/ 7 LIBRARY
,_ "I Michigan State
University

 

This is to certify that the
dissertation entitled

CLIMATE TRENDS AND SYNOPTIC PATTERNS
ASSOCIATED WITH MAJOR PRECIPITATION EVENTS
OF THE SOUTHERN APPALACHIANS

presented by

Ryan P. Shadbolt

has been accepted towards fulfillment
of the requirements for the

PhD. degree in Geography

 

 

r-
(t It’d/LR I

w v

/' (Major Professor’s Signature

I l

 

('

Date

MSU is an Affirmative Action/Equal Opportunity Employer

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:IProj/Acc&Pres/ClRC/DateDuo.Indd

7 7 V , i A ‘ fi _

CLIMATE TRENDS AND SYNOPTIC PATTERNS ASSOCIATED WITH MAJOR
PRECIPITATION EVENTS OF THE SOUTHERN APPALACHIANS

By

Ryan P. Shadbolt

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Geography

2009

ABSTRACT

CLIMATE TRENDS AND SYNOPTIC PATTERNS ASSOCIATED WITH MAJOR
PRECIPITATION EVENTS OF THE SOUTHERN APPALACHIANS

W

MmR$MWh

This study focuses on the southern Appalachians, a region of pronounced
biodiversity. While environmental change in the area is documented, our understanding
of associated regional climate change remains less certain. For this reason, surface
observations of minimum, maximum, and mean temperature, total precipitation, and total
snowfall from 463 stations within the region were obtained from the National Climatic
Data Center's Cooperative Observer Program to facilitate the construction of a baseline
climatology focused on temporal trends and spatial clusters.

F irst-order temporal trends show that the southern Appalachian region
experienced statistically significant cooling trends during 1931-2006, although two-phase
regression results also reveal that statistically significant warming trends emerged during
recent decades. This warming was most evident in wintertime and minimum
temperatures with January temperatures warming as much as 4.00°C between the mid
1970s and 2006. Overall, diurnal and annual temperature ranges decreased, while annual
precipitation increased over the study period. Snowfall totals decreased at low and mid
elevations and a statistically significant increase occurred at high elevations. The recent
warming and decreasing temperature ranges for this locale support trends presented
within recent reports from Intergovernmental Panel on Climate Change and the United

States Global Change Research Program. The effectiveness of the cluster analysis was

limited by the availability and spatial resolution of the observation stations. However, the
resulting clusters still depicted relationships between stations of similar geographic
locations and elevations and up to five climate regimes were defined for the region.
Additionally, a study of synoptic patterns during 1437 major precipitation events
spanning 1979-2006 revealed up to 11 common synoptic patterns. The synoptic patterns
were created by objective, subjective, and hybrid clustering approaches. The results from
the objective approach differed from the subjective approach, and patterns uncovered by
the hybrid approach had similarities to both the objective and subjective approaches.
Since results varied depending on the cluster approach used, it is recommended that
future studies should not rely solely on one approach when clustering synoptic patterns.
A temporal analysis showed that some synoptic patterns became more or less
frequent over the 1979-2006 period and some trends were statistically significant.
Precipitation events associated with low-pressure and frontal systems became more
frequent, whereas precipitation events associated primarily with orographic lifting or
thermodynamics became less common. The results from this dissertation build upon
those from previous studies and provide a more complete summary of the region's recent
climate, one that should help inform future climate studies of the region and inform those

who manage the region's ecosystems.

ACKNOWLEDGEMENTS

Many individuals helped make the completion of this dissertation and my Ph.D.
program possible. I am grateful to all Geography faculty, staff, and graduate students at
Michigan State University and have enjoyed being a member of the department. I wish
to thank the department's willing staff: Sharon Ruggles, Judy Reginek, and Claudia
Brown. Tech supporters, Jim Brown and Wilson Ndovie, frequently assisted me with
computer issues. The virtual course coordinator, Beth Weisenborn, was of great
assistance during my time instructing the online version of the Geography of the US. and
Canada course. The past and present graduate program coordinators, Dr. Randy Schaetzl
and Dr. Antoinette WinklerPrins were always happily available to lend advice during the
course of my MA. and Ph.D. programs.

During my time at MSU, I was fortunate to receive funding from a variety of
sources. I am appreciative of the Geography Department for the teaching and research
assistantship opportunities I was awarded through the years. I am also thankful for the
research opportunities offered to me by Dr. Julie Winkler, Dr. Joe Messina, and Dr.
Ashton Shortridge. Dr. Jay Charney, Dr. Brian Potter, Dr. Warren Heilman, and Xindi
Bian also graciously offered me research positions at the US. Forest Service office
during some summer months as well.

In particular, completing this dissertation would not have been possible without
the participation of my five committee members. Dr. Jeff Andresen, Dr. Catherine Yansa,
and Dr. Jay Chamey each provided valuable recommendations during the early stages of

solidifying the objectives for this research project. Dr. Pang-Ning Tan provided essential

iv

suggestions for the cluster analysis portions of the study. My advisor, Dr. Jay Harman,
deserves a large portion of the credit for sparking my interest in the southern Appalachian
region, and ultimately, the initiation of this study. I have had the pleasure to accompany
him on three separate field trips to the region and was taught to view it through the eyes
of a geographer. His thoughtful questions, suggestions, and dedicated editing helped
improve the text of this document with each passing revision.

Lastly, and most importantly, I wish to thank my family and friends for their
patience and never-ending support throughout the years of my PhD. program. My wife,
Jennifer Lamb, has been beside me every step of the way and I look forward to the future
travels of our journey. I am certain at least a few trips to the southern Appalachians will

be included in the itinerary.

TABLE OF CONTENTS

LIST OF TABLES .............................................................................................................. x
LIST OF FIGURES .......................................................................................................... xi
CHAPTER 1
INTRODUCTION .............................................................................................................. 1
1.1. Overview and Statement of Problem ............................................................................ 1
1.2. Ecological Change in the Region .................................................................................. 4
1.3. Research Goals .............................................................................................................. 6
1.4. Organization of the Dissertation ................................................................................... 7
CHAPTER 2
LITERATURE REVIEW .................................................................................................. 8
2.1. General Climatologies of the Region ............................................................................ 8
2.2. Synoptic Patterns Impacting Southern Appalachian Precipitation ............................. 13
2.3. Methods of Cluster Analysis ....................................................................................... 19
2.3.1. Overview ...................................................................................................... 19
2.3.2. Hierarchical Clustering Techniques ............................................................ 2O
2. 3. 3. Non-Hierarchical Clustering Techniques .................................................... 21
2. 3. 4. Number of Clusters ...................................................................................... 22
2.3.5. Distance Measures ....................................................................................... 23
2.4. Relevant Cluster Studies ............................................................................................. 25
2.5. Research Questions and Objectives ............................................................................ 31
CHAPTER 3
STUDY AREA, DATA, AND METHODS ..................................................................... 34
3.1. Study Area ................................................................................................................... 34
3.2. Data for the Baseline Climatology .............................................................................. 35
3.3. Data for the Synoptic Climatology ............................................................................. 37
3.4. Procedures for the Baseline Climatology ................................................................... 39
3.4.1. Acquiring the Data and Quality Control ..................................................... 39
3.4.2. Cluster Analysis ........................................................................................... 43
3.5. Procedures for the Synoptic Climatology ................................................................... 46
3. 5. 1. Acquiring the Data and Quality Control ..................................................... 46
3.5.2. Clustering Synoptic Patterns ....................................................................... 51
CHAPTER 4
RESULTS AND DISCUSSION OF THE TEMPORAL ANALYSIS .......................... 55
4.1. Overview ..................................................................................................................... 55
4.2. Temporal Results of Temperature ............................................................................... 57
4. 2. 1. January ........................................................................................................ 57
4.2.2. April ............................................................................................................. 59

vi

4. 2. 3. July ............................................................................................................... 60

4.2.4. October ........................................................................................................ 62
4.2.5. Annual .......................................................................................................... 63
4.3. Temporal Results of Precipitation and Snowfall ......................................................... 65
4.3.1. January ........................................................................................................ 65
4. 3. 2. April ............................................................................................................. 68
4.3.3. July ............................................................................................................... 70
4.3.4. October ........................................................................................................ 71
4.3.5. Annual .......................................................................................................... 73
4.4. Discussion ................................................................................................................... 77
CHAPTER 5
RESULTS AND DISCUSSION OF THE SPATIAL ANALYSIS ................................. 82
5.1. Overview ..................................................................................................................... 82
5.2. Spatial Results of Temperature ................................................................................... 84
5.2.1. January ........................................................................................................ 84
5.2.2. July ............................................................................................................... 87
5.3. Spatial Results of Precipitation ................................................................................... 90
5. 3. 1. January ........................................................................................................ 90
5.3.2. July ............................................................................................................... 93
5.4. Spatial Results of Snowfall ......................................................................................... 96
5. 4. 1. January ........................................................................................................ 96
5.5. Spatial Results for All Variables ................................................................................. 99
5. 5. 1. January ........................................................................................................ 99
5.5.2. July ............................................................................................................. 103
5.6. Discussion ................................................................................................................. 107
CHAPTER 6
RESULTS AND DISCUSSION OF THE SYNOPTIC CLIMATOLOGY ............... 112
6.1. Overview ................................................................................................................... 112
6.2. Objective Clustering Results ..................................................................................... 113
6.2.1. Clusters 2-5 and 7-9 ................................................................................... 113
6.2.2. Clusters 1, 6, 10, and 11 ............................................................................ 120
6.3. Subjective Clustering Results ................................................................................... 124
6.3.1. Closed Low (CL) ........................................................................................ 124
6.3.2. Northerly, Behind Low, No Fronts (NbLnF) .............................................. 126
6.3.3. Southerly, Ahead of Low, No Fronts (SaLnF) ............................................ 128
6.3.4. Northerly, No Law, No Fronts (NnLnF) ..................................................... 129
6.3.5. Southerly, No Low, No Fronts (SnLnF) ...................................................... 130
6.3.6. Bermuda High (BI-I) ................................................................................... 132
6. 3. 7. North-South Oriented Front (nsF) ............................................................. 134
6.3.8. Northerly, East- West Oriented Front (NewF) ............................................ 136
6.3.9. Southerly, East- West Oriented Front (SewF) ............................................. 137
6.3.10. Zonal Flow (Z) ......................................................................................... 138
6.3.1]. Weak Flow (W) ......................................................................................... 140
6.4. Speculations Regarding Lifting Mechanisms ........................................................... 141

vii

6.5. Discussion ................................................................................................................. 146
6. 5. 1. Advantages and Disadvantages of Subjective and Objective Clustering. .146

6.5.2. Consideration of a Hybrid Clustering Approach ....................................... 149
6.5.3. Comparison with Previous Studies ............................................................ 150
6.6. Summary ................................................................................................................... 153
CHAPTER 7
SUMMARY AND CONCLUSION ............................................................................... 155
7.1. Summary ................................................................................................................... 155
7.2. Future Climate Change Impacts ................................................................................ 164
7.3. Recommendations for Future Research .................................................................... 166
7.4. Final Remarks ........................................................................................................... 169
APPENDICES ................................................................................................................ 171
APPENDIX A
ADDITIONAL RESULTS FROM THE TEMPORAL ANALYSIS .......................... 171
A. 1. Temporal Results of Temperature ............................................................................ 171
AL]. February .................................................................................................... 171
A. 1.2. March ........................................................................................................ 172
A.I.3. May ............................................................................................................ 173
A.I.4. June ........................................................................................................... 174
A.I.5. August ........................................................................................................ 175
A.I.6. September .................................................................................................. 176
A]. 7. November ................................................................................................... 177
A.I.8. December ................................................................................................... 178
A2. Temporal Results of Precipitation and Snowfall ...................................................... 179
A21. February .................................................................................................... 179
A.2.2. March ........................................................................................................ 181
A2. 3. May ............................................................................................................ 183
AZ. 4. June ........................................................................................................... 185
A25. August ........................................................................................................ 186
A2. 6. September .................................................................................................. 187
AZ. 7. November ................................................................................................... 188
A2. 8. December ................................................................................................... 190
APPENDIX B
ADDITIONAL RESULTS FROM THE SPATIAL ANALYSIS ................................ 192
B. 1. Spatial Results of Temperature ................................................................................. 192
B. 1.1. April ........................................................................................................... 192
B. 1.2. October ...................................................................................................... 195
B2. Spatial Results of Precipitation ................................................................................ 198
B. 2. 1. April ........................................................................................................... 198
B. 2. 2. October ...................................................................................................... 201
83. Spatial Results of Snowfall ...................................................................................... 204
B. 3. 1. April ........................................................................................................... 204

viii

B. 3. 2. October ...................................................................................................... 207

84 Spatial Results for All Variables ............................................................................... 210
B4]. April ........................................................................................................... 210
B. 4. 2. October ...................................................................................................... 213
APPENDIX C
GRIB FILES ................................................................................................................... 216
C. 1. Automation of GRIB downloads .............................................................................. 216
C2. Uncompressing GRIB files using WGRIB .............................................................. 216
APPENDIX D
ADDITIONAL RESULTS FROM THE SYNOPTIC CLIMATOLOGY ................. 218
D. 1. Maps of 700 and 300 hPa Variables from the Objective Clustering Approach ....... 218
D2. Maps of 700 and 300 hPa Variables from the Subjective Clustering Approach ...... 224
APPENDIX E
SYNOPTIC PATTERNS FROM THE HYBRID APPROACH ................................ 230
El. Maps of 850 hPa Variables from the Hybrid Clustering Approach .......................... 230
E2. Maps of 700 and 300 hPa Variables from the Hybrid Clustering Approach ............ 238
LITERATURE CITED .................................................................................................. 243

ix

LIST OF TABLES

Table 3.1. Possible flag values included with monthly COOP observations ..................... 41
Table 3.2. Possible flag 1 values included with daily COOP observations ....................... 47
Table 3.3. Possible flag 2 values included with daily COOP observations ....................... 48

Table 3.4. Possible number of precipitation events. Event criteria include the percentage
of available stations reporting (values along the topmost row) and amount of precipitation
received in a day (values along the leftmost column). The criteria of 0.5 in (12.70 mm)
or more of precipitation received by 25 percent or more of the available stations was

selected for this study resulting in 1437 precipitation events (shown in bold) .................. 49
Table 4.1. Temporal change in minimum temperature (°C) 1931-2006 ........................... 56
Table 4.2. Temporal change in maximum temperature (°C) 1931-2006 .......................... 56
Table 4.3. Temporal change in mean temperature (°C) 1931-2006 .................................. 56
Table 4.4. Temporal change in total precipitation (mm) 1931-2006 ................................. 56
Table 4.5. Temporal change in total snowfall (mm) 1931-2006 ....................................... 56

Table 5.1. Number of stations with at least 90 percent of data available for minimum,
maximum, mean temperature, total precipitation, and total snowfall ................................ 83

LIST OF FIGURES

Figure 1.1. Southern Appalachian study area. Shaded elevation values are included as
meters above sea level ......................................................................................................... 1

Figure 1.2. Study region with gray areas depicting federally—owned lands. Locations of
Great Smoky Mountains National Park and nearby national forests are labeled ................ 2

Figure 3.1. Study area bounded by 34.50°N to 36.55°N and 81.50°W to 84.75°W.
Geographic provinces are labeled with approximate boundaries included. Elevation is
shaded in gray based on meters above sea level ................................................................ 35

Figure 3.2. Study area with elevation shaded in gray. Plus signs represent locations of
the 463 monthly COOP stations ......................................................................................... 36

Figure 3.3. Study area with elevation shaded in gray. Plus signs represent locations of
the 151 daily COOP stations .............................................................................................. 38

Figure 3.4. Study area with elevation shaded in gray and station locations plotted by
elevation category. Black plus signs represent low-elevation stations located below 500
m. Dark gray triangles represent mid-elevation stations located between 500 and 999 m.
Light gray diamonds represent high-elevation stations located at 1000 m or above ......... 43

Figure 3.5. Subset of NARR grid points selected for objective cluster analysis .............. 52

Figure 4.1. January minimum (bottom line), maximum (top line), and mean (center line)
temperature in °C. A linear fit was calculated for each variable over 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 - 999 m; top
right), high elevations (2 1000 m; bottom left), and all elevations (bottom right) ............ 58

Figure 4.2. April minimum (bottom line), maximum (top line), and mean (center line)
temperature in °C. A linear fit was calculated for each variable over 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 -— 999 m; top
right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right) ............ 60

Figure 4.3. July minimum (bottom line), maximum (top line), and mean (center line)
temperature in °C. A linear fit was calculated for each variable over 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right) ............ 61

Figure 4.4. October minimum (bottom line), maximum (top line), and mean (center line)
temperature in °C. A linear fit was calculated for each variable over 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (2 1000 m; bottom left), and all elevations (bottom right) ............ 63

xi

Figure 4.5. Annual values of minimum (bottom line), maximum (top line), and mean
(center line) temperature in °C. A linear fit was calculated for each variable over 193 l-
2006. Results are included for low elevations (< 500 m; top left), mid elevations (500 -
999 m; top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom
right) ................................................................................................................................... 64

Figure 4.6. Intra-annual trends of minimum (bottom line), maximum (top line), and
mean (center line) temperature in °C versus months of the year. Values were averaged
over the entire 1931-2006 period. Results are included for low elevations (< 500 m; top
left), mid elevations (500 — 999 m; top right), high elevations (2 1000 m; bottom left),
and all elevations (bottom right) ........................................................................................ 65

Figure 4.7. January precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right)......66

Figure 4.8. January snowfall in mm. A linear fit was calculated for 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right) ............ 67

Figure 4.9. April precipitation in mm. A linear fit was calculated for 1931-2006. Results
are included for low elevations (< 500 m; top lefi), mid elevations (500 —- 999 111; top
right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right) ............ 68

Figure 4.10. April snowfall in mm. A linear fit was calculated for 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 - 999 m; top
right), high elevations (2 1000 m; bottom left), and all elevations (bottom right) ............ 69

Figure 4.11. July precipitation in mm. A linear fit was calculated for 1931—2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right) ............ 70

Figure 4.12. October precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right) ...... 71

Figure 4.13. October snowfall in mm. A linear fit was calculated for 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right) ............ 72

Figure 4.14. Annual values for precipitation in mm. A linear fit was calculated for 1931-
2006. Results are included for low elevations (< 500 m; top left), mid elevations (500 —
999 m; top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom
right) ................................................................................................................................... 73

xii

Figure 4.15. Annual values for snowfall in mm. A linear fit was calculated for 1931-
2006. Results are included for low elevations (< 500 m; top left), mid elevations (500 —
999 m; top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom
right) ................................................................................................................................... 74

Figure 4.16. Intra-annual trends of total precipitation in mm versus months of the year.
Values were averaged over the entire 1931-2006 period. Results are included for low
elevations (< 500 m; top left), mid elevations (500 — 999 m; top right), high elevations (Z
1000 m; bottom left), and all elevations (bottom right) ..................................................... 76

Figure 4.17. Intra-annual trends of total snowfall in mm versus months of the year.
Values were averaged over the entire 1931-2006 period. Results are included for low
elevations (< 500 m; top left), mid elevations (500 — 999 m; top right), high elevations (Z
1000 m; bottom left), and all elevations (bottom right) ..................................................... 77

Figure 5.1. January during 1935-1944 with 5 clusters for mean temperature, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Elevation in meters is shaded in gray ............... 85

Figure 5.2. January during 1965-1974 with 5 clusters for mean temperature, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Elevation in meters is shaded in gray ............... 86

Figure 5.3. January during 1995-2004 with 5 clusters for mean temperature, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Elevation in meters is shaded in gray ............... 87

Figure 5.4. July during 1935-1944~with 5 clusters for mean temperature , where Cluster 1
(dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Elevation in meters is shaded in gray .................. 88

Figure 5.5. July during 1965-1974 with 5 clusters for mean temperature, where Cluster 1
(dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Elevation in meters is shaded in gray .................. 89

Figure 5.6. July during 1995-2004 with 5 clusters for mean temperature, where Cluster 1
(dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Elevation in meters is shaded in gray .................. 90

Figure 5.7. January during 1935-1944 with 5 clusters for total precipitation, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Elevation in meters is shaded in gray ............... 91

Figure 5.8. January during 1965-1974 with 5 clusters for total precipitation, where

Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Elevation in meters is shaded in gray ............... 92

xiii

Figure 5.9. January during 1995-2004 with 5 clusters for total precipitation, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Elevation in meters is shaded in gray ............... 93

Figure 5.10. July during 1935-1944 with 5 clusters for total precipitation, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Elevation in meters is shaded in gray .................. 94

Figure 5.11. July during 1965-1974 with 5 clusters for total precipitation, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Elevation in meters is shaded in gray .................. 95

Figure 5.12. July during 1995-2004 with 5 clusters for total precipitation, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Elevation in meters is shaded in gray .................. 96

Figure 5.13. January during 1935-1944 with 5 clusters for total snowfall, where Cluster 1
(dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Elevation in meters is shaded in gray .................. 97

Figure 5.14. January during 1965-1974 with 5 clusters for total snowfall, where Cluster 1
(dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Elevation in meters is shaded in gray .................. 98

Figure 5.15. January during 1995-2004 with 5 clusters for total snowfall, where Cluster 1
(dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Elevation in meters is shaded in gray .................. 99

Figure 5.16. January during 1935-1944 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation, and total
snowfall), where Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4
(triangle), and Cluster 5 (diamond) represent station locations. Elevation in meters is
shaded in gray .................................................................................................................. 100

Figure 5.17. January during 1965-1974 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation, and total
snowfall), where Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4
(triangle), and Cluster 5 (diamond) represent station locations. Elevation in meters is
shaded in gray .................................................................................................................. 102

Figure 5.18. January during 1995-2004 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation, and total
snowfall), where Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4
(triangle), and Cluster 5 (diamond) represent station locations. Elevation in meters is
shaded in gray .................................................................................................................. 103

xiv

Figure 5.19. July during 1935-1944 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation), where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Elevation in meters is shaded in gray ............. 104

Figure 5.20. July during 1965-1974 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation), where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Elevation in meters is shaded in gray ............. 106

Figure 5.21. July during 1995-2004 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation), where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Elevation in meters is shaded in gray ............. 107

Figure 5.22. Precipitation climatology for January during 1961-1990. Values are
reported in mm (adapted from Gaffin and Hotz, 2000; p. 6) ........................................... 109

Figure 5.23. Precipitation climatology for July during 1961-1990. Values are reported in
mm (adapted from Gaffin and Hotz, 2000; p. 7) .............................................................. 110

Figure 6.1. Precipitation events by year (1979-2006). Precipitation events were defined
as 25 percent or more of available stations reporting 0.5 in (12.7 mm) or more of
precipitation in a day ........................................................................................................ 113

Figure 6.2. Cluster 2. Temperature contoured every 5°C (top left), dewpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level ................................................................................... 114

Figure 6.3. Cluster 3. Temperature contoured every 5°C (top left), dewpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level ................................................................................... 115

Figure 6.4. Cluster 4. Temperature contoured every 5°C (top left), dewpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level ................................................................................... 116

Figure 6.5. Cluster 5. Temperature contoured every 5°C (top left), dewpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level ................................................................................... 117

Figure 6.6. Cluster 7. Temperature contoured every 5°C (top left), dewpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level ................................................................................... 118

XV

Figure 6.7. Cluster 8. Temperature contoured every 5°C (top left), deWpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level ................................................................................... 119

Figure 6.8. Cluster 9. Temperature contoured every 5°C (top left), dewpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level .................................................................................. 120

Figure 6.9. Cluster 1. Temperature contoured every 5°C (top left), dewpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level .................................................................................. 121

Figure 6.10. Cluster 6. Temperature contoured every 5°C (top left), dewpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level .................................................................................. 122

Figure 6.11. Cluster 10. Temperature contoured every 5°C (top left), dewpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level .................................................................................. 123

Figure 6.12. Cluster 11. Temperature contoured every 5°C (top left), deWpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level .................................................................................. 124

Figure 6.13. Closed low (CL). Temperature contoured every 5°C (top left), dewpoint
contoured every 5°C (top right), height contoured every 30 m (bottom left), and wind
vectors (bottom right) at the 850 hPa level ...................................................................... 126

Figure 6.14. Northerly, behind low, no fronts (NbLnF). Temperature contoured every
5°C (top left), dewpoint contoured every 5°C (top right), height contoured every 30 m
(bottom left), and wind vectors (bottom right) at the 850 hPa level ................................ 127

Figure 6.15. Southerly, ahead of low, no fronts (SaLnF). Temperature contoured every
5°C (top left), dewpoint contoured every 5°C (top right), height contoured every 30 m
(bottom left), and wind vectors (bottom right) at the 850 hPa level ................................ 129

Figure 6.16. Northerly, no low, no fronts (NnLnF). Temperature contoured every 5°C
(top left), dewpoint contoured every 5°C (top right), height contoured every 30 m (bottom
left), and wind vectors (bottom right) at the 850 hPa level ............................................. 130

Figure 6.17. Southerly, no low, no fronts (SnLnF). Temperature contoured every 5°C

(top left), dewpoint contoured every 5°C (top right), height contoured every 30 m (bottom
left), and wind vectors (bottom right) at the 850 hPa level ............................................. 132

xvi

Figure 6.18. Bermuda High (BH). Temperature contoured every 5°C (top left), dewpoint
contoured every 5°C (top right), height contoured every 30 m (bottom left), and wind
vectors (bottom right) at the 850 hPa level ...................................................................... 134

Figure 6.19. North-south oriented front (nsF). Temperature contoured every 5°C (top
left), dewpoint contoured every 5°C (top right), height contoured every 30 m (bottom
left), and wind vectors (bottom right) at the 850 hPa level ............................................. 135

Figure 6.20. Northerly, east-west oriented front (NewF). Temperature contoured every
5°C (top left), dewpoint contoured every 5°C (top right), height contoured every 30 m
(bottom left), and wind vectors (bottom right) at the 850 hPa level ................................ 137

Figure 6.21. Southerly, east-west oriented front (SewF). Temperature contoured every
5°C (top left), dewpoint contoured every 5°C (top right), height contoured every 30 m
(bottom left), and wind vectors (bottom right) at the 850 hPa level ................................ 138

Figure 6.22. Zonal flow (Z). Temperature contoured every 5°C (top left), dewpoint
contoured every 5°C (top right), height contoured every 30 m (bottom left), and wind
vectors (bottom right) at the 850 hPa level ...................................................................... 139

Figure 6.23. Weak flow (W). Temperature contoured every 5°C (top left), dewpoint
contoured every 5°C (top right), height contoured every 30 m (bottom left), and wind
vectors (bottom right) at the 850 hPa level ...................................................................... 141

Figure 6.24. Vertical velocity (in Pa/s) at 700 hPa for the BH pattern. Negative values
represent areas of upward motion and positive values represent areas of downward
motion .............................................................................................................................. 143

Figure 6.25. Vertical velocity (in Pa/s) at 700 hPa for the CL pattern. Negative values
represent areas of upward motion and positive values represent areas of downward
motion .............................................................................................................................. 144

Figure 6.26. Vertical velocity (in Pa/s) at 700 hPa for the SaLnF pattern. Negative
values represent areas of upward motion and positive values represent areas of downward
motion .............................................................................................................................. 144

Figure 6.27. Vertical velocity (in Pa/s) at 700 hPa for the SewF pattern. Negative values
represent areas of upward motion and positive values represent areas of downward
motion .............................................................................................................................. 145

Figure 6.28. Vertical velocity (in Pa/s) at 700 hPa for the SnLnF pattern. Negative

values represent areas of upward motion and positive values represent areas of downward
motion .............................................................................................................................. 145

xvii

Figure 6.29. Vertical velocity (in Pa/s) at 700 hPa for the W pattern. Negative values
represent areas of upward motion and positive values represent areas of downward
motion .............................................................................................................................. 146

Figure A. 1. February minimum (bottom line), maximum (top line), and mean (center
line) temperature in °C. A linear fit was calculated for each variable over 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 in;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right)....171

Figure A.2. March minimum (bottom line), maximum (top line), and mean (center line)
temperature in °C. A linear fit was calculated for each variable over 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right) .......... 172

Figure A.3. May minimum (bottom line), maximum (top line), and mean (center line)
temperature in °C. A linear fit was calculated for each variable over 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (2 1000 m; bottom left), and all elevations (bottom right) .......... 173

Figure A.4. June minimum (bottom line), maximum (top line), and mean (center line)
temperature in °C. A linear fit was calculated for each variable over 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (2 1000 m; bottom left), and all elevations (bottom right) .......... 174

Figure A.5. August minimum (bottom line), maximum (top line), and mean (center line)
temperature in °C. A linear fit was calculated for each variable over 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right) .......... 175

Figure A.6. September minimum (bottom line), maximum (top line), and mean (center
line) temperature in °C. A linear fit was calculated for each variable over 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 - 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right)....176

Figure A.7. November minimum (bottom line), maximum (top line), and mean (center
line) temperature in °C. A linear fit was calculated for each variable over 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (2 1000 m; bottom left), and all elevations (bottom right)....177

Figure A.8. December minimum (bottom line), maximum (top line), and mean (center
line) temperature in °C. A linear fit was calculated for each variable over 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (2 1000 m; bottom left), and all elevations (bottom right)....178

xviii

Figure A.9. February precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 - 999 m;
top right), high elevations (_>_ 1000 m; bottom left), and all elevations (bottom right)....179

Figure A.10. February snowfall in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right)....180

Figure A.11. March precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 - 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right)....18l

Figure A.12. March snowfall in mm. A linear fit was calculated for 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (2 1000 m; bottom left), and all elevations (bottom right) .......... 182

Figure A.13. May precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 - 999 m;
top right), high elevations (2 1000 m; bottom left), and all elevations (bottom right)....183

Figure A.14. May snowfall in mm. A linear fit was calculated for 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (_>_ 1000 m; bottom left), and all elevations (bottom right) .......... 184

Figure A. 1 5. June precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 —— 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right)....185

Figure A.16. August precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right)....186

Figure A.17. September precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (2 1000 m; bottom left), and all elevations (bottom right)....187

Figure A. 1 8. November precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right)....188

Figure A.19. November snowfall in mm. A linear fit was calculated for 1931-2006.

Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right)....189

xix

Figure A.20. December precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (2 1000 m; bottom left), and all elevations (bottom right)....190

Figure A.21. December snowfall in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (2 1000 m; bottom left), and all elevations (bottom right)....19l

Figure B. 1. April during 1935-1944 with 5 clusters for mean temperature, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Mean temperature values for Clusters 1-5 were
4.24, 9.65, 15.48, 12.47, and 12.16°C. Elevation in meters is shaded in gray ................ 192

Figure 8.2. April during 1965-1974 with 5 clusters for mean temperature, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Mean temperature values for Clusters 1-5 were
16.22, 15.06, 14.11, 12.89, and 10.16°C. Elevation in meters is shaded in gray ............ 193

Figure B.3. April during 1995-2004 with 5 clusters for mean temperature, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Mean temperature values for Clusters 1-5 were
14.41, 12.88, 15.68, 10.81, and 619°C. Elevation in meters is shaded in gray .............. 194

Figure B.4. October during 1935-1944 with 5 clusters for mean temperature, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Mean temperature values for Clusters 1-5 were
7.79, 17.06, 13.99, 15.73, and 11.64°C. Elevation in meters is shaded in gray .............. 195

Figure 8.5. October during 1965-1974 with 5 clusters for mean temperature, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Mean temperature values for Clusters 1-5 were
16.29, 13.80, 14.93, 10.29, and 12.62°C. Elevation in meters is shaded in gray ............ 196

Figure B.6. October during 1995-2004 with 5 clusters for mean temperature, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Mean temperature values for Clusters 1-5 were
13.66, 11.59, 15.02, 16.27, and 833°C. Elevation in meters is shaded in gray .............. 197

Figure B.7. April during 1935-1944 with 5 clusters for total precipitation, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Precipitation values for Clusters 1-5 were 105.42,
147.59, 120.09, 89.45, and 74.75 mm. Elevation in meters is shaded in gray ............... 198

XX

Figure B.8. April during 1965-1974 with 5 clusters for total precipitation, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Precipitation values for Clusters 1-5 were 85.44,
111.18, 98.80, 138.72, and 120.93 mm. Elevation in meters is shaded in gray .............. 199

Figure 8.9. April during 1995-2004 with 5 clusters for total precipitation, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Precipitation values for Clusters 1-5 were 170.35,
103.61, 87.67, 134.18, and 120.04 mm. Elevation in meters is shaded in gray ............. 200

Figure B.10. October during 1935-1944 with 5 clusters for total precipitation, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Precipitation values for Clusters 1-5 were 77.80,
87.15, 54.87, 63.34, and 108.94 mm. Elevation in meters is shaded in gray ................. 201

Figure 3.1]. October during 1965-1974 with 5 clusters for total precipitation, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Precipitation values for Clusters 1-5 were
105.25, 158.56, 88.71, 125.91, and 72.51 mm. Elevation in meters is shaded in gray.. 202

Figure B.12. October during 1995-2004 with 5 clusters for total precipitation, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Precipitation values for Clusters 1-5 were 49.77,
77.14, 64.73, 97.19, and 120.84 mm. Elevation in meters is shaded in gray ................. 203

Figure B.13. April during 1935-1944 with 5 clusters for total snowfall, where Cluster 1
(dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Snowfall values for Clusters 1-5 were 77.22,
26.42, 61.72, 7.50, and 0.12 mm. Elevation in meters is shaded in gray ....................... 204

Figure B.14. April during 1965-1974 with 5 clusters for total snowfall, where Cluster 1
(dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Snowfall values for Clusters 1-5 were 5.93, 16.26,
10.24, 33.66, and 0.19 mm. Elevation in meters is shaded in gray ................................ 205

Figure B.15. April during 1995-2004 with 5 clusters for total snowfall, where Cluster 1
(dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Snowfall values for Clusters 1-5 were 127.5],
12.71, 0.11, 78.99, and 212.26 mm. Elevation in meters is shaded in gray .................... 206

Figure B.16. October during 1935-1944 with 5 clusters for total snowfall, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Snowfall values for Clusters 1-5 were 17.02, 5.08,
30.48, 7.62, and 0.00 mm. Elevation in meters is shaded in gray .................................. 207

xxi

Figure B.17. October during 1965-1974 with 5 clusters for total snowfall, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Snowfall values for Clusters 1-5 were 0.00, 0.76,
6.86, 0.00, and 0.00 mm. Elevation in meters is shaded in gray .................................... 208

Figure B. 18. October during 1995-2004 with 5 clusters for total snowfall, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Snowfall values for Clusters 1-5 were 1.27, 0.00,
0.00, 0.00, and 0.00 mm. Elevation in meters is shaded in gray .................................... 209

Figure B.19. April during 1935-1944 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation, and total
snowfall), where Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4
(triangle), and Cluster 5 (diamond) represent station locations. Minimum temperature
values for Clusters 1-5 were 4.72, 8.06, 6.98, 5.48, and 543°C. Maximum temperature
values for Clusters 1-5 were 19.46, 22.82, 18.47, 20.93, and 19.96°C. Mean temperature
values for Clusters 1-5 were 12.09, 15.44, 12.73, 13.21, and 12.71°C. Precipitation
values for Clusters 1-5 were 107.66, 116.73, 115.90, 82.59, and 133.06 mm. Snowfall
values for Clusters 1-5 were 8.48, 0.08, 15.49, 0.29, and 10.29 mm. Elevation in meters
is shaded in gray ............................................................................................................... 210

Figure B.20. April during 1965-1974 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation, and total
snowfall), where Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4
(triangle), and Cluster 5 (diamond) represent station locations. Minimum temperature
values for Clusters 1-5 were 8.41, 5.46, 6.98, 4.72, and 660°C. Maximum temperature
values for Clusters 1-5 were 23.17, 20.10, 21.86, 21.34, and 20.51°C. Mean temperature
values for Clusters 1-5 were 15.81, 12.79, 14.43, 13.04, and 13.57°C. Precipitation
values for Clusters 1-5 were 94.35, 87.88, 131.72, 111.89, and 114.70 mm. Snowfall
values for Clusters 1-5 were 0.09, 1.11, 0.66, 1.09, and 6.63 mm. Elevation in meters is
shaded in gray .................................................................................................................. 211

Figure B.21. April during 1995-2004 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation, and total
snowfall), where Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4
(triangle), and Cluster 5 (diamond) represent station locations. Minimum temperature
values for Clusters 1-5 were 7.07, 6.77, 6.29, 4.53, and 603°C. Maximum temperature
values for Clusters 1-5 were 20.97, 21.69, 21.53, 20.28, and 18.36°C. Mean temperature
values for Clusters 1-5 were 14.03, 14.24, 13.93, 12.42, and 12.21°C. Precipitation
values for Clusters 1-5 were 103.58, 99.20, 121.35, 111.15, and 121.20 mm. Snowfall
values for Clusters 1-5 were 2.62, 0.97, 0.00, 3.06, and 41.38 mm. Elevation in meters is
shaded in gray .................................................................................................................. 212

xxii

Figure 322. October during 1935-1944 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation, and total
snowfall), where Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4
(triangle), and Cluster 5 (diamond) represent station locations. Minimum temperature
values for Clusters 1-5 were 7.83, 9.07, 8.26, 6.79, and 662°C. Maximum temperature
values for Clusters 1-5 were 23.82, 21.30, 23.75, 22.08, and 19.65°C. Mean temperature
values for Clusters 1-5 were 15.83, 15.19, 16.01, 14.50, and 13.13°C. Precipitation
values for Clusters 1-5 were 58.77, 80.51, 80.82, 68.08, and 91.69 mm. Snowfall values
for Clusters 1-5 were 0.00, 3.40, 0.00, 5.08, and 2.12 mm. Elevation in meters is shaded
in gray .............................................................................................................................. 213

Figure B.23. October during 1965-1974 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation, and total
snowfall), where Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4
(triangle), and Cluster 5 (diamond) represent station locations. Minimum temperature
values for Clusters 1-5 were 7.09, 9.16, 5.07, 8.27, and 580°C. Maximum temperature
values for Clusters 1-5 were 22.99, 22.64, 21.00, 21.59, and 19.12°C. Mean temperature
values for Clusters 1-5 were 15.05, 15.92, 13.05, 14.94, and 12.45°C. Precipitation
values for Clusters 1-5 were 89.19, 116.47, 84.67, 81.53, and 137.00 mm. Snowfall
values for Clusters 1-5 were 0.00, 0.00, 0.00, 0.00, and 0.54 mm. Elevation in meters is
shaded in gray .................................................................................................................. 214

Figure B.24. October during 1995-2004 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation, and total
snowfall), where Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4
(triangle), and Cluster 5 (diamond) represent station locations. Minimum temperature
values for Clusters 1-5 were 8.17, 6.84, 6.94, 8.59, and 597°C. Maximum temperature
values for Clusters 1-5 were 22.63, 21.87, 21.08, 20.63, and 19.52°C. Mean temperature
values for Clusters 1-5 were 15.42, 14.36, 14.03, 14.63, and 12.76°C. Precipitation
values for Clusters 1-5 were 76.96, 61.28, 59.00, 91.53, and 100.66 mm. Snowfall values
for Clusters 1-5 were 0.00, 0.00, 0.00, 0.09, 0.00 mm. Elevation in meters is shaded in
gray .................................................................................................................................. 215

Figure D. 1. Cluster 1. Maps in the top row from left to right are temperature (contoured
every 5°C) wind vectors, deWpoint (contoured every 5°C), and height (contoured every 30
m) at the 700 hPa level. Maps in the bottom row from left to right are height (contoured

every 120 m), wind vectors, and isotachs (contoured every 10 m/s) at the 300 hPa level.
.......................................................................................................................................... 218

Figure D.2. Cluster 2. Maps in the top row from left to right are temperature (contoured
every 5°C) wind vectors, dewpoint (contoured every 5°C), and height (contoured every 30
m) at the 700 hPa level. Maps in the bottom row from left to right are height (contoured
every 120 m), wind vectors, and isotachs (contoured every 10 m/s) at the 300 hPa level.
.......................................................................................................................................... 219

xxiii

Figure D.3. Cluster 3. Maps in the top row from left to right are temperature (contoured
every 5°C) wind vectors, deWpoint (contoured every 5°C), and height (contoured every 30
m) at the 700 hPa level. Maps in the bottom row from left to right are height (contoured
every 120 m), wind vectors, and isotachs (contoured every 10 m/s) at the 300 hPa level.
.......................................................................................................................................... 219

Figure D.4. Cluster 4. Maps in the top row from left to right are temperature (contoured
every 5°C) wind vectors, dewpoint (contoured every 5°C), and height (contoured every 30
m) at the 700 hPa level. Maps in the bottom row from left to right are height (contoured
every 120 m), wind vectors, and isotachs (contoured every 10 m/s) at the 300 hPa level.
.......................................................................................................................................... 220

Figure D.5. Cluster 5. Maps in the top row from left to right are temperature (contoured
every 5°C) wind vectors, dewpoint (contoured every 5°C), and height (contoured every 30
m) at the 700 hPa level. Maps in the bottom row from left to right are height (contoured
every 120 m), wind vectors, and isotachs (contoured every 10 m/s) at the 300 hPa level.
.......................................................................................................................................... 220

Figure D.6. Cluster 6. Maps in the top row from left to right are temperature (contoured
every 5°C) wind vectors, dewpoint (contoured every 5°C), and height (contoured every 30
m) at the 700 hPa level. Maps in the bottom row from left to right are height (contoured
every 120 m), wind vectors, and isotachs (contoured every 10 m/s) at the 300 hPa level.
.......................................................................................................................................... 221

Figure D.7. Cluster 7. Maps in the top row from left to right are temperature (contoured
every 5°C) wind vectors, dewpoint (contoured every 5°C), and height (contoured every 30
m) at the 700 hPa level. Maps in the bottom row from left to right are height (contoured
every 120 m), wind vectors, and isotachs (contoured every 10 m/s) at the 300 hPa level.
.......................................................................................................................................... 221

Figure D.8. Cluster 8. Maps in the top row from left to right are temperature (contoured
every 5°C) wind vectors, dewpoint (contoured every 5°C), and height (contoured every 30
m) at the 700 hPa level. Maps in the bottom row from left to right are height (contoured
every 120 m), wind vectors, and isotachs (contoured every 10 m/s) at the 300 hPa level.
.......................................................................................................................................... 222

Figure D.9. Cluster 9. Maps in the top row from left to right are temperature (contoured
every 5°C) wind vectors, dewpoint (contoured every 5°C), and height (contoured every 30
m) at the 700 hPa level. Maps in the bottom row from left to right are height (contoured
every 120 m), wind vectors, and isotachs (contoured every 10 m/s) at the 300 hPa level.
.......................................................................................................................................... 222

xxiv

Figure D.10. Cluster 10. Maps in the top row from left to right are temperature
(contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and height
(contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to right
are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10 m/s)
at the 300 hPa level 1 ........................................................................................................ 223

Figure D.11. Cluster 11. Maps in the top row from left to right are temperature
(contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and height
(contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to right
are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10 m/s)
at the 300 hPa level .......................................................................................................... 223

Figure D.12. Closed low (CL). Maps in the top row from left to right are temperature
(contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and height
(contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to right
are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10 m/s)
at the 300 hPa level .......................................................................................................... 224

Figure D.13. Northerly, behind low, no fi'onts (NbLnF). Maps in the top row from left to
right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row
from left to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level ................................................................. 225

Figure D. 14. Southerly, ahead of low, no fronts (SaLnF). Maps in the top row fi'om left
to right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row
from left to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level ................................................................. 225

Figure D.15. Northerly, no low, no fronts (NnLnF). Maps in the top row from left to
right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row
from left to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level ................................................................. 226

Figure D.16. Southerly, no low, no fronts (SnLnF). Maps in the t0p row from left to
right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row
from left to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level ................................................................. 226

XXV

Figure D.17. Bermuda High (BH). Maps in the top row from left to right are
temperature (contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and
height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to
right are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10
m/s) at the 300 hPa level .................................................................................................. 227

Figure D.18. North-south oriented front (nsF). Maps in the top row from left to right are
temperature (contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and
height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to
right are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10
m/s) at the 300 hPa level .................................................................................................. 227

Figure D.19. Northerly, east-west oriented front (NewF). Maps in the top row from left
to right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row
from left to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level ................................................................. 228

Figure D.20. Southerly, east-west oriented front (SewF). Maps in the top row from left
to right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row
from left to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level ................................................................. 228

Figure D.21. Zonal flow (Z). Maps in the top row from left to right are temperature
(contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and height
(contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to right
are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10 m/s)
at the 300 hPa level .......................................................................................................... 229

Figure D.22. Weak flow (W). Maps in the top row from left to right are temperature
(contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and height
(contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to right
are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10 m/s)
at the 300 hPa level .......................................................................................................... 229

Figure B]. Closed low (CL). Temperature contoured every 5°C (top left), dewpoint
contoured every 5°C (top right), height contoured every 30 m (bottom left), and wind
vectors (bottom right) at the 850 hPa level ...................................................................... 230

Figure E.2. Northerly, behind low, no fronts (NbLnF). Temperature contoured every 5°C

(top left), dewpoint contoured every 5°C (top right), height contoured every 30 m (bottom
left), and wind vectors (bottom right) at the 850 hPa level ............................................. 231

xxvi

Figure B.3. Southerly, ahead of low, no fronts (SaLnF). Temperature contoured every
5°C (top left), dewpoint contoured every 5°C (top right), height contoured every 30 m
(bottom left), and wind vectors (bottom right) at the 850 hPa level ................................ 232

Figure E.4. Southerly, no low, no fronts (SnLnF). Temperature contoured every 5°C (top
left), dewpoint contoured every 5°C (top right), height contoured every 30 m (bottom
left), and wind vectors (bottom right) at the 850 hPa level ............................................. 233

Figure E.5. Bermuda High (BH). Temperature contoured every 5°C (top left), dewpoint
contoured every 5°C (top right), height contoured every 30 m (bottom left), and wind
vectors (bottom right) at the 850 hPa level ...................................................................... 234

Figure E.6. North-south oriented front (nsF). Temperature contoured every 5°C (top
left), dewpoint contoured every 5°C (top right), height contoured every 30 m (bottom
left), and wind vectors (bottom right) at the 850 hPa level ............................................. 235

Figure E.7. Northerly, east-west oriented front (NewF). Temperature contoured every
5°C (top left), dewpoint contoured every 5°C (top right), height contoured every 30 m
(bottom left), and wind vectors (bottom right) at the 850 hPa level ................................ 236

Figure E.8. Southerly, east-west oriented front (SewF). Temperature contoured every
5°C (top left), dewpoint contoured every 5°C (top right), height contoured every 30 m
(bottom left), and wind vectors (bottom right) at the 850 hPa level ................................ 237

Figure E.9. Closed low (CL). Maps in the top row from left to right are temperature
(contoured every 5°C) wind vectors, deWpoint (contoured every 5°C), and height
(contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to right
are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10 m/s)
at the 300 hPa level .......................................................................................................... 238

Figure E. 1 O. Northerly, behind low, no fronts (NbLnF). Maps in the top row from left to
right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row
from left to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level ................................................................. 239

Figure E.11. Southerly, ahead of low, no fronts (SaLnF). Maps in the top row from left
to right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row
from left to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level ................................................................. 239

xxvii

Figure B.12. Southerly, no low, no fronts (SnLnF). Maps in the top row from left to right
are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every 5°C),
and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row from
left to right are height (contoured every 120 m), wind vectors, and isotachs (contoured
every 10 m/s) at the 300 hPa level ................................................................................... 240

Figure E.13. Bermuda High (BH). Maps in the top row from left to right are temperature
(contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and height
(contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to right
are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10 m/s)
at the 300 hPa level .......................................................................................................... 240

Figure E.14. North-south oriented front (nsF). Maps in the top row from left to right are
temperature (contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and
height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to
right are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10
m/s) at the 300 hPa level .................................................................................................. 241

Figure B. 15. Northerly, east-west oriented front (N ewF). Maps in the top row from left
to right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row
from left to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level ................................................................. 241

Figure B.16. Southerly, east-west oriented front (SewF). Maps in the top row from left
to right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row
from left to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level ................................................................. 242

xxvfii

CHAPTER 1

INTRODUCTION

1.1. Overview and Statement of Problem

With a relief greater than 1500 m in many areas, the southern Appalachian region
offers a variety of microclimates with weather conditions comparable to those spanning
from Georgia to Canada (Figure 1.1). Thousands of plant and animal species exist in the
region, some of which live no place else in the world. Located along the TN and NC
border, Great Smoky Mountains National Park (GSMNP; see Figure 1.2), for example,
contains the largest expanse of old—growth forest east of the Mississippi River. According
to the National Park Service (NPS), an estimated 100,000 species of living organisms

reside within GSMNP and new species are frequently discovered (NPS, 2009).

 

    

Elevation (m)

North
Carollna

   

35N ,, ‘ _7 . .2 . ............... .........

 

 

 

 

82W

Figure 1.1. Southern Appalachian study area. Shaded elevation values are included as
meters above sea level.

   
 
 

 

 
  
  

 

Daniel Boone NF
36M 7_ . - - 7.
Great Smoky, > _’ ’ 5 :5;
Cherokee NF T: V ..
ChafinhoocheeNF
N

 

 

 

84W 83W 82W

Figure 1.2. Study region with gray areas depicting federally-owned lands. Locations of
Great Smoky Mountains National Park and nearby national forests are labeled.
While the region has demonstrated resilience over time, it is vulnerable to natural
and anthropogenic disturbances. Such factors as climatic extremes and trends, wildfires,
air pollution, deforestation, and insects all affect the dynamics of the vegetation in ways
that are often poorly understood. Possible regional impacts resulting from climate
change are also receiving attention (e.g., Kupfer and Cairns, 1996; Delcourt and
Delcourt, 2001; Delcourt, 2002). Recent general circulation model (GCM) output
summarized in the Fourth Assessment Report by the lntergovemmental Panel on Climate
Change (IPCC) suggests warming trends for the southeastern US, although the models
vary on the magnitude of warming. In addition, while some models predict wetter
conditions in the southeastern US. by 2100, other models suggest drier conditions (IPCC,

2007a).

Analyses of global climate have shown a gradual temperature increase during the
twentieth century ranging between 056°C and 092°C, which may continue at an
increasing rate during the twenty-first century (IPCC, 2007a). Considering the 1901-
2008 and 1970-2008 periods, the US. Global Change Research Program (USGCRP)
found that annual temperature in the southeast U.S. warmed by 0.17 and 089°C,
respectively (USGCRP, 2009). IPCC model output suggests that the 2090-2099 average
global temperature will be between 1.1°C and 64°C warmer than the 1980-1999 average
global temperature. Although a different ending time period was used, values for the
southern Appalachian region suggest that the 2080-2099 period will be 30°C to 35°C
warmer than 1980-1999 (IPCC, 2007a). When compared to temperature, model
projections of precipitation from the Fourth Assessment Report display less agreement.
In general, precipitation is expected to increase in mid-latitude regions (e.g., the Great
Lakes) while decreasing in regions closer to the equator (e.g., Mexico). The southern
Appalachians lie in between these two extremes, and climate models disagree over
whether precipitation will increase, decrease, or remain close to current levels (IPCC,
2007a). The USGCRP did not make firm projections regarding the region's precipitation,
although the report did point out that precipitation in the region did increase over the
1901-2008 period. In recent decades, rainfall was received in heavier amounts, but
during fewer overall events. The result created the opportunity for more drought periods
during the past few decades (USGCRP, 2009)

Amidst this uncertainty about firture climates, an elaboration of temperature and
precipitation trends of the recent past, as well as an in—depth description of synoptic

patterns affecting the region's precipitation, remain lacking in the scientific literature.

These limitations constrain the ability of climate scientists to put climate projections
emerging from GCMs into a scientific context. As a result, the ability to assess the likely
consequences on the regional moisture climatology of changing synoptic patterns
projected by model output is compromised, a problem made even worse by climatic
complexity introduced by the extreme topographic variety within the region. Thus, I
propose to construct a synoptic climatological inventory of the southern Appalachian
region in order to elaborate the region’s complex moisture climatology. A baseline
temporal and spatial analysis of the region's temperature and precipitation will also be
included to provide a context in which to compare results from global and regional

climate model projections.

1.2. Ecological Change in the Region

Although debate exists over whether or not the southern Appalachians served as a
refuge for plant and animal species during the last glacial advance (Deevey, 1949; Braun,
1950; Delcourt, 2002), there is consensus that over time the area has become a place of
high biodiversity. However, the species richness of the southern Appalachians could be
adversely impacted by a period of accelerated warming. A temperature increase of 15°C
to 25°C (well within the range presented by the IPCC) is expected to put approximately
20 to 30 percent of the world’s species at risk for extinction (IPCC, 2007b). Poleward
and upward shifts in habitats are expected, leaving high-elevation species and endemic
species (those found nowhere else) of the southern Appalachians particularly at risk
(Delcourt, 2002). These high-elevation environments may be thought of as “sky islands”

with similar ecosystems existing on each isolated mountain peak. Although sky island

terminology has commonly been used to describe a species-rich area near the union of the
Arizona, New Mexico, and Mexico borders (e.g., Heald, 1967; Kupfer et al., 2005;
Skroch, 2009), the southern Appalachian region also possesses a species-rich, high-
elevation island characteristic and has also received attention (e.g., Kupfer and Cairns
1996; Allen and Kupfer, 2000; 2001; Delcourt and Delcourt, 2001; Delcourt, 2002).
Future warming trends may result in low-elevation species moving upward and
eventually replacing high-elevation species since those species currently adapted to the
highest elevations would have already reached the physical limits of their respective
ranges.

Natural and anthropogenic causes have had detrimental effects on wildlife in the
region already. Some species, such as Fraser fir (Abiesfraseri) are endemic to region.
The tree species has been impacted greatly by the exotic balsam woolly adelgid insect
(Busing et al., 1988), as well as by high ozone levels (Aneja et al., 1991), and acid rain
(Aneja et al., 1992). With the addition of increasing temperatures, Fraser fir is one
example of a species that could disappear entirely by 2100 (Delcourt, 2002). Some of the
most biodiverse and unique areas of the southern Appalachian high elevations are grassy
balds (Mark, 1958; Houk, 2009). These balds, and likely all of the wildlife they contain,
are also expected to continue disappearing as temperatures increase. In addition to the
loss of diversity, a longstanding view has been that old-growth forests are carbon-neutral
environments. However, two recent studies proposed that old-growth forests actually
continue to sequester carbon in boreal, temperate (Luyssaert et al., 2008), and tropical
(Lewis et al., 2009) climates. If true, any loss of the region's old-growth forest due to a

warming climate could contribute to a feedback cycle and lead to additional warming.

1.3. Research Goals

The rationale for this study has two main considerations. First, studies of the
region’s ecology suggest that species richness is being affected by recent climate change
and the trend will continue if global warming continues at the projected pace.
Biodiversity is what sets the southern Appalachians, and particularly GSMNP, apart from
other areas nearby, and it may be in jeopardy given this warming. The possible extinction
of high-elevation and endemic species only raises the biological stakes. Second, climate
projection results presented in the IPCC’s Fourth Assessment Report and the USGCRP's
2009 report are tentative and uncertain for this region and a review of past climatological
literature for the region suggests that additional research is needed to improve our
scientific knowledge of the region's climate.

In order to address the need for additional climate research for this region, my
specific goals for this dissertation include: 1) examining the region's temperature and
precipitation values over time to highlight any temporal trends and test any existing
trends for statistical significance, 2) spatially analyzing the region's station observations
to note whether or not multiple distinct climate regimes exist, and if so, how many, 3)
conducting a climatological inventory of synoptic patterns occurring during major
precipitation events within the region using several approaches and testing the frequency
trends of each synoptic pattern for statistical significance, and 4) comparing the results
from the various approaches used.

To improve current knowledge of the region's future climate, researchers must
first fully understand the region's climate of the recent past. As a result, I deemed a

thorough investigation of recent climate of the southern Appalachians as necessary. In

particular, climate scientists need an updated and expanded climatology of temperature,
precipitation, and snowfall, as well as a broadened understanding of the synoptic origins
of precipitation in order to obtain knowledge of conditions during the recent past. With
the results provided here, those of us in the climate community should be better equipped
to recognize future changes in the climate and synoptic patterns that influence the

species-rich southern Appalachians.

1.4. Organization of the Dissertation

In Chapter 2, I provide a review of relevant literature covering previous baseline
and synoptic Climatologies, the history and methods of cluster analysis used in
climatological research, and climate studies that used objective or subjective clustering
techniques. Also within Chapter 2, I provide a list of specific research questions and
objectives that I will address through this study. Chapter 3 includes a description of the
study area and the datasets used. In that chapter I also describe my methods for carrying
out the temporal and spatial baseline climatology and the synoptic climatology. In
Chapter 4, the results and discussion of the temporal analysis are presented. The results
and discussion of the spatial analysis are presented in Chapter 5. Results and discussion
of the synoptic climatology are presented in Chapter 6. Lastly, in Chapter 7, I revisit the
research questions, provide recommendations for future research, and share final

remarks. Appendices A-E follow the dissertation and include supplemental material.

CHAPTER 2

LITERATURE REVIEW

2.1. General Climatologies of the Region

According to Koppen’s well-known climate classification system, the southern
Appalachian region is included within the Cfa, or “humid subtropical,” category.
Typically, Cfa regions are located between 25° and 40° latitude, along eastern sections of
continents (Ahrens, 2003). These regions are characterized by hot, humid summers (i.e.,
dewpoints often exceed 23°C) and mild winters (i.e., average temperature of the coldest
month below 18°C, and above -3 °C). Precipitation is frequent and widespread. Annual
totals usually range between 800 and 1650 mm (Ahrens, 2003). However, the complex
topography of the southern Appalachians likely encompasses more climatic zones than
just Cfa. Topographic shape, slope, aspect, and elevation create a variety of
microclimates (Hack and Goodlett, 1960), and thus, the region experiences temperature
and precipitation averages that can differ from those discussed above.

Using recorded precipitation data spanning 1936-1949, Whittaker (1952) showed
that the southern Appalachian region had some of the highest annual precipitation totals
in the eastern US. Precipitation in valley regions (e.g., Gatlinburg, TN, Elkmont, TN,
and Maryville, TN) ranged from 50-60 inches (1270-1524 mm) annually and increased
with elevation up to 80 inches (2032 mm) or more. Whittaker’s dataset also revealed that
precipitation maxima typically occurred during late winter/early spring and late summer,
while minima occurred in late spring and early autumn.

Smallshaw (1953) presented three precipitation studies conducted by the

Tennessee Valley Authority (TVA) and the US. Weather Bureau. Annual precipitation
values across the southern Appalachian region ranged from 39 inches (991 mm) in the
valleys to 93 inches (2362 mm) on the ridges. The first study included stations scattered
throughout the Newfound Gap region of GSMNP during the years 1946-1950.
Precipitation totals increased abruptly with elevation up to about 3000-3 500 ft (914-1067
m) and then continued to increase with elevation, but at a slower rate. However, the
second and third studies demonstrated that not all areas in the Appalachians followed the
pattern of increasing precipitation with increasing elevation. Snake Mountain, NC, and
Clinch Mountain, VA, both received less precipitation near the peaks than in the
surrounding valleys. In both cases, the smaller amount of precipitation was attributed to
“a combination of updraft and carry-over of moisture-laden air because of the steep upper
slope and the narrow crest of the ridge[s]” (Smallshaw, 1953; p. 587). In addition, a
funneling effect could have occurred as wind passed the mountain peaks. The funneling
effect may have accelerated wind speeds and impacted the amount of precipitation the
gages collected. As a result, the TVA moved precipitation gages from peaks in order to
avoid any future funneling effect.

Shanks (1954) used data from four stations within GSMNP during the years 1946-
1950 to describe how temperature and precipitation varied with increasing elevation
within the park. Stations included were: Park Headquarters (445 m), Alum Cave parking
lot (1158 m), Newfound Gap (1524 m), and Clingman’s Dome (1920 m). Over the five-
year period, annual precipitation at the lowest station averaged 57.8 inches (1468.1 mm).
Precipitation increased with increasing elevation, with the highest station averaging 90.9

inches (2308.9 mm). The five—year mean annual temperature profile for the stations

showed that the average rate of decrease with altitude was 2.23°F/1000 ft (4.06°C/1000
m), which was similar to the results found by others (see Table 2 provided in Hicks,
1979). These lapse rate values tended to be less than the “normal lapse rate” usually
given as 6.5°C/1000 m (Ahrens, 2003). However, the humid conditions and low
continentality of the southeastern US. probably accounted for the region’s smaller lapse
rate (Hicks, 1979). Shanks found July to be the warmest month at all elevations, while
the coldest month varied between November and December for low elevations and March
for the highest station. These classic works by Whittaker (1952), Smallshaw (1953), and
Shanks (1954) continue to be cited in reference to southern Appalachian climate. For
similar classic overviews of the region’s vegetation, the reader is referred to Braun (1950)
and Whittaker (1956).

Mark (1958) created a climatology spanning a 26-year period (1931-1956) and
included three stations: Asheville, NC (671 m), Banner Elk, NC (1131 m), and Mount
Mitchell, NC (2022 m). Daily mean temperature fell to a minimum in January for
Asheville (46°C) and Banner Elk (1.5°C), while the minimum for Mount Mitchell
(-2.6°C) was in February. In agreement with Shanks (1954), Mark (1958) found that the
daily mean temperature maximum occurred in July for all three stations (Asheville,
23.6°C; Banner Elk, 19.3°C; Mount Mitchell, 15.2°C). Annual precipitation increased
with increasing elevation (Asheville, 956 mm; Banner Elk, 1301 mm; Mount Mitchell,
1830 mm). Substantial precipitation occurred at all times of the year, with dry periods in
late spring and autumn. Snowfall began in October and continued through May, with
maximum snowfall occurring in February for Asheville (annual total of 79 mm), January

for Banner Elk (annual total of 224 mm), and March for Mount Mitchell (annual total of

10

378 mm). On average, Asheville received precipitation 128 days per year, while Banner
Elk received precipitation on 117 days, and Mount Mitchell 145 days.

In order to study intense rainfall in GSMNP, Bogucki (1972) used rain gages at 20
locations during 1937-1968. Station elevations ranged from 280 to 1935 m. A total of
293 intense rain events were recorded, with 71 percent of events occurring between April
and September. Intense rain events were defined as one or more inches received in one
hour, or defined as three or more inches received in 24 hours. Annual average
precipitation ranged from 55.19 inches (1401.83 mm) in Gatlinburg, TN, to 75.73 inches
(1923.54 mm) at Clingman’s Dome. Additional results were inconclusive as neither total
precipitation nor elevation had a direct relationship on the number of intense rainfalls
recorded at a particular gage.

In the surrounding area of GSMNP, Hicks (1979) used 1975-1976 summer
observations from nine communities: Gatlinburg, TN (443 m and 445 m), Oconaluftee,
NC (640 m), Cataloochee, NC (798 m), Sevierville, TN (283 m), Pigeon Forge, TN (314
m), Jones Cove, TN (360 m), Pittman Center, TN (411 m), Cosby, TN (524 m), and
Cataloochee Branch, NC (1463 m). In addition, five stations were used within
Greenbrier Cove: low-elevation cove (760 m), high-elevation cove (1040 m), oak-heath
deciduous (1040 m), oak-heath deciduous (1030 m), and hemlock (1030 m). Linear
relationships were determined based on temperature and precipitation data. The resulting

temperature relationship with elevation was expressed as:

y = —0.0046x + 23.68 (where r = —O.98).

ll

The resulting precipitation relationship with elevation was expressed as:

y = 0.065x + 62.58 (where r = 0.72).

Note that these relationships were based on stations at elevations less than 1500 m. As
indicated by Smallshaw (1953), at higher elevations precipitation tended to increase
upward at a slower rate suggesting that the true relationship between precipitation and
elevation may not be linear.

More recently, Bolstad et al. (1998) used 13 local stations located at the Coweeta
Hydrological Lab near Otto, NC, as well as 35 regional stations from the National
Climatic Data Center (N CDC) to analyze temperature lapse rates in order to create a
regression model. Station elevations ranged from 239 to 1967 m. Lapse rates varied
between 3°C/1000 m and 7°C/1000 m depending on time of day. Daily maximum
temperatures above canopy on south-facing slopes were 1.4°C warmer on average
compared to north-facing slopes. Regional regression models provided more accurate
estimates compared to models created from local observations or estimates created from
kriging.

Lastly, Gaffrn and Hotz (2000) used 30 years (1961-1990) of station data from the
NCDC’s C00perative Observing Program (COOP) to create a spatial climatology of
precipitation. The highest average amount of annual precipitation (>2000 mm) was
received in the southwest portion of NC, near SC and GA. An annual precipitation
minimum (1000 mm) was present near Asheville, NC, and was probably the result of

downslope flow. During winter months, the spatial pattern of one maximum was most

12

prominent. However, from May through August a second maximum was present near the
area of GSMN P. In agreement with other studies, the driest times of the year occurred in
spring and autumn, while the wettest time of year was summer.

While many climatological studies have been conducted within this region, many
are older studies that often utilized short periods of record and/or incorporated data from
a small number of stations (e.g., Whittaker, 1952; Smallshaw, 1953; Shanks, 1954; Mark,
1958; Bogucki, 1972; Hicks, 1979; Bolstad et al., 1998). Gaffin and Hotz (2000)
presented the longest spatial climatology of precipitation and used a larger station
network, but temporal variability was not discussed. Of the studies listed, only one
station covered snowfall (Mark, 195 8). Although these past studies are informative, they
have not adequately covered recent decades, which in turn has not allowed for a sufficient
and up-to-date temporal and spatial climatology of the region's temperature, precipitation,

and snowfall.

2.2. Synoptic Patterns Impacting Southern Appalachian Precipitation

Miller’s (1946) classic study presented two types of cool-season Atlantic Coast
cyclones. Using data spanning 1929-1939 (October to April only) from 208 cyclone
cases, he classified the cyclones as Type A or Type B. Type A cyclones frequently
formed along the eastern coast. Usually, a large high-pressure system and cold air mass
were present over much of the continental US, while a warm air mass was located over
the western Atlantic where the cyclones originated. Precipitation was mostly limited to
the areas ahead of the warm front and within the comma cloud region. Due to the

maritime moisture involved, these cyclones had the potential to result in heavy

l3

precipitation; however, the cyclone movement was typically to the northeast, so in many
cases they did not impact the southern Appalachians. The annual average frequency of
Type A cyclones reached a maximum in November, decreased to a minimum in January,
and then had a secondary maximum in March.

Miller's (1946) Type B cyclones formed over the southeastern US. along the
southern portion of another mature cyclone, which was usually occluded and positioned
near the Great Lakes region. Like Type A cyclones, precipitation with Type B cyclones
could exist ahead of the warm front and within the comma surrounding the low, but also
along the cold front and with the existing mature cyclone to the north. Therefore,
precipitation with Type B cyclones was more widespread. Type B cyclones were more
likely to have a warm sector that traversed the southern Appalachians. Another common
feature of Type B cyclones was cold-air damming, where a shallow layer of cold air was
present on the eastern side of the Appalachians (Richwein, 1980; Bell and Bosart, 1988).
The annual average frequency of Type B cyclones reached a maximum in January,
decreased to a minimum in February, and then had a secondary maximum in March.

In rare cases, some Atlantic Coast cyclones exhibited characteristics similar to
both Type A and Type B cyclones. Bosart’s (1975) addendum to Miller (1946) included
three more cyclone types studied during the years 1964-1972. Type C cyclones were
similar to Type B except that the secondary low was absent and a strong, cold anticyclone
was present to the east. Type D cyclones were similar to Type A, though they were
weaker systems that did not intensify during the northeast propagation. The cyclones
were more elongated and usually short-lived. Lastly, Type E cyclones were characterized

by an east-west frontal zone orientation with the low located north of the southern

14

Appalachians and south of New England. The systems moved rapidly eastward and an
anticyclone was present to the northeast. While Miller (1946) and Bosart (1975) both
highlighted the importance of Atlantic Coast cyclones on climate of the eastern US, they
did not, however, discuss other cyclone types that also impact the region.

Klein (1957), Reitan (1974), and Zishka and Smith (1980) presented frequencies
and origins of cyclones for North America. One important result from these papers was
the noted decrease in cyclone frequency during the middle portion of the last century.
Whether or not a direct link existed between the decreases in cyclone frequency and
documented global cooling of the same time period is unclear, as these authors did not
address it, though subsequent studies suggested such a relationship was present (van
Loon and Williams 1976a, 1976b; Harvey 1978; Diaz and Quayle, 1978).

Over the 1951-1970 period, Reitan (1974) found 4.4 cyclone events on average
traversing the southern Appalachian region during January, 2.9 events in April, 0.9 events
in July, and 1.8 in October, depicting an annual pattern. The cyclones were of the three
classic types: Alberta cyclones, Colorado cyclones, and Atlantic Coast cyclones. Results
between Reitan (1974) and Klein (1957) were similar; however, Klein’s earlier time
periods (1901-1914 and 1924-1937) revealed the Alberta cyclone type as the most
frequent, whereas Reitan’s results showed Colorado cyclones as more common. In both
cases, though, these cyclone types were more likely to affect the southern Appalachians
during winter and early spring, whereas during the warmer portion of the year, the
cyclone origins and paths were located farther north. Zishka and Smith’s (1980) results
of cyclogenesis and cyclone paths were similar to that of Reitan (1974), though Zishka

and Smith focused only on January and July (1950-1977). The authors found that more

15

cyclones were present during January and also that cyclones exhibited a lower minimum
sea-level pressure on average during January (988 hPa for January and 1001 hPa for
July).

O’Handley and Bosart (1996) described the impact of southern Appalachian
topography on cyclones. Studying 40 winter cyclones spanning 1979-1986 (December
through March), they discovered that frontal speed decreased by about 40 percent and
that fronts often developed a “kink” in their shape just east of the southern Appalachians.
In addition, thermal and pressure gradients were enhanced by the mountains, but once the
fronts passed the mountains, the fronts became weaker.

Konrad (1997) studied 312 heavy rainfall events and found five patterns
associated with heavy rainfall within the southeastern United States, although his patterns
were more on meso-scale phenomena and he did not include composite maps of his
patterns. Heavy precipitation events were defined as at least 50 mm of rainfall observed
at one or more stations within a 6-hour period. In most cases, high mixing ratios were
present at 700 hPa, while warm air advection was also present at 850 hPa. Pattern 1 had
boundary-layer convergence about 400 km northwest of the heavy rain area. A low-
pressure system with trailing cold front was probably present. Moisture and warmth
moved in from the Gulf of Mexico and had the highest values at the point of greatest
rainfall. Pattern 1 was most common in the western portion of the southern
Appalachians.

Pattern 2 was common with slow-moving cold fronts. Boundary-layer
convergence was located to the immediate northwest of the heavy rain location. Mean

winds were stronger for this pattern compared to the others. High moisture and

16

temperature advection values were, again, found near the point of highest precipitation.
Pattern 2 systems were most common in the western portion of the Appalachians, and
were also most frequent in May and June, as upper-level winds were still relatively
strong.

Weaker winds, with boundary-layer convergence occurring 100 km southeast of
the rainfall center, were common with Pattern 3. A weak, stationary frontal boundary was
often present. This pattern was most common during the summer months, as upper-level
winds were weaker. Frequent southerly/southeasterly flow resulted in frontal
overrunning. Pattern 3 was common along the eastern and southern portions of the area.

Pattern 4 was often associated with an elongated area of low-level convergence.
A low-pressure center moved north/northeast from the Gulf region. Moisture and warm
air advection maxima were present to the east of the boundary-layer convergence as a
result of southerly flow just east of the approaching low. Like Pattern 3, Pattern 4 was
also most frequent in summer and along the eastern and southern portions of the region.

Pattern 5 was associated with two regions of convergence. The first was located
to the northwest of the rainfall center and the other was located more than 400 km to the
southeast. This pattern was less common and no well-defined surface features were
identified. A weak front, with a northwest to southeast configuration, was sometimes
present. The south/southwesterly flow was responsible for high moisture and
temperature advection values near the point of highest rainfall.

With a focus on 500 hPa heights and surface fronts, Gaffrn and Hotz (2000)
discussed four subjective synoptic patterns associated with heavy precipitation in the

region. The heavy rain events were determined by either 3 inches (76.2 mm) of rain in 6

l7

hours and/or 4 inches (101.6 mm) of rain in 12 hours. The authors named the first pattern
“Synoptic,” which consisted of a strong low-pressure system with warm and cold fronts
that traversed the area. Low-level winds were southerly, with southwesterly winds above
850 hPa. The Synoptic pattern was most common during the cold portion of the year, as
the stronger westerlies played a larger role in determining the weather experienced in the
southern Appalachians. The second pattern was “Frontal” and consisted of an east/west
stationary front positioned over the region. The Frontal pattern was more common in the
warmer portion of the year. Westerly winds tended to be weaker, which allowed for a
stronger topographic influence on rainfall. The third pattern was “Meso High.” The
Bermuda High had a substantial influence on how much southerly airflow moved through
the region. Precipitation was most frequent with Meso Highs when a stationary front was
located to the north of the southern Appalachians. Lastly, “Tropical” patterns were
simply heavy precipitation events associated with tropical cyclones that affected the
region during summer and autumn months.

Diem (2006) supplied a summer precipitation and circulation analysis; however,
his study area was limited to the north and central portions of GA. He found that
prolonged wet periods were associated with mid-tropospheric troughing west of the study
region, while dry periods had the trough located to the east. For wet conditions, large
positive anomalies of 500 hPa and 850 hPa heights were located off the northeast coast,
as negative height anomalies were present over the Midwest. During dry periods, Diem
(2006) found that the height anomalies were located in the same general areas, but were
of opposite sign. The study area was usually under an anticyclone or on the western side

of a trough. The anticyclone was either of continental origin or an extension of the

18

Bermuda High. Spanning 1953-2002, Diem (2006) found that wet summers occurred
during the 19605 and the early 19905. Dry summers occurred during the mid 19505, the
mid 19705 to mid 19805, and in the late 19905.

Whereas many classic cyclone studies exist for North America (e.g., Miller, 1946;
Klein, 1957; Reitan, 1974; Bosart, 1975; Zishka and Smith, 1980), only Gaffrn and Hotz
(2000) specifically addressed synoptic weather patterns impacting the southern
Appalachians, and even that study focused solely on heavy precipitation events. Using
surface fronts and 500 hPa heights, Gaffin and Hotz (2000) concluded with four
subjectively-derived synoptic types (Synoptic, Frontal, Meso-High, and Tropical). Diem
(2006) presented a 50-year temporal climatology of precipitation and circulation analysis,
but his study area was geographically limited to a portion of GA. The studies discussed
still do not cover origins of precipitation (synoptic or otherwise) received in the region
during smaller precipitation events. The past studies also considered few synoptic

variables and levels, which may have also limited the number of synoptic patterns found.

2.3. Methods of Cluster Analysis
2. 3. 1. Overview

In climatological research, cluster analysis is one preferred approach used to
group items exhibiting similar characteristics. Cluster analysis can be used to objectively
create synoptic patterns or climate regimes, both of which I have expressed as an interest
to this study. Although cluster analysis was first introduced in the 19305 (e.g., Tyron,
1939), it was not until the middle portion of the twentieth century that technological

improvements assisted in shortening the time required for computation. Prior to the

19

quantitative revolution in the 19605, clusters were often formed either by entirely
subjective means (visual examination) or by a rule-based algorithm (e.g., Koppen, 1923;
Thornthwaite, 1931). The first climatological studies utilizing objective cluster analysis
techniques surfaced during the late 19605 (e.g., Skaggs, 1969). Objective approaches
have received more attention since that time, although subjective techniques still remain
common in climate research for defining regions (e. g., Shadbolt et al., 2006; Walters et
al., 2008) and synoptic patterns (e.g., Gaffin and Hotz, 2000; Walters and Winkler, 2001;
Walters et al., in preparation). In Section 2.3, I provide information regarding the
different clustering techniques available, methods used for determining the correct
number of clusters, and some of the available distance measures used for clustering. In
Section 2.4, I then provide a review of previous climate studies that used cluster analysis

and are relevant to this study.

2. 3. 2. Hierarchical Clustering Techniques

Many objective cluster analysis techniques are available, but most can be broadly
categorized as either hierarchical or non-hierarchical (i.e., partitional) in nature (Gong
and Richman, 1995; Tan et al., 2006; Wilks, 2006). Hierarchical clustering begins by
assuming each individual entity (i.e., data point) is a cluster, where the total number of
entities and the initial number of clusters is k. Step 1 merges the two most similar entities
together to create a new cluster and the total number of clusters then becomes k-l. This
agglomerative process continues with each step until either all entities have been
clustered into one cluster or the user terminates the clustering at a specified step to create

a desired number of clusters. A hierarchical cluster result can be illustrated as a tree

20

diagram or dendrogram and allows the user to consider nested or overlapping clusters if
desired (Gong and Richman, 1995; Tan etal., 2006; Wilks, 2006).

Of hierarchical approaches available, those most commonly used are single
linkage, complete linkage, group average, and Ward's method. Single linkage defines the
proximity of two clusters by the linear distance between the two closest entities located in
different clusters. In contrast, complete linkage defines the proximity as the distance
between the farthest two points in different clusters. Group average considers all entities
within each cluster by using the average pairwise proximity of all entity pairs between
two clusters: Ward's method represents each cluster by a centroid, but also minimizes the
sum of squared distances of each entity from its respective cluster centroid (Gong and

Richman, 1995; Tan et al., 2006; Wilks, 2006).

2.3.3. Non-Hierarchical Clustering Techniques

The most widely—used non-hierarchical cluster approach is K-means (Gong and
Richman, 1995; Tan et al., 2006; Wilks, 2006). Generally, K-means partitions entities
into a user-specified number of clusters (k). K-means considers data centroids (K-medoid
can also be used if a user strongly prefers using median instead of mean). To begin the
K-means approach, the user must first determine the desired number of clusters (k) and
the centroid locations. These locations (often referred to as “seed values”) can be
randomly chosen, although random initialization may yield different cluster results each
time the analysis is performed. Poor initial centroid placement can also result in clusters
that may not be particularly representative of the data, so users must consider initial

centroid placement carefully when using K-means. Once the initial centroids are chosen,

2]

the user creates k clusters by assigning each entity to the closest centroid. Once all
entities have been assigned, the centroids are computed again based on the entities within
each cluster. In the next step, entities are assigned to the updated centroids and then the
centroids are updated yet again. The process continues until either the centroids no
longer change or the user terminates the clustering after a predetermined number of
iterations (Gong and Richman, 1995; Tan et al., 2006; Wilks, 2006).

An extension to the K-means approach is bisecting K-means. In some ways the
bisecting approach resembles a reversal of hierarchical clustering. Initially, all entities
are clustered together. During the second step, the cluster is bisected to create two
clusters. The third step involves one cluster being selected and bisected to create a total
of three clusters. The process continues until k-l bisections have occurred and k clusters
have been created. Of course, this process creates the problem of determining which
cluster to bisect at each step. Two approaches include either always selecting the largest
cluster or selecting the cluster with the largest sum of squared errors. Once a cluster has
been selected for bisection, the centroid is removed and two new centroid values must be

determined (Tan et al., 2006).

2.3.4. Number of Clusters

While non-hierarchical clustering techniques suffer from the problems of
approximating centroid location, neither non-hierarchical nor hierarchical approaches
have a straightforward way of determining the correct number of clusters within a given
dataset. Numerous statistical measures have been developed and some literature has been

devoted exclusively to the subject (e.g., Milligan and Cooper, 1985). Most approaches

22

involve calculating a particular statistic and then plotting those values on the vertical axis
and the number of clusters on the horizontal axis. Ideally, a curve with a distinct peak or
bend will result, or the curve may reach asymptotic behavior. The point at which the
peak, bend, or asymptote occurs usually corresponds to the optimum number of clusters.
Unfortunately, a distinct signal in the curve is not always present, which still leaves a
subjective judgment call up to the researcher. To complicate the issue further, with so
many different statistical measures available and none of them perfect, determining the
best measure is difficult, particularly when different measures may provide a variety of
results.

Milligan and Cooper (1985) examined 30 different statistics using artificial
datasets with known numbers of distinct and non-overlapping clusters. Some statistics
measured the cohesiveness of entities within each cluster, while others measured how
different clusters were from each other, and some measured a combination of both.
Results from Milligan and Cooper's (1985) study showed that all 30 statistics were not
always perfect, with the number of clusters often under- or over-estimated. Such results
are not encouraging given that the number of clusters is usually not known in advance,
and, in a real world situation, clusters tend to be overlapping. With no known superior
measure available, the best option is to consider multiple measures and compare results.
If a map or plot of the data can be examined visually, that step may be helpful as well,

although it can become cumbersome with large or multiple datasets.

2. 3. 5. Distance Measures

Cluster analysis assigns entities into clusters by considering the distance (or

23

similarity) between all entities. Although many distance measures exist (e.g., Euclidean,
Mahalanobis, City Block, Supremum, Chebyshev, and Correlative), no universal measure
is used, so a researcher must determine which distance measure is best for a given dataset
(Mimmack et al., 2001; Tan et al., 2006). The most popular measure used in
climatological research has been Euclidean distance. A review by Gong and Richman
(1995) revealed that roughly 85 percent of climatological cluster studies used Euclidean
distance. Euclidean distance is calculated as the shortest possible distance between two
entities. The calculation is repeated between all entities to create a distance vector or
matrix, where if the number of entities is m, the resulting distance vector will have

m * (m — 1) / 2 elements or the distance matrix will be an m x m symmetric matrix with
zeros down the diagonal. Although Euclidean distance is easily calculated and frequently
used, users should be aware that the measure does include two assumptions when two or
more variables are to be clustered. First, Euclidean distance does not consider possible
correlations between variables. Therefore, variables are weighted equally even when
correlated variables are present. Secondly, Euclidean distance assumes that all variables
are based on the same unit scale, which is frequently not the case (e.g., when considering
temperature and precipitation). As a result, some variables may be inappropriately
weighted more than others during clustering (Mimmack et al., 2001).

Although a thorough review highlights the popularity'of Euclidean distance in
climatological research (e.g., Gong and Richman, 1995), some studies have reported the
benefits of using Mahalanobis distance and/or calculating principle components prior to
using Euclidean distance when two or more variables are being clustered (e.g., F ovell and

Fovell, 1993; Stephenson, 1997; Mimmack et al., 2001). To avoid the assumptions of the

24

~‘ ., "a

«pr—w _.

 

Euclidean distance measure one may first use principle components analysis and then
calculate Euclidean distance based on those component scores. The component scores
can also be standardized so that unit scales are consistent. Alternatively, Mahalanobis
distance can be used, since it attributes less weight to highly correlated variables and is
able to account for variables of differing unit scales. However, in order to calculate
Mahalanobis distance, one must first be able to invert the covariance matrix of the
dataset. This criterion can be met only if the original data matrix is “full rank.” To be
full rank, a matrix cannot have more variables than observations and also no variable can
be a linear combination of other variables (Mimmack et al., 2001). In events when a
matrix is not full rank, it is possible to use either a pseudo-inverse that results in a
truncated Mahalanobis distance (see Stephenson, 1997), or the principle components
approach described previously can be used. Now that cluster methods have been

discussed, past cluster studies that are relevant to this study will now be reviewed.

2.4. Relevant Cluster Studies

Stooksbury and Michaels (1991) clustered 449 stations within the southeastern
US. based on 36 monthly temperature variables and 36 monthly precipitation variables
spanning 1951-1980. Variables were standardized to give all variables equal weight.
States included in the study were VA, TN, NC, SC, GA, FL, AL, and MS. The authors'
hypothesis was that the conventional climate divisions that are based on county and state
borders do not adequately capture homogeneous climate zones and that a cluster analysis
may provide better results. Stooksbury and Michaels (1991) used Euclidean distance and

a two-step clustering approach beginning with hierarchical clustering to determine the

25

initial number of clusters and corresponding centroids. A partitioning cluster approach
was then used to create a 24-c1uster result and a 31-cluster result. Over 95 percent of
stations were found to cluster with at least one spatially adjoining station. Although
climatological differences between the clusters were not discussed in depth by
Stooksbury and Michaels (1991), the southern Appalachian region was found to have at
least four clusters. The “Western Appalachian & Valleys” cluster included the eastern
side of TN and northern GA. The “Eastern Appalachian” cluster included a narrow swath
along the extreme western edge of NC. The “Southside Virginia & Western North
Carolina” cluster included the more interior portion of western NC. Lastly, the “Interior
South Atlantic” cluster spanned portions of NC, SC, and northeastern GA.

Spanning 1949-1987, Gong and Richman (1995) considered summer precipitation
for central and eastern North America in an effort to compare an assortment of cluster
techniques. In most cases only two clusters existed for the southern Appalachian region.
One cluster included TN, western NC, and northern GA, while a second cluster included
the rest of NC, SC, and GA.

Fovell and F ovell (1993) used principle components analysis, group average
linkage, and the 344 climate divisions created for the contenninous US. by the NCDC in
order to improve upon Koppen's climate classification. The results of their study
included 12 variables for temperature and 12 variables for precipitation. Cluster results
included 14 and 25 clusters. Principle components analysis was used to eliminate
correlation between variables and the values were standardized so that all variables had
the same unit scale. Euclidean distance was then calculated. A climate division in

northwest SC was omitted from analysis because it was considered a poor data point.

26

Fovell and F ovell's (1993) 14-cluster result showed eastern TN and western NC clustered
together in the “E. Central” cluster, but the cluster included other areas as far away as MI.
Another cluster (“Southeast”) included eastern NC, SC, GA, and most the of the
remaining southeastern states. Their 25-cluster result again clustered eastern TN and
western NC into the E. Central category, which still included states much farther north.
Eastern portions of NC, SC, and GA were clustered into a Southeast cluster. Northern
GA, AL, and western TN were clustered into a second Southeast category. Clusters along
the Gulf of Mexico and Atlantic coast were warmer and wetter than inland clusters. Of
the inland clusters, the area including northern GA, AL, and western TN received
substantial precipitation, especially in winter and spring. In the end, Fovell and Fovell
(1993) classified most the southern Appalachian region as Koppen's Cfa category, with a
few of the higher elevation regions categorized as Cfb.

Winkler (1992) studied hourly precipitation across the US. during 1967-1983 in
order to determine any regional patterns in the diurnal receipt of heavy precipitation.
Euclidean distance and average linkage were employed in a hierarchical cluster analysis.
The data were segregated by season and then clustered. During winter, eastern TN, and
most of NC, SC, and GA were clustered together. Heavy precipitation was found to be
most common around sunrise. Spring clustered northeast TN and northwest NC together
with a distinct nocturnal maximum and midday minimum. Southeast TN, southwest NC,
the northern portions of SC and GA were also clustered together, but had no distinct
diurnal pattern. For summer, a large eastern cluster including the southern Appalachian
region had a prominent maximum during the mid-afternoon hours. Lastly, during autumn

the majority of the southern Appalachian region fell into a cluster that still exhibited a

27

 

 

mid-afternoon maximum, but it was not as distinct as it was during summer. A portion of
central and eastern TN also belonged to a cluster with a more midday precipitation
maximum.

Although Davis and Rogers (1992) did not focus directly on the southern
Appalachian region, they did use cluster analysis to create a spring and summer (April-
September) synoptic climatology for VA. In their study, surface data were used from
Washington, DC, and synoptic data at the 850, 700, and 500 hPa levels from the
surrounding area were also used. Davis and Rogers (1992) used the two-step clustering
approach described earlier by Stooksbury and Michaels (1991) to create nine clusters.
Cluster 1 represented a Bermuda High situation and was most common in July and
August. A 500 hPa ridge was located inland with a trough located off the Atlantic coast.
Cluster 2 had an anticyclone located over VA with an upper-level ridge present to the
west. The situation was most common in late spring. Cluster 3 represented the passage
of a surface cyclone to the north of VA during late spring months. A common mid-to-late
summer occurrence was categorized as Cluster 4, where a deep, warm-core anticyclone
was present. Cluster 5 had a surface low and large amplitude, short-wave trough located
over New England or eastern Canada and was most common in late spring. Cluster 6
occurred almost entirely in April and was depicted as a recent cold frontal passage over
the region. Occurring in spring, Cluster 7 was associated with an extra-tropical cyclone
off the east coast and an anticyclone approaching from the west. Cluster 8 occurred in
mid-to-late summer with a surface anticyclone to the northeast of VA resulting in onshore
flow at low levels. Lastly, Cluster 9 occurred mostly in June and July, but was still

common in other months. Westerly flow aloft was the most common quality associated

28

with Cluster 9.

A consensus clustering technique was presented by F ovell (1997), where the
author created a 4-cluster result for temperature and a 6-cluster result for precipitation
and then overlaid the two maps, ultimately creating 14 clusters. A higher-order approach
using seven temperature clusters and 15 precipitation clusters resulted in 26 consensus
clusters. The author used data from the NCDC's 344 climate divisions spanning 1931-
1981. Euclidean distance was used, but the author attempted to break the distance value
into several terms including a component for mean, seasonality, and coseasonality. With
the data standardized, the first two terms disappeared, leaving the coseasonality (or
correlation) term. Group average linkage was the clustering approach used. For the 14-
cluster result, eastern TN and western NC were clustered together. The remaining portion
of NC, SC, and GA were also clustered together as a more wetter and warmer region.
The 26-cluster result again had eastern TN and western NC as a cluster. Central NC was
its own cluster. Eastern NC and all of SC and GA were also clustered. The differences
between the clusters were not thoroughly discussed. In fact, Fovell (1997) concluded that
perhaps the southeast region had been over-clustered and that merging of some the
existing clusters could be justified.

When studying heavy precipitation for the southeastern US, Konrad (1997) used
cluster analysis. He first clustered a variety of temperature, moisture, and lifting indices
into three primary variables represented by vectors. The “A” vector represented the 500
hPa wind vector, while the “B” vector represented the area of greatest boundary-layer
convergence or low-pressure center. Lastly, the “C” vector represented the nearest ridge

of boundary-layer convergence or front. The cluster analysis used was the same two-step

29

cluster approach described previously by Stooksbury and Michaels (1991). The five
resulting cluster patterns were discussed previously in Section 2.2 of this chapter.

Walters and Winkler (2001) presented a climatology of 260 southerly low-level jet
cases for central North America during the 1991-1992 warm seasons. The study did not
directly focus on the southern Appalachian region, but the classification technique is still
relevant because the authors subjectively clustered southerly low-level jets based on
visual inspection of streamline curvature, latitudinal extent, and orientation of confluence
and deformation zones. Twelve distinct southerly low-level jet patterns were defined. A
subsequent study by Walters et al. (in preparation) followed the same approach to
categorize 546 northerly low-level jet cases during 1991 into 16 different patterns.

Lastly, Walters et a1. (2008) used principle components analysis and cluster
analysis to investigate low-level jets and create regions with similar characteristics.
Cluster techniques included using standardized principle component analysis scores.
Three different hierarchical approaches were used (Ward's minimum variance, average
linkage, and complete linkage). Variables included from each station in the clustering
were low-level jet frequency, average jet speed, and average jet height, with each based
on an eight-point compass. Results from the various cluster approaches were compared
to visually outline clustered stations and create low-level jet regions. Regions were
created for the annual scale, but also for each season. The southern Appalachian region
was just off the eastern edge of the study area (Nashville, TN, was the easternmost station
in the study).

Of these cluster analysis studies, each clustered relatively large areas. The

southern Appalachian region was often limited to a small number of stations or climate

30

division data points. Past studies covered temperature and/or precipitation, but snowfall
was never included. Clusters were often created over a time period, but no consideration
of possible temporal variability of the clusters was discussed. Most relevant studies for
the region discussed above relied primarily on objective clustering techniques.
Historically, however, subjective clustering techniques have been common in
climatological research (e.g., Koppen, 1923; Thornthwaite, 1931), and some recent
studies continue to use subjective approaches for defining climate regions or clustering
synoptic patterns (e.g., Walters and Winkler, 2001; Walters et al., 2008; Walters et al., in
preparation). A side-by-side comparison of objectively and subjectively-derived results

was not attempted or discussed by the studies cited.

2.5. Research Questions and Objectives

In view of this survey of relevant literature cited here in Chapter 2, I wish to

address four research questions in this study that previous studies have not fully

answered. I will refer back to these research questions throughout subsequent chapters.

1) What is the temperature and precipitation climatology of the southern Appalachians?

2) How many climate regimes exist in this region?

3) What are the synoptic origins of precipitation here?

4) To what extent does the answer to Question 3 depend on whether an objective,
subjective, or hybrid clustering approach is used

31

The purpose for Question 1 is to build upon studies referred to in Section 2.1.
Question 2 also harks back to Section 2.1, but should also help bolster the previous
results shared in Section 2.4 regarding the region's climate regimes. Studies in Section
2.2 have not fully described the region's synoptic climatology, hence the need for
Question 3. Lastly, Question 4 was born out of results discussed in Section 2.4 in that
those studies did not consider comparisons of various clustering approaches.

In order to answer the research questions, I will address a number of smaller
objectives along the way. Objectives for the temporal and spatial analysis include: 1)
selecting the region boundaries and time period of interest, 2) selecting and collecting the
climate data, 3) making decisions about the temporal and spatial resolution of the data, 4)
assessing the quality of the data and making adjustments in the dataset as needed, 5)
analyzing the data for temporal and spatial trends, and 6) summarizing results.

Objectives for the synoptic climatology include: 1) selecting and collecting the
appropriate data, 2) making decisions about the temporal and spatial resolution of the
data, 3) assessing the quality of the data and making analytical adjustments, as needed, 4)
determining criteria for a precipitation event, 5) plotting resulting synoptic pattern for
each precipitation event, 6) summarizing the events into clusters based on objective,
subjective, and hybrid approaches, 7) comparing and contrasting the results from the
various approaches, and 8) analyzing the fi'equency of synoptic patterns during past years
to monitor temporal variability.

As I will present in Chapters 4-6, my study considers additional stations and a
longer time period, allowing for a more comprehensive look at temporal and spatial

relationships that the historic studies may not have been able to obtain. My cluster study

32

also focuses solely on climate data from stations located within the southern Appalachian
region and the immediate surrounding area, allowing for a cluster representation of the
area with a higher spatial resolution. The synoptic climatology created by objective,
subjective, and hybrid approaches yields additional synoptic patterns not previously
covered. However, before the specifics of the results are presented, I will first describe

the study area, datasets, and methods used in Chapter 3.

'33

CHAPTER 3

STUDY AREA, DATA, AND METHODS

3.1. Study Area

The southern Appalachians are often defined as the mountainous region in the
eastern United States that remained unglaciated during the last glacial advance. Most
studies focus on the southern states of TN, NC, SC, and GA, although other states (e.g.,
AL, MS, KY, VA, WV, MD, OH, PA, and NY) are sometimes considered part of the
region, as well. For this study, a bounding rectangle of 34.50°N to 36.55°N and 81 .50°W
to 84.75°W including portions of TN, NC, SC, and GA was used to denote the study area
of interest (Figure 1.1). The bounding coordinates yield a rectangular area and although
the southern Appalachian mountain chain itself is not rectangular (the chain is oriented in
a southwest to northeast direction), a rectangular area is advantageous when gridded data
are used for map creation.

The study area has diverse topography including the Piedmont, Blue Ridge, and
the Ridge and Valley provinces associated with the southern Appalachians (Figure 3.1).
The Piedmont generally corresponds to the low-lying area below 500 m in the
southeastern portion of the study area. The Blue Ridge includes areas above 500 m
running from southwest to northeast. The Ridge and Valley province includes the low-
lying areas on the northwest side of the Blue Ridge. A small portion of the Appalachian
Plateau is present in the very northwestern portion of the map where elevations above
500 m are present (Figure 3.1). The complex topography of the region often leads to

orographic lifting and enhanced rainfall totals in the area. In fact, some parts of the

34

southern Appalachian region receive more precipitation than any other in the eastern US.
Large portions of the study area are federally-owned. As illustrated in Chapter 1, these
lands include GSMNP, and the Cherokee, Nantahala, Pisgah, Chattahoochee, Sumter, and

Daniel Boone National Forests (Figure 1.2).

 

Elevation (m)

ISBN '

35N

 

 

 

N

 

84W 83W 82W
Figure 3.1. Study area bounded by 34.50°N to 36.55°N and 81 .50°W to 84.75°W.
Geographic provinces are labeled with approximate boundaries included. Elevation is
shaded in gray based on meters above sea level.
3.2. Data for the Baseline Climatology
Climate data are generated by the Cooperative Observer Program (COOP) and are

archived by the NCDC. Monthly variables initially considered here included: minimum
temperature, maximum temperature, mean temperature, mean soil temperature, total
precipitation, total snowfall, and total evaporation. Beginning in 1890, the COOP

network is the largest and oldest weather observation network within the United States.

35

Volunteers use instruments provided by the National Weather Service (NWS) to record
the information, which is then conveyed to the NCDC, where the data undergo quality
control prior to being made available to the public (Winkler, 2004). More than 18,000
US. COOP stations with monthly data exist, 463 of which reside within the region
selected for this study (Figure 3.2). Station data for the COOP network are available for
free online at: http://cdo.ncdc.noaa.gov/CDO/cdo for users with an academic affiliation.
Downloads include a file with the observation list, a file with the station list (includes
station latitude, longitude, and elevation values), an inventory file, and an information

file, which describes how to interpret the observation file.

 

Elevation (m)

  

36N "

 

 

35N __ ' 3

Figure 3.2. Study area with elevation shaded in gray. Plus signs represent locations of
the 463 monthly COOP stations.

 

Elevation data used were from the National Elevation Dataset (NED) and were
obtained at no cost from the United States Geological Survey (USGS) using the Seamless

Data Distribution System (SDDS; available online at: http://seamless.usgs.govl. The

36

SDDS provides a seamless, high-resolution elevation dataset for the US. The dataset
itself is a 1:24,000 scale Digital Elevation Model (DEM) for the 48 contenninous states.
The projection is Geographic, with a resolution of one are second. The horizontal datum
is NAD83 and the vertical datum is NAVD88. Elevation units are in meters above sea

leveL

3.3. Data for the Synoptic Climatology

Daily precipitation totals are also available from the COOP dataset provided by
the NCDC. Although precipitation data were available for earlier years, I chose the study
period for the synoptic climatology as 1979-2006 to coincide with data availability of the
North American Regional Reanalysis (NARR) dataset. Approximately 19,000 COOP
stations have daily surface variables available with 151 of those included within my
region of study (Figure 3.3). As was the case with the monthly dataset, station data for
the daily COOP network are available online (http://cdoncdcnoaa.gov/CDO/cdo1 at no
cost for those with an academic affiliation. An observation list, station list (includes
station latitude, longitude, and elevation values), an inventory file, and an information file

that describes how to interpret the observation file are all included with the download.

37

 

Elevation (m)

36M ‘ ‘

 

35N

    

_ .+
4337' .'f‘" .+ + + 1 41—
84W 83W 82W

    

 

 

 

Figure 3.3. Study area with elevation shaded in gray. Plus signs represent locations of
the 151 daily COOP stations.
The NARR dataset, created by the National Centers for Environmental Prediction

(N CEP), is available for 1979 to the present (Mesinger et al., 2006). The NARR was
preceded by the global reanalysis dataset that was available four times daily on a 25° x
2.5° grid for 17 atmospheric levels (Kalnay et al., 1996). In contrast, NARR data are
available eight times daily on the 32 km Eta grid for 45 vertical levels. A map of the
NARR grid domain is available at: http://nomads.ncdc.noaa.gov/images/grid-ZZl .gif.
NARR values are integrated from rawinsonde, dropsonde, pibals, aircraft, surface, and
satellite observations, to arrive at a gridded dataset. Temperature, wind, moisture, and
other variables are available for free download at:
http://www.emc.ncep.noaa.gov/mmb/rreanU. Due to the large size of the files, the data

are available in a compressed, GRIdded Binary (GRIB) format.

38

3.4. Procedures for the Baseline Climatology
3. 4. 1. Acquiring the Data and Quality Control

To begin the analysis, I downloaded monthly COOP surface station data. For
downloads, users can request an entire state, specific station(5), specific climate
division(s), specific county(s), or a range of stations based on COOP identification
numbers. To test for continuity, I initially attempted separate downloads by state, county,
and climate division. An inspection of the station list showed that county information
was often not provided and climate division information was frequently incorrect.
Therefore, unless the COOP identification numbers of the desired stations are known in
advance, one should download stations for an entire state. The variables I initially
selected for this study were monthly and annual values of minimum temperature
(MMNT), maximum temperature (MMXT), mean temperature, (MNTM), mean soil
temperature (MOyz), total precipitation (TPCP), total snowfall (TSNW), and total
evaporation (TEVP). The year range was selected as 1890-2006 since 1890 was the first
year COOP data were available and 2006 was the most recent complete year at the time
of download. The download process was repeated for each of the four states in the study
region (TN, NC, SC, and GA).

All data were processed using Interface Description Language (IDL; available for
purchase at: hgp://rsinc.com/idl_/). The IDL source code for this project became quite
lengthy, so it is not provided in this dissertation, but I can provide an electronic copy
upon request. The National Elevation Dataset was downloaded as eight separate files. I
combined the files using the IDL program eleVation.pro. The dataset had a resolution of

10,980 rows and 15,304 columns within the study area. However, using such a large

39

array slowed IDL execution time considerably during subsequent map creation. As will
be discussed, I created numerous maps for this project, so the need for speedy processing
time had to be balanced against the necessity for maintaining a sufficient elevation data
resolution. After experimentation, 1 determined a compromise of counting every 50th
elevation point. In the end, the elevation array used had 220 rows and 307 columns.

I created clean1.pro for the purpose of extracting only the stations within the
study region from the TN, NC, SC, and GA files. Also included within clean1.pro was
code to: 1) convert station latitude/longitude locations from minutes into decimal degrees,
2) count the total number of stations, which was 463 in this case, 3) create maps of the
study area, elevation, and station locations (Figures 3.2), and 4) check the observations
for errors. Although observations undergo quality control by the NCDC prior to release,
some errors do remain present in the dataset, of which users must be cognizant. In some
cases, more than one entry for an observation can be found. Frequently, these multiple
entries consist of data entries and missing entries and a user can “combine” the two or
more entries to create one complete observation. However, at other times the distinction
of entries is not so clear (e.g., the entries report a non-missing value for the same month).
To eliminate this problem, in all cases I only considered the first observation entry.
Whenever a subsequent entry was found for the same observation, it was deleted.
Regarding variables, although only MMNT, MXNT, MNTM, MOyz, TPCP, TSNW, and
TEVP were requested, some entries for M034 (i.e., mean soil temperature taken in bare
ground at a depth of 50 cm) and M054 (i.e., mean soil temperature taken in sod at a
depth of 50 cm) were present in the data and were deleted. In the observation files, after

each monthly data entry, a flag value is provided (Table 3.1).

40

Table 3.1. Possible flag values included with monthly COOP observations.

£l_ag T. fiscfiition T

A Accumulated amount. Value may include data from previous month(s).
B

E

l

M

S

T

+

 

 

Adjusted Total. Monthly value based on proportional available data.

 

Estimatedmonthlr oranm'tmaL _ M- L _-_. _
Monthly means or totals based on incomplete time series.

Missing data element.

Amount is accumulating and will be included in a subsequent value.

I
|
"L.“y- -L. .. __ _ _-___ _,__,Il

 

 

 

Irasepfrzrecicitatien 91 §QQWE|.|;XaIt£=£0099-

__“ _ ,_ _ - - -___ _ _ _.-.____ _.__l

Phenomenon occurred on several days. J
Flag not needed. l

 

 

 

 

For consistency within the dataset, unless the flag was equal to “M,” “T,” or
blank, the data entry was reset to missing (“-99999”). Although observations were
available in some cases back to 1890, most stations had a later starting date. In addition,
despite having been checked for errors and missing observations, the data were
compromised by changes in station location and stations with short periods of record that
needed to be addressed. For these reasons, I created clusterl .pro to initially analyze data
in 10-year overlapping periods (e.g., 1930-1939, 1935-1944, etc., up through 1995-2004).
Some stations had a short period of record in comparison to the entire study period
chosen, and using IO-year overlapping periods allowed for many of those stations to be
included when they would otherwise have been unusable for further analysis. Stations
reporting observations 100 percent of the time were optimum, but not the norm. For this
study, stations with observations available 90 percent or more of the time during a 10-
year period were considered “good” stations. Stations with fewer than 90 percent of
observations available were not used.

My analysis within clusterl .pro revealed that the mean soil temperature (MOyz)
variable did not have a single station where data were available 90 percent or more of the

time for a single 10-year period. In addition, the total evaporation (TEVP) variable never

41

had more than seven stations with 90 percent of observations available in a given decade.
As a result, I omitted those two variables from further analysis. Data characterizing other
variables (minimum, maximum, and mean temperature, total precipitation, and total
snowfall) were available 90 percent or more of the time at some stations beginning
shortly after 1900, although it was not common until after 1930. Spatial analysis results
presented here are for years spanning 1930-2004 when 10-year overlapping periods were
used. In Chapter 5 I will present results from the decades of 1935-1944, 1965-1974, and
1995-2004 in order to provide spatial depictions for the beginning, middle, and end of the
record.

I converted all variables to metric units (°C for temperature variables and mm for
precipitation variables) and then used simple correlation analysis to determine any linear
temporal trends and test the resulting trends for statistical significance. Linear trends
were calculated for all stations, by month, and by station elevation during 1931-2006. In
many cases the series were not linear, so I performed a two-phase regression analysis (see
Solow, 1987) in order to highlight years when statistically significant “breaks” in the
trends were common.

Annual values are included within the COOP dataset, but values were rarely
reported. Therefore, I calculated annual values as the mean of value from all 12 months
of the year when all months had a non-missing value. Elevation categories were grouped
as low (< 500 m), mid (500-999 m), and high (_>_ 1000 m). Of the 463 monthly stations,
231 were in the low category, 167 were in the in mid category, and 65 were in the high
category. Figure 3.4 includes station locations by the respective elevation grouping, as

created by elev.pro. I will summarize the temporal results in Chapter 4 using mid-season

42

months (January, April, July, and October) to represent seasonal values (other months are
included in Appendix A). Spatial results for January and July are presented in Chapter 5

and results for April and October are included in Appendix B. Results will also be

summarized by elevation category.

Elevation (m)

 

   

36N I _ . ,1. ' _
: ‘s - , 1000

35N

 

 

N

 

 

Figure 3.4. Study area with elevation shaded in gray and station locations plotted by
elevation category. Black plus signs represent low-elevation stations located below 500
m. Dark gray triangles represent mid-elevation stations located between 500 and 999 m.

Light gray diamonds represent high-elevation stations located at 1000 m or above.

3.4.2. Cluster Analysis
Although IDL provides commands for both non-hierarchical (CLUSTER_WTS)

and hierarchical (CLUSTER_TREE) approaches, the approach I used here was a
bisecting K-means algorithm, which borrows some of its methodology from both
hierarchical and non-hierarchical methods. As discussed in Section 2.3.3., bisecting K-
means obtains k clusters by first assuming all stations are within one cluster and then

splits the cluster into two. Next, one cluster is selected for splitting and the process

43

continues until k clusters are reached (Tan et al., 2006).

IDL does not directly provide source code to perform the bisecting K-means
approach, but I was still able to perform the task using IDL and other software. To
initiate clustering, I needed to provide a pairwise distance measure between the climate
values at all stations. Euclidean distance is often used, but it is based on the assumptions
that all variables use the same scale and are not correlated. In cases where variables are
of varying scales and/or variables are correlated, a more appropriate choice is
Mahalanobis distance (Tan et al., 2006; Wilks, 2006). Specifically, the Mahalanobis

distance (D) between two stations (vectors) x and y is defined as:

D2 = [x — yiTisr'rx — y]

S is the covariance matrix of the data. All Mahalanobis distance calculations were
completed within clusterl .pro. Distance values were converted into similarity values

using:

sim = exp(-DZ)

Next, I performed bisecting K-means clustering using CLUstering TOolkit software

(CLUTO; available online for free at:

ht_tp://glaros.dtc.umn.edu/gkhome/cluto/cluto/download). The download includes a

manual that gives a thorough description of how to use the various options and

parameters within CLUTO (Karypis, 2003). In this case, similarity matrices based on

44

Mahalanobis distances were provided, so the CLUTO “scluster” command was

appropriate with the general syntax:

scluster [optional parameters] GraphFile NClusters

GraphFile was the name of the file storing the similarity matrix and NClusters was the
desired number of clusters. An example of a specific line command (to be typed as one

line) I used is included as follows:

./scluster -clmethod=rbr -clustfile=cluster9/Jan_l 9301939.xls.clustering.10 Jan_l 9301939.xls 10

In this example, an optional call to “-clmethod” was used to signal that repeating
bisections were to be used and a call to “-clustfile” was used to create a specific output
file name. I performed the clustering for January during 1930-1939 and stations were
bisected nine times to create 10 clusters. I created script files with edits made to the
above command so that I could automate execution for all months and annual values, for
all 10-year overlapping periods, for all climate variables individually and combined, and
for 1-10 clusters.

As discussed in Chapter 2, one challenge of cluster analysis is determining the
number of clusters that best represents the data. Some quantitative approaches are
available and CLUTO provides seven different statistical measurements. These values
either measure the cohesiveness within clusters, the difference between clusters, or some

combination of both. The results from each measurement were compared. Plotting these

45

values against the number of clusters results in a curve. Preferably the curve will reach
asymptotic behavior and level off. The point at which the curve levels off highlights
where additional clusters is not beneficial. I created such plots using cluster3.pro. The
resulting plots from the statistics were not always consistent. At times, the resulting plots
continued increasing beyond 10 clusters and did not level off. Other instances had a
leveling off after the inclusion of only a few clusters. However, in most cases leveling
off occurred at approximately five clusters. These results were further supported by the
cluster maps themselves (created using cluster2.pro), because mapping five clusters often
still revealed distinct patterns amongst the stations, with the addition of more clusters
making the cluster maps more difficult to interpret. For these reasons, and to make the

comparison of the maps easier, only the 5-cluster results will be shown in Chapter 5.

3.5. Procedures for the Synoptic Climatology
3.5.1. Acquiring the Data and Quality Control

In order to determine which specific meteorological (synoptic) scenarios account
for precipitation in the region, I used COOP precipitation observations at the daily time
scale. Similar to the monthly data, daily data were downloaded for the entire states of
TN, NC, SC, and GA. In this case, the daily record of precipitation (PRCP) was the only
desired variable. The year range was selected as 1979-2006 because the NARR dataset
begins in 1979, and although data were available for 2007 at the time of download, much
of the data values were still preliminary and had not all been fully checked via the
NCDC's quality control process.

Again, data were processed using IDL. I created the program clean2.pro for the

46

purpose of extracting the daily stations within the study area. The program completed the
following tasks: I) converted station latitude/longitude locations from minutes into
decimal degrees, 2) counted the number of stations, which was 151 in this case, and 3)
created a map of the study area, elevation, and stations locations (Figure 3.3). Next, I
created precip.pro to check the observations for errors and determine the number and
dates of precipitation events. The dataset was checked for any remaining preliminary
observations, multiple observations for the same day, observations reporting the wrong
variable (anything other than PRCP), observations reported in the wrong units, and
observations reporting dates in an incorrect order. One preliminary observation was
found and removed from further consideration. No instances of the other errors were
found. One peculiarity discovered: some stations reported the time of observation as
hour 24 when reported time is supposed to range between 0 and 23. Two flag values

were provided with each observation (Tables 32-33).

Table 3.2. Possible flag 1 values included with daily COOP observations.

1
|
l
l
l
l
1

:2 l

VT—JWZHFWJ’L
‘ 93

Descmtion

Accumulated amount since last measurement.

Accumulated amount includes estimated values (since last measurement).
-l-ESFi-rr-Iatcd swag W W- _ W- - WWW -
Value has been manually validated.

Missing data value (data value = -99999 for missing).

Included in a subsequent value. (data value = “00007)” or “-99999”).
Trace (data value = 00000 for a trace).

Expert system edited value, not validated.

Expert system approved edited value.

Flag not needed.

 

 

 

.JL. 1

 

_l
.

 

 

 

 

 

 

 

47

Table 3.3. Possible flag 2 values included with daily COOP observations.

lag ml Description _ U ' ' 1
Valid data element.

Valid data element (from “unknown” source, pre-1982).

Invalid data element (subsequent value replaces original value).

Invalid data element (no replacement value follows).

Validity unknown (not checked).

gNon-numeric data value has been replaced by a deciphered numeric value.
Substituted TOBS for TMAX or TMIN.

Time shifted value.

Precipifltion estimated from snowfall.

Transposedggits. _.
Changed units. ‘.
Adjusted TMAX or TMIN by a multiple of + or -10 degrees.
Changed algebraie sign

Moved decimal point.

Rescalingo_ther than F, G, or H.

Subjectively derived value.

-n§|

 

 

 

 

 

 

l

 

 

 

 

 

 

 

l

 

 

l

Extracted from an acoumulated value. 5
Switched TMAX and/or TMIN. i
Switched TOBS with TMAX or TMIN. '
W Substitution of “3 nearest station mean.’ _
Switched snow and precipitation data value
Added snowfall to snow depth.

HSwitched snowfall and__ snow _d_epth
Precipitation not reported; estimated as “O. ”
Manually edited value.

Failed internal consistency check.

Failed areal consistency check (beginningOct. I992).

Manually edited value derived by any of the procedures noted by Flags A- R.

 

 

 

Zarmh—zo'nm'unm>v~ewN—O

i
l

 

 

 

l
__ - .-___-_WW_W - .._____.._.__l
l
l

 

l

 

'I'
*

 

 

Cir-l CD 710 '5 Oi

 

 

 

*
Ir

 

Accumulated, estimated, and invalid observations were not desired, so I
considered only those observations meeting the following flag criteria: Flag 1 had to be
equal to “J,” “M,” “T,” “),” or blank, while Flag 2 had to be equal to “0,” “1,” or blank.

If any other flag values were present, the observation was discarded by resetting the
observation to missing (“-99999”) and Flag 1 to “M.”

I determined precipitation events within precip.pro. Previous studies of the region
have focused on “heavy” or “intense” precipitation events and used criteria such as: 3
inches (76.2 mm) or more of precipitation received at one or more stations in a day

(Bogucki, 1972), 50 mm of precipitation received at one or more stations in a 6-hour

4s

lil‘lilflll‘lli‘llllllli

period (Konrad, 1997), or 3 inches (76.2 mm) of rain received in 6 hours and/or 4 inches
(101.6 mm) of rain received in 12 hours (Gaffin and Hotz, 2000). However, my
suspicion was that much of the precipitation received in the southern Appalachian region
may actually result from smaller and lighter rain events. In fact, by considering the
amount of precipitation received and the percentage of available stations reporting
precipitation, I was able to determine that at least a trace of precipitation was received at
one or more of the stations on 9096 out of the possible 10,227 days (88.9 percent) during
the 1979-2006 study period (Table 3.4). Less stringent criteria were considered in order
to be more inclusive of smaller events, but due to computation restraints, I ultimately
chose my specification as 25 percent or more of the available stations reporting 0.5
inches (12.7 mm) or more of precipitation in a day. Given these criteria, over the 1979-
2006 time period 1437 precipitation events were found (Table 3.4). A more in-depth
discussion of the computation and virtual memory issues encountered will be covered in

Section 3.5.2.

Table 3.4. Possible number of precipitation events. Event criteria include the percentage
of available stations reporting (values along the topmost row) and amount of precipitation
received in a day (values along the leftmost column). The criteria of 0.5 in (12.70 mm)
or more of precipitation received by 25 percent or more of the available stations was
selected for this study resulting in 1437 precipitation events (shown in bold).

 

1 >0% 25% 210% 215% 220% 225% 230% 235% 240% 245% 250%
>00 in (0.00 mm) 9096 7238 6464 5905 5390 4921 4546 4151 3824 3496 3231
20.1 in (2.54 mm) 8126 5981 5135 4500 3967 3534 3176 2859 2591 2329 2097
£202 in (5.08 mm) 7594 5219 4311 3649 3164 2771 2455 2135 1900 1663 1463
320.3 in (7.62 mm) 7164 4621 3698 3084 2602 2209 1881 1636 1436 1244 1093
'20.4 in (10.16 mm) 6793 4117 3185 2562 2121 1762 1494 1278 1108 971 827
20.5 in (12.70 mm) 6426 3703 2743 2140 I719 1437 1190 1023 867 752 636
I206 in (15.24 mm) 6080 3272 2311 1764 1426 1158 982 824 702 577 482
I207 in (17.78 mm) 5776 2877 1978 1483 1176 951 787 658 542 439 371
20.8 in (20.32 mm) 5483 2534 1684 1245 971 795 635 521 421 354 287
£203 in (22.86 mm) 5194 2202 1429 1040 808 643 517 432 336 284 218
.21 .0 in (25.40 mm) 4931 1922 1212 866 672 529 422 339 282 219 I74

 

 

 

 

49

In contrast to the once-daily COOP data, NARR data are available eight times
daily at 0000, 0300, 0600, 0900, 1200, 1500, 1800, and 2100 UTC. Because the daily
COOP observation time varied amongst stations, only the last NARR observation of the
day (2100 UTC) was considered in this study. NARR files are split by time period, so
365 files (or 366 for leap years) needed to be downloaded for each year. Despite being
compressed into GRIB format, the files were still quite large and contained values for
many variables at many vertical levels and at a horizontal resolution of 32 km. For this
study, I wished to download the following variables: height, temperature, specific
humidity, vertical velocity, and the u and v wind components at 850 hPa and 700 hPa and
height, and the u and v wind components at 300 hPa. Fortunately, software exists for
automating the download and variable extraction process. Instructions on how to
complete these tasks are available at:
ht_tp://www.cm.ncep.noaa.gov/products/wesley/ fast downloading grib.html. The
instructions guide the user through the process of downloading the required software
(perl, grep, cURL, get_inv.pl, and get_grib.pl). A sample script is also provided to assist
in the automation process. A portion of the revised script that I created is included in
Appendix C. Once all files were downloaded in GRIB format, an approach for
uncompressing the data was needed. Various options for uncompressing GRIB files are
available, but a fairly straightforward approach is available at:
hgpzflwwwcpc.ncepnoaa.gov/products/wesley/wgg'b.htrnl. Software called “WGRIB”
is available in the C programming language. I downloaded and compiled the WGRIB
source code manually, but some compiled versions are available for direct download.

Appendix C provides information regarding WGRIB and uncompressing GRIB files.

50

3. 5. 2. Clustering Synoptic Patterns

Objective and subjective approaches can be used for clustering synoptic patterns.
I considered three different schemes for this study: 1) an objective approach, in which I
created all clusters objectively using IDL and CLUTO, 2) a subjective approach, in which
I visually examined and clustered all 1437 cases, and 3) a hybrid approach, wherein I
initially created 50 objective clusters and then subjectively combined the clusters further.

For the objective approach I utilized CLUTO software to objectively cluster the
synoptic patterns. CLUTO requires an input matrix of similarity values based on
pairwise distances. The distance measure I used was Mahalanobis distance and I created
the IDL program cluster4.pro for the calculations. To begin, a matrix with dimensions
Events x Variables was required, where Events = 1437 (the number of precipitation
events) and Variables = 1,450,095 (15 variables at all 96,673 grid points). In addition,
some matrices required a dimension, M, where M = (Events * (Events — 1)) / 2. In this
case, if Events = 1437, then M = 1,031,766. Unfortunately, these large numbers created
execution problems due to memory allocation for each matrix. In order to remain within
the computation and virtual memory limits of the IDL software and computer, it was
necessary for me to decrease the overall size of the input dataset. l experimented with
matrix dimensions until values were reached that the software and computer could handle
successfully. In the end, I chose seven of the original 15 variables (850 hPa height,
temperature, and dewpoint, 700 hPa height, temperature, and dewpoint, and 300 hPa
height). In addition, rather than use the entire 349 x 277 horizontal grid provided within

the NARR dataset, I selected a much smaller 8 x 9 subset (Figure 3.5).

51

 

A

N

 

 

 

 

Figure 3.5. Subset of NARR grid points selected for objective cluster analysis.

I created the program extractpro to extract the subset of grid points and variables.
Also included in the program were conversions for Kelvin to Celsius temperature and
specific humidity to dewpoint. Once the Mahalanobis distances and similarity values
were created successfully, the values were used within CLUTO. Up to 100 clusters were
created using the repeated bisections method discussed earlier. In addition, clusters were
created using each of the seven statistical measures provided by CLUTO (Karypis, 2003;
Zhao and Karypis, 2003) in an effort to determine the optimum number of clusters.
Unfortunately, the results from the statistical measures were inconclusive because the
plots (created by cluster5.pro) did not display a distinct pattern even when up to 100
objective clusters were used, so in the end I chose to run the objective analysis using the

same number of clusters that was found during the subjective analysis.

52

For the subjective approach, I created maps for all 1437 events using events.pro.
The synoptic patterns for each precipitation event were printed and then subjectively
categorized by me to create clusters. Separate maps were created for each variable at the
850, 700, and 300 hPa level. I considered all levels and variables as I subjectively
clustered, but the synoptic patterns were most distinct at the 850 hPa level. In order to
quickly count the number of cases within all subjective and objective clusters, as well as
determine seasonal distributions for all clusters, I created the programs cluster6.pro and
cluster7.pro. Only the 850 hPa maps for the subjective and objective approaches will be
shown in Chapter 6, but the 700 and 300 hPa maps are included in Appendix D.

In the hybrid approach, I used CLUTO and events.pro to initially cluster the
synoptic patterns from the 1437 precipitation events into 50 objective clusters and create
maps for each. Much like the subjective approach, I then visually clustered those 50
further, in order to end with a smaller number of clusters. The results from this hybrid
approach were then compared to the subjective and objective results. All maps for the
hybrid approach are included in Appendix E.

For each cluster, the associated synoptic data from all cases were averaged to
create a composite map (using objectivepro, subjective.pro, and hybrid.pro).
Comparison of the resulting synoptic patterns was done mostly in a qualitative way by
describing the differences between the synoptic patterns. However, some quantitative
comparisons are provided by referencing contour values on the maps, for example.
Additionally, I calculated frequency values in Cluster7.pro in order to show which
synoptic patterns were more common than others and to highlight temporal changes that

occurred over the 1979-2006 period. Frequency values were calculated for three

53

overlapping periods of 1979-1992, 1986-1999, and 1993-2006. In addition, trends of the
annual frequency values were tested for statistical significance. Results of the

dissertation will now be presented in Chapters 4-6.

54

CHAPTER 4

RESULTS AND DISCUSSION OF THE TEMPORAL ANALYSIS

4.1. Overview

First-order changes for all variables during 1931-2006 are reported in Tables 4.1-
4.5. In this chapter, I will present and discuss results for January, April, July, October,
and Annual. Additional figures for other months are included in Appendix A. Please
note that the ranges of y-axis values on the figures are not always the same for all
months. I adjusted the y-axis values based on the inter- or intra-annual variability of
temperature and precipitation. The results of this chapter build upon those from previous
studies and provide new temporal information to more thoroughly answer Question 1

posed in Section 2.5.

55

00.Hom 0H.0v vo.NI 0H.0I 00.0 00.0 00.0 00.0 No.0 00.00 0N.0H ma.Hma 00.00H swam

mm.N0I mv.mmt 00.0I 0N.0I 00.0 00.0 00.0 00.0 Nv.0 mm.m Nv.mvl 0H.0HI 00.0H Uflz

mm.NmI 0v.OMI v0.01 00.0 HH.0I 00.0 00.0 00.0 00.0 00.0 mm.mNI 00.mHI N0.a BOA

0N.mvt 00.0NI mm.mI va.0| 00.0I 00.0 00.0 00.0 00.0 00.0H 00.0NI mm.mt ma.mH Had

HMSCCKN UGO >02 #00 Qmm 05¢ H5O SSH: \nmz HQAN HOE Dwm EMU >OHM
080:2 :55 3305 .59 a cause Ramsay .2. 23

mo.mmH 0v.NHI Mv.vm mm.m mm.mm N0.mm1 om.mvi no.00 00.na 00.0H 00.0 mm.mi no.0 coax

ov.Hm mv.omi mm.mv 00.HI 0H.mm mm.mHI ov.Nm1 >V.>m mm.mm 00.0 mm.0HI om.mr mH.H 0a:

mm.om 00.0HI 00.0N mw.HH 00.mv RN.NHI va.omi mm.Hm mm.mm H>.H mm.mi om.mHI hm.mi 300

mH.mm vm.mHI mm.mm >>.0 0>.>v vv.mHI N0.>NI N0.Nm mm.mm Hm.H >>.mi mm.HHI vm.vi Had

am5cc< 000 >02 000 00m 05¢ H50 550 >02 H00 H02 000 :00 >0Hm
88-32 €80 Swansea 33 5 ago Baa“; .3. see

00.HI >0.NI mv.0i 0N.NI NH.NI HH.HI mm.0I hH.NI mh.HI H0.HI 00.HI 00.HI mn.mt cmflm

HN.0I m>.0I mm.0 n0.0r 00.0I ma.0 mm.0 mm.01 vm.0i >H.0 vv.0 mm.or 0H.NI 0H2
NH.HI NH.HI m0.0 mm.HI om.HI mm.0I 00.0: mm.HI vm.HI mn.0t mm.0r 00.0: 00.NI 30A
00.01 «0.01 00.0 >H.HI 0H.HI ma.0t MH.01 mm.HI mo.HI 0N.0I 00.0 00.0: mv.NI Ham
Hm5ccd 000 >02 000 00m 05¢ H50 :50 >02 H00 H02 000 000 >0Hm

003-30. Gov 2820088 :38 E 00030 3.80800. .00 0300.

mo.m1 mo.mi m>.HI mo.vi H>.m1 mm.mi mm.HI am.MI va.mr mh.HI mm.HI Hm.mi 0H.vi nmflm

mm.01 am.oi m>.o vm.HI NN.HI ao.or Ho.or >m.ai mm.ot mm.0 m>.0 vm.0| m>.HI 0H2

mm.HI mm.0i 00.0: Hfi.mr mH.NI mm.oi 00.0: o>.NI N0.NI Nv.or 00.0I mm.oi ov.mi 304

Hm.0I 00.01 NN.0 om.HI m>.HI 00.0I Hm.01 0H.NI mm.HI 00.0I mm.o No.0: NN.NI Had

Hm5cc¢ 000 >02 poo 00m 05¢ H50 :50 >02 50¢ mm: 000 500 >0Hm
.88- _ m2 600 sausage Sagas a oases ESSA .3. 29¢

o>.or 0m.HI an.o mm.or H0.0I v0.0I 00.0 mw.or No.01 mm.or mv.0I vo.HI Nv.m1 20H:

ma.or 00.5: mm.o mH.ou m5.or mm.o ea.o 40.0- mm.or Hm.or mo.o om.ou Hm.mr 052

mm.or 04.0: 05.0 mo.HI 00.5- mm.o- mH.oI mm.0t no.5: 00.5- mm.ou 50.5: mm.mr 305

em.or mH.HI 00.0 mm.o- we.or m5.o am.o m«.or 00.01 mm.ou efi.o- we.ou no.~u 00¢

Hm5cc< 000 >02 000 00m 05¢ H50 :50 >02 50¢ H02 000 500 >0Hm
.0000. 0 m9 600 8380088 85800:: S owqfio 3000800. .34 030,—.

56

4.2. Temporal Results of Temperature
4. 2. 1. January

Time series plots for January show a very distinct pattern in minimum, maximum,
and mean temperature (Figure 4.1). A distinct warming trend was present at all
elevations beginning in the late 19705 and continued through the end of the series. In
fact, results from a two-phase regression analysis highlight that statistically significant
breaks (p S 0.10) in the time series occurred most often in the mid-to-late 19705 for cool
season months and support the documented 1976/1977 climate shift found by others
(Trenberth, 1990; IPCC, 2007a). In many cases, January temperatures warmed by as
much as 4.00°C or more over the last few decades. Nonetheless, linear values show that
an overall cooling trend occurred at all elevations over the entire 1931-2006 period. High
elevations experienced the greatest amount of cooling with a drop of 342°C for minimum
temperature, 416°C for maximum temperature, and 3.78°C for mean temperature, all of
which were statistically significant at the p 5 0.01 level. Low-elevation stations
experienced the second-most cooling with changes of 282°C for minimum temperature
(significant at the p S 0.05 level), 246°C for maximum temperature (significant at the p S
0.05 level), and 264°C for mean temperature (significant at the p _<_ 0.01 level). Lastly,
mid elevations had temperature drops of 261°C for minimum temperature (significant at
the p _<_ 0.05 level), 1.75°C for maximum temperature (significant at the p S 0.10 level),

and 216°C for mean temperature (significant at the p S 0.05 level), respectively.

57

Low Elevations Mid Elevations

 

 

   

 

 

 

 

 

 

 

 

 

   

 

 

 

 

 

g 20 V I j g 20 ' ' T I I I

‘6 '5 _

E T)

o o

.E .E

2 8

3 D

+4 +-'

E E

0 4)

Q. Q.

g —10- g

I— -15 AL ._L I I I I" —15 L I I I
1940 1960 1980 2000 194—0 1960 1980 2000

Years Yeors
High Elevations All Elevations

07 20 ' ' ' or 20 ' ' '

3 3

E 15- E _

a) 0

O 10 O

C C

._ 5 ..

8 2

3 0 3

E E

o —5 a)

Q. Q

s -10 s

1'— —15 I I . I I 1— —15 I I I I
1940 1960 1980 2000 1940 1960 1980 2000

Years Years

Figure 4.1. January minimum (bottom line), maximum (top line), and mean (center line)
temperature in °C. A linear fit was calculated for each variable over 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

While the 1931-2006 linear trends suggest that cooling occurred, the trends are
also heavily influenced by the period of extended warmth at the beginning of the record.
Although warm values were present at the end of the series as well, at the time of
analysis, data were available through 2006 only. At the time of this writing (2009),
however, preliminary data suggest that the years 2007 and 2008 were also warm and their
addition will likely impact any subsequent long-term study. These monthly value results
also suggest that substantial decreases in the daily temperature range may be present

(e.g., temperature range values for high elevations were statistically significant at the p S

0.05 level), which could have an effect on the region's plant and animal species that rely

58

on (or are vulnerable to) variation in diurnal temperature range.

4. 2.2. April

Compared to January, temporal changes in temperature for April were more
modest (Figure 4.2). Over 1931-2006, low-elevation stations had a decrease in minimum
temperature (106°C; statistically significant at the p S 0.05 level), maximum temperature
(042°C; not statistically significant), and mean temperature (073°C; not statistically
significant). High-elevation stations also experienced some cooling across the board with
028°C for minimum temperature (not statistically significant), 1.79°C for maximum
temperature (significant at the p S 0.01 level), and 101°C for mean temperature (not
statistically significant). Changes at mid elevations were not statistically significant with
a decrease in minimum temperature of 021°C. Maximum temperature increased by
053°C and mean temperature increased by 017°C. Such results suggest an overall
increase in the diurnal temperature range at low elevations (not statistically significant))
and mid elevations (significant at p S 0.10 level) and a decrease in temperature range for

high-elevation stations (significant at p _<_ 0.01 level).

59

Low Elevations Mid Elevations

 

 

 

 

 

 

 

 

'3 25: l - I I ' '3 25 I I l '

a zo— - a zoM/lrWVM/Mxlwd-

o u

s Isévval/NMWU- .s 15- —

g _ 3 WW

3 10‘ ‘ 3 10' ‘

E - E

3 5W- 8 SMWVW-

E E

+9 o . . . . I33 o W . . .
1940 1950 1980 2000 1940 1960 1950 2000

Years Years

High Elevations All Elevations

 

 

Temperature in Celsius
or
1 1
Temperature In Celsrus
on
g
l

 

 

 

 

 

 

.0 MW‘NWV 10- -
o I I I 0 l I I I ‘
194-0 1960 1980 2000 1940 1960 1980 2000
Years Years

Figure 4.2. April minimum (bottom line), maximum (top line), and mean (center line)
temperature in °C. A linear fit was calculated for each variable over 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (2 1000 m; bottom left), and all elevations (bottom right).

4.2.3. July

Much like April, most temporal changes in temperature for July were smaller than
they were during January (Figure 4.3). Two-phase regression results suggest that
statistically significant breaks occurred in the 19605, which corresponds to a period of
cooler temperatures. Warm season temperatures have tended to rise since that time.
Long-term deviations were not necessarily consistent across each variable or elevation.
Low-elevation stations experienced cooling for all variables with 019°C for minimum

temperature (not statistically significant), 099°C for maximum temperature (significant at

the p S 0.05 level), and 058°C for mean temperature (not statistically significant). Mid-

60

elevation stations had warming for minimum temperature (047°C) and mean temperature
(025°C), and slight cooling for maximum temperature (001°C), none of which were
statistically significant. Stations at high elevations had a small amount of warming for
minimum temperature (008°C; not statistically significant) and more substantial cooling
for maximum temperature (199°C; significant at the p S 0.01 level) and mean

temperature (095°C; significant at the p S 0.05 level).

 

 

 

 

 

 

 

 

Low Elevations Mid Elevations
'02) 35 I l ' ' ' V I ' '3 35 +1 I I ' l ‘
O U
E zswwdwvv- E 25- -
e 8 WWW
*3 ZOW‘ 3 20" -
E E
3 15— 3 15W-
; i s
I— 10 . . . . I— 10 . . I I

1940 1960 1950 2000 1940 1960 1980 2000
Years Years
High Elevations All Elevations

 

 

35 ‘
”W‘-
20" ‘

15W1

10 1 J_ I I 1 O 4_1 A L I I
1940 1960 1980 2000 1940 1960 1980 2000
Yea rs Yea rs '

 

 

 

 

Temperature ln Celsrus
N
o
1
Temperature In CeISIus

 

 

Figure 4.3. July minimum (bottom line), maximum (top line), and mean (center line)
temperature in °C. A linear fit was calculated for each variable over 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (2 1000 m; bottom left), and all elevations (bottom right).

In general, the overall trends for April (Figure 4.2) and July (Figure 4.3) were

similar. Low- and mid-elevation stations exhibited relatively stable temperatures over

61

time, with a very small amount of cooling during 1931-2006. In contrast, stations at high
elevations had maximum temperatures during spring and summer decrease steadily over
1931-2006. Mean temperature at those locations also decreased to a lesser extent, while
minimum temperatures remained stable. Diurnal temperature ranges during July
decreased at low elevations (statistically significant at the p _<_ 0.10 level), mid elevations
(not statistically significant), and high elevations (statistically significant at the p S 0.01

level).

4. 2. 4. October

October (Figure 4.4) experienced regional cooling throughout the 1931-2006
period. High-elevation stations had cooling of 053°C for minimum temperature (not
statistically significant), 4.03°C for maximum temperature (statistically significant at the
p _<_ 0.01 level), and 226°C for mean temperature (statistically significant at the p S 0.0]
level). Stations at mid elevations encountered cooling of minimum temperature (013°C;
not statistically significant), maximum temperature (124°C; statistically significant at the
p S 0.01 level), and mean temperature (067°C; not statistically significant), as well.
Low-elevation stations cooled by 1.05°C for minimum temperatures (not statistically
significant), 2.11°C for maximum temperatures (statistically significant at the p S 0.01
level), and 158°C for mean temperatures (statistically significant at the p S 0.01 level),
respectively. With minimum temperatures cooling less than maximum temperatures,
diurnal temperature ranges likely decreased over the period at all elevations. In fact,
temperature range values over time were statistically significant for low and mid

elevations at the p S 0.10 level and at the p S 0.01 level for high elevations.

62

Low Elevations Mid Elevations

 

 

Temperature 1n Celslus
5
Temperature in Celsms
G

 

 

 

 

0 1 1 A; 41
1940 1960 1980 2000 1940 1960 1980 2000
Years Years
High Elevations All Elevations

 

 

 

 

Temperature 1n Celslus
-' N
01 0
Temperature in Celsius
-' N N
or o tn

 

 

 

 

O 41 L L

1940 1960 1980 2000 1940 1960 1980 2000
Years Years

 

 

Figure 4.4. October minimum (bottom line), maximum (top line), and mean (center line)
temperature in °C. A linear fit was calculated for each variable over 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (2 1000 m; bottom left), and all elevations (bottom right).

4. 2. 5. Annual

Averages over the entire year showed a general cooling trend over the 1931-2006
period at all elevations (Figure 4.5). Minimum temperatures at low elevations dropped
by 093°C (statistically significant at the p S 0.01 level), mid elevations dropped by
019°C (not statistically significant), and high elevations dropped by 070°C (statistically
significant at the p S 0.05 level). Maximum temperatures cooled 128°C at low elevations
(statistically significant at the p S 0.01 level), 025°C at mid elevations (not statistically

significant), and 3.05°C at high elevations (statistically significant at the p S 0.01 level)

while mean temperatures decreased by 1.12°C at low elevations (statistically significant

63

at the p S 0.01 level), 021°C at mid elevations (not statistically significant), and 1.78°C at

high elevations (statistically significant at the p S 0.01 level). These results show that the

diurnal temperature range decreased at all elevations with low-elevation values

statistically significant at the p S 0.10 level and high-elevation values significant at the p

S 0.01 level.

Temperature in Celsius

20
15
10

Temperature in Celsius

Low Elevations

 

25’
20.

15.

10'

1 I ' I T ' V I I

‘
‘
yMAVAC/MM~®JvVv~”qunfl/‘
D ‘
n d

1A.A_A1AAAA

W
h

a
D ‘
h 1
p
I.

1
.1

 

 

l A A A 1 J g 1 LA L_._l

 

1940 1960 1980 2000
Years

High Elevations

 

251

Y I v v v—r v v v ' v v *v—y v

 

 

4 L AL A L J A A L l A A A l A

 

1940 1960 1980 2000
Years

Mid Elevations
ZOW‘:

15

 

10

 

 

Temperature in Celsius

 

1940 1960 1980 2000

 

 

 

 

Years
All Elevations
a) 25 F V v V V T v 1 v v v v v f 1 v .
.3 Z I
(0 - J
33 20 WW 1
U : 1
E 15 W ‘
2 I I
3 L j
g 10: 1
1... _ ,
a 5 W1
,6, a a
’— 0 . A l g‘ L4 A A A L A L A l A ‘
1940 1960 1980 2000
Years A

Figure 4.5. Annual values of minimum (bottom line), maximum (top line), and mean
(center line) temperature in °C. A linear fit was calculated for each variable over 1931-
2006. Results are included for low elevations (< 500 m; top left), mid elevations (500 —

999 m; top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom

right).

In terms of seasonal trends presented in Figures 4.1-4.4 , the results here generally

suggest that autumn and winter months warmed in recent decades, while spring and

64

summer months cooled. These results imply that the overall annual cycle of temperature
also may have decreased with time. At all elevations over the 1931-2006 period, the
annual maximum for all temperature variables was reached in July and the minimum was
reached in January (Figure 4.6). This figure also highlights the trend of generally

decreasing temperature values with increasing elevation.

Low Elevations Mid Elevations

 

 

   

 

 

 

 

 

 

 

 

   

 

 

 

 

 

 

g 40; ’ * r - - g +0§ ' ' ‘
'_q—_) : £5 E
d) ' (D
o - o
E .5
8 8
3 B
8 8
<1) <1)
0. Q.
g é a
1—._10:....J..... 1—_10:...A,.....
12.314513789101112 123456789101112
Months Months
High Elevations All Elevations
040:--...---.-: ”40----.-
.2 g s .2 a
, ; tn :
% 30:- 1 s 30:
o 1 o
E 20; E 20;
8 8 f
3 1o; 3 103
E E E’. E
8. QE 8. 0:
g g ;
1—_105..r,..1191 1——10:..........
123456789101112 12.114515789101112
Months Months

Figure 4.6. Intra-annual trends of minimum (bottom line), maximum (top line), and
mean (center line) temperature in °C versus months of the year. Values were averaged
over the entire 1931-2006 period. Results are included for low elevations (< 500 at; top
left), mid elevations (500 — 999 m; top right), high elevations (Z 1000 m; bottom left),
and all elevations (bottom right).

4.3. Temporal Results of Precipitation and Snowfall

4. 3. 1. January

First-order changes in total precipitation and total snowfall from 1931-2006 are

65

reported in Tables 4.4 and 4.5. Two-phase regression results did not highlight any
common breaks in the time series for precipitation or snowfall. January precipitation for
low—elevation stations decreased by 8.57 mm (Figure 4.7), while snowfall increased by
1.02 mm (Figure 4.8), neither of which was statistically significant. Precipitation and
snowfall increased for mid and high stations with precipitation going up by 1.13 mm at
mid elevations (not statistically significant) and 9.67 mm at high elevations (not
statistically significant). Snowfall increased by 17.08 mm at mid elevations (not
statistically significant) and 196.87 mm at high elevations (statistically significant at the p

S 0.01 level).

Low Elevations Mid Elevations

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

400 400
E E
E 300 - E 300 -
E .E
g C
.5 200- -f-3 200 '
B B
.a .a
‘6 100 '0 100
“13 8
l [L
O I I l I O I I I I
1940 _1960 198C] 2000 1940 1960 1980 2000
Years Years
High Elevations All Elevations
400 ' ' 400 ' ' ' '
E E
E 300 -
E .E
S S
.g ,3 200 -
B 3
EL 29 100
0 0
l‘.’ E
(L 11.
O I I I l 0 I I I I_._
1940 1960 1980 2000 1940 1960 1980 2000
Years Years

Figure 4.7. January precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (2 1000 m; bottom left), and all elevations (bottom right).

66

Low Elevations Mid Elevations

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1 200 l I I I 1200 I I I I

E 1000 - ' E 1000 ' -
E 800 - - E -
E .5
= 600 ' ' = '
E .E’
z 400 — — g -
C C
w 200111.01 M i m i

0 it Am I

1940 1960 1980 2000 1940 1960 1980 2000
Years Years
High Elevations All Elevations
1200 ' ‘ ' 1200 ' ' '
E E 1000 ' -
E E 800 — -
E .5
= : SUD ' '
2 E
g 35 400 - -
C C
m m 200 WW‘
I I I 0 I I I I
1940 1960 1980 2000 1940 1960 1980 2000
Years Years

Figure 4.8. January snowfall in mm. A linear fit was calculated for 1931-2006. Results
are included for low elevations (< 500 111; top left), mid elevations (500 — 999 m; top
right), high elevations (2 1000 m; bottom left), and all elevations (bottom right).

67

4.3.2. April

April precipitation and snowfall increased at all elevations over the 1931-2006
period (Figures 4.9-4.10). Statistically insignificant increases of 1.71 mm (low
elevations), 0.80 mm (mid elevations), and 16.85 mm (high elevations) occurred for
precipitation, while 4.07 mm (low elevations), 9.32 mm (mid elevations), and 77.60 mm
(high elevations) occurred for snowfall. The high-elevation snowfall increase was

statistically significant at the p 5 0.10 level.

Low Elevations Mid Elevations

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

400 400
E E
300 - E 300 -
.E .E
C I:
3 200 - {-3 200 '
B B
E 100 E 100
0 - D -
9 e
D. [L
O I I A I J. O I I I I I
1940 1960 1980 2000 1940 1960 1980 2000
Years Years
High Elevations All Elevations
400 ' ' 400 2 ' '
.E .E 300 —
C C
.3 .3 200 -
B B
25 25
E E 100
D. l
0 I I I I 0 AL I A_ I I
1940 1960 1980 2000 1940 1960 1980 2000
Years Years

Figure 4.9. April precipitation in mm. A linear fit was calculated for 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (2 1000 m; bottom left), and all elevations (bottom right).

68

 

‘Iullli 111i.

 

 

Low Elevations Mid Elevations

 

 

 

 

 

 

 

 

 

 

 

 

 

 

500 ' 500 ' ' ' ' '
E 400- E 400 -
E E
.E 2500 - .E 300 -
e e
'3 200 - 3 200 -
2 2
U) 100 - - m 100 -
0 I I A AA I 0‘ IA ll ' . IA
1940 1960 1980 2000 1940 1960 1980 2000
Years Years
High Elevations All Elevations
500 ' ' SUD ' ‘ ' '
E 400 - - E 400 - .
E E
.E 500 - .E 500 - .
=0 g
“5 200 - .3 200 - -
8 8
U) 100 M ' U) 100 ' AA '
O I l A I O )\ _JA AA A I HAM
194-0 1960 1980 2000 1940 1960 1980 2000
Years Years /..

Figure 4.10. April snowfall in mm. A linear fit was calculated for 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 In; top
right), high elevations (Z 1000 m; bottom lefi), and all elevations (bottom right).

69

4. 3.3. July

No snowfall values were reported for July. July precipitation for 1931-2006
decreased at all elevations (Figure 4.11). Low-elevation stations dropped 20.14 mm (not
statistically significant). Meanwhile, mid elevations decreased by 32.40 mm (statistically
significant at the p S 0.10 level) and high elevations decreased by 43.36 mm (statistically

significant at the p S 0.05 level).

Low Elevations Mid Elevations

 

 

 

 

 

 

 

 

 

 

400 ' ' ' 400

E E
E 300 - E 300 -
.E .E
C C
.5—3 zoo- 3 200 '
B B
'a ‘a
'6 100 ‘5 100
93 2
Q Q.

0 I I I I O I I I I

1940 1960 1950 2000 1940 1960 19230 2000
Years Years
High Elevations All Elevations
400 ' ' ' 400 ' ' '

E E
E 500- E 300-
E .E
C I:
.3 200 .3 200-
B B
'a ‘a
'6 100 ‘6 100
9 2
D. D.

0 o

 

 

 

 

 

 

1940 1960 1950 2000 1940 1960 1980 2000
Years Years

Figure 4.11. July precipitation in mm. A linear fit was calculated for 1931-2006. Results

are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

70

4. 3. 4. October

Total precipitation over the study period for October (Figure 4.12) increased at

low elevations (11.88 mm), decreased at mid elevations (1.00 mm), and increased at high

elevations (8.55 mm). Snowfall totals (Figure 4.13) remained relatively steady with a

small increase at low elevations of 0.09 mm, and a decrease at mid and high elevations of

0.27 and 0.19 mm, respectively. None of the precipitation or snowfall values for October

were statistically significant.

Precipitation in mm

Precipitation in mm

400

 

400 7

 

 

Low Elevations

1940 1960 1980 2000
Years

High Elevations

Precipitation in mm

 

 

v—fiv

._I_

1940 1960 1950 2000

Yea rs

Precipitation in mm

 

Mid Elevations

 

400

300

200

100

  

 

 

 

1940 1960 1980 2000
Years

All Elevations

 

400 V'

 

 

 

194-0 1960 1950 2000
Years

Figure 4.12. October precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (2 1000 m; bottom left), and all elevations (bottom right).

Low Elevations Mid Elevations

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

500E ' I l I ' 1 I ' g 500% ' I r I j r I I V E

E 4003 5 5 4003 i
E E a E E i
.5 5003 j .5 3003 3
'5 2003 g -; 2003 3
2 3 s g a .
‘D 100% - (n 1003 -
05....-.5-5.5--..2 oi.-.----.---5...-a

1940 1960 1980 2000 1 940 1960 1950 2000
Years Years
High Elevations All Elevations

500E v I . v v y v v v v - u v y V g 500 fir vvvvvvvvvv fig

E 400i i 5 400 g
E a E
.5 5003 3 .5 300 3
1; 2003' ‘g “a? 200 q
8 a a g a
In 1003 h j to 100 g
OEALAAA .MAEA .AI A.AAI I A: 0 4 . __. I_. . . l . AAA 4 A:

1940 1960 1980 2000 1940 1960 1980 2000
Years Years

Figure 4.13. October snowfall in mm. A linear fit was calculated for 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (2 1000 m; bottom left), and all elevations (bottom right).

72

4. 3. 5. Annual

When considering the values after averaging all 12 months, total precipitation
(Figure 4.14) increased at all elevations over the study period (60.55 mm for low, 81.40
mm for mid, and 125.03 mm for high), although none of the values trends were
statistically significant. Snowfall (Figure 4.15) decreased at low elevations (52.59 m;
not statistically significant) and mid elevations (92.59 m; not statistically significant),
while increasing substantially at high elevations (561.76 mm; statistically significant at

the p S 0.01 level).

 

 

        

 

 

 

 

 

 

     

 

 

 

   

 

 

Low Elevations Mid Elevations
E 2400 ‘ E ‘
E 2200 - E -
5 2000 ‘ E '
§ 1600 - é -
,3 1500 - g —
% 1400 - % -
2 S
a 1200 - a_ _
1000 - - - 1000 - L 1 1
1940 1960 1980 2000 1940 1960 1980 2000
Years Years
High Elevations All Elevations
E 5 2400 -
E E 2200 -
E E- 2000 _
.5 5 1800 -
.3. ,3 1500 -
% % 1400 —
0 Q)
at L: 1200
1000 . . . . . . 1000 . . . . . .
1940 1960 1950 2000 1940 1960 1980 2000
Years Years

Figure 4.14. Annual values for precipitation in mm. A linear fit was calculated for 1931-
2006. Results are included for low elevations (< 500 m; top left), mid elevations (500 —
999 m; top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom
right).

73

Low Elevations Mid Elevations

 

 

2000 - - 2000 - 1

  

 

 

 

 

 

 

 

 

 

        

 

 

 

 

 

 

E E

E 1500- - E 1500- .
s 5

E 1000' - E. 1000- .
2 i
8 8

m 5003mm- m 500 i

0 A r k 1 i 0 J l 1 1
1940 1960 1980 2000 1940 1960 1950 2000
Years Years
High Elevations All Elevations

2000, ' 2000' "
E E

E 1500 - E 1500- -
.E .E

E 1000’ - E 1000- —
¥ 3
O O

1% 500, - (7', 500 .

o . . . . o . . . .
1940 1960 1980 2000 1940 1960 1950 2000
Years Years

Figure 4.15. Annual values for snowfall in mm. A linear fit was calculated for 1931-
2006. Results are included for low elevations (< 500 m; top left), mid elevations (500 —
999 m; top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom
right).

Trends during 1931-2006 suggest that total precipitation diminished in winter
months and the months of July and August. Annual precipitation values increased,
however, and during months spanning April through June and September through
November (Table 4.4 and Figures 4.7, 4.9, 4.11, 4.12, and 4.14). Annual snowfall totals
decreased at low and mid elevations, while increasing at high elevations (Table 4.5 and
Figures 4.8, 4.10, 4.13, and 4.15). Even at high elevations, snowfall decreased during
some autumn months, while a shift toward additional snowfall during spring seemed to
be the norm. When looking ahead, preliminary data for 2007 and 2008 suggest that much

of the southeastern US. continues to be in a dry period. In some cases, Figures 4.7-4.15

74

already display this trend at the end of the series, as values often dropped off during the
last year or two.

Intra-annual trends for 1931-2006 show that values of both precipitation and
snowfall generally increased with increasing elevation and the snow season spanned from
October until the following May (Figures 4.16-4.17). The annual snowfall maximum
was experienced during January for stations at low and mid elevations. For high-
elevation stations, the snowfall maximum was reached in February. In terms of total
precipitation, low- and mid-elevation stations had a maximum in March with a secondary
maximum in July. In contrast, high elevations had the largest maximum in July and a
secondary maximum in March. At all elevations, the driest month of the year was
October. A secondary precipitation minimum was present in April for low and mid
elevations. High elevations actually had a second minimum in February and a third in

April.

75

Low Elevations Mid Elevations

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

200.... ..... 200m
E . I E I .
150.- 4 E1501— -
.E . I .E . I
g - i C W
.3100:- 1 3100." 1
B - 1 B - l
'a t ‘ 'a : 1
'13 50- a ‘13 50' ‘
B C 93 z i
0.. [L -
0 . . . . . 0 . . . - . . - a .
123456789101112 123456789101112
Months Months
High Elevations All Elevations
2001 - - - - ' - - * ~ - . 200. - - - - ~ - .
E E i E i :
c . : 1: . .
.3100; : .3100: 1
3 - B 1 «
f9: 50: : Zé 50C 1
0 ' ‘ 0 ’ ‘
8 i ‘ 8 I :
D. - 1 0.. ' 1
o- . . - . . L L . . . 1 or . . - - a . L . . 1
123456789101112 123456789101112
Months Months

Figure 4.16. Intra-annual trends of total precipitation in mm versus months of the year.
Values were averaged over the entire 1931-2006 period. Results are included for low
elevations (< 500 m; top left), mid elevations (500 — 999 m; top right), high elevations (_>_
1000 m; bottom left), and all elevations (bottom right).

76

Low Elevations Mid Elevations

 

 

 

 

 

 

 

 

 

   

 

 

 

 

 

I500 ' ' ' r ‘ - ' '1 500: ‘ ' ' *
: 1 :
E 2501‘ ‘1 E 250;"
E 200:- ~1 E 200:-
.E : 1 .E :
=150I- : =150C~
.2 t : .2 :
3100; -, 3100:
E : : E :
m 50E\ i w 50x
o: . . . . . . . . M 0: . L L . . . L 4
123456789101112 123456789101112
Months Months
High Elevations All Elevations
300;**' 300,"*'*'~*-
E 250. -3 E 250.5
E 200: -1 E 200;»
.E : ‘ .E l
:150: =150Z'
E : .E’ :
3100; 3100;
C : C _
V’ 50; °° 50\
0’ . . . . L . L . 019 . 2 . . . L
123456789101112 12.314515789101112
Months Months

Figure 4.17. Intra-annual trends of total snowfall in mm versus months of the year.
Values were averaged over the entire 1931-2006 period. Results are included for low
elevations (< 500 m; top left), mid elevations (500 - 999 m; top right), high elevations (Z
1000 m; bottom left), and all elevations (bottom right).
4.4. Discussion
In general, these results support Kéippen's classification of this region as humid
subtropical (Cfa). As a whole, the region experienced relatively hot and humid summers
and mild winters (i.e., average temperature of the coldest month was below 18°C, and
above -3°C). Annual precipitation generally fell within Koppen's specified range of 800
to 1650 mm. However, the results here were more similar to those found by Gaffin and
Hotz (2000) in that the annual totals of precipitation I found were usually 1000 mm or

greater even at low-elevation stations. Furthermore, in agreement with the previous

studies discussed in Chapter 2 (e.g., Whittaker, 1952; Smallshaw, 1953; Shanks, 1954;

77

Mark, 1958; Bogucki, 1972; Hicks, 1979; Gafiin and Hotz, 2000), the results here also
suggest that precipitation totals tended have a positive relationship with elevation.
Across all elevations, the annual range in precipitation generally ranged from 1000 to
1700 mm. However, local-scale topographical variations in elevation, slope, shape, and
aspect likely contributed to some values exceeding these averages. For instance, some
high-elevation stations had precipitation totals of 2000 mm or more. Similar ranges and
values were found in previous studies by Whittaker (1952), Smallshaw (1953), Shanks
(1954), Mark (1958), Bogucki, (1972), Hicks (1979), and Gaffin and Hotz (2000).

Whittaker (1952) found that the annual precipitation peaks occurred during late
winter/early spring and late summer, while the minima occurred in late spring and early
autumn. Mark (195 8) reported that dry periods usually occurred during late spring and
autumn months. Gaflin and Hotz (2000) found spring and autumn to be the driest
seasons while summer was the wettest season. Although Whittaker (1952), Mark (195 8),
and Gaffin and Hotz (2000) did not specify exact months, their broad results do agree
with the intra-annual trends I illustrated in Figure 4.16. Diem's (2006) results show that
dry periods during summer months in GA spanned the mid 19503, the mid 19703 through
the mid 19803, and again during the late 1990s, while wet periods spanned the 19603 and
early 19903. My July maps in Figure 4.1] support those results. In addition, my results
illustrated that wet summer periods were present from roughly the mid 19303 through
1950, the 19603, and during the early 20003. Observations at the very end of the series in
Figure 4.11 and preliminary data not shown suggest that the region is currently in a dry
period.

The only previous study to discuss snowfall was Mark (1958). Much like my

78

results in Figure 4.17, he found that snowfall began in October and continued through
May. His results relied on the three stations of Asheville, NC (671 m), Banner Elk, NC
(1131 m), and Mount Mitchell, NC (2022 m), to represent low, mid, and high elevations.
Mark's (1958) results showed that maximum snowfall occurred in February for Asheville
(79 mm), January for Banner Elk (224 mm), and March for Mount Mitchell (378 mm).
In contrast, I found that both low- and mid-elevation stations generally received peak
snowfall during January, while high-elevations stations peaked in February (Figure 4.17),
and the average snowfall values I found were somewhat lower (53 mm for low, 82 mm
for mid, and 247 mm for high). However, the representative stations that Mark (195 8)
used do not coincide exactly with the elevation categories I specified. Following my
scheme, Asheville, NC, would be a mid-elevation station, whereas both Banner Elk, NC,
and Mount Mitchell, NC, would both be in the high-elevation category. These
differences probably account for at least some of the variation in snowfall totals between
our respective studies.

As was the case in previous studies (e.g., Shanks, 1954; Mark, 1958; Hicks, 1979;
Bolstad et al., 1998), I found temperatures to decrease with increasing elevation (Figures
4146). In agreement with Shanks (1954) and Mark (1958), I found July to be the
warmest month at all elevations (Figure 4.6). I also found January to be the coolest
month for all elevations. In contrast, Shanks (1954) found the coolest months at low
elevations to be November and December and March to be coldest for high elevations,
but Mark (1958) found the coldest temperatures for high elevations during February. In
general, I found that minimum, maximum, and mean temperatures cooled overall during

the I931-2006 period, but warming did occur throughout the last few decades. My

79

temperature values also suggest that the diurnal temperature range decreased, which is in
agreement with results found by others in global studies (Karl et al., 1993; Easterling et
al., 1997; IPCC, 2007a). In addition, my results show that cool-season temperatures
warmed more than warm-season temperatures during the past few decades, which
coincides with results from the USGCRP's 2009 report for the southeastern US. This
result suggests that the range of the annual temperature cycle decreased. Other studies by
Balling et a1. (1998), the IPCC (2007a), and Stine et a1. (2009) found this same general
trend occurring globally.

In reference to Question 1, my results thus far have generally supported the
findings reported by others, but some small differences are present. For instance, my
precipitation values were somewhat higher than those classified by the Koppen
classification for the region. I was able to be more specific than Whittaker (1952), Mark
(1958), and Gaffin and Hotz (2000) in pinpointing the timing of annual precipitation
maxima and minima. My snowfall values were somewhat lower than Mark's (195 8), but
this result may be explained by his inclusion of only three stations. Although my data
agree with others regarding the timing of the annual temperature maximum (July), my
minimum temperature results did differ from others (Shanks, 1954; Mark, 1958). My
speculation is that these differences result from the fewer stations and shorter study
periods utilized in earlier studies. My results show that temperature values generally had
an inverse relationship with elevation, whereas precipitation and snowfall totals had a
positive relationship with elevation.

In addition to building upon past studies, my study provides new information as

well to assist with answering Question 1. Diurnal and annual temperature ranges in the

80

region were found to be shrinking over the study period. Total precipitation increased
over the study period. Snowfall totals decreased at low and mid elevations during most
months. Snowfall totals increased at high elevations, where a decrease in snow totals

during some autumn months were offset by a shift toward increased springtime snowfall.

81

CHAPTER 5

RESULTS AND DISCUSSION OF THE SPATIAL ANALYSIS

5.1. Overview

Table 5.1 shows the number of stations for each month (and annually) within each
overlapping decade when at least 90 percent of the data were available for all variables
(minimum, maximum, and mean temperature, total precipitation, and total snowfall).
Few stations were available until after 1930. Although 463 monthly stations were used
throughout the study, during the 1930-1939 decade 30 or fewer stations frequently
reported data. In general, with each passing decade the number increased and in the final
decade (1995-2004), the number of stations surpassed 70. Some noticeable exceptions to
this trend were January and February during the 1955-1964 and 1960-1969 decades when
fewer than 30 stations were available. During those decades, fewer stations reported
snowfall totals and no known technological changes in snowfall measurement during that

time have been documented.

82

Table 5.1. Number of stations with at least 90 percent of data available for minimum,
maximum, mean temperature, total precipitation, and total snowfall.

 

 

 

 

.__ P909919" FebatA2_M.XLJJ.L19|-Aus.53913903331,999L159.
1390-1399ooooooooooooo
1395-19040000000000000
1900-19091100011100000
1905-19141100011000010
191019191100010000000
191519241011010111000
192019290011111111100
192549340111111111110
1930-1939 29 3o 27 27 23 23 29 26 23 25 29 26 11
1935-1944 33 42 41 37 36 39 33 4o 33 33 39 35 24
1940-1949 37 39 41 39 33 39 33 39 34 35 37 32 26
1945-1954 40 33 39 44 44 43 45 46 43 45 37 35 24
1950-1959 37 37 46 55 53 55 55 55 54 54 44 34 21
l955-l964 20 22 34 54 53 54 57 59 60 59 53 23 13
1960-1969 22 23 33 66 64 64 64 66 66 65 53 32 15
1965-1974 32 44 so 75 77 73 75 76 76 74 6o 51 23
1970-1979 41 42 52 69 70 7o 69 63 63 67 57 52 31
1975-1934 44 41 55 67 67 66 65 66 65 66 64 53 32
1930-1939 47 43 66 68 67 67 67 67 63 63 67 61 34
1935—1994 65 67 71 72 7o 72 72 72 71 71 7o 66 55
1990-1999 71 71 7o 68 72 71 72 7o 70 69 72 69 61
199572004,__72_ 74 _73 72 72 71 f7o_ 7o 74 ._74_ 75,73” 64_

l
1

My initial intent for using lO-year overlapping periods was to not only study the
stations in a spatial context, but also to determine whether any temporal change or
variability was noticeable amongst the clusters over time. However, because the overall
number and location of available stations changed substantially during the study period,
the number and location of stations were probably the largest factors creating noticeable
change amongst the clusters. For instance, few stations existed during early years of the
study (the northwestern portion of the study area in particular) and inclusion of additional
stations in later decades likely affected the clusters.

Cluster results for minimum, maximum, and mean temperature were quite similar,

so only mean temperature is presented here. The 5-cluster result is shown and the 1935-

83

1944, 1965-1974, and 1995-2004 decades are used to represent periods near the
beginning, middle, and end of the record. January and July maps are shown here to
represent cool and warm season clusters. April and October clusters for representing the
transition seasons are not discussed in this chapter, but the maps for those months are
included in Appendix B. Results within this chapter yield spatial information to assist in
answering Questions 1 and 2 from Section 2.5.

During the analysis, stations were ranked based on the “strength” of the
corresponding cluster. When referring to the strongest cluster, I mean that stations within
that cluster were statistically most similar. On the maps, stations in Cluster 1 (i.e., the
strongest cluster) are depicted as a black dot. Stations in Cluster 2 are very dark gray
plus signs. Stations in Cluster 3 are dark gray squares. Those in Cluster 4 are light gray
triangles, and those in Cluster 5 (i.e., the weakest cluster) are very light gray diamonds.
Although cluster strength is important, its evaluation is more crucial in studies using
many clusters. Since I am only presenting five clusters in this study, in most cases all of
the clusters are important, so I will not overly emphasize cluster strength or rank. As I
will show, some groups of stations tended to cluster together frequently, but did not
necessarily always cluster together at the same rank each time (e.g., in one time period a
group of stations may be within Cluster 1 and in another time period the same stations

may be within Cluster 3).

5.2. Spatial Results of Temperature

5. 2. I . January

During 1935-1944, 38 stations were available (Figure 5.1). Results for most maps

84

suggest cluster relationships with latitude and/or elevation. Cluster 3 (square) was
located in the southeast and low-lying Piedmont portion of the area and was the warmest
(612°C) cluster. Cluster 2 (plus sign) was at slightly higher elevations mostly along the
southeast-facing side of the Blue Ridge (5.10°C). Cluster 4 (triangle) stations were
comprised of the mid-elevation stations and had mean temperatures of 3.71°C. Cluster 5
(diamond) stations were located at mid-to-high elevations and had a lower temperature
(1.87°C). Lastly, Cluster 1 (dot) consisted of one station at a high elevation that had the

lowest temperature (-3.04°C).

 

Elevation (m)

36N '

 

35N

 

 

 

 

Figure 5.1. January during 1935-1944 with 5 clusters for mean temperature, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Elevation in meters is shaded in gray.

Thirty-two stations were available during the 1965-1974 period (Figure 5.2). In

the southeast portion, Cluster 2 was the warmest (6.16°C). Cluster 1 had some stations at

85

slightly higher elevations along the Blue Ridge (474°C). Cluster 4 was comprised of
mostly low- and mid-level stations (3.07°C). Cluster 3 stations were located at higher
elevations (088°C) in the northeast portion of the study area, and Cluster 5 stations were

generally at the highest elevations (-0.80°C).

    
 

 

Elevation (m)

61“.
.41

t

  

.‘ ;.,|,1 a. H1.

36N : E" ........... V ‘ I

 

., u 0
35,, ., .......... 1

 

 

 

84W 83W 82W

Figure 5.2. January during 1965-1974 with 5 clusters for mean temperature, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Elevation in meters is shaded in gray.

For the 1995-2004 period, 72 stations were available (Figure 5.3). Cluster 2 was
the warmest cluster (585°C) and was generally located in the southeast and lowest-lying
areas. Cluster 1 was also mostly scattered along the southeast side of the Blue Ridge at
slightly higher elevations (4.40°C). Cluster 3 was a large cluster comprised of more
northerly and mid-level stations (300°C). Cluster 5 was also northerly with mid- to high-

level stations (126°C). Lastly, Cluster 4 consisted of only two stations, but they were

86

high-elevation stations (-2.60°C).

 

Elevation (m)

   

 

 

 

84W 83W 82W

Figure 5.3. January during 1995-2004 with 5 clusters for mean temperature, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Elevation in meters is shaded in gray.

5. 2. 2. July

During 1935-1944 for July, 38 stations were available (Figure 5.4). In this case,
Cluster 1 was the coldest cluster (15.09°C) and included one high-elevation station.
Cluster 4 was the second coolest cluster (20.89°C) and consisted of mid- to high-
elevation stations. Mid-level stations were included in Cluster 2 and were somewhat
warmer (23.00°C). Stations in low- to mid-elevation areas were in Cluster 3 (24.67°C).
The warmest stations (26.04°C) were in Cluster 5. Most of these stations were located in
the southeast area, but a few others lay at low elevations on the northwest side of the Blue

Ridge.

87

 

Elevation (m)

   

 

 

 

. _ : ('5
84W 83W 82W

Figure 5.4. July during 1935-1944 with 5 clusters for mean temperature , where Cluster 1
(dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Elevation in meters is shaded in gray.

Seventy-five stations reported during the 1965-1974 decade (Figure 5.5). The
warmest cluster (25.38°C) was Cluster 3, with all stations at low elevations and most
located to the southeast. Cluster 5 was also a warm cluster (23.95°C) and was comprised
of stations at low and mid elevations on both sides of the Blue Ridge. Cluster 1 included
mid-level and northerly stations (22.60°C), whereas Cluster 2 included mid- and high-
elevation stations (21 .51°C). The coolest cluster was Cluster 4, where temperatures

averaged 19.13°C. Stations in Cluster 4 were generally at high elevations.

88

 

Elevation (m)

 

 

 

 

 

A.
Figure 5.5. July during 1965-1974 with 5 clusters for mean temperature, where Cluster 1
(dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Elevation in meters is shaded in gray.

For 1995-2004, 70 stations were available for July (Figure 5.6). The warm cluster
(26.12°C), consisting of stations to the southeast and low elevations, was Cluster 4. Low-
and mid-elevation stations were included in Cluster 3 and had an average temperature of
24.65°C. Stations at mid and high locations were grouped in Cluster 2 and were

somewhat cooler (22.86°C). High-elevation stations were grouped into Clusters 1 and 5,

with temperatures of 20.45°C and 16.03°C, respectively.

89

 

Elevation (m)

 

 

 

 

 

84W 83W 82W

Figure 5.6. July during 1995-2004 with 5 clusters for mean temperature, where Cluster 1
(dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Elevation in meters is shaded in gray.

5.3. Spatial Results of Precipitation
5. 3. 1 January

Unlike temperature, cluster patterns expressed in terms of precipitation tended to
be less regionally homogeneous, which supports previous literature (Root and Schneider,
2001). For January during 1935-1944, there were 38 stations present (Figure 5.7).
Cluster 1 had the lowest precipitation (90.49 mm). The cluster was located near the TN
and NC border in low and mid elevations. Cluster 5 also had low precipitation (110.23
mm), but had stations scattered at varying locations and elevations. Cluster 3 had stations
in low and mid elevations (129.29 mm). Cluster 2 had more precipitation (149.15 mm)
and was mostly located on the south side of the Blue Ridge at low and mid elevations.

Highest precipitation totals were associated with Cluster 4 (183.50 mm) at four stations

90

near the union of the NC, SC, and GA borders located on the south side of the Blue Ridge

at mid elevations.

 

Elevation (m)

36N g} '

 

35N

 

 

 

 

Figure 5.7. January during 1935-1944 with 5 clusters for total precipitation, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Elevation in meters is shaded in gray.
Thirty-two stations were reporting during January 1965-1974 (Figure 5.8). The

driest locations were associated with Cluster 5 (77.37 mm). Four of the stations in that
cluster were located within a low-lying valley that cuts through the Blue Ridge in a
northwest to southeast direction toward Asheville, NC. This valley was conveniently
used for the placement of Interstate-40 and ultimately connects Knoxville, TN, and
Asheville, NC. This “I-40 corridor” generally received less precipitation than the much
higher elevations to its southwest and northeast. Cluster 2 was comprised of mid- and
high-elevation stations mostly at northerly locations and was also dry (96.06 mm).

Cluster 3 included some southeastern stations at low elevations, two stations in the Ridge

91

and Valley province and two stations at higher elevations (110.60 mm). Some low and
mid stations were in Cluster 4 (131.06 mm). Lastly, Cluster 1 had the highest

precipitation total (153.81 mm) and was located in northern GA and near the SC and NC

borders.

Elevation (m)

 

 

 

 

 

84W 83W 82W

Figure 5.8. January during 1965-1974 with 5 clusters for total precipitation, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Elevation in meters is shaded in gray.

During 1995-2004, 72 stations were available for January (Figure 5.9). Clusters 1
and 3 had the lowest precipitation totals (107.62 mm and 124.62 mm) and were located in
low-lying northerly locations, along the I-40 corridor, and at low—lying southeastern
locations. Cluster 2 was scattered at locations on both sides of the Blue Ridge (142.95
mm). Higher totals were present for Cluster 4 (161.50 mm) at mid-elevations locations
along the Blue Ridge. The highest totals were associated with Cluster 5 (196.02 m),

where most stations were located in northeast GA and southwest NC.

92

 

Elevation (111)

 

 

 

 

 

Figure 5.9. January during 1995-2004 with 5 clusters for total precipitation, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Elevation in meters is shaded in gray.

5. 3. 2. July

During 1935—1944 there were 38 stations reporting for July (Figure 5.10). The
driest locations were included in Cluster I mostly at low elevations (129.48 mm). Cluster
3 was mostly at mid elevations (149.67 mm), whereas Cluster 2 stations were at mid and
high elevations (176.19 mm). Cluster 4 had two stations at mid-to-high elevations
(210.19 mm) and Cluster 5 had three stations (249.43 mm). Two stations from Cluster 5
were located near the union of NC, SC, and GA and the third was located at a higher

elevation.

93

 

Elevation (m)

 

 

 

 

 

Figure 5.10. July during 1935-1944 with 5 clusters for total precipitation, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Elevation in meters is shaded in gray.

Lower precipitation values were reported at 75 stations during 1965-1974 (Figure
5.11). In this case, the lowest values were included in Cluster 3 (96.01 mm) along the I-
40 corridor and in the southeastern portion of the region. Low values were also present
in Cluster 1 (127.76 mm) and Cluster 4 (115.29 mm) at low and mid elevations. Cluster
2 was also at low and mid elevations, but had somewhat higher precipitation totals

(144.13 mm). The wettest region was Cluster 5 (166.92 mm) and mostly included

stations in southwest NC and northeast GA.

94

Elevation (m)

 

  

 

 

84W 83W 82W

Figure 5.11. July during 1965-1974 with 5 clusters for total precipitation, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Elevation in meters is shaded in gray.

Seventy stations reported during July for the 1995-2004 period (Figure 5.12).
Cluster 4 was the driest cluster (94.24 mm) with stations located in the low-lying
southeast portion of the area and some stations located at mid-to-high elevations along
the Blue Ridge. Stations within Cluster 2 (107.39 mm) and Cluster 3 (121.45 mm) were
mostly scattered at low and mid elevations. Cluster 1 had stations along the Blue Ridge
and also within the Ridge and Valley province (138.17 mm). Cluster 5 had the highest

totals (167.30 mm), but also had stations scattered in along the Blue Ridge and the Ridge

and Valley province.

95

 

Elevation (m)

  

 

 

 

Figure 5.12. July during 1995-2004 with 5 clusters for total precipitation, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Elevation in meters is shaded in gray.
5.4. Spatial Results of Snowfall
5.4.1. January

During 1935-1944, 38 stations reported snowfall (Figure 5.13). Stations that
received the lowest totals (36.61 mm) were located either in the southeastern portion of
the region or within the 1-40 corridor and were included in Cluster 2. Cluster 3 included
stations at low and mid elevations (56.27 mm). Stations mostly at mid elevations of the
Blue Ridge were in Cluster 4 (75.81 mm). The highest values were generally reported at

the highest elevations within Cluster 1 (258.28 mm) and Cluster 5 (127.04 mm).

96

Elevation (m)

36N '

35N

 

 

 

 

84W 83W 82W

Figure 5.13. January during 1935-1944 with 5 clusters for total snowfall, where Cluster 1
(dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Elevation in meters is shaded in gray.

Somewhat higher values were reported from 75 stations during 1965-1974 (Figure
5.14). Cluster 1 included the lowest values (54.18 mm). Stations within that cluster were
all southern stations in either GA or SC. Cluster 5 and Cluster 2 had much higher values
(102.41 mm and 139.21 mm) and included stations at mid elevations along the I-40
corridor and two stations in the Ridge and Valley province. Cluster 4 included stations at
northern locations and mid-to-high elevations (187.92 mm). The highest totals were

reported in Cluster 3 (289.50 mm) and included four northeast stations at high elevations.

97

 

Elevation (m)

36N * ‘J

 

35N

 

 

 

 

Figure 5.14. January during 1965-1974 with 5 clusters for total snowfall, where Cluster 1
(dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Elevation in meters is shaded in gray.

Seventy stations reported snowfall during January for 1995-2004 (Figure 5.15).
Small totals were reported in Cluster 1 (9.58 mm). Stations in that cluster were mostly at
low elevations and/or at southern locations. Somewhat more snowfall was received by
Cluster 2 stations (61.01 mm). Many of those stations were also located at low
elevations, although a large percentage was located in the Ridge and Valley province of
TN. Cluster 3 was comprised of northern stations and stations along the I-40 corridor
(113.00 mm). The highest totals were reported by stations at high elevations and
northeast locations. Those stations were included in Cluster 4 (239.02 mm) and Cluster 5

(516.51 mm).

98

 

Elevation (m)

   

 

 

 

84W 83W 82W

/
Figure 5.15. January during 1995-2004 with 5 clusters for total snowfall, where Cluster 1
(dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Elevation in meters is shaded in gray.
5.5. Spatial Results for All Variables
5. 5. 1. January
Clusters based on all variables (minimum, maximum, and mean temperature, total
precipitation, and total snowfall) were also considered. Cluster 2 was the coldest cluster
(-3.59, 2.30, 819°C for minimum, mean, and maximum temperature) and largely
included stations along the Blue Ridge (Figure 5.16). Cluster 2 also had precipitation and
snowfall values of 115.13 and 93.41 mm. Cluster 5 consisted of high-elevation stations
and was also cool (-2.76, 2.76, 826°C for minimum, mean, and maximum temperature),
but also had the highest precipitation (147.37 mm) and snowfall (121.82 m) values.
Cluster 3 stations were mostly located near the I-40 corridor and had the lowest

precipitation values (-2.16, 3.93, 10.03°C for minimum, mean, and maximum

99

temperature, 95.78 mm for precipitation, and 66.88 mm for snowfall). Stations in Cluster
4 were at low and mid elevations and had relatively warm and wet conditions (-1.34,
4.95, and 11.25°C for minimum, mean, and maximum temperature, 142.28 mm for
precipitation, and 61.92 mm for snowfall). Lastly, Cluster 1 included southerly stations
at low elevations. Cluster 1 was the warmest of the clusters (0.54, 5.96, and 11.3 8°C for
minimum, mean, and maximum temperature), and had moderate precipitation (141.12

mm), but low snowfall receipt (47.71 mm).

 

Elevation (m)

36N '

 

35N

 

 

 

 

84W 83W 82W

Figure 5.16. January during 1935-1944 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation, and total
snowfall), where Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4

(triangle), and Cluster 5 (diamond) represent station locations. Elevation in meters is
shaded in gray.

January clusters during 1965-1974 did not express any strong spatial patterns as
all clusters had stations at varying locations and elevations (Figure 5.17). Cluster 1 had

northerly stations with relatively cool temperatures and the highest snowfall total (-2.81,

100

2.53, and 784°C for minimum, mean, and maximum temperature, 100.32 mm for
precipitation, and 201.88 mm for snowfall). Cluster 3 had the highest precipitation and
lowest snowfall values (—1.86, 3.66, and 916°C for minimum, mean, and maximum
temperature, 139.40 mm for precipitation, and 94.30 mm for snowfall). Cluster 4 had the
highest mean and maximum temperatures (-2.22, 4.10, and 10.39°C for minimum, mean,
and maximum temperature, 104.20 mm for precipitation, and 131.28 mm for snowfall).
Cluster 2 had several stations in the I-40 corridor and the lowest precipitation value
(-1.76, 3.98, and 969°C for minimum, mean, and maximum temperature, 87.84 mm for
precipitation, and 103.46 mm for snowfall). The lowest minimum temperatures were
associated with Cluster 5 (-3.24, 2.80, and 881°C for minimum, mean, and maximum

temperature, 124.52 mm for precipitation, and 139.52 mm for snowfall).

101

 

36N _" '

35N

 

 

 

 

Figure 5.17. January during 1965-1974 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation, and total
snowfall), where Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4
(triangle), and Cluster 5 (diamond) represent station locations. Elevation in meters is
shaded in gray.

The 1995-2004 decade for January was comprised of five distinct climate regimes
(Figure 5.18). Stations in Cluster 1 were all southeastern stations at low elevations within
the Piedmont. This was the warmest cluster (-1.11, 5.35, and 11.77°C for minimum,
mean, and maximum temperature). Cluster 1 was also relatively dry (120.30 mm for
precipitation and 34.84 mm for snowfall). Cluster 2 mostly included stations along the I-
40 corridor and stations to the north of that area. Overall, Cluster 2 was the driest cluster
(110.04 mm for precipitation), although it did experience the second highest snowfall
totals (73.41 mm). Cluster 3 was located near the NC, SC, and GA borders and had the
highest precipitation total (178.39 mm), but lowest snowfall total (25.80 mm). Stations

in Cluster 4 were located at low and mid elevations on the north and south sides of the

102

Blue Ridge. Moderate precipitation (139.53 mm) and low snowfall (48.12 mm) were
associated with Cluster 4. Clusters 2-4 were close in terms of temperature, having
minimum temperatures of -2.98, -3.07, and 245°C, mean temperatures of 3.01 , 3.32, and
308°C, and maximum temperatures of 8.98, 9.68, and 859°C, respectively. Lastly,
Cluster 5 mostly included stations at northerly and high-elevation locations. This regime
was the coldest (-4.49, 1.48, and 743°C for minimum, mean, and maximum
temperature), received moderate precipitation (134.87 mm), and had the highest snowfall

totals (183.41 mm).

 

Elevation (m)

 

 

 

 

 

84W 83W 82W

1,”.

Figure 5.18. January during 1995-2004 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation, and total
snowfall), where Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4
(triangle), and Cluster 5 (diamond) represent station locations. Elevation in meters is
shaded in gray.

5.5.2. July

Since snowfall was not reported during July, clusters were based on the other four

103

variables. During July 1935-1944 (Figure 5.19) the coolest and wettest cluster was
Cluster 5 (14.92, 20.15, and 25.38°C for minimum, mean, and maximum temperature,
and 218.51 mm for precipitation). Stations in that cluster were all located along the Blue
Ridge and ofien at high elevations. Cluster 2 was also cool (16.11, 22.71, and 29.32°C
for minimum, mean, and maximum temperature), but had the lowest precipitation totals
(146.67 mm). Cluster 1 (18.97, 25.23, and 31.49°C for minimum, mean, and maximum
temperature, and 152.37 mm for precipitation), Cluster 3 (18.72, 24.68, and 30.55°C for
minimum, mean, and maximum temperature, and 159.17 mm for precipitation), and
Cluster 4 (19.21, 25.48, and 31 .76°C for minimum, mean, and maximum temperature,
and 157.30 mm for precipitation) were all located at mostly low elevations and had

similar temperature and precipitation values.

 

Elevation (m)

33N ,.

 

35N

 

 

 

84W 83W 82W

 

Figure 5.19. July during 1935-1944 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation), where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Elevation in meters is shaded in gray.

104

Station coverage for July during 1965-1974 (Figure 5.20) was much better
compared to the 1935-1944 period (Figure 5.19). The wettest cluster was Cluster 1
(15.85, 22.38, and 28.90°C for minimum, mean, and maximum temperature, and 161.84
mm for precipitation), where many of the stations were located near the intersection of
NC, SC, and GA. The driest cluster was Cluster 5 (17.76, 23.12, and 28.46°C for
minimum, mean, and maximum temperature, and 118.96 mm for precipitation), although
the stations were scattered between high and low elevations. Cluster 4 was mostly
comprised of stations along the Blue Ridge and had the lowest minimum and mean
temperatures (15.47, 22.07, and 28.64°C for minimum, mean, and maximum temperature,
and 124.87 mm for precipitation). Clusters 2 and 3 were the warmest clusters and were
located primarily in the lowest-lying areas. Cluster 2 had a minimum, mean, and
maximum temperature of 18.15, 24.53, and 30.89°C, with a precipitation value of 128.88
mm. Meanwhile, Cluster 3 had minimum, mean, and maximum temperatures of 17.86,

23.99, and 17.86°C, and a precipitation value of 146.43 mm.

105

 

 

 

 

 

Figure 5.20. July during 1965-1974 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation), where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Elevation in meters is shaded in gray.

Regimes for July (Figure 5.21) during 1995-2004 were less distinct than was the
case for January (Figure 5.18). Cluster 1 consisted of stations at varying locations, but
most were at low and mid elevations. These locations had moderate temperatures (17.59,
23.81, 29.98°C for minimum, mean, and maximum temperature) and the second highest
precipitation receipt (128.72 mm). Cluster 2 stations were located at low elevations on
both sides of the Blue Ridge, had the warmest mean and maximum temperatures (18.59,
25.01, and 31.40°C for minimum, mean, and maximum temperature), and lower
precipitation (119.94 mm). Cluster 3 stations were at low and mid elevations. Minimum,
mean, and maximum temperatures at these locations were 19.34, 24.96, and 30.55°C,

while precipitation was the highest of all clusters (133.09 mm). Cluster 4 stations were

106

located at mid elevations, had cooler temperatures (17.01 , 23.13, and 29.48°C for
minimum, mean, and maximum temperature), and the lowest precipitation total (107.71
mm). Lastly, Cluster 5 stations were generally located at the highest elevations. While
the precipitation total was not the highest (127.55 mm), the cluster was the coolest

(15.57, 21.13, and 26.65°C for minimum, mean, and maximum temperature).

 

Elevation (m)

 

 

 

 

84W 83W 82W

Figure 5.21. July during 1995-2004 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation), where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Elevation in meters is shaded in gray.

5.6. Discussion
Results in this chapter furnish information to help answer Questions 1 and 2. The
clusters for all temperature variables and months were closely associated with elevation,

and to a lesser extent, latitude. This finding is further supported by the results I presented

in Figure 4.6 of Chapter 4 and by previous studies (Shanks, 1954; Mark, 1958; Hicks,

107

1979; Bolstad et al., 1998). Stations in the warmest regime were located within the
Piedmont province of SC at lower elevations. At times the regime also included low-
elevation stations within the Ridge and Valley province of TN. A second regime closely
followed the base of the Blue Ridge on both the southeast and northwest sides. A third
regime was often located at mid elevations, a fourth was located at mid-to-high
elevations, and the fifth (and coldest) regime was located at the highest elevations along
the Blue Ridge. Often the highest elevation regime was comprised of one or more
stations in the northeast portion of the study region, despite the presence of similar high
elevations in the central portion of the study region (in the area of GSMNP and its
neighboring national forests). This result may occur because the federal lands and the
surrounding area do not appear to be as well represented by stations in the high elevations
(especially during the early portions of the record), perhaps due to the underdevelopment
of the region.

Widespread snowfall was generally received only during winter months. Some
snowfall was received during spring and autumn months but was primarily limited to the
highest elevations, and cluster results for those months generally reflected that (see
Appendix B). Regimes for snowfall were less distinct than those for temperature.
However, snowfall totals were usually greatest at the high elevations and lowest in
southern and low-elevation locations.

Although my results in Chapter 4 and the results of others (e.g., Whittaker, 1952;
Smallshaw, 1953; Shanks, 1954; Mark, 1958; Bogucki, 1972; Hicks, 1979; Gaffin and
Hotz, 2000) generally showed that total precipitation varies directly with elevation, that

relationship was not always obvious from the spatial patterns displayed by the cluster

108

maps (Figures 5.7-5.12) and may be the result of windward and leeward topographic
properties at the higher elevations. Three consistent regimes for precipitation did appear,
however, and tended to be more visible during January than July. The first was a
maximum located near the union of the NC, SC, and GA borders. Elevation in that area
increases sharply from south to north. Any southerly airflow over that region could result
in abrupt orographic lifting and precipitation. Gaffin and Hotz (2000) found a similar
precipitation maximum in that region that persisted throughout the year (Figures 5.22-
5.23). I also found a second maximum located at high-elevation, northeastern stations or
along the northwestern side of the Blue Ridge. A minimum generally occurred along the
[-40 corridor and at stations to the north of that corridor. Secondary minimum values also
occurred at low-lying areas both in the Piedmont province and areas in the northwestern

part of the region. These additional maxima and minima were also broadly confirmed by

 

Gaffin and Hotz (2000).
1'0'1'5 {6'2 “762 s
.1016
. 7e2'
. .. 15.24 .' .' 1016
127i) 2." -' .
' '» 1.2.7.0

 

 

 

 

Figure 5.22. Precipitation climatology for January during 1961-1990. Values are
reported in mm (adapted from Gaffin and Hotz, 2000; p. 6).

109

 

i524
.. '_ 152.4
152.4. ' E1016"

. ' '. ' 177.8
,,12_7.0. 4524-,

 

 

 

Figure 5.23. Precipitation climatology for July during 1961-1990. Values are reported in
mm (adapted from Gaffin and Hotz, 2000; p. 7).

Actual precipitation values calculated in my study were comparable to those
reported by Gafiin and Hotz (2000) as well. For January, they found a minimum of 3
inches (76.2 mm) and a maximum of 6 inches (152.4 mm). July values ranged between 4
and 8 inches (101.6 to 177.8 mm). My 1965-1974 decade overlapped with their study
and used the same COOP dataset, and my January (77.4 to 153.8 mm) and July (96.0 to
166.9 mm) values were quite similar. My values during 1935-1944 for January (90.5 to
183.5 mm) and July (129.5 to 249.4 mm), as well as January for 1995-2004 (107.6 to
196.0 mm), were somewhat greater than those reported by Gaffin and Hotz (2000).
However, my July values during 1995-2004 were similar (94.2 to 167.3 mm).

Although cluster analysis has been used in previous climate studies that included
the southern Appalachian region (e.g., Stooksbury and Michaels, 1991; Winkler, 1992;
Fovell and Fovell, 1993; Gong and Richman, 1995; Fovell, 1997), all of those studies
considered much larger areas. As a result, the southern Appalachian region was usually
clustered into four or fewer clusters. The cluster work of Stooksbury and Michaels

(1991) was most relevant in that they focused on the southeast U.S. Using temperature

110

and precipitation, the authors found four clusters for the southern Appalachians: Western
Appalachian & Valleys (included eastern TN and northern GA), Eastern Appalachian
(included western edge of NC along the Blue Ridge), Southside Virginia & Western
North Carolina (included western NC), and Interior South Atlantic (included portions of
NC, SC, and GA). Unfortunately, Stooksbury and Michaels (1991) did not discuss the
differences between the clusters in regards to the temperature and precipitation
characteristics. In contrast, my maps presented in this chapter illustrated cluster results
for the region, backed by the temperature, precipitation, and snowfall values to allow
further comparison of the regimes.

In summary, my results presented in Chapters 4 and 5 suggest that temperature
tends to have an inverse relationship with elevation, whereas precipitation and snowfall
typically have a direct relationship with elevation, lending to the idea that the region's
topography greatly affects the variety of microclimates in the region. The warmest
region was the southeastern, low-lying Piedmont and the coolest temperatures were found
at the high peaks, particularly in the northeastern portion of the region. Highest
precipitation totals were near the union of NC, SC, and GA where a steep topographic
gradient exists. Low precipitation values were present in an area of lower elevations that
I have referred to as the 1-40 corridor. Snowfall totals were greatest in the highest-
elevation areas. In this chapter, I provided spatial representations of temperature and
snowfall that have not been provided in previous studies. My spatial representation of
precipitation was similar to that of Gaffin and Hotz (2000), although I was able to include
data for years beyond those included in their study. Their spatial study considered 1961-

1990, whereas mine covered the years 1930-2004.

111

CHAPTER 6

RESULTS AND DISCUSSION OF THE SYNOPTIC CLIMATOLOGY

6.1. Overview

Over the 1979-2006 period, 1437 precipitation events were determined. The
frequency of events revealed an increase over time, although the correlation was not
statistically significant (Figure 6.1). Eleven clusters were created from the objective
classification of the precipitation events (Section 6.2). Although they were based on
variables at the 850 hPa level (height, temperature, and dewpoint), 700 hPa level (height,
temperature, and dewpoint), and 300 hPa level (height), only the 850 hPa variables are
shown here (maps of 700 and 300 hPa variables are included in Appendix D). Eleven
clusters also emerged from the subjective (visual) classification (Section 6.3).
Subjectively derived clusters were based on variables at the 850 hPa level (height,
temperature, dewpoint, and wind), 700 hPa level (height, temperature, dewpoint, and
wind), and 300 hPa level (height and wind). The 850 hPa variables are shown in this
chapter (maps of 700 and 300 hPa variables are included in Appendix D). A short
discussion regarding lifting associated with the subjective clusters is also included
(Section 6.4). A hybrid clustering approach will be discussed within Section 6.5 and the
results from that method are included in Appendix E. The results of this chapter will help

answer Questions 3 and 4 presented in Section 2.5.

112

Precipitation Events by Year
BO me...

70 P
60 “
50 '
40 '
30 '
20 .-. ......

1980 1990 2000 2010
Years

 

Number of events

 

 

 

Figure 6.1. Precipitation events by year (1979-2006). Precipitation events were defined
as 25 percent or more of available stations reporting 0.5 in (12.7 mm) or more of
precipitation in a day.
6.2. Objective Clustering Results
6. 2. I . Clusters 2-5 and 7-9
The 11 resulting objective clusters exhibit a large amount of similarity. In fact,
the 11 clusters can be visually categorized further into two primary groups. The first
group consists of most of the cases and includes Clusters 2-5 and 7-9 (Figures 6.2-6.8).
Cluster 2 (345 total cases) and Cluster 3 (319 total cases) depict a warm-season situation
with weak temperature and height gradients, warm/moist conditions (1 0-15°C
temperatures and 5-10°C dewpoints), and southwesterly airflow (Figures 6.2-6.3).
Cluster 4 (228 total cases), Cluster 5 (208 total cases), and Cluster 7 (102 total cases)
occurred most frequently during transition seasons and had stronger gradients, cooler
temperatures (5-10°C temperatures), and sometimes a trough present over the study
region (Figures 6466). Cluster 8 (100 total cases) and Cluster 9 (99 total cases) were
most common during the cool season (Figures 6.7-6.8). Tighter gradients, cooler
temperatures (O-5°C temperatures), and a trough over the study region were common

with cases within those clusters.

113

 

 

 

 

Figure 6.2. Cluster 2. Temperature contoured every 5°C (top lefi), dewpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level.

114

“'2'

 

r
’ o v... T
1530:“; Ru . r 1) .?~/

Figure 6.3. Cluster 3. Temperature contoured every 5°C (top left), dewpoint contoured
every 5°C (top right), height contoured every 30 at (bottom left), and wind vectors
(bottom right) at the 850 hPa level.

115

 

 

 

 

 

 

 

 

 

 

 

Figure 6.4. Cluster 4. Temperature contoured every 5°C (top left), dewpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level.

116

 

 

Figure 6.5. Cluster 5. Temperature contoured every 5°C (top lefi), dewpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level.

117

 

- _ 1530

_—
---.'

Figure 6.6. Cluster 7. Temperature contoured every 5°C (top left), dewpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level.

118

 

Figure 6.7. Cluster 8. Temperature contoured every 5°C (top left), dewpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level.

119

 

Figure 6.8. Cluster 9. Temperature contoured every 5°C (top left), dewpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level.

6. 2.2. Clusters 1, 6, 10, and II

The remaining clusters consisted of a small number of cases (Figure 6.9-6.12).
Cluster 1 (2 total cases), Cluster 6 (3 total cases), Cluster 10 (27 total cases), and Cluster
11 (4 total cases) were similar. In each of the objective clusters, an east-west temperature
gradient and strong winds alofi (> 30 m/s) were often present. Most of these cases
occurred during winter. Clusters l, 6, and 10 had particularly low dewpoint values

(< -15°C), but these cases occurred infrequently. Of all 11 objective clusters, the

120

frequency of the cases within each objective cluster revealed that Clusters 1, 3, 6, 8, 10,
11 increased over time with Cluster 3 statistically significant at the p S 0.10 level and
Cluster 6 significant at the p S 0.05 level. Clusters 2, 4, 5, 7, and 9 decreased in

frequency over time, but none of the trends were statistically significant.

 

Figure 6.9. Cluster 1. Temperature contoured every 5°C (top left), dewpoint contoured
every 5°C (top right), height contoured every 30 at (bottom left), and wind vectors
(bottom right) at the 850 hPa level.

121

 

0-

Figure 6.10. Cluster 6. Temperature contoured every 5°C (top left), dewpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level.

122

 

Figure 6.11. Cluster 10. Temperature contoured every 5°C (top lefi), dewpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level.

123

 

l‘x ‘57 in L1 DEL—«(f
K d l?“ .51; _V___]

l
EST—"l“! EJ\4__H

 

  
  

 

9» , i Vi" _ _
_ x

1560 ’ ‘x 1‘ ~31 w: 0 v-g

-. ,. a... .3. . «.2 $22

Figure 6.12. Cluster 11. Temperature contoured every 5°C (top left), dewpoint contoured
every 5°C (top right), height contoured every 30 m (bottom left), and wind vectors
(bottom right) at the 850 hPa level.

6.3. Subjective Clustering Results
6.3.1. Closed Low (CL)

Compared to the objective approach, clusters resulting from the subjective
approach tended to be more distinct from one another. “Closed Low” (CL) was a small
cluster with 61 total cases (Figure 6.13). For all cases, a closed low was located directly
over the southern Appalachian region. Generally, weak cyclonic winds were present at

all levels and no discernible fronts were present. Temperatures over the region at 850

124

hPa were usually 10-15°C and dewpoints were 5-10°C. The cluster had two subtypes.
The first was a meso-low or low of tropical origin with closed height contours only at the
850 hPa level. The second was a closed (i.e., cut-oft) low of extra-tropical origin that
reached up to 300 hPa. CL occurred most frequently during the warm season with 13.1
percent of cases occurring during winter, 24.6 percent during spring, 39.3 percent during
summer, and 23.0 percent during autumn. Although CL was an infrequent pattern during
precipitation events, its occurrence increased over time (statistically significant at the p S
0.10 level). During 1979-1992 it accounted for 3.1 percent of the total cases, 4.5 percent

of the cases during 1986-1999, and 5.3 percent of the the cases during 1993-2006.

125

 

 

Figure 6.13. Closed low (CL). Temperature contoured every 5°C (top left), dewpoint
contoured every 5°C (top right), height contoured every 30 m (bottom left), and wind
vectors (bottom right) at the 850 hPa level.

6. 3.2. Northerly, Behind Low, N0 Fronts (NbLnF)

“Northerly, Behind Low, No Fronts” (NbLnF) had 127 totals cases and was a
common springtime precipitation pattern, with 21.3 percent of occurrences in winter, 37.8
percent occurring in spring, 21.3 occurring in summer, and 19.7 occurring in autumn
(Figure 6.14). Precipitation occurring with NbLnF was associated with a low-pressure
system to the northeast of the study region. Cyclonic flow circled around from behind

the low bringing northerly or northwesterly flow into the region. No fronts over the study

126

area were obvious, winds were generally weak at 300 hPa (10-20 m/s), and 850 hPa
temperatures were generally 5-10°C, while dewpoints were O-5°C. Like CL, NbLnF also
increased in frequency (7.8, 8.1, and 9.9 percent) during the three overlapping periods

and the trend was statistically significant at the p S 0.10 level.

 

 

Figure 6.14. Northerly, behind low, no fronts (NbLnF). Temperature contoured every
5°C (top left), dewpoint contoured every 5°C (top right), height contoured every 30 m
(bottom left), and wind vectors (bottom right) at the 850 hPa level.

127

6.3.3. Southerly, Ahead of Law, No Fronts (SaLnF)

A synoptic pattern with 211 total cases was “Southerly, Ahead of Low, No Fronts”
(SaLnF; Figure 6.15). The pattern was common during all seasons (26.1 percent during
winter, 29.4 during spring, 21.8 during summer, and 22.7 during autumn) and during all
three overlapping time periods (14.8, 14.9, and 14.5 percent of total cases) with no
statistically significant trend in frequency. The cluster was associated with a low-
pressure system located to the west/northwest of the study region. Airflow was generally
cyclonic and southerly or southwesterly at all levels. No fronts were present over the
study region; however, the region was sometimes within the warm sector of an
approaching low-pressure system. Southerly flow was responsible for relatively warm
(10-1 5°C) temperatures and (5-10°C) dewpoints over the region at 850 hPa. Upper-level

winds at 300 hPa were still generally weak (~30 m/s).

128

 

R~\/A

Figure 6.15. Southerly, ahead of low, no fronts (SaLnF). Temperature contoured every
5°C (top left), dewpoint contoured every 5°C (top right), height contoured every 30 m
(bottom left), and wind vectors (bottom right) at the 850 hPa level.
6. 3. 4. Northerly, No Low, N0 Fronts (NnLnF)

“Northerly, No Low, No Fronts” (NnLnF) occurred 159 times (Figure 6.16).
Northwesterly flow at all levels was the most common feature. No low-pressure center
or fronts were present, although a trough axis at the 850 hPa level was often present near
the study area. A ridge was usually present to the west of the study region. Winds were
weak, temperatures were roughly 10°C, and dewpoints were around 5°C. NnLnF was

most common in summer (34.0 percent), but did occur during other seasons (15.7 for

129

winter, 28.3 for spring, and 22.0 for autumn). During 1979-1992, NbLnF accounted for
10.0 percent of cases, 12.4 percent of cases during 1986-1999, and 12.1 percent of cases

during 1993-2006 with no statistically significant trend in frequency.

 

t 0 -~-
"\ w_} Jr“. 1' '
i- n .- ~‘1 W n .- \ A?
Figure 6.16. Northerly, no low, no fronts (NnLnF). Temperature contoured every 5°C

(top left), dewpoint contoured every 5°C (top right), height contoured every 30 m (bottom
left), and wind vectors (bottom right) at the 850 hPa level.

6. 3.5. Southerly, N0 Low, No Fronts (SnLnF)
Like NnLnF, the “Southerly, No Low, No Fronts” (SnLnF) pattern was most

common during summer (Figure 6.17). There were 186 total cases with 37.1 percent of

130

those during summer. Winter had 16.7 percent, spring had 21.5 percent, and autumn had
24.7 percent. Winds were generally weak at all levels and were southerly or
southwesterly. No low-pressure center or fronts were present. Temperatures at 850 hPa
were warm (10-1 5°C). Dewpoints were also high (~10°C) as a tongue of high dewpoint
values seemed to be associated with the southerly flow. The frequency of SnLnF did
decrease over the study period (14.5 percent'of cases during 1979-1992, 13.3 percent
during 1986-1999, and 11.4 percent during 1993-2006) although the decrease was not

statistically significant.

131

 

Figure 6.17. Southerly, no low, no fronts (SnLnF). Temperature contoured every 5°C
(top left), dewpoint contoured every 5°C (top right), height contoured every 30 m (bottom
left), and wind vectors (bottom right) at the 850 hPa level.

6. 3. 6. Bermuda High (BH)

“Bermuda High” (BH) was similar to SnLnF with the exception that anti-cyclonic
flow associated with the Bermuda High was more apparent in the height fields and wind
vectors (Figure 6.18). Often easterly flow was present to the south of the study region
with anti-cyclonic turning around the Bermuda High, leading to southwesterly flow over

the southern Appalachians. Winds were usually weak, 850 hPa temperatures were 10-

132

15°C and dewpoints were greater than 10°C. Of the 184 cases, 12.0 percent occurred
during winter, 24.5 percent during spring, 40.2 percent during summer, and 23.4 percent
during autumn. The frequency of the BH pattern has diminished over time (16.8 percent
of cases during 1979-1992, 11.8 percent during 1986-1999, and 8.9 percent of cases
during 1993-2006) and the trend was statistically significant and the p S 0.01 level. Most
BH cases were associated with a strong summertime Bermuda High, although a small
minority of cases were associated with a meso-high present over the southeastern US.

That subtype was most common during autumn months.

133

 

 

1560 t

\
.‘Q ~.

I." . i > ‘
, \ 9v frag;
‘. ° *1»

Figure 6.18. Bermuda High (BH). Temperature contoured every 5°C (top left), dewpoint
contoured every 5°C (top right), height contoured every 30 m (bottom left), and wind
vectors (bottom right) at the 850 hPa level.

6.3.7. North-South Oriented Front (nsF)

The most common synoptic pattern associated with precipitation in the region was
“North-South Oriented Front” (nsF), where I found a total of 3 12 cases (Figure 6.19).
Occurrence was most common during winter months (44.2 percent). Spring and autumn
both had 27.6 percent and summer only had 0.6 percent of the cases. The overall
occurrence of nsF increased over the three overlapping time periods (19.4, 19.7, and 24.0

percent of cases), but the trend was not statistically significant. The nsF pattern was

134

primarily characterized by a north-south oriented cold front located over the study region.
The front was usually present in the temperature and dewpoint fields at both 850 and 700
hPa. A trough was also present in the height and wind fields at all levels. Upper-level

winds were strong (40 m/s or greater) with a jet streak centered over the study region.

 

Figure 6.19. North-south oriented front (nsF). Temperature contoured every 5°C (top
left), dewpoint contoured every 5°C (top right), height contoured every 30 m (bottom
left), and wind vectors (bottom right) at the 850 hPa level.

135

6. 3. 8. Northerly, East- West Oriented Front (NewF)

The “Northerly, East-West Oriented Front” (NewF) pattern was similar to the nsF
pattern except that the cold front was oriented in a more east-west manner (Figure 6.20).
Again, the front was present in the temperature and dewpoint fields for both 850 and 700
hPa. A trough was present over the region at all levels, but the overall flow was more
northerly or northwesterly. Upper-level winds at 300 hPa were strong (40-50 m/s) and
the study region was located in a region of low-level convergence under the right rear
portion of an upper-level jet streak (Harman, 1991; Carlson, 1998). NewF was less
common than nsF with only 88 total cases. Of those cases, 51.1 percent occurred during
winter, 28.4 during spring, 4.5 during summer, and 15.9 during autumn. Although it was
not statistically significant, the frequency of NewF did increase by a small amount over
time. The occurrence during the three time periods (6.1, 6.3, and 6.2) did not capture the

small increase.

136

     

' ‘*~144df’{" ‘ 2
1470;):‘lr.

; i

\L_
11500 3.. {AM

Figure 6.20. Northerly, east-west oriented front (NewF). Temperature contoured every
5°C (top left), dewpoint contoured every 5°C (top right), height contoured every 30 m
(bottom lefi), and wind vectors (bottom right) at the 850 hPa level.
6. 3. 9. Southerly, East- West Oriented Front (SewF)

“Southerly, East-West Oriented Front” (SewF) was similar to nsF and NewF
(Figure 6.21). An east-west cold front was present, but airflow was primarily southerly
or southwesterly over the region. Upper-level winds were strong (40-50 m/s) and the
region was again located under the right rear of a jet streak. SewF was an uncommon
pattern and occurred just 37 times. During 1979-1992, it occurred during 3.0 percent of

cases, 3.1 percent during 1986-1999, and 2.2 percent during 1993-2006 and an overall

137

decreasing trend was not statistically significant. Most of the cases were during winter
(70.3 percent), although a few occurrences occurred in spring (10.8 percent), summer

(2.7 percent), and autumn (16.2 percent) as well.

73’- "' ""- n- 3,12. 4'
r” . r 44

 

7'
”214‘ .731 .4.

Figure 6.21. Southerly, east-west oriented front (SewF). Temperature contoured every
5°C (top left), dewpoint contoured every 5°C (top right), height contoured every 30 m
(bottom left), and wind vectors (bottom right) at the 850 hPa level.

6.3.10. Zonal F low (Z)

The “Zonal” (Z) pattern was generally one of west-to-east flow at all levels

138

(Figure 6.22). It was an infrequent pattern with only 51 total cases. Of those, 43.1
percent occurred in winter, 25.5 percent in spring, 17.6 percent in summer, and 13.7
percent in autumn. The frequency trend of the Z pattern did not change much over time
and was statistically insignificant (3.2 percent of cases during 1979-1992, 4.2 of cases
during 1986-1999, and 3.8 of cases during 1993-2006). Temperatures at 850 hPa were
usually 5-10°C and dewpoints were 0-5°C. Upper-level winds at 300 hPa were relatively

weak (~30 m/s).

 

\1 \\
fix: v:_?\ ti

 

Figure 6.22. Zonal flow (Z). Temperature contoured every 5°C (top left), dewpoint
contoured every 5°C (top right), height contoured every 30 m (bottom left), and wind
vectors (bottom right) at the 850 hPa level.

139

6. 3. 11. Weakflow (W)

Finally, the “Weak” (W) cluster was a collection of 21 remaining cases that did
not fully fit any of the previous 10 clusters (Figure 6.23). Cases within the cluster
generally had weak airflow circulation over the study area, with no distinct meso-high or
meso-low directly over the region. A few of the cases had easterly flow with a low-
pressure center to the south and a high-pressure center to the north. Most of the cases
occurred during summer (57.1 percent) with warm temperatures and dewpoints. Winter
had 4.8 percent of the cases, spring had 9.5 percent, and autumn had 28.6 percent. The
frequency of these cases increased slightly over time (1.3 percent of cases during 1979-
1992, 1.5 percent of cases during 1986-1999, and 1.6 percent of cases during 1993-2006)

although the trend was not statistically significant.

140

 

 

\ x. i i ’
L‘J n 3"- ~ VR‘L }
Figure 6.23. Weak flow (W). Temperature contoured every 5°C (top left), dewpoint

contoured every 5°C (top right), height contoured every 30 m (bottom left), and wind
vectors (bottom right) at the 850 hPa level.

,/
A _ ‘\ /./.

6.4. Speculations Regarding Lifting Mechanisms

Although lifting mechanisms associated with the synoptic patterns were not
explicitly considered by the clustering approaches used in this study, the topic is still
important to determine the origins of precipitation mentioned in Question 3. The dataset
resolutions (once daily COOP observations and NARR data only at 2100 UTC) used here
were not necessarily best suited for distinguishing the time or region of lifting since the

process can be of short duration or favor certain times of day not captured by the

141

observations. In addition, the NARR data have a horizontal resolution of 32 km, and
although that is high resolution as far as climate data are concerned, lifting in this region
frequently occurs on a more local scale as a result of the complex topography and
warm/moist conditions. Some COOP stations provide hourly data and NARR data are
available every 3 hours, which could assist a subsequent study in determining the timing
of lifting associated with the above synoptic patterns, but for now, horizontal data
resolution remains an obstacle.

Despite the data resolution shortcomings, I was able to create maps of vertical
velocity (in Pa/s) at the 700 hPa level. For the subjective clusters, some amount of lifting
occurred with the BH, CL, SaLnF, SewF, SnLnF, and W patterns (Figures 624-629).
With the exception of the SewF pattern, these patterns were most often summertime
events. Southerly flow was associated with the BH, SaLnF, SewF, and SnLnF types and
may have resulted in orographic lifting due to the abrupt increase in elevation along the
southern side of the Blue Ridge. Upward motion for the CL pattern makes sense given
that a closed low was located over the study area with that pattern. Frontal lifting could
be responsible for the upward motion associated with the SewF pattern. Additional
lifting may have resulted from the presence of a 300 hPa jet streak located over the study
area, which would have promoted lower-tropospheric lifting (Harman, 1991; Carlson,
1998). No lifting was found within the study region for the other patterns, although it
still could have occurred at a different time of day or at a scale that was not captured by
the 32 km horizontal data. In particular, frontal lifting may have been common with the
nsF and NewF patterns (vertical velocity maps not shown). Likewise, some lifting may

have been present with objective Clusters 1, 3, 4, 6, 7, 9, and 11 (vertical velocity maps

142

not shown), as some of those patterns were frontal in nature.

 

 

 

 

 

84W 83W 82W

Elevation (m)

 

Figure 6.24. Vertical velocity (in Pa/s) at 700 hPa for the BH pattem. Negative values
represent areas of upward motion and positive values represent areas of downward

motion.

143

 

Elevation (m)

   

 

 

 

84W 83W 82W

Figure 6.25. Vertical velocity (in Pa/s) at 700 hPa for the CL pattern. Negative values
represent areas of upward motion and positive values represent areas of downward
motion.

 

Elevation (m)

   

 

 

 

84W 83W 82W

Figure 6.26. Vertical velocity (in Pa/s) at 700 hPa for the SaLnF pattern. Negative
values represent areas of upward motion and positive values represent areas of downward
motion.

 

Elevation (m)

 

35N

 

 

 

 

84W 83W 82W

Figure 6.27. Vertical velocity (in Pa/s) at 700 hPa for the SewF pattern. Negative values
represent areas of upward motion and positive values represent areas of downward
motion.

 

Elevation (m)

  

36N ;'

35N

 

 

 

84W 83W 82W

Figure 6.28. Vertical velocity (in Pals) at 700 hPa for the SnLnF pattern. Negative
values represent areas of upward motion and positive values represent areas of downward
. motion.

145

 

Elevation (m)

 

 

 

 

 

84W 83W 82W

Figure 6.29. Vertical velocity (in Pa/s) at 700 hPa for the W pattern. Negative values
represent areas of upward motion and positive values represent areas of downward
motion.

6.5. Discussion
6.5.1. Advantages and Disadvantages of Subjective and Objective Clustering

As I outlined near the end of Chapter 2, one of the primary objectives of this
study was to create a climatology of the synoptic patterns associated with precipitation in
the southern Appalachian region. My intent was to analyze the 1437 precipitation events
and categorize them into groups sharing similar synoptic features. In this chapter, I
provided the results from objective and subjective clustering approaches. Each approach
has strengths and weaknesses, and as I showed, the results between the two approaches
differed substantially.

The objective clustering process using IDL and CLUTO provided a hands-free

and automated way of clustering the synoptic patterns that was far faster than a subjective

146

approach could ever hope to achieve. The process considered each grid point, variable,
and atmospheric level equally to eliminate any bias during the clustering. Unfortunately,
this method of eliminating bias also had a potential drawback of giving grid points near
the edge of the map equal weight as grid points directly over my study area. In addition,
the objective approach may not have been capable of determining how neighboring grid
points were related and smaller-scale features were not detected as frequently. As
discussed in Chapter 3, the objective approach was hindered by limitations in the
software and virtual memory of the computer used. The process of experimenting with
smaller grid arrays and fewer variables was perhaps the most time-consuming step of the
entire project. Once 1 determined the appropriate grid sizes and variables, the clustering
process moved quickly. The coarse resolution of the data grid (8 x 9) may explain why
some smaller features and gradients were not fully captured by the objective approach
resulting in many similar cluster patterns. In addition, statistics generated by the CLUTO
software provided inconclusive guidance regarding the appropriate number of clusters, so
I chose to use 11, which was the same number determined by the subjective approach. In
the end, the 11 objective clusters were not nearly as distinct as the subjective clusters. In
fact, the objective clusters broadly fit into two main categories, as stated earlier. If,
however, the coarse grid used and the inconclusive results regarding the appropriate
number of clusters did not hinder the outcome of the cluster patterns, the results from the
objective analysis suggest that many distinct synoptic patterns may not affect the
precipitation of the region. Such a result would suggest that local features like
topography, heat, and moisture have a larger impact on the region's precipitation

climatology.

147

In contrast to the objective approach, the subjective approach allowed me to
visually examine each individual case. My general knowledge of synoptic climatology
assisted me as I searched for synoptic features such as fronts, high- and low-pressure
centers, troughs, ridges, and jet streaks. The benefit of this process was that I considered
each case in a holistic way, considering the relationships between values at neighboring
grid points. The process also required me, the researcher, to be fully involved in the
clustering process. Although it was a time-consuming approach, the act of examining
each case provided me the opportunity to monitor the synoptic patterns during
precipitation events for my region of study over a 28-year span. A first-hand experience
such as that cannot be obtained during objective clustering.

The downsides of the subjective clustering included the time involved, user bias,
and user error. I made my best attempt to consider the various cases with an open mind,
but certainly my background in synoptic climatology and my review of previous studies
of the region may have led me to have some preconceived ideas about what synoptic
patterns to expect. In addition, the human eye is likely drawn to certain spatial features
(e.g., fronts), certain atmospheric levels, or variables and may lead to bias during the
clustering process. For example, I often found myself focusing initially on the
temperature and height maps at 850 hPa simply because they were the first two maps 1
saw based on the way I printed out the maps. If I had arranged the maps in a way that
had the 700 or 300 hPa maps first, my results may not have been exactly the same. In
order to avoid as much user bias and error as possible in my visual typing, 1 went through
several iterations during the clustering process. Once my clusters were established, 1

reviewed the clusters multiple times to check for continuity. If one or more cases did not

148

fit in, I would move those cases to a more appropriate cluster. The results from the
subjective approach did provide 11 distinct synoptic patterns that resembled traditional
synoptic features, suggesting that multiple synoptic patterns often have an impact on the

region's precipitation climatology.

6. 5. 2. Consideration of a Hybrid Clustering Approach

Given the advantages and disadvantages of subjective and objective clustering
approaches and the differing results presented here, I am unable to conclude that one
approach is superior to the other. My suggestion is that others consider both subjective
and objective approaches and if possible use multiple objective approaches to create an
ensemble of results. Another option I explored in this study was the idea of a hybrid
clustering approach, where I used the objective approach to initially create 50 clusters. 1
then visually examined the maps for those clusters and subjectively clustered them
further to ultimately reach eight clusters (maps are included in Appendix E). The overall
resulting patterns had many similarities to the patterns from the subjective approach. The
eight resulting hybrid clusters were categorized as BH, CL, NbLnF, SaLnF, SnLnF, nsF,
NewF, and SewF. No discernible hybrid patterns resembling NnLnF, Z, or W were
found. Of the eight hybrid clusters, BH, nsF, NewF, and SewF had nearly the same
number of cases as was found during the subjective approach, whereas the CL, NbLnF,
and SaLnF clusters had fewer cases. The largest deviation in terms of the number of
cases was the SnLnF cluster, which had 749 cases under the hybrid approach but 186
cases with the subjective approach. Since the hybrid CL, NbLnF, and SaLnF clusters

were so few in number, the patterns were likely comprised of outlying cases (e.g., the

149

hybrid CL and SaLnF patterns had very low dewpoint values present) and may not fully
represent an average CL, NbLnF, or SaLnF case.

The hybrid approach itself was an experiment and the initial choice of creating 50
objective clusters was simply arbitrary. Increasing the number of initial clusters would
allow for more clusters to be found during the subjective process. A future study of the
whole cluster analysis issue may wish to explore this topic more thoroughly in order to
determine how many initial objective clusters to create. For instance, initially using 50
clusters to represent 1437 cases essentially meant reducing the number of cases to 3.5
percent of the original total. Perhaps using a smaller or larger value, such as 1 percent
(14 clusters in this case), 2.5 percent (36 clusters in this case), 5 percent (72 clusters in

this case) or 10 percent ( 144 clusters in this case), would have been a better option.

6. 5. 3. Comparison with Previous Studies

In regards to the cyclone studies covering the southern Appalachian region (e.g.,
Miller, 1946; Klein, 1957; Reitan, 1974; Bosart, 1975; Zishka and Smith, 1980), the
synoptic pattern that is most relevant here is nsF, since it possesses the typical Norwegian
cyclone model cold front. Because I was analyzing the synoptic situation only near the
time of precipitation, I did not monitor the synoptic patterns through time to detect the
initial timing and location of cyclogenesis. Thus, the nsF pattern may contain cyclones of
Colorado, Alberta (Zishka and Smith, 1980), and Atlantic (Miller, 1946; Bosart, 1975)
origins. My nsF pattern most closely resembles Miller's (1946) Type B cyclones, Gaffin
and Hotz's (2000) Synoptic pattern, and Davis and Rogers's (1992) Cluster 3. Miller

(1946) found Type B cyclones to be most common during the winter season, which

150

agrees with my findings for the nsF pattern. The works of Klein (1957), Reitan (1974),
and Zishka and Smith (1980) also supported the idea that cyclones in the region were
more common during winter, although Davis and Rogers (1992) found their type to be
most common in late spring. Zishka and Smith (1980) also found cyclone frequency to
decrease during the middle portion of the last century, and some authors speculated in
subsequent studies about the potential impact cooling temperatures during that time may
have had on that decrease in cyclone frequency (van Loon and Williams 1976a, 1976b;
Harvey 1978; Diaz and Quayle, 1978). In contrast, the time period for this study (1979-
2006) began shortly after Zishka and Smith's (their study covered 1950-1977). I found
that the frequency for the nsF pattern increased over the time period, although the trend
was not statistically significant. Other cyclone-related patterns such as CL and NbLnF
also increased in frequency and were statistically significant at the p S 0.10 level. As I
discussed in Chapter 4, this time period was accompanied by warming temperatures,
which echoes the idea posed by others (e.g., van Loon and Williams 1976a, 1976b;
Harvey 1978; Diaz and Quayle, 1978) that cyclone frequency in the region may be
related to (though not necessarily caused by) rising temperatures. If true, as temperatures
continue to rise in the twenty-first century, cyclogenesis and/or cyclone passage in the
region may become increasingly more common (IPCC, 2007a; USGCRP, 2009). The
increase in cyclone frequency during 1979-2006 also coincided with a statistically
insignificant increase in the number of precipitation events, as defined by this study. The
USGCRP found precipitation in the southeastern US. to increase over 1970—2008, but the
rainfall was received over fewer events, making droughts more common (USGCRP,

2009)

151

The NewF and SewF patterns found in this study resembled the Type E pattern
discussed by Bosart (1975) and the Frontal pattern discussed by Gafl'ln and Hotz (2000).
I was able to distinguish two different patterns based on the low-level flow of the study
region and general location of the east-west front. Both Bosart (1975) and Gaffin and
Hotz (2000) appear to have grouped these cases together. I found both the NewF and
SewF patterns to be most common during the winter months. Although Gaffin and Hotz
(2000) considered only heavy rain events, they found their Frontal pattern to be most
common during summer. Bosart (1975) did not discuss the seasonal frequency of his
Type E pattern.

Of my remaining synoptic patterns, Gaffin and Hotz (2000) had a Meso-High
pattern that was similar to my BH pattern. Our studies agree in that those patterns were
most common during summer. We both grouped Bermuda High and meso-high cases
together, although we opted to name our groups differently. I found Bermuda High cases
to be more common during summer with meso-highs being more common during
autumn. During a streamline analysis, Bryson (1966) found mesa-highs to be common
during autumn and winter (September through February). Cluster 1 of Davis and Rogers
(1992) was also similar to my BH pattern. They found the pattern to be most common
during July and August.

Gaffin and Hotz's (2000) Tropical pattern also shares some similarities to my CL
pattern, although their pattern only included tropical lows. In contrast, my CL category
included tropical lows and extra-tropical cut-off lows. Similar to my CL pattern, Gaffin
and Hotz's Tropical pattern was common during summer and autumn, although I did find

some cases occurring during spring as well. Davis and Rogers's (1992) Cluster 6 was

152

similar to my NbLnF pattern. I found that pattern to be most common in spring, and their
results agreed, with Cluster 6 most common in April. Their Cluster 9 also resembled
some features of my Z and BH patterns. Davis and Rogers (1992) found Cluster 9 to
have anticyclonic flow near the surface and zonal flow aloft. No previous studies
thoroughly covered the remaining patterns I found in this study (N nLnF, SaLnF, SnLnF,
Z, and W).

O’Handley and Bosart (1996) discussed the possible effect that the southern
Appalachian topography has on cyclones and fronts. Frontal speed and orientation were
found to be affected by the mountains with the fronts developing a “kink” in shape.
Because I did not monitor the events over time, 1 cannot comment regarding the speed of
the fronts, but, in many cases, my synoptic patterns did have kinks or distortions in the
fronts near the mountainous areas. Frequently, the height fields had a weak trough over
the Appalachians (e.g., nsF, NewF, and NnLnF) and the temperature and/or dewpoint
fields showed possible impacts (e.g., nsF, NewF, and Z) from the mountains. As I
discussed earlier, temporal and spatial trends for temperature, precipitation, and snowfall

all appeared to be greatly affected by the region's complex topography.

6.6. Summary

I found 11 distinct clusters using subjective clustering and of those some have
been covered in previous studies, but others (e.g., NnLnF, SaLnF, SnLnF, Z, and W) were
not covered. An objective clustering approach was also used to determine 11 clusters.
Patterns resulting from the objective approach were quite similar and could be
qualitatively grouped into two primary categories. I also recommended a combination of

an objective method to cluster the cases into a predetermined number of clusters with a

153

subjective (visual) step to combine them further into a smaller number of clusters. In this
hybrid approach, I used 50 initial clusters, but recommended more investigation
regarding the appropriate initial number is needed. I was able to reduce the 50 clusters
down to eight. The resulting hybrid clusters showed many similarities to the subjective

clusters.

154

CHAPTER 7

SUMMARY AND CONCLUSIONS

7.1. Summary

In this dissertation, I presented an expanded climatology of temperature and
precipitation for the southern Appalachian region. I undertook this study, in part, because
most climatological studies of the region are now dated, and, given its biological
importance, I felt that an updated description of the region’s recent climate was needed.
While we already know that the region is a warm temperate location that receives
substantial precipitation, my long-term study provided the opportunity to highlight trends
in these climatic conditions, both temporally and spatially.

The need for a thorough climate study of the region was supported further by
model projections summarized by the Intergovernmental Panel on Climate Change in its
Fourth Assessment Report (IPCC, 2007a) and a recent national report from the US.
Global Change Research Program (USGCRP, 2009). The IPCC and USGCRP reports
briefly covered the southern Appalachians and projected that the region will experience
warming temperatures during this century, but precipitation projections were much more
uncertain (IPCC, 2007a; USGCRP, 2009).

In Section 2.5, I shared four research questions to be addressed in this dissertation
study. In this final chapter, 1 will now revisit the research questions and summarize the
results presented in the previous three chapters in an effort to answer each research

question.

155

1) What is the temperature and precipitation climatology of the southern Appalachians?

Regarding the temporal aspect of Question 1, temperatures were warm during the
19303 and 19403 and then decreased until the mid 19603 to the late 19703. A two-phase
regression analysis unveiled that statistically significant breaks were common spanning
the mid 19603 through the late 19703, showing some support of a 1976/1977 climate shift
documented by others (Trenberth, 1990; IPCC, 2007a). In recent decades, a warm trend
began that continued until the end of the record. A linear representation of the data
revealed an overall cooling trend during the 1931-2006 period. This trend was probably a
statistical reflection of extended warmth during the 19303 and 19403 and cooling during
the middle portion of the record. Cooling spanning the 19403 to the 19703 was common
nationally as sulfur dioxide levels from coal power plants increased nationwide (Delcourt
and Delcourt, 2001). Although introduced legislation has led to nationally decreasing
sulfur dioxide levels since the 19703, the increasing population and energy demands,
coupled with the region's reliance on coal energy, has kept sulfur dioxide levels high in I
the southern Appalachian region and is probably an explanatory factor for the region's
statistically significant cooling trends over the 1931-2006 period (Delcourt and Delcourt,
2001)

Warming in recent decades was often more strongly developed in the minimum
temperature data than it was for maximum temperatures, suggesting that diurnal
temperature ranges decreased over the period. In particular, a decreasing diurnal
temperature range at high elevations was often statistically significant at the p S 0.01

level. Also, in recent decades, cool season temperatures tended to increase (e.g., by more

156

than 4.00°C for January), whereas warm season temperatures remained stable or even
cooled, suggesting that the annual range of temperature also decreased. At all elevations,
temperatures reached an annual maximum during July and a minimum during January.
Temperatures were found to be affected by elevation, with temperature values decreasing
as elevation increased (see Figure 4.6). Temperature values for the region generally
support the Koppen classification of this region as Cfa or htunid subtropical (i.e., average
temperature of the coldest month was below 18°C, and above -3°C). The recent warming
trends also provide some support for the possible projected trends discussed by the
IPCC's Fourth Assessment Report and the recent report from the USGCRP (IPCC, 2007a;
USGCRP, 2009).

In general, annual precipitation totals increased over the 1931-2006 period, but
the values were not statistically significant. Precipitation increased during April through
June and September through November, but decreased during most other months.
Snowfall totals decreased at low- and mid-elevation stations during most months, but
neither trend was statistically significant. In contrast, snowfall increased at high-
elevation stations during the period of record and was significant at the p S 0.01 level. At
the high elevations, snowfall decreased during October and November, but increased
during other winter and spring months, suggesting that more snow was received in late
autumn through the following spring. Intra-annual values of precipitation reached
maximum values during March and July with minima common in October and April.

The typical snow season spanned October through May at high elevations with a
somewhat shorter snow season at lower elevations. Maximum snowfall totals occurred

during January for low and mid elevations and during February for high elevations. With

157

IPCC (2007a) and USGCRP (2009) results remaining uncertain regarding future
precipitation trends in the region, the precipitation and snowfall trends found here
provide useful information if the trends continue. Spatial aspects of the region's

temperature, precipitation, and snowfall climatology lead to Question 2.

2) How many climate regimes exist in this region?

Based on the horizontal resolution of station data, I found temperature to be
associated with elevation, and to a lesser extent, latitude. In the end, up to five climate
regimes were recognized, and the resulting patterns varied somewhat based on what
variable was clustered. Clusters for minimum, maximum, and mean temperature were
quite similar. The warmest regime contained stations at elevations generally below 500
m, many of which were in the Piedmont province in the southeastern portion of the study
area. The second-warmest regime included stations at an elevation of approximately 500
m on both sides of the Blue Ridge. The third-warmest regime included mid-elevation
stations. The fourth warmest regime included mostly mid-to-high elevations, and the
filth warmest (i.e., coldest) regime included stations at the highest elevations.

Precipitation and snowfall generally had fewer than five climate regimes. A
precipitation maximum was often present near the confluence of the NC, SC, and GA
borders. A second maximum was sometimes present in high elevations along the
northeastern portion of the study area and along the northwestern (or TN) side of the Blue
Ridge. A precipitation minimum was usually present in the vicinity of the I-40 corridor,

which I defined as the area of lower elevations cutting through the Blue Ridge from

158

northwest to southeast. Snowfall was usually received in small amounts for low- and
mid-elevation stations, and only two or three snowfall regimes were generally present,
with most low-elevation and high-elevation stations clustering separately.

Some amount of intra-annual and inter-annual variability amongst the clusters
was present. Cluster patterns were frequently more distinct during winter than summer
months. I suspect that varying snowfall totals and a stronger temperature gradient across
elevations were largely the reasons for the variability. Nonetheless, most variation in
cluster patterns was probably due to the changes in station availability. Specifically, the
number of stations in TN and around the area of GSMNP increased over time. Near the
beginning of the record, usually only about 30 stations across the study region were
available for clustering, but by the last decade more than 70 stations were typically
available. The inclusion of more stations near the end of the record made it easier to

visualize cluster patterns.

3) What are the synoptic origins of precipitation here?

Using a subjective clustering approach, I found 11 distinct synoptic patterns that
led to southern Appalachian precipitation. Common summertime patterns included
precipitation associated with a strong “Bermuda High” (BH), a “Closed Low” (CL)
situated over the study region, “Southerly, No Low, No Fronts” (SnLnF), or “Northerly,
No Low, No Fronts” (NnLnF). A few cases of “Weak Flow” (W) were also most
common during summer. A “Northerly, Behind Low, No Fronts” (NbLnF) pattern of

northerly flow wrapping around the backside of a low-pressure center had its greatest

159

frequency during spring months. In general, warm season patterns exhibited weaker
winds and weaker temperature, dewpoint, and height gradients. On the other hand, cool
season patterns were frequently frontal in nature. Of those, one pattern had a typical
Norwegian cyclone model cold front with a southwest-northeast or “North-South
Oriented Front” (nsF). A second “Northerly, East-West Oriented Front” (NewF) pattern
was associated with northerly flow and an east-west oriented cold front located over the
study area (N ewF). A third “Southerly, East-West Oriented Front” (SewF) pattern had
southerly flow and an east-west oriented cold front. A “Zonal Flow” (Z) wind pattern of
west-to-east winds was most common during winter. Lastly, a “Southerly, Ahead of Low,
No Fronts” (SaLnF) pattern associated with southerly flow in front of an advancing low-
pressure center was common during all seasons.

Of these 11 subjective synoptic patterns, six of the patterns were associated with a
low-pressure center, frontal zone, and/or an upper-level jet streak (CL, NbLnF, SaLnF,
nsF, NewF, and SewF). The number of cases included within those patterns was 836,
accounting for roughly 58 percent of the total number of precipitation events. These
events favored cooler months. The remaining five patterns (NnLnF, SnLnF, BH, Z, and
W) were responsible for 601 precipitation events (42 percent of the total). None of those
patterns had a low-pressure system, front, or jet streak present. Many of those cases
occru'red during warmer months and the precipitation was likely due to orographic lifting
resulting from complex topography and other surface irregularities combined with the
pronounced warm and moist summer conditions of the region. The CL (statistically
significant at the p S 0.10 level), NbLnF (statistically significant at the p S 0.10 level),

nsF, NewF, NnLnF, SaLnF, and W patterns increased in frequency over the study period,

160

while the BH (statistically significant at the p S 0.01 level), SnLnF, SewF, and Z patterns
decreased in frequency. This overall result suggests that precipitation events associated
with cyclone activity increased during 1979-2006, whereas precipitation events resulting
from orographic lifting and/or surface heating decreased in frequency during the same
period. Although I did not specifically cover the quantity of precipitation received with
each synoptic pattern, it is likely that precipitation resulting from cyclonic events was
probably more widespread and often impacted all elevations. Precipitation events
associated with primarily orographic patterns, however, were probably limited to mid and
high elevations. Thus, I may speculate that the trend toward more events associated with
cyclonic systems should have resulted in more precipitation received at low elevation
areas, which is consistent with results summarized in Table 4.4.

The impact of defining a precipitation event as 25 percent of stations reporting 0.5
in (12.7 mm) in a day should not be overlooked when considering the results above. As
shown in Table 3.4, this definition still included just 16 percent of the possible
precipitation events. Precipitation in the region was often received in smaller amounts by
a smaller percentage of stations. Although the criteria used here were less stringent than
those used by others (e. g., Bogucki, I972; Konrad, 1997; Gaffin and Hotz, 2000), the
precipitation events as defined by this study still had a bias toward “major” precipitation
events and may in turn have a bias toward synoptic origins of precipitation rather than
origins from local topography, heat, and moisture. Nonetheless, the results above
illustrate that the combination of heat, moisture, and complex topography within the
region appear to be important factors controlling the number of precipitation events and

the amount of precipitation received as well. Hypothetically speaking, if the southern

161

Appalachian Mountains were removed while all other aspects of the region's climate
controls were unaltered, the number of precipitation events, the overall amount of
precipitation received, and moisture present in the region, would all likely be less than
what is experienced currently. Ranges in precipitation would be more uniform across the
region with most locales receiving 1000-1300 mm annually. However, given the region's
complex terrain and its impact on precipitation, some high-elevation stations receive
more than 2000 mm annually. These large precipitation values in high elevations also
affect the lower-lying valleys through downslope drainage of rainfall and snow melt, as
well as through the resulting increases in humidity. The significant amount of overall
moisture brought about by the region's varying topography is at least partially responsible
for the lush and diverse environments found there. Consequently, because the
topography keeps a substantial moisture supply anchored in the region, we may have
some assurance that the region's moisture may be less susceptible to any changes in

synoptics during projected changes in global climate.

4) To what extent does the answer to Question 3 depend on whether an objective,
subjective, or hybrid clustering approach is used?

The discussion above for Question 3 covered the results from the subjective
clustering approach in detail. However, that approach tended to have a bias toward more
and distinct synoptic patterns. In contrast, the results from both the objective and hybrid
clustering approaches suggested that fewer than 11 clusters may be present and/or that
clusters may not be as distinct from one another as those that were subjectively derived.

On the one hand, this finding suggests that perhaps precipitation in the region is not

162

affected by synoptic-scale origins as much as the subjective clustering results tend to
reveal. On the other hand, however, the finding may be the result of statistical
shortcomings in the objective and hybrid approaches used. In regards to the objective
approach, statistics for determining the appropriate number of clusters produced
inconclusive results, so I ultimately created 11 objective clusters, which was the same
number of clusters revealed by the subjective approach.

The resulting 11 objective clusters were less distinct from one another than the
subjective clusters were. As shown in Chapter 6, I was able to broadly categorize the 11
objective clusters into non-frontal and frontal groups. The non-frontal group was
comprised of most of the cases and some of the clusters highlighted seasonality with
varying gradients and temperatures values. The small minority of frontal cases suggested
that most precipitation in the region originates locally. This result should be interpreted
carefully though since the objective approach utilized a much smaller data grid in order to
remain within the computational limits of the software and computer. The use of the
smaller grid (8 x 9) may at least partially explain why the objective clusters (and
subsequent hybrid clusters) were less distinct than the subjective clusters.

Lastly, I proposed the option of a hybrid approach, wherein I used an objective
analysis to first cluster the cases into 50 clusters. I then printed the maps for those 50
clusters and subjectively clustered them further to ultimately arrive at eight clusters. Of
those eight patterns, many closely resembled those derived from the subjective approach.
However, the “Closed Low” (CL) pattern uncovered by the hybrid approach had only two
cases, and the “Southerly, Ahead of Low, No Fronts” (SaLnF) pattern had only one case,

so those two patterns should be considered with caution. The hybrid “Southerly, No Low,

163

No Fronts” (SnLnF) also had far more cases (749) than its similar subjective SnLnF
pattern (186). The decision to begin with 50 objective clusters was an arbitrary choice.
If future studies consider a hybrid approach, it may be beneficial to experiment with

increasing or decreasing the number of initial objective clusters.

7.2. Future Climate Change Impacts

Although this dissertation project was not a climate modeling study, the results
still provide useful information regarding the region's climate in recent decades and shed
some light on possible future climatic trends in the region. In regards to precipitation, an
overall trend toward wetter conditions was present during 1931-2006, although drier
conditions existed near the end of the study period and in the years since. In general,
moister conditions may help moderate temperatures near the ground, particularly in
shaded areas, which could offset the effects of increasing temperatures during the twenty-
first century. Consequently, this moister trend could help decrease negative impacts on
the biodiversity of the region. However, if the recent dry period becomes a new climate
regime, then negative impacts on the region's diversity could be more severe in the years
to come if low— and mid-elevation species adapt by moving upslope and habitats for high-
elevation species decrease.

My results showed that trends in temperature and temperature range were not
always consistent by elevation, season, or time of day. Many statistically significant
trends were present in the results of this study, but even if those trends continue, they may
not necessarily translate to significant ecological change. Conversely, one or more

statistically insignificant trends could have a dire impact on a species if the species is

164

already approaching a tipping point. Whether species will adapt by simply moving
upward in elevation or adapt in some other way is unclear. However, the loss of grassy
bald environments does suggest that sky island shrinkage is occurring. If plant and/or
animal species are found to be affected in negative ways, those interested in policy
decisions regarding conservation may wish to explore potential mitigation measures that
could be taken to minimize impacts, such as the grassy bald restoration projects
conducted in GSMNP on Gregory Bald and Andrew's Bald (Houk, 2009).

Some have countered that such restoration projects are never-ending battles
against nature that ought not be pursued. Another option includes preserving or creating
ecological “corridors” to allow species migration during changing climate conditions
(Delcourt and Delcourt 2001; Delcourt, 2002). This option may have some promise
given that the Blue Ridge of the southern Appalachian region is already largely connected
by various federally-owned properties, creating a large and continuous expanse of
potential habitat. However, some studies suggest that such conservation corridors are not
always effective when political conflicts arise (Goldman, 2009) or during times of rapid
ecological change when species simply do not have enough time to adequately migrate
through the corridor (Delcourt and Delcourt 2001; Delcourt, 2002). In addition, even
with corridors present, high-elevation sky islands still remain isolated from other
potential habitats, making migration for those species unlikely. Based on knowledge of
past migration speeds, Delcourt and Delcourt (2001) and Delcourt (2002) have
recommended that corridors should be used concurrently with transferring vulnerable
species from current sites to new suitable sites in an effort to speed up migration times

and improve the survival chances of sky island species. Such experiments of “assisted

165

migration” are already underway within British Columbia (Marris, 2009).

7.3. Recommendations for Future Research

For this project, my goal was to provide an updated and elaborated baseline
climatology and synoptic climatology for the southern Appalachian region. I believe I
have been largely successful in that task, but nonetheless, areas remain where I feel
additional research is needed. First, although station availability increased substantially
over the length of my study period, I still recommend that additional COOP stations be
added. Particularly, underdeveloped regions, such as GSMNP and the neighboring
national forests have fewer stations compared to the surrounding area. Additional
stations in those regions may help future projects better understand climate change and
variability at those locations.

Second, for the spatial analysis and determination of climate regimes, I focused
on five variables (minimum, maximum, and mean temperature, total precipitation, and
total snowfall). I chose those variables largely because they were often used in previous
studies, so it provided me the ability to compare my results to previous work. Future
studies may wish to consider some of the other available variables. I suspect that many
of those variables have later starting dates or are reported infrequently, as 1 found this to
be the case with total evaporation and mean soil temperature. The availability of these
variables may improve over time.

Third, although I was able to create a synoptic climatology of the area using three
different approaches (objective, subjective, and hybrid), each approach did have some

shortcomings. Of the seven statistical measures available within CLUTO for determining

166

the appropriate number of objective clusters, none of them provided a conclusive answer
regarding the correct number. An exploration of additional statistics, such as those
studied by Milligan and Cooper (1985), may be an improvement. Furthermore, the
objective analysis was hindered by the computational limitations of the IDL software and
the computer I used. Perhaps those with the resources available to handle the large
NARR dataset in an objective analysis could provide a better cluster representation of the
region since they could utilize a data grid with a higher resolution and/or could include
additional synoptic variables. They may also be able to easily conduct a variety of
objective analyses in order to create an ensemble of cluster results from which
conclusions could be drawn. The hybrid clustering approach I introduced here mayalso
be improved by further experimentation regarding the initial number of objective clusters
to create. I started with 50 clusters to represent 1437 cases in this study, but some
additional experimentation may improve the hybrid approach.

Fourth, although I did briefly discuss lifting mechanisms for the various synoptic
patterns, I feel that some additional research is possible in this area. I used daily COOP
precipitation data and NARR data only at 2100 UTC. However, data for some COOP
stations are available hourly and NARR data are currently available every 3 hours. A
more in-depth analysis, perhaps also including a study of vorticity and divergence, may
provide a more thorough depiction of the lifting that occurs with each synoptic system.

Fifth, not only did my temperature results suggest warming trends during the past
few decades, but some of my results also pointed toward decreasing ranges in diurnal
temperature, as well as in the annual temperature cycle. Because my expertise is not in

plant or animal biology, I was unable to adequately cover the potential impacts that the

167

trends in climate have had or could have on the biogeography of the region. Vegetation
changes have been documented, however (e.g., Mark, 1958; Allen and Kupfer, 2000;
2001; Delcourt and Delcourt, 2001; Delcourt, 2002), and I hope that additional
biogeographical studies of the region's species can shed some light on the potential
impacts that changes in temperature range may have on vegetation. In this study,
decreasing temperature ranges were most obvious at high elevations with trends often
being statistically significant at the p S 0.01 level, making it crucial to understand
whether or not the ecological sky islands of the region are indeed shrinking as others
have suggested (Allen and Kupfer, 2000; 2001; Delcourt and Delcourt, 2001; Delcourt,
2002). Delcourt and Delcourt (2001) and Delcourt (2002) discussed sky island shrinkage
primarily in terms of temperature and projected that the region's boreal forest could
disappear by 2100 or even earlier. However, changes in temperature and temperature
range are likely not the only factors that can contribute to or hinder sky island shrinkage.
Kupfer and Cairns (1996) suggested that precipitation, soil depth, soil acidity, soil
moisture, wind, growing season length, frost-free season length, air pollution, and insect
pests are just some of the additional variables that may play a role in whether or not the
sky islands of the region will shrink.

Additional areas of research may include more substantial studies of drought,
wildfires, or frequency of temperature and precipitation extremes in the region.
Decreased precipitation and warmer conditions over the most recent years have coincided
with an increased frequency of natural and human-caused fires (Great Smoky Mountains
Association, 2007; 2009). Modeling studies may also be able to create projections

regarding the risk of fires in the region. If the studies find that fire risk is expected to

168

increase, then local decision makers will have the opportunity to prepare in advance by
increasing the number of fire towers, increasing personnel, or purchasing additional

equipment, if they choose to do 30.

7.4. Final Remarks

Within current discussions of climate change, the southern Appalachian region
has gained little attention compared to other regions around the world. In this study, I
have provided new evidence indicating that this region has experienced statistically
significant temporal trends in temperature, precipitation, and snowfall during decades of
the recent past (1931-2006). In particular, I determined that the diurnal and annual
ranges of temperature decreased over the period and that significant trends were most
common at the highest elevations (_>_ 1000 m). Statistically significant increases in high-
elevation snowfall also occurred. An objective cluster analysis yielded five common
climate regimes for temperature in the region. Fewer than five regimes were common for
precipitation and snowfall, although all variables exhibited a relationship to the region's
topography. My review of precipitation events over 1979—2006 was more inclusive of
smaller events than past studies and revealed up to 11 different synoptic origins of
precipitation for. the region. A comparison of objective, subjective, and hybrid clustering
approaches revealed differing results, suggesting that some uncertainty still remains.
However, of those synoptic patterns found, some have experienced statistically
significant trends in frequency, which may shed some light on the region's future
precipitation trends.

As I have discussed, the significant trends revealed by this study have coincided

with negative changes in the region's biodiversity, although the causal connections among

169

these changes remains unclear. Once diminished, the species richness of the area may not
return. If the region's species diversity is something we value as a society, we need to
understand the processes that affect it, such as natural and anthropogenic climate change.
It is my hope that with this dissertation I have built a stronger foundation from which

researchers can move forward in that endeavor.

170

APPENDIX A

ADDITIONAL RESULTS FROM THE TEMPORAL ANALYSIS

A.1. Temporal Results of Temperature

A.I.l. February

 

 

   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Low Elevations Mid Elevations
g 20 fit '—'i I l ‘ I g 20 I 7 ' l l V ' l
E - E -
0 a)
O O
.E ' .E '
2 - 8 '
3 3
E - g _
d) d)
o. 0- _
§ ' E
r— —10 . . . l 1— —10 . . . .
1940 1960 1980 2000 1940 1960 1980 2000
Years Years
High Elevations All Elevations
.3 2° ' ' .3 2° ' '
0) ~ _ (I) _
8 15 g 15
.E _ .E 10- _
8 - 8 5 -
3 3
E - E o -
a a
s ' s '5‘ '
i— _10 , . . i— —10 . . . l .
1940 1960 1950 2000 1940 1960 1980 2000
Years Years

Figure A. 1. February minimum (bottom line), maximum (top line), and mean (center
line) temperature in °C. A linear fit was calculated for each variable over 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 - 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

171

A. 1.2. March

Low Elevations

 

 

 

 

 

 

 

 

 

 

3 25 ' 3
.6 _ .5
76 T)
0 O
.E .E
8 8
3 ' 3
E _ E
O 0
Q C).
E - E
0 V
H h
1940 1960 1950 2000
Years
High Elevations
a) 25 ' r ' v7
3 3
.6 _ _ .(7)
6 E
o o
.E .E
e a
3 3
+4 +4
E E
0 0
Q. CL
8 s
13 -10 . . . . mm . . l—
1940 1960 1980 2000
Years

—10

 

Mid Elevations

 

 

 

 

l I I I

 

1940 1960 1980 2000
Yea rs

All Elevations

 

 

 

 

1940 1960 1980 2000
Yea rs

/4.

Figure A.2. March minimum (bottom line), maximum (top line), and mean (center line)
temperature in °C. A linear fit was calculated for each variable over 1931-2006. Results

are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top

right), high elevations (2 1000 m; bottom left), and all elevations (bottom right).

172

A.I.3. May

Low Elevations Mid Elevations
35 . . . . . , .

 

 

 

 

 

 

Temperature in Celsius
N
0
Temperature In Celsrus
N
o

 

 

15* ‘ 15 i
,OMWWWWMMM- lows/Mum-
5 - - n - 5 - . L i '
1940 1960 1980 2000 1940 1960 1930 2000
Years Years
High Elevations All Elevations

 

 

 

 

 

 

Temperature in Celsms
N
o
5 1
l I 1
Temperature in Celsius
N
o
l g: l
l l

 

 

15 Ari/WW 4
10 W‘ 10 WW-
5 I I 5 l r 1 4
1940 1960 1980 2000 1940 1960 1980 2000
Years Years

Figure A.3. May minimum (bottom line), maximum (top line), and mean (center line)
temperature in °C. A linear fit was calculated for each variable over 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

173

A.I.4. June

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Low Elevations Mid Elevations

.3 35 l I I I g 35 ' I 'fi " ‘I V V ' r I
5 ”WW“ 5 25- i
3 20‘ ‘ 2 EW-
E 15» - E 15W-
3 3
g 10- - g 10- -
E 5 l . . J 11’ 5 . . . A .

1940 1960 1950 2000 1940 1960 1980 2000

Years Years
High Elevations All Elevations

m 35 ' ' m 35 ' ' ' ' '
a 2 -
(I! — _ _
3 5° § ”WWW/W
E 25W_ '5 25W:
3 20' - 1'3 20- -,
E 15' - E 15 —,
3 WW 3 i
g 1oL - g 10- -.
11’ 5 . .4 . . +9 5 - mm . . . ‘

1940 1960 1980 2000 1940 1960 1930 2000

Years Years 1,],

Figure A.4. June minimum (bottom line), maximum (top line), and mean (center line)
temperature in °C. A linear fit was calculated for each variable over 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

174

A. 1.5. August

Temperature in Celsius

Temperature in Celsius

35
30

25
2O

15
10

Low Elevations

 

 

 

 

1940

1960

19813

Yea rs

2000

 

 

High Elevations

 

 

1940

1960

1 980

Yea rs

2000

Temperature in Celsius

Temperature in Celsius

35
30

25
20

10

35
30
25
20

15
10

Mid Elevations

 

 

 

 

1940 [1960 1980 2000
Years

 

All Elevations
W ,

1

 

 

r I_L J_._

 

1940 1960 1980 2000
Years

/.

Figure A.5. August minimum (bottom line), maximum (top line), and mean (center line)
temperature in °C. A linear fit was calculated for each variable over 1931-2006. Results

are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top

right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

A. 1. 6. September

Low Elevations Mid Elevations

 

 

 

 

 

 

 

 

g 35 I l filfi’ g 35 I I V V I ' ' I

g 30. - g 30 - -

o 0 WWW

E 25 - - 5 25 -

E 15W— E 15— -

a a W 2

g 10 - ' g 10 ' g

Q) 0

l— 5 r I r r l— 5 r r r ._g
1940 1960 1980 2000 1940 1960 1930 2000

Years Years

All Elevations

 

 

High Elevations

Temperature in Celsms
N] N
o 01
5
I I
Temperature in Celsms
N N U
0 on o
g g
A l I I

 

 

 

 

 

 

1940 1960 1950 2000 194-0 1960 1980 2000
Years Years ,4

Figure A.6. September minimum (bottom line), maximum (top line), and mean (center
line) temperature in °C. A linear fit was calculated for each variable over 1931-2006.
Results are included for low elevations (< 500 at; top left), mid elevations (500 — 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

176

A. I. 7. November

Low Elevations

Mid Elevations
25 ' ' ' ' ' '

 

 

 

 

 

 

Temperature In Celsrus
O 01

Temperature in Celsius
O 01

—5 I I I I
1940 1960 1980 2000 1940 1960 1930 2000
Years Years
High Elevations All Elevations
. . 25 . . . .

 

 

 

 

20' ‘
15 '
10W

—5 4 I I I

 

 

 

 

Temperature in Celsrus
O 01
I l

Temperature in Celsius
O 01
g;
l I

—5 . . . .
1940 1960 1980 2000 1940 1960 1980 2000
Years Years .4;

 

 

Figure A.7. November minimum (bottom line), maximum (top line), and mean (center
line) temperature in °C. A linear fit was calculated for each variable over 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

177

A. 1. 8. December

Low Elevations Mid Elevations

l ' r

 

 

    
  

 

 

 

 

 

Temperature in Celsius
Temperature in Celsius

     

 

 

 

 

—10 I I I
1940 1960 1950 2000 1940 1.960 1950 2000
Years Years
High Elevations All Elevations
20 ' ' ' ' 20 ' ' '
15 — - 15 r -

_5— ‘ _

I I
Temperature in Celsrus
U"! o
r
I I . .

 

 

 

 

Temperature in Celsius

 

 

 

-10 I I t ;_L I I --10 I L A I l I
1940 1960 1980 2000 1940 1960 1980 2000
Years Years 4,,

Figure A.8. December minimum (bottom line), maximum (top line), and mean (center
line) temperature in °C. A linear fit was calculated for each variable over 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

178

A.l. Temporal Results of Precipitation and Snowfall

A.2.1. February

Low Elevations

 

400

I500 '

200 '

100

Precipitation in mm

 

I

l l

I l

L

 

 

1940

1960 1980
Yea rs

High Elevations

2000

 

400'

300 '

200'

100

Precipitation in mm

 

I

I

 

 

Figure A.9. February precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m
top right), high elevations (2 1000 m; bottom left), and all elevations (bottom right).

1940

1960 1980
Years

2000

Precipitation in mm

Precipitation in mm

179

400

300 '

200 '

100

300 '

200 '

100

Mid Elevations

 

 

 

 

 

 

 

 

1940 1960 1980 2000
Years
All Elevations
1940 1960 1980 2000
Years

3

 

Low Elevations

 

 

 

 

 

 

 

 

Mid Elevations

 

 

 

 

 

 

 

 

iooo ' ' ' ' 1000 ' ' ' n '

E 800 E 300 - J
E E

.E 600 ' .E 600 - ‘

E :3 '

3 3 400 ~ -

2 2 ‘

(n (n 200 ' r

o ‘ N l

1940 1960 1930 2000 1940 1960 1930 2000
Years Years
High Elevations All Elevations
1000 ' ' 1000 ' '

l

E 800 E 800 - 1

E E «

.E 600 .E 500 - ‘1

E E i

3 400 3 400 - 1

2 8 i

(n 200 m 200 M «

O I I 4 I O i V I ' .

1940 1960 1980 2000 194-0 1960 1980 2000
Years Years

.‘\\

Figure A.10. February snowfall in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

180

A. 2. 2. March

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Low Elevations Mid Elevations
400 l I ' l I

E E

E E

«E .E

C C

.9 .9

-I-I +4

B B

.5 ‘a

.6 .6

B E

[L D.

0 I I I L O I I I I
1940 1960 1980 2000 1940 1960 1980 2000
Years Years
High Elevations All Elevations
400 ' ' 400 ' '
E E
300 '

.E .E

s s

1.: ' ‘p 200 -

B B

:53 E

g E 100

D. [L

o I I I . I O L l l I L+
1940 1960 1980 2000 1940 1960 1980 2000
Years Years

Figure A. l 1. March precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

181

 

 

Low Elevations Mid Elevations

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1000 1000 r T '

E 800+ ~ E 800 - _
E E .
.s 500 _ 1 .E 500 - -
fi e

3; 400 - a «i 400 - q
8 8 l
m 200 - A l m 200 - i

o MAE/t . ‘M l A A o 1 MA
1940 1960 1980 2000 1940 1960 1980 2000
Years Years
High Elevations All Elevations
1ooo ' ' iooo ' '

E 800 E 800 - -
E E

.E 500 .E -
=0 e

'5 400 *5 -
2 2

(n 200 (n L» _

0 I I I l I MAW
1940 1960 1930 2000 1940 1960 1980 2000
Years Years

I
//,

Figure A. 12. March snowfall in mm. A linear fit was calculated for 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (2 1000 m; bottom left), and all elevations (bottom right).

182

AZ. 3. May

400

100

Precipitation in mm

400

Precipitation in mm

Low Elevations

 

300 '

200 '

 

L

 

 

1940

1960 1950
Yea rs

2000

 

300 '

200'

100'

 

High Elevations

I I

 

 

1940

1960 1980
Years

2000

Precipitation in mm

Precipitation in mm

Mid Elevations

 

 

 

 

 

 

 

 

400 I l ' V I ' T
500 -
200 '
100
O I I I I
1940 1960 1980 2000
Years
All Elevations
400 ' ' T
300 -
200'
100
0 I I I l
1940 1960 1980 2000
Years

Figure A.13. May precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top lefi), mid elevations (500 —— 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

183

Low Elevations Mid Elevations

 

 

 

 

 

 

 

 

500 E I I ' I ' I ' g 500 E T I ' r T Ii' ' I I ' i

E 400% 4 E 4003 i

E i a E

.5 5003 j .E 5003 i

1% E i E E i

'3 200; g -; 200; i
3 8

w 100 ;- -g u) 100 E' _

O E I I . . I I ....... J J i O E ..... I . I I I A Mr L . i

1940 1960 1980 2000 1940 1960 1980 2000
Years Years

High Elevations All Elevations

 

 

 

 

 

 

 

 

500g 500; "r" ' g

E 400i i E 400} g
-E 300? j .5 5003 -
‘5 i i ‘e E 5
“5 200%- ': “5 200?- -
3 E a 3 E i
l” 100? j in 1003 g
0 EA . I A .A a , A - _: - A . ‘_ i 0 E . n A ‘ A . 4 g L i A A 44 i

1940 1960 1950 2000 1940 1960 1950 2000
Years Years

Figure A.14. May snowfall in mm. A linear fit was calculated for 1931-2006. Results
are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m; top
right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

184

A. 2. 4. June

Low Elevations

 

 

 

 

 

400 ' ‘ ' ' ' '
E
E
.E
C
.9
+4
3
'E
'6
8
D.

0 I I I l

1940 1960 1980 2000
Years
High Elevations
400 ' '

E
E
.E
C
.9
E
‘a
'6
E
D.

0 I I I I

 

 

 

1940 1960 1980 2000
Years

Figure A. l 5. June precipitation in mm. A linear fit was calculated for 1931-2006.

Precipitation in mm

Precipitation in mm

Mid Elevations

 

 

 

 

 

 

 

 

400 ' '
300 '
200 '
100
O I I I I
1940 1960 1950 2000
Years
All Elevations
400 ' '
0 I L J I
194-0 1960 1950 2000
Years

[4'

Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;

top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

185

A. 2. 5. August

Low Elevations

 

 

 

 

 

 

 

 

 

500 l l I l
E
E 400 -
.E
c 500 '
.9
+4
_‘é’ 200 -
.9
0
g 100 '
D.
0 I L I I
1940 1960 1980 2000
Years
High Elevations
500 ' '
E
E
.E
C
.9
3c}
'6
‘5
B
D.
0 I I I I
1940 1960 1980 2000
Years

Precipitation in mm

Precipitation in mm

500

400 '
500 '
200 '
100 '

500

400 '
300 '
200 '

100'

Mid Elevations

 

 

I I v I ' I

 

I I I I

 

1940 1960 1980 2000
Years

All Elevations

 

 

 

J L L

 

1940 1960 1980 2000
Years

Figure A.16. August precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

186

A. 2. 6. September

Low Elevations Mid Elevations

 

 

 

 

 

   

 

 

 

 

   

500 ' ' ' 5007
E E
E 400' E
.E .E
I: 500' I:
.2 .9
+w +I
g 200- ,3
.9 .9
‘9’ 100— §
0. D.
0 I I I I o I I I I
1940 1960 1950 2000 1940 1960 1950 2000
Years Years
High Elevations All Elevations

500 ' ' 500 - .
E E
E E
E .E
I: C
.2 .9
E E
‘a .5
.5 .6
2 E
(L Q.

o I I I O J J

 

 

 

 

 

 

1940 1960 1980 2000
Yang Yars fl

1940 1960 1980 2000

Figure A. 17. September precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 —- 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

187

A2 7. November

Low Elevations

 

 

 

 

 

 

 

 

Mid Elevations

 

 

 

 

 

 

 

 

400 ' ' ' ' 400 '
300 '
E .E
C C
.g .5-3 200 -
B B
‘a ‘5.
'6 ‘6 100
8 E
[L 0.
o I I I I O I I I I
1940 1960 1950 2000 194—0 1960 1950 2000
Years Years
High Elevations All Elevations
400’ ' ' 400 ' '
500-
E .E
s s
.5 .5 200-
B 3
IE it
§ § 100
[1. EL
0 I I I I 0 I I I I
1940 1960 1980 2000 1940 1960 1980 2000
Years Years

'2

Figure A. 1 8. November precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

188

500

Snowfall in mm

100

500

Snowfall in mm

Low Elevations

 

400 '
300 '
200 '

' I I l I

 

M: M -A_.A_‘_I A

 

 

1940 1960 1980 2000
Years

 

400 '
500 '
200 '

100'

High Elevations

 

I I

 

 

1940 1960 1950 2000
Years

Snowfall in mm

Snowfall in mm

500

400 '
500 '

200
‘I 00

500

400 '

500 '

200 '

100

 

Mid Elevations

I I 'T'i 1+

 

 

mail __.-'

 

194—0 1960 1980 2000
Years

 

All Elevations

 

 

AAML .AAA. AA

 

1940 1960 1980 2000
Years

Figure A.19. November snowfall in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

189

A. 2. 8. December

Precipitation in mm

Precipitation in mm

Low Elevations

 

400

 

I

I

'fir'

I

I

 

I

 

 

1940

1960

1980

Yea rs

High Elevations

2000

 

400

500'

200

100'

 

I

I

I

 

 

1940

1960

1980

Yea rs

2000

Precipitation in mm

Precipitation in mm

Mid Elevations

 

400

300'

200 '

100'

 

I

 

I

 

1940

1960

1950

Yea rs

All Elevations

2000

 

400

500 '

200'

100'

 

 

 

1940

1960

1980

Yea rs

2000

Figure A.20. December precipitation in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 m;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

190

 

 

Low Elevations Mid Elevations

 

   

 

 

    

 

 

 

 

700' fi ' fi '7‘ I I 700 V I I I
E ' E
E E
.E .E
E E
p 3
O O
C C
m w
1940 1960 1980 2000 1940 1960 1980 2000
Years Years
High Elevations All Elevations
700 ' ' ' ' 700 ' ' ' '
600
E E
E 500 E
.E 400 .E
E 300 E
3 z
I?) 200 5
100
O r . .
1940 1960 1980 2000 1940 1960 1980 2000
Years Years

Figure A.21. December snowfall in mm. A linear fit was calculated for 1931-2006.
Results are included for low elevations (< 500 m; top left), mid elevations (500 — 999 111;
top right), high elevations (Z 1000 m; bottom left), and all elevations (bottom right).

191

APPENDIX B

ADDITIONAL RESULTS FROM THE SPATIAL ANALYSIS

B.1. Spatial Results of Temperature

 

 

 

 

 

8.1.]. April
Elevation (m)
36N
35N I:
N

 

I . :
84W 83W 8 2W

Figure B]. April during 1935-1944 with 5 clusters for mean temperature, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Mean temperature values for Clusters 1-5 were
4.24, 9.65, 15.48, 12.47, and 12.16°C. Elevation in meters is shaded in gray.

192

 

Elevation (m)

36N

 

35N

 

 

 

 

84W 83W 82W

.é

Figure B.2. April during 1965-1974 with 5 clusters for mean temperature, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Mean temperature values for Clusters 1-5 were
16.22, 15.06, 14.11, 12.89, and 10.16°C. Elevation in meters is shaded in gray.

193

 

Elevation (m)

36N '

 

35N

 

 

 

N

 

84W 83W 82W

Figure B.3. April during 1995-2004 with 5 clusters for mean temperature, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Mean temperature values for Clusters l-5 were
14.41, 12.88, 15.68, 10.81, and 6.19°C. Elevation in meters is shaded in gray.

194

B. I. 2. October

 

 

 

 

 

84W 83W 82W

[I

Figure B.4. October during 1935-1944 with 5 clusters for mean temperature, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Mean temperature values for Clusters 1-5 were

7.79, 17.06, 13.99, 15.73, and 11.64°C. Elevation in meters is shaded in gray.

195

 

Elevation (m)

36N ‘

35N

 

 

84W 83W 82W

.4

Figure B.5. October during 1965-1974 with 5 clusters for mean temperature, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Mean temperature values for Clusters 1-5 were

16.29, 13.80, 14.93, 10.29, and 12.62°C. Elevation in meters is shaded in gray.

196

 

Elevation (m)

  

 

 

Figure 8.6. October during 1995-2004 with 5 clusters for mean temperature, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Mean temperature values for Clusters 1-5 were

13.66, 11.59, 15.02, 16.27, and 833°C. Elevation in meters is shaded in gray.

197

B.2. Spatial Results of Precipitation

 

 

B. 2. 1. April
Elevation (m)
36N '
35N f
N

 

 

 

 

84W 83W 82W

Figure 8.7. April during 1935-1944 with 5 clusters for total precipitation, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Precipitation values for Clusters 1-5 were 105.42,
147.59, 120.09, 89.45, and 74.75 mm. Elevation in meters is shaded in gray.

198

 

Elevation (m)

  

36N E '

35N

 

 

A”.

Figure B.8. April during 1965-1974 with 5 clusters for total precipitation, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Precipitation values for Clusters 1-5 were 85.44,
111.18, 98.80, 138.72, and 120.93 mm. Elevation in meters is shaded in gray.

199

 

Elevation (m)

  

36N :27

 

 

35N , ' . .- ;,: ............. g ......... :

 

84W 83W 82W

xx?

Figure B.9. April during 1995-2004 with 5 clusters for total precipitation, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Precipitation values for Clusters 1-5 were 170.35,
103.61, 87.67, 134.18, and 120.04 mm. Elevation in meters is shaded in gray.

200

B. 2. 2. October

 

Elevation (m)

   

 

 

 

84W 83W 82W

./

Figure B. 10. October during 1935-1944 with 5 clusters for total precipitation, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Precipitation values for Clusters 1-5 were 77.80,

87.15, 54.87, 63.34, and 108.94 mm. Elevation in meters is shaded in gray.

201

 

 

 

 

 

Elevation (m)

 

é

Figure B.11. October during 1965-1974 with 5 clusters for total precipitation, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Precipitation values for Clusters 1-5 were
105.25, 158.56, 88.71, 125.91, and 72.51 mm. Elevation in meters is shaded in gray.

202

 

Elevation (m)

  

 

 

 

84W 83W 82W

,4

Figure B.12. October during 1995-2004 with 5 clusters for total precipitation, where
Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster
5 (diamond) represent station locations. Precipitation values for Clusters 1-5 were 49.77,

77.14, 64.73, 97.19, and 120.84 mm. Elevation in meters is shaded in gray.

203

B.3. Spatial Results of Snowfall

 

 

 

 

B. 3. 1. April
Elevation (m)
35N I:
N

 

 

A?

Figure B.13. April during 1935-1944 with 5 clusters for total snowfall, where Cluster 1
(dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Snowfall values for Clusters 1-5 were 77.22,
26.42, 61.72, 7.50, and 0.12 mm. Elevation in meters is shaded in gray.

204

Elevation (m)

36M 2:. '

 

35N

 

 

 

 

84W 83W W

Figure B.14. April during 1965-1974 with 5 clusters for total snowfall, where Cluster 1
(dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Snowfall values for Clusters 1-5 were 5.93, 16.26,
10.24, 33.66, and 0.19 mm. Elevation in meters is shaded in gray.

205

 

Elevation (m)

 

35N

 

 

 

84W 83W 82W

Figure 8.15. April during 1995-2004 with 5 clusters for total snowfall, where Cluster 1
(dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Snowfall values for Clusters 1-5 were 127.5],
12.71, 0.11, 78.99, and 212.26 mm. Elevation in meters is shaded in gray.

206

B. 3. 2. October

 

Elevation (m)

   

 

 

 

84W 83W 82W

A.

Figure B. 16. October during 1935-1944 with 5 clusters for total snowfall, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Snowfall values for Clusters 1-5 were 17.02, 5.08,
30.48, 7.62, and 0.00 mm. Elevation in meters is shaded in gray.

207

 

Elevation (m)

36N ,_,

 

35N

 

 

 

 

,4

Figure B.17. October during 1965-1974 with 5 clusters for total snowfall, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Snowfall values for Clusters 1-5 were 0.00, 0.76,
6.86, 0.00, and 0.00 mm. Elevation in meters is shaded in gray.

208

 

Elevation (m)

 

 

 

 

 

84W 83W 82W

.4

Figure 3.18. October during 1995-2004 with 5 clusters for total snowfall, where Cluster
1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4 (triangle), and Cluster 5
(diamond) represent station locations. Snowfall values for Clusters 1-5 were 1.27, 0.00,
0.00, 0.00, and 0.00 mm. Elevation in meters is shaded in gray.

209

B.4. Spatial Results for All Variables

 

 

 

 

 

B. 4. 1. April
Elevation (m)
36N i
35N :
N

 

Figure B. 19. April during 1935-1944 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation, and total
snowfall), where Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4

(triangle), and Cluster 5 (diamond) represent station locations. Minimum temperature
values for Clusters 1-5 were 4.72, 8.06, 6.98, 5.48, and 543°C. Maximum temperature
values for Clusters 1-5 were 19.46, 22.82, 18.47, 20.93, and 19.96°C. Mean temperature
values for Clusters 1-5 were 12.09, 15.44, 12.73, 13.21, and 12.71°C. Precipitation
values for Clusters 1-5 were 107.66, 116.73, 115.90, 82.59, and 133.06 mm. Snowfall
values for Clusters 1-5 were 8.48, 0.08, 15.49, 0.29, and 10.29 mm. Elevation in meters

is shaded in gray.

210

 

  

Elevation (m)

36N

35N

 

 

 

Figure B.20. April during 1965-1974 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation, and total
snowfall), where Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4

(triangle), and Cluster 5 (diamond) represent station locations. Minimum temperature
values for Clusters 1-5 were 8.41 , 5.46, 6.98, 4.72, and 6.60°C. Maximum temperature
values for Clusters 1-5 were 23.17, 20.10, 21.86, 21.34, and 20.51"C. Mean temperature
values for Clusters 1-5 were 15.81, 12.79, 14.43, 13.04, and 13.57°C. Precipitation
values for Clusters l-5 were 94.35, 87.88, 131.72, 111.89, and 114.70 mm. Snowfall
values for Clusters 1-5 were 0.09, 1.11, 0.66, 1.09, and 6.63 mm. Elevation in meters is
shaded in gray.

211

 

  

Elevation (m)

 

 

 

84W 83W 82W

Figure 8.2]. April during 1995-2004 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation, and total
snowfall), where Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4

(triangle), and Cluster 5 (diamond) represent station locations. Minimum temperature
values for Clusters 1-5 were 7.07, 6.77, 6.29, 4.53, and 6.03°C. Maximum temperature
values for Clusters 1-5 were 20.97, 21.69, 21.53, 20.28, and 18.36°C. Mean temperature
values for Clusters 1-5 were 14.03, 14.24, 13.93, 12.42, and 12.21°C. Precipitation
values for Clusters 1-5 were 103.58, 99.20, 121.35, 111.15, and 121.20 mm. Snowfall
values for Clusters 1-5 were 2.62, 0.97, 0.00, 3.06, and 41.38 mm. Elevation in meters is
shaded in gray.

212

B. 4. 2. October

 

Elevation (m)

   

 

 

 

. : N
83W 82W ,
Figure 822. October during 1935-1944 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation, and total
snowfall), where Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4
(triangle), and Cluster 5 (diamond) represent station locations. Minimum temperature
values for Clusters 1-5 were 7.83, 9.07, 8.26, 6.79, and 6.62°C. Maximum temperature
values for Clusters 1-5 were 23.82, 21.30, 23.75, 22.08, and 19.65°C. Mean temperature
values for Clusters 1-5 were 15.83, 15.19, 16.01, 14.50, and 13.13°C. Precipitation
values for Clusters 1-5 were 58.77, 80.51, 80.82, 68.08, and 91.69 mm. Snowfall values
for Clusters 1-5 were 0.00, 3.40, 0.00, 5.08, and 2.12 mm. Elevation in meters is shaded
In gray.

213

 

Elevation (m)

 

 

 

 

 

,4:
Figure 323. October during 1965-1974 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation, and total
snowfall), where Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4
(triangle), and Cluster 5 (diamond) represent station locations. Minimum temperature
values for Clusters 1-5 were 7.09, 9.16, 5.07, 8.27, and 580°C. Maximum temperature
values for Clusters 1-5 were 22.99, 22.64, 21.00, 21.59, and 19.12°C. Mean temperature
values for Clusters 1-5 were 15.05, 15.92, 13.05, 14.94, and 12.45°C. Precipitation
values for Clusters 1-5 were 89.19, 116.47, 84.67, 81.53, and 137.00 mm. Snowfall
values for Clusters 1-5 were 0.00, 0.00, 0.00, 0.00, and 0.54 mm. Elevation in meters is
shaded in gray.

214

 

Elevation (m)

36N . '

 

35N

 

 

 

 

84W 83W 82W

Figure 824. October during 1995-2004 with 5 clusters for all variables (minimum
temperature, maximum temperature, mean temperature, total precipitation, and total
snowfall), where Cluster 1 (dot), Cluster 2 (plus sign), Cluster 3 (square), Cluster 4
(triangle), and Cluster 5 (diamond) represent station locations. Minimum temperature
values for Clusters 1-5 were 8.17, 6.84, 6.94, 8.59, and 597°C. Maximum temperature
values for Clusters 1-5 were 22.63, 21.87, 21.08, 20.63, and 19.52°C. Mean temperature
values for Clusters 1-5 were 15.42, 14.36, 14.03, 14.63, and 12.76°C. Precipitation
values for Clusters 1-5 were 76.96, 61.28, 59.00, 91 .53, and 100.66 mm. Snowfall values
for Clusters 1-5 were 0.00, 0.00, 0.00, 0.09, 0.00 mm. Elevation in meters is shaded in
gray.

215

APPENDIX C

GRIB FILES

C.1. Automation of GRIB downloads

I used the following sample script to download NARR data for the month of
January, 1979 at the 2100 UTC time period. The variables HGT, TMP, SPFH, VVEL,
UGRD, and VGRD were extracted for the 850 and 700 hPa levels. HGT, UGRD, and
VGRD were extracted for the 300 hPa level. All files are saved into a folder named
“1979/” located in the working directory. I created similar scripts with minor edits to

download data for other months and years.

#llbin/sh

# download selected variables from NARR 19790101-002 merged-a data from NCDC
yymm=197901

date=l9790101

hr=21

while [ $date != "19790132" ]

do
a="http://nomads.ncdc.noaa.gov/data/narr/$yymm/$date/narr-a_221_${date}_${hr} OO_OOO"
b="narr-a_221_${date}_${hr}00"

./get_ inv. pl $a. inv | egrep " :(HGTITMPISPFH|VVEL|UGRD|VGRD):850 mb" |
./get_grib. p1 $a. grb 1979/${b}_ 850mb. grb

./get_ inv. pl $a. Inv | egrep " ::(HGT|TMP|SPFHIVVELIUGRDIVGRD) 700 mb" |
./get_gr1bpl $a. grb l979/${b}_700mb. grb

./get_inv.pl $a.inv | egrep ":(HGT|UGRD|VGRD):3OO mb" 1

./get_grib.pl $a.grb 1979/${b}_300mb.grb

date=‘expr $date + 1‘

echo "$date"

done

echo "finished downloading variables for $yymm"

C.2. Uncompressing GRIB files using WGRIB

Instructions for downloading, compiling, and using WGRIB are located at:

216

hgp://www.cpc.ncep.noaa.gov/products/weslev/wgrib.html. Once I downloaded and
compiled WGRIB, I used the following command to uncompress 850 hPa data for

January 1, 1979.

./wgrib.x 1979/narr-a_22 l_|9790101_2100_850mb.grb -d all -text —h -o 1979/narrl9790101_850mb.txt

The command uncompresses the file and saved it as a .txt formatted file. I used similar

commands with minor edits to uncompress files for other days, months, years, and levels.

217

APPENDIX D

ADDITIONAL RESULTS FROM THE SYNOPTIC CLIMATOLOGY

D.l. Maps of 700 and 300 hPa Variables from the Objective Clustering Approach

 

Figure D.1. Cluster 1. Maps in the top row from left to right are temperature (contoured
every 5°C) wind vectors, dewpoint (contoured every 5°C), and height (contoured every 30
m) at the 700 hPa level. Maps in the bottom row from left to right are height (contoured
every 120 m), wind vectors, and isotachs (contoured every 10 m/s) at the 300 hPa level.

218

 

Figure D.2. Cluster 2. Maps in the top row from left to right are temperature (contoured
every 5°C) wind vectors, dewpoint (contoured every 5°C), and height (contoured every 30
m) at the 700 hPa level. Maps in the bottom row from left to right are height (contoured
every 120 m), wind vectors, and isotachs (contoured every 10 m/s) at the 300 hPa level.

 

Figure D.3. Cluster 3. Maps in the top row from left to right are temperature (contoured
every 5°C) wind vectors, dewpoint (contoured every 5°C), and height (contoured every 30
m) at the 700 hPa level. Maps in the bottom row from left to right are height (contoured
every 120 m), wind vectors, and isotachs (contoured every 10 m/s) at the 300 hPa level.

219

 

Figure D.4. Cluster 4. Maps in the top row from left to right are temperature (contoured
every 5°C) wind vectors, dewpoint (contoured every 5°C), and height (contoured every 30
m) at the 700 hPa level. Maps in the bottom row from left to right are height (contoured
every 120 m), wind vectors, and isotachs (contoured every 10 m/s) at the 300 hPa level.

 

Figure D.5. Cluster 5. Maps in the top row from left to right are temperature (contoured
every 5°C) wind vectors, dewpoint (contoured every 5°C), and height (contoured every 30
m) at the 700 hPa level. Maps in the bottom row from left to right are height (contoured
every 120 m), wind vectors, and isotachs (contoured every 10 m/s) at the 300 hPa level.

220

 

Figure D.6. Cluster 6. Maps in the top row from left to right are temperature (contoured
every 5°C) wind vectors, dewpoint (contoured every 5°C), and height (contoured every 30
m) at the 700 hPa level. Maps in the bottom row from left to right are height (contoured
every 120 m), wind vectors, and isotachs (contoured every 10 m/s) at the 300 hPa level.

 

Figure D.7. Cluster 7. Maps in the top row from left to right are temperature (contoured
every 5°C) wind vectors, dewpoint (contoured every 5°C), and height (contoured every 30
m) at the 700 hPa level. Maps in the bottom row from left to right are height (contoured
every 120 m), wind vectors, and isotachs (contoured every 10 m/s) at the 300 hPa level.

221

 

Figure D.8. Cluster 8. Maps in the top row from left to right are temperature (contoured
every 5°C) wind vectors, dewpoint (contoured every 5°C), and height (contoured every 30
m) at the 700 hPa level. Maps in the bottom row from left to right are height (contoured
every 120 m), wind vectors, and isotachs (contoured every 10 m/s) at the 300 hPa level.

 

Figure D.9. Cluster 9. Maps in the top row from left to right are temperature (contoured
every 5°C) wind vectors, dewpoint (contoured every 5°C), and height (contoured every 30
m) at the 700 hPa level. Maps in the bottom row from left to right are height (contoured
every 120 m), wind vectors, and isotachs (contoured every 10 m/s) at the 300 hPa level.

222

 

Figure D.10. Cluster 10. Maps in the top row from left to right are temperature
(contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and height
(contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to right
are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10 m/s)
at the 300 hPa level.

 

Figure D.11. Cluster 11. Maps in the top row from left to right are temperature
(contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and height
(contoured every 30 m) at the 700 hPa level. Maps in the bottom row from lefi to right
are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10 m/s)
at the 300 hPa level.

223

D.2. Maps of 700 and 300 hPa Variables from the Subjective Clustering Approach

 

Figure D. 12. Closed low (CL). Maps in the top row from left to right are temperature
(contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and height
(contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to right
are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10 m/s)
at the 300 hPa level.

224

 

Figure D.13. Northerly, behind low, no fronts (NbLnF). Maps in the top row from left to
right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row
from left to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level.

 

Figure D.14. Southerly, ahead of low, no fronts (SaLnF). Maps in the top row from left
to right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row
from left to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level.

225

 

Figure D.15. Northerly, no law, no fronts (N nLnF ). Maps in the top row from left to
right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row
from lefi to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level.

 

Figure D. 16. Southerly, no low, no fronts (SnLnF). Maps in the top row from lefi to
right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row

from left to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level.

226'

 

Figure D.17. Bermuda High (BH). Maps in the top row from left to right are
temperature (contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and
height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to
right are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10
m/s) at the 300 hPa level.

 

 

Figure D.18. North-south oriented front (nsF). Maps in the top row from lefi to right are
temperature (contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and
height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to
right are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10
m/s) at the 300 hPa level.

227

 

Figure D.19. Northerly, east-west oriented front (NewF). Maps in the top row from left
to right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row
from left to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level.

 

Figure D.20. Southerly, east-west oriented front (SewF). Maps in the top row from left
to right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row
from left to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level.

228

 

Figure D.21. Zonal flow (Z). Maps in the top row from left to right are temperature
(contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and height
(contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to right
are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10 m/s)
at the 300 hPa level.

 

Figure D22. Weak flow (W). Maps in the top row from left to right are temperature
(contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and height
(contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to right
are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10 m/s)
at the 300 hPa level.

229

APPENDIX E

SYNOPTIC PATTERNS FROM THE HYBRID APPROACH

E.1. Maps of 850 hPa Variables from the Hybrid Clustering Approach

 

I I
‘ a - " '1530 - _ fist-Er 3:.) J»
a.)

«-15

Figure E. 1. Closed low (CL). Temperature contoured every 5°C (top left), dewpoint
contoured every 5°C (top right), height contoured every 30 at (bottom left), and wind
vectors (bottom right) at the 850 hPa level.

230

 

 

Figure B.2. Northerly, behind low, no fronts (NbLnF). Temperature contoured every 5°C
(top left), dewpoint contoured every 5°C (top right), height contoured every 30 m (bottom
left), and wind vectors (bottom right) at the 850 hPa level.

231

I

:5

L—I’
MS‘
reg-‘51}

 

Figure E.3. Southerly, ahead of low, no fronts (SaLnF). Temperature contoured every
5°C (top left), dewpoint contoured every 5°C (top right), height contoured every 30 m
(bottom left), and wind vectors (bottom right) at the 850 hPa level.

232

 

 

Figure B.4. Southerly, no low, no fronts (SnLnF). Temperature contoured every 5°C (top
left), dewpoint contoured every 5°C (top right), height contoured every 30 m (bottom
left), and wind vectors (bottom right) at the 850 hPa level.

233

 

 

._“‘a

75x? 1"
4‘ r ’

r45}

57 " ‘ ,
,5 ‘33.

 

 

Figure E.S. Bermuda High (BH). Temperature contoured every 5°C (top left), dewpoint
contoured every 5°C (top right), height contoured every 30 m (bottom left), and wind
vectors (bottom right) at the 850 hPa level.

234

 

Figure E.6. North-south oriented front (nsF). Temperature contoured every 5°C (top
lefi), dewpoint contoured every 5°C (top right), height contoured every 30 m (bottom
left), and wind vectors (bottom right) at the 850 hPa level.

235

 

 

Figure 13.7. Northerly, east-west oriented front (NewF). Temperature contoured every
5°C (top left), dewpoint contoured every 5°C (top right), height contoured every 30 m
(bottom left), and wind vectors (bottom right) at the 850 hPa level.

236

 

Figure E.8. Southerly, east-west oriented front (SewF). Temperature contoured every
5°C (top left), dewpoint contoured every 5°C (top right), height contoured every 30 m
(bottom left), and wind vectors (bottom right) at the 850 hPa level.

237

E.2. Maps of 700 and 300 hPa Variables from the Hybrid Clustering Approach

 

Figure E.9. Closed low (CL). Maps in the top row from lefi to right are temperature
(contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and height
(contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to right
are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10 m/s)
at the 300 hPa level.

238

 

Figure E.10. Northerly, behind low, no fronts (NbLnF). Maps in the top row from left to
right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row

from left to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level.

 

Figure E.11. Southerly, ahead of low, no fronts (SaLnF). Maps in the top row from left
to right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row
from left to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level.

239

 

Figure B.12. Southerly, no low, no fronts (SnLnF). Maps in the top row from lefi to right
are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every 5°C),
and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row from
left to right are height (contoured every 120 m), wind vectors, and isotachs (contoured

every 10 m/s) at the 300 hPa level.

 

Figure B.13. Bermuda High (BH). Maps in the top row from left to right are temperature
(contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and height
(contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to right
are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10 m/s)
at the 300 hPa level.

240

 

Figure B.14. North-south oriented front (nsF). Maps in the top row from left to right are
temperature (contoured every 5°C) wind vectors, dewpoint (contoured every 5°C), and
height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row from left to
right are height (contoured every 120 m), wind vectors, and isotachs (contoured every 10
m/s) at the 300 hPa level.

 

Figure E.15. Northerly, east-west oriented front (N ewF). Maps in the top row from left
to right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row
from lefi to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level.

241

 

Figure B.16. Southerly, east-west oriented front (SewF). Maps in the top row from left
to right are temperature (contoured every 5°C) wind vectors, dewpoint (contoured every
5°C), and height (contoured every 30 m) at the 700 hPa level. Maps in the bottom row
from left to right are height (contoured every 120 m), wind vectors, and isotachs
(contoured every 10 m/s) at the 300 hPa level.

242

LITERATURE CITED

Ahrens CD (2003) Meteorology today, 7th ed. Pacific Grove, California: Brook/Cole —
Thomson Learning.

Allen TR, Kupfer JA (2000) Application of spherical statistics to change vector analysis
of Landsat data: Southern Appalachian spruce-fir forests. Remote Sensing of the
Environment, 74, 482-493.

Allen TR, Kupfer JA (2001) Spectral response and spatial pattern of Fraser fir mortality
and regeneration, Great Smoky Mountains, USA. Plant Ecology, 156, 59—74.

Aneja VP, Businger S, Li Z, Claibom CS, Murthy A (1991) Ozone climatology at high
elevations in the southern Appalachians. Journal of Geophysical Research, 96(D 1),
1007-1021.

Aneja VP, Robarge WP, Claibom CS, Murthy A, Soo-Kim D, Li Z (1992) Chemical
climatology of high elevation spruce-fir forests in the southern Appalachian mountains.
Environmental Pollution, 75, 89-96.

Balling RC, Michaels PJ, Knappenberger PC (1998) Analysis of winter and summer
warming rates in gridded temperature time series. Climate Research, 9, 175-181.

Bell GD, Bosart LF (1988) Appalachian cold-air damming. Monthly Weather Review,
116,137-161.

Bogucki DJ (1972) Intense rainfall in Great Smoky Mountains National Park. Journal of
the Tennessee Academy of Science, 47(3), 93-97.

Bolstad PV, Swifi L, Collins F, Régniére J (1998) Measured and predicted air
temperatures at basin to regional scales in the southern Appalachian mountains.

Agricultural and Forest Meteorology, 91, 161-176.

Bosart LF (1975) New England coastal frontogenesis. Quarterly Journal of the Royal
Meteorological Society, 101, 957-978.

Braun EL (1950) Deciduous forests of eastern North America. New York: Hafner Press.

Bryson RA (1966) Air masses, streamlines, and the boreal forest. Geographical Bulletin,
8, 228-269.

Busing RT, Clebsch EEC, Eagar CC, Pauley EF (1988) Two decades of change in the

Great Smoky Mountains spruce-fir forest. Bulletin of the Torrey Botanical Club, 115(1),
25-31.

243

Carlson TN (1998) Mid-latitude weather systems. Boston, Massachusetts: The American
Meteorological Society.

Davis R, Rogers RF (1992) A synoptic climatology of severe storms in Virginia. The
Professional Geographer, 44(3), 319-332.

Deevey ES (1949) Biogeography of the Pleistocene: Part I: Europe and North America.
Bulletin of the Geological Society of America, 60, 1315-1416.

Delcourt PA, Delcourt HR (2001) Living well in the age of global warming. White River
Junction, Vermont: Chelsea Green Publishing Company.

Delcourt HR (2002) Forests in peril. Blacksburg, Virginia: The McDonald and
Woodward Publishing Company.

Diaz HF, Quayle RG (1978) The 1976-77 winter in the contiguous United States in
comparison with past records. Monthly Weather Review, 106(10), 1393-1421.

Diem JE (2006) Synoptic-scale controls of summer precipitation in the southeastern
United States. Journal of Climate, 19, 613-621.

Easterling DR, Horton B, Jones PD, Peterson TC, Karl TR, Parker DE, Salinger MJ,
Razuvayev V, Plummer N, Jamason P, F olland CK (1997) Maximum and minimum
temperature trends for the globe. Science, 277, 364-367

Fovell RG (1997) Consensus clustering of US. temperature and precipitation data.
Journal of Climate, 10, 1405-1427.

Fovell RG, Fovell M-YC (1993) Climate zones of the contenninous United States
defined using cluster analysis. Journal of Climate, 6(11), 2103-2135.

Gaflin DM, Hotz DG (2000) A precipitation and flood climatology with synoptic features
of heavy rainfall across the southern Appalachian mountains. National Weather Digest,
24(3), 3-15.

Goldman, M (2009) Constructing connectivity: Conservation corridors and conservation
politics in East African rangelands. Annals of the Association of American Geographers,
99(2), 335-359.

Gong X, Richman MB (1995) On the application of cluster analysis to growing season
precipitation data in North America east of the Rockies. Journal of Climate, 8(4), 897-
931.

Great Smoky Mountains Association (2007) A big year for fires. Bearpaw, F all-Winter
2007-2008, 4.

244

Great Smoky Mountains Association (2009) Spring fires scorch hundreds of acres.
Smokies Guide, Summer 2009, 6.

Hack JT, Goodlett JC (1960) Geomorphology and forest ecology of a mountain region in
the central Appalachians. Geological Survey Professional Paper, 347, 1-66.

Harman JR (1991) Synoptic climatology of the westerlies: Processes and patterns.
Washington, DC: Association of American Geographers.

Harvey WS (1978) Trends and variations of mean temperature in the lower troposphere.
Monthly Weather Review, 106, 413-416.

Heald W (1967) Sky islands. Princeton, New Jersey: Van Nostrand.

Hicks DJ (1979) Forest microclimate in the Great Smoky Mountains. Journal of the
Tennessee Academy of Science, 54(3), 84-88.

Houk R (2009) “A” is for azalea. Smokies Life Magazine, 3(1), 10-17.

Intergovernmental Panel on Climate Change (2007a) Climate change 2007: The physical
science basis. Cambridge, New York: Cambridge University Press.

Intergovernmental Panel on Climate Change (2007b) A contribution of Working Group 11
to the fourth assessment report of the Intergovernmental Panel on Climate Change:
Summary for policymakers.
<http://www.ipcc.ch/pdf/assessment-report/ar4/wg2/ar4-wg2-spm.pdf>.

Karl TR, Jones PD, Knight RW, Kukla G, Plummer N, Razuvayev V, Gallo KP, Lindseay
J, Charlson RJ, Peterson TC (1993) A new perspective on recent global warming:

Asymmetric trends of daily maximum and minimum temperature. Bulletin of the
American Meteorological Society, 74(6), 1007-1023.

Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D, Gandin L, Iredell M, Saha S,
White G, Woollen J, Zhu Y, Chelliah M, Ebisuzaki W, Higgins W, Janowiak J, Mo KC,
Ropelewski C, Wang J, Leetmaa A, Reynolds R, Jenne R, Joseph D (1996) The
NCEP/N CAR 40-year reanalysis project, Bulletin of the American Meteorological
Society, 77(3), 437-471.

Karypis G (2003) CLUTO: A clustering toolkit. Technical Report: #02-01 7. University
of Minnesota, Department of Computer Science, Minneapolis, MN 55455.

Klein WH (1957) Principle tracks and mean frequencies of cyclones and anticyclones in
the Northern Hemisphere. Weather Bureau Research Paper No. 40, 60 pp., US.
Department of Commerce, NOAA, Washington, DC.

Konrad CE (1997) Synoptic-scale features associated with warm season heavy rainfall
over the interior southeastern United States. Weather and Forecasting, 12, 557-571.

245

Koppen W (1923) Die klimate der erde; Grundriss der klimakunde. De Gruyter.

Kupfer JA, Cairns DM (1996) The suitability of montane ecotones as indicators of global
climatic change. Progress in Physical Geography, 20(3), 253-272.

Kupfer JA, Balmat J, Smith JL (2005) Shifis in the potential distribution of sky island
plant communities in response to climate change. In Connecting mountain islands and
desert seas: Biodiversity and management of the Madrean Archipelago 11. Edited by GJ
Gottfried, BS Gebow, LG Eskew, and CB Edminster. pp. 485-490. Fort Collins,
Colorado: US. Department of Agriculture, Forest Service, Rocky Mountain Research
Station.

Lewis SL, Lopez-Gonzalez G, Sonke B, Affum-Baffoe K, Baker TR, Ojo LO, Phillips
OL, Reitsma JM, White L, Comiskey JA, Djuikouo K M-N, Ewango CEN, Feldpausch
TR, Hamilton AC, Gloor M, Hart T, Hladik A, Lloyd J, Lovett JC, Makana J-R, Malhi Y,
Mbago FM, Ndangalasi HJ, Peacock J, Feb KS-H, Sheil D, Sunderland T, Swaine MD,
Taplin J, Taylor D, Thomas SC, Votere R, W611 H (2009) Increasing carbon storage in
intact African tropical forests. Nature, 457, 1003-1006.

Luyssaert S, Schulze E-D, Bomer A, Knohl A, Hessenmoller D, Law BE, Ciais P, Grace J
(2008) Old-growth forests as global carbon sinks. Nature, 455, 213-215.

Mark AF (1958) The ecology of the southern Appalachian grass balds. Ecological
Monographs, 28(4), 293-336.

Marris E (2009) Planting the forest of the future. Nature, 459, 906-908.
Mesinger F, DiMego G, Kalnay E, Mitchell K, Shafran PC, Ebisuzaki W, Jovié D,
Woollen J, Rogers E, Berbery EH, Ek MB, Fan Y, Grumbine R, Higgins W, Li H, Lin Y,

Manikin G, Parrish D, Shi W (2006) North American regional reanalysis, Bulletin of the
American Meteorological Society, 87(3), 343-360.

Miller JE (1946) Cyclogenesis in the Atlantic coastal region of the United States.
Journal of Meteorology, 3, 31-44.

Milligan GW, Cooper MC (1985) An examination of procedures for determining the
number of clusters in a data set. Psychometrika, 50, 159-179.

Mimmack GM, Mason SJ, Galpin J S ( 2001) Choice of distance matrices in cluster
analysis: Defining regions. Journal of Climate, 14, 2790-2797.

National Park Service (2009) Nature & Science.
<http://www.nps.gov/grsm/naturescience/index.htm>.

O’Handley C, Bosart LF (1996) The impact of the Appalachian Mountains on cyclonic
weather systems. Part I: A climatology. Monthly Weather Review, 124, 1353-1373.

246

Reitan CH (1974) Frequencies of cyclones and cyclogenesis for North America, 1951-
1970. Monthly Weather Review, 102, 861-868.

Richwein BA (1980) The darnming effect of the southern Appalachians. National
Weather Digest, 5(1), 2-12.

Root TL, Schneider SH (2001) Climate change: Overview and implications for wildlife.
In Wildlife responses to climate change. Ed. S.H. Schneider and TL. Root. Washington:
Island Press. 1-56.

Shadbolt RP, Waller EA, Messina JP, Winkler JA (2006) Source regions of lower-
tropospheric airflow trajectories for the lower peninsula of Michigan: A 40-year air mass
climatology. Journal of Geophysical Research, 11], D21 117, doi:
10.1029/2005JD006657.

Shanks RE (1954) Climates of the Great Smoky Mountains. Ecology, 35(3), 354-361.

Skaggs RH (1969) Analysis and regionalization of the diurnal distribution of tornadoes in
the United States. Monthly Weather Review, 97, 103-115.

Skroch M (2009) The sky islands of North America. International Journal of Wilderness,
15(1), 9-14.

Smallshaw J (1953) Some precipitation-altitude studies of the Tennessee Valley
Authority. American Geophysical Union — Transactions, 34(4), 583-588.

Solow AR (1987) Testing for climate change: An application of the two-phase regression
model. Journal of Climate and Applied Meteorology, 26, 1401-1405.

Stephenson DB (1997) Correlation of spatial climate/weather maps and the advantages of
using the Mahalanobis metric in predictions. Tellus, 49A, 513-527.

Stine AR, Huybers P, Fung IY (2009) Changes in the phase of the annual cycle of surface
temperature. Nature, 457, 435-440.

Stooksbury DE, Michaels PJ (1991) Cluster analysis of southeastern US. climate
stations. Theoretical and Applied Climatology, 44(3-4), 143-150.

Tan P-N, Steinbach M, Kumar V (2006) Introduction to data mining. Boston,
Massachusetts: Pearson Education, Inc.

Thomthwaite CW (1931) The climates Of North America, according to a new
classification. Geographical Review, 21, 633-655.

Trenberth KE (1990) Recent observed interdecadal climate changes in the northern
hemisphere. Bulletin of the American Meteorological Society, 71(7), 988-993.

247

Tyron RC (1939) Cluster analysis. Edwards Brothers.

United States Global Change Research Program (2009) Global climate change impacts
in the United States. Cambridge, New York: Cambridge University Press.

van Loon H, Williams J (1976a) The connection between trends of mean temperature and
circulation at the surface: Part 1. Winter. Monthly Weather Review, 104, 365-380.

van Loon H, Williams J (1976b) The connection between trends of mean temperature and
circulation at the surface: Part 11. Summer. Monthly Weather Review, 104, 1003-1011.

Walters CK, Winkler JA (2001) Airflow configurations of warm season southerly low-
level wind maxima in the Great Plains. Part 1: spatial and temporal characteristics and
relationship to convection. Weather and Forecasting, 16(5), 513-530.

Walters CK, Winkler JA, Shadbolt RP, van Ravensway J, Bierly GD (2008) A long-term
climatology of southerly and northerly jets for the central United States. Annals of the
Association of American Geographers, 98(3), 521-552.

Walters CK, Winkler JA, van Ravensway J, Shadbolt RP (in preparation) Airflow patterns
of northerly low-level jets in central North America. Part I: Climatological analysis.

Whittaker RH (1952) A study of summer foliage insect communities in the Great Smoky
Mountains. Ecological Monographs, 22(1), 1-44.

Whittaker RH (1956) Vegetation of the Great Smoky Mountains. Ecological
Monographs, 26(1), 1-80.

Wilks DS (2006) Statistical methods in the atmospheric sciences. 2nd ed. Burlington,
Massachusetts: Academic Press.

“Winkler JA (1992) Regional patterns of the diurnal properties of heavy hourly
precipitation. Professional Geographer, 44, 127-146.

Winkler JA (2004) The impact of technology upon in situ atmospheric observations and
climate science. In Geography and technology, ed. S.D. Brunn, S.L. Cutter, and J .W.
Harrington, Jr. Norwell, Massachusetts: Kluwer Academic Publishers.

Zhao Y, Karypis G (2003) Criterion functions for document clustering: Experiments and
analysis. Technical Report #01 -40. University of Minnesota, Department of Computer
Science /Army HPC Research Center, Minneapolis, MN 55455.

Zishka ZM, Smith PJ (1980) The climatology of cyclones and anticyclones over North

America and surrounding ocean environs for January and July, 1950-77. Monthly
Weather Review, 108(4), 387-401.

248