to above average

36 precipitation in the American Southwest and Mexico and below average precipitation in the Pacific Northwest, Midwest, Great Lakes, and Southeast United States (Mantua and

Hare 2002). Positive phases tend to coincide with cooler temperatures in the eastern U.S. and Midwest and warmer temperatures in the western U.S. and Canada (Mantua and Hare

2002). By itself, PDO does not exert a major influence on the Southeast. However, when positive phases of PDO and AMO coincide, the Southeast is likely to experience major drought. Conversely, drought is unlikely during negative phases of both PDO and

AMO (McCabe et al. 2004). The relationship between PDO and tree growth in the western U.S. is well established (Gray et al. 2003) but the links between the PDO-AMO synergy and tree growth in the eastern U.S. are not yet fully understood.

3.1.4 Vegetation

The southern Appalachians are highly acclaimed for their biodiversity. Over

10,000 species have been identified in Great Smoky Mountains National Park, about 100 of which are native tree species (Whittaker 1956; NPS 2008). The biodiversity of the park is attributed to the northeast-southwest orientation of the southern Appalachians, the wide range of elevations and habitats, and the abundant precipitation the region receives

(Whittaker 1956; Clark 2001; NPS 2008). The orientation of the mountains was particularly critical during past glacial periods, allowing species to migrate south as the glaciers advanced and north as the glaciers receded. Because the southern Appalachians were not glaciated, the area served as a refuge for northern species whose habitats were glaciated. The spruce-fir forests found on Clingmans Dome and other mountain peaks in the park are examples of relict plant communities common in the southern Appalachians

37 during the last glacial period, but now limited to high elevations (Walker 1991; Clark

2001). The topographic variation in the park also encourages biodiversity, and particularly plant diversity. Because temperature and precipitation vary greatly between low and high elevations, plant species native to both the northern and southern U.S. are found in the park (Madden et al. 2004).

Over 95% of Great Smoky Mountains National Park is forested, and 15–25% of the forested land is considered old-growth or virgin, depending on the source (Pyle 1988;

Madden et al. 2004; NPS 2008). The most common forest types in the park are submesic to mesic oak/hardwood forest (20% of the total area), subxeric to xeric oak/hardwood forest (15%), southern Appalachian cove forest (15%), and southern Appalachian northern hardwood forest (14%) (Madden et al. 2004). Xeric pine woodlands account for just 8.9% of the total area of the park and are mainly found in the westernmost part of the park (Madden et al. 2004). This forest type is characterized by xeric slopes and ridgelines dominated by yellow pines and oaks, with an understory of white pine, red maple, rhododendron, and mountain laurel (Callaway et al. 1987; Harrod et al. 1998).

Fire suppression in xeric pine woodlands has led to drastic changes over the past century, including increased biomass, denser undergrowth, and lack of yellow pine regeneration

(Harmon et al. 1984; Harrod et al. 1998; South and Buckner 2003). At some sites, southern pine beetle (Dendroctonus frontalis Zimmerman) infestations have hastened changes, leading to premature mortality of yellow pines and regeneration of maples and white pine in canopy gaps (Lafon and Kutac 2003; Land and Rieske 2006; Waldron et al.

2007).

3.1.5 Land Use History

Prior to European settlement, the southern Appalachians were home to the

Cherokee tribe of Native Americans. The Cherokee set occasional fires, often to lure game into open areas for hunting (Van Lear and Waldrop 1989). Evidence of fire in the southern Appalachians during pre-Columbian times has been established through the presence of both soil and sedimentary charcoal (Delcourt and Delcourt 1997; Delcourt and Delcourt 1998; Welch 1999; Hass 2008). Using pollen and charcoal analysis,

Delcourt and Delcourt (1997, 1998) found that natives’ selective use of fire affected pre-

Columbian forest composition in the southern Appalachians by encouraging growth and regeneration of fire-tolerant or fire-dependent tree species including oaks and pines.

The first Euro-American settlers arrived in the Great Smoky Mountains in the late

18th century (Pyle 1988). In the early period of settlement, human-related disturbance consisted mainly of wagon roads and land clearance for pasture, agriculture, and building construction. Several large valleys in the region, including Cades Cove and

Cattaloochee, were almost completely cleared for pasture (Bratton et al. 1979; NPS

2008). Settlement was focused in valleys, at low elevations, and on the edges of what is now the national park (Pyle 1988). During the 19th and early 20th centuries, the fire return interval was shorter than in the pre-settlement era, likely because of a combination of naturally ignited fires, human-set fires for land clearance, and post-logging fires

(Harmon 1982; Pyle 1988).

The corporate logging boom in the Great Smoky Mountains was relatively short, beginning around the turn of the 20th century and ending within park boundaries by

1940. Still, approximately 40% of the park was logged by large-scale mechanized

39 logging operations (Pyle 1988). According to logging maps, over half of the North

Carolina side of the national park was corporately logged, while only about a quarter of the land in Tennessee was logged. Pyle (1988) suggested that this occurred because many large tracts of land were available to logging corporations from land speculators in

North Carolina. In Tennessee, many of the large tracts had previously been subdivided and sold off to individual settlers. On both sides of the park, it was common for half to two-thirds of a watershed to be cleared within 10–15 years through corporate logging

(Pyle 1988).

Corporate logging does not account for all of the land affected by logging in Great

Smoky Mountains National Park. Many small unmechanized logging operations also existed in the 19th and early 20th centuries (Pyle 1988; NPS 2008). These operations were focused in coves and at low elevations due to access issues. Unmechanized operations may have been more selective in their logging than corporate operations that cut over entire watersheds. It has been difficult for researchers to identify areas of unmechanized, early-style logging because it occurred in small patches rather than large swathes of land (Pyle 1988).

According to maps from logging corporations, the westernmost end of the park was not commercially logged. Cades Cove was settled and cleared, but the majority of this end of the park was subject only to “diffuse disturbance,” including small-scale logging operations on lower slopes, widespread livestock grazing on ridgetops and in old fields, human-set fires, and tanbark gathering (Pyle 1988). Very little of the westernmost end of the park was undisturbed by early settlers or loggers, but Pyle (1988) noted many large, old trees within areas of diffuse disturbance, particularly in the Abrams Creek,

Parsons Branch, and Panther Creek watersheds. When the National Park Service took over management of the Great Smoky Mountains National Park in 1934, a policy of fire suppression was instituted. Because of active fire suppression, the fire interval in the

20th century was much longer than during settlement and pre-Columbian times (Harmon

1982; Harrod et al. 2000).

3.2 Microsite Characteristics

3.2.1 Gold Mine Trail

The Gold Mine Trail site was located along flat or gentle slopes at low elevations (460–

600 m) along the western border of the park (Table 3.1; Figure 3.1). The trail extends south from the park’s western boundary near Look Rock Campground and Top of the

World Estates to Cooper Road Trail in the Abrams Creek watershed. Species present in the canopy at this site include shortleaf pine, pitch pine, Virginia pine (Pinus virginiana

L.), white oak (Quercus alba L.), black oak (Quercus velutina Lamb.), chestnut oak

(Quercus prinus L.), northern red oak (Quercus rubra L.), scarlet oak (Quercus coccinea

Münchh.), and eastern hemlock (Tsuga Canadensis (L.) Carr.) (Grissino-Mayer et al.

2007; White 2007). The understory contains white pine (Pinus strobus L.), red maple

(Acer rubrum L.), rosebay rhododendron (Rhododendron maximum L.), and mountain laurel (Kalmia latifolia L.) (Grissino-Mayer et al. 2007). The yellow pine canopy showed significant damage from southern pine beetle infection.

Also, fire scars were present on several yellow pine individuals. Prior to the creation of the national park, this area was owned by the Morton Butler Lumber

Table 3.1 Location and elevation of the five sites for chronology development within Great Smoky Mountains National Park.

Site Latitude, Longitude Elevation Gold Mine Trail N35°38.20' W83°54.77' 460–600 m

Cooper Road Trail N35°36.93' W83°51.29' 575–700 m

Shaw Grave Gap N35°30.12' W83°56.77' 550–600 m

Maynard Creek N35°30.41' W83°57.93' 400–500 m

Gregory Ridge Trail N35°32.94' W83°49.89' 700–750 m

Figure 3.1 Map of the five study sites on the western end of Great Smoky Mountains National Park, Tennessee.

Company, but the presence of many large, old trees suggests that it was not completely logged (Pyle 1988). However, the site was likely subject to diffuse human disturbance in the forms of farming and livestock grazing. The soils at Gold Mine Trail are composed of loamy residuum from metasedimentary rocks. In comparison to the other sites in this study, the soils at Gold Mine Trail are deeper, reaching bedrock at 75–100 cm (Web Soil

Survey 2007).

3.2.2 Cooper Road Trail

Cooper Road Trail is located just north of Cades Cove in the Abrams Creek watershed (Table 3.1; Figure 3.1). In the 19th century, the trail served as a wagon road leading settlers in and out of Cades Cove (Bratton et al. 1979). The site was unquestionably subject to some human disturbance, but it was not logged corporately

(Pyle 1988). The canopy at Cooper Road Trail consists of yellow pines (mainly pitch pine, Virginia pine, and shortleaf pine), chestnut oak, white oak, and white pines and eastern hemlock at the lowest elevations. The understory is dominated by white pine and dense mountain laurel, indicative of fire suppression in the 20th century. The trees directly adjacent to the trail did not show evidence of fire, but many trees along the northeast ridgetop had been scarred by fire. Soils at this site are well-drained and rocky, mainly derived from hard metasandstone. The depth to bedrock is 50–100 cm (Web Soil

Survey 2007).

3.2.3 Shaw Grave Gap

Shaw Grave Gap is located on the southwestern extreme of the national park, near the Tennessee-North Carolina border (Table 3.1; Figure 3.1). The site marks the grave of

Bas Shaw, a Union soldier who was captured and killed during the Civil War by

Confederate troops. This indicates that the study site, located in Parsons Branch watershed, was subject to some human disturbance. According to Pyle (1988), the watershed was not logged corporately, but was settled diffusely and cannot be classified as virgin forest. At Shaw Grave Gap, yellow pines were sampled on steep, xeric south- and southwest-facing slopes between 550 and 650 m in elevation. The canopy is relatively sparse and mainly consists of pitch, shortleaf, and Virginia pine, red oak, and black oak. Blueberry (Vaccinium spp.) is prevalent in the understory, along with white pine and red maple saplings and seedlings. Several large, old yellow pines had triangular scars at the base, indicating that one or more fires occurred at the site over the past two centuries. Soils at Shaw Grave Gap are rocky loams with very low water storage capacity. The depth to bedrock is 50–100 cm (Web Soil Survey 2007).

3.2.4 Maynard Creek

The Maynard Creek site is located in the southwestern corner of the park, three to five kilometers west of Shaw Grave Gap (Table 3.1; Figure 3.1). Maynard Creek is a small tributary of Tabcat Creek, which ultimately flows south into the Little Tennessee

River. The Tabcat Creek watershed was not logged corporately, but it was likely subject to scattered human disturbance in the form of early-style logging and livestock grazing

(Pyle 1988). Shortleaf pine, pitch pine, Virginia pine, and several oaks (Quercus spp.)

44 dominate the canopy at MCR. The understory includes blueberry, scattered mountain laurel, white pine, sassafras (Sassafras albidum (Nutt.) Nees), and red maple. The yellow pines present in the canopy at Maynard Creek are generally smaller and appear younger than those at Shaw Grave Gap, indicating that perhaps the site was once cut over. Also, very few yellow pines were fire-scarred. Soils at Maynard Creek are less rocky and more stable than at Shaw Grave Gap, likely because the slopes are gradual rather than steep.

3.2.5 Gregory Ridge Trail

Gregory Ridge Trail is located in the Abrams Creek watershed south of Cades

Cove (Table 3.1; Figure 3.1). Gregory Ridge Trail is the highest in elevation of the five sites, at 700–750 m. The environment also appears the most xeric of the five sites, although this was not measured quantitatively. The canopy is dominated by shortleaf pine, pitch pine, and Table Mountain pine (Pinus pungens Lamb.) at the highest elevations. Chestnut oak, blackjack oak (Quercus marilandica Münchh.), northern red oak, and black oak are also present in the overstory. The understory includes red maple, white pine, sassafras, sourwood (Oxydendrum arboreum (L.) DC.), and black gum (Nyssa sylvatica Marsh.) Little to no yellow pine regeneration has occurred at this site, and many of the mature yellow pines have been affected or killed by southern pine beetle.

The site showed evidence of fire in the form of fire-scarred trees and charred wood.

CHAPTER 4

METHODS: FIELD, LABORATORY, AND STATISTICAL ANALYSIS

4.1 Field Methods

Tree cores were collected from five sites on the western end of Great Smoky

Mountains National Park for chronology development (Table 4.1). The Gold Mine Trail chronology was developed from yellow pine cores collected in 2005–2007 by members of the Laboratory of Tree-Ring Science at the University of Tennessee. I selected four additional sites at which to develop yellow pine chronologies (Gregory Ridge Trail, Shaw

Grave Gap, Maynard Creek, and Cooper Road Trail) and obtained a research permit from the National Park Service to sample within Great Smoky Mountains National Park. Sites were selected based on the Principle of Site Selection, which states that trees growing on xeric slopes or in rocky soils are more likely to be sensitive to climate than trees growing in flat, mesic environments (Fritts 1976). Sampling at Gregory Ridge Trail was completed in fall 2007, while sampling at Shaw Grave Gap, Maynard Creek, and Cooper

Road Trail was completed in summer and fall 2008. The Gold Mine Trail chronology was developed primarily from shortleaf pine samples, with a small number of cores from pitch pine and Virginia pine. The Shaw Grave Gap and Maynard Creek chronologies included only shortleaf pine, while the Gregory Ridge Trail chronology included shortleaf pine, Table Mountain pine, and pitch pine (Figure 4.1). The Cooper Road Trail chronology was developed mainly from pitch pine samples, with a small number of shortleaf pine cores (Table 4.1).

Table 4.1 Chronologies used to examine climate-tree growth relationships.

Site Chronology Species Date Sampled Gold Mine Trail P. echinata, P. rigida 2005–2007

Cooper Road Trail P. rigida, P. echinata Fall 2008

Shaw Grave Gap P. echinata Summer 2008

Maynard Creek P. echinata, P. rigida Summer 2008

Gregory Ridge Trail P. echinata, P. pungens, P. rigida Fall 2007

Figure 4.1 At the Gregory Ridge Trail site, Laboratory of Tree-Ring Science student Ian Feathers cores a shortleaf pine.

At each of the sites, I targeted yellow pines with a diameter at breast height

(DBH) in excess of 20 cm. I also examined each tree for characteristics of old age, attempting to sample the oldest trees to construct chronologies that span multiple centuries. Particular characteristics that I targeted included: large DBH, heavy branches, isolation from other pines, erratic growth forms, and fire or other scarring on the base of the tree (Schulman 1954; Wagener and Schulman 1954). Two core samples were extracted from each tree at approximately 1 m height using Haglof increment borers.

Cores were taken at a right angle to the slope (or parallel to the contours) to avoid the reaction wood that conifers often develop on the downslope side of the stem (Grissino-

Mayer 2003). The increment borer often hit rot before it reached the pith of the tree. In this case, the core was extracted immediately upon hitting rot. After extraction, all samples were stored in paper straws and labeled with site and core ID, species, date,

DBH, and any identifying features. I collected a total of 210 cores: 110 from Shaw

Grave Gap, 40 from Maynard Creek, 60 from Cooper Road Trail, and 68 from Gregory

Ridge Trail. After coring, samples were transported to the Laboratory of Tree-Ring

Science at the University of Tennessee, Knoxville for processing and analysis.

4.2 Laboratory Methods

4.2.1 Processing and Preparation

After drying for at least 24 hours, cores were mounted on wooden core mounts using Elmer’s glue and masking tape. Mounts were labeled with site and core ID, species, DBH, date extracted, and the initials of the corer. Samples were then prepared for analysis by sanding each core with progressively finer sandpaper (ANSI 80, 120, 220,

320, and 400 grit) to smooth the surface and reveal individual cell boundaries (Orvis and

Grissino-Mayer 2002). Each core was then examined under a stereoscopic microscope and each ring was assigned a calendar year by counting backwards from the most recent year’s growth, if bark was present on the core. If the end of the core did not include bark, each ring was assigned a floating date, beginning with year one at the pith or the earliest ring present. After ring counting, the widths of all rings on all cores were measured to

0.001 mm on a Velmex measuring system with Measure J2X software.

4.2.2 Crossdating

The samples from each site were then crossdated using a combination of visual, graphical (skeleton plotting), and statistical crossdating techniques (Yamaguchi 1991;

Stokes and Smiley 1996). The technique of crossdating involves assigning a calendar year to each individual ring by matching patterns of wide and narrow rings across samples. Samples from the same site will generally possess similar growth patterns that reflect the past climate of the region (Fritts 1976). I mainly used statistical crossdating to construct the chronologies, but visual and graphical techniques were used to correct misdated samples. Dating accuracy was checked with COFECHA 2.1. This software uses segmented time series analysis and calculates numerous correlation coefficients to determine if ring-width patterns match across samples (Holmes 1983; Grissino-Mayer

2001). I used 40-year segments lagged by 20 years to test the correlations between each ring width series and the average of all the other series in the chronology. Correlation coefficients below the critical threshold of 0.3665 (P < 0.01) were flagged by

COFECHA. If a higher correlation could be achieved by shifting the flagged segment by

10 or fewer years, COFECHA suggested a dating adjustment. With each flagged segment, I first looked at the correlation coefficients and determined the specific period that was problematic within the series. I then examined the core under a microscope, looking for false rings, micro-rings, shadow rings, or breaks in the core that may have led to misdating. If necessary, I created a skeleton plot (graphical crossdating) of the core to determine the precise time period when an offset occurred. If correlation coefficients were extremely low throughout the series with no evidence of misdating, the sample was eliminated from the chronology because it did not exhibit a strong regional climate signal.

4.2.3 Standardization

Once crossdating was established, the raw ring-width measurements were standardized using the program ARSTAN (Cook 1985). Standardization converts the ring widths into dimensionless indices, removes age-growth trends, and allows the raw data to be averaged into a single index chronology for each site (Fritts 1976). The raw data were detrended conservatively, using linear regression lines or negative exponential curves, rather than cubic smoothing splines, which may remove low-frequency climate information along with age-growth trends and disturbance events (Figure 4.2) (Cook and

Peters 1981; Cook et al. 1990; Carrer and Urbinati 2006). Three different outputs were created for each site: standard (STD), residual (RES), and ARSTAN (ARS) chronologies

(Cook 1985). Because the standard chronologies at each of the five sites correlated best with climate, I chose to analyze the consistency of the climate response using the standard chronologies only.

Figure 4.2 Using the program ARSTAN, the majority of the raw ring-width series were detrended with negative exponential curves. The top graph shows a raw ring-width series from Gold Mine Trail and the negative exponential curve fitted to the line, while the bottom shows the standardized tree-ring indices.

I also compared several descriptive statistics generated by COFECHA and

ARSTAN across the five site chronologies: average interseries correlation, mean sensitivity, and first-order serial autocorrelation. Average interseries correlation is a measure of how well the samples within a chronology correlate. A high average interseries correlation (greater than 0.40) suggests accurate crossdating and a common signal, likely a regional climate response (Grissino-Mayer 2001). Mean sensitivity reflects year-to-year changes in tree growth. Samples from a site with a high mean sensitivity have considerable variation in ring widths, while those from a site with low mean sensitivity are complacent and are unlikely to exhibit a strong climate response

(Holmes 1983; Grissino-Mayer 2001). In the Southeast, a mean sensitivity of 0.15–0.20 is common, but a mean sensitivity between 0.25 and 0.35 is best for crossdating and climate analysis. First-order serial autocorrelation was also compared across sites to determine if biological persistence is an issue in any of the chronologies. If serial autocorrelation is high (close to 1.0), the climate signal in a chronology may be obscured by year-to-year persistence in tree growth (Grissino-Mayer 2001).

4.3 Climate Analysis

4.3.1 Climate Data

Climate data were obtained from the National Climatic Data Center (NCDC) for

Eastern Tennessee (Division 4001) from 1895 to 2007. The divisional dataset consists of observations averaged from all of the weather stations within the region. I selected the following variables for analysis: monthly mean minimum temperature, monthly total precipitation, and monthly Palmer Drought Severity Index (PDSI). PDSI is a drought

52 index calculated using precipitation, temperature, and soil moisture measurements.

Positive values indicate wet conditions, while negative values indicate drought (NCDC

2008). I chose to analyze the relationship between tree growth and monthly average minimum temperature, rather than average temperature, because a previous study by

Grissino-Mayer et al. (2007) found that minimum temperatures more strongly correlate with yellow pine tree growth in the southern Appalachians. My preliminary analysis supported this finding as well. The precipitation and PDSI data used in the analysis were divisional data, while the monthly mean minimum temperatures were from a single weather station: the McGhee Tyson Airport, south of Knoxville in Blount County,

Tennessee (NCDC 2008). I selected this weather station because it is the closest weather station to the study sites that has early 20th century data available (beginning in 1910).

I also used the following data on ocean-atmospheric teleconnections as potential drivers of tree growth in the region:

• AMO: North Atlantic Sea Surface Temperature Anomalies (SSTAs) based on a

5 º latitude by 5 º longitude global SST grid, available 1856–2008 (Kaplan et al.

1998; NOAA 2009).

• PDO: a monthly PDO index derived from SSTAs in the North Pacific Ocean,

poleward of 20 ºN latitude, available 1900–2008 (Mantua 2009).

• ENSO: monthly values of the Southern Oscillation Index (SOI), derived from

sea level pressure differentials between Tahiti and Darwin, Australia, available

1887–2008 (Troup 1965; Queensland Government Environmental Protection

Agency 2009).

• NAO: monthly NAO index values based on standardized sea level pressure

anomalies in the north Atlantic, available 1874–2005 (Rogers 1984; Rogers

2005).

4.3.2 DENDROCLIM2002

The first step of climate analysis was to determine which climate variables were most significantly related to yellow pine growth at each of the five sites. I used two complementary techniques, correlation analysis and response function analysis, to test the growth response to climate. Both types of analysis were performed in

DENDROCLIM2002, a dendroclimatic analysis program developed by Franco Biondi

(Biondi 1997; Biondi and Waikul 2004). The benefit of using DENDROCLIM2002, as opposed to other programs such as PRECON (Grissino-Mayer and Fritts 1995), is that bootstrapped confidence intervals are calculated for both correlation and response functions. This ensures accurate testing for significance of variables (Biondi and Waikul

2004). A weakness of DENDROCLIM2002, however, is that it does not develop a bioclimatic model of tree growth with r2 values that show how much of the variance in tree growth is explained by the independent climate variables in response function analysis (Fritts et al. 1971; Biondi 1997).

4.3.3 Correlation Analysis

Within DENDROCLIM2002, I calculated Pearson product-moment correlation coefficients between each year’s ring-width index and monthly climate variables from

June of the previous growing season to November of the current growing season, as tree

54 growth is influenced by past as well as present growing conditions (Fritts 1976).

Correlation coefficients were calculated using the period from 1930 to the most recent year of growth. While instrumental records of temperature, precipitation, PDSI, and

SSTs extend back prior to 1930, they may be less reliable and include more outlier observations. In correlation analysis, high r-values indicate a positive relationship between a given climate variable and tree growth, whereas negative r-values suggest an inverse relationship (Fritts 1976).

4.3.4 Response Function Analysis

Developed by Fritts, this technique uses principal components multiple regression to eliminate the effects of interdependence among climate variables (Fritts et al. 1971;

Fritts 1976). Like correlation functions, response function analysis indicates which monthly climate variables significantly affect tree growth. This technique has been criticized for incorrect estimates of error and overstated significance levels in older software versions, but DENDROCLIM2002 and recent versions of PRECON (Grissino-

Mayer and Fritts 1995) use bootstrapped confidence intervals to comprehensively test for significance (Biondi and Waikul 2004). In PRECON, a model of tree-growth is developed based on climatic variables, while in DENDROCLIM2002, the significant variables are highlighted, but a model is not developed (Fritts et al. 1971; Biondi and

Waikul 2004).

4.3.5 Moving Correlation Analysis

After the initial correlation and response function analyses, I performed moving correlation analysis between growth and monthly precipitation, temperature, and PDSI at moving 36-year intervals using DENDROCLIM2002. I selected a window length of 36 years because Biondi and Waikul (2004) suggest a window of at least twice the number of predictor months. The length of the intervals was constant, but they were progressively shifted forward in time by one year. For precipitation and PDSI, the period of analysis was 1895 to the end of each chronology (2005, 2006, or 2007). For temperature, the analysis began in 1910 when minimum temperature data were first recorded at the McGhee Tyson Airport weather station. The objective of this type of analysis was to test for temporal changes in the climate-tree growth relationship (Biondi

1998; Biondi and Waikul 2004). This includes not only complete shifts to which variables a tree, or stand of trees, responds, but also changes in the strength of that response. A consistent response to a given climate variable indicates that a dataset may have potential for climate reconstruction. The results of moving correlation analysis will be displayed individually for each variable (monthly mean minimum temperature, monthly precipitation, and PDSI) using the standard graphical outputs from

DENDROCLIM2002. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient. For example, correlations above 0.40 are shaded dark red, while correlations between 0.30 and 0.40 are a lighter shade of red.

Analyzing the changing relationships between tree growth and climate has a long history within dendroclimatology (Van Deusen 1987; Visser and Molenaar 1987; Smith

1990), but has become increasingly important since the identification of the divergence problem (Briffa et al. 1998a). In the 1980s, Van Deusen (1987) and Visser and Molenaar

(1987) applied the Kalman filter (Kalman 1960) to dendrochronology. The Kalman filter has been used to create tree-ring chronologies from numerous ring-width series and to generate predictors of tree growth (based on climate variables) that can change over time

(Van Deusen 1987). This technique indicates whether or not the significance levels of specific climate predictors are stable over time and has been used to analyze tree growth- climate relationships in the eastern U.S. (Smith 1990). The Kalman filter is rarely used in dendroclimatology today, but the early work of Van Deusen (1987) and Visser and

Molenaar (1987) suggested that the climate variables that affect tree growth today might not be the same variables that affected past tree growth. Now, calibration periods for dendroclimatic reconstructions are thoroughly tested for stationarity (Biondi and Waikul

2004; D’Arrigo et al. 2008). If the climate response varies significantly throughout the calibration period, the reliability of the reconstruction is compromised.

4.3.6 Wavelet Analysis Wavelet analysis indicates the dominant modes of frequency in a temporal dataset by breaking a time series into wavelets of different wavelengths and frequencies

(Torrence and Compo 1998). In the final segment of my climate analysis, I used this technique to determine the dominant frequencies in each of the five chronologies over time. The dominant frequencies in the tree-ring data were then compared to the periodicities of climate oscillations known to affect weather and climate, and thus tree growth, in the Southeast (AMO, PDO, ENSO, NAO). Wavelet analysis can also reveal whether or not the dominant modes of frequency in a dataset remain stable over time. In

57 the case of tree-ring data, instability in the dominant frequency modes suggests inconsistencies in the climate-tree growth relationship over time.

4.3.7 Composite Chronology

After climate analysis of each individual site chronology, the chronologies with a common climate signal were averaged together to create a composite yellow pine chronology. Because the climate responses across the five sites were similar, all five chronologies were used to create the composite chronology. Correlation, response function, moving correlation, and wavelet analyses were then performed on the composite chronology as well. The purpose of this step was to determine the dominant climate variables to which yellow pines in Great Smoky Mountains National Park respond, and also to identify changes in climate response over time.

CHAPTER 5

RESULTS: CHRONOLOGY CONSTRUCTION AND CLIMATE ANALYSIS

5.1 Chronology Construction

I constructed ring-width chronologies at four sites (Cooper Road Trail, Shaw

Grave Gap, Maynard Creek, and Gregory Ridge Trail), while Lisa LaForest developed the fifth site chronology (Gold Mine Trail). I collected a total of 260 cores across the four sites, but I used only 203 measured ring-width series to develop the chronologies

(Table 5.1). Several measured series were not included because they contained unclear rings due to excess resin or numerous breaks in the core. Also, some series were omitted because they did not correlate with others at the same study site. These trees may have been responding to factors other than regional climate, such as localized disturbance or gap-phase dynamics (Cook 1987).

The most extensive sampling took place at Gold Mine Trail. The site chronology contained 187 series from 117 yellow pine trees (mainly shortleaf pine with some pitch pine.) The remaining four chronologies included between 34 and 80 series from 24 to 47 individual trees (Table 5.1). The Maynard Creek chronology included the fewest series

(34 series from 24 trees). Although not as extensive as Gold Mine Trail, the average interseries correlation was high (0.52), suggesting a common climate signal. The

Maynard Creek chronology was also the shortest in length, spanning from 1838 to 2007, while the Gold Mine Trail chronology was the longest (1684–2006). Average interseries correlations at all of the sites were high, varying from 0.52 at Maynard Creek to 0.54 at

Table 5.1 Characteristics of the five chronologies, including the number of individual trees and series used in chronology construction, the time span of each chronology, the average interseries correlation, mean sensitivity, and first-order serial autocorrelation.

Number of Site Chronology Time Span A.I.C. M.S. Autocorrelation Trees/Series Gold Mine Trail 117/187 1684–2006 0.53 0.27 0.75

Cooper Road Trail 30/49 1808–2007 0.53 0.26 0.74

Shaw Grave Gap 47/80 1733–2007 0.54 0.25 0.78

Maynard Creek 24/34 1838–2007 0.52 0.26 0.77

Gregory Ridge Trail 27/40 1836–2005 0.52 0.27 0.77

Shaw Grave Gap (Table 5.1). The trees at each of the sites were moderately sensitive, with mean sensitivities ranging from 0.25 at Shaw Grave Gap to 0.27 at Gregory Ridge

Trail. These are fairly typical values for sloped sites in the Southeast (Grissino-Mayer

2001). First-order serial autocorrelation for each of the five chronologies was relatively high, suggesting annual growth was affected by previous years’ growth (biological persistence) (Grissino-Mayer 2001). Because autocorrelation can obscure the climate signal in a ring-width chronology, standardization is used to reduce the degree of autocorrelation in each series.

The five chronologies show relatively similar patterns of suppression and release during the common period (1895–2005) (Figure 5.1). The Gold Mine Trail chronology shows erratic growth during the early period (1730–1740), likely due to the low sample size (n = 10). Because no other chronologies extend this far back with adequate sample depth, the growth patterns cannot be compared. All of the chronologies show below- average ring widths in the 1890s, mid-1960s, and late 1980s (Figure 5.1). Since 2000, tree growth was about average at four of the sites, but the Gregory Ridge Trail chronology showed a release event characterized by above-average ring widths. Because none of the other nearby chronologies showed a release, it was likely site-specific rather than related to regional climate trends.

5.2 Gold Mine Trail

5.2.1 Temperature, Precipitation, and PDSI

Tree growth (standardized ring-widths) at Gold Mine Trail was correlated more with monthly mean minimum temperature than precipitation or PDSI (Table 5.2). The

2.0 A 1.5

1.0

0.5

0.0 2.0 B 1.5

1.0

0.5 0.0

2.0 C

1.5

1.0

0.5

0.0 2.0 D 1.5

1.0

0.5

0.0 2.0 E 1.5

1.0

0.5

0.0 1725 1775 1825 1875 1925 1975 2025

Figure 5.1 Standard chronologies at each of the five sites, shown during the periods when each chronology includes at least 10 ring-width series. A: Gold Mine Trail, B: Cooper Road Trail, C: Shaw Grave Gap, D: Maynard Creek, E: Gregory Ridge Trail.

Table 5.2 In ranked order, the five monthly variables most strongly correlated with growth at each site (and whether the relationship is positive or negative.) At Cooper Road Trail, Maynard Creek, and Gregory Ridge Trail, the strongest correlations were with PDSI. At Gold Mine Trail and Shaw Grave Gap, winter temperature was most significant.

Gold Cooper Shaw Maynard Gregory Composite Mine Road Trail Grave Gap Creek Ridge Trail Chronology Trail Jan. T May PDSI Jan. T May PDSI Feb. PDSI Jan. T (+) (+) (+) (+) (+) (+) Feb. T May P Feb. T June PDSI July PDSI Feb. T (+) (+) (+) (+) (+) (+) June T June PDSI Feb. PDSI Feb. PDSI May PDSI Feb. PDSI (+) (+) (+) (+) (+) (+) Feb. PDSI July PDSI May PDSI Aug. PDSI Aug. PDSI May PDSI (+) (+) (+) (+) (+) (+) May T Dec. PDSI March T July PDSI June PDSI June PDSI (+) (+) (+) (+) (+) (+)

63 most significant variables were January (r = 0.52, P < 0.05) and February temperature (r

= 0.46, P < 0.05) (Figure 5.2). Warmer winters were associated with above average yellow pine growth, while colder winters tended to coincide with below average annual growth, supporting the findings of Grissino-Mayer et al. (2007). Current March, May,

June, and October temperatures, along with previous June temperature, were also positively correlated with annual growth (Table 5.2). February precipitation and PDSI were weakly positively correlated with tree growth. While moisture availability in the late winter may affect tree growth, it did not have as strong a relationship with growth as mean minimum temperature. Overall, correlation analysis showed that January and

February were the most significant months for yellow pine growth at Gold Mine Trail.

The response function coefficients complemented the results of correlation analysis, but fewer monthly variables were significant (Figure 5.3). Again, winter minimum temperatures (January, February, and March) were the most significant variables, suggesting that warmer winters lead to increased annual growth. October temperature and previous November temperature also significantly affected growth.

Warmer Octobers led to above average growth, likely because the growing season was lengthened. Conversely, warmer previous Novembers were associated with below average growth in the following year. October temperature was also significant in correlation analysis. In response function analysis, no monthly PDSI or precipitation variables were significant at Gold Mine Trail, indicating that temperature, rather than moisture availability, drives yellow pine growth at this site (Figure 5.3).

0.55 * * 0.45 0.35 * ** * * * 0.25 * 0.15 0.05

-0.05 -0.15 J J A S O N D J F M A M J J A S O N -0.25 Figure 5.2 Bootstrapped correlations between the Gold Mine Trail chronology and monthly mean minimum temperature, precipitation, and PDSI from the previous June to the current November (1930–2006) (P < 0.05).

0.35 * * 0.25 * *

0.15

0.05

-0.05 -0.15 J J A S O N D J F M A M J J A S O N -0.25 * Temp. Precip. PDSI

Figure 5.3 Bootstrapped response function coefficients at Gold Mine Trail. January, February, and March temperatures were the strongest predictors of growth, while previous November and current October temperatures were weakly significant (1930– 2006) (P < 0.05).

5.2.2 Moving Correlation Analysis

Using DENDROCLIM2002, I performed correlation analysis over moving 36- year intervals to identify changes in the climate-tree growth relationship over time. This analysis indicated a strengthened response to winter temperature (particularly January and February) in the latter half of the 20th century (Figure 5.4). From 1910 to 1950,

January and February average low temperatures were not significantly related to tree growth. Since 1950, however, winter temperatures have been strongly associated with annual tree growth at Gold Mine Trail (Figure 5.4). The influence of October temperature on tree growth also shifted in recent years. Before about 1970, the relationship between October temperature and tree growth was insignificant, but from

1970 to 2006, October temperature was significantly positively associated with growth

(Figure 5.4).

Although temperature was the main driver of tree growth at Gold Mine Trail, the relationships between growth and precipitation and PDSI also shifted over time (Figures

5.5 and 5.6). In the first half of the 20th century, May, June, and July precipitation and

PDSI were positively correlated with growth. The relationship between growing season precipitation and growth weakened mid-century, and February precipitation became more significant (Figure 5.5). The relationship with growing season PDSI also weakened mid- century, but again became significant in the 1980s. From 1990 to 2006, moving correlation analysis suggested a strengthening of the response to both PDSI and May precipitation (Figures 5.5 and 5.6).

Figure 5.4 Results of moving correlation analysis between temperature and the Gold Mine Trail chronology, using 36-year moving intervals. Monthly temperature variables are listed on the y-axis, beginning with previous June in the lower left corner and ending with current November in the upper left corner. Last years of moving intervals are listed on the x-axis. For example, a grid square marked 1946 represents the period from 1910 to 1946. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient.

Figure 5.5 Results of moving correlation analysis between monthly precipitation and the Gold Mine Trail chronology, using 36-year moving intervals. Monthly variables are listed on the y-axis, beginning with previous June in the lower left corner and ending with current November in the upper left corner. Last years of moving intervals are listed on the x-axis. For example, a grid square marked 1946 represents the period from 1910 to 1946. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient.

Figure 5.6 Results of moving correlation analysis between monthly PDSI and the Gold Mine Trail chronology, using 36-year moving intervals. Monthly variables are listed on the y-axis, beginning with previous June in the lower left corner and ending with current November in the upper left corner. Last years of moving intervals are listed on the x-axis. For example, a grid square marked 1946 represents the period from 1910 to 1946. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient.

5.2.3 Ocean-Atmospheric Teleconnections

Correlation analysis between monthly oscillation index values and the Gold Mine

Trail chronology suggested a complex relationship between ocean-atmospheric teleconnections, regional climate, and yellow pine tree growth. Of the four variables analyzed (Atlantic Ocean SSTAs, SOI, and NAO and PDO indices), the Gold Mine Trail chronology had the strongest relationship with Atlantic Ocean SSTAs (Figure 5.7).

However, tree growth was more strongly correlated with previous summer and fall

SSTAs than with current growing season values.

NAO index values, based on Atlantic Ocean sea level pressure differentials between the Icelandic Low and the Azores High, were also related to tree growth at Gold

Mine Trail (Figure 5.7). Previous November NAO values were negatively associated with growth (r = –0.25, P < 0.05), while January (r = 0.26, P < 0.05), February (r = 0.25,

P < 0.05), and August values (r = 0.28, P < 0.05) were significantly positively correlated to growth, suggesting that positive phases of the NAO tended to coincide with, and possibly cause, increased annual growth.

The Gold Mine Trail chronology was less associated with climate variability in the Pacific Ocean than the Atlantic Ocean. The Southern Oscillation Index (SOI), a measure of the strength of ENSO, was weakly positively correlated with tree growth during the previous December (r = 0.21, P < 0.05), current April (r = 0.22, P < 0.05), and current July (r = 0.26, P < 0.05) (Figure 5.7). Higher SOI values (La Niña years) were weakly associated with increased growth, while lower values (El Niño years) tended to coincide with below average annual growth. Still, the correlation coefficients between

0.5 * * * A * * * * 0.4 * * * * * * * * * 0.3 * * 0.2 0.1 0.0 -0.1 J J A S O N D J F M A M J J A S O N -0.2 -0.3

0.5 B 0.4

0.3 * * * 0.2 0.1 0.0 -0.1 -0.2 J J A S O N D J F M A M J J A S O N -0.3 * *

0.5 C 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 J J A S O N D J F M A M J J A S O N -0.3 * * *

0.5 D 0.4 0.3 * * * * * 0.2 0.1 0.0 J J A S O N D J F M A M J J A S O N -0.1 -0.2 -0.3 Figure 5.7 Bootstrapped correlation coefficients between monthly oscillation index values and the Gold Mine Trail standard chronology (1930–2006). A: Atlantic Ocean SSTAs. B: NAO index. C: PDO index. D: Southern Oscillation Index (SOI) (P < 0.05).

71 growth and the SOI were weaker than correlations with NAO index values and Atlantic

Ocean SSTAs.

Of the four oscillations, PDO had the weakest relationship with tree growth at

Gold Mine Trail (Figure 5.7). Correlations with monthly PDO index values were weakly negative during the winter and current growing season. No significant relationship existed between previous year PDO values and tree growth. January (r = –0.21, P <

0.05), February (r = –0.25, P < 0.05), and August (r = –0.24, P < 0.05) were the three months in which the relationship with PDO was significant. Overall, however, the correlation coefficients for PDO were weak, suggesting only a tenuous relationship between Pacific Ocean SSTAs and yellow pine growth in the Southeast.

5.2.4 Wavelet Analysis

Wavelet analysis of the Gold Mine Trail chronology showed a 60–100 year oscillation dominant from about 1810 to 2000 (Figure 5.8). This periodicity was not significant in the early part of the chronology (1731–1810), indicating a possible change in the climate response in the early 19th century.

5.3 Cooper Road Trail

5.3.1 Temperature, Precipitation, and PDSI

The Cooper Road Trail chronology was positively correlated with monthly PDSI values from June of the previous year to November of the current year (Figure 5.9). The highest correlations were with May (r = 0.49, P < 0.05), June (r = 0.44, P < 0.05), and

July PDSI (r = 0.39, P < 0.05), indicating that yellow pine growth at Cooper Road Trail

Figure 5.8 Results of wavelet analysis of the Gold Mine Trail chronology (1731–2006): (a) the standard chronology at Gold Mine Trail, (b) wavelet power spectra for the chronology, (c) the global wavelet power spectrum. Crosshatches represent the cone of uncertainty. Black lines outline the significant periodicities. All analyses used the Morlet wavelet and zero-padded series (Torrence and Compo 1998).

0.55 * * * 0.45 * * * 0.35 * * * * * * * * * * * * * * * 0.25 *

0.15

0.05

-0.05 J J A S O N D J F M A M J J A S O N -0.15

-0.25 Temp. Precip. PDSI

Figure 5.9 Bootstrapped correlations between the Cooper Road Trail chronology and monthly mean minimum temperature, precipitation, and PDSI from the previous June to the current November (1930–2007) (P < 0.05).

0.35 *

0.25 * *

0.15

0.05

-0.05

-0.15 J J A S O N D J F M A M J J A S O N

-0.25 Temp. Precip. PDSI Figure 5.10 Bootstrapped response function coefficients at Cooper Road Trail. May precipitation was the strongest predictor of growth, while February and March temperatures were weakly significant (1930–2007) (P < 0.05).

74 was linked to growing season moisture availability and low drought severity (Table 5.2).

However, PDSI was not significant in any month in response function analysis (Figure

5.10). The three monthly variables that explained tree growth were May precipitation and February and March temperature, but the response function coefficients for these variables were relatively low (0.30 or less). May precipitation also correlated well with growth (r = 0.47, P < 0.05), whereas February (r = 0.28, P < 0.05) and March temperature

(r = 0.26, P < 0.05) were only weakly significant (Figure 5.9). The differing results from correlation analysis and response function analysis suggest that 20th century tree growth at Cooper Road Trail was not driven by one primary climate variable or climate in one particular season. Rather, yellow pine growth was linked to temperature, precipitation, and PDSI. February and March temperature and May precipitation were the only variables significant in both analyses, suggesting that winter mean minimum temperatures and spring precipitation are most important for yellow pine growth at this site.

5.3.2 Moving Correlation Analysis

Moving correlation analysis revealed some inconsistencies in the climate response of yellow pines at Cooper Road Trail over time. The positive relationship with March temperature was relatively stable throughout the period of analysis (1910–2007) (Figure

5.11), as was the positive relationship between May precipitation and growth (Figure

5.12). However, the response to several other climate variables changed over time

(Figures 5.11, 5.12, and 5.13). In the first half of the 20th century, summer temperatures were inversely related to growth, but the correlation weakened around 1950. Also, the

Figure 5.11 Results of moving correlation analysis between temperature and the Cooper Road Trail chronology, using 36-year moving intervals. Monthly temperature variables are listed on the y-axis, beginning with previous June in the lower left corner and ending with current November in the upper left corner. Last years of moving intervals are listed on the x-axis. For example, a grid square marked 1946 represents the period from 1910 to 1946. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient.

Figure 5.12 Results of moving correlation analysis between monthly precipitation and the Cooper Road Trail chronology, using 36-year moving intervals. Monthly variables are listed on the y-axis, beginning with previous June in the lower left corner and ending with current November in the upper left corner. Last years of moving intervals are listed on the x-axis. For example, a grid square marked 1946 represents the period from 1910 to 1946. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient.

Figure 5.13 Results of correlation analysis between monthly PDSI and the Cooper Road Trail chronology, using 36-year moving intervals. Monthly variables are listed on the y-axis, beginning with previous June in the lower left corner and ending with current November in the upper left corner. Last years of moving intervals are listed on the x-axis. For example, a grid square marked 1946 represents the period from 1910 to 1946. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient.

78 moving analysis showed no relationship between winter temperatures and growth prior to

1960. From the 1970s to 2000, the response to winter temperature remained strong

(Figure 5.11). Relationships with April and October precipitation also varied. From

1895 to about 1960, April and October precipitation were significantly related to growth, but after 1950, the relationships weakened and became insignificant (Figure 5.12). The moderately strong correlations with PDSI were relatively stable over time, but during a brief period (1955–1970), growth was not associated with PDSI. Around 1970, the correlations with growing season PDSI again strengthened and remained moderately strong until 2007 (Figure 5.13).

5.3.3 Ocean-Atmospheric Teleconnections

Tree growth at Cooper Road Trail was significantly positively correlated with

NAO but was not associated with Atlantic Ocean SSTAs, PDO index values, or SOI

(Figure 5.14). January (r = 0.35, P < 0.05) and February (r = 0.38, P < 0.05) NAO were the most significant variables, suggesting that positive phases of the NAO (reduced sea level pressure differentials between the Azores High and Icelandic Low) tended to coincide with increased tree growth, while negative phases tended to coincide with below average growth. Previous and current August NAO values were also weakly positively correlated with growth. Atlantic Ocean SSTAs, PDO, and SOI were insignificant in all months in correlation analysis. However, correlations with previous year values of these three variables were actually stronger than correlations with current year values, suggesting that climatic or biological persistence may have caused a lag in the growth response to climate oscillations.

0.45 A 0.35 0.25 0.15 0.05 -0.05 -0.15 J J A S O N D J F M A M J J A S O N -0.25 0.45 B * 0.35 * * 0.25 0.15 0.05 -0.05 -0.15 -0.25 J J A S O N D J F M A M J J A S O N

0.45 C 0.35

0.25

0.15

0.05

-0.05 J J A S O N D J F M A M J J A S O N -0.15

-0.25 0.45 D 0.35 0.25 0.15

0.05 -0.05 J J A S O N D J F M A M J J A S O N -0.15 -0.25 Figure 5.14 Bootstrapped correlation coefficients between monthly oscillation index values and the Cooper Road Trail chronology (1930–2007). A: Atlantic Ocean SSTAs. B: NAO index. C: PDO index. D: Southern Oscillation Index (SOI) (P < 0.05).

5.3.4 Wavelet Analysis

Like the Gold Mine Trail chronology, yellow pine growth at Cooper Road Trail exhibited a low-frequency 55–70 year oscillation that was strongest in the middle of the chronology (1900–2000) (Figure 5.15). Two shorter oscillations (5–10 year and 15–30 year) were also present in the latter half of the 20th century but were not dominant. The simultaneous emergence of the higher frequency patterns and the decline of the low frequency oscillation may indicate a change in either climate itself or the response to climate by yellow pines at Cooper Road Trail around 1950.

5.4 Shaw Grave Gap

5.4.1 Temperature, Precipitation, and PDSI

The Shaw Grave Gap chronology was significantly correlated with January (r =

0.48, P < 0.05) and February minimum temperatures (r = 0.49, P < 0.05) (Table 5.2;

Figure 5.16). Also, response function analysis identified these monthly variables as significant, along with March temperature (Figure 5.17). These results were similar to the climate-growth relationships at Gold Mine Trail. At Shaw Grave Gap, correlation analysis and response function analysis both suggested that temperature was the main driver of growth, but other monthly variables were related to growth as well. February and May precipitation were weakly but significantly correlated with growth, as were previous December to current November PDSI values (Figure 5.16). February precipitation was the only non-temperature variable identified as significant in response function analysis. The highest correlations with PDSI were in late winter and spring

(February to June). While growth at Shaw Grave Gap was weakly related to precipitation

Figure 5.15 Results of wavelet analysis of the Cooper Road Trail chronology (1878– 2007): (a) the standard chronology at Cooper Road Trail, (b) wavelet power spectra for the chronology, (c) the global wavelet power spectrum. Crosshatches represent the cone of uncertainty. Black lines outline the significant periodicities. All analyses used the Morlet wavelet and zero-padded series (Torrence and Compo 1998).

0.55 * * 0.45 * * * * 0.35 * * * * * * * * * * * * 0.25 *

0.15

0.05

-0.05

-0.15 J J A S O N D J F M A M J J A S O N -0.25 Temp. Precip. PDSI

Figure 5.16 Bootstrapped correlations between the Shaw Grave Gap chronology and monthly mean minimum temperature, precipitation, and PDSI from the previous June to the current November (1930–2007) (P < 0.05).

0.35 * * * 0.25 *

0.15

0.05

-0.05 J J A S O N D J F M A M J J A S O N

-0.15

-0.25 Temp. Precip. PDSI

Figure 5.17 Bootstrapped response function coefficients at Shaw Grave Gap (1930– 2007). January, February, and March temperatures and February precipitation were the strongest predictors of growth (P < 0.05).

83 and PDSI, winter mean minimum temperatures (January, February, and March) exerted the strongest influence on tree growth since 1930.

5.4.2 Moving Correlation Analysis

At Shaw Grave Gap, no monthly climate variables remained significant throughout the entire 20th century in moving correlation analysis. The most stable variables were January and February temperature (Figure 5.18). Prior to 1930, these monthly variables were insignificant, possibly due to unreliable observations in the early part of the record. Since about 1935, mean minimum temperatures in January and

February were significantly positively correlated with growth at Shaw Grave Gap (Figure

5.18). Although winter temperature was significant throughout most of the period of analysis, the relationship grew even stronger in the latter half of the 20th century. Also, toward the end of the chronology, previous November temperature emerged as significantly inversely correlated to growth, indicating a possible change in the climate response of yellow pines at Shaw Grave Gap around this time (Figure 5.18).

The relationship between tree growth and precipitation was less stable than the temperature-tree growth relationship. Precipitation in several months was significant only for brief periods, likely based on chance rather than an actual relationship with growth (Figure 5.19). In response function analysis, February precipitation was identified as a driver of growth, but the moving correlation analysis showed that the relationship weakened around 1950 and remained weak for the rest of the century.

Similarly, May precipitation was significantly positively correlated with growth until mid-century. The correlation then weakened before re-intensifying around 1980. The

Figure 5.18 Results of moving correlation analysis between temperature and the Shaw Grave Gap chronology, using 36-year moving intervals. Monthly

temperature variables are listed on the y-axis, beginning with previous June in the lower left corner and ending with current November in the upper left corner. Last years of moving intervals are listed on the x-axis. For example, a grid square marked 1946 represents the period from 1910 to 1946. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient.

Figure 5.19 Results of moving correlation analysis between monthly precipitation and the Shaw Grave Gap chronology, using 36-year moving intervals. Monthly variables are listed on the y-axis, beginning with previous June in the lower left corner and ending with current November in the upper left corner. Last years of moving intervals are listed on the x-axis. For example, a grid square marked 1946 represents the period from 1910 to 1946. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient.

86 unstable relationship between monthly precipitation and tree growth suggests that precipitation was not a primary driver of growth, but did influence growth during certain periods. The relationship with PDSI was similar. Although correlations with PDSI were significant in several months, the relationship was not stable over time. In the mid-20th century, relationship with PDSI weakened for several decades (1960–1980) before becoming significant again (Figure 5.20). Overall, the temperature-tree growth relationship was relatively stable at Shaw Grave Gap, while the relationships with precipitation and PDSI there were not consistent over time.

5.4.3 Ocean-Atmospheric Teleconnections

The Shaw Grave Gap chronology was not strongly correlated with any of the four teleconnection variables. A weak positive relationship existed between Atlantic Ocean

SSTAs and growth, particularly during the late summer and fall (both previous and current) (Figure 5.21). Winter NAO index values were also associated with growth.

Both January (r = 0.31, P < 0.05) and February NAO (r = 0.31, P < 0.05) were significantly positively correlated with annual growth. Positive phases of NAO tended to occur during years of above average growth, while negative phases tended to coincide with below average annual growth. No monthly PDO or SOI values were significant in correlation analysis, indicating that Pacific Ocean SSTAs and pressure patterns did not have a strong influence tree growth at this site. Despite the insignificant relationships, however, PDO was weakly negatively correlated with growth, while ENSO was weakly positively correlated with growth (Figure 5.21).

Figure 5.20 Results of correlation analysis between monthly PDSI and the Shaw Grave Gap chronology, using 36-year moving intervals. Monthly variables are listed on the y-axis, beginning with previous June in the lower left corner and ending with current November in the upper left corner. Last years of moving intervals are listed on the x-axis. For example, a grid square marked 1946 represents the period from 1910 to 1946. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient.

0.45 A 0.35 * * * * * * 0.25 * 0.15

0.05

-0.05 J J A S O N D J F M A M J J A S O N -0.15

-0.25 0.45 B 0.35 * * 0.25

0.15

0.05

-0.05

-0.15 J J A S O N D J F M A M J J A S O N -0.25 0.45 C 0.35

0.25

0.15

0.05

-0.05

-0.15 J J A S O N D J F M A M J J A S O N -0.25

0.45 D 0.35

0.25

0.15

0.05

-0.05 J J A S O N D J F M A M J J A S O N -0.15

-0.25 Figure 5.21 Bootstrapped correlation coefficients between monthly oscillation index values and the Shaw Grave Gap chronology (1930–2007). A: Atlantic Ocean SSTAs. B: NAO index. C: PDO index. D: Southern Oscillation Index (SOI) (P < 0.05).

5.4.4 Wavelet Analysis

Wavelet analysis of the Shaw Grave Gap chronology showed several oscillations of varying frequencies present over the past two centuries (Figure 5.22). A low-frequency

60–80 year oscillation was dominant through most of the chronology, particularly during the 20th century. The early part of the chronology exhibited a 20–30 year oscillation, but this periodicity faded around 1915. It re-intensified around 1955, before again fading at the end of the century. From about 1980 to 2000, a short, higher frequency 5–10 year oscillation was also present.

5.5 Maynard Creek

5.5.1 Temperature, Precipitation, and PDSI

The Maynard Creek chronology had moderately strong correlations with both mean minimum temperatures and PDSI (Figure 5.23). The highest correlations were between spring and summer PDSI and tree growth, although PDSI was significant (P <

0.05) from the previous June through the current November (Table 5.2). In particular,

May (r = 0.53, P < 0.05) and June PDSI (r = 0.50, P < 0.05) were most correlated with growth, suggesting that late spring moisture availability was an important driver of growth. Despite the strong correlations, however, PDSI was not a significant predictor of growth in any month based on response function analysis (Figure 5.24). Rather, this analysis identified winter temperatures as positive predictors of tree growth. Winter mean minimum temperatures were also moderately correlated with growth (January: r =

0.42, P < 0.05; February: r = 0.40, P < 0.05; March: r = 0.29, P < 0.05) (Figure 5.23). In

Figure 5.22 Results of wavelet analysis of the Shaw Grave Gap chronology (1860– 2007): (a) the standard chronology at Shaw Grave Gap, (b) wavelet power spectra for the chronology, (c) the global wavelet power spectrum. Crosshatches represent the cone of uncertainty. Black lines outline the significant periodicities. All analyses used the Morlet wavelet and zero-padded series (Torrence and Compo 1998).

0.55 * * * * 0.45 * * * * * * * * * * * 0.35 * * * * * * * * 0.25 *

0.15

0.05

-0.05

-0.15 J J A S O N D J F M A M J J A S O N -0.25 Temp. Precip. PDSI

Figure 5.23 Bootstrapped correlations between the Maynard Creek chronology and monthly mean minimum temperature, precipitation, and PDSI from the previous June to the current November (1930–2007) (P < 0.05).

0.45

0.35

* * 0.25 * * *

0.15

0.05

-0.05

-0.15 J J A S O N D J F M A M J J A S O N

-0.25 Temp. Precip. PDSI Figure 5.24 Bootstrapped response function coefficients at Maynard Creek (1930– 2007). January, February, and March temperatures and February and May precipitation were the strongest predictors of growth (P < 0.05).

92 addition, February and May precipitation were significant in response function analysis, but were relatively weakly correlated with growth (February: r = 0.26, P < 0.05; May: r =

0.37, P < 0.05). Because response function analysis uses principal components multiple regression to eliminate the effects of interdependence among variables, PDSI was likely eliminated as a predictor because it is inherently autocorrelated, and it is also associated with precipitation and temperature. Overall, winter temperature and spring precipitation and moisture availability most strongly affected yellow pine growth at Maynard Creek, but growth was not driven by a single variable or one particular month or season.

5.5.2 Moving Correlation Analysis

Moving correlation analysis showed that the response to temperature and precipitation at Maynard Creek changed during the 20th century, while the response to current year PDSI was relatively stable. Winter mean minimum temperatures were not significantly correlated with growth before about 1955, but they became significant and remained so throughout the rest of the 20th century (Figure 5.25). In addition, moving correlation analysis suggested a strong negative relationship between previous November temperatures and growth beginning around 1970. Although previous November temperature was not significant in the stationary correlation analysis or response function analysis, its relationship with growth strengthened in the last 30–40 years of the chronology. A similar trend was identified at Shaw Grave Gap.

The relationship between precipitation and growth at Maynard Creek was also unstable over time (Figure 5.26). Although February and May precipitation were

Figure 5.25 Results of moving correlation analysis between temperature and the Maynard Creek chronology, using 36-year moving intervals. Monthly temperature variables are listed on the y-axis, beginning with previous June in the lower left corner and ending with current November in the upper left corner. Last years of moving intervals are listed on the x-axis. For example, a grid square marked 1946 represents the period from 1910 to 1946. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient.

Figure 5.26 Results of moving correlation analysis between monthly precipitation and the Maynard Creek chronology, using 36-year moving intervals. Monthly variables are listed on the y-axis, beginning with previous June in the lower left corner and ending with current November in the upper left corner. Last years of intervals are listed on the x-axis. For example, a grid square marked 1946 represents the period from 1910 to 1946. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient.

95 identified as significant influences of growth by response function analysis, they were not consistently significantly correlated with growth throughout the 20th century. The relationships with precipitation in both months weakened in the 1950s. Precipitation in several other months was significant during brief periods of time (1–5 years), but these may have been spurious (significant based on chance alone). The relationship between growth and growing season PDSI was relatively stable throughout the 20th century, but it did weaken briefly between 1955 and 1980 (Figure 5.27).

5.5.3 Ocean-Atmospheric Teleconnections

Correlation analysis indicated that yellow pine growth at Maynard Creek was related to winter NAO index values (Figure 5.28). In particular, February NAO index values were moderately correlated with growth (r = 0.39, P < 0.05). Along with

February, January NAO (r = 0.31, P < 0.05) was also positively correlated with the standard chronology. Growth at Maynard Creek was not significantly associated with

Atlantic Ocean SSTAs, PDO, or SOI values (Figure 5.28).

5.5.4 Wavelet Analysis

Wavelet analysis of the Maynard Creek chronology showed no dominant oscillations prior to the mid-20th century (Figure 5.29). Around 1950, a 20–30 year oscillation emerged and was dominant for the rest of the century. A higher-frequency, 6–

10 year periodicity was also present from about 1975 to 2005. Unlike several of the other chronologies, tree growth at Maynard Creek did not exhibit a low frequency multidecadal oscillation of over 50 years (Figure 5.29).

Figure 5.27 Results of correlation analysis between monthly PDSI and the Maynard Creek chronology, using 36-year moving intervals. Monthly variables are listed on the y-axis, beginning with previous June in the lower left corner and ending with current November in the upper left corner. Last years of moving intervals are listed on the x-axis. For example, a grid square marked 1946 represents the period from 1910 to 1946. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient.

0.45 A 0.35 0.25

0.15

0.05

-0.05 J J A S O N D J F M A M J J A S O N -0.15

-0.25

0.45 B * 0.35 * 0.25 0.15 0.05 -0.05 -0.15 J J A S O N D J F M A M J J A S O N -0.25 0.45 C 0.35 0.25 0.15 0.05 -0.05 -0.15 J J A S O N D J F M A M J J A S O N -0.25

0.45 D 0.35

0.25 *

0.15

0.05

-0.05 J J A S O N D J F M A M J J A S O N -0.15

-0.25 Figure 5.28 Bootstrapped correlation coefficients between monthly oscillation index values and the Maynard Creek chronology (1930–2007). A: Atlantic Ocean SSTAs. B: NAO index. C: PDO index. D: Southern Oscillation Index (SOI) (P < 0.05).

Figure 5.29 Results of wavelet analysis of the Maynard Creek chronology (1895– 2007): (a) the standard chronology at Maynard Creek, (b) wavelet power spectra for the chronology, (c) the global wavelet power spectrum. Crosshatches represent the cone of uncertainty. Black lines outline the significant periodicities. All analyses used the Morlet wavelet and zero-padded series (Torrence and Compo 1998).

5.6 Gregory Ridge Trail

5.6.1 Temperature, Precipitation, and PDSI

Tree growth at Gregory Ridge Trail was strongly correlated with winter and growing season PDSI (Figure 5.30). The most significant months were February (r =

0.46, P < 0.05), May (r = 0.40, P < 0.05), June (r = 0.39, P < 0.05), July (r = 0.42, P <

0.05), and August (r = 0.39, P < 0.05) (Table 5.2). Despite the strong relationship, PDSI was not significant in response function analysis, likely because the autocorrelation between monthly values was removed (Figure 5.31). Correlations with temperature were weakly significant in several months, including January (r = 0.29, P < 0.05), February (r

= 0.34, P < 0.05), and March (r = 0.26, P < 0.05). These correlations suggested a winter temperature signal, similar to but weaker than the signals identified in the Shaw Grave

Gap and Gold Mine Trail chronologies. October temperature was also significant in correlation analysis (r = 0.26, P < 0.05) (Figure 5.30). Response function analysis identified March and October temperature as significant, as well as February precipitation

(Figure 5.41). February precipitation was also positively correlated with growth (r =

0.30, P < 0.05). The results from correlation analysis differed from the results of response function analysis. While correlation analysis suggested a strong link between growing season PDSI and growth, response function analysis identified winter and fall temperature and late winter precipitation as the primary climatic drivers of yellow pine growth at Gregory Ridge Trail. This indicates that both growing season moisture conditions and winter mean minimum temperatures influence growth.

100

0.55 * 0.45 * * * * * * * 0.35 * * * * * * * * * * * 0.25 * * *

0.15

0.05

-0.05 J J A S O N D J F M A M J J A S O N -0.15

-0.25 * Temp. Precip. PDSI

Figure 5.30 Bootstrapped correlations between the Gregory Ridge Trail chronology and monthly mean minimum temperature, precipitation, and PDSI from the previous June to the current November (1930–2005) (P < 0.05).

0.45

0.35

* 0.25 * * * 0.15

0.05

-0.05 J J A S O N D J F M A M J J A S O N

-0.15

-0.25 Temp. Precip. PDSI Figure 5.31 Bootstrapped response function coefficients at Gregory Ridge Trail (1930– 2005). February precipitation and February, March, and October temperatures were significant predictors of growth (P < 0.05).

101

5.6.2 Moving Correlation Analysis

Moving correlation analysis indicated several mid-century shifts in the climate- tree growth relationship at Gregory Ridge Trail. In this analysis, the temperature-tree growth relationship was insignificant prior to about 1950–1960 (Figure 5.32). Around mid-century, the positive correlations between growth and January, February, and

October temperatures began to strengthen and were significant throughout the rest of the analysis. Similarly, previous November temperature became significantly negatively correlated with growth around 1950 (Figure 5.32).

The growth relationships with precipitation and PDSI also shifted mid-century.

February precipitation was not significantly correlated with growth in the first half of the century, but became significant and remained so throughout the latter half of the century

(Figure 5.33). As the relationship with February precipitation strengthened, previous

October and current May and July precipitation faded and were no longer significantly correlated with growth in the second half of the century. Also, moving correlations with

PDSI showed a 10–20 year period in the middle of the 20th century during which PDSI was insignificant (Figure 5.34). Similar trends were identified at Gold Mine Trail,

Cooper Road Trail, and Shaw Grave Gap. Around 1970, PDSI again became significantly positively correlated with growth, but the significant months shifted to earlier in the growing season. In the first half of the century, summer and fall PDSI were most associated with growth, but since 1970, winter and spring PDSI have been most significant.

102

Figure 5.32 Results of moving correlation analysis between temperature and the Gregory Ridge Trail chronology, using 36-year moving intervals. Monthly temperature variables are listed on the y-axis, beginning with previous June in the lower left corner and ending with current November in the upper left corner. Last years of moving intervals are listed on the x-axis. For example, a grid square marked 1946 represents the period from 1910 to 1946. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient.

103

Figure 5.33 Results of moving correlation analysis between monthly precipitation and the Gregory Ridge Trail chronology, using 36-year moving intervals. Monthly variables are listed on the y-axis, beginning with previous June in the lower left corner and ending with current November in the upper left corner. Last years of moving intervals are listed on the x-axis. For example, a grid square marked 1946 represents the period from 1910 to 1946. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient.

104

Figure 5.34 Results of correlation analysis between monthly PDSI and the Gregory Ridge Trail chronology, using 36-year moving intervals. Monthly variables are listed on the y-axis, beginning with previous June in the lower left corner and ending with current November in the upper left corner. Last years of moving intervals are listed on the x-axis. For example, a grid square marked 1946 represents the period from 1910 to 1946. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient.

105

5.6.3 Ocean-Atmospheric Teleconnections

The Gregory Ridge Trail chronology was not associated with SOI or PDO index values, suggesting that Pacific Ocean climate variability did not have a strong effect on yellow pine growth at this site (Figure 5.35). However, NAO index values and Atlantic

Ocean SSTAs were significantly correlated with growth. This indicates that Atlantic

Ocean temperature and pressure patterns likely influence tree growth at Gregory Ridge

Trail. SSTAs from the previous August (r = 0.42, P < 0.05), September (r = 0.44, P <

0.05), October (r = 0.46, P < 0.05), and November (r = 0.48, P < 0.05) were significantly correlated with annual growth (Figure 5.35). Current fall SSTAs were also moderately positively correlated with growth. The weakest correlations with SSTAs were during the current spring and summer.

NAO index values were also significantly correlated to growth (Figure 5.35). In the previous November (r = –0.20, P < 0.05) and current October (r = –0.41, P < 0.05) and November (r = –0.41, P < 0.05), NAO was inversely related to growth. In January and February, however, the relationship between NAO and growth was significant and positive. The correlation results suggested that NAO might have affected growth differently in winter and fall.

5.6.4 Wavelet Analysis

From 1830 to 1890, yellow pine growth at Gregory Ridge Trail did not oscillate at any particular frequency (Figure 5.36). However, the wavelet power spectrum showed a low-frequency 60–80 year oscillation present from about 1890 to 2000. In the latter half of the 20th century, this low-frequency oscillation shortened to a 40–70 year oscillation.

106

0.55 A * * * * * * * * * * * * * 0.35 * * * * * 0.15

-0.05 J J A S O N D J F M A M J J A S O N

-0.25

-0.45 0.55 B 0.35 * * 0.15

-0.05

-0.25 * J J A S O N D J F M A M J J A S O N -0.45 * 0.55 * C 0.35

0.15

-0.05

-0.25 J J A S O N D J F M A M J J A S O N -0.45

0.55 D

0.35

0.15

-0.05 J J A S O N D J F M A M J J A S O N -0.25

-0.45 Figure 5.35 Bootstrapped correlation coefficients between monthly oscillation index values and the Gregory Ridge Trail chronology (1930–2005). A: Atlantic Ocean SSTAs. B: NAO index. C: PDO index. D: Southern Oscillation Index (SOI) (P < 0.05).

107

Figure 5.36 Results of wavelet analysis of the Gregory Ridge Trail chronology (1830–2005): (a) the standard chronology at Gregory Ridge Trail, (b) wavelet power spectra for the chronology, (c) the global wavelet power spectrum. Crosshatches represent the cone of uncertainty. Black lines outline the significant periodicities. All analyses used the Morlet wavelet and zero-padded series (Torrence and Compo 1998).

108

The oscillation faded around 2000, likely because of the erratic growth at the end of the chronology.

5.7 Composite Chronology

5.7.1 Temperature, Precipitation, and PDSI

After climate analysis of each individual site chronology, the index values of the five chronologies were averaged to create a composite chronology. The composite chronology exhibited strong correlations with January (r = 0.45, P < 0.05), February (r =

0.45, P < 0.05), and March (r = 0.32, P < 0.05) mean minimum temperatures (Table 5.2;

Figure 5.37). January, February, and March temperatures were also significant in response function analysis (Figure 5.38). Precipitation was significantly correlated with growth in February (r = 0.29, P < 0.05) and May (r = 0.34, P < 0.05) (Figure 5.37).

These months were weakly significant in response function analysis as well. Growth was strongly positively correlated with PDSI, particularly in February (r = 0.45, P < 0.05),

May (r = 0.45, P < 0.05), June (r = 0.41, P < 0.05), and July (r = 0.39, P < 0.05).

Together, the results of correlation and response function analyses suggest that this network of chronologies has a strong winter temperature signal, as well as a weaker, but still significant, growing season precipitation and moisture availability signal.

5.7.2 Moving Correlation Analysis

Moving correlation analysis of the composite chronology exhibited several of the same patterns found in the individual chronologies. Winter temperatures were not significantly correlated with growth in the early part of the 20th century, but became

109

0.55

* 0.45 * * * * * * * * * * * 0.35 * * * * * * * 0.25 * *

0.15

0.05

-0.05 J J A S O N D J F M A M J J A S O N -0.15

-0.25 Temp. Precip. PDSI

Figure 5.37 Bootstrapped correlation coefficients between the composite chronology and monthly mean minimum temperature, precipitation, and PDSI from the previous June to the current November (1930–2007) (P < 0.05).

0.35

* * * 0.25 * * 0.15

0.05

-0.05

-0.15 J J A S O N D J F M A M J J A S O N

-0.25 Temp. Precip. PDSI

Figure 5.38 Bootstrapped response function coefficients for the composite chronology (1930–2007). January, February, March temperature, and February and May precipitation were significant predictors of growth (P < 0.05).

110 significant around 1950 (Figure 5.39). From 1910 to 1935, June temperature was significantly negatively correlated with growth, but the relationship weakened mid- century. In recent decades, previous November temperature and current October temperature have emerged as significant. In terms of precipitation, May, July, and previous October were significant months, but these relationships also faded mid-century

(Figure 5.40). In the 1960s, February precipitation emerged as a significant climate variable. Shortly after, the relationship between growth and May precipitation re- intensified and again became significant. The relationship with growing season PDSI was strong until about 1960, when the correlations became insignificant (Figure 5.41).

PDSI remained insignificant until about 1980, when it reemerged as significantly positively correlated with tree growth.

5.7.3 Ocean-Atmospheric Teleconnections

The composite chronology was most strongly associated with Atlantic Ocean

SSTAs and NAO index values. Growth was significantly positively correlated with

SSTAs during both the previous and current growing season (Figure 5.42). The composite chronology was also positively correlated to January (r = 0.33, P < 0.05) and

February (r = 0.36, P < 0.05) NAO index values. Previous November (r = –0.23, P <

0.05) and current August (r = 0.23, P < 0.05) NAO index values were weakly significant.

The composite chronology was not significantly associated with PDO index values, but the relationship tended negative. Conversely, the relationship between SOI values and growth was weakly positive. July (r = 0.22, P < 0.05) was the only month in which SOI was significantly correlated to the composite chronology.

111

Figure 5.39 Results of correlation analysis between monthly mean minimum temperature and the composite chronology, using 36-year moving intervals. Monthly variables are listed on the y-axis, beginning with previous June in the lower left corner and ending with current November in the upper left corner. Last years of moving intervals are listed on the x-axis. For example, a grid square marked 1946 represents the period from 1910 to 1946. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient.

112

Figure 5.40 Results of moving correlation analysis between monthly precipitation and the composite chronology, using 36-year moving intervals. Monthly variables are listed on the y-axis, beginning with previous June in the lower left corner and ending with current November in the upper left corner. Last years of moving intervals are listed on the x-axis. For example, a grid square marked 1946 represents the period from 1910 to 1946. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient.

113

Figure 5.41 Results of correlation analysis between monthly PDSI and the composite chronology, using 36-year moving intervals. Monthly variables are listed on the y-axis, beginning with previous June in the lower left corner and ending with current November in the upper left corner. Last years of moving intervals are listed on the x-axis. For example, a grid square marked 1946 represents the period from 1910 to 1946. Intervals during which a variable was not significantly correlated to tree growth are shaded green, while significant correlations are colored according to the strength of the correlation coefficient.

114

0.45 A 0.35 * * * * * * * * * * 0.25 *

0.15

0.05

-0.05 J J A S O N D J F M A M J J A S O N -0.15 -0.25 0.45 B * 0.35 *

0.25 * 0.15 0.05

-0.05 J J A S O N D J F M A M J J A S O N -0.15

-0.25 * 0.45 C 0.35 0.25 0.15 0.05

-0.05 -0.15 J J A S O N D J F M A M J J A S O N -0.25 0.45 D 0.35 0.25 * 0.15 0.05

-0.05 J J A S O N D J F M A M J J A S O N -0.15 -0.25 Figure 5.42 Bootstrapped correlation coefficients between monthly oscillation index values and the composite chronology (1930–2005). A: Atlantic Ocean SSTAs. B: NAO index. C: PDO index. D: Southern Oscillation Index (SOI) (P < 0.05).

115

5.7.4 Wavelet Analysis

Wavelet analysis of the composite chronology (1850–2007) showed a 64–80 year oscillation present in the data (Figure 5.43). This low-frequency oscillation was the only significant periodicity identified. It was relatively constant throughout the chronology, but was strongest from about 1880 to 1980.

116

Figure 5.43 Results of wavelet analysis of the composite chronology (1850–2007): (a) the composite chronology, (b) wavelet power spectra for the chronology, (c) the global wavelet power spectrum. Crosshatches represent the cone of uncertainty. Black lines outline the significant periodicities. All analyses used the Morlet wavelet and zero- padded series (Torrence and Compo 1998).

117

CHAPTER 6 DISCUSSION OF YELLOW PINE CLIMATE-GROWTH RELATIONSHIPS

6.1 Growth Response to Climate

Despite their close proximity and similar site characteristics, the five yellow pine chronologies from Great Smoky Mountains National Park showed slightly different responses to 20th century climate. In an earlier study, Grissino-Mayer et al. (2007) identified a winter temperature signal in a preliminary chronology from Gold Mine Trail, but no strong precipitation or PDSI signal. Based on these results, I expected to find a winter temperature signal at each of the five sites. At Gold Mine Trail, winter temperature was indeed the dominant variable to which yellow pines responded. The

Shaw Grave Gap chronology also showed a strong winter temperature signal. At

Maynard Creek, winter temperature was again the dominant variable, but the response to spring precipitation was strong as well. The Gregory Ridge Trail chronology showed a winter temperature signal, a winter precipitation signal, and a fall temperature signal.

These signals were of roughly equal strength. At Cooper Road Trail, spring precipitation was the dominant variable to which trees responded, but a relatively strong winter temperature signal was present as well. Overall, this network of chronologies exhibited a winter temperature signal and a weaker late winter/spring precipitation signal.

6.1.1 Growth Response to Temperature

All five of the chronologies indicated some degree of sensitivity to winter mean minimum temperatures, although using monthly rather than seasonal data may have weakened the winter temperature signal. Seasonal climate data usually have stronger

118 relationships with tree growth than monthly data (Blasing et al. 1981), because trees often respond to climate patterns for extended periods of sequential months. The western side of Great Smoky Mountains National Park is the driest part of the park, but the temperature response was still greater than the precipitation or PDSI response at Gold

Mine Trail, Shaw Grave Gap, and Maynard Creek. At Cooper Road Trail and Gregory

Ridge Trail, the temperature response was only moderately strong. The composite chronology, an average of the five individual site chronologies, exhibited a strong winter temperature signal and weaker growing season precipitation and PDSI signals.

Although Grissino-Mayer et al. (2007) identified a strong winter temperature signal at Gold Mine Trail, few other studies have found yellow pines in the Southeast to be primarily sensitive to winter temperature. In northern Georgia, Grissino-Mayer and

Butler (1993) found growing season precipitation to be the primary climatic driver of shortleaf pine growth. They also identified an inverse relationship between growth and temperature through much of the growing season (April, June–September). Interestingly, growing season temperature did not influence growth at any of the five sites in Great

Smoky Mountains National Park, perhaps because summer high temperatures are less extreme or because soil moisture is more abundant in the Great Smoky Mountains than in the Piedmont region of northern Georgia. Despite the primary role of precipitation,

Grissino-Mayer and Butler (1993) also identified a positive relationship between January temperature and growth. However, this relationship was not as strong as the winter temperature response in Great Smoky Mountains National Park. In central North

Carolina, Friend and Hafley (1989) found late winter/early spring mean monthly air temperature to affect shortleaf pine growth. Stambaugh and Guyette (2004) found that

119 winter minimum temperatures affected growth of shortleaf pine in the Ozarks of

Arkansas and southern Missouri. This relationship was also evident in the five chronologies from Great Smoky Mountains National Park. Warmer winters cause yellow pines to break dormancy earlier in the spring and thus put on more wood during the growing season (Fritts 1976). Minimum winter temperatures are likely more important than mean winter temperatures because a certain temperature threshold must be reached for the trees to break dormancy (Perry 1971).

Growth at Gold Mine Trail and Gregory Ridge Trail was positively influenced by

October temperature. The other three chronologies showed a similar but statistically insignificant relationship. Warmer Octobers lead to wider annual growth rings by lengthening the growing season and delaying the onset of dormancy. Phenologically, the positive association between growth and mean minimum October temperature implies that yellow pines in Great Smoky Mountains National Park do not reach true dormancy until late October or November.

Unlike October temperature, previous November temperature negatively affected growth at Gold Mine Trail and Gregory Ridge Trail. Similar patterns existed at Shaw

Grave Gap and Maynard Creek, but the relationships were insignificant. Warmer previous Novembers were associated with decreased annual growth, while cooler previous Novembers tended to coincide with increased growth. Grissino-Mayer and

Butler (1993) found a weaker version of this response in shortleaf pines in northern

Georgia as well. This relationship is an example of biological persistence or preconditioning, related to food storage and the timing of cambial dormancy and reactivation (Perry 1971; Fritts 1976). During fall, trees accumulate food and energy to

120 use when cambial activity begins again in the early spring. A warm November leads to increased production of growth hormones and thus new growth, which ultimately takes away from the carbohydrate reserves to be used in the spring during cambial reactivation.

If a warm November is followed by a cold December, cambial dormancy may be stimulated rapidly, without adequate time for storage and preparation (Perry 1971; Fritts

1976). Because of biological persistence, cooler temperatures in previous months tend to favor yellow pine growth in the following year.

6.1.2 Growth Response to Precipitation and PDSI

Overall, the growth responses to precipitation and PDSI were weaker than the response to temperature, but still significant. Growing season PDSI (mainly spring and early summer) was positively related to growth. High PDSI values indicate low drought stress and abundant soil moisture. These conditions are particularly beneficial in the spring and early summer, when trees put on most of their annual cambial growth (Fritts

1976). Precipitation also affected yellow pine growth at several sites. February and May precipitation influenced growth at Gregory Ridge Trail, Maynard Creek, and Cooper

Road Trail, and were positively but insignificantly related to growth at Gold Mine Trail and Shaw Grave Gap. These results differed from those of several previous studies. In northern Georgia, shortleaf pines responded to spring and summer (May–July) precipitation (Grissino-Mayer and Butler 1993). Similarly, Friend and Hafley (1989) found that late growing season soil moisture content influenced shortleaf pine growth in central North Carolina. In the Ozarks, Stambaugh and Guyette (2004) identified May,

June, and July PDSI as the strongest influences on shortleaf pine growth.

121

The responses to precipitation and PDSI were much weaker in Great Smoky

Mountains National Park. Winter (February) and spring (May) precipitation were significant in this study, while late growing season precipitation and PDSI were more influential in other studies. A potential explanation for this is that the sites in Great

Smoky Mountains National Park are relatively resistant to drought, despite their location on the drier side of the park. Another possible explanation is that soil moisture is relatively stable throughout the year, as long as precipitation is abundant in the late winter (February) and spring (May).

6.1.3 Differences in the Climate Response Between Sites

Differences in the climate response between the five sites in Great Smoky

Mountains National Park support the idea that there is a geography to climate-tree growth relationships (Fritts 1976). Yellow pines at Gold Mine Trail and Shaw Grave Gap were the most responsive to temperature and least responsive to precipitation and PDSI, while trees at Cooper Road Trail and Gregory Ridge Trail were the most responsive to precipitation and PDSI and least responsive to temperature. The Maynard Creek chronology exhibited a relatively weak response to precipitation, PDSI, and winter temperature.

Microclimatic differences likely explain some of the variation in climate response between sites. At 700–750 m, Gregory Ridge Trail had the highest elevation of the five sites. Trees were sampled on steep, south-facing xeric slopes, and soils were relatively thin and rocky, likely with lower water storage capacity than sites with deeper, more mature soils. Cooper Road Trail had the second highest elevation (575–700 m) of the

122 five sites, though the south-facing slopes were less steep than the slopes at Gregory Ridge

Trail. Slope, elevation, and soils all affect microclimate and may account for the increased sensitivity to precipitation at Cooper Road Trail and Gregory Ridge Trail.

Gold Mine Trail, which exhibited the strongest temperature response and the weakest precipitation/PDSI response, was also the lowest in elevation, at 400–600 m. The soils at

Gold Mine Trail were fairly deep and stable, and trees were sampled on gradual rather than steep slopes. Despite its location in the drier part of the park, the thick, stable soils at Gold Mine Trail are likely able to hold more moisture than the thin, rocky soils at

Gregory Ridge Trail. These differences may at least partially explain the enhanced temperature response and muted precipitation and PDSI responses at Gold Mine Trail.

Microclimate alone, however, does not fully explain the differences in climate response between sites. For example, Shaw Grave Gap exhibited a strong winter temperature signal and a relatively weak precipitation signal. The soils at Shaw Grave

Gap were extremely thin and rocky, and trees were sampled on steep south-facing xeric slopes. The aspect, elevation, and soils were similar to Gregory Ridge Trail. Because of these site characteristics, I expected to find a precipitation or PDSI signal rather than a temperature signal at Shaw Grave Gap. Correlation and response function analyses, however, revealed that winter temperature was the primary driver of growth. While microclimate may have affected growth patterns at Shaw Grave Gap to some degree, other factors such as endogenous or exogenous disturbances, stand age, and species- specific responses may have also contributed to the temperature sensitivity. Endogenous forms of disturbance (such as tree falls) and exogenous forms (such as fire) can affect climate response. In climate analysis, disturbance may contribute to noise in the data that

123 obscures the climate signal (Cook 1987). Thus, the climate signal at Shaw Grave Gap may have been affected by disturbance or stand-level dynamics.

Also, it is well established that growth rates change as trees age (Fritts 1976), but not as much is known about changes in climate response with age. Still, age-dependent climate responses have been found in white spruce, European larch, and stone pine in

Alaska and the European Alps (Szeicz and MacDonald 1994; Szeicz and MacDonald

1995; Carrer and Urbinati 2004). Although this study did not explicitly address tree age or stand dynamics, the Gold Mine Trail chronology had the oldest samples, while the

Maynard Creek chronology consisted mainly of 100–150 year old trees. At Shaw Grave

Gap, many trees over 200 years old were present. The discrepancies in average tree age between sites may have contributed to the differing climate responses. However, this study did not explicitly measure tree age because many of the cores hit rot before the pith of a tree. Therefore, the effects of age on growth response cannot be discounted.

Although the three yellow pine species sampled for this study crossdate well, the climate response of each species may differ slightly. The Shaw Grave Gap, Gold Mine

Trail, and Maynard Creek chronologies consisted solely or mainly of shortleaf pine samples. The Cooper Road Trail chronology was mainly pitch pine with some shortleaf pine samples, and the Gregory Ridge Trail chronology was a mix of shortleaf pine, pitch pine, and Table Mountain pine. While the climate response of shortleaf pine has been studied elsewhere in the Southeast (Cleaveland 1975; Friend and Hafley 1989; Grissino-

Mayer and Butler 1993; Stambaugh and Guyette 2004), less is known about the climate- growth relationships of pitch pine. In the Virginia Piedmont, Copenheaver et al. (2002) found that pitch pine and Virginia pine growth were correlated with late fall temperature

124 and summer precipitation. This suggests that pitch pine may have a slightly different climate response than shortleaf pine, despite their high correlations when crossdating. In this study, the differing degrees of temperature and precipitation sensitivity may be influenced by microclimate, age-dependent responses, or species-specific responses.

6.1.4 Relationships to Ocean-Atmospheric Teleconnections

Yellow pine growth in Great Smoky Mountains National Park was related to ocean-atmospheric teleconnections that affect the Southeast. However, not all of the chronologies responded identically to the four variables analyzed (Atlantic Ocean

SSTAs, NAO index, PDO index, and SOI). Growth at Gold Mine Trail, Shaw Grave

Gap, and Gregory Ridge Trail was significantly positively related to Atlantic Ocean

SSTAs, while growth at Cooper Road Trail and Maynard Creek was only weakly related to SSTAs. The composite chronology was significantly correlated to Atlantic Ocean

SSTAs. Atlantic Ocean SSTAs are used in the calculation of the AMO indices. High

SSTAs occur during positive or warm phases of the AMO, while low SSTAs occur during negative or cool phases. Above average growth tended to coincide with high

SSTAs or warm phases of the AMO. This was unexpected, as warm phases lead to slight increases in drought frequency and severity in the Southeast (Kerr 2000; Enfield et al.

2001; McCabe et al. 2004). However, because yellow pine growth in the Great Smoky

Mountains National Park is not extremely drought- or moisture-sensitive, warm phases may actually favor growth through higher fall temperatures. For example, the AMO entered a cool phase around 1960, marked by low SSTAs and a slight decrease in drought risk for the Southeast (Kerr 2000). At the same time, growth became increasingly related

125 to October temperature. Through reductions in drought frequency and severity and slight increases in late growing season temperature, the AMO phase change may have contributed to the strengthening of the fall temperature response. Interestingly, previous fall AMO index values correlated better with annual growth than current growing season values. This suggests that warm Atlantic SSTs and the associated climatic effects may allow trees to accumulate and store more carbohydrate reserves for growth in the following year.

Winter NAO index values (January and February) were significantly positively related to all five of the individual site chronologies as well as the composite chronology.

NAO affects winter climate in both North America and Europe. Positive phases of the

NAO generally bring mild, wet conditions to the eastern U.S., while negative phases bring colder temperatures to the region (van Loon and Rogers 1978; Lamb and Peppler

1987). Positive phases of the NAO are characterized by increased pressure differentials between the Azores High and the Icelandic Low. Because the pressure gradient forces air from high pressure to low pressure centers, positive phases of the NAO tend to strengthen southerly winds and increase the flow of warm, moist air over the Southeast (Hurrell

1996; Hurrell and van Loon 1997). The positive relationships between winter NAO values and growth, and winter temperatures and growth, suggest that NAO may be the mechanism behind the positive growth response to winter temperatures. In other words, positive NAO phases slightly increase both winter temperatures and moisture conditions in the southern Appalachian region. In turn, these conditions encourage earlier cambial reactivation and rapid early season growth in yellow pines.

126

PDO index values and SOI values were only weakly related to yellow pine growth in Great Smoky Mountains National Park. PDO tended to be negatively but insignificantly correlated with growth, while the relationship with ENSO tended to be positive but insignificant. This weak relationship is likely because PDO does not strongly affect the climate of the Southeast on its own, though simultaneous positive phases of the PDO and positive phases of the AMO often lead to drought (McCabe et al.

2004). This relationship is reflected in the weak inverse relationship between PDO and growth. The weak positive relationship between ENSO and growth indicates that La

Niña events may slightly favor yellow pine growth. During La Niña periods, winters in the Southeast tend to be slightly warmer and wetter than average. However, the relationship between ENSO and growth was significant in July only, indicating that

Pacific Ocean pressure patterns do not usually influence yellow pine growth in the

Southeast. Still, particularly strong El Niño or La Niña events may affect growth in certain years. The relationships between tree growth and Pacific Ocean temperature and pressure patterns are weaker than the relationships between growth and Atlantic Ocean climate variability. This is likely because Great Smoky Mountains National Park is in the eastern U.S. and is influenced more by air masses that originate in the Gulf of Mexico and the Atlantic Ocean than by air masses that form over the Pacific Ocean.

Wavelet analysis showed low-frequency, 55–90 year oscillations in the Gold

Mine Trail, Cooper Road Trail, Shaw Grave Gap, and Gregory Ridge Trail chronologies, as well as the composite chronology. The Maynard Creek chronology exhibited a shorter multidecadal oscillation of 20–30 years. The AMO oscillates at a 65–80 year period, which corresponds well with the dominant frequency modes in the tree-ring data.

127

Interestingly, Maynard Creek was the site at which AMO was least correlated with growth, and the Maynard Creek chronology was also the only chronology without a low- frequency oscillation of 50 years or greater. It is likely that the low-frequency, 55–90 year oscillation present in the tree-ring data is indeed related to hemispheric-scale circulation patterns.

6.2 Temporal Consistency of Climate-Growth Relationships

Moving correlation and response function analyses revealed inconsistencies in the climate-tree growth relationship over time, most notably a shift in the mid-20th century, between 1935 and 1955. At all sites and in the composite chronology, the correlation between growth and winter temperature strengthened in the latter half of the 20th century.

As winter temperature became increasingly related to growth, summer temperature became an unimportant factor. Simultaneously, the relationships between precipitation and PDSI and tree growth weakened. Growing season precipitation (particularly May,

June, and July) influenced tree growth in the first half of the century, but the relationship began to weaken around 1950. Fall temperatures became increasingly important toward the end of the 20th century. Current October temperature became significantly positively related to growth, while previous November temperature became negatively related to growth. These relationships were both insignificant prior to about 1960, but strengthened from 1960 to the end of the century. To summarize, yellow pine growth became less associated with late spring and summer moisture conditions in the latter half of the 20th century, while winter and fall temperatures became more significant drivers of growth.

128

Several factors may have contributed to the apparent mid-century shift in the climate response. These include: an AMO phase change, decreased drought frequency in the latter part of the century, a change in yellow pine phenology, age-dependent climate responses, data quality, and atmospheric pollution. Around 1960, the AMO shifted from a warm to cool phase that lasted through the 1990s (Gray et al. 2004). However, the shift in the climate response appears to have begun shortly before 1960. Also, a warm AMO phase began around 2000, but no shifts in the climate response in any of the five chronologies were apparent. Still, a phase change of AMO may have contributed to the increasing importance of temperature and decreasing significance of precipitation.

Negative or cool phases of the AMO are associated with increased precipitation in the

Southeast, while positive or warm phases tend to increase drought frequency and severity. In the second half of the 20th century, yellow pines in Great Smoky Mountains

National Park may have been less responsive to precipitation and PDSI simply because of fewer or less extreme droughts. A similar trend was identified in shortleaf pines in northern Georgia (Grissino-Mayer and Butler 1993). Around 1963, the trees became less sensitive to both precipitation and summer temperature. However, Grissino-Mayer and

Butler (1993) did not determine if the weakening response to moisture conditions was coupled with an increased sensitivity to winter temperature. They concluded that the changes in growth were likely caused by non-climatic factors such as acidic deposition, tree age, stand dynamics, or anthropogenic disturbance. In the Ozarks, shortleaf pines also showed heightened sensitivity to winter temperature toward the end of the 20th century (Stambaugh and Guyette 2004). In the first half of the century, the trees were responsive to growing season precipitation and PDSI, but winter and early spring

129 temperatures became increasingly important as the century progressed. Stambaugh and

Guyette (2004) attributed this to fewer droughts from 1960 to 1990. The climate response of yellow pines in Great Smoky Mountains National Park may have been similarly influenced by changes in drought frequency and severity in the second half of the century, possibly linked to an AMO phase change.

A change in the phenology of yellow pines in the southern Appalachians may have also contributed to the mid-20th century shift in climate response, but little evidence exists to support or refute this hypothesis. Cambial activity may now begin earlier in the spring or continue later into the fall than it did in the first half of the century. A lengthening of the growing season could explain the absence of an October temperature signal prior to 1960 and the recent strengthening of the winter temperature signal.

However, if the growing season lengthened, a stronger precipitation signal would be expected due to increased potential evapotranspiration, and this is not seen in the climate responses. Also, any theoretical change in phenology would require a causal agent: a change in climate. If a phenological change did occur, the underlying climatic change may have been related to the AMO phase change or to anthropogenic impacts on global climate.

A third possible explanation for the mid-century shift in climate response is that the effects of climate on tree growth may vary with tree age. Although it is well established that tree age affects growth rates (Fritts 1976), most dendroclimatological studies have assumed that climate-tree growth relationships are independent of tree age.

However, Carrer and Urbinati (2004) found that the climate responses of two species from the European Alps were related to tree age, but the species exhibited differing

130 degrees of age dependency. Overall, the climate signal was maximized when the trees were greater than 200 years old. Based on these findings, Carrer and Urbinati (2004) suggested that sampling that does not explicitly address tree age could result in an age- biased climate response. When sampling in Great Smoky Mountains National Park, age was addressed only implicitly; I sought to sample the trees that appeared oldest based on visual clues such as heavy lower branches or erratic growth forms (Schulman 1954;

Wagener and Schulman 1954). It is possible that yellow pines become increasingly sensitive to temperature and less drought sensitive with age, but I would then expect the shifts in climate response to occur at different times at each site. In this case, however, the shifts all occur around 1950–1960. Still, the effects of age on the climate response of yellow pines cannot be ruled out, as little is known about the relationship between tree age and the climate response of these particular species.

The quality of the climate data may have also played a role in the mid-century shift in climate response. In the latter half of the 20th century, the climate data used in the study were likely of better quality, more accurate, and more spatially extensive

(precipitation and PDSI) than in the first half of the century. The year 1930 is commonly used as a cut-off year for climate data reliability. After 1930, temperature, precipitation, and PDSI data may be more accurate due to improved technology and a greater number of observations within the climate division. Lower quality data early in the analysis may have weakened the temperature response in the first half-century, thus creating an apparent shift in the climate response. While this cannot be ruled out as a potential cause of the mid-century shift, it does not explain the weakening of the PDSI signal around

1960 and the re-intensification around 1980. Also, the change in climate response would

131 likely be more gradual if data quality caused the shift because the quality of the data improved gradually rather than suddenly. Still, poor data quality in the early 20th century may have contributed to an apparent mid-century shift in climate response at Great

Smoky Mountains National Park.

Atmospheric pollution may have also been a factor in the shifting response to climate. In the 20th century, acidic deposition (resulting from atmospheric pollution) has been a major problem in the southeastern U.S., and particularly East Tennessee. Copper mining and smelting began in the Copper Basin of southeast Tennessee in the 1850s and continued into the mid-20th century (Poste 1932; Foehner 1980; Harden and Matthews

2000). When copper ore was roasted, large quantities of sulfur dioxide were released into the atmosphere and deposited by rain as sulfuric acid. Baes and McLaughlin (1984) found increased trace metals and suppressed growth from 1863 to 1912 in tree rings of shortleaf pines growing near Cades Cove in Great Smoky Mountains National Park.

They inferred that the increased metal content and growth suppression were related to air pollution from copper smelting operations 88 km downwind from Cades Cove in the

Copper Basin. Baes and McLaughlin (1984) also found increased concentrations of trace metals, coupled with decreased growth rates beginning in the 1970s. Growth-trend declines likely caused by air pollution or acidic deposition have also been identified in conifers throughout the eastern U.S. (Ashby and Fritts 1972; Adams et al. 1985; LeBlanc et al. 1987). It is possible that acidic deposition not only influenced growth rates, but also masked or altered the shortleaf pine growth response to climate in the 20th century.

132

6.3 Evidence of Divergence

Divergence is defined as a weakening of the temperature-tree growth relationship in the latter part of the 20th century at traditionally temperature-sensitive sites, particularly in the high latitudes (Briffa et al. 1998a; D’Arrigo et al. 2008). Given this definition, the yellow pine chronologies from Great Smoky Mountains National Park do not show evidence of divergence, because the relationship between temperature and tree growth did not weaken. However, the climate-tree growth relationships were not stable during the past century. In the first half of the 20th century, the temperature signal was weak at all of the five sites. After 1950–1960, the temperature signal strengthened and winter mean minimum temperatures strongly affected growth. While this shift in climate response coincides with the widespread divergence noted in the high latitudes, no significant weakening of the temperature-tree growth relationship occurred at the five sites in Great Smoky Mountains National Park. The winter and fall temperature signals became more apparent, indicating convergence between instrumental temperatures and yellow pine growth rather than divergence.

Because the mechanism for divergence is still unclear, it is impossible to determine if the same factors causing divergence in the high latitudes also caused the shift in climate response in the Southeast. Global dimming (a decrease in solar radiation due to atmospheric pollution) is a prominent explanation for divergence (D’Arrigo et al.

2008). Because global dimming is strongest at the poles, it is unlikely that this phenomenon would also account for or contribute to a shift in climate response in the southern Appalachians. However, while global dimming itself likely was not an issue in this study, increased atmospheric pollution and the associated acidic deposition may have

133 played a role in the mid-century shift in climate-growth relationships in Great Smoky

Mountains National Park. A second prominent explanation for divergence is that trees in the high Northern latitudes have experienced temperature-induced drought stress during the past half-century due to anthropogenic climate change (Barber et al. 2000; D’Arrigo et al. 2008). In contrast, yellow pines in the southern Appalachians have experienced decreased drought sensitivity since 1960, possibly related to an AMO phase change or anthropogenic climate change. Both the divergence in the high latitudes and the convergence seen in Great Smoky Mountains National Park may have been caused by a changing climate, but the resultant growth responses to climate differ.

6.4 Potential for Reconstruction

Although yellow pine growth in Great Smoky Mountains National Park was influenced primarily by winter temperature, the relationship was not stable over the course of the 20th century. To reliably reconstruct a particular climate variable, its relationship with growth must remain stable over time. Therefore, this network of chronologies should be used with caution for climate reconstruction because of the temporal inconsistencies in the climate responses. Using a preliminary version of the

Gold Mine Trail chronology, Grissino-Mayer et al. (2007) reconstructed past winter temperature back to the 18th century. While the Gold Mine Trail chronology had the strongest response to winter temperature, the moving analysis revealed that even this response changed over time. Therefore, the Gold Mine Trail chronology should be used with caution for future reconstructions. The other four chronologies are also not ideal for climate reconstruction because their relationships with climate are unstable. At Gregory

134

Ridge Trail and Cooper Road Trail, no single climate variable accounts for the majority of the variance in annual growth. Thus, it may be difficult, though not impossible, to isolate the winter temperature signal from other climate signals and non-climatic noise in the data.

Although this network of chronologies is not ideal for reconstruction, it does provide insight into climate-growth relationships over time. In particular, this network indicates that a similar mid-century shift in climate response occurred in yellow pine chronologies throughout the Southeast (Georgia, Tennessee, Arkansas, and Missouri).

Also, the inconsistencies in climate response over time raise questions about the reliability of previous climate reconstructions. In the future, the temporal stability of the climate response in all chronologies or networks of chronologies should be tested as a first step to climate reconstruction.

135

CHAPTER 7 CONCLUSIONS AND FUTURE RESEARCH

The overall purpose of this research was to explore patterns in climate-tree growth relationships in the southern Appalachians over time. In recent years, the divergence problem has called into question the reliability of tree-ring based climate reconstructions

(Briffa et al 1998a; D’Arrigo et al. 2008). Dendroclimatologists now emphasize that consistent responses to climate are necessary to reliably reconstruct past climate variables such as temperature or precipitation. Because so few temperature-sensitive stands exist outside of the high latitudes, little research has been conducted on the stability of temperature responses in more temperate zones. If responses to temperature are consistently strong and show no evidence of divergence, tree-ring chronologies from temperate regions may prove useful for temperature reconstruction. Grissino-Mayer et al. (2007) identified a strong winter temperature signal in a shortleaf pine chronology from Great Smoky Mountains National Park. My research built upon this finding and explored the strength and stability of relationships between climate and yellow pine growth at five sites on the western side of Great Smoky Mountains National Park. In addition, this study confronted the divergence problem and assessed the potential of temperate region tree-ring chronologies to reconstruct past temperature. This final chapter outlines the main conclusions of this study in response to the four research questions and hypotheses put forth in Chapter One. I also suggest several directions for future dendroclimatological research in the southeastern U.S., and more specifically, in the southern Appalachian Mountains region.

136

7.1 Major Conclusions

1. Winter temperature was the primary climatic influence on yellow pine growth, while fall temperature and winter and early spring precipitation were secondary climatic influences.

Winter mean minimum temperatures were statistically significant drivers of growth at all five of the sites within Great Smoky Mountains National Park. Therefore, the null hypothesis that no significant relationship exists between climate and tree growth is rejected, while the alternate hypothesis that temperature and precipitation are related to tree growth is accepted. Warmer average minimum temperatures led to above average annual yellow pine growth, while colder winter temperatures led to below average growth. Fall temperature and winter and early spring precipitation also affected growth at some of the sites. Lower temperatures in November encouraged growth in the following growing season, an example of preconditioning. Warmer October months contributed to above-average growth by lengthening the growing season. Precipitation in late winter and spring affects growth to a lesser degree than temperature. Wetter winters and springs were associated with increased annual growth at Gregory Ridge Trail,

Cooper Road Trail, and Maynard Creek, but these relationships were weak at Shaw

Grave Gap and Gold Mine Trail.

2. Yellow pine growth was affected by Atlantic Ocean sea surface temperature and sea level pressure patterns (AMO and NAO).

Winter NAO index values were significantly correlated with growth at all sites.

AMO index values were also significant at three of the five sites. At the remaining two

137 sites, AMO was insignificant but tended to be positive. NAO was also positively related to growth, indicating that positive phases of both oscillations led to above average annual growth. NAO is the primary driver of winter temperature in the southern Appalachians, and positive phases are associated with the warm, wet winters that favor yellow pine growth. Positive or warm phases of the AMO are characterized by warm Atlantic SSTs.

Therefore, warm phases likely favor growth through increased accumulation and storage of carbohydrate reserves for growth in the following year. Also, four of the five chronologies exhibited low-frequency multidecadal oscillations of 55–90 years that may be related to AMO. Unlike Atlantic Ocean climate patterns, Pacific Ocean temperature and pressure patterns (PDO and ENSO) did not have a strong influence on yellow pine growth in Great Smoky Mountains National Park.

4. Climate-growth relationships were non-uniform between sites.

Although climate responses were similar at each of the five sites, certain sites exhibited stronger temperature responses, while other sites were more moisture-sensitive.

The Gold Mine Trail, Shaw Grave Gap, and Maynard Creek chronologies showed the strongest relationships to temperature, while the Cooper Road Trail and Gregory Ridge

Trail chronologies were the most responsive to precipitation and moisture conditions.

Differences in climate response between sites were likely related to microclimate and site conditions such as aspect, slope, and soil moisture. Species-specific responses, stand- level dynamics, and disturbance may have also influenced the differences in climate response between sites.

138

5. All five chronologies from Great Smoky Mountains National Park, as well as other shortleaf pine chronologies from the Southeast, showed a major shift in climate- growth relationships in the mid-20th century.

Moving correlation and response function analyses showed shifts in the climate response at all five sites around 1950–1960. Therefore, the null hypothesis that the climate-growth relationship remained stable over time can be rejected. The alternate hypothesis, which stated that the climate response changed over time, is accepted. The response to winter temperature strengthened in the latter half of the 20th century, while the response to growing season precipitation and moisture conditions weakened. Fall temperatures also became increasingly important toward the end of the century. Similar changes were identified in studies of shortleaf pine in northern Georgia and the Ozarks

(Grissino-Mayer and Butler 1993; Stambaugh and Guyette 2004). The cause of the mid- century shift in climate-growth relationships is unknown, but it may have been related to an AMO phase change (warm to cool), anthropogenic warming, changes in the phenology of yellow pines, age-specific climate responses, data quality issues, atmospheric pollution, or to other non-climatic factors.

6. None of the five chronologies showed evidence of divergence between temperature and tree growth. Rather, the temperature-tree growth relationships all strengthened in the latter half of the 20th century.

The null hypothesis, which stated that these chronologies show no evidence of divergence, is accepted. Divergence, or a weakening of the temperature-tree growth relationship in the past half-century, was not apparent in the five yellow pine

139 chronologies analyzed in this study. Instrumental records of minimum winter temperatures and predicted values based on ring-width index values actually showed a converging trend beginning around 1950–1960. Although divergence in the high latitudes and convergence in the Southeast occurred simultaneously, it is unlikely that the direct causal agents were the same. For example, global dimming is a popular explanation for divergence (D’Arrigo et al. 2008), but it is not a probable cause of the shift in the Southeast because global dimming has been greatest at the poles and weaker in temperate zones. Despite the differences, this study does have implications for the divergence problem. Changes in climate-growth relationships may be more common than previously believed. All five of the chronologies developed for this research showed major shifts in climate response over the course of one century. In Great Smoky

Mountains National Park, the climate variables that have influenced yellow pine growth in recent decades are not the same factors that drove growth a century ago. This suggests a re-thinking of the Principle of Uniformitarianism, as applied to dendroclimatology

(Fritts 1976). Future research might benefit from examining shifts in climate response that occurred in the late 19th or early 20th century to determine if any large-scale shifts occurred prior to widespread human impacts on the atmosphere and climate.

7. Because the climate-growth relationships are not stable over time, these chronologies should be used with caution when reconstructing past climate.

The climate response of yellow pines in Great Smoky Mountains changed over time. In the early 20th century, precipitation and growing season moisture availability affected tree growth. By the end of the century, winter temperature was the primary

140 driver of growth. Because of the shifting nature of the climate response, it is impossible to determine which climate variables affected tree growth prior to the period of instrumental temperature and precipitation records. Therefore, the five chronologies from this study should be used with caution for climate reconstruction. Performing tests on climate responses over time and setting standards for the stability of chronologies will increase confidence in tree-ring-based climate reconstructions in the future.

7.2 Future Research

This study suggests several new directions for dendroclimatological research in mid-latitude forests. The five yellow pine chronologies from Great Smoky Mountains

National Park and two shortleaf pine chronologies from northern Georgia (Grissino-

Mayer and Butler 1993) and the Ozarks (Stambaugh and Guyette 2004) all showed a similar shift in climate-growth relationships in the mid-20th century. Further analysis of climate-tree growth relationships in the southeastern U.S. is necessary to determine if a similar shift occurs across the region. Determining its spatial extent is the first step toward understanding the causes of this shift in growth response to climate. Also, temporal patterns in climate response should be analyzed for long-lived tree species other than yellow pines. Thus far, the mid-century shift has only been identified in yellow pine chronologies. The shift may have been unique to yellow pines, or it may have occurred widely across species lines. The International Tree-Ring Data Bank (ITRDB) includes many chronologies developed from various species in the Southeast. These existing chronologies, along with new chronologies developed in the region, should be examined for a similar mid-century shift in the climate-growth relationship.

141

Future research might also investigate the relationship between tree age and climate response. Carrer and Urbinati (2004) explored this issue in the European Alps and found species-specific age effects for European larch and stone pine. Research in the

Southeast might explore the relationship between tree age and climate response for particular species or groups of species such as yellow pines. In this study, the effects of tree age on climate-growth relationships could not be quantified because targeted sampling is necessarily age-biased. Future studies might explicitly sample trees of varying ages within the same species and compare the climate responses between even- aged groups to determine if climate-growth relationships are a function of individual tree ages.

The relationship between climate oscillations and tree growth over time represents a third opportunity for future dendroclimatologial research in the southern Appalachians.

In this study, I did not analyze the relationships between oscillations and growth using moving correlation analysis. Therefore, it is unclear whether the strong associations between growth and AMO and NAO were stable over time or changed with phase changes. Analyzing these relationships over moving intervals might help to identify the cause of the shifts in temperature and precipitation response. Also, if the relationships between tree growth and NAO or AMO are indeed stable and significant throughout the

20th century, this network of chronologies could provide useful information on how these ocean-atmospheric teleconnections affected the climate of the southern Appalachians over the past few centuries.

Overall, more research is needed on temporal changes in climate-growth relationships. The divergence problem has spurred such research at high latitude sites,

142 but few temperate chronologies have been thoroughly tested for consistent responses to climate. In particular, the southeastern U.S. is an exceptional location for such research because of the many existing chronologies available in the ITRDB and the variety of tree species. It is possible that the mid-century shift in climate response identified in this study represents a major shift found in chronologies of various species throughout the region, but more research is necessary on both the spatial and taxonomic extent of this phenomenon.

143

REFERENCES

144

Abakumova, G., E. Feigelson, V. Russak, and V. Stadnik. 1996. Evaluation of long-term changes in radiation, cloudiness, and surface temperature on the territory of the former Soviet Union. Journal of Climate 9: 1319–1327.

Adams, H.S., S.L. Stephenson, T.J. Blasing, and D.N. Duvick. 1985. Growth-trend declines of spruce and fir in Mid-appalachian subalpine forests. Environmental and Experimental Botany 25(4) 315–325.

Ashby, W.C., and H.C. Fritts. 1972. Tree growth, air pollution, and climate near LaPorte, Indiana. Bulletin of the American Meteorological Society 53(3): 246–251.

Baes, C.F., and S.B. McLaughlin. 1984. Trace elements in tree rings: Evidence of recent and historical air pollution. Science 224: 494–497.

Barber, V.A., G.P. Juday, and B.P. Finney. 2000. Reduced growth of Alaskan white spruce in the twentieth century from temperature-induced drought stress. Nature 405: 668–673.

Bennett, H.H. 1921. The Soils and Agriculture of the Southern States. MacMillan Company, New York. 399 pp.

Biondi, F. 1997. Evolutionary and moving response functions in dendroclimatology. Dendrochronologia 15: 139–150.

Biondi, F., and K. Waikul. 2004. DENDROCLIM2002: A C++ program for statistical calibration of climate signals in tree-ring chronologies. Computers and Geosciences 30: 303–311.

Blasing, T.J., D.N. Duvick, and D.C. West. 1981. Dendroclimatic calibration and verification using regionally averaged and single station precipitation data. Tree- Ring Bulletin 41: 37–43.

Bratton, S.P., M.G. Hickler, and J.H. Graves. 1979. Trail erosion patterns in Great Smoky Mountains National Park. Environmental Management 3(5): 431–445.

Briffa, K.R. 2000. Annual climate variability in the Holocene: interpreting the message of ancient trees. Quaternary Science Reviews 19: 87–105.

Briffa, K.R., T.S. Bartholin, D. Eckstein, P.D. Jones, W. Karlén., F.H. Schweingruber, and P. Zetterberg. 1990. A 1,400-year tree-ring record of summer temperatures in Fennoscandia. Nature 346: 434-439.

Briffa, K.R., P.D. Jones, and F.H. Schweingruber. 1992. Tree-ring density reconstructions of summer temperature patterns across western North America since 1600. Journal of Climate 5(7): 735–754.

145

Briffa, K.R., F.H. Schweingruber, P.D. Jones, T.J. Osborn, S.G. Shiyatov, and E.A. Vaganov. 1998a. Reduced sensitivity of recent tree-growth to temperature at high northern latitudes. Nature 391: 678–682.

Briffa, K.R., F.H. Schweingruber, P.D. Jones, T.J. Osborn, I.C. Harris, S.G. Shiyatov, E.A. Vaganov, H. Grudd, and J. Cowie. 1998b. Trees tell of past climates: but are they speaking less clearly today? Philosophical Transactions: Biological Sciences 353: 65–73.

Bryant, B., and J.C. Reed. 1970. Structural and metamorphic history of the Southern Blue Ridge. In: Fisher, G.W., F.J. Pettijohn, J.C. Reed, Jr., and K.N. Waver, eds. Studies of Appalachian Geology: South and Central. Interscience Publishers, New York: 213–225.

Buntgen, U., J. Esper, D.C. Frank, K. Nicolussi, and M. Schmidhalter. 2005. A 1052-year tree-ring proxy for Alpine summer temperatures. Climate Dynamics 25(2–3): 141–153.

Callaway, R.M., E.C. Clebsch, and P.S. White. 1987. A multivariate analysis of forest communities in the western Great Smoky Mountains National Park. American Midland Naturalist 118(1): 107–120.

Campbell, C.C. 1960. Birth of a National Park in the Great Smoky Mountains. University of Tennessee Press, Knoxville. 154 pp.

Carrer, M., and C. Urbinati. 2004. Age-dependent tree-ring growth responses to climate in Larix decidua and Pinus cembra. Ecology 85(3): 730–740.

Carrer, M., and C. Urbinati. 2006. Long-term change in the sensitivity of tree-ring growth to climate forcing in Larix decidua. New Phytologist 170: 861–872.

Christopherson, R.W. 2006. Geosystems. Pearson Prentice Hall, Upper Saddle River, New Jersey. 689 pp.

Clark, S.H.B. 2001. Birth of the mountains: The geologic story of the southern Appalachian Mountains. U.S. Geological Survey General Interest Publication. 23 pp.

Cleaveland, M.K. 1975. Dendroclimatic Relationships of Shortleaf Pine (Pinus echinata Mill.) in the South Carolina Piedmont. M.S. Thesis. Clemson University, Clemson. 189 pp.

Cook, E.R. 1985. A Time Series Approach to Tree-Ring Standardization. PhD. Dissertation. University of Arizona, Tucson. 171 pp.

146

Cook, E.R. 1987. On the disaggregation of tree-ring series for environmental studies. In: G.C. Jacoby, Jr. and J.W. Hornbeck, eds., Proceedings of the International Symposium on Ecological Aspects of Tree-Ring Analysis, 17–21 Aug. 1986, Tarrytown, New York, U.S. Department of Energy: 522–542.

Cook, E.R., and G.C. Jacoby, Jr. 1979. Evidence for quasi-periodic July drought in the Hudson Valley, New York. Nature 282(5737): 390–392.

Cook, E.R., and K. Peters. 1981. The smoothing spline: a new approach to standardizing forest interior tree-ring width series for dendroclimatic studies. Tree-Ring Bulletin 41: 45–53.

Cook, E.R., K. Briffa, S. Shiyato, and V. Mazepa. 1990. Tree-ring standardization and growth-trend estimation. In: E.R. Cook and L.A. Kairukstis, eds. Methods of Dendrochronology: Applications in the Environmental Sciences. Kluwer Academic Publishers, Dordrecht, the Netherlands: 104–123.

Cook, E.R., R.D. D’Arrigo, and K. Briffa. 1998. A reconstruction of the North Atlantic Oscillation using tree-ring chronologies from North America and Europe. The Holocene 8(1): 9–17.

Cook, E.R., D.M. Meko, D.W. Stahle, and M.K. Cleaveland. 1999. Drought reconstructions for the continental United States. Journal of Climate 12(4): 1145– 1162.

Copenheaver, C.A., L.E. Grinter, J.H. Lorber, M.A. Neathour, and M.P. Spinney. 2002. A dendroecological and dendroclimatic analysis of Pinus virginiana and Pinus rigida at two slope positions in the Virginia Piedmont. Castanea 67: 302–315.

D’Arrigo, R.D, G.C Jacoby, and R. Free. 1992. Tree-ring width and maximum latewood density at the North American treeline: parameters of climatic change. Canadian Journal of Forest Research 22: 1290–1296.

D’Arrigo, R.D., R.K. Kaufmann, N. Davi, G.C. Jacoby, C. Laskowski, R.B. Myeni, and P. Cherubini. 2004. Threshold for warming-induced growth decline at elevational tree line in the Yukon Territory, Canada. Global Biogeochemical Cycles 18. doi:10.1029/2004GB002249.

D’Arrigo, R.D., R. Wilson, B. Liepert, and P. Cherubini. 2008. On the ‘Divergence Problem’ in Northern Forests: A review of the tree-ring evidence and possible causes. Global and Planetary Change 60: 289–305.

Davi, N.K., G.C. Jacoby, and G.C. Wiles. 2003. Boreal temperature variability inferred from maximum latewood density and tree-ring width data, Wrangell Mountain region, Alaska. Quaternary Research 60(3): 252–262.

147

Davis, D.E. 2000. Where there are Mountains: An Environmental History of the Southern Appalachians. University of Georgia Press, Athens. 320 pp.

Delcourt, H.R., and P.A. Delcourt. 1997. Pre-Columbian Native American use of fire on southern Appalachian landscapes. Conservation Biology 11(4): 1010–1014.

Delcourt, P.A., and H.R. Delcourt. 1998. The influence of prehistoric human-set fires on oak-chestnut forests in the southern Appalachians. Castanea 63(3): 337–345.

Driscoll, W.W., G.C. Wiles, R.D. D’Arrigo, and M. Wilmking. 2005. Divergent tree growth response to recent climate warming, Lake Clark National Park and Preserve, Alaska. Geophysical Research Letters 32. doi:10.1029/2005GL02425.

Enfield, D.B., A.M. Mestas-Nuñez and P.J. Trimble. 2001. The Atlantic Multidecadal Oscillation and its relation to rainfall and river flows in the continental U.S. Geophysical Research Letters 28(10): 2077–2080.

Esper, J., and D.C. Frank. 2004. Climate reconstructions: Low-frequency ambition and high-frequency ratification. Eos: Transactions of the American Geophysical Union 85(12): 113–120.

Esper, J., F.H. Schweingruber, and M. Winiger. 2002. 1300 years of climatic history for Western Central Asia inferred from tree-rings. The Holocene 12(3): 267–1277.

Foehner, N.L. 1980. The Historical Geography of Environmental Change in the Copper Basin. M.S. Thesis. University of Tennessee, Knoxville.

Frank, D., U. Büntgen, R. Böhm, M. Maugeri, and J. Esper. 2007. Warmer early instrumental measurements versus colder reconstructed temperatures: shooting at a moving target. Quaternary Science Reviews 26: 3298–3310.

Frank, D., R. Wilson, and J. Esper. 2005. Synchronous variability changes in Alpine temperature and tree-ring data over the past two centuries. Boreas 34(4): 498– 505.

Friend, A.L., and W.L Hafley. 1989. Climatic limitations to growth in loblolly and shortleaf pine: A dendroclimatological approach. Forest Ecology and Management 26: 113–122.

Fritts, H.C. 1976. Tree Rings and Climate. Academic Press, New York. 567 pp.

Fritts, H.C., T.J. Blasing, B.P. Hayden, and J.E. Kutzbach. 1971. Multivariate techniques for specifying tree-growth and climate relationships and for reconstruction anomalies in paleoclimate. Journal of Applied Meteorology 10(5): 845–864.

148

Gray, S.T., J.L. Betancourt, C.L. Fastie, and S.T. Jackson. 2003. Patterns and sources of multidecadal oscillations in drought-sensitive tree-ring records from the central and southern Rocky Mountains. Geophysical Research Letters 30. doi:10.1029/2002GL016154.

Gray, S.T., L.J. Graumlich, J.L. Betancourt, and G.T. Pederson. 2004. A tree-ring based reconstruction of the Atlantic Multidecadal Oscillation since 1567 A.D. Geophysical Research Letters 31. doi:10.1029/2004GL019932.

Gaffin, D.M., D.G. Hotz, and T.I. Getz. 2002. An evaluation of temperature variations around the Great Smoky Mountains National Park and their associated synoptic weather patterns. NOAA Technical Memorandum: NWS SR-221.

Grissino-Mayer, H.D. 2001. Evaluating crossdating accuracy: A manual and tutorial for the computer program COFECHA. Tree-Ring Research 57(2): 205–221.

Grissino-Mayer, H.D. 2003. A manual and tutorial for the proper use of an increment borer. Tree-Ring Research 59(2): 63–79.

Grissino-Mayer, H.D., and D.R. Butler. 1993. Effects of climate on growth of shortleaf pine (Pinus echinata Mill.) in northern Georgia: a dendroclimatic study. Southeastern Geographer 33(1): 65–81.

Grissino-Mayer, H.D., and H.C. Fritts. 1995. Dendroclimatology and dendroecology in the Pinaleño Mountains. Pages 100–120 in C.A. Stock and R.S. Hoffmann, eds., Storm over a Mountain Island: Conservation Biology and the Mt. Graham Affair. University of Arizona Press, Tucson, Arizona.

Grissino-Mayer, H.D., A.P. Williams, L.B. LaForest, W.H. Brenton, P.W. Clark, S.G. Mortenson, and M.D. Spond. 2007. Influence of winter temperatures on growth of shortleaf pine in Great Smoky Mountains National Park, Tennessee, U.S.A. 2007 North American Dendroecological Fieldweek (NADEF) Report. 30 pp.

Harden, C.P., and L. Matthews. 2000. Rainfall response of degraded soil following restoration in the Copper Basin, Tennessee, U.S.A. Environmental Management 26(2): 163–174.

Harmon, M.E. 1982. Fire history of the westernmost portion of Great Smoky Mountains National Park. Bulletin of the Torrey Botanical Club 109(1): 74–79.

Harmon, M.E., S.P. Bratton, and P.S. White. 1984. Disturbance and vegetation response in relation to environmental gradients in the Great Smoky Mountains. Plant Ecology 55(3): 129–139.

Harrod, J., P.S. White, and M.E. Harmon. 1998. Changes in xeric forests in western Great Smoky Mountains National Park, 1936–1995. Castanea 63(3): 346–360.

149

Harrod, J.C., M.E. Harmon, and P.S. White. 2000. Post-fire succession and 20th century reduction in fire frequency on xeric southern Appalachian sites. Journal of Vegetation Science 11(4): 465–472.

Hass, A. 2008. Fire History in the Southern Appalachian Region, U.S.A., From Sedimentary Charcoal. M.S. Thesis. University of Tennessee, Knoxville.

Holmes, R.L. 1983. Computer-assisted quality control in tree-ring dating and measurement. Tree-Ring Bulletin 43: 69–78.

Hurrell, J.W. 1996. Influence of variations in extratropical wintertime teleconnections on Northern Hemisphere temperatures. Geophysical Research Letters 23: 665–668.

Hurrell, J.W., and H. van Loon. 1997. Decadal variations in climate associated with the North Atlantic Oscillation. Climatic Change 36: 301–326.

IPCC (Intergovernmental Panel on Climate Change). 2001. Climate Change 2001: The Scientific Basis. Contribution of working group I to the third assessment report of the IPCC. Cambridge University Press, Cambridge, England.

Jacoby, G.C., I.S. Ivanciu, and L.D. Ulan. 1988. A 263-year record of summer temperature for northern Quebec reconstructed from tree-ring data and evidence of a major climatic shift in the early 1800s. Palaeogeography, Palaeoclimatology, Palaeoecology 64: 69–78.

Jones, P.D., K.R. Briffa, T.P. Barnett, and S.F.B. Tett. 1998. High-resolution palaeoclimatic records for the last millennium: interpretation, integration and comparison with General Circulation Model control-run temperatures. The Holocene 8(4): 455–471.

Kalman, R.E. 1960. A new approach to linear filtering and prediction problems. Journal of Basic Engineering 82(1): 35–45.

Kaplan, A., M.A. Cane, Y. Kushnir, A.C. Clement, M.B. Blumenthal, and B. Rajagopalan. 1998. Analyses of global sea surface temperature 1856–1991. Journal of Geophysical Research 103: 567–589.

Kerr, R.A. 2000. A North Atlantic climate pacemaker for the centuries. Science 288: 1984–1985.

King, P.B., R.B. Newman, and J.B. Hadley. 1968. Geology of the Great Smoky Mountains National Park: Tennessee and North Carolina. U.S. Geological Survey Professional Paper 587. 23 pp.

150

Lafon, C.W., and M.J. Kutac. 2003. Effects of ice storms, southern pine beetle infestation, and fire on Table Mountain pine forests of southwestern Virginia. Physical Geography 24(6): 504–519.

LaMarche, V.C., Jr., and C.W. Stockton. 1974. Chronologies from temperature sensitive Bristlecone pines at upper treeline in the western United States. Tree-Ring Bulletin 34: 21-45.

Lamb, P.J., and R.A. Peppler. 1987. North Atlantic Oscillation: concept and an application. Bulletin of the American Meteorological Society 68(10): 1218–1225.

Land, C.W., and L.K. Rieske. 2006. Interactions among prescribed fire, herbivore pressure, and shortleaf pine (Pinus echinata) regeneration following southern pine beetle (Dendroctonus frontalis) mortality. Forest Ecology and Management 235: 260–269.

Latif, M., and T.P. Barnett. 1994. Causes of decadal climate variability over the North Pacific and North America. Science 266: 634–637.

LeBlanc, D.C., D.J. Raynal, and E.H. White. 1987. Acidic deposition and tree growth I: The use of stem analysis to study historical growth patterns. Journal of Environmental Quality 16(4): 325–333.

Lester, B., A. Khiel, and D. Thomas. 2007. Soil organic matter at different elevations in the Great Smoky Mountains National Park. Proceedings of the 2007 Meeting of the American Society of Agronomy, 4–8 November 2007, New Orleans.

MacDonald, G.M., and R.A. Case. 2005. Variations in the Pacific Decadal Oscillation over the past millenium. Geophysical Research Letters 32. doi:10.1029/2005GL022478.

Madden, M., R. Welch, T. Jordan, P. Jackson, R. Seavey, and J. Seavey. 2004. Digital vegetation maps for the Great Smoky Mountains National Park. Final report to the U.S. Department of Interior, National Park Service. Cooperative Agreement 1443-CA-5460-98-019.

Mann, M.E., R.S. Bradley, and M.K. Hughes. 1998. Global-scale temperature patterns and climate forcing over the past six centuries. Nature 392: 779–787.

Mantua, N.J. 2009. PDO Index. Data archived at: http://jisao.washington.edu/pdo/ PDO.latest

Mantua, J.J., S.R. Hare, Y. Zhang, J.M. Wallace and R.C. Francis. 1997. A Pacific interdecadal climate oscillation with impacts on salmon production. Bulletin of the American Meteorological Society 78(6): 1069–1080.

151

Mantua, N.J., and S.R. Hare. 2002. The Pacific Decadal Oscillation. Journal of Oceanography 58: 33–44.

McCabe, G.J., M.A. Palecki, and J.L. Betancourt. 2004. Pacific and Atlantic Ocean influences on multidecadal drought frequency in the United States. Proceedings of the National Academy of Sciences of the United States 101(12): 4136–4141.

McIntyre, S. 2005. Some thoughts on disclosure and due diligence in Climate Science. Climate Audit. http://climateaudit.org/index.php?p=66. Accessed 5 Dec 2008.

McIntyre, S., and R. McKitrick. 2005a. Hockey sticks, principal components, and spurious significance. Geophysical Research Letters 32. doi:10.1029/2004GL021750.

McIntyre, S., and R. McKitrick. 2005b. The M&M critique of the MBH98 Northern hemisphere climate index: update and implications. Energy and Environment 16(1): 69–100.

Moberg, A., D.M. Sonechkin, K. Holmgren, N.M. Datsenko, and W. Karlen. 2005. Highly variable Northern Hemisphere temperatures reconstructed from low- and high-resolution proxy date. Nature 4332: 613–617.

Moore, H.L. 1988. A Roadside Guide to the Geology of the Great Smoky Mountains National Park. University of Tennessee Press, Knoxville. 178 pp.

NCDC. 2008. National Climatic Data Center. U.S. Department of Commerce, National Oceanic and Atmospheric Administration, Asheville, North Carolina. http://www.ncdc.noaa.gov. Accessed 21 Dec 2008.

National Park Service. 2008. Great Smoky Mountains National Park. http://www.nps.gov/grsm. Accessed 2 Dec 2008.

National Weather Service. 2003. ENSO impacts on the U.S.: Tennessee. http://www.cpc.noaa.gov/products/predictions/threats2/enso/elnino/tn_bar.html. Accessed 23 Jan 2009.

NOAA. 2009. National Oceanic and Atmospheric Administration. Kaplan SST V2 data. Data archived at: http://www.cdc.noaa.gov/

Oberhuber, W., W. Kofler, K.Pfeifer, A. Seeber, A. Gruber, and G. Wieser. 2008. Long- term changes in tree-ring-climate relationships at Mt. Patscherkofel (Tyrol, Austria) since the mid-1980s. Trees 22: 31–40.

Orvis, K.H., and H.D. Grissino-Mayer. 2002. Standardizing the reporting of abrasive papers used to surface tree-ring samples. Tree-Ring Research 58(1/2): 47–50.

152

Park, W.K., and F.W. Telewski. 1993. Measuring maximum latewood density by image analysis at the cellular level. Wood and Fiber Science 25(4): 326–332.

Perry, T.O. 1971. Dormancy of trees in winter. Science 171: 29–36.

Philander, S.G.H. 1983. El Niño-Southern Oscillation phenomena. Nature 302: 295–301.

Poste, E.R. 1932. The Tennessee Copper Basin. Industrial and Engineering Chemistry 24(6): 690–694.

Pyle, C. 1988. The type and extent of anthropogenic vegetation disturbance in the Great Smoky Mountains before National Park Service acquisition. Castanea 53(3): 183–196.

Queensland Government Environmental Protection Agency. 2009. Southern Oscillation Index. Data archived at: http://www.longpaddock.qld.gov.au/ SeasonalClimateOutlook/SouthernOscillationIndex/SOIDataFiles/

Rogers, J.C. 1984. The association between the North Atlantic Oscillation and the Southern Oscillation in the Northern Hemisphere. Monthly Weather Review 112: 1999–2015.

Rogers, J.C. 2005. The Rogers (1984) North Atlantic Oscillation Index. Data archived at: http://polarmet.mps.ohio-state.edu/NAO/

Ropelewski, C.F., and M.S. Halpert. 1986. North American precipitation and temperature patterns associated with the El Niño/Southern Oscillation (ENSO). Monthly Weather Review 114(12): 2352–2363.

Schulman, E. 1954. Longevity under adversity in conifers. Science 119: 396–399.

Smith, E.R. 1990. Temporal and spatial variability in tree growth response to climate. PhD. dissertation, University of Tennessee, Knoxville. 391 pp.

South, D.B., and E.R. Buckner. 2003. The decline of southern yellow pine timberland. Journal of Forestry: 30–35.

Stambaugh, M.C., and R.P. Guyette. 2004. Long-term growth and climate response of shortleaf pine at the Missouri Ozark Forest Ecosystem Project. Proceedings of the 14th Central Hardwood Forest Conference: 448–458.

Stokes, M.A., and T.L. Smiley. 1996. An Introduction to Tree-Ring Dating. University of Arizona Press, Tucson. 73 pp.

153

Szeicz, J.M., and G.M. MacDonald. 1994. Age-dependent tree-ring growth responses of subarctic white spruce to climate. Canadian Journal of Forest Research 24(1): 120–132.

Szeicz, J.M., and G.M. MacDonald. 1995. Dendroclimatic reconstruction of summer temperatures in northwestern Canada since A.D. 1638 based on age-dependent modeling. Quaternary Research 44(2): 257–266.

Torrence, C., and G.P. Compo. 1998. A practical guide to wavelet analysis. Bulletin of the American Meteorological Society 79: 61–78.

Troup, A.J. 1965. The ‘southern oscillation.’ Quarterly Journal of the Royal Meteorological Society 91: 490–506.

U.S. Geological Survey. 2004. A Tapestry of Time and Terrain. http://tapestry.usgs.gov/. Accessed 12 Jan 2009.

Van Desusen, P.C. 1987. Some applications of the Kalman filter to tree-ring analysis. In: G.C. Jacoby, J.W. Hornbeck, eds., Proceedings of the International Symposium on Ecological Aspects of Tree Ring Analysis, 17–21 August 1986, Tarrytown, New York, U.S. Department of Energy: 566–578.

Van Lear, D.H., and T.A. Waldrop. 1989. History, uses, and effects of fire in the Appalachians. U.S. Forest Service General Technical Report SE-54. 30 pp. van Loon, H., and J.C. Rogers. 1978. The seesaw in winter temperatures between Greenland and Northern Europe. Monthly Weather Review 106: 296–310.

Visser, H., and J. Molenaar. 1987. Time-dependent responses of trees to weather variations: an application of the Kalman filter. In: G.C. Jacoby, J.W. Hornbeck, eds., Proceedings of the International Symposium on Ecological Aspects of Tree Ring Analysis, 17–21 August 1986, Tarrytown, New York, U.S. Department of Energy: 579–590.

Wagener, W.W., and E. Schulman. 1954. Communications: longevity under adversity in conifers. Science 119(3103): 883–884.

Waldron, J.D., C.W. Lafon, R.N. Coulson, D.M. Cairns, M.D. Tchakerian, A. Birt and K.D. Klepzig. 2007. Simulating the impacts of southern pine beetle and fire on the dynamics of xerophytic pine landscapes in the southern Appalachians. Applied Vegetation Science 10(1): 53–64.

Walker, G.T. 1924. Correlations in seasonal variations of weather, IX. Memoirs of the Indian Meteorology Department 24: 275–332.

154

Walker, S.L. 1991. Great Smoky Mountains: The Splendor of the Southern Appalachians. Elan Publishing, Charlottesville, 63 pp.

Web Soil Survey. 2007. Natural Resources Conservation Service, U.S. Department of Agriculture. http://websoilsurvey.nrcs.usda.gov. Accessed 20 Jan 2009.

Welch, N.T. 1999. Occurrence of fire in southern Appalachian yellow pine forests as indicated by macroscopic charcoal in soil. Castanea 64(4): 310–317.

White, P.B. 2007. Dendroclimatological Analysis of Oak Speices in the Southern Appalachian Mountains. Bachelor’s Honors Thesis, University of Tennessee, Knoxville. 69 pp.

Whittaker, R., 1956. Vegetation of the Great Smoky Mountains. Ecolological Monographs 26: 1–80.

Wilmking, M., R.D. D'Arrigo, G.C. Jacoby, and G.P. Juday. 2005. Increased temperature sensitivity and divergent growth trends in circumpolar boreal forests. Geophysical Research Letters 32. doi:10.1029/2005GL023331.

Yamaguchi, D.K. 1991. A simple method for cross-dating increment cores from living trees. Canadian Journal of Forest Research 21: 414–416.

155

APPENDICES

156

Appendix A1. Gold Mine Trail COFECHA Output Summary Statistics.

Corr //------Unfiltered ------\\ //---- Filtered -----\\ No. No. No. with Mean Max Std Auto Mean Max Std Auto AR Seq Series Interval Years Segmt Flags Master msmt msmt dev corr sens value dev corr () ------1 GMX005A 1854 2006 153 8 0 .581 1.33 5.38 .815 .862 .211 2.89 .506 -.019 1 2 GMX005B 1853 2006 154 8 0 .686 1.48 5.22 .804 .829 .205 2.66 .458 .000 3 3 GMX006B 1910 2006 97 5 0 .566 2.52 7.77 1.937 .906 .286 2.58 .423 .009 2 4 GMX007A 1882 2006 125 6 1 .398 2.20 5.48 1.193 .860 .208 2.63 .434 -.019 1 5 GMX007B 1883 2006 124 6 1 .454 2.53 5.63 1.430 .869 .193 2.64 .378 .000 1 6 GMX008A 1863 2006 144 7 0 .474 1.37 4.17 .807 .815 .295 2.84 .505 -.009 3 7 GMX009A 1847 2006 160 8 3 .459 1.44 4.16 .771 .827 .236 2.70 .428 .014 1 8 GMX009B 1860 2006 147 7 0 .571 1.81 5.66 1.076 .830 .277 2.68 .432 -.062 4 9 GMX074B 1731 1800 70 4 0 .597 1.40 2.75 .419 .390 .237 2.69 .501 .021 1 10 GMC001A 1847 2006 160 8 1 .509 1.66 5.53 1.023 .821 .253 2.57 .352 -.002 1 11 GMC010B 1873 2006 134 7 0 .608 1.36 3.27 .692 .808 .259 2.70 .527 -.004 1 12 GMC016A 1863 2004 142 7 3 .384 1.30 3.55 .803 .706 .370 2.63 .406 -.003 1 13 GMC022A 1876 2006 131 7 5 .308 .93 1.99 .457 .733 .282 2.83 .432 .002 1 14 GMC022B 1863 2006 144 7 2 .435 1.05 2.78 .513 .640 .320 3.04 .462 -.004 1 15 GMC033A 1871 2006 136 7 0 .551 1.16 2.57 .492 .674 .285 2.56 .421 -.013 1 16 GMC033B 1868 2006 139 7 0 .597 1.42 3.54 .627 .753 .243 2.65 .372 -.027 1 17 GMC035A 1858 2006 149 8 1 .530 1.49 3.52 .772 .795 .264 2.78 .524 -.045 1 18 GMC035B 1862 2006 145 7 1 .549 1.16 3.73 .655 .810 .260 2.78 .443 -.013 1 19 GMC047A 1867 2006 140 7 2 .443 1.15 2.79 .577 .740 .295 2.89 .601 .043 1 20 GMC047B 1874 2006 133 7 3 .437 .92 2.24 .445 .696 .251 2.87 .512 .011 1 21 GMC048A 1870 2005 136 7 2 .518 1.04 2.82 .612 .846 .267 2.74 .430 -.002 1 22 GMC048B 1873 2006 134 7 0 .569 1.33 3.71 .761 .791 .272 2.81 .433 -.033 1 23 GMT001A 1780 2004 225 11 2 .543 1.13 3.23 .602 .756 .305 2.61 .369 -.016 1 24 GMT001B 1804 2004 201 10 0 .556 1.12 2.70 .533 .741 .249 2.48 .276 -.013 1 25 GMT002A 1729 2004 276 14 1 .525 .10 .38 .079 .875 .237 2.57 .357 -.006 3 26 GMT002B 1753 2004 252 13 0 .536 1.02 3.29 .559 .821 .213 2.61 .405 .029 1 27 GMT002C 1796 2004 209 11 3 .442 1.50 4.61 .699 .577 .303 2.80 .369 -.019 1 28 GMT003A 1838 2004 167 9 0 .554 1.26 3.48 .687 .768 .246 2.68 .377 -.006 1 29 GMT003B 1852 2004 153 8 0 .633 1.26 3.07 .679 .837 .214 2.72 .408 -.023 1 30 GMT003C 1815 1997 183 9 0 .508 1.13 2.50 .464 .698 .249 2.70 .423 -.024 2 31 GMT004A 1780 1982 203 10 0 .646 1.04 2.24 .483 .695 .270 2.66 .398 .030 1 32 GMT004B 1780 1983 204 10 1 .565 .72 1.48 .307 .651 .275 2.76 .464 .014 1 33 GMT005A 1905 2004 100 5 0 .535 1.29 3.10 .599 .734 .261 2.79 .523 -.014 1 34 GMT005B 1910 2004 95 5 0 .598 1.87 4.37 .871 .665 .300 2.61 .472 -.060 1 35 GMT006A 1895 2004 110 6 0 .560 .91 2.54 .568 .730 .341 2.86 .542 .054 1 36 GMT006B 1763 2001 239 12 5 .411 .69 1.63 .322 .695 .306 2.67 .434 -.024 2

157

Appendix A1. continued.

Corr //------Unfiltered ------\\ //---- Filtered -----\\ No. No. No. with Mean Max Std Auto Mean Max Std Auto AR Seq Series Interval Years Segmt Flags Master msmt msmt dev corr sens value dev corr () ------37 GMT007A 1882 2004 123 6 1 .559 .18 .47 .076 .671 .275 2.52 .362 .020 1 38 GMT007B 1887 2004 118 6 2 .468 2.02 5.78 1.367 .801 .283 2.39 .290 -.044 3 39 GMT008A 1883 2004 122 6 0 .540 .14 .28 .051 .545 .291 2.49 .426 -.011 1 40 GMT008B 1875 2004 130 7 0 .510 .10 .36 .071 .814 .292 2.57 .472 .001 2 41 GMT009A 1815 1943 129 7 0 .511 .81 1.70 .324 .805 .197 2.74 .493 .049 1 42 GMT009B 1826 1987 162 8 1 .548 .79 2.41 .284 .570 .230 2.70 .503 -.007 2 43 GMT010A 1812 1962 151 8 0 .527 .86 2.56 .529 .877 .212 2.77 .493 -.008 1 44 GMT010C 1770 1906 137 7 1 .597 .84 2.56 .425 .774 .209 2.78 .523 -.012 1 45 GMT011A 1802 1926 125 6 1 .625 .83 1.43 .299 .643 .230 2.64 .507 .056 1 46 GMT011B 1810 1980 171 9 1 .629 .75 1.76 .274 .659 .233 2.77 .440 .047 1 47 GMT015A 1888 2000 113 6 0 .531 1.61 2.91 .616 .770 .226 2.65 .499 -.048 2 48 GMT016A 1884 1925 42 2 0 .502 2.66 4.16 .711 .613 .204 2.38 .443 .207 1 49 GMT017A 1907 1963 57 3 0 .656 2.13 3.87 .886 .809 .241 2.45 .436 .059 1 50 GMT017B 1920 2001 82 4 0 .676 1.90 3.72 .691 .712 .242 2.72 .404 .048 2 51 GMT018A 1753 2004 252 13 0 .678 .93 4.92 .690 .863 .252 2.69 .384 -.016 1 52 GMT018B 1741 2004 264 13 0 .672 .11 .50 .088 .878 .249 2.74 .479 -.013 1 53 GMT019A 1761 1850 90 4 0 .533 .93 2.64 .442 .705 .257 2.89 .498 -.008 3 54 GMT019B 1754 2004 251 13 2 .542 .08 .33 .060 .829 .265 2.69 .426 -.019 1 55 GMT020A 1896 2004 109 6 1 .453 .16 .40 .082 .775 .295 2.50 .418 -.065 1 56 GMT020B 1895 2003 109 6 1 .513 .17 .38 .081 .777 .292 2.71 .492 .002 1 57 GMT021A 1900 2004 105 5 0 .638 .20 .41 .076 .661 .266 2.52 .345 -.068 2 58 GMT021B 1917 2004 88 5 0 .681 2.07 4.54 .804 .749 .239 2.50 .435 -.036 2 59 GMT022A 1890 2004 115 6 1 .464 .17 .50 .119 .884 .273 2.49 .386 -.057 1 60 GMT022B 1882 1999 118 5 0 .615 1.58 5.21 .800 .824 .216 2.71 .454 -.012 3 61 GMT023A 1934 2004 71 4 0 .576 .21 .55 .105 .742 .317 2.71 .567 .024 1 62 GMT023B 1936 2004 69 4 0 .529 .21 .57 .116 .766 .327 2.71 .454 .000 1 63 GMT024A 1887 2004 118 6 0 .453 .22 .65 .162 .852 .258 2.72 .367 -.030 1 64 GMT024B 1884 2004 121 6 1 .607 .19 .67 .154 .852 .328 2.58 .487 -.003 1 65 GMT026A 1857 2004 148 8 0 .622 1.17 3.10 .579 .557 .358 2.70 .424 -.040 2 66 GMT026B 1857 2004 148 8 0 .632 1.04 3.09 .558 .607 .359 2.70 .450 -.004 1 67 GMT027A 1914 2004 91 5 0 .613 1.75 3.22 .663 .633 .269 2.74 .604 -.069 1 68 GMT027B 1906 2004 99 5 0 .650 .22 .58 .104 .739 .256 2.85 .559 -.003 1 69 GMT028A 1908 2004 97 5 0 .524 .13 .52 .082 .743 .334 2.92 .527 -.006 1 70 GMT028B 1904 2004 101 5 0 .600 .13 .33 .067 .759 .343 2.61 .481 .006 2 71 GMT029A 1892 2004 113 6 0 .470 .10 .24 .047 .684 .283 2.78 .470 .010 3 72 GMT029B 1878 2004 127 7 0 .422 .09 .27 .061 .763 .344 2.90 .554 -.053 2

158

Appendix A1. continued.

Corr //------Unfiltered ------\\ //---- Filtered -----\\ No. No. No. with Mean Max Std Auto Mean Max Std Auto AR Seq Series Interval Years Segmt Flags Master msmt msmt dev corr sens value dev corr () ------73 GMT030A 1927 2004 78 4 1 .384 .12 .24 .057 .614 .296 2.51 .469 -.132 3 74 GMT030B 1927 2004 78 4 1 .390 .10 .25 .052 .587 .310 2.43 .348 -.061 1 75 GMT031A 1900 2004 105 5 1 .499 1.41 3.77 .915 .870 .267 2.70 .524 -.046 1 76 GMT032A 1899 2004 106 6 1 .468 .21 .40 .069 .766 .187 2.67 .505 .003 4 77 GMT032B 1885 2004 120 6 1 .571 .18 .36 .075 .773 .236 2.64 .551 .020 1 78 GMT033A 1937 2004 68 4 0 .587 1.29 2.40 .474 .579 .285 2.49 .382 .011 2 79 GMT034A 1940 1990 51 2 1 .410 .22 .49 .097 .753 .204 2.79 .561 -.022 1 80 GMT035A 1902 2004 103 5 0 .620 1.97 5.23 .828 .650 .277 2.72 .433 -.103 2 81 GMT035B 1895 2004 110 6 0 .622 .21 .45 .080 .732 .218 2.65 .512 .027 1 82 GMT036A 1760 2004 245 12 0 .528 .08 .41 .055 .873 .228 2.61 .422 .007 1 83 GMT036B 1818 2004 187 10 0 .715 .07 .16 .028 .670 .238 2.83 .339 .015 2 84 GMT037A 1856 1971 116 6 0 .608 1.94 5.38 1.152 .794 .295 2.62 .470 -.018 1 85 GMT037B 1860 1980 121 6 0 .612 1.80 5.12 1.044 .797 .271 2.67 .402 .017 1 86 GMT038A 1883 2005 123 6 0 .549 2.52 6.22 1.219 .741 .261 2.59 .363 -.070 1 87 GMT038B 1883 2005 123 6 2 .396 2.22 6.08 1.138 .727 .316 2.56 .415 .007 1 88 GMT041A 1970 2005 36 1 0 .502 2.45 4.72 .889 .653 .249 2.62 .668 -.139 2 89 GMT042A 1968 2005 38 1 1 .355 1.49 2.91 .669 .837 .190 3.06 .678 .042 1 90 GMT042B 1968 2005 38 1 0 .573 1.47 2.75 .594 .863 .153 2.92 .614 -.009 1 91 GMT043A 1877 1953 77 4 0 .573 1.79 4.74 .815 .531 .340 2.69 .524 -.027 1 92 GMT044A 1863 1963 101 5 0 .678 1.12 2.59 .515 .710 .299 2.65 .394 -.027 1 93 GMT045A 1861 2005 145 7 0 .589 1.02 4.01 .692 .900 .261 2.59 .363 -.012 1 94 GMT045B 1861 2005 145 7 0 .519 .84 3.75 .511 .815 .259 2.68 .557 .006 1 95 GMT047A 1902 2005 104 5 0 .484 1.92 4.48 .806 .785 .246 2.66 .466 -.051 2 96 GMT047B 1902 1950 49 2 0 .516 2.58 4.88 1.014 .652 .285 2.52 .459 -.021 1 97 GMT048A 1909 2005 97 5 2 .421 .81 2.04 .405 .756 .282 2.66 .410 -.013 1 98 GMT048B 1909 2005 97 5 0 .544 1.08 2.35 .452 .613 .286 2.72 .499 .013 3 99 GMT050A 1943 2005 63 3 0 .548 1.87 3.59 .663 .677 .239 2.54 .423 -.043 1 100 GMT050B 1943 2005 63 3 0 .507 1.58 2.71 .564 .682 .251 2.87 .627 .037 1 101 GMT051A 1897 2005 109 6 0 .553 2.11 4.81 .692 .538 .240 2.90 .602 -.006 1 102 GMT051B 1885 2005 121 6 0 .616 1.35 2.47 .381 .489 .217 2.79 .439 -.037 3 103 GMT053A 1850 2005 156 8 1 .589 1.51 7.72 1.094 .728 .247 3.00 .479 -.022 1 104 GMT053B 1865 2005 141 7 1 .634 1.19 3.61 .756 .875 .245 2.71 .406 -.002 1 105 GMT054A 1856 2005 150 8 2 .551 1.55 4.25 .906 .842 .253 2.57 .358 .019 1 106 GMT054B 1872 2005 134 7 0 .647 1.26 3.69 .671 .802 .257 2.74 .478 -.071 1 107 GMT055A 1860 1940 81 4 0 .611 1.87 4.37 .852 .663 .261 2.67 .428 -.072 1 108 GMT055B 1862 2005 144 7 0 .556 2.19 5.85 1.130 .710 .295 2.53 .336 -.022 1

159

Appendix A1. continued.

Corr //------Unfiltered ------\\ //---- Filtered -----\\ No. No. No. with Mean Max Std Auto Mean Max Std Auto AR Seq Series Interval Years Segmt Flags Master msmt msmt dev corr sens value dev corr () ------109 GMT056A 1862 2005 144 7 2 .456 1.76 4.38 .784 .759 .234 2.65 .439 -.052 1 110 GMT056B 1890 2005 116 6 0 .457 1.09 3.04 .487 .669 .277 2.60 .399 -.067 1 111 GMT057A 1862 2005 144 7 0 .548 .95 3.07 .627 .870 .246 2.70 .621 -.051 1 112 GMT057B 1862 2005 144 7 0 .656 .90 2.57 .522 .781 .256 2.75 .474 -.003 1 113 GMT058A 1874 2005 132 7 0 .565 .64 2.88 .443 .728 .331 2.72 .488 -.015 1 114 GMT058B 1885 2005 121 6 0 .595 .79 2.07 .367 .725 .286 2.54 .384 -.061 1 115 GMT059A 1862 2005 144 7 0 .593 1.20 3.62 .616 .676 .326 2.73 .540 -.006 1 116 GMT059B 1862 2005 144 7 0 .519 .81 2.99 .433 .599 .346 2.78 .421 -.042 3 117 GMT060A 1868 1998 131 6 0 .639 .77 1.82 .408 .723 .344 2.78 .544 -.010 2 118 GMT060B 1868 1940 73 4 0 .585 1.04 2.53 .491 .620 .313 2.67 .504 .118 1 119 GMT061A 1928 2005 78 4 0 .421 1.63 4.31 .997 .800 .265 2.75 .486 -.038 2 120 GMT063A 1948 2005 58 3 1 .375 2.59 5.67 .878 .654 .195 2.91 .526 -.091 1 121 GMT063B 1948 2005 58 3 0 .352 2.39 5.58 .765 .648 .199 2.79 .631 .027 1 122 GMT064A 1904 2005 102 5 0 .475 .84 2.06 .342 .702 .266 2.89 .583 -.045 2 123 GMT064B 1904 2005 102 5 0 .488 .87 2.85 .519 .656 .359 2.65 .391 -.066 1 124 GMT065A 1941 2005 65 3 1 .331 1.91 6.27 .913 .525 .289 2.70 .568 .063 1 125 GMT066A 1930 2005 76 4 1 .541 1.10 4.31 .848 .765 .378 2.43 .347 -.080 1 126 GMT066B 1940 2005 66 3 1 .334 1.16 2.62 .500 .646 .344 2.61 .538 -.049 1 127 GMT101A 1845 1950 106 5 1 .384 1.46 2.99 .630 .629 .279 2.66 .440 .005 1 128 GMT101B1 1842 1915 74 3 1 .365 1.71 3.81 .706 .710 .247 2.78 .621 -.125 1 129 GMT101B2 1841 1908 68 3 0 .638 1.91 5.17 .883 .768 .215 2.69 .477 -.047 1 130 GMT102 1850 1918 69 3 0 .513 1.55 3.06 .692 .772 .261 2.45 .460 .109 1 131 GMT103 1869 1967 99 5 0 .593 1.70 3.83 .843 .679 .290 2.80 .530 .018 1 132 GMT104 1880 1963 84 4 0 .561 1.36 3.50 .661 .656 .314 2.76 .573 -.016 1 133 GMT105 1885 2005 121 6 1 .497 2.02 4.60 .820 .711 .257 2.60 .360 -.028 1 134 GMT106 1841 1927 87 4 0 .477 1.09 3.51 .676 .794 .308 2.70 .416 -.042 1 135 GMT107 1806 1928 123 6 1 .399 .87 1.56 .298 .610 .238 2.73 .450 .060 1 136 gmt107s 1934 1999 66 3 0 .528 1.26 4.30 .954 .902 .196 2.85 .563 -.067 1 137 GMT108 1883 2000 118 6 0 .487 .62 1.91 .345 .601 .343 2.83 .493 -.010 2 138 GMT110 1877 2001 125 7 4 .406 1.28 3.84 .807 .839 .252 2.48 .334 -.064 2 139 GMT111 1864 1999 136 6 1 .516 1.07 3.82 .837 .893 .339 2.76 .508 -.004 1 140 GMT111B 1864 2001 138 7 2 .576 1.41 5.47 1.022 .842 .297 2.64 .452 -.030 2 141 GMT113 1922 2001 80 4 1 .247 1.74 5.36 .983 .590 .335 2.65 .554 -.032 1 142 GMT114 1920 2001 82 4 1 .489 2.10 5.92 1.028 .665 .286 2.51 .417 -.031 2 143 GMT119A 1851 2001 151 8 0 .702 1.49 6.54 1.024 .897 .209 2.70 .382 -.010 1 144 GMT119B 1851 2001 151 8 2 .527 1.87 5.74 .876 .801 .221 2.50 .332 -.021 1

160

Appendix A1. continued.

Corr //------Unfiltered ------\\ //---- Filtered -----\\ No. No. No. with Mean Max Std Auto Mean Max Std Auto AR Seq Series Interval Years Segmt Flags Master msmt msmt dev corr sens value dev corr () ------145 GMT121 1856 1984 129 7 3 .381 1.40 4.48 .874 .804 .257 2.58 .382 .031 4 146 GMT122 1866 2000 135 7 2 .532 1.87 4.91 1.098 .754 .302 2.80 .468 -.034 1 147 GMT123 1778 1968 191 10 1 .584 .82 2.13 .407 .701 .300 2.55 .324 -.013 2 148 GMT123B 1871 1940 70 4 2 .346 1.16 3.76 .750 .842 .288 2.94 .661 .002 1 149 GMT124 1790 1928 139 7 0 .515 .91 2.52 .375 .644 .257 2.76 .483 .022 1 150 GMT125 1729 1954 226 11 3 .501 1.20 4.40 .665 .788 .283 2.55 .347 .023 1 151 GMT126A 1739 1928 190 10 2 .484 1.22 3.22 .550 .784 .234 2.82 .400 .030 1 152 GMT126B 1737 1927 191 10 3 .470 1.13 4.03 .712 .929 .193 2.54 .338 -.021 2 153 GMT127A 1829 1947 119 6 0 .443 .65 1.37 .319 .707 .343 2.66 .484 -.045 1 154 GMT127B 1764 1958 195 9 2 .547 1.04 2.39 .446 .687 .311 2.51 .393 -.026 1 155 GMT128 1785 1938 154 7 1 .451 1.50 4.76 .868 .842 .222 2.67 .365 .008 1 156 GMT129 1890 2005 116 6 0 .485 .58 2.44 .369 .850 .240 2.73 .506 .018 1 157 GMT130A 1884 2000 117 6 3 .300 2.04 6.21 1.221 .849 .235 2.41 .296 .027 1 158 GMT130C 1850 1925 76 4 1 .394 2.93 6.60 1.398 .770 .293 2.67 .455 .037 1 159 GMT131 1854 2001 148 8 2 .534 1.06 2.38 .507 .869 .194 2.53 .353 .028 1 160 GMT132 1896 1998 103 5 0 .676 1.11 2.13 .485 .747 .284 2.65 .553 -.049 1 161 GMT133 1823 1999 177 8 0 .649 1.22 3.35 .663 .835 .256 2.52 .367 -.025 3 162 GMT135 1836 1950 115 6 2 .341 .94 3.39 .586 .788 .293 2.87 .487 -.010 1 163 GMT136 1873 1990 118 6 4 .343 1.68 4.48 1.107 .837 .269 2.84 .533 -.014 2 164 GMT137 1782 1908 127 6 1 .577 1.80 4.46 .756 .648 .261 2.76 .436 -.059 2 165 GMT137B 1850 1927 78 4 0 .636 1.15 2.46 .430 .618 .287 2.68 .510 .041 1 166 GMT138 1890 1989 100 5 3 .354 1.65 6.20 1.198 .719 .295 2.63 .531 -.059 2 167 GMT139 1855 1990 136 7 4 .261 1.44 4.39 .655 .640 .282 3.01 .483 .043 1 168 GMT140 1864 1930 67 3 0 .482 1.70 3.20 .739 .698 .340 2.42 .469 -.014 1 169 GMT141 1807 1952 146 7 4 .353 .95 2.45 .444 .683 .278 2.74 .526 .094 1 170 GMT145 1900 1999 100 4 0 .494 1.72 4.33 .697 .532 .327 2.62 .438 -.059 2 171 GMT149A 1848 2000 153 8 0 .590 1.46 3.73 .701 .716 .281 2.76 .539 -.071 2 172 GMT149B 1848 2000 153 8 0 .594 1.27 3.60 .745 .739 .296 2.71 .426 -.007 1 173 GMT152 1712 1801 90 5 2 .304 1.10 3.26 .411 .399 .263 2.86 .484 .060 1 174 GMX301AR 1697 1930 234 12 0 .534 1.11 3.25 .519 .770 .241 2.57 .392 -.014 1 175 GMX301B 1697 1890 194 10 2 .526 1.21 3.78 .630 .770 .277 2.58 .337 .007 1 176 GMX3021R 1794 1916 123 6 2 .419 1.26 2.70 .559 .739 .257 2.58 .378 .001 1 177 GMX3022 1800 1933 134 6 3 .313 1.43 3.56 .611 .729 .266 2.65 .508 -.036 1 178 GMX305A 1731 1897 167 8 0 .457 .60 1.81 .338 .756 .338 2.72 .514 -.014 2 179 GMX305B 1726 1917 192 9 5 .336 .55 2.01 .338 .805 .315 2.82 .429 -.029 2 180 GMX307 1684 2006 323 16 6 .366 .86 6.47 .919 .903 .291 2.81 .426 .004 2

161

Appendix A1. continued.

Corr //------Unfiltered ------\\ //---- Filtered -----\\ No. No. No. with Mean Max Std Auto Mean Max Std Auto AR Seq Series Interval Years Segmt Flags Master msmt msmt dev corr sens value dev corr () ------181 GMX309A 1731 1897 167 8 1 .564 1.00 2.03 .388 .689 .272 2.59 .375 .014 1 182 GMX309B 1733 1862 130 7 0 .562 1.30 3.06 .518 .770 .205 2.46 .372 -.025 1 183 GMX401 1859 1930 72 4 0 .514 1.73 3.15 .509 .433 .235 2.67 .536 .039 1 184 GMX403 1837 1970 134 7 0 .629 1.26 3.91 .634 .799 .241 2.66 .421 -.008 1 185 GMX404 1837 1991 155 8 0 .681 1.24 4.17 .718 .794 .296 2.71 .402 .037 4 186 GMX405 1742 1902 161 8 1 .498 1.05 2.75 .523 .752 .275 2.88 .529 -.038 2 187 GMX406 1814 1999 186 9 0 .576 1.27 3.89 .899 .915 .259 2.69 .497 .046 1 ------Total or mean: 23910 1203 151 .525 1.13 7.77 .590 .750 .269 3.06 .445 -.010

162

Appendix A2. Cooper Road Trail COFECHA Output Summary Statistics.

Corr //------Unfiltered ------\\ //---- Filtered -----\\ No. No. No. with Mean Max Std Auto Mean Max Std Auto AR Seq Series Interval Years Segmt Flags Master msmt msmt dev corr sens value dev corr () ------1 COO001A 1817 2007 191 10 2 .465 1.15 4.28 .784 .884 .225 2.81 .518 -.025 1 2 COO001B 1808 2007 200 10 3 .362 1.07 2.99 .652 .891 .187 2.84 .384 -.015 1 3 COO002A 1860 2006 147 7 1 .473 1.07 2.70 .548 .714 .303 2.65 .470 -.048 1 4 COO002B 1888 2007 120 6 0 .598 1.69 3.80 .924 .813 .224 2.56 .465 -.006 1 5 COO003A 1877 2007 131 7 1 .436 1.46 4.50 1.113 .903 .233 2.59 .375 -.061 1 6 COO003B 1887 2002 116 6 0 .460 1.62 3.95 .873 .824 .212 2.59 .444 -.017 3 7 COO005A 1924 2006 83 4 0 .462 .75 2.74 .530 .802 .328 2.74 .409 -.051 1 8 COO005B 1878 2007 130 7 0 .551 1.29 3.92 .696 .719 .278 2.76 .406 .024 1 9 COO006A 1865 2007 143 7 1 .560 2.63 5.90 1.330 .736 .281 2.55 .405 .020 1 10 COO006B 1873 2007 135 7 0 .457 1.53 5.67 .939 .770 .278 2.64 .449 .020 1 11 COO007A 1874 2007 134 7 2 .548 1.98 5.07 .950 .769 .283 2.58 .477 .052 2 12 COO007B 1871 2007 137 7 0 .533 1.79 5.46 .849 .718 .311 2.73 .488 .019 1 13 COO008A 1917 2007 91 5 1 .504 1.49 4.15 .855 .875 .214 2.50 .373 .015 4 14 COO008B 1916 2007 92 5 0 .461 1.34 4.35 1.012 .911 .255 2.49 .317 -.050 2 15 COO009A 1879 2007 129 7 2 .509 1.28 3.94 .678 .757 .252 2.63 .468 -.036 2 16 COO009B 1895 2007 113 6 1 .398 1.81 4.86 .954 .748 .236 2.61 .418 -.026 2 17 COO010A 1898 2007 110 6 0 .490 1.25 4.34 .952 .883 .289 2.54 .454 -.068 1 18 COO010B 1883 2007 125 6 0 .591 1.65 5.09 .986 .855 .271 2.37 .296 .015 1 19 COO011A 1896 2007 112 6 0 .479 .87 2.09 .380 .591 .278 2.89 .595 -.027 4 20 COO011B 1899 2006 108 6 0 .529 1.62 3.94 .724 .693 .261 2.71 .501 .005 2 21 COO012A 1887 2007 121 6 1 .576 1.04 2.29 .437 .628 .287 2.79 .463 -.035 3 22 COO012B 1888 2007 120 6 0 .602 1.33 6.28 .782 .260 .400 2.86 .438 .067 1 23 COO013A 1897 2007 111 6 0 .721 1.65 3.16 .535 .479 .242 2.72 .481 .071 1 24 COO013B 1894 2007 114 6 0 .649 1.48 3.42 .578 .594 .275 2.69 .505 .019 1 25 COO014A 1890 1990 101 5 0 .577 1.45 4.18 1.095 .889 .329 2.68 .477 -.012 1 26 COO015B 1897 2003 107 6 0 .657 1.71 4.67 1.044 .790 .306 2.63 .478 .068 1 27 COO018A 1917 2007 91 5 0 .642 1.58 4.51 .737 .786 .253 2.78 .486 .115 1 28 COO018B 1932 2003 72 4 0 .624 1.92 3.05 .474 .444 .213 2.54 .466 .110 1 29 COO019A 1914 2007 94 5 0 .597 2.51 6.99 .911 .661 .242 2.66 .437 .036 1 30 COO019B 1924 2007 84 4 0 .643 1.75 3.26 .627 .676 .219 2.67 .551 -.074 1 31 COO020A 1918 2007 90 5 0 .549 2.07 5.30 .802 .762 .212 2.67 .566 -.036 1 32 COO021A 1931 2007 77 4 0 .546 1.59 3.28 .633 .722 .220 2.62 .574 -.081 1 33 COO021B 1946 2007 62 3 0 .468 1.28 2.25 .489 .791 .232 2.44 .433 -.020 1 34 COO022A 1919 2007 89 5 3 .411 2.00 4.20 .651 .668 .222 2.53 .533 .050 3 35 COO022B 1928 2007 80 4 0 .500 2.01 3.60 .647 .659 .213 2.63 .490 -.060 1 36 COO023A 1930 2007 78 4 0 .560 1.72 3.05 .459 .532 .208 2.52 .441 .023 2

163

Appendix A2. continued.

Corr //------Unfiltered ------\\ //---- Filtered -----\\ No. No. No. with Mean Max Std Auto Mean Max Std Auto AR Seq Series Interval Years Segmt Flags Master msmt msmt dev corr sens value dev corr () ------37 COO023B 1931 2007 77 4 1 .561 1.70 3.39 .519 .579 .200 2.88 .593 -.006 2 38 COO024A 1904 2006 103 5 0 .457 1.08 5.26 .972 .775 .327 3.11 .588 -.011 1 39 COO025A 1926 2007 82 4 0 .519 1.95 4.36 .918 .860 .179 2.84 .532 -.045 1 40 COO025B 1926 2007 82 4 1 .353 2.15 4.88 1.062 .830 .234 2.71 .604 -.024 2 41 COO026A 1918 2007 90 5 0 .472 2.09 7.11 1.137 .853 .225 2.89 .508 -.009 1 42 COO026B 1929 2007 79 4 0 .662 1.75 5.84 .770 .646 .222 2.75 .560 .004 1 43 COO027A 1915 2007 93 5 0 .583 1.64 7.17 1.507 .838 .331 2.70 .470 -.017 1 44 COO027B 1915 2007 93 5 0 .532 2.18 5.72 1.197 .753 .290 2.61 .358 -.008 4 45 COO028A 1928 2007 80 4 0 .649 2.25 9.48 1.224 .517 .328 2.79 .496 -.049 1 46 COO028B 1915 2007 93 5 0 .689 2.35 5.01 .785 .570 .232 2.81 .491 -.070 1 47 COO029A 1859 2007 149 8 2 .440 .85 4.88 .783 .825 .337 2.75 .488 .015 2 48 COO029B 1909 2006 98 5 2 .457 .94 3.17 .638 .879 .253 2.73 .535 .029 1 49 COO030B 1920 2007 98 4 0 .534 2.08 4.45 .758 .730 .182 2.62 .397 -.003 1 ------Total or mean: 5245 272 24 .527 1.58 9.48 .819 .740 .260 3.11 .468 -.004

164

Appendix A3. Shaw Grave Gap COFECHA Output Summary Statistics.

Corr //------Unfiltered ------\\ //---- Filtered -----\\ No. No. No. with Mean Max Std Auto Mean Max Std Auto AR Seq Series Interval Years Segmt Flags Master msmt msmt dev corr sens value dev corr () ------1 SGG002A 1860 2007 148 7 2 .465 1.03 3.23 .702 .883 .220 2.56 .341 -.030 1 2 SGG002B 1860 1970 111 5 1 .533 1.09 3.42 .790 .880 .239 2.57 .434 .014 3 3 SGG003A 1883 1979 97 4 0 .518 1.67 2.81 .627 .769 .194 2.78 .580 .036 3 4 SGG003B 1864 2007 144 7 1 .474 1.44 2.93 .596 .813 .194 2.79 .459 -.075 1 5 SGG004B 1890 2007 118 6 0 .498 .79 1.78 .329 .794 .212 2.79 .543 .007 3 6 SGG005A 1820 2007 188 9 0 .592 1.17 2.77 .580 .757 .257 2.57 .365 -.045 2 7 SGG005B 1860 1970 111 5 1 .456 .73 1.86 .363 .832 .236 2.68 .478 -.028 1 8 SGG006A 1817 2007 191 10 2 .463 1.05 3.02 .446 .692 .238 2.91 .468 .019 1 9 SGG006B 1733 2007 275 10 4 .379 .88 2.51 .393 .778 .232 2.58 .364 .014 1 10 SGG007A 1872 1998 127 6 0 .523 .51 1.38 .242 .724 .249 2.94 .569 .010 2 11 SGG009A 1917 2003 87 5 0 .595 1.96 4.47 1.072 .890 .235 2.58 .468 -.013 3 12 SGG009B 1921 2007 87 4 0 .586 2.06 5.94 .956 .791 .228 2.64 .375 .038 1 13 SGG010A 1896 2007 112 6 3 .438 .91 2.48 .633 .895 .246 2.74 .546 -.002 1 14 SGG010B 1869 2007 139 7 2 .502 .72 1.94 .415 .855 .247 2.58 .513 -.020 3 15 SGG012A 1895 2007 113 6 1 .514 .82 1.83 .331 .714 .245 2.81 .512 .025 1 16 SGG014A 1911 2007 97 5 2 .470 2.31 3.99 .770 .775 .187 2.50 .456 .003 1 17 SGG014B 1919 2004 86 5 2 .462 1.79 4.06 .853 .833 .212 2.76 .461 .060 3 18 SGG015A 1944 2007 64 3 0 .493 1.57 4.06 .904 .857 .198 2.55 .406 -.055 1 19 SGG015B 1943 2007 65 3 1 .414 2.48 5.41 1.181 .871 .204 2.57 .528 -.001 4 20 SGG016A 1931 2007 77 4 0 .508 2.14 4.98 1.178 .883 .180 2.89 .523 -.005 1 21 SGG016B 1932 2007 76 4 0 .642 2.74 5.44 1.293 .874 .171 2.60 .567 -.021 1 22 SGG017A 1922 2007 86 4 0 .809 1.69 4.59 .965 .889 .197 2.50 .401 -.014 1 23 SGG017B 1922 2007 86 4 0 .705 1.91 4.61 .972 .903 .177 2.75 .537 .022 1 24 SGG018A 1920 2007 88 4 0 .649 2.34 5.28 .978 .854 .179 2.59 .427 -.039 2 25 SGG018B 1920 2005 86 4 0 .494 1.97 5.13 1.109 .858 .224 2.58 .455 -.023 1 26 SGG019A 1925 2007 83 4 2 .509 1.88 5.66 1.563 .893 .248 2.58 .423 -.047 1 27 SGG019B 1924 2007 84 4 1 .461 1.18 4.21 1.075 .900 .299 2.71 .523 .045 2 28 SGG020A 1911 2007 97 5 0 .556 1.38 4.28 .822 .863 .202 2.73 .444 -.018 1 29 SGG020B 1910 2007 98 5 0 .583 1.50 4.25 1.190 .892 .280 2.48 .324 -.001 2 30 SGG022A 1891 1979 89 4 0 .518 1.25 3.63 .751 .770 .252 2.61 .432 -.023 1 31 SGG023A 1933 2007 75 4 0 .596 1.77 4.11 .832 .796 .296 2.65 .478 .076 1 32 SGG023B 1934 2007 74 4 0 .391 2.00 6.57 1.179 .857 .257 2.56 .426 .005 2 33 SGG024A 1922 2007 86 4 0 .589 1.90 4.73 1.163 .889 .227 2.46 .367 -.014 1 34 SGG024B 1922 2007 86 4 0 .562 2.02 4.76 1.142 .874 .223 2.44 .382 -.069 1 35 SGG025A 1899 2007 109 6 0 .555 1.07 2.66 .464 .737 .270 2.33 .316 -.035 1 36 SGG025B 1935 2007 73 4 2 .512 1.54 4.02 .642 .778 .223 2.53 .410 -.111 3

165

Appendix A3. continued.

Corr //------Unfiltered ------\\ //---- Filtered -----\\ No. No. No. with Mean Max Std Auto Mean Max Std Auto AR Seq Series Interval Years Segmt Flags Master msmt msmt dev corr sens value dev corr () ------37 SGG026B 1918 2007 90 5 0 .566 1.79 4.81 .893 .792 .229 2.45 .364 -.036 1 38 SGG027A 1889 2007 119 6 0 .671 1.38 4.07 .896 .825 .314 2.82 .477 .015 1 39 SGG027B 1882 2007 126 6 0 .632 1.69 4.61 .824 .768 .293 2.45 .328 -.041 1 40 SGG028B 1867 2007 141 7 0 .639 1.50 3.87 .773 .778 .272 2.60 .418 -.026 2 41 SGG029A 1910 2007 98 5 0 .614 2.04 5.14 .755 .519 .267 2.50 .371 .028 1 42 SGG029B 1911 2007 97 5 0 .619 2.00 4.31 .692 .488 .266 2.49 .348 .020 1 43 SGG031A 1894 2007 114 6 0 .595 1.88 3.28 .681 .644 .259 2.37 .275 .032 1 44 SGG031B 1906 2007 102 5 1 .425 1.70 3.34 .638 .710 .220 2.56 .432 -.033 1 45 SGG032A 1908 2007 100 5 1 .470 .88 3.23 .613 .840 .296 2.63 .380 .000 1 46 SGG032B 1909 2007 99 5 0 .588 1.09 3.46 .646 .772 .326 2.78 .380 .079 1 47 SGG034A 1910 2007 98 5 0 .471 14.78 47.02 11.275 .907 .288 3.48 .496 .067 1 48 SGG034B 1925 2004 80 4 0 .469 1.26 3.58 .903 .895 .239 2.71 .489 -.025 3 49 SGG035A 1854 2007 154 8 1 .473 1.22 2.78 .498 .788 .201 2.83 .512 -.035 1 50 SGG035B 1869 2007 139 7 0 .646 1.27 3.90 .746 .876 .210 2.85 .481 .047 1 51 SGG036A 1850 2007 158 8 0 .668 1.15 2.62 .519 .824 .227 2.61 .388 .022 1 52 SGG036B 1867 2007 141 7 1 .609 1.64 4.72 .840 .738 .253 2.63 .427 .054 3 53 SGG038A 1864 2007 144 7 1 .481 .86 2.80 .471 .687 .355 2.51 .297 -.073 2 54 SGG038B 1907 2007 101 5 0 .488 .61 2.30 .417 .778 .358 2.70 .471 -.110 2 55 SGG039B 1884 2007 124 6 1 .458 1.55 5.39 .783 .726 .321 2.54 .353 -.043 1 56 SGG041A 1800 2007 208 10 2 .350 .86 2.64 .467 .737 .285 2.75 .399 .026 1 57 SGG042A 1891 2007 117 6 0 .579 1.20 3.72 .851 .627 .399 2.59 .404 -.028 1 58 SGG042B 1891 2007 117 6 0 .443 1.21 3.50 .636 .704 .246 2.57 .417 -.010 1 59 SGG043A 1948 2007 60 3 0 .518 2.08 4.12 .681 .645 .209 2.67 .449 .071 1 60 SGG043B 1950 2007 58 3 0 .472 3.23 5.35 .996 .607 .210 2.59 .454 .024 1 61 SGG044B 1895 2007 113 6 1 .522 1.85 4.06 .824 .767 .248 2.68 .374 -.013 4 62 SGG045A 1875 2007 133 7 2 .476 1.43 4.71 .868 .670 .378 2.42 .324 -.042 5 63 SGG046A 1923 2007 85 4 0 .524 1.33 3.46 .769 .848 .248 2.85 .591 -.028 3 64 SGG047A 1916 2007 92 5 0 .457 1.59 5.02 1.336 .918 .268 2.67 .442 -.004 3 65 SGG047B 1916 2007 92 5 0 .483 1.39 3.60 .790 .892 .179 2.66 .375 -.019 1 66 SGG048B 1916 2005 90 5 0 .479 1.94 4.11 1.034 .843 .237 2.65 .526 .015 1 67 SGG049A 1929 2007 79 4 0 .530 2.34 5.75 1.305 .828 .207 2.64 .580 -.050 1 68 SGG049B 1929 2007 79 4 0 .651 2.83 5.50 1.385 .902 .170 2.60 .391 -.004 1 69 SGG050A 1924 2007 84 4 0 .699 1.65 4.15 .968 .893 .196 2.62 .535 .092 1 70 SGG050B 1924 2007 84 4 0 .586 1.83 4.78 .983 .862 .195 2.61 .516 .014 1 71 SGG051A 1870 2007 138 7 0 .547 1.42 3.35 .635 .706 .283 2.61 .373 .034 1 72 SGG051B 1869 2007 139 7 1 .596 1.14 2.77 .504 .659 .294 2.45 .303 .003 1

166

Appendix A3. continued.

Corr //------Unfiltered ------\\ //---- Filtered -----\\ No. No. No. with Mean Max Std Auto Mean Max Std Auto AR Seq Series Interval Years Segmt Flags Master msmt msmt dev corr sens value dev corr () ------73 SGG052A 1844 2007 164 8 0 .583 1.24 3.61 .624 .756 .302 2.46 .369 -.049 2 74 SGG052B 1843 2007 165 8 0 .594 1.40 3.18 .647 .712 .273 2.51 .345 -.041 1 75 SGG053A 1908 2007 100 5 0 .495 .97 2.49 .472 .763 .270 2.59 .352 -.009 1 76 SGG053B 1923 2007 85 4 0 .622 1.89 6.13 1.019 .785 .232 2.60 .388 -.053 1 77 SGG054A 1892 2007 116 6 0 .702 1.30 3.40 .829 .809 .347 2.59 .369 -.051 1 78 SGG054B 1893 2007 115 6 0 .667 1.66 4.95 .887 .761 .322 2.57 .398 -.021 1 79 SGG055A 1845 2007 163 8 0 .561 1.51 4.53 .726 .627 .314 2.57 .369 -.045 2 80 SGG055B 1915 2007 93 5 0 .593 1.97 5.31 .962 .784 .263 2.52 .370 -.068 1 ------Total or mean: 8807 437 39 .539 1.60 47.02 .866 .784 .254 3.48 .423 -.009

167

Appendix A4. Maynard Creek COFECHA Output Summary Statistics.

Corr //------Unfiltered ------\\ //---- Filtered -----\\ No. No. No. with Mean Max Std Auto Mean Max Std Auto AR Seq Series Interval Years Segmt Flags Master msmt msmt dev corr sens value dev corr () ------1 MCR001B 1945 2007 63 3 0 .509 1.56 6.00 1.005 .793 .256 2.72 .492 -.035 1 2 MCR002A 1943 2007 65 3 1 .499 1.18 3.53 .605 .769 .285 2.68 .492 -.105 1 3 MCR003A 1962 2007 46 2 0 .443 2.12 3.89 .548 .503 .189 2.64 .468 .019 1 4 MCR003B 1963 2007 45 2 1 .470 2.97 4.98 .796 .521 .193 2.47 .442 -.024 1 5 MCR004B 1887 2007 121 6 0 .586 1.53 3.21 .555 .640 .244 2.74 .539 -.015 1 6 MCR005B 1934 2007 74 4 0 .457 1.92 5.90 1.520 .889 .231 2.46 .498 -.014 1 7 MCR006A 1894 2007 114 6 0 .476 1.13 3.81 .800 .868 .279 2.68 .468 .065 1 8 MCR006B 1894 2005 112 6 0 .479 1.01 3.98 .850 .894 .297 2.78 .495 -.012 1 9 MCR007A 1926 2007 82 4 0 .380 .75 1.56 .254 .335 .296 2.94 .517 .039 1 10 MCR008B 1838 2007 170 8 1 .392 .86 2.49 .498 .857 .219 2.66 .402 .000 1 11 MCR009B 1900 2007 108 5 3 .339 .94 2.35 .487 .775 .293 2.69 .425 -.039 3 12 MCR010A 1920 2007 88 4 2 .379 1.38 4.11 .860 .745 .325 2.56 .426 -.048 1 13 MCR011A 1892 2007 116 6 0 .641 1.30 3.40 .829 .809 .347 2.59 .369 -.051 1 14 MCR011B 1893 2007 115 6 0 .712 1.66 4.95 .887 .761 .322 2.57 .398 -.021 1 15 MCR012A 1845 2007 163 8 1 .475 1.51 4.53 .726 .627 .314 2.57 .369 -.045 2 16 MCR012B 1915 2007 93 5 0 .568 1.97 5.31 .962 .784 .263 2.52 .370 -.068 1 17 MCR013A 1856 2007 152 8 0 .637 1.44 4.34 .651 .704 .265 2.58 .351 -.007 1 18 MCR014A 1928 2007 80 4 0 .586 1.65 3.01 .563 .693 .215 2.70 .498 .022 1 19 MCR014B 1896 2007 112 6 0 .594 2.08 3.57 .735 .642 .242 2.43 .294 .018 1 20 CAL001A 1917 2003 87 5 1 .529 1.96 4.47 1.072 .890 .235 2.58 .468 -.013 3 21 CAL001B 1921 2007 87 4 0 .554 2.06 5.94 .956 .791 .228 2.64 .375 .038 1 22 CAL002A 1937 2007 71 4 0 .481 2.45 5.29 .966 .704 .206 2.63 .536 .072 1 23 CAL003A 1922 2007 86 4 0 .548 1.90 4.73 1.163 .889 .227 2.46 .367 -.014 1 24 CAL003B 1922 2007 86 4 0 .573 2.02 4.76 1.142 .874 .223 2.44 .382 -.069 1 25 CAL004B 1920 2007 88 4 0 .456 1.47 3.03 .782 .826 .230 2.69 .481 -.059 2 26 CAL005A 1874 1965 92 5 0 .666 .88 2.69 .439 .597 .316 2.88 .530 -.072 3 27 CAL006A 1910 2007 98 5 0 .468 14.78 47.02 11.275 .907 .288 3.48 .496 .067 1 28 CAL006B 1925 2004 80 4 0 .465 1.26 3.58 .903 .895 .239 2.71 .489 -.025 3 29 CAL007A 1930 2007 78 4 0 .598 1.52 4.06 .861 .884 .196 2.64 .566 .090 1 30 CAL007B 1924 2007 84 4 0 .602 1.83 4.78 .983 .862 .195 2.61 .516 .014 1 31 CAL008A 1943 2007 65 3 2 .352 .86 2.53 .629 .881 .232 2.88 .535 -.091 1 32 CAL009A 1908 2007 100 5 1 .400 .97 2.49 .472 .763 .270 2.59 .352 -.009 1 33 CAL009B 1923 2007 85 4 0 .601 1.89 6.13 1.019 .785 .232 2.60 .388 -.053 1 34 CAL010B 1930 2007 78 4 1 .519 1.09 2.78 .757 .854 .281 2.56 .554 .008 1 ------Total or mean: 3184 159 14 .517 1.90 47.02 1.102 .768 .260 3.48 .442 -.013

168

Appendix A5. Gregory Ridge Trail COFECHA Output Summary Statistics.

Corr //------Unfiltered ------\\ //---- Filtered -----\\ No. No. No. with Mean Max Std Auto Mean Max Std Auto AR Seq Series Interval Years Segmt Flags Master msmt msmt dev corr sens value dev corr () ------1 GRT1010A 1824 2006 183 9 0 .509 .96 21.05 1.819 .239 .313 3.20 .424 .017 1 2 GRT1010B 1834 2006 173 9 0 .627 1.07 5.88 1.033 .781 .314 2.78 .380 .007 1 3 GRT1013A 1895 2006 112 6 1 .433 1.58 4.00 1.060 .849 .284 2.77 .477 -.010 1 4 GRT1024B 1828 2006 179 9 0 .545 1.24 3.94 .860 .922 .223 2.86 .573 .001 1 5 GRT1032A 1952 2006 55 3 1 .463 2.82 5.01 .763 .617 .182 2.64 .531 .029 4 6 GRT1018B 1836 2006 171 9 1 .518 1.07 3.66 .691 .847 .283 2.94 .552 .084 1 7 GRT1O24A 1816 2006 191 10 2 .585 .95 3.93 .828 .901 .260 2.79 .443 .101 1 8 GRT1011A 1921 2006 86 4 0 .627 1.17 3.00 .509 .681 .288 2.70 .585 .068 1 9 GRT1011B 1920 2006 87 4 0 .685 1.28 3.24 .524 .658 .261 2.70 .546 .016 1 10 GRT1023B 1864 2002 139 7 0 .546 1.46 3.44 .658 .779 .243 2.67 .516 .000 1 11 GRT1005B 1887 2006 120 6 1 .488 1.90 4.03 .663 .629 .244 2.54 .365 .041 3 12 GRT1012A 1894 2006 113 6 0 .463 1.14 3.81 .798 .867 .279 2.71 .499 .054 1 13 GRT1012B 1894 2005 112 6 0 .596 1.01 3.98 .850 .894 .297 2.78 .495 -.012 1 14 GRT1007A 1883 2006 124 6 0 .636 1.81 4.51 1.011 .816 .269 2.72 .420 .015 1 15 GRT1007B 1902 2006 105 5 0 .559 1.56 4.27 1.116 .871 .318 2.58 .452 -.004 1 16 GRT1002A 1913 2001 89 5 1 .419 2.11 5.65 1.058 .691 .289 2.64 .536 -.079 2 17 GRT1008A 1894 2005 112 6 0 .496 1.39 3.03 .560 .693 .247 2.71 .506 -.035 1 18 GRT1015B 1925 1987 63 3 1 .528 1.58 4.32 1.046 .809 .322 2.76 .591 -.003 2 19 GRT1027A 1906 2006 101 5 1 .410 2.38 5.22 .994 .816 .206 2.70 .512 -.002 1 20 GRT1030A 1928 1997 70 3 0 .485 1.54 2.67 .572 .774 .231 2.58 .522 .157 1 21 GRT1016A 1894 2006 113 6 3 .414 1.67 3.58 .639 .718 .206 2.69 .445 .090 1 22 GRT1034A 1900 2006 107 5 0 .518 1.20 4.01 .522 .718 .203 2.69 .429 -.034 1 23 GRT1034B 1900 2006 107 5 1 .503 1.05 2.57 .355 .650 .222 2.74 .530 -.032 1 24 GRT1008B 1879 2006 128 7 1 .556 1.24 5.13 .819 .814 .262 2.85 .548 -.030 1 25 GRT1009A 1908 1995 88 4 0 .624 1.02 2.18 .488 .761 .259 2.74 .449 .049 1 26 GRT1009B 1908 1994 87 4 0 .538 1.03 2.13 .486 .739 .248 2.83 .524 .047 1 27 GRT1013B 1900 2006 107 5 0 .499 1.46 3.62 .875 .881 .209 2.73 .453 -.064 1 28 GRT1014A 1901 2006 106 5 0 .508 1.53 3.81 .903 .894 .262 2.76 .612 .085 1 29 GRT1014B 1883 2006 124 6 0 .534 2.19 4.78 1.099 .817 .264 2.73 .474 -.029 1 30 GRT1020A 1829 2006 178 9 0 .532 .84 4.67 .611 .737 .281 2.66 .366 .001 1 31 GRT1020B 1830 1928 99 5 0 .575 .77 2.87 .451 .380 .316 2.69 .520 -.029 1 32 GRT1017A 1815 2006 192 10 5 .399 .76 2.88 .462 .887 .245 2.77 .481 -.016 2 33 GRT1026A 1779 1969 191 8 0 .517 1.19 4.94 .831 .703 .333 2.56 .350 .008 1 34 GRT1003A 1891 2006 116 6 1 .426 .94 2.23 .521 .722 .365 2.73 .481 -.048 1 35 GRT1005A 1881 2006 126 6 0 .570 1.85 4.28 .806 .810 .228 2.55 .413 .040 1 36 GRT1033B 1847 2001 155 8 0 .596 .74 2.17 .438 .824 .260 2.57 .423 -.034 1

169

Appendix A5. continued.

170

VITA

Christine Biermann grew up in Syracuse, New York and learned to love the outdoors during summers on Otisco Lake. In 2007, she graduated from the State

University of New York at Geneseo with a Bachelor of Arts in Geography and a minor in

English. She was first introduced to tree-ring science in her undergraduate coursework and at the 2007 North American Dendroecological Fieldweek in Great Smoky Mountains

National Park. The following August, she joined the Geography program at the

University of Tennessee and began working toward a Master of Science degree. Along the way, she served as a Teaching Assistant for both Geography of the Natural

Environment and World Regional Geography. In 2007, Christine was awarded a J.

Wallace and Katie Dean Fellowship from the College of Arts and Sciences at the

University of Tennessee. The following year, she began a three-year Graduate Research

Fellowship funded by the National Science Foundation. Christine received her Master of

Science degree in Geography from the University of Tennessee in August 2009. After graduation, Christine will begin work on a doctoral degree in Geography at The Ohio

State University.

171