LOSS OF URBAN FOREST CANOPY AND THE RELATED EFFECTS ON SOUNDSCAPE AND HUMAN DIRECTED ATTENTION
ROBERT JAMES PAUL LAVERNE
Bachelor of Science in Biological Science University of Michigan – Dearborn April 1978
Bachelor of Science in Forestry Michigan Technological University February 1980
Master of Science in Natural Resources - Remote Sensing University of Michigan May 1985
submitted in partial fulfillment of requirements for the degree DOCTOR OF PHILOSOPHY IN URBAN STUDIES AND PUBLIC AFFAIRS at the CLEVELAND STATE UNIVERSITY NOVEMBER 2016
©COPYRIGHT BY ROBERT JAMES PAUL LAVERNE 2016
We hereby approve this dissertation for
ROBERT JAMES PAUL LAVERNE
Candidate for the Doctor of Philosophy in Urban Studies and Public Affairs degree
for the Maxine Goodman Levin College of Urban Affairs
and the CLEVELAND STATE UNIVERSITY
College of Graduate Studies
______Dissertation Chairperson, Wendy A. Kellogg, Ph.D.
Department of Urban Studies Date______
______Dissertation Committee Member, Sanda Kaufman, Ph.D.
Department of Urban Studies Date______
______Dissertation Committee Member, Helen Liggett, Ph.D.
Department of Urban Studies Date______
______Dissertation Committee Member, Nicholas Zingale, Ph.D.
Department of Urban Studies Date______
______Dissertation Committee Member, William C. Sullivan, Ph.D.
Department of Landscape Architecture, University of Illinois Date______
Date of Defense: November 21, 2016
DEDICATION
Ad majorem Dei gloriam
To the greater glory of God
ACKNOWLEDGEMENTS
My sincere thanks to my employer, the Davey Tree Expert Company for supporting my education and encouraging me to pursue this research.
I am grateful for the assistance of Mr. Dru Sabatello, Village Forester and Ms. Ashley
Karr, Arborist from the Village of Arlington Heights, Illinois. Their maps and information regarding the Emerald Ash Borer management program was essential.
Dr. Dean Hawthorne of the Cornell Lab of Ornithology provided knowledge and guidance on the use of the Raven sound analysis program, for which I am thankful.
Dr. Sara Incera Burkert of the Cleveland State University Psychology Department generously provided her time, knowledge and guidance in the use of the MouseTracker program. I am thankful and most fortunate to have encountered Sara at precisely the time
I needed help.
Since enrolling at Cleveland State University I have been extremely fortunate to have experienced the knowledge, insight, encouragement and support of the faculty, staff and fellow students. I am most grateful to my dissertation committee including:
Sanda Kaufman – Cleveland State University
Wendy Kellogg (Chair) - Cleveland State University
Helen Liggett - Cleveland State University
William Sullivan – University of Illinois
Nicholas Zingale - Cleveland State University
I am especially thankful for the guidance and friendship of Dr. Wendy Kellogg. Life has purpose. We come to life and cross paths with other people, seemingly by chance but likely by good fortune. These companions in life’s journey help us to form our experiences, insights, understanding, and beliefs. With these essential elements comes wisdom, and from wisdom comes a pathway to faith. I have been blessed with travelling convergent paths with others willing to help me find my way and show me how to assist others. These are a few of the many that have helped build the foundation for my life and eventually for this dissertation:
Ruth Ann and Albert James Laverne, for sharing with me their wonder of nature.
The teachers throughout my education, who encouraged and inspired me to learn,
especially those who shared their passion for discovery and enlightenment.
My friends in school – those who suffered with me through years of
underachieving and those who celebrated with me as learning grew to be splendid
and essential.
My family, who continually define for me the meaning of pride.
My wife, Donna Jean, who makes life delightful and rich.
And also the many people who have enhanced my faith in God, the Creator of this miraculous world. To all, my sincere and deep thanks. LOSS OF URBAN FOREST CANOPY AND THE RELATED EFFECTS ON
SOUNDSCAPE AND HUMAN DIRECTED ATTENTION
ROBERT JAMES PAUL LAVERNE
ABSTRACT
The specific questions addressed in this research are: Will the loss of trees in residential neighborhoods result in a change to the local soundscape? The investigation of this question leads to a related inquiry: Do the sounds of the environment in which a person is present affect their directed attention?
An invasive insect pest, the Emerald Ash Borer (Agrilus planipennis), is killing millions of ash trees (genus Fraxinus) throughout North America. As the loss of tree canopy occurs, urban ecosystems change (including higher summer temperatures, more stormwater runoff, and poorer air quality) causing associated changes to human physical and mental health. Previous studies suggest that conditions in urban environments can result in chronic stress in humans and fatigue to directed attention, which is the ability to focus on tasks and to pay attention. Access to nature in cities can help refresh directed attention. The sights and sounds associated with parks, open spaces, and trees can serve as beneficial counterbalances to the irritating conditions associated with cities. This research examines changes to the quantity and quality of sounds in Arlington Heights,
Illinois. A series of before-and-after sound recordings were gathered as trees died and were removed between 2013 and 2015. Comparison of recordings using the Raven sound analysis program revealed significant differences in some, but not all measures of sound
vii attributes as tree canopy decreased. In general, more human-produced mechanical sounds (anthrophony) and fewer sounds associated with weather (geophony) were detected. Changes in sounds associated with animals (biophony) varied seasonally.
Monitoring changes in the proportions of anthrophony, biophony and geophony can provide insight into changes in biodiversity, environmental health, and quality of life for humans. Before-tree-removal and after-tree-removal sound recordings served as the independent variable for randomly-assigned human volunteers as they performed the
Stroop Test and the Necker Cube Pattern Control test to measure directed attention. The sound treatments were not found to have significant effects on the directed attention test scores. Future research is needed to investigate the characteristics of urban soundscapes that are detrimental or potentially conducive to human cognitive functioning.
viii
TABLE OF CONTENTS
Page
ABSTRACT ...... vii
LIST OF TABLES ...... xix
LIST OF FIGURES ...... xxiv
CHAPTER
I. INTRODUCTION
Title ...... 1
Threats to Urban Ecosystems and the Associated Effect on Human Well-being
...... 1
Urban Forests, Emerald Ash Borer, and Changing Communities ...... 2
Soundscapes ...... 6
Noise ...... 7
Directed Attention ...... 10
Trees, Sounds, and People ...... 11
Research Questions ...... 12
Components of this Dissertation Document ...... 13
II. LITERATURE REVIEW ...... 15
ix
Directed Attention ...... 16
The Benefits of Urban Forests and Access to Nature to Human Health ...... 20
Introduction...... 20
Urban forests and human physical health...... 22
Exercise and muscle-powered transportation ...... 23
Human physical health recovery through mental well-being...... 24
The view from home and work ...... 27
Well-being and noise...... 29
Physical & mental response following exposure to nature and urban settings. 29
Theories regarding the human response to nature ...... 30
Spiritual / emotional enrichment...... 35
Soundscape Ecology ...... 39
Noise attenuation by vegetation...... 45
Ecological consequences of anthropogenic noise...... 47
The effect of anthrophony, biophony and geophony on humans...... 50
Perception of visual and auditory stimuli...... 58
Summary ...... 59
III. COMMENTARY ON THE LITERATURE: HOW DID I GET HERE?...... 61
IV. RESEARCH DESIGN ...... 76
Research Questions ...... 76
x
Evaluation of soundscape attributes and their importance to humans ...... 78
Research Process and Conceptual Framework...... 81
Selection of a Subject Community...... 83
Selection of Sound Recording Sites ...... 85
Recording of soundscapes ...... 86
Selection of Sound Analysis Software ...... 87
Selection of Sound Analysis Methods ...... 87
Selection of Sound Recording Treatments ...... 88
Selection of Tests for Directed Attention ...... 89
Selection of Demographic Information ...... 90
Selection of Stroop Test Platform ...... 91
Selection of Participants for the Tests for Directed Attention ...... 92
Analysis of the Tests for Directed Attention ...... 93
Necker Cube Pattern Control ...... 94
Stroop Test ...... 95
Summary ...... 96
V. METHODS PART 1A – SOUNDSCAPE DATA COLLECTION ...... 100
Locating a Host Community ...... 100
Meeting with local officials in the Chicago metropolitan area ...... 104
Sound Recording Equipment ...... 106
xi
Sound recorder options and settings ...... 107
Calibration of sound recording equipment ...... 108
Selection of Sound Recording Sites ...... 109
Measurement of Tree Canopy Cover ...... 125
Sound Recording Data Collection...... 130
Field notes ...... 132
Spatially matched sound recordings ...... 135
Temporally matched sound recordings ...... 136
Change in Tree Canopy Not Attributable to Emerald Ash Borer ...... 138
Soundscape Data Collection Limitations ...... 139
VI. METHODS PART 1B – SOUNDSCAPE ANALYSIS ...... 143
Sound Recording Interpretation and Spectrogram Analysis ...... 143
Sound analysis using the Raven program...... 143
Choosing sound presentation options within Raven...... 144
Choosing sound measurement options within Raven...... 147
Testing differences between calibration recordings...... 152
Testing differences between simultaneous field recordings...... 158
Slice-and-Dice analysis of initial sound recording pair:
2,000 Hz slices of Average Power (dB)...... 158
Conversion of measures of sound intensity from decibels (dB) to
millivolts (mV) ...... 161
xii
Slice-and-Dice analysis of initial sound recording pair:
2,000 Hz slices of Average Power (mV)...... 162
Sound signature analysis of initial sound recording pair...... 165
Measuring silence...... 180
Selection of sound recording pairs to be analyzed...... 183
Slice-and-Dice analysis of sound recording pairs...... 186
Statistical Analysis ...... 188
One-third octave slice-and-dice t-Tests...... 188
Sound signature t-Tests...... 190
Proportional statistics and graphs...... 193
What was Learned During the Sound Analysis Process? ...... 194
VII. RESULTS PART 1 – ANALYSIS OF SOUND RECORDNGS ...... 197
Introduction ...... 197
Comparison of Sound Recording Pairs ...... 199
Third octave slice-and-dice analysis...... 200
Evaluating the method of analysis: The process of adding and subtracting
decibels...... 206
Manual identification of individual sounds and statistical analysis...... 209
Plotting the data...... 212
Proportional analysis...... 216
Proportional changes to soundscape types relative to time...... 224
xiii
Proportional changes to soundscape types relative to frequency...... 225
Proportional changes to soundscape types relative to energy...... 227
Proportional changes to soundscape types relative to entropy...... 235
Proportional changes to soundscape types summary...... 240
Analysis of sound attribute mean values...... 244
t-Test for Delta Time by sound type...... 245
t-Test for Delta Frequency by sound type...... 248
t-Test for Delta Time X Delta Frequency by sound type...... 251
t-Test for Center Frequency by sound type...... 253
t-Test for Peak Frequency by sound type...... 256
t-Test for Average Power (mV) by sound type ...... 258
t-Test for Peak Power (mV) by sound type ...... 260
t-Test for Energy (mV) by sound type ...... 263
t-Test for Average Entropy by sound type ...... 265
t-Test for Aggregate Entropy by sound type ...... 268
t-Test for Maximum Entropy by sound type ...... 270
t-Test for Delta Entropy by sound type ...... 272
Summary of t-Test analysis for mean values of sound attributes...... 274
Soundscape Analysis Summary ...... 281
Summary of one-third octave slice-and-dice analysis ...... 282
xiv
Summary of total values for four sound attributes ...... 284
Summary of proportional values for four sound attributes ...... 287
Summary of mean values for four sound attributes ...... 291
Summary of mean values for eight sound attributes in silence samples ...... 294
Common patterns revealed by sound recording analysis ...... 296
Areas for Improvement ...... 298
What Matters? ...... 300
VIII. METHODS PART 2 – TESTS FOR DIRECTED ATTENTION ...... 303
The Stroop Effect ...... 303
Administering the Stroop Test ...... 304
Necker Cube Pattern Control ...... 308
Enlisting Test Participants ...... 309
Selecting the Sound Recordings for Play During Tests for Directed Attention ...... 310
Administering the Tests for Directed Attention ...... 313
Setting and introduction ...... 313
Purpose of the testing design...... 314
Arithmetic mental fatigue task ...... 315
Spell check mental fatigue task ...... 317
10-minute rest period ...... 317
Stroop Test ...... 318
xv
Necker Cube Pattern Control test ...... 320
Limitations to the Experimental Design ...... 321
Methods for Analysis of Data ...... 322
Stroop Test analysis ...... 323
Necker Cube Pattern Control analysis ...... 326
IX. RESULTS PART 2 – TESTS FOR DIRECTED ATTENTION ...... 327
Participant Demographics ...... 327
Results of Necker Cube Pattern Control Test ...... 329
Stroop Test Results ...... 331
Stroop Test response time results...... 334
Stroop Test maximum deviation time results...... 339
Stroop Test maximum deviation distance results...... 341
Stroop Test area under the curve results...... 344
Stroop Test percent incorrect answers results...... 346
Tests for Directed Attention Results by Demographics ...... 349
Directed Attention test scores by gender and omitting sound treatment...... 350
Directed Attention test scores by age and omitting sound treatment...... 351
Directed Attention test scores by community type and omitting sound
treatment...... 352
Summary ...... 354
xvi
X. DISCUSSION AND CONCLUSIONS ...... 356
Introduction ...... 356
Summary of Important Observations and Findings ...... 358
Urban forestry ...... 358
Soundscape ecology ...... 359
Soundscape data collection ...... 359
Soundscape data analysis ...... 360
Tests for directed attention ...... 365
Interpretation of Findings ...... 367
Soundscape characteristics relative to tree canopy ...... 367
Tests for directed attention ...... 369
Contributions to the Literature ...... 373
Contributions to Practice ...... 375
Sound data analysis methods ...... 376
Soundscapes as a monitoring method ...... 378
Design of tests for directed attention ...... 379
Limitations ...... 381
Analysis of sound recordings ...... 381
Enlisting volunteers for tests for directed attention ...... 382
Design of test for directed attention sessions ...... 383
xvii
Demographic and geographic limitations ...... 386
Potential improvements to methods ...... 388
Recommendations for Future Research ...... 389
Urban forestry ...... 389
Emerald Ash Borer ...... 389
The value of landscape trees ...... 390
Soundscape ecology ...... 391
Human health and social concerns ...... 394
Conclusions ...... 397
Change in the tree canopy cover ...... 398
Change in the soundscapes...... 398
Change in measures of directed attention ...... 400
Closing Thoughts ...... 402
REFERENCES ...... 407
APPENDICES ...... 424
A. Sound recording locations, dates and recording length ...... 425
B. Example of a sound recording field data sheet...... 428
xviii
LIST OF TABLES
Table Page
I. Summary of sound recording dates and times ...... 131
II. a) Examples of positioning variables within Excel according to Variance
value ...... 156
b) Examples of positioning variables within Excel according to Variance
value ...... 156
III. a) Examples of t-Test for 2,000 Hz slices of side-by-side sound recordings ..158
b) Examples of t-Test for 2,000 Hz slices of side-by-side sound recordings ..158
IV. F-Test and t-Test results for two 2,000 Hz “slices” that test for significant
differences in variances (F-Test) and means (t-Test) of average power (dB) of
simultaneous sound recordings at Prindle Ave. and Stratford Rd. on September 5,
2014...... 161
V. F-Test and t-Test results for two 2,000 Hz “slices” that test for significant
differences in variances (F-Test) and means (t-Test) of average power (mV) of
simultaneous sound recordings at Prindle Ave. and Stratford Rd. on September 5,
2014...... 163
VI. a) Spatially matched sound recording pairs (same location; before-and-after
dates)...... 185
b) Temporally matched sound recording pairs (Same time; different locations
for many trees and few trees) ...... 185
xix
VII. Frequency ranges (Hz) of one third octaves used in slice-and-dice analysis...... 186
VIII. t-Test matrix for combinations of sound types and sound attributes...... 192
IX. Mean Energy values for each of the 32 third octave frequency bands calculated
from logarithmic decibel (dB) values and from logarithmic values first converted
to non-logarithmic millivolt (mV) values before conversion back to decibels ...208
X. Count of identified sounds and sound choruses for the May 19, 2015 sound
recordings at Stratford Rd. and Prindle Ave...... 245
XI. Delta Time sound attribute t-Test results for anthrophony, biophony, and
geophony sound types for the temporally matched sound recordings at Stratford
Rd. and Prindle Ave. on May 19, 2015...... 247
XII. Delta Frequency sound attribute t-Test results for anthrophony, biophony, and
geophony sound types for the temporally matched sound recordings at Stratford
Rd. and Prindle Ave. on May 19, 2015...... 250
XIII. Delta Time X Delta Frequency sound attribute t-Test results for anthrophony,
biophony, and geophony sound types for the temporally matched sound
recordings at Stratford Rd. and Prindle Ave. on May 19, 2015 ...... 252
XIV. Center Frequency sound attribute t-Test results for anthrophony, biophony, and
geophony sound types for the temporally matched sound recordings at Stratford
Rd. and Prindle Ave. on May 19, 2015...... 254
XV. Peak Frequency sound attribute t-Test results for anthrophony, biophony, and
geophony sound types for the temporally matched sound recordings at Stratford
Rd. and Prindle Ave. on May 19, 2015...... 257
xx
XVI. Average Power (mV) sound attribute t-Test results for anthrophony, biophony,
and geophony sound types for the temporally matched sound recordings at
Stratford Rd. and Prindle Ave. on May 19, 2015 ...... 259
XVII. Peak Power (mV) sound attribute t-Test results for anthrophony, biophony, and
geophony sound types for the temporally matched sound recordings at Stratford
Rd. and Prindle Ave. on May 19, 2015...... 262
XVIII. Energy (mV) sound attribute t-Test results for anthrophony, biophony, and
geophony sound types for the temporally matched sound recordings at Stratford
Rd. and Prindle Ave. on May 19, 2015...... 264
XIX. Average Entropy sound attribute t-Test results for anthrophony, biophony, and
geophony sound types for the temporally matched sound recordings at Stratford
Rd. and Prindle Ave. on May 19, 2015...... 266
XX. Aggregate Entropy sound attribute t-Test results for anthrophony, biophony, and
geophony sound types for the temporally matched sound recordings at Stratford
Rd. and Prindle Ave. on May 19, 2015...... 269
XXI. Maximum Entropy sound attribute t-Test results for anthrophony, biophony, and
geophony sound types for the temporally matched sound recordings at Stratford
Rd. and Prindle Ave. on May 19, 2015...... 271
XXII. Delta Entropy sound attribute t-Test results for anthrophony, biophony, and
geophony sound types for the temporally matched sound recordings at Stratford
Rd. and Prindle Ave. on May 19, 2015...... 273
XXIII. Bird, amphibian, insect, and mammal species identified by sound in the pairs of
sounds recordings...... 298
xxi
XXIV. Example of a problem set used for the arithmetic task for inducing mental fatigue
in participants before the tests for directed attention ...... 316
XXV. Matrix for ANOVA analysis comparing MouseTracker scores for groups and sub-
groups of participants by sound treatment ...... 325
XXVI. Matrix for ANOVA analysis comparing Necker Cube scores for groups and sub-
groups of participants by sound treatment ...... 326
XXVII. Necker Cube Pattern Control Test - Mean values of perceived change in cube
perspective during 3-minute timed trial ...... 330
XXVIII. Analysis of Variance (ANOVA) results for MouseTracker response time for
congruent color matches for all participants ...... 336
XXIX. Analysis of Variance (ANOVA) results for MouseTracker response time for
incongruent color matches for all participants ...... 337
XXX. Stroop Test - Mean values of response time for congruent color matches and
incongruent color matches and ANOVA P-value results ...... 338
XXXI. Stroop Test - Mean values of maximum deviation from straight line time
(milliseconds) for congruent color matches and incongruent color matches and
ANOVA P-value results ...... 340
XXXII. Stroop Test - Mean values of maximum deviation from straight line distance
(bins) for congruent color matches and incongruent color matches and ANOVA
P-value results ...... 342
XXXIII. t-Test P-values for mean values of maximum deviation distance of the suburban
community type subgroup for the six sound treatment pairings ...... 343
xxii
XXXIV. Stroop Test - Mean values of area under curve (AUC) for congruent color
matches and incongruent color matches and ANOVA P-value results ...... 345
XXXV. Stroop Test - Mean values of number of incorrect answers for congruent color
matches and incongruent color matches and ANOVA P-value results ...... 348
XXXVI. t-Test results for mean scores of Necker Cube Pattern Control and Stroop Test
variables for participant population sorted by gender...... 350
XXXVII. t-Test results for mean scores of Necker Cube Pattern Control and Stroop Test
variables for participant population sorted by age ...... 351
XXXVIII. ANOVA results for mean scores of Necker Cube Pattern Control and Stroop Test
variables for participant population sorted by community type ...... 352
XXXIX. Post hoc t-Test results for mean scores of Stroop Test variables for participant
population sorted by community type...... 353
xxiii
LIST OF FIGURES
Figure Page
1. Emerald Ash Borer adult (Agrilus planipennis) ...... 4
2. Ash tree infested with Emerald Ash Borer ...... 4
3. Research task flow chart for each of the three areas of study and resultant
professional and scholarly audiences and applications...... 83
4. The City of Arlington Heights, Illinois uses signs tied to ash street trees to
provide residents with information on Emerald Ash Borer and the city’s plan for
dealing with the insect pest ...... 102
5. Portion of Village of Arlington Heights ash trees map (2011) ...... 103
6. Aerial photo of portion of Arlington Heights ...... 103
7. Canterbury Street in Arlington Heights, Illinois ...... 104
8. Locations of three neighborhoods in Arlington Heights where ash trees were
scheduled to be removed during the winter of 2013-14 ...... 106
9. a) A Zoom H4n digital sound recorder; and…
b) deployed at the intersection of Shirra Ct. and Raleigh St. in Arlington Heights,
Illinois ...... 107
10. Location of Greenbrier neighborhood in Arlington Heights, Illinois ...... 111
11. Street trees in the Greenbrier neighborhood of Arlington Heights, Illinois ...... 112
xxiv
12. a) Shirra Court as photographed on September 4, 2013 before ash trees were
removed, and …
b) as photographed on May 17, 2014 after ash trees were removed ...... 113
13. a) Huron Street as photographed on September 4, 2013 before ash trees were
removed, and …
b) as photographed on July 16, 2014 after some ash trees were removed ...... 114
c) Huron Street as photographed on April 9, 2016 after additional ash trees were
removed...... 115
14. a) The traffic island at the Huron St. location includes a group of juniper shrubs
in which the sound recorder was concealed to prevent tampering or theft on July
15, 2014 and…
b) on April 9, 2016 ...... 116
15. A whole-tree harvester was used in the Greenbrier neighborhood to remove ash
trees ...... 117
16. The locations within the Greenbrier and Carousel Park neighborhoods where
sound recordings were gathered ...... 118
17. a) Prindle Avenue as photographed on July 17, 2013, and…
b) as photographed on July 16, 2014 ...... 119
c) Prindle Avenue as photographed on August 13, 2015...... 120
18. a) East Fleming Drive as photographed on October 8, 2013, and…
b) as photographed on July 15, 2014 ...... 121
19. a) Stratford Road as photographed on October 19, 2014, and…
b) as photographed on May 18, 2015 ...... 122
xxv
20. a) Suffield Dr. as photographed on May 9, 2014, and…
b) as photographed on September 3, 2014 ...... 123
21. Publicly-owned trees that were treated to protect from Emerald Ash Borer attack
are marked with tags ...... 124
22. a) May 18, 2015 180-degree panoramic photo of the recording site at Stratford
Rd. and…
b) May 18, 2015 180-degree panoramic photo of the recording site at Prindle
Ave...... 125
23. a) Pre-tree removal aerial image from Mapquest of the Shirra Ct. area of the
Greenbrier neighborhood and…
b) Post-tree removal aerial image from Google Maps of the Shirra Ct. area of the
Greenbrier neighborhood ...... 127
24. a) Pre-tree removal aerial image from Mapquest of the Huron St. area of the
Greenbrier neighborhood and…
b) Post-tree removal aerial image from Google Maps of the Huron St. area of the
Greenbrier neighborhood ...... 128
25. a) Pre-tree removal aerial image from Mapquest of the Prindle Ave. – Stratford
Rd. – Suffield Ave. area of the Carousel Park neighborhood and…
b) Post-tree removal aerial image from Google Maps of the Prindle Ave. –
Stratford Rd. – Suffield Ave. area of the Carousel Park neighborhood ...... 129
26. Unattended sound recording equipment was placed on private property with the
homeowner’s permission ...... 132
xxvi
27. A one-hour recording of the soundscape adjacent to the Illinois Route 53 freeway
was collected on May 6, 2015...... 137
28. Hundreds of trees in Arlington Heights were damaged or destroyed during a wind
storm on September 5, 2014, further contributing to the loss of tree canopy in the
community ...... 138
29. Raven Pro Sound Analysis Software spectrogram of stereo channels of a 10-
second sound sample...... 146
30. Sixty-second spectrogram showing 2,000 Hz slices and 30 second dices used in
the sound recorder calibration analysis ...... 154
31. Aerial photo of the Carousel Park neighborhood in Arlington Heights, IL...... 159
32. Comparison of slice-and-dice t-Test results for Average Power reported in
decibels and milliwatts for the sound recordings at Stratford Rd. and Prindle Ave.
on September 5, 2014 ...... 164
33. a) Spectrogram of tree leaves rustling in the wind – an example of geophony
(Gw) ...... 169
b) Spectrogram of a cicada buzzing and a house wren calling, both examples of
biophony (Bi & Bb) ...... 169
c) Spectrogram of a cedar waxwing, a northern cardinal and a mourning dove,
examples of biophony (Bb), and background traffic noise, an example of
anthrophony (Ax) ...... 170
d) Spectrogram of human voices, an example of anthrophony (Av) ...... 170
e) Spectrogram of overlapping sound signatures ...... 170
34. a and b) Spectrograms of the sound of a passing car ...... 175
xxvii
35. a and b) Spectrogram of several bird songs including a house wren, cardinal and
robin ...... 177
36. a) A spectrogram of stereo channels with associated data table for a ten-second
sound segment simultaneously recorded on September 5, 2014 on Stratford
Road ...... 179
b) A spectrogram of stereo channels with associated data table for a ten-second
sound segment simultaneously recorded on September 5, 2014 on Prindle
Ave...... 179
37. a) A one-second sample of relative silence at the Shirra Ct. recording site in
Arlington Heights...... 181
b) Spectrogram of a one-second sample of relative silence adjacent to the Route
59 freeway in Arlington Heights...... 182
c) Relative silence as measured at the Seney National Wildlife Refuge ...... 182
38. Spectrogram of stereo channels including the one third octave frequency
slices...... 187
39. Summary graphic of the third octave slice-and-dice analysis for the sound
recording pair of Prindle Ave. and Stratford Rd. recorded on September 5,
2014...... 189
40. t-Test summary for sound recording pair of Stratford Rd. and Prindle Ave. on
September 5, 2014 ...... 192
41. Sum of the Delta Time sound attribute values for Stratford Rd. (green - many
trees) and Prindle Ave (blue - few trees) recorded on September 5, 2014 ...... 194
xxviii
42. a) One third octave slice-and-dice analysis summary for spatially matched
recording pairs and…
b) One third octave slice-and-dice analysis summary for temporally matched
recording pairs ...... 202
43. Predicted (not actual) results for slice-and-dice analysis for Energy (dB) in areas
with many trees and few trees ...... 203
44. a1 through e2) Sum the sound attribute data for Delta Time, Delta Frequency,
Energy reported in decibels, Energy converted to millivolts, and Aggregate
Entropy by sound type and sub-type for spatially matched and temporally
matched sound recordings...... 212
45. a through e) Proportional summaries for the sound attributes Delta Time, Delta
Frequency, Energy (dB), Energy (mV), and Aggregate Entropy for the temporally
matched sound recording pair gathered on May 19, 2015 at Stratford Rd. and
Prindle Ave ...... 217
46. a through f) Proportional summaries for the sound attributes Delta Time, Delta
Frequency, Energy (dB), Energy (mV), and Aggregate Entropy for the six
spatially matched sound recording pairs ...... 222
47. a through h) Proportional summaries for the sound attributes Delta Time, Delta
Frequency, Energy (dB), Energy (mV), and Aggregate Entropy for the eight
temporally matched sound recording pairs ...... 223
48. a through f) Total sound Energy (mV) by sound type (anthrophony, biophony,
and geophony) and sub-type for the six spatially matched sound recording
pairs ...... 229
xxix
49. a through h) Total sound Energy (mV) by sound type (anthrophony, biophony,
and geophony) and sub-type for the eight temporally matched sound recording
pairs ...... 231
50. a, b & c) Cooper’s hawks and occupied nests were frequently observed in large
ash trees ...... 234
51. a through f) Spectrogram examples of aggregate entropy values of various sound
signatures ...... 238
52. Summary of the sound attribute t-Test results for anthrophony, biophony, and
geophony sound types for the temporally matched sound recordings at Stratford
Rd. and Prindle Ave. on May 19, 2015...... 274
53. a through f) t-Test comparison of mean sound attribute values by sound type
(anthrophony, biophony, and geophony) and sub-type for the six spatially
matched sound recording pairs ...... 277
54. a through h) t-Test comparison of mean sound attribute values by sound type
(anthrophony, biophony, and geophony) and sub-type for the eight temporally
matched sound recording pairs ...... 279
55. Summary of total values for the sound attributes Delta Time, Delta Frequency,
Energy (mV), and Aggregate Entropy for all individual sound selections in the
spatially matched recording pairs and the temporally matched recording pairs ..284
56. Summary of sound attribute proportions for Delta Time, Delta Frequency, Energy
(mV), and Aggregate Entropy for the spatially matched recording pairs and the
temporally matched recording pairs...... 288
xxx
57. Summary of sound attribute mean values for Delta Time, Delta Frequency,
Energy (mV), and Aggregate Entropy for the spatially matched recording pairs
and the temporally matched recording pairs ...... 292
58. Summary of eight sound attribute mean values for silence samples for the
spatially matched recording pairs and the temporally matched recording pairs ..294
59. Example of an instruction screen from the Stroop Test ...... 306
60. a) Example of a congruent word / color match in which the color of the print
matches the meaning of the word and…
b) Example of an incongruent word / color match in which the color of the print
does not match the meaning of the word ...... 306
61. Example of computer mouse trajectory for an incongruent word / color
combination...... 307
62. The optical illusion cube graphic used in the Necker Cube Pattern Control test 308
63. Practice screen in the Stroop Test ...... 318
64. Example of a Stroop Test practice slide in which the color appears as a pattern,
not as letters ...... 319
65. Residential community type by gender for participants of tests for directed
attention...... 328
66. Birth year by gender for participants of tests for directed attention ...... 328
67. Participant count by gender and sound treatment experienced while taking tests
for directed attention ...... 329
68. Sample of one MouseTracker analysis screen ...... 332
xxxi
69. Maximum deviation (MD) and area under the curve (AUC) formed by an
idealized response trajectory and an actual computer mouse trajectory within the
MouseTracker program ...... 333
xxxii
CHAPTER I
INTRODUCTION
“We are, in actuality, students of that greater order known as nature. It is
into nature that man vanishes”
- Loren Eiseley
Title: Loss of urban forest canopy and the related effects on soundscape and human directed attention.
Threats to Urban Ecosystems and the Associated Effect on Human Well-being
Sensory stimuli within dense urban areas can wear a person down. The high energy sights, loud sounds, unpleasant odors and frequent chaos can be stressful, and chronic exposure to conditions that may approach sensory overload may have negative effects on human health and cognitive ability. Many studies have considered the detrimental effects of urban conditions, as will be detailed in the literature review section of this dissertation.
At the same time, access to the relatively calm conditions of outdoor nature areas has been shown to relieve some of the negative effects associated with urban settings. A
1 basic question that emerges is “How does access to nature affect human health?” The research undertaken in this dissertation explores a specific aspect of access to nature as conditions in a community change over time.
To begin, it is helpful to frame the project in common knowledge and then review the literature specifically relevant to this research.
Urban Forests, Emerald Ash Borer, and Changing Communities
Natural resources, including trees within cities serve a wide range of environmental, economic, and social benefits that affect the quality of life for humans. Improvement to air quality, temperature moderation, and sound attenuation are three of many well- documented functions of urban forests. Natural resource managers, urban foresters, urban planners, and even sociologists, psychologists, and medical professionals are beginning to understand how to integrate natural resources into cities and effectively improve living conditions. Unfortunately, many urban forests in the United States are under attack. It seems that once every generation an insect pest or pathogen arrives that decimates a portion of the North American forest. In the early 20th century a fungus known as chestnut blight (Cryphonectria parasitica) destroyed millions of American chestnut (Castanea dentata) trees. In the mid 20th century another introduced fungus,
Ophiostoma ulmi caused the destruction of most American elm (Ulmus americana) trees.
Today, the Emerald Ash Borer (Agrilus planipennis), an invasive insect from Asia is killing millions of trees in the genus Fraxinus.
2
Emerald Ash Borer is believed to have arrived in the United States sometime in the late 1990’s, probably embedded in the wood of packing crates or pallets shipped from somewhere in its native distribution in Asia. Emerald Ash Borer attacks ash trees in
Asia, but the trees have co-evolved with the insect and have built varying degrees of resistance. In 2002 the beetle was discovered in southeast Michigan west of Detroit and east of Ann Arbor. Around this same time the beetle was also found in Windsor, Ontario.
Initial attempts at eradicating the local populations failed as the beetle quickly spread throughout the region, largely through the assistance of unknowing humans transporting ash logs and firewood.
The adult beetles are metallic green and are approximately ½ inch (1 cm) in length
(Figure 1). The Emerald Ash Borer flies to a healthy ash tree in the adult stage and feeds on the foliage, but does little damage. The adult beetles mate and the female lays eggs on the bark of the trees’ branches and trunk. The eggs hatch and the tiny worm-like larvae chew their way into the bark and establish themselves in the cambium tissue of the tree just under the bark. The larvae feed on the cambium tissue and disrupt the flow of photosynthates through the phloem and the flow of water through the outer xylem. The larvae grow and pupate, while they overwinter in the tree. The following spring the adult beetles emerge from the host tree by chewing small “D”-shaped exit holes that are visible on the bark. As additional generations of larvae feed the damage to the cambium worsens and the tree begins to decline. Symptoms include progressive dieback in the tree
3 crown, watersprout development from the base of the tree, splitting of the bark, and eventual death of the tree (Figure 2).
Figure 1: Emerald Ash Borer adult
(Agrilus planipennis)
Figure 2: Ash tree infested with
Emerald Ash Borer. Symptoms include
progressive crown dieback, watersprout development on trunk and lower branches, splitting of bark, and eventual death of the tree.
Many communities throughout North America have plentiful populations of ash trees.
Ironically, ash trees were heavily planted in the 1950’s and ‘60’s following the loss of millions of elm trees (Ulmus) as a result of Dutch elm disease. Since its discovery,
Emerald Ash Borer has killed millions of ash trees throughout eastern North America and has spread as far west as Colorado.
4
While the loss of trees has seriously damaged natural forest ecosystems and urban ecosystems alike, the spread of the insect offers a rare opportunity to observe predictable change over time. Since Emerald Ash Borer targets only ash trees (although there is evidence it may also reproduce in white fringetree (Chioanthus virginicus)), and since many communities have inventories of publicly-maintained trees, it is relatively easy to predict where trees will be lost, and forecast the extent of the damage. The infestation has moved from its original outbreak site in a radiating pattern, and placement of traps in advance of the spread have assisted in announcing the arrival of the pest as the wave progresses. Therefore the predictable spread of the insect and the knowledge of the location and numbers of ash trees in the path of the spread allow researchers to take advantage of this experimental opportunity to observe change of urban forest canopy over time.
Because the spread of Emerald Ash Borer was closely monitored, and the vast destruction of ash trees had been witnessed in southeast Michigan and beyond, it was certain that as the insect pest spread there would be substantial and dramatic changes to tree canopy cover in many areas of the United States and Canada. As this study began in early 2013, it was known that Emerald Ash Borer was present in northeast Illinois, including Chicago and its suburbs. Street tree inventories maintained by local communities showed that ash trees comprised 15% to over 20% of the urban forest in those communities. At the neighborhood block scale it was known that there were some streets that had close to 100% of the street tree population comprised of ash species. In these smaller areas it was possible that over 50% of the total tree canopy was in danger of being destroyed by Emerald Ash Borer within the next 1 to 2 years. At the same time
5 there existed streets with few ash trees, so there would be areas where the tree canopy would change very little in the following 2 years. This situation created a favorable and interesting opportunity to study the loss of a large proportion of trees in a city over time.
As the urban forest experienced this loss of tree canopy, related changes to the urban ecosystem were expected to take place. As fewer trees provided less shade the urban heat island effect would likely increase. As less surface area of foliage was present to intercept precipitation, stormwater runoff would likely increase. As less surface area of foliage was present to absorb and allow deposition of air pollutants, air quality would likely be affected. The literature review section of this dissertation explores the many research studies that quantify the environmental, economic and social benefits of urban forests.
The initial idea for the research described in this dissertation began as a pair of simple
“I wonder” type questions: “I wonder if the sounds heard in communities will change as the loss of trees occurs, and if so, I wonder if the change in sounds will affect people in some way?” The spread of Emerald Ash Borer provided a rare experimental opportunity to observe before-and-after treatment and control conditions that could reveal answers to these questions.
Soundscapes
The study of landscapes, known as landscape ecology considers the living organisms that occupy a landscape, and the non-living resources that support (or deplete) those organisms including natural resources such as soil, water, and air as well as modifications
6 to natural systems including agriculture and the built components of the landscape such as buildings and transportation systems. Maps are an effective tool for studying landscapes.
The study of soundscape ecology emerges from landscape ecology and considers the characteristic of the landscape that we (and other organisms) perceive as sound. In the same way that maps interpret the characteristics of a landscape, sound recordings and their analysis can serve as valuable tools for understanding landscapes and how they change. One of the most notable changes to a landscape as it becomes impacted by human development is the increase in sounds from machines, which may lead to changes in the sounds we hear (or no longer hear) associated with nature.
Noise.
All information that reaches the human brain enters through one of the five senses, and our senses have evolved to function optimally in the range of conditions found on earth. Similar to how the senses work optimally in defined energy ranges for light, sound and temperature, the human brain works optimally within a range of sensory stimuli.
Sensory deprivation and sensory overload can cause the human brain to malfunction.
The word “noise” is universally understood to be a descriptor of unwanted sound.
There are other applications for the word such as in electronic energy that interferes with communication devices, but even in these instances “noise” is usually unwanted and in some cases harmful.
7
Noise, relative to sound energy has been described as any sound that interferes with communication between a source and a receiver. Sound energy from human mechanical devices like vehicles may interfere with birds attempting to attract a mate or announce the presence of a predator. Both of these situations have implications for the growth or depletion of the bird population. Sound energy from the same vehicles may interfere with your conversation with another person perhaps causing you or your colleague to misunderstand what you are each trying to convey. Unwanted sound (noise) can reduce the effectiveness of communication and therefore lead to those organisms that rely on information conveyed by sound to make mistakes. The bird that doesn’t hear the warning of the nearby predator may make the mistake of not taking cover, and the human that doesn’t hear the approaching siren may make the mistake of proceeding into an intersection, both of which may lead to unfortunate but avoidable consequences.
Noise does not only disrupt active and purposeful communication between a source and a listener, but it may also disrupt the passive experience of a landscape. Background noise from distant traffic may interfere with an owl’s ability to hear movement of prey in the dark, affecting the owl’s ability to find food. Sounds from overhead aircraft may detract from the experience of a visitor to a national park who seeks separation from human development. More commonly, seemingly inconsequential yet distracting sounds may pull a person’s attention from an important task, or an uncontrollable barrage of chronic unwanted sounds may cause harmful physiological, psychological, or emotional problems. Have you ever felt your temper rise as your neighbor’s noisy party goes on, and on….and on? Then you understand the implications.
8
The majority of studies that consider urban soundscapes have focused on unwanted, disruptive sound energy and how to manage noise. These studies primarily focus on the perceived “loudness” of noise, which requires measuring the amount of energy carried in the sound waves and the methods of reducing the energy. Many studies have considered the sound attenuation properties of vegetation and have concluded that sound energy that travels between a sound source and a receiver can be reduced (but rarely eliminated) by placing a dense and thick layer of vegetation (or some other type of barrier) between the sound source and the receiver. This is useful information when there are point sources of noise and there is sufficient space between the sound source and the receivers in which barriers can be erected. But this is of limited value in cities where there are numerous sources of mechanical sound energy, many of which are mobile, and numerous sound receivers in the form of humans, also many of which are mobile. Is the solution to unwanted sound and the problems it causes likely to be found by only seeking to manage noise? Perhaps not.
Considering only the extremes of sound energy in the form of noise is missing a larger and more important understanding of how sounds affect humans. A body of research suggests that the reason our senses function as they do is because over the course of millions of years the human body has evolved to synchronize with the conditions encountered in the environment. Many evolutionary anthropologists believe that those individuals whose senses detected (and whose brains interpreted) environmental conditions that indicated available food and shelter, and avoided threats survived longer and reproduced more successfully.
9
In some urban environments the quantity and quality of sensory stimuli may require the human brain to work outside of the “comfort zone” within which human cognitive function performs most efficiently. Many people who have experienced driving in urban rush hour traffic know the feeling of stress and distraction. Fatigue affects our ability to focus, concentrate and accomplish the tasks that are important to us. This ability to concentrate is known as directed attention.
Directed Attention
Directed attention has been defined as “The effortful, conscious process of bringing cognitive resources to bear in order to focus on selected stimuli, while avoiding distraction from unrelated perceptual inputs. Assessment of this ability involves tests used to measure concentration, impulse inhibition and memory” (Bratman, Hamilton, &
Daily, 2012, Page 121). Directed attention requires and consumes a person’s energy because it focuses attention on a narrow task and builds filters or attentional blinders against all of the competing stimuli in the surrounding environment and in the persons’ thoughts that would distract attention away from the task at hand (Basu, 2015). At some point rest from this exertion of directed attention is needed, and that rest and replenishment can be realized when the attentional power of the brain is allowed to relax
(S. Kaplan & R. Kaplan, 1982). There is a different kind of attention called involuntary attention that does not consume energy to concentrate but rather is believed to replenish the energy consumed in directed attention. Involuntary attention, fascination, wonder,
10 day-dreaming, letting-your-mind-wander, are all descriptors of the beneficial and refreshing occasions when your brain gets a break from focusing on a specific task.
Trees, Sounds and People
As the discipline of soundscape ecology has emerged from the study of landscape ecology, a greater understanding of the components of soundscapes has also emerged.
Sound energy, which is perceived as loudness is just one attribute of the complex sensation of hearing in humans. “Loudness” is somewhat synonymous with sound volume or sound quantity. As explored in this study, there are many other sound attributes that more accurately describe a sound’s quality, including pitch, complexity, duration, and perhaps most importantly the nature of the sound source.
It is clear that excessive noise is harmful. In the same way that involuntary attention can counteract the energy-depleting effect of over-used directed attention, can sound of a different quality than noise be beneficial? A simple example is music. If it is of the variety that is pleasing to an individual (a highly subjective preference), music can bring joy, sadness, excitement, or relaxation. Certainly sounds in the form of music can affect our emotions. Just as directed attention can become fatigued, and involuntary attention
(fascination, wonder) can serve to replenish directed attention, there are sounds that can encourage and support involuntary attention. It has been determined through multiple research studies that views of nature, whether experienced as a person moves through a natural setting, or views trees or greenspace out of one’s window, or even views artwork depicting natural scenes, can have therapeutic value such as reducing symptoms of
11 physiological stress or mental fatigue. Therefore, it seems possible or quite probable that soundscapes within urban forests, and in particular the sounds of nature may also enhance involuntary attention, thereby allowing the mind to wander, to engage effortlessly in wonder, and in doing so assist in the replenishment of directed attention. These are the pathways of inquiry that led to this research.
Research Questions
The hypotheses for this research have evolved from these first curious questions regarding trees, sounds and people. The initial research questions were developed into several components including:
1. Will the percentage of tree canopy cover change as the Emerald Ash Borer population
increases in northeast Illinois?
2. Will the quantity and quality of sounds change in areas that have significant changes to
tree canopy cover?
3. Will the changes in the quantity and quality of sounds affect human directed attention?
To investigate these questions, soundscape measurements were recorded in treatment
(tree loss) and control (no tree loss) urban areas both before and after trees are removed.
Leading into the study I anticipated that more human-produced mechanical sounds
(anthrophony) would be detected and fewer sounds associated with wildlife (biophony) and weather (geophony) would be heard as trees were removed. To test these assumptions I collected and analyzed spatially matched pairs (same location separated in time by 1 year) and temporally matched pairs (same time in different locations with different levels of tree canopy) of before-tree-removal and after-tree-removal sound
12 recordings. As results emerged, I turned my focus to the next line of inquiry: As changes to the urban soundscape occur, would there be associated changes to measures of human directed attention? To investigate this question the pairs of sound recordings served as the independent variables for randomly-assigned human volunteers as they performed a pair of standardized tests to measure directed attention. This research advances the current knowledge of how environmental conditions, specifically soundscapes, affect humans.
The research also provides an important link between measurable changes in an urban environment and measurable changes in human cognitive functioning. The research explores overlapping concerns among natural resource managers, urban planners, sociologists, health care providers and public policy officials. Improving our understanding of how urban environments affect physical, mental and emotional health allows practitioners in these disciplines to provide people with better access to nature in cities resulting in improved living conditions.
Components of this Dissertation Document
The chapters that follow present this dissertation research project from its conception through its results and final conclusions. Chapter 2 presents a review of the relevant literature that builds the foundation for the dissertation inquiry. Chapter 3 is a commentary on the body of knowledge outlined in the literature review and builds the connective ideas that become the hypotheses. Chapter 4 presents the research design, including a discussion of the assumptions and choices that were made while designing the project. Chapter 5 presents the methods for data collection relative to the soundscape
13 portion of this study and chapter 6 discusses the analysis of those soundscape data, followed by Chapter 7 that presents the results of the soundscape data analysis. Then the dissertation shifts focus to the topic of directed attention. Chapter 8 outlines the methods used for gathering data for tests for directed attention followed by the results of the analysis of those data, which are found in Chapter 9. Chapter 10 concludes the study with an exploration of the conclusions and discussion in terms of policies and procedures of what was learned during the research and recommendations for future research.
14
CHAPTER II
LITERATURE REVIEW
The research conducted in this study considers the interaction between trees in cities
(the urban forest), how those trees affect the composition of sounds in cities, and how those sounds affect a particular aspect of human cognition, namely directed attention. As will be discussed in Chapter 4 – Research Design, this is the order in which the research gathers data, analyses results, and proceeds to the next line of inquiry, namely how the urban forest is affected by an invasive insect, then how the changes to the urban forest affect the urban soundscape, and finally how the changes to the urban soundscape may affect human cognition. For the benefit of understanding why this order of reasoning developed it is most efficient to begin with an understanding of the component of human cognition that the research will consider, namely human directed attention. Therefore, the literature review is organized in these basic groups. First is a discussion of directed attention – what it is and why it is important to humans; next is a brief review of how access to natural settings in urban forests affects human health; and finally an exploration of soundscape ecology and how sounds reveal information about an environment.
15
Directed Attention
Consider when you were a teenager and you had some unfortunate news that you needed to tell your father. Perhaps you got a speeding ticket – nothing major - and in your mind there were extenuating circumstances that needed to be understood. From previous experience with such situations you knew that the timing of your confession would influence how the bad news was received and how sternly your punishment would be determined. Dad worked hard and when he was tired he had little patience. Dad liked to golf on his days off and was usually in a good mood on those days. If only you could catch dad in a good mood then he might be patient enough to hear your entire story and hopefully render a reasonable judgment to your mistake.
Acting reasonably is important at many scales. Acting reasonably is important in personal relationships, in work relationships, and in government. People tend to act more reasonably when their basic needs (food and shelter) are met, when they do not feel threatened, and they are willing to communicate (including purposeful listening). People tend to act less reasonably and less patiently when they are tired (Sullivan, 2015).
Stephen Kaplan and Rachel Kaplan explored the question of “Why do people act reasonably or unreasonably?” in a seminal paper titled “Health, Supportive Environments and the Reasonable Person Model” (S. Kaplan & R. Kaplan, 2003). The Reasonable
Person Model posits that a person is more likely to act reasonably if three needs for information about a person’s environment are met: the ability to build mental models, the ability to act effectively, and the ability to act meaningfully. Mental models or mental roadmaps result when our brains receive information and patterns are recognized. If
16 there is sufficient information the brain stores the relevant pieces (filtering out those elements that are irrelevant) and then is able to retrieve the information the next time a similar pattern is encountered. This is as basic as being able to recognize moods on faces of people never encountered before and as complex as navigating through multi-layered problems. Acting effectively involves the brain’s ability to receive constant waves of information in the form of sensory stimuli and to be able to make sense of it. Acting effectively requires the brain to sift through the stimuli, determine what is important and should be paid attention to and what is unimportant background information that can be ignored. Acting meaningfully – in a manner that contributes to an individual’s goals and to the larger good of others – is made possible by effectively processing incoming information to the brain, using the information to build accurate mental models, and using those models to navigate through life (Basu & Kaplan, 2015; R. Kaplan & Basu,
2015).
All three of the components of the Reasonable Person Model – building mental models, being effective, and acting meaningfully – are affected by a person’s ability to pay attention to their surroundings, to determine what elements of their surroundings are important and what elements are not important, and to put the important information to good use. This ability to pay attention has been termed “directed attention” (S. Kaplan &
Berman, 2010). Directed attention has been defined as “The effortful, conscious process of bringing cognitive resources to bear in order to focus on selected stimuli, while avoiding distraction from unrelated perceptual inputs. Assessment of this ability involves tests used to measure concentration, impulse inhibition and memory” (Bratman,
Hamilton, & Daily, 2012, Page 121). Because of the focus on selected stimuli this form
17 of attention is also frequently referred to as selective attention. Directed attention requires and consumes a person’s energy. Paying close attention to a detailed task such as completing ones’ income taxes or listening attentively to a lecture on calculus can lead to a feeling of mental fatigue. Directed attention consumes energy not only because it focuses attention on a narrow task, but more importantly directed attention builds filters or attentional blinders against all of the competing stimuli in the surrounding environment and in the persons’ thoughts that would distract attention away from the task at hand (Basu, 2015). At some point rest from this exertion of directed attention is needed, and that rest and replenishment can be realized when the attentional power of the brain is allowed to relax (S. Kaplan & R. Kaplan, 1982). According to Drs. Rachel and
Stephen Kaplan “Directed attention can be thought of as one means of achieving focus in a confusing environment. The achievement of clarity and focus, in turn, can result in the reduction of elimination of the pain generated by uncertainty and confusion” (R. Kaplan
& S. Kaplan, 1995, Page 188).
Attention Restoration Theory posits that when directed attention becomes fatigued that it can be restored by engaging in a different form of attention – involuntary attention
(S. Kaplan, 1995). Unlike directed attention, involuntary attention does not feel like work. Involuntary attention occurs when the mind is allowed to wander in a direction that is pleasing to the person. Involuntary attention can have the opposite effect on fatigue as does directed attention. Given the opportunity to relax and explore calming and interesting thoughts the fatigue caused by over-exertion of directed attention can be reversed. The degree to which a person experiences fatigue from extended use of directed attention and the degree to which involuntary attention is fostered and allowed to
18 replenish a fatigued mind is affected by a persons’ surroundings. And some types of surroundings tend to encourage involuntary attention and relief from fatigue more than others. This examination of fatigue and recovery relative to a person’s surroundings inspired me to consider the components of outdoor environments that have brought a sense of “peace and quiet” to my own experiences.
Involuntary attention, also called “fascination” by Rachel and Stephen Kaplan, allows directed attention to rest and recover. There are many settings or activities that allow fascination to take over from directed attention, and natural settings seem to have certain advantages.
In addition to fascination, Stephen Kaplan suggests that there are three additional components that enhance a settings’ effectiveness as being restorative; the setting must provide a feeling of “being away”; the setting must have extent, that is it must be interesting enough to pull a person’s attention away from thoughts that cause fatigue; and the setting must be compatible with a person’s interests (S. Kaplan, 1995).
Dr. Kaplan quotes Fredrick Law Olmsted’s thoughts on restoration, saying that natural scenery “employs the mind without fatigue and yet exercises it; tranquilizes it and yet enlivens it; and thus, through the influence of the mind over the body, give the effect of refreshing rest and reinvigoration to the whole system” (S. Kaplan, 1995, Page 174;
Olmsted, 1995, Page 22). These investigations and observations by Frederick Law
Olmsted and later by Stephen Kaplan are enlightening, and yet seem intuitive. My own intuition tells me that sensory stimuli in outdoor settings that convey a feeling of safety allow relaxation to occur which in turn allows the mind to wander, while settings that
19 convey frequent warning signals or concern of being lost initiate feelings of alertness or worry.
A growing body of research considers the benefits of access to nature for humans, some of which specifically addresses the effects of nature on human directed attention, and some of which considers alternate psychological and physiological functions.
The Benefits of Urban Forests and Access to Nature to Human Health
Introduction.
The sites of most of the early settlements in North America were chosen because of their access to natural resources, and for many this meant proximity to navigable water.
Villages along rivers and lakes rapidly grew into towns and cities because the water body meant goods could be transported to the location and trade could occur. New York,
Boston, Philadelphia, Baltimore, Buffalo, Montreal, Toronto, Pittsburgh, Cleveland,
Cincinnati, Detroit, Chicago, Minneapolis, St. Louis, New Orleans and many other cities were located on waterways and grew from there. Many other communities were established near sources of natural resources needed for industry, such as iron, copper, coal, and timber, and other communities were located as trade centers for agriculture. The settlements of North America were developed largely because of some tie to natural resources.
Unfortunately, many of the same water sources that were essential for transportation also became convenient sewers for eliminating human and industrial waste. Almost
20 without exception, all of the major industrial cities experienced tremendous pollution problems during the late 1800’s through the mid 1900’s. Much of the waste that didn’t wind up in the surface waters was incinerated and sent up smokestacks and into the air.
In her book “Healing Spaces” Dr. Ester Sternberg writes of the deplorable pollution of
19th century European and American cities and the tremendous negative health effects caused by unclean water and air. Communicable diseases like tuberculosis and cholera were rampant, and the infant mortality rate in many cities was twice that of the surrounding countryside (Sternberg, 2009).
Conditions in cities improved immensely when antibiotics were discovered, sanitary conditions improved, and urban population density relaxed. There eventually came to be a health advantage to living in cities as compared to rural areas due to proximity to health care.
It is estimated that today approximately 80% of the population in Canada and the
United States live in urban areas (U.S. Census, 2000). With so many people it is appropriate to look deeper into the relationship between humans in cities and the environment. Polluted air and water can have harmful physical effects on humans, but are there other, possibly more subtle interactions between humans and the environment that also affect the quality of life?
Quality of life for humans is certainly a function of physical health and also a function of mental and emotional health. It is helpful also to look at the published research in this area, and to understand the methods of inquiry that researchers use to test the correlation between the quality of urban ecosystems and the quality of life for humans. There is a wealth of literature that describes the many environmental and economic benefits of
21 urban forests. For the purposes of this dissertation research I will focus primarily on the benefits of urban forests that are relevant to human health, particularly human cognitive functions.
Urban forests and human physical health.
Urban ecosystems are the collection of natural resources that support life in the communities where people live. Urban ecosystems include all of the non-living resources including air, water and soil plus all of the living organisms, including humans, that utilize these resources. Among the living organisms that provide the greatest range of environmental, economic and social benefits to humans in cities are trees, which explains the great interest in quantifying and qualifying the benefits of urban forests.
It is well documented that urban forests improve air quality by reducing pollutants
(Beckett, Freer-Smith, & Taylor, 2000), reduce extreme summer temperatures (Laverne
& Lewis, 1996, 2000), and reduce stormwater runoff (Mateo, Randhir, & Bloniarz,
2006), which in turn reduces surface runoff of pollutants and reduces the frequency and severity of flooding events. All of these benefits have direct implications toward human health and safety. Increasingly research is revealing that in addition to these direct effects on human health associated with air, water and temperature there are aspects of urban forests that may improve your disposition, your mood, your desire to be active, and even your ability to concentrate.
22
Exercise and muscle-powered transportation.
Air and water quality are variables that affect human health through the physiological uptake of substances. Human health is also affected by leisure resources that lead to exercise, and here also is an area influenced by urban forests and access to nature. Where urban forest effects on air and water quality directly affect human physiological function, urban forest effects relative to exercise affect a person’s physiological function through their psychological desire to be outside. Trees and access to nature provide a more inviting setting in which to be physically active. According to Robert Marans (2003)
“environmental and urban amenities are related to community quality and individual activities, satisfactions, and physical health. Environmental amenities include both natural recreation resources (NRR) such as lakes, rivers, wetlands forests and park land and the quality of the ambient environment (EQ) including air and water, noise, and solid and hazardous waste” (Marans, 2003, Pages 75 & 77). Public spaces including parks and streets that include appropriate landscaping, summer shade, and environments where people feel safe to participate in muscle-powered transportation (walking, running, cycling) have been shown to be well used in communities throughout the world.
Landscapes that invite exercise can lead to greater physical activity that in turn may lead to weight loss which has been shown to improve conditions associated with heart disease, diabetes, and respiratory disease (Mitchell & Popham, 2008). Can the process of improving one’s physical condition also affect one’s mental and emotional functioning?
“Health” in a broader sense is not limited to physiological functions, but can also be considered as the state of a persons’ complete physical, mental and social well-being.
23
There is a growing body of evidence that suggests that access to nature does indeed affect physical, mental and social functions (Kuppuswamy, 2009). Humans develop a positive or negative “sense of place” associated with the areas where they live and work, which can have an important impact on how a person perceives the quality of their life. Howard
Frumkin lists nature contact, buildings, public spaces, and urban form as the four aspects of the built environment that foster a “sense of place” for residents of cities (Frumkin,
2003).
Many people, perhaps most people feel some positive feeling associated with being outdoors and near “nature.” Of what value (not necessarily in financial terms) is this feeling, and how does it benefit individuals and the community? John Dwyer and his colleagues comment, “The strong ties between people and trees cannot be explained by increased property values, reductions in air pollutants, and moderation in temperature.
The psychological ties between people and trees defy easy quantification, yet few would deny their existence or their profound implications for urban forest management”
(Dwyer, Schroeder, & Gobster, 1994, Pages 137 – 138). Mental health and well-being can be affected in a variety of settings and can result in subtle or dramatic changes in humans. The following discussion considers the difficulty in quantifying the effects of nature on mental health.
Human physical health recovery through mental well-being.
Can access to nature help people suffering with medical problems? A number of research studies have considered this question. Possibly the most ground-breaking
24 research and certainly one of the most frequently cited projects was conducted by Roger
Ulrich in early 1980’s. He studied patients in a Pennsylvania hospital who had all undergone a type of gall bladder surgery. Following the surgery patients were randomly assigned to hospital rooms that either had a window that had a view of greenspace with trees or had a window with a view of an adjacent building. Ulrich found several differences in measures of recovery from surgery between the two groups. The difference in length of hospital stay was significantly shorter for the patients housed in rooms with a view of greenspace. There was a slightly lower incidence of negative reports from nurses regarding patients housed in rooms with a view of greenspace, but the difference was not statistically significant. There was a statistically significant difference in analgesic (pain management) doses between the groups for recovery days 2 through 5 (but not significant for days 1, 6 or 7) with the patients having a view of greenspace needing less pain management medication. There was no significant difference in amount of anxiety drugs administered, or weighted score of minor post- surgical complications (Ulrich, 1984).
In another study, Cimprich and Ronis (2003) considered 157 women who had recently been diagnosed with breast cancer. The patients were divided into intervention (test) and non-intervention (control) groups. Those in the intervention group got all of the standard medical care plus they participated in a home-based program in which they spent at least
120 minutes per week exposed to the natural environment. Activities could be active such as hiking in the woods or passive such as watching birds or rain through a window.
The activity was determined by the patient. The non-intervention group got all of the regular medical care but did not actively seek out activities tied to nature. Both groups
25 were tested on measures of the cognitive capacity to direct attention (CDA) shortly after diagnosis but before surgery, and again after surgery. The intervention group showed greater recovery of capacity to direct attention. Why is this important? In addition to being better able to cope with daily issues, patients with an improved capacity to direct attention should be able to follow the sometimes complicated instructions of post-surgical regimens including taking proper doses of medication at instructed times.
Other studies in hospitals have found that patients with access to healing gardens within the hospital grounds report a greater sense of well-being and a higher level of satisfaction with the hospital experience than patients without access to healing gardens
(Hartig & Marcus, 2006; Sternberg, 2009). The term “healing” might be deceiving in that the gardens are not intended to cure a disease directly. But rather the gardens are intended to facilitate some movement toward improvement of health, whether that be physical health, psychological health or spiritual health. Terminally ill patients may find a garden as a place for contemplation. Visitors and medical staff may find a garden as a place for stress recovery.
A study conducted by Sherman, Varni, Ulrich, and Malcarne (2005) found the healing gardens at one hospital to be used predominantly by visitors and staff. The study considered three healing gardens located at a children’s cancer hospital in southern
California. Sixty hours of behavioral observations were recorded (20 hours in each garden). In addition to observing people physically in the gardens, observers recorded whether window blinds were open or closed for hospital rooms that had a view of the gardens. The researchers were surprised to find that overall, only 4% of the users of the three gardens were patients – the remaining 96% of users were either hospital staff or
26 visitors to the hospital. A small number of patient evaluations were conducted and the researchers found lower (although not statistically significant) distress scores across all six measured domains (anxiety, sadness, anger, worry, fatigue, and pain).
The view from home and work.
A number of important studies have been conducted in the network of public housing facilities in Chicago. Many of these units are essentially identical in size and layout. A benefit of studying the population of these units is that residents are randomly assigned to the locations, which provides a convenient experimental design allowing researchers to hold demographic variables constant while considering the effects of landscape variables.
One of the studies based in the Chicago Public Housing developments considered the effect of the view from an apartment window on measures of adolescent emotional development. University of Illinois researchers, Faber-Taylor, Kuo, and Sullivan outlined three forms of self-discipline: concentration, inhibiting initial impulses, and delaying gratification. They also identified the risks associated with weakness in each of the three areas (academic underachievement, juvenile delinquency, and teenage pregnancy respectively). They suggest that each of the three forms is influenced by directed attention. As a test to determine if visual access to nature from a home window affects these measures of emotional development, mothers of adolescents were asked,
“How much of the view from your window is of nature?” and “How much of the view from your window is man-made?” Then standardized tests were given to the children to test concentration, inhibition of initial impulses, and delay of gratification. The
27 researchers found that the view of nature from a student’s home positively affects concentration, inhibiting initial impulses, and delaying gratification, but only for girls.
The same tests were inconclusive for boys. The researchers state: “If near-home nature can provide a daily, easily accessible means of supporting impulse inhibition and delay of gratification in a setting where individuals are likely to be chronically mentally fatigued…the implications for individuals, families, and society may be enormous”
(Faber Taylor, Kuo, & Sullivan, 2002, Page 60).
Rachel Kaplan conducted a study that asked residents of apartment buildings to describe the view from their apartment window and then collected information on the resident’s degree of satisfaction with their neighborhood. She found that views of trees were important to resident perceptions of being “at peace” and not being “distracted”; views of landscaped areas and gardens were important to “effective functioning.” A view of a busy street had a negative influence on satisfaction with the neighborhood (R.
Kaplan, 2001).
In a similar study also conducted by Rachel Kaplan questionnaires were sent to employees of office buildings. In the questionnaire were photographs of various office complex landscapes, and the participants were asked to rank them for their personal preference. The questionnaire also asked about what they liked and didn’t like about the building grounds where they worked. The results showed that office workers had a greater preference for views of large trees than for manicured lawns. The results also suggested that respondents would find more natural landscapes that included native plants such as prairie grasses acceptable replacements for mown turf, as long as a walkway or path was provided (R. Kaplan, 2007).
28
In summary, views from within buildings to outdoor greenspaces can provide brief restorative experiences that interrupt depletion of directed attention and relieve feelings of stress (van den Berg, Hartig, & Staats, 2007).
Well-being and noise.
Quality of life in urban neighborhoods is affected by many variables including levels of noise. Noise from dense traffic, construction, and emergency vehicles is reflected by hardscape surfaces. Loud, persistent noise is known to negatively affect reported levels of stress in humans (Nassiri et al., 2013), but recent studies suggest that more pleasing sounds may positively affect human health (Alvarsson, Wiens, & Nilsson, 2010). The effects of sound on human quality of life are explored more fully in the soundscape ecology section of this literature review.
Physical & mental response following exposure to nature and urban settings.
Terry Hartig and his colleagues have done several studies that combine asking participants about their sense of well-being and measuring physical indicators of attention recovery during and after exposure to natural landscapes and urban settings. They have found that those people who are subjected to nature following stressful events show superior ability to recover as compared to people subjected to urban settings (van den
Berg et al., 2007).
29
Hartig, Mang, and Evans (1991) speculate that exposure to nature may actually serve as a proactive inoculation against the effects of stress, meaning that access to nature may better prepare a person to deal with stress after the outdoor experience as well as helping a person restore attention following exposure to stressful situations.
Theories regarding the human response to nature.
The many research studies summarized above show compelling evidence that access to nature for urban dwellers provides an array of physiological and psychological benefits. But how does this beneficial effect work? There are two main camps of thought that address that question. As previously discussed in the preceding section on directed attention, the first functional theory came from Rachel Kaplan and Stephen
Kaplan. They suggest that access to nature functions by restoring a person’s capacity to pay attention to tasks.
A number of other researchers have conducted studies that support the theory of restoration of directed attention including the studies conducted at the Chicago public housing developments by Kuo and Sullivan and their colleagues, the study conducted by
Cimprich and her colleagues on recovery of breast cancer patients (Cimprich & Ronis,
2003), and others (Laumann, Garling, & Stormark, 2003).
Frances Kuo and William Sullivan offer supporting discussion for this theory. They state that mental fatigue leads to an inability to focus, or what Stephen Kaplan refers to as depleted directional attention. Kuo and Sullivan suggest that mental fatigue leads to a heightened propensity for anger and violence, and that three symptoms of mental fatigue
30 might contribute to aggression. The first symptom is a diminished ability for cognitive or effortful processing. This means that fatigue causes an individual to be more emotional and less rational when evaluating a situation. The second symptom is heightened irritability. The authors note that urban environments are frequently crowded, noisy and hot – all of which can contribute to irritability. The third symptom is impulsivity – acting without reasoning the situation through. Each of these three symptoms – impaired effortful cognitive processing, irritability and impulsivity - have all been shown to be linked to aggression. So if mental fatigue (depleted directional attention) leads to aggression and violence, then methods of mitigating mental fatigue such as access to nature may result in decreased violence in urban populations (Kuo & Sullivan, 2001a).
The other primary theory found in the literature was offered by Roger Ulrich and his colleagues. Instead of restoring a person’s ability to pay attention, Ulrich believes that access to nature works as a counterbalance to stress, and stress is a serious problem for residents in crowded cities.
To demonstrate this Roger Ulrich and his colleagues conducted a study where 120 college students were subjected to watching a video that was about preventing industrial accidents and included blood and gore. Then 40 of these students watched a video of urban traffic and pedestrians, 40 watched a video of people walking in a mall, and 40 watched a video with an outdoor scene with trees and a stream. Physiological measurements were recorded during all sessions that recorded heart rate, pulse, skin conductance (sweating) and facial muscle tension. After viewing each of the videos verbal responses from participants were also recorded. Questions inquired about levels of fear, anger/aggression, positive effects, sadness, and attentiveness/interest. The people
31 that watched the nature video recovered from stress quicker as measured by both physiological indicators and verbal responses. This suggests that recovery from stress has both psychological and physiological components. It’s important to note that recovery happened within minutes, suggesting that brief exposure to nature in cities, like a drive through a tree-lined street or lunch in a park can have restorative effects (Ulrich et al., 1991).
Several other studies offer evidence that supports this theory. In one study twelve
Japanese male students were divided into two groups. One group sat for 15 minutes in a forest and the other group sat during the same period in a city. Then each subject was measured for pulse rate, blood pressure, and cortisol concentration in saliva (an index of stress response). The next day the two groups switched locations and repeated the procedure. Subjects were also questioned about their level of comfort (if they felt soothed and refreshed). Significant differences were found in lower cortisol concentrations, lower pulse rates, and lower diastolic blood pressure for forest viewers.
There was no significant difference in systolic blood pressure between forest viewers and urban viewers. Participants also reported feeling more comfortable, soothed, and refreshed after viewing the forest than after viewing the city (Lee, Park, Yuko, Takahide,
& Yoshifumi, 2009).
As will be discussed in greater depth in an upcoming section on soundscapes, the restorative benefits to stress are not limited to the visual experience of entering natural landscapes. Sounds can also affect measures of stress. According to soundscape ecologist Bernie Krause (2015) “Anecdotal evidence strongly suggests that natural soundscapes and biophonies, in particular, may lower stress indicator levels
32
(glucocorticoid enzyme, heart rate, blood pressure, and so on) in humans far more successfully than environments saturated with music, because music is culturally based and may actually produce the result opposite from what it was clinically intended or hoped for” (Krause, 2015, page 47).
Ulrich and his colleagues offer a contrast between their theory that stress reduction is the primary benefit of access to nature and the Kaplan’s belief that attentional restoration is involved. One aspect they point to is their belief that stress recovery is a biological evolutionary function that needed to occur quickly. Nature acting as a restorative environment (as Rachel and Stephen Kaplan suggest) implies working more at the emotional level of modern humans that are chronically over-stimulated, as opposed to an evolutionary psychophysiological process that developed through evolution (Sherman et al., 2005).
Ester Sternberg also writes about the human response to stress and supports Ulrich’s theory of stress reduction through access to nature. In her book chapter with Krista
Tippet (Tippett & Sternberg, 2010) and also in her book Healing Spaces (Sternberg,
2009) Dr. Sternberg describes “stress” as an emotional affliction that affects physiological responses, which was not recognized within the medical community until the 1950’s. It is a very recent advancement that science has accepted the mind-body connection and has come to understand how human emotions affect human health. Dr.
Sternberg explains how hormones that suppress inflammation and turn down the immune system are released during times of stress. When a person has prolonged periods of stress, such as in a difficult marriage or when caring for an Alzheimer’s patient, the suppressed immune system may be unable to adequately guard against infection. The
33 body’s reduced ability to produce inflammation can also lead to arthritis and other problems. She believes that access to nature can help reduce the effects of stress.
The difference between the two theories may seem subtle, but Ulrich’s theory of stress reduction is based on a person’s physiological response to irritating and pressure-filled situations. Rachel and Stephen Kaplan’s theory of attentional restoration is based on a person’s psychological response to the same types of situations. In a detailed comparison of the Kaplan’s Attention Restoration Theory and Ulrich’s Stress Reduction Theory
Gregory Bratman, Paul Hamilton and Gretchen Daily (2012) state:
As Ulrich stresses the importance of the evolutionary aspects of response to
environment, he tends to emphasize affective and stress-related components of
the individual’s relationship with landscapes. The Kaplans’ theory is centered
more on effects on cognition. Thus Ulrich emphasizes the importance of a
reduction in arousal, with physiological evidence showing decreased stress
levels in subjects when viewing natural versus urban images. This contrasts
with ART [Attention Restoration Theory], which is more concerned with a
replenishment of attentional capacities (Bratman, Hamilton, & Daily, 2012,
Page 125).
Stated simply, one theory is based in the body’s response to urban pressures and the other is based on the mind’s response. Terry Hartig and his colleagues have presented several papers that span the gap between these theories (Hartig et al., 1991; van den Berg et al., 2007). By measuring aspects of both stress and attentional fatigue the research suggests that impacts on attention and stress may happen at different times and that there may be different causal pathways for the positive impact of both attention restoration and
34 stress recovery. The comprehensive review of the theories of restorative benefits of nature by Bratman, Hamilton, and Daily (2012) provides an in-depth discussion of many published studies related to this topic.
Spiritual / emotional enrichment.
“…people possess an inherent inclination to affiliate with the natural world
that encompasses the quest for empirical understanding as well as the search
for spiritual meaning and transcendence…” (Kellert, 2002, Page 51)
The quantification of urban forest benefits began with measuring attributes of the environment such as air quality and water quality and projecting what these benefits mean to human health and well-being. It is more difficult to capture the effects of access to nature by measuring responses in humans themselves. Within the quest to measure the human body’s response to nature it is relatively easy to record physiological attributes such as heart rate and even hormone levels. It seems to be somewhat more difficult to measure the human psychological response to nature, but measures of cognitive function can be recorded. Researchers continue to look at the relationship between nature and the human body and the human mind. What is left?
An area that many humans speak of frequently and openly is how being in nature affects them emotionally and even spiritually. Most of us have personally heard someone, perhaps even ourselves say something like “I feel closer to God when I’m in the forest” or “I feel a sense of peace and harmony with Nature.” We plant trees in memory of relatives and friends who have died. Almost every faith that humans practice
35 from Buddhism to Islam to Judaism and Christianity have teachings on the place of nature and trees in faith. And yet how can research demonstrate such a correlation when there is little scientific evidence for God? (Although some would argue plenty of evidence and others would argue no evidence. That is a debate outside of this discussion.) How does a researcher first find the human soul and then measure it before and after exposure to a tree?
An interesting start to this quest is presented by Gretel Van Wieren (2008). This is not a research paper but rather an opinion that provides perspective on the connection between the work of ecological restoration and spiritual enrichment. It is interesting (and perhaps to the author’s credit) that throughout the paper there is no mention of any type of religious belief (such as Christian, Jewish, Buddhist, etc.) The premise of the paper is that people (especially volunteers and non-professionals) who become involved with environmental restoration projects (such as brownfield clean-up or stream restoration) gain from a feeling of reconnection between the human spirit and nature. The implication is that through an act that feels like assisting in healing, humans can feel like they are doing something of moral value, and gain spiritual enrichment. This happens because there is direct connection (not just observation) from humans to nature, and it also provides a common connection between like-minded people. The author writes: “… restoration activity can enable a de-centering, or loss, of the self that comes through the realization that humans are dependent on and interdependent with larger Nature. An important part of this decentering is the feeling of human limitation, and, in turn, humility that restorationists often experience” (Van Wieren, 2008, Page 248). This insight
36 suggests that there is value in discovering that we (humans) are not the center of the universe around which all of Creation spins, but rather we are part of the mix.
Perhaps the question should set aside aspects of spirituality and instead focus on what improves us as compassionate humans (recognizing that compassion can be demonstrated by people of all faiths and of no faith). This seems possible since many attempts have been made at considering the effect of access to nature on crime (which might be considered as the opposite of compassion).
The title of one research paper asks an interesting question: “Can Nature Make Us
More Caring?” (Weinstein, Przybylski, & Ryan, 2009). The authors begin by making a distinction between intrinsic and extrinsic value sets. Intrinsic values or aspirations deal with goals that satisfy basic psychological growth (such as intimacy and community).
Intrinsic values are not fully self-centered. Extrinsic values or aspirations focus on externally valued goods believed to bring positive regard from others (money, image, fame). Extrinsic values are self-centered. The premise of this paper is stated thusly:
…when people are in contact with natural scenes or living objects they will
demonstrate a more intrinsic value set, orienting them to greater connection
and a focus on others. In contrast, exposure to non-natural and artificial
environments will elicit more extrinsic goals… Given the literature discussed
here, we suggest that natural environments, unlike human-made
environments, can increase valuing of intrinsic aspirations and decrease
valuing of extrinsic aspirations because natural environments create
experiences fostering autonomy and nature relatedness (Weinstein,
Przybylski, & Ryan, 2009, Page 1316).
37
The paper then describes several studies conducted to test this hypothesis. The first study involved 98 people who were shown photographs of landscapes of either urban or natural settings. The participants were then provided questionnaires designed to assess their level of immersion into the scenes, their level of positive affect, and their life aspirations. Regression analysis was used. The results revealed that participants exposed to images of natural environments valued intrinsic aspirations and devalued extrinsic aspirations. The second study involved 112 people using similar methods but added questions related to connectedness to nature and autonomy. The results were similar as in Test 1. The third study involved 85 participants using similar methods to the first two tests but added a decision task. Results were similar to the first two tests plus people were found to be more generous when exposed to the nature scenes than the urban scenes. The fourth study involved 75 participants. Instead of exposing participants to images of nature or urban scenes, participants were interviewed in either an office with 4 large plants or an office with no plants. Results were consistent with previous studies – the simple addition of plants to the office setting resulted in higher valuing of intrinsic aspirations while the office with no plants resulted in participants reporting higher valuing of extrinsic aspirations. The presence of plants was also positively correlated with feelings of more autonomy and generous decision-making.
The results are interesting in that all four tests gave consistent results relative to intrinsic and extrinsic aspirations. They are also interesting in that exposure to nature increased intrinsic aspirations and exposure to urban settings increased extrinsic aspirations, suggesting that not only is exposure to nature beneficial to human feelings of generosity, but exposure to urban settings is harmful. The bottom line is summarized by
38 the final sentence of the paper: “Our results suggest that to the extent our links with nature are disrupted, we may also lose some connection with each other. This relation, if sustained, has broad implications for environmental psychology and community design.”
(Weinstein, Przybylski, & Ryan, 2009, Page 1328)
As will be discussed in the section “The effect of anthrophony, biophony and geophony on humans” Moser (1988) found that the intensity of urban sounds may also affect human propensity for helpful behavior.
Soundscape Ecology
Much of the research concerning the effects of urban conditions on humans has focused on what we see of cities, either out of our windows or how we respond to images of landscapes with varying degrees of vegetation and built elements. Vision however, provides a limited field of sensory stimuli, where hearing is broader and more far- reaching. Based in the study of landscape ecology, the emerging field of soundscape ecology allows us to investigate landscapes with sound. A pioneer in the field of soundscape ecology, Krause notes:
In a matter of seconds, a soundscape reveals much more information from
many perspectives, from quantifiable data to cultural inspiration. Visual
capture implicitly frames a limited frontal perspective of a given spatial
context. But soundscapes widen that scope to a full 360-degree hemisphere –
completely enveloping us. Based on the data these records show, accurate
projections about habitat sustainability can be made concerning the effects of
39
human enterprise, like resource extraction and land transformation (Krause
2015, page 126).
On the topic of the value of listening to a landscape Krause notes: “A picture may be worth a thousand words, but a natural soundscape is worth a thousand pictures” (Krause
2015, page 43).
The term “soundscape” was coined by urban planner Michael Southworth who wrote about the characteristics of sounds in Boston’s public spaces (Southworth, 1969). The term was then made popular by R. Murray Schafer, a Canadian musician and composer who initially became interested in natural sounds as inspiration for music. The more he listened to sounds in the environment, the more he recognized how sounds reflected the characteristics of the place, and how humans impacted the sounds of a place. In the introduction to his seminal book The Soundscape: The Sonic Environment and the Tuning of the World, Schafer (1994) writes:
The soundscape of the world is changing. Modern man is beginning to inhabit
a world with an acoustic environment radically different than any he has
hitherto known. These new sounds, which differ in quality an intensity from
those of the past, have alerted many researchers to the dangers of an
indiscriminant and imperialistic spread of more and larger sounds into every
corner of man’s life (Schafer, 1994, Page 3).
Schafer also moved beyond the perspective of urban sounds as unwanted noise to be abated. He recognized that in addition to noise pollution an urban soundscape had favorable sound components that helped define the spaces. Schafer introduced three
40 components of soundscapes: keynote sounds, signals, and soundmarks. Keynote sounds can be thought of as the background sounds of a place. Schafer writes:
Even though keynote sounds may not always be heard consciously, the fact
that they are ubiquitously there suggests the possibility of a deep and
pervasive influence on our behavior and moods. The keynote sounds of a
given place are important because they help outline the character of men
living among them. The keynote sounds of a landscape are those created by
its geography and climate: water, wind, forests, plains, birds, insects and
animals. Many of these sounds may possess archetypal significance; that is,
they may have imprinted themselves so deeply on the people hearing them
that life without them would be sensed as a distinct impoverishment. They
may even affect the behavior or life style of a society (Schafer, 1994, pages 9-
10)
Schafer defines signals as foreground sounds that are listened to consciously. Some of these signals carry messages, such as sirens, that demand attention. Soundmarks are acoustic landmarks, such as church bells that help define a community (Schafer, 1994).
According to Dr. Brian Pijanowski of Purdue University’s Laboratory of Human
Environment Modeling and Analysis, “Schafer shifted emphasis away from just the noise aspect, which had been the main focus when thinking about soundscapes up until that point, to considering more positive soundscape design” (Hawkins, 2012, Page 6).
Bernie Krause, who has invested his career in recording soundscapes around the world commented on the range of disciplines that the study of soundscapes affects: “…aspects of the soundscape inform disciplines as far-reaching as medicine, religion, politics,
41 music, architecture, dance, natural history, literature, poetry, biology, anthropology, and environmental studies” (Krause 2015, page 7). Krause also recognizes the implications for urban studies: “…as our lives became more urban-centered, the connections to those guiding beacons of the natural world began to lose their significance and consequently grew to be scarcely acknowledged” (Krause 2015, page 8).
Most recently it has been Dr. Pijanowski and his colleagues that have advanced the field of soundscape ecology. In an interview with Ecologist magazine, Pijanowski commented on the diversity of professional disciplines who have collaborated to study soundscapes:
It's based on a lot of work that's been done by bioacoustics experts - who have
been studying birdsong and vocalization and communication in animals for
decades. We're also building upon the work of acoustic ecologists who have
turned their ears to natural sounds and who are often musicians. They have
provided us with a rich vocabulary to begin to think about these natural
soundscapes. Then there are cognitive psychologists who know how
vertebrates process acoustic information and how certain kinds of sounds can
give us an emotional response….. soundscape is fully a reflection of the
landscape (Hawkins, 2012, Page 5).
Up until Schafer’s work, sounds in urban environments had been studied mostly quantitatively from the perspective of monitoring the intensity of noise and seeking methods of attenuating the offensive or irritating sounds. As Schafer identified both negative and positive components to urban soundscapes, categories of sounds were identified. Krause (2015) explains the three categories of sounds within a soundscape:
42
The first is geophony, the non-biological natural sounds produced in any
given habitat, like wind in the trees or grasses, water in a stream, waves at the
ocean shore, or movement of the earth. The second is biophony, the collective
sound produced by all living organisms that reside in a particular biome. And
last is anthropophony1, or all of the sounds we humans generate. Some of
these sounds are controlled, like music, language, or theatre. But most of
what humans produce is chaotic or incoherent – sometimes referred to as
noise (Krause, 2015, page 12).
For the purpose of research outlined in this dissertation, anthrophony will be divided into the following sub-categories:
• Anthrophony Type 1: sounds from human voices and music
• Anthrophony Type 2: sounds from human mechanical devices.
The proportions of biophony, geophony, and anthrophony in a soundscape, and particularly the diversity of sounds within the biophony may be indicators of the health of an ecosystem (Farina, Lattanzi, Malavasi, Pieretti, & Piccioli, 2011; Hawkins, 2012), which may in turn may also have implications for the physical and mental health of the people within the ecosystem. The change in proportion of sound categories over time may be indicators of changing ecosystem health (Pijanowski et al., 2011).
1. Most of the soundscape ecology literature employs the term “anthrophony” to refer to human-sourced sounds. However due to the Greek origins of the prefix “anthro” meaning caves, and not human, the more accurate term for human-sourced sounds is “anthropophony.” For consistency with the majority of published literature on the topic I will continue to use the more popular if less accurate spelling of anthrophony.
43
According to Dr. Pijanowski, “The idea is to study the patterns of all of these, how they occur and emerge in different landscapes around the world. From this we can learn about ecosystems and how they function, and how these ecosystems might be threatened by the measure of anthrophony, because anthrophony provides us with a reflection of the amount of human activity that occurs in a landscape” (Hawkins, 2012, Page 5).
There is a fundamental difference in the function of biophony and anthrophony - type
2 (the mechanical sounds). Animals including birds, amphibians, insects and mammals use sound to communicate for the purposes of mating, territory defense, location, and predator alert. Human mechanical sounds contain little or no communication value but can overpower and interfere with biophony (Farina et al., 2011; Pijanowski et al., 2011)
(as will be discussed in a following section). The increase in human mechanical noise has prompted some bird species to alter the intensity and/or pitch with which they sing.
Other species, particularly amphibians and insects that cannot alter intensity or pitch may experience reductions in reproductive rates (Farina et al., 2011). If urban noise can interfere with the functions of animals, can it also interfere with the functions of humans?
As will be discussed in the upcoming section titled “The effect of anthrophony, biophony and geophony on humans” the answer is yes, but the full extent to which a soundscape affects human cognitive function, both negatively and positively is not yet fully understood (Schulte-Fortkamp & Fiebig, 2006). This dissertation seeks to advance the study of sounds within a landscape as an indicator of ecosystem health, and since ecosystems support human life, how soundscape ecology might be used as an indicator of human quality of life.
44
Noise attenuation by vegetation.
According to Wiley (2015) sound attenuates during transmission from the source to the receiver in three ways. First is attenuation as the energy spreads from the source. As the sound moves the energy covers an increasingly large area and the energy becomes diluted. The second type of sound attenuation is absorption, which happens when the energy from a sound wave is transferred as minute movements of particles within the medium (such as air or water) through which the sound wave travels. Vibrations induced in vegetation by absorption are very small when compared to absorption by the medium through which a sound wave travels. The third type of sound attenuation is by scattering of sound waves by physical objects including trees. Foliage, branches and trunks of trees reflect sound energy in many directions reducing the amount of sound energy that reaches a receiver.
The degree of sound energy scattering is strongly dependent on frequency, and therefore wavelength. If the wavelength of the sound energy is much smaller than the diameter of the object it encounters, most of the sound energy will be reflected backwards and perceived by the listener as an echo. When the wavelength is approximately equal (from one-tenth to ten times the diameter) to the size of the object it encounters the reflection of the sound waves becomes much more scattered in many directions. Trunks and large branches of trees scatter sound most effectively in the wavelengths above 1,000 Hz, broad-leaf foliage scatters sounds above 2,000 Hz to 4,000
Hz, and conifer foliage scatter sounds most effectively above 4,000 to 8,000 Hz. These are the frequency ranges within which many animals and birds communicate by sound.
45
Numerous studies have considered the noise attenuation properties of trees and other vegetation. In general, the density of the vegetation and the height of the vegetation are the variables that affect noise attenuation the most. Dense vegetation at the same height as the source of the noise and the receiver reduced the amount of sound energy at the receiver more effectively than taller trees (Fang & Ling, 2003).
In his book Urban Forest Acoustics, Bucur (2006) explains the sound attenuation variables of individual trees: “Scattering effectiveness is consistent with the geometry of the scatterers such as trunk, branches, and leaves. The bigger the scatterer, the lower the frequency at which the scattering phenomenon becomes effective” (Bucur 2006, page
53). He continues to then explain the sound attenuation variables of stands of trees:
…the main dendrological, and physical characteristics of the stand effecting
excess attenuation in a tree belt…are: the biomass of the stand, the structure of
the stand in a horizontal plane (size and shape of the canopy) and the quality
of the surfaces (size and shape of the leaves and needles, soil). The
characteristics allow admitting that mixed stands composed of coniferous and
deciduous trees and bushes would be the most effective for noise attenuation
(Bucur 2006, page 54).
Kragh (1981) found that rows of trees 3 to 25 meters deep and positioned between a road and the sound receiver were no better at dampening traffic noise than grass-covered ground except for frequencies above 2,000 Hz. A later study (Samara & Tsitsoni, 2011) came to a different conclusion. A row of pine trees positioned between a roadway and the sound receiver significantly improved noise attenuation, suggesting that the species of vegetation and the arrangement of the vegetation affects the degree of noise attenuation.
46
Another study found that noise attenuation of groups of trees, especially in the lower frequencies associated with traffic noise could be significantly improved by positioning the trees in a repeatable pattern similar to a periodic lattice. Specific patterns of vegetation may act at dampening specific ranges of sound energy (Martínez-Sala et al.,
2006).
Bucur (2006) summarizes the extent to which urban trees have the potential to attenuate sound: “In urban areas, trees can be used as noise buffers, able to reduce noise by 5 – 10 dB, if some general recommendations are respected (plant trees near the noise source, plant trees / shrubs with dense foliage as close as possible, plant belt trees of 7 –
17 m wide, etc.)” (Bucur 2006, page 127).
Ecological consequences of anthropogenic noise.
Numerous studies have considered the effects of human sounds on animals. Many animals including some amphibians, insects, mammals, and certainly birds use vocalizations for a variety of purposes. These important functions may include location of a mate, territory announcement and defense, alarm of predation, announcement of food location, and communication between group members. It has been well-documented that human mechanical sounds may disrupt these functions both spatially and temporally
(Barber et al., 2011; Brumm, 2006; Farina et al., 2011; H. Slabbekoorn & den Boer-
Visser, 2006; Hans Slabbekoorn & Ripmeester, 2008).
Krause (2015, pages 12-13) describes sound partitioning by organisms such as cicadas in which they “carve out unique acoustical spaces where they can communicate without
47 their voices being masked by others…..Contours of the landscape itself can help determine how different organisms adjust their vocalizations to accommodate to those acoustic permutations.” He goes on to describe how “each type of organism evolved to vocalize within a specific bandwidth – based on either frequency or time. That, in turn, shed light on the bioacoustic relationships between all of the organisms present in a particular biome. In other words, in order to be heard, vocal organisms must find appropriate temporal or acoustic niches where their utterances are not buried by other signals” (Krause 2015, pages 39-40).
A study in Puerto Rico found that traffic noise did not affect the calls of frogs and toads, but bird species diversity and occurrence, especially with those species with songs and calls in lower frequencies, was diminished in proximity to roadways (Herrera-Montes
& Aide, 2011).
Several studies have considered the change in land use associated with urbanization, particularly with increasing volumes of vehicle traffic, on bird species diversity and vocalization. Bird biodiversity as measured by recording birdsong was found to be inversely correlated with anthrophony, and that both were correlated with human land use along an urban-rural gradient (Joo, Gage, & Kasten, 2011). A European study found far less bird diversity in cities than in rural areas. For those species found in both urban and rural areas, the songs of urban birds were shorter, faster and sung at higher minimum frequencies than for their rural counterparts (H. Slabbekoorn & den Boer-Visser, 2006).
Similar results were found for song sparrows (Melospiza melodia) in Oregon (Wood,
Yezerinac, & Dufty, 2006).
48
Red-winged blackbirds (Agelaius phoeniceus) were found to change the timing and complexity of their calls as anthropogenic background noise from traffic increased. In urban areas red-winged blackbirds sang more frequently at mid-day when traffic sounds diminished and less during morning and evening traffic commute periods as compared to their rural counterparts. More traffic noise also was associated with less-complex birdsong (Cartwright, Taylor, Wilson, & Chow-Fraser, 2014).
Proximity to noisy industrial compressors was found to be associated with a decrease in nesting rates for gray flycatchers (Empidonax wrightii), but an increase in the survival rates for chicks in those nests, presumably because the noise also affected the presence of the main predator of flycatcher nests, the western scrub-jay (Aphelocoma californica)
(Francis, Paritsis, Ortega, & Cruz, 2011).
Goodwin and Shriver (2011) studied eight forest-breeding bird species in areas with little and with much traffic background noise. They found that the traffic noise, which is most prevalent in lower frequencies, masked the birdsong that occupies the same frequency range. Those bird species that sing or call in these lower frequencies were ten times less likely to be found in the areas with greater traffic noise.
In what frequency ranges is there the greatest potential for anthrophony to interfere with bird song? Bucur (2006) reports: “The frequency of vocalization depends on body size…It is also accepted that the larger the wavelength relative to the size of the bird, the lower the intensity of the emitted sound. For everyday life, a practical solution to produce intense sounds and avoid losses due to attenuation and distortion leads to the optimum frequency. For the majority of song birds, for which the body size is between
49 several centimeters and < 0.5 m, the optimum frequency range is between 1 kHz and 6 kHz” (Bucur 2006, page 140).
The effect of anthrophony, biophony and geophony on humans.
The word “noise” generally refers to undesirable sounds, and therefore is a subjective term. Noise is defined as “unwanted or meaningless sound that apart from auditory adverse health effects may distract attention from cues that are important for task performance” (Nassiri et al., 2013, Page 87).
According to Szalma and Hancock (2011, Page 682) “Noise is a pervasive and influential source of stress. Whether through the acute effects of impulse noise or the chronic influence of prolonged exposure, the challenge of noise confronts many who must accomplish vital performance duties in its presence.” They also state “noise increases levels of general alertness/activation and attentional selectivity. It does not influence performance speed, but it reduces performance accuracy and short- term/working memory performance” and “noise has been found to increase the mental workload imposed by a given task environment, thereby reducing the cognitive resources available for allocation to task performance” (Szalma & Hancock, 2011, Page 683).
Wiley (2015, Page 9) describes noise as “sound that has no interest for us yet it makes sounds of interest hard to hear.” Therefore one definition of noise is sound energy that carries no information and interferes with communication between individuals or groups.
Many animals, particularly birds but also some insects, amphibians and mammals, communicate through sound to establish territories, attract mating partners and announce
50 danger. Noise that interferes with this communication is usually associated with humans but may also come from weather such as excessive wind or precipitation. According to
Wiley one appropriate measure of noise is the amount of mistakes made by a receiver due to the corruption of the information within an auditory communication caused by noise
(such as a pair of potentially mating birds being unable to locate one another due to excessive vehicle traffic, or you bringing home the wrong brand of laundry detergent due to noise that interferes with the phone call from your spouse).
Numerous studies have documented the negative effects of industrial noise on the performance and health of workers. In an effort to determine what types of noise were most damaging, Nassiri et al., (2013) considered several variables of mechanical sound including sound pressure level (“loudness”), noise schedule (the timing of sounds) and noise type (treble vs. bass). Their study found that intermittent noise (as opposed to continuous) at high pressure levels (loud) resulted in the greatest decrease in human performance.
Schapkin, Falkenstein, Marks, and Griefahn (2006) subjected participants to a quiet night of sleep and also a night of sleep accompanied by traffic noise and found that performance on several tests given the following day to measure cognitive performance and inhibitory brain activity differed in the two groups. While the nocturnal traffic noise affected task performance and inhibitory brain activity to varying degrees, it was reported that overall the psychological costs for inhibitory functioning associated with sleep disturbance from noise were measurable, even when no change in task performance was observed.
51
A study conducted near the Aukland International Airport in New Zealand found that residents experienced noise-related sleep disturbance and had associated health problems and reported diminished quality of life (Shepherd, Welch, Dirks, & Mathews, 2010).
The severity of negative effects of traffic noise may be dependent on human population demographics including gender. Belojevic, Evans, Paunovic, and Jakovljevic
(2012) found that exposure to traffic noise was associated with decreased measures of executive functioning in male school children but not significantly in females. The researchers believe that in that age group (7 to 11 years) boys may be more susceptible to chronic stressors including noise than are girls. Cohen, Glass, and Singer (1973) report that children living on the lower levels of apartment buildings and therefore exposed to greater levels of outside traffic noise showed greater impairment of ability to discriminate between sounds and impaired reading achievement compared to children with lower exposure to traffic sounds.
Alvarsson et al. (2010) recorded physiological measures of stress in humans that were exposed to either noisy environments (human mechanical sounds) or natural sounds and found that recovery from stress tests tended to be faster in the population exposed to natural sounds compared to those exposed to mechanical sounds.
Most previous work on mitigating the negative effects of urban noise has focused on reducing offensive, unwanted sounds – usually those associated with traffic. Urban vegetation, including trees, has been shown to reduce noise levels (Fang & Ling, 2003) and therefore reduce the level of human stress. But it seems that simply reducing noise levels is not the full story. Increasingly the focus of research is changing to also include preserving or enhancing favorable sounds. Quality of urban life seems to be affected by
52 the setting in which a person finds quiet. Krause (2015, page 113) notes: “Although there is considerable evidence of negative impact from incoherent and loud acoustic signals, the emotional consequences of natural soundscapes on urban-living humans may have a large positive effect on our sense of well-being.”
One study questioned 500 residents of apartment buildings next to busy roads in
Sweden. The premise of the study is that some of the apartments faced the noisy roads but also had a face to a quiet side away from the road. Other apartments faced busy streets on two sides and had no quiet side. The apartments were also judged for their distance from local parks with quiet places. Residents were asked about the noise conditions and psychosocial conditions (tired, irritated and angry, stressed). The accessibility to local greenspaces had a greater influence on resident’s perception of noise and psychosocial conditions than did the presence or absence of a quiet side of their apartment. The conclusions of the study are: “availability to nearby green areas can moderate or buffer the effects of chronic-noise exposure on health and well-being.… significantly less residents with ‘better’ availability to green areas exhibited stress-related psychosocial symptoms than residents with ‘poorer’ availability to green area” (Gidlöf-
Gunnarsson & Öhrström, 2007, Page 123).
Some sounds have been found to have detrimental effects on humans while other sounds have been found to be beneficial. Not all people consistently perceive all sounds to be favorable or unfavorable. The effects of various sounds on humans can be measured qualitatively, by gathering people’s opinions on various sounds for example, or quantitatively, by measuring people’s physiological or emotional responses to various sounds. Dzhambov and Dimitrova (2015) report that the single most important variable
53 in determining the degree of noise annoyance for residents of Plovdiv, Bulgaria was the distance between a person’s residence and the closest green space. It seems that those residents who could easily reach the relative tranquility of a nearby greenspace reported less overall noise annoyance with their occupation of the city.
In general, tranquil (quieter) urban spaces are believed to provide more restorative experiences for humans, and therefore contribute to health and quality of life. The acoustical criteria for perceived tranquility obviously include relatively low energy
(quieter) sounds, but also the absence of not-fitting sounds. Other attributes of favorable soundscapes include contributing to a sense of place, enhancing interactions with landscape elements, and revelation of wildlife (Dumyahn & Pijanowski, 2011).
The degree to which humans find sounds to be pleasing or objectionable seems to be related to the context of the setting in which the sounds are heard. Variables that can affect listeners’ perception of an urban soundscape include the activity of the listener, daily, weekly and seasonal variations in sounds, the type and intended use of the space, and architectural, cultural and historic characteristics of the space (Cain et al., 2008).
Irwin, Hall, Peters, and Plack (2011) controlled recorded sound level (“loudness”) but varied recordings by setting and content. Participants in the study listened to recordings of different urban sounds while researchers monitored brain activity and heart rate.
Urban sounds reported as being pleasant including the dawn chorus of birds and a string quartet playing music were found to elicit responses in different areas of the brain and were associated with lower heart rates than were most mechanical sounds, which were reported as less pleasant. The conclusion of the study is that “urban soundscapes with
54 similar loudness can have dramatically different effects on the brain’s response to the environment” (Irwin, Hall, Peters, and Plack, 2011, Page 258).
Anderson, Mulligan, Goodman, and Regen (1983) report that listeners found recorded sounds of nature to enhance scenes of wooded areas more than scenes of urban areas where the nature sounds may have seemed out of place. Similarly traffic sounds were reported to be more undesirable when shown with wooded scenes than when shown with urban scenes.
One study considered the effect of several levels of urban noise, including heavy construction noise, on willingness of people to help others. In areas of varying degrees of mechanical sounds researchers provided opportunities for passers-by to offer altruistic actions ranging from responding to requests for directions, to picking up dropped keys, to attempting to contact the owner of a found medical card. Quieter situations resulted in greater willingness for passers-by to communicate and to engage in helpful actions than did noisier situations. This may be because noise creates more stress in individuals and/or effectively reduces the sphere of attentiveness in individuals. Researchers report that “noise appears to be the most important component of overload, affecting both the subjects’ attentiveness in implicit helping demands, as well as the refusal to engage in verbal interaction” (Moser, 1988, Page 287).
Several studies have considered opinions of visitors to national parks relative to experienced sounds. In general visitors prefer biophony and geophony (sounds of nature) over anthrophony (sounds from humans). Sounds associated with air and ground traffic are reported to be particularly irritating (Benfield, Bell, Troup, & Soderstrom, 2010).
55
In a study conducted in Italian urban parks visitors reported their park experience to be good or excellent even when sound pressure level (perceived as “loudness”) was nearly twice as high as guidelines set for quiet areas. Researchers believe this is true because of the presence of trees, natural features and the relative tranquility of the parks compared to the surrounding city (Brambilla, Gallo, Asdrubali, & Alessandro, 2013). A study conducted in parks in the United Kingdom questioned park visitors about preferences of sound and found that most people preferred natural sounds over mechanical sounds. Sound level, both objective and perceived, also was important to reported satisfaction – quieter parks were generally preferred over louder parks (Irvine et al., 2009). A similar study in Hong Kong found that visitors to parks preferred sounds associated with nature over sounds associated with heavy vehicles or motor bikes (Tse et al., 2012).
There are many components of a soundscape that can be categorized as anthrophony, biophony, and geophony. Some researchers have investigated which elements within each of these categories may be most detrimental and lead to fatigue, and which may help to reverse these effects. As previously discussed, many studies have considered the detrimental effects of the mechanical noise element of anthrophony. Several studies are now considering the beneficial elements of biophony. Birdsong is one aspect of a soundscape that seems to be perceived by many people as pleasing. Ratcliffe,
Gatersleben, and Sowden (2013) report that study participants who listened to a range of soundscape recordings rated birdsong as the component most commonly associated with perceived stress recovery and attention restoration, although no measures for stress or attention fatigue were employed. The researchers recommend that “future studies should
56 quantitatively examine the potential of a variety of bird sounds to aid attention restoration and stress recovery, and how these might be predicted by acoustic, aesthetic, and associative properties, in order to better understand how and why sounds such as birdsong might provide restorative benefits” (Ratcliffe, Gatersleben, and Sowden, 2013,
Page 221).
The seminal book Silent Spring by Rachel Carson (1962) alerted us to the detrimental effects of over-use of insecticides. While the impetus of the book focused attention on the threat to human health, it also proved to fuel a rapidly growing environmental awareness movement. The title to the book is in reference to the diminished amount of birdsong, which was a result of the decline of many bird species due to DDT poisoning of insects that were then consumed by songbirds (Carson, 1962). Loss of species affects ecosystems in many ways, and the loss of aesthetically, emotionally and perhaps spiritually valued characteristics of some species, including song birds, may affect humans. The renowned ecologist Stephen Kellert writes:
A world without warblers would be mute and barren, lacking the richness of
sound, color, the promise of hope, rebirth, and transcendence. Their
exuberant passage reaffirms connections with the miracle of tenuous life.
Their diminution contracts our tiny world of organized and purposeful matter
and spirit; without them, the edge of a more universal deadness and
dissolution advances (Kellert, 2002, Page 53).
57
Perception of visual and auditory stimuli.
Human perception of a soundscape surprisingly is not simply a function of the energy received and interpreted by the auditory system. It seems that what we see also affects how we interpret sounds. Non-acoustical criteria including the presence of natural elements such as trees within sight can affect the perception of sound (De Coensel, Boes,
Oldoni, & Botteldooren, 2013).
Payne (2013) introduced the Perceived Restorativeness Sound Scale (PRSS) that is used to “assess park visitor’s perceptions of a soundscape’s potential to provide psychological restoration” (Payne, 2013, Page 255). Using this method in several locations it was determined that rural soundscapes were perceived as highest in restorative potential followed by urban parks, and non-park urban soundscapes were determined to have the least potential for attention restoration.
A study conducted in Belgium gathered opinions from residents on their level of noise annoyance as experienced from within their home. Interestingly the researchers found that reported noise annoyance was not simply a function of the sound that originated around the homes but also a function of what the residents could see from their living room window. Eight percent of people who could see vegetation from their window reported some level of noise annoyance while 34% of people without a view of vegetation reported some level of noise annoyance. All of the homes faced a roadway and had similar levels of measured sound levels outside the home (Van Renterghem &
Botteldooren, 2016).
58
These studies suggest that sound and vision are not perceived independently in humans. Our perception of sound is influenced by what we see and is further influenced by whether the information arriving through our eyes and ears is consistent and contextually matched. Therefore the role of vegetation in attenuating unwanted sounds is not limited to the ability of foliage to dampen sound energy, but it seems that the visual masking of the sound source affects our perception of the sound as well (Bucur, 2006).
Summary
It is well documented that the urban environment affects human physical, mental and emotional health. The mechanisms include increasing (or decreasing) stress, which is primarily a physiological function. Our ability to focus and concentrate (directed attention) is also affected, which is primarily a psychological function. Urban conditions void of the sights and sounds associated with nature can contribute to mental fatigue and access to nature can help restore human attention. The amount of vegetated greenspace present in a neighborhood may even affect the ability of residents to build meaningful social networks. Neighborhoods with publicly accessible greenspace benefit by providing inviting spaces where people can gather and communicate. Most research in the area of access to nature in cities has focused on the visual aspect of “greenness,” but greenness can also be represented by sound. The discipline of soundscape ecology is revealing that the composition of sounds in a landscape can be a reliable indicator of biodiversity as well as a source of damage or wellness to human health. As Krause
(2015, page 151) notes: “Natural sounds that define the field of soundscape ecology are
59 the voices we need to heed closely. For they are balanced somewhere between creation and destruction – and we silence them at our own peril.”
How do all of these diverse topics of urban forests, human health, and soundscape ecology connect into the research of this dissertation? Personal experience and an extensive review of literature suggests that many (perhaps most but not all) people find improvement to physical, mental and/or emotional health when they spend time outdoors, preferably in a setting that includes elements of the natural world. The beneficial components of the optimal outdoor experience obviously differs greatly for individuals – some will prefer a walk on a tropical beach while a few would rather climb a mountain and many may simply prefer occasionally planting flowers around their house. There is no universal recipe for a beneficial landscape. Perhaps there are similar components, however. My interest led me to investigate the role that sounds play in the mix of common landscape elements that allow humans to find some level of “peace and quiet” in their preferred outdoor experience.
60
CHAPTER III
COMMENTARY ON THE LITERATURE: HOW DID I GET HERE?
The previous chapter reviewed an extensive selection of papers published in peer reviewed journals as well as books from respected authors. The majority of these are reports on research studies that attempt to understand the world and human’s place in it in very thin slices of inquiry that include human directed attention, the benefits for humans of access to nature, and soundscape ecology. But how did each of these researchers find their respective topics and hypotheses? What foundation of inquiry supports their research? And on a personal level, how did I get to this interest in trees, sounds and people? Can a look back beyond the current science help to organize the literature and provide context to this study?
All inquiry is subjective. Many, but not all of the publications explored in the previous literature review describe experiments and surveys that follow proper scientific protocols designed to bring reliable objective repeatable methods to the quest for understanding our world. But in the end, all inquiry is subjective, because all inquiry begins with wonder.
61
Ideas are sparked in individual imaginations that lead to moments of “I wonder…”
Occasionally this leads to a sharing of ideas among colleagues and possibly a search of the literature to find what is already known, and eventually if the idea gains merit and the person embodying the imagination has the good fortune to have access to resources, a study is built to formally explore a hypothesis using sound scientific methods. But in the end, or at the beginning, it is the subjective and personal moment of “I wonder…” that fuels discovery.
This chapter will be somewhat unconventional in the format of doctoral dissertations and will temporarily stray from the preferred third-person presentation of foundational literature and gathered data and will attempt to explain how it is that I came to these questions regarding trees, sounds and people. In doing so I will build a bridge between the bodies of knowledge related to the topic of this study and show how the generous knowledge of those who preceded me has shaped the hypotheses of this study and the methods I have chosen to employ in the study.
I believe my literature review unknowingly began around 11 or 12 years of age. I was an undistinguished student known mostly to my teachers as a daydreamer who was easily distracted by events outside of the classroom window. There once was a day that I had ridden my bicycle to the corner drug store with the intent of buying the latest copy of The
Hockey News, about the only text that held my interest at that time in life. While walking to the front register with The Hockey News in hand I passed a display of paperbacks and saw a small blue book with the picture of a bird on it. To this day I do not understand why I stopped to look at the book. The title was The Immense Journey and the subtitle read “An imaginative naturalist explores the mysteries of man and nature.” The author
62 was Loren Eiseley (1946). I didn’t have enough money for both the book and the Hockey
News, and for some reason I chose the book, and invested $1.95 in something important.
Eiseley (1907 – 1977) was the University Professor of Anthropology and History of
Science at the University of Pennsylvania. In The Immense Journey, Eiseley alternates between chapters that are discussions of topics related to anthropology and stories of his life and career. In one such story, Eiseley describes a journey into a remote area of the
American west on a quest for dinosaur bones. He was also tasked with capturing some live birds from the region to be sent back to a zoo. Following his orders, he entered a deserted cabin at night where he suspected birds were roosting and was successful at nabbing a male sparrow hawk (kestrel) that elected to attack the intruder’s grasping hand while the female bird escaped through a hole in the broken roof. Eiseley was pleased with the capture if not with the bloodied hand. The next day, Eiseley was preparing the cage in which the bird would be shipped back to the zoo when he describes an unexpected action that he impulsively took:
In the morning with the change that comes on suddenly in that high
country, the mist that had hovered below us in the valley was gone. The sky
was a deep blue, and one could see for miles over the high outcroppings of
stone. I was up early and brought the box in which the little hawk was
imprisoned out onto the grass where I was building a cage. A wind as cool as
a mountain spring ran over the grass and stirred my hair. It was a fine day to
be alive. I looked up and all around at the hole in the cabin roof out of which
the other little hawk had fled. There was no sign of her anywhere that I could
see.
63
“Probably in the next county by now,” I thought cynically, but before beginning work I decided I’d have a look at my last night’s capture.
Secretively, I looked all around the camp and up and down and opened the box. I got him right out in my hand with his wings folded properly and I was careful not to startle him. He lay limp in my grasp and I could feel his heart pound under the feathers but he only looked beyond me and up.
I saw him look that last look away beyond me into a sky so full of light that
I could not follow his gaze. The little breeze flowed over me again, and nearby a mountain aspen shook all of its tiny leaves. I suppose I must have had an idea then of what I was going to do, but I never let it come up into my consciousness. I just reached over and laid the hawk in the grass.
He lay there a long minute without hope, unmoving, his eyes still fixed on that blue vault above him. It must have been that he was already so far away in heart that he never felt the release from my hand. He never even stood. He just lay with his breast against the grass.
In the next second after that long minute he was gone. Like a flicker of light, he had vanished with my eyes full on him, but without actually seeing even a premonitory wing beat. He was gone straight into that towering emptiness of light and crystal that my eyes could scarcely bear to penetrate.
For another long moment there was silence. I could not see him. The light was too intense. Then from far up somewhere a cry came ringing down.
I was young then and had seen little of the world, but when I heard that cry my heart turned over. It was not the cry of the hawk I had captured, for, by
64
shifting my position against the sun, I was now seeing further up. Straight out
of the sun’s eye, where she must have been soaring restlessly above us for
untold hours, hurled his mate. And from far up, ringing from peak to peak of
the summits over us, came a cry of such unutterable and ecstatic joy that it
sounds down across the years and tingles among the cups on my quiet
breakfast table.
I saw them both now. He was rising fast to meet her. They met in a great
soaring gyre that turned to a whirling circle and a dance of wings. Once more,
just once, their two voices joined in a harsh wild medley of question and
response, struck and echoed against the pinnacles of the valley. Then they
were gone forever somewhere into those upper regions beyond the eyes of
men. (Eiseley, 1946, Pages 190 – 192)
That story, among others in Eiseley’s book, bridged a gap that I am still trying to understand. The story seems to be less about two birds and much more about a man (a distinguished scientist at that) and an experience in nature that deeply moved him in a positive way – and in recounting the story moved my young imagination as well.
My father worked in an auto factory. He didn’t seem to enjoy life very much but one source of pleasure was feeding whatever wildlife came into our backyard. Our small house on the edge of Detroit shared a property line with an undeveloped woodlot approximately 15 acres in size. My mom would make peanut butter and jelly sandwiches and dad would take them into the backyard and feed the local raccoons. Dad always kept the birdfeeder full and mom called the birds by name. They didn’t experience nature to the same magnitude as Dr. Eiseley, and neither did I, but watching my parents seek some
65
“peace and quiet” in our backyard made me believe that something good happens when people interact with nature. For my parents the woods offered peace (a lack of turmoil) and quiet (a lack of noise) that certainly brought quality to their lives. I took that lesson to heart and climbed over the back fence and explored the woods early and often.
I suspect that many people who are inspired to formally investigate a hypothesis begin their quest through attentive observation of their own surroundings and reading the observations of other like-minded explorers. Many of the foundation stones of this doctoral dissertation research come from the observations of people immersed in (not removed from) the natural environment. John Muir effectively championed the preservation of wilderness lands by writing of his experiences in these lands. In his 1894 book The Mountains of California, Muir includes a chapter titled “A wind storm in the forests” in which he writes: “There is always something deeply exciting, not only in the sounds of winds in the woods, which exert more or less influence over every mind, but in their waterlike flow as manifested by the movements of the trees” (Muir, 2014, Page 7).
Having myself worked in forestry and experienced prolonged periods of solitude in northern forests, Muir’s description of sounds of winds in the woods influencing minds certainly rang true for me. The observations and subsequent writings of Muir undoubtedly influenced many scientists to embark on studies related to wilderness preservation, landscape ecology, and the benefits to humans of access to nature. In turn,
Muir and many others were influenced by observers of nature who preceded them. Muir was an avid reader of Ralph Waldo Emerson and Henry David Thoreau. In the opening paragraph to his 1862 essay “Walking,” Thoreau writes of the human place within nature and not outside of it: “I wish to speak a word for Nature, for absolute freedom and
66 wilderness, as contrasted with a freedom and culture merely civil, - to regard man as an inhabitant, or a part and parcel of Nature, rather than a member of society” (Emerson,
Thoreau, & Elder, 1991, Page 71). In an introduction to this essay, John Elder writes:
“Thoreau understood that wilderness is not dependent upon a vast, unsettled tract of land.
Rather, it is a quality of awareness, and openness to the light, to the seasons, sharpened by the dialogue between settlements and the untracked lands around them, between human language and the rich communication of the senses” (Emerson et al., 1991, Page xvi). This implies that the benefits of access to nature can be found not only in the wilderness but also in human settlements, if only the seeking person is possessed with a
“quality of awareness.” This strikes a similar note with the research conducted by Rachel and Stephen Kaplan that found that access to nature, even in urban spaces, can give pleasure, that visits to natural landscapes or even encounters with individual trees are satisfying to experience, and that natural settings support human functioning. Outdoor experiences provide a context in which humans can manage information effectively, and access to natural environments fosters the recovery from mental fatigue. These natural settings permit tired individuals to regain effective functioning and they refresh mental fatigue and replenish directed attention (R. Kaplan & S. Kaplan, 1995). So it is that the subjective observations of an attentive person with a quality of awareness may inspire others to conduct objective, science based research that withstands the scrutiny of peer review.
One of the most influential Americans that would promote the access of nature within cities is Frederick Law Olmsted. The early design of settlements and towns in the United
States and Canada condensed commercial, industrial, and residential areas leaving little
67 open greenspace beyond the town square. Recreation in nature was viewed as an activity primarily for the upper class and required a trip into the surrounding countryside. In the mid-1800’s Frederick Law Olmsted had a much different vision as he designed parks, most notably New York’s Central Park, to be fit into the interior of cities. He viewed urban greenspaces to be essential spaces for retreat for all classes and categories of people, but perhaps mostly for the working people who had little opportunity to escape the congestion and pollution of cities.
In 1870 Olmsted gave a speech to the American Social Science Association in Boston, and the following year published the text in the Association’s Journal of Social Science.
In the paper Olmsted describes not only the recreational value of parks, but indeed of the benefits to the mental and emotional health of urban-dwellers:
For this purpose neither of the forms of ground we heretofore considered are
at all suitable. We want a ground to which people may easily go after their
day’s work is done, and where they may stroll for an hour, seeing, hearing,
and feeling nothing of the bustle and jar of the streets, where they shall, in
effect, find the city put far away from them. We want the greatest possible
contrast with the streets and the shops and the rooms of the town which will
be consistent with convenience and the preservation of good order and
neatness. We want, especially, the greatest possible contrast with the
restraining and confining conditions of the town, those conditions which
compel us to walk circumspectly, watchfully, jealously, which compel us to
look closely upon others without sympathy. Practically, what we most want is
a simply, broad, open space of clean greensward, with sufficient play of
68
surface and a sufficient number of trees about it to supply a variety of light
and shade. This we want as a central feature. We want depth of wood enough
about it not only for comfort in hot weather, but to completely shut out the
city from our landscapes. These are the distinguishing elements of what is
properly called a park (Twombly, 2010, pages 230 – 231).
It seems that Olmsted envisioned, and perhaps inspired the research that would take place a century later that investigates the value of access to nature for the human propensity for compassion when he noted the urban-dweller’s unfortunate environment “which compels us to look closely upon others without sympathy.”
While Frederick Law Olmsted is most frequently recognized as the preeminent
American landscape architect that designed and built grand urban parks, even Olmsted recognized that greenspaces need not be large or defined by recognizable boundaries. He envisioned extending the greenness of parks and the accompanying benefits beyond the urban park boundaries and into the streets themselves. In 1871 he wrote “Would trees for seclusion and shade and beauty, be out of place, for instance, by the side of certain of our streets? It will, perhaps, appear to you that it is hardly necessary to ask such a question, as throughout the United States trees are commonly planted at the sides of streets”
(Twombly, 2010, page 221). Clearly Olmsted understood that the benefits of access to nature need not be confined within park boundaries but should be incorporated into the urban environment in which humans are included, and not apart from.
Further speaking of the benefits of streets lined with trees, Olmsted wrote:
The change both of scene and of air which would be obtained by people
engaged for the most part in the necessarily confined interior commercial
69
parts of the town, on passing into a street of this character after the trees had
become stately and graceful, would be worth a good deal. If such streets were
still made broader in some parts, with spacious malls, the advantage would be
increased. If each of them were given the proper capacity, and laid out with
laterals and connections in suitable directions to serve as a convenient trunk-
line of communication between two large districts of the town or the business
center and the suburbs, a very great number of people might thus be placed
every day under influences counteracting those with which we desire to
contend (Twombly, 2010; pages 222 – 223).
In an 1881 paper Olmsted lists some of the maladies with which urban-dwellers contend that may be aided by access to greenspace, including “vivial exhaustion,”
“nervous irritation,” “constitutional depression,” “excessive materialism,” “loss of faith and lowness of spirit” (Twombly 2010; page 284).
It seems that considering scale and accessibility is important in designing greenspaces, parks and neighborhoods that encourage frequent use and facilitate communication among neighbors. Parks are successful when they draw the community in and neighborhoods discover a common space that reflects their bond.
If urban landscapes that include elements of nature encourage use of common spaces, can soundscapes function in a similar fashion? Can urban soundscapes function as sonic parks or refuges from excessive noise?
The problem with sound (and conversely the joy of sound) is that it doesn’t stay in place very well. If you don’t like what your neighbor’s yard looks like you can erect a fence or plant a row of shrubs along the property line to provide a visual screen. If you
70 don’t like the noise coming from your neighbor’s yard the fence won’t help much. For neighborhoods to function without constant friction among residents, recognition of this wandering property of sound must be respected. Loud late-night parties or early Sunday morning lawn mowing are discouraged. A “Disturbing the Peace” violation frequently means acting in a noisy manner. So at the neighborhood scale we have perhaps unknowingly accepted the concept of sonic open space or auditory common ground.
In his book The Soundscape: Our Sonic Environment and the Tuning of the World, R.
Murray Schafer (1994) introduces the term of keynote sounds. Keynote sounds may be considered the background sound signatures in a soundscape. They do not have to be listened to consciously but they convey information none-the-less. Schafer (page 9) states “the fact that they are ubiquitously there suggests the possibility of a deep and pervasive influence on our behavior and moods. The keynote sounds of a place are important because they help outline the character of men living among them.” This suggests that communities announce their identity not only by their pattern of land use or their population demographics, but also by the properties of their soundscape. Not surprisingly, residents of communities do not always agree on the sounds that are purposely introduced to a soundscape. Some (but not all) people find comfort in the long-held tradition of ringing of church bells on Sunday mornings but object to the call to prayer of local mosques. Perhaps more importantly are the unwanted sonic side effects of transportation and industry. Communities near airports certainly are affected by the sounds associated with the airport. Communities adjacent to high volume travel corridors or train tracks are certainly affected by the sounds associated with cars, trucks, and trains.
Numerous studies have documented the harmful effects of living or working in “noisy”
71 environments. Are these sounds examples of harmful “noise” or rather non-threatening signatures of the places where we live?
Schafer’s typology of sound is from the listener’s perspective. Schafer hears the landscape relative to what humans interpret in their surroundings. Others, including
Bernie Kraus, interpret landscapes from the perspective of the source (and not as much from the listener). By coming to sounds from this perspective one can learn more about the composition of the organisms (including humans) that occupy the landscape, and monitor changes in the landscape by monitoring and grouping sounds according to their source. Others including R. Haven Wiley (2015), understand sounds as tools for communication between organisms. Any sounds that interfere with communication diminish the information that is carried in the sound and decrease the degree of understanding and effective response in the receiver. Noise therefore can be defined as any sound energy that interferes with audio communication. Noise, that is, unwanted sound is relative to the source and the receiver, and one organism’s noise may be another organism’s music. Seth Horowitz (2012) comes at sound not as much as a communication tool between source and receiver, but rather focuses on how sound shapes the perception of the world for the receiving human. Certainly communication is part of this, but sound throughout a landscape, whether actively communicative or passively informative, will shape a person’s understanding of that landscape, which brings us back around to Schafer’s concept of keynote sounds.
Can we design or at least partially control soundscapes on a neighborhood or community scale? Should we as community residents or urban planners be aware of, or actively working to improve shared sonic spaces? If there are benefits to establishing and
72 maintaining properly-scaled neighborhood parks, might there also be benefits to considering the quality of soundscape greenspaces? How might those benefits be measured?
Since sound is not confined to property boundaries, might some of the benefits of greenspaces be realized beyond the boundaries of the park by way of the sounds that emanate from the space? Can beneficial sounds be enjoyed like the pleasant smells from the neighborhood bakery for those fortunate to live downwind? These are all “I wonder…” type questions that may serve as foundations to more formal modes of inquiry, such as this doctoral dissertation.
Returning to our starting point of this discussion raises yet another question: If all inquiry begins with wonder, is it possible to design communities that foster wonder and reduce wonder-robbing fatigue and distractions? Recall the definition of directed attention posited by Bratman, Hamilton & Daly (2012, page 122): “The effortful, conscious process of bringing cognitive resources to bear in order to focus on selected stimuli, while avoiding distraction from unrelated perceptual inputs.” It is clear that directed attention is both the effort put forth at paying attention to a specific task and the effort put forth at avoiding distraction. Therefore, urban spaces that unnecessarily cause alarm and distraction wear on directed attention and cause fatigue. Conversely, urban spaces that foster involuntary attention, also called fascination by the Kaplans (and wonder by this researcher) serve in replenishing fatigued directed attention and allow a person to act more reasonably.
Stephen Kaplan’s Reasonable Person Model, as well as many other attempts at explaining our place in nature, have their foundations built on the subjective inquiries and
73 observations of insightful wonderers like Muir, Thoreau, Eiseley, Olmsted and Krause.
By studying the bridge to the past insights of these and other passionate thinkers provides context to the body of science-based knowledge from which this dissertation research launches.
All of these authors, scientists and wonderers have unique views of our human place in nature. What did Loren Eiseley’s story of the hawk mean for me? Perhaps he initially saw the abandoned cabin as a symbol for nature – that the animals were in nature and that humans perhaps once had occupied the cabin but were now outside of nature. Eiseley went into the cabin (a former home for his predecessors) as an intruder, in the dark and not understanding what he would find. He captured one bird for the purpose of putting it into a cage for his own purposes. The next morning, in the light he saw things differently
– he was enlightened. He instinctually (not through purposeful thought that he cared to admit) saw the bird not as an object to dominate but rather as another organism that shares in nature, and he came to marvel in the tenacity with which the hawk’s mate persisted despite his actions. With each of many readings of this book I have come to appreciate our place within the natural world and come to better understand the costs of separation from nature. This is my personal bridge from “I wonder…” to a scientific (yet subjective) inquiry into trees, sounds and people.
Wonder is valuable for the sake of wonder itself, but it is of greater value if it leads to improvement, whether that is in sustainable use of natural resources, cure for disease, or improvement of living conditions in cities. Therefore, the purpose for this research is not simply to explore how sounds are related to trees, but rather how we might use the
74 components of urban soundscapes as informants of the health of a landscape and how that supports the health and quality of life for humans.
75
CHAPTER IV
RESEARCH DESIGN
Research Questions
As discussed in Chapter 3 - Commentary on the Literature, I was influenced to a great degree by writers who described their experiences and observations in nature. Henry
David Thoreau, Ralph Waldo Emerson, John Muir, Frederick Law Olmsted, Thomas
Merton, Loren Eiseley, Rachel Carson, Aldo Leopold, Barry Lopez, and others described what they saw, heard and felt while experiencing landscapes, and all of them without exception commented on the damage to natural systems that humans were causing due to pollution and unsustainable consumption.
These observers of nature led me to publications reporting on scientific studies related to the human place within natural ecosystems. From these writings I learned the important issues that had been investigated and the many questions that remain. One of the questions that these sources of inquiry has raised for me is “Is there an optimal recipe of environmental conditions that promotes human well-being with which to design a city?” Of course the answer is “No” given the wide range of preferences and values of
76 the people who choose to live in cities. Choice implies a preference for urban conditions, however many people do not choose to live in cities but rather are forced to live there because of access to employment or services (including health care). So it cannot be assumed in all cases that urban environmental conditions attract the people who prefer those conditions.
It may be easier to answer an opposite-pointing question, such as “Is there a recipe of environmental conditions that degrades human well-being which are characteristically found in cities?” Certainly poor air quality, poor water quality, overcrowding, and uncontrollable sensory stimuli including excessive noise would be elements of a city that tends to degrade human well-being.
There are seemingly endless questions that may be posed about favorable and unfavorable conditions in cities. I have chosen in this dissertation research to focus on trees, sounds and people, and specifically how changes in urban forests affect the sounds we hear, and in turn how the sounds we hear affect our perception of the environment around us. As previously stated, the specific questions that caused me to design this dissertation research in the manner that I have chosen are:
4. Will the percentage of tree canopy cover change as the Emerald Ash Borer
population increases in northeast Illinois?
5. Will the quantity and quality of sounds change in areas that have significant
changes to tree canopy cover?
6. Will the changes in the quantity and quality of sounds affect human directed
attention?
77
Each of these questions investigates separate but related issues. The hypotheses that emerge from these questions are stated as:
1. The percentage of tree canopy cover will change in northeast Illinois.
2. The soundscape will change as tree canopy cover changes
a. The quantity of sounds will change as tree canopy cover changes
b. The quality of sounds will change as tree canopy cover changes
3. Human directed attention will change as the quantity and/or quality of sounds change.
The null hypotheses that was tested in the dissertation research was then stated as:
1. The percentage of tree canopy cover does not change in northeast Illinois.
2. The soundscapes do not change as tree canopy cover changes
a. The quantity of sounds do not change as tree canopy cover changes
b. The quality of sounds do not change as tree canopy cover changes
3. Human directed attention does not change as the quantity and/or quality of sounds
change.
To test these hypotheses, soundscape measurements were recorded in treatment (tree loss) and control (no tree loss) urban areas both before and after trees are removed. Since very few people occupy houses specifically for their proximity to ash trees, the treatment can be considered random.
Evaluation of soundscape attributes and their importance to humans.
When listening to a soundscape, can the individual sounds be evaluated relative to their salience or importance to the human listener? Several variables of individual sounds within a soundscape can be considered:
78
1. Loudness – The intensity of a sound (“loudness” as perceived by the human ear)
is referred to as a sound’s power spectral density, and can be estimated by
measuring the energy or power that the sound carries. This energy is usually
expressed using the unit of decibel. The decibel scale is logarithmic.
2. Time – The duration of a sound may influence the importance as perceived by the
human ear. This is a complex relationship however, because a very short sound,
as in the crack of a whip or bark of a dog may be perceived with a much higher
sense of urgency or sense of being startled than perhaps a long persistent sound.
In fact a long uniform sound may be filtered out by the brain as being not
important and therefore ignorable. Because of this, the duration of a sound is
unlikely to be a predictable indicator of perceived importance by a listener – a
sound of long duration is not always more important than a sound of short
duration, and vice versa.
3. Frequency – The pitch or frequency of a sound, or perhaps more importantly the
varying pitch of a sound may be perceived as salient by a listener. A sound that
widely fluctuates in pitch, such as a siren may be more noticeable than a sound
that is has little variation in pitch. Insect calls such as crickets and some cicadas
are usually more uniform in pitch, sometimes very monotonic, as compared to the
song of a bird. The frequency range as well as the complexity of pitch variation
seem to influence the level of interest to many human listeners.
4. Complexity and entropy – Some sounds are not very complex. The monotonic
sound of a single bell that vibrates within a very narrow frequency range is not
complex, but it is highly ordered and focused along that narrow frequency. The
79
sound of leaves rustling in the wind or water rushing over a waterfall also seems
to be low in complexity, but unlike the bell, the sound occupies a wide frequency
range, and there is not much of a pattern within the sound. As will be discussed
in Chapter 7 – Results Part 1 in greater detail, entropy is a measure of disorder.
While both the bell and the wind or waterfall sounds are similar in that they do
not seem complex, they are opposite in terms of entropy. The wind and waterfall
sounds are highly disordered (high entropy) and the bell sound is highly ordered
(low entropy).
5. Sound source – The source of a sound as interpreted by the human listener may be
the single most important factor to the salience of that sound. Sounds may be
very loud (such as a car horn) or almost imperceptibly quiet (such as the hiss of
air leaking from a tire) and either may be interpreted with great urgency if the
sound source is believed to be dangerous. Sounds with similar high energy or
loudness (such as loud music) or softness (such as wind rustling leaves) may be
interpreted with less urgency or even comfort. In general, and as previously
discussed in this Literature Review section, studies have shown that sounds
originating from mechanical sources (such as from tools or vehicles) are generally
perceived less favorably than sounds originating from nature (such as birds or
water). However examples to the contrary are plentiful.
Because there are only trends, and few if any straight line relationships between sound variables (loudness, duration, frequency, entropy, and source) and human perception, it is difficult to quantify these variables is such a way as to predict the level of irritability or comfort of a soundscape for humans. We can however break the elements of a
80 soundscape into components and in doing so determine if one soundscape differs from another, and if so examine what aspects of sound affect human perception. Computer- based sound analysis programs accomplish the task of measuring these individual sound attributes (Charif, Waack, & Strickman, 2010).
Research Process and Conceptual Framework
The general research design throughout this dissertation study follows three distinct but related areas of study: urban forestry, soundscape ecology, and human cognitive function. Within each of these areas of study the specific research questions outlined above lead to a series of research tasks designed to gather appropriate data and implement appropriate analysis methods with which to arrive at answers to the research questions.
Therefore the research design included data collection focused on changes to an urban forest. The analysis of those data led to additional data collection focused on changes to sound attributes and sound types during the period of changing tree canopy. Based on the results of the analysis of sound data, which confirmed significant changes in elements of the soundscape, additional data were gathered concerning the potential effects that various soundscape treatments have on humans as they participated in tests for directed attention.
What professional and scholarly audiences find application to this line of inquiry about trees, sounds and people? Numerous practitioners in urban forest management, urban planning, public health and related fields will find value in this research design and the subsequent findings. One frequent problem is the lack of communication between
81 practitioners; urban foresters frequently don’t understand the challenges and objectives of urban planners and public works directors (and vice versa), and public health and safety officials don’t understand how their challenges and objectives can be met through the assistance of natural resource managers. This function of cross-pollination of ideas between disciplines for the advancement of knowledge frequently falls within the realm of research carried out by the academic community. College and university scholars, with support from government and non-government organizations, are frequently well- suited to advance interdisciplinary science and foster communication between disciplines.
Figure 3 outlines the research task flow chart, the areas of study, and the professional audience and applications for this dissertation research.
82
Figure 3: Research task flow chart for each of the three areas of study and resultant professional and scholarly audiences and applications.
Selection of a Subject Community
The search for a subject community in which to base the study was driven by the advance of the Emerald Ash Borer and the density of the ash tree population that the invasive insect was likely to destroy. The objective was to find a community with a large ash tree population where the Emerald Ash Borer was known to be present but before many trees had died and been removed. The city of Detroit had suffered the Emerald
Ash Borer infestation in the early to late 2000’s and the vast majority of ash trees in southeast Michigan were dead before this study began. Scouting trips to Cleveland,
83
Columbus, and Minneapolis / St. Paul revealed that the EAB infestation was coming, but mass removal of ash trees was still several years away. The Chicago region was heavily stocked with ash trees and was in the early to mid-stages of heavy EAB infestation with tree mortality already apparent in some communities.
The Davey Resource Group provided contact information for city foresters and municipal arborists in approximately 20 communities around the City of Chicago.
Communities to the south such as Tinley Park were found to have heavy infestations with more than 50% of the ash trees already dead. Communities to the north such as
Lincolnshire were found to have EAB infestations but with relatively low proportions of ash trees in the urban forest. The Village of Arlington Heights had very favorable conditions for this research including:
High proportion of ash trees (> 35%) planted along streets and in parks
Detailed street tree inventory with mapping of all known ash trees
Known presence of EAB but relatively low tree mortality to date
An active program of preventative systemic insecticide injections that would
preserve ash trees in some areas, creating a control condition that could be
compared to the treatment areas of tree removal
A pending ash tree removal program that would rapidly remove ash trees in
targeted neighborhoods
A cooperative Village forestry staff willing to share information on the ash tree
removal schedule
84
Selection of Sound Recording Sites
Sites for collecting sound recordings were selected by first inspecting tree inventory maps and receiving information from the Village arborist on which neighborhoods were scheduled for tree removal. Once inside, the neighborhood sites were sought where the sound recorder mounted on a tripod could be located on the strip of grass between the curb and sidewalk, which technically is Village property.
One challenge to be managed was the fact that only one researcher was present during most of the recording dates, and two sound recorders needed to be in operation, meaning that one of the recorders would be unsupervised. It was not favorable to leave a digital sound recorder alone near a public sidewalk and road, so one of the recorders was frequently concealed in a low group of shrubs in a cul-de-sac island on Huron Street.
The recorder would be located before dawn and retrieved in the afternoon. Fortunately the recorder was undetected and undisturbed during each session. At another location on
Stratford Rd. I approached a homeowner and asked permission to set up the sound recorder on their property near their front porch. The digital recorder on the tripod was clearly visible to passers-by but proved to be safe by locating it near the house. Three sites (Shirra Ct., Huron St. and Prindle Ave.) were selected because they experienced heavy tree removal (treatment sites) and three sites (Stratford Rd., Suffield Dr. and East
Fleming Dr.) were selected because they experienced virtually no tree removal (control sites). One site, Wilke Rd. was selected because of its proximity to the freeway.
85
Recording of soundscapes.
The collection of sound recordings was accomplished with few complications or problems. Having limited experience with operating digital sound recorders going into the study meant that I had a modest learning curve to climb. This resulted in a few occasions early in the study that the two sound recorders were not calibrated to the exact same recording settings. This was easily correctable when one recorder was set to a higher recording sample resolution format (48 kHz / 24 bit) than the other (48 kHz / 16 bit) through a simple mathematical transformation. The more difficult (but not impossible) inconsistency to correct were the few occasions when the record volume was at different settings on the recorders. There is no published transformation to correct this problem. After discovering the inconsistency I made side-by-side calibration recordings with the two recorders set first to identical record volume settings and then to the different settings used in a few of the early recordings and developed an approximate transformation for the power and energy measurements (sound attributes associated with time, frequency and entropy were negligibly impacted by slight differences in sound recording settings). But for the purposes of this study, recordings made without identical sound recordings levels were not used in the soundscape analysis.
There were no other problems suffered related to the collection of the sound recordings. The sound recorders proved to be durable and reliable. One of the recorders was slightly damaged when a wind gust tipped the tripod on which it was mounted, and the recorder landed on a concrete curb. A side-by-side calibration test was made
86 immediately following this event and the two recorders were found to still be consistent in the measures for sound attributes.
Sound recordings are distorted during periods of moderate to high winds. The foam wind screens reduce some of this distortion, but not all of it. The sound recorders can also be damaged if subjected to water. Therefore sound recording sessions were limited to periods of relative calm and rain-free weather, meaning that geophony (sounds associated with weather) are under-represented in the sound recordings.
Selection of Sound Analysis Software
Several sound analysis programs were available. The Raven program produced by the
Cornell Lab of Ornithology was selected because it was specifically designed for use in analyzing soundscapes, and because the Cornell Lab offered training classes for its use.
Selection of Sound Analysis Methods
Sound recordings were arranged in matched pairs. The same sound attributes were compared in each pair using the same sound types. Each sound attribute for each sound type was compared once for each pair of sound recordings, and no sound recordings appeared in more than one pairing. Because of this exclusive one-on-one paring statistical analysis using t-Tests was appropriate. Ad hoc corrections such as the
Bonferroni correction were not needed.
87
When comparing proportions of sound types between pairs of sound recordings, calculating Z scores and using the Z scores to estimate p-values was determined to be the proper statistical tool.
Microsoft Excel equipped with the data analysis pack was used for statistical analysis.
Selection of Sound Recording Treatments
When approaching the design of the test sessions, the first important question to address was how to select the sound recordings to be used as treatments during the sessions for tests of directed attention. On the one hand, using a wide variety of recordings from both the spatially and temporally matched pairs would provide the most complete sample of many tree versus few tree recordings. On the other hand using many recordings during the test sessions would inject another layer of variability into the tests, and as a result there might only be several test participants who were subjected to each of the many recordings. To limit this additional variable it was decided to use only one of the matched pairs of recordings during the test sessions. The cost of limiting variability on the recordings was the possibility of selecting a pair of recordings that were not representative of true changes to a soundscape as a result of tree canopy loss.
It was anticipated that the change in the soundscape attributable to tree canopy loss may be so subtle (even if proven to be statistically significant for multiple sound attributes) that the effect on directed attention may be small, if at all. Therefore it was decided to add a third sound recording to the treatment list. This third recording was gathered near the freeway that is located on the western edge of Arlington Heights. This
88 sound recording is noticeably different in sound quantity and quality from the matched pair of recordings, even to the casual listener. This distinctly different recording was added to test if a radically different sound treatment would lead to significantly different scores on tests for directed attention.
Finally, as experiments benefit from having both treatments and controls, a control situation of having no sound recordings played during the test sessions was added. The three treatment recordings and the one control were randomly assigned to the test sessions. More than 50 test subjects participated in each of the three sound treatments and the no-sound control situation.
Selection of Tests for Directed Attention
An important question in the research design was how to measure the effect of soundscapes on humans. One option considered was to conduct a series of qualitative interviews and seek the insights and opinions of residents of Arlington Heights.
Previous studies on the effects of access to nature on directed attention reported the use of several standardized psychological tests including Necker Cube Pattern Control,
Stroop Test, Digit Span Forward and Backward, and Trail-Making Test.
Unsure of which method or tests to use, I sought out the advice of Dr. Bernadine
Cimprich at the University of Michigan. Dr. Cimprich had conducted several research projects involving medical patients and access to nature. Patients were randomly assigned to groups who either spent regular time outside or did not. The patient populations were given tests for directed attention. In my interview with Dr. Cimprich I
89 asked about what comments and insights the patients provided about their outdoor experiences and how these experiences were perceived relative to their health. To my surprise, Dr. Cimprich responded, “I don’t know. I didn’t ask.” I asked why that information was not collected and she responded that those opinions were subjective, qualitative and were difficult to quantify. The tests for directed attention however were based on measurable quantitative scores that could be statistically analyzed. Using this advice I chose to also follow the methods employing tests for directed attention so as to be consistent with previously published studies.
I would have preferred to use more than two tests for directed attention, but I discovered that very few volunteers were willing to invest more than an hour of their time for this study. Therefore in order to keep the entire test session, including pre-test stressing, a recovery period, and the tests for directed attention to a maximum of 60 minutes in duration, only two tests for directed attention were used.
Selection of Demographic Information
Based on a review of the literature it was known that numerous variables can affect a person’s response to sounds. Age is an obvious variable if for no other reason that as most people age their sensitivity to higher frequency sounds diminishes. Perhaps one of the most important variables is the environment in which a person lives and/or works.
Humans tend to acclimate to the acoustic conditions that normally surround them, so it was believed to be important to record information about the population density of the participant’s place of residence. Some research studies reported differences in scores for
90 tests for directed attention and measures of self-control that were attributable to gender, so this demographic information was also gathered.
Selection of Stroop Test Platform
There are a variety of methods of administering the Stroop Test ranging from displaying paper examples of congruent and incongruent color matches and recording the verbal response of the participant to computer-based versions of the test. With the help of several computer programmers at the Davey Tree Expert Company I attempted to build a computer-based Stroop Test but found that the results lacked several important capabilities. Our version of the test only recorded how many of the color test pairs were answered correctly or incorrectly and the entire time elapsed from the start to the finish of the test. It did not record the response time for each color pair. I felt it was important to be able to compare the response times for congruent and incongruent color matches for each pair, and it was important to be able to exclude incorrect answers from the timed analysis. At the time we were finishing the build of our prototype program, Dr. Conor
McLennan of the Cleveland State University Psychology Department introduced me to his Ph.D. student Sara Incera, who had built a version of the Stroop Test using the
MouseTracker program. The MouseTracker program offered several key advantages over other versions of the Stroop Test, namely the ability to measure the response time of each color pair presentation with great precision, and the ability to easily randomize the presentation of congruent and incongruent color pairs. Based on these features, the
MouseTracker program was selected to administer the Stroop Test.
91
The Necker Cube Pattern control test uses only one variable to monitor, namely the frequency with which the cube line drawing appears to flip perspective within a given time period. Since this test relies on the participant’s response to an optical illusion, it is limited by the self-reporting by the participant. Therefore the Necker Cube Pattern
Control test was easily built using the Microsoft PowerPoint program to present repeated identical slides of the cube line drawing. At the end of the 3-minute timed period the researcher simply noted the number of times the participant had advanced the slides.
Selection of Participants for the Tests for Directed Attention
All of the participants for the tests for Directed Attention were volunteers. To enlist as many willing volunteers as possible, I set up testing stations at events in many locations, including:
The Davey Institute of Tree Sciences – Kent, Ohio
Davey Resource Group utility arborist training
o Richmond, Virginia
o Des Moines, Iowa
o Baltimore, Maryland
The Ohio Chapter of the International Society of Arboriculture annual conference
– Sandusky, Ohio
Michigan Technological University School of Forest Resources and
Environmental Science – Houghton, Michigan
The Davey Institute of Grounds Management – Kent, Ohio
92
The Chautauqua Institution – Chautauqua, New York
The International Society of Arboriculture 2016 annual conference – Fort Worth,
Texas
I found that I was most successful at attracting participants at events where I was speaking or teaching. Attendees at these events where more likely to have an interest in tree-related topics and seemed grateful for the delivery of information that I had already shared.
On several occasions I attempted to arrange a convenient time and location at the
Arlington Heights public library to administer the tests for Directed Attention to local residents, but the library would not provide space to a Arlington Heights non-resident, and I was unable to enlist the support of a local resident who would arrange for the test room and remain in the room at all times during the testing.
Participants were randomly assigned to one of three recorded sound treatments or the control situation of no sound recordings played during the trials.
Analysis of the Tests for Directed Attention
Unlike the analysis of sound attributes that compared a series of one-on-one matched pairs of data, the analysis of the tests for directed attention simultaneously compare the results of the tests with three different sound treatments and an additional control setting of no sound. Therefore, due to the multiple simultaneous comparisons the Analysis of
Variance (ANOVA) method of statistical analysis was employed. ANOVA indicates when there is a statistically significant difference between at least one of the pairs in a
93 multiple pair comparison, but the method does not indicate in which pair or pairs the difference exists. When the ANOVA analysis did indicate a significantly significant difference was found, a series of t-Tests was used to test each individual comparison of test results from the treatment and control sound recording trials. However this creates the undesirable situation of multiple comparisons within t-Tests, which dilutes the desired confidence interval that in all tests was 95% (alpha = 0.05). To rectify this problem the
Bonferroni correction was applied to the alpha level.
Necker Cube Pattern Control.
The analysis of the Necker Cube Pattern Control test involved determining if there are significant differences between the values of the single variable that was measured, namely the number of perceived perspective shifts of the cube during the three-minute test period relative to the sound treatment experienced during the session. Since there are three sound treatments (and one no-sound control) and the task is to compare each sound treatment to every other sound treatment and control, analysis of variance (ANOVA) is an appropriate tool for this task.
ANOVA was conducted on the Necker Cube scores for the four groups of participants as defined by the sound treatments. These groups were also divided into subgroups determined by three criteria:
Gender (Male, Female)
Age (born before 1980, born during or after 1980)
Community type by population density (Rural, Suburban, Urban)
94
Therefore, including the primary grouping by sound treatment and the sub-groups by sound treatment, 8 ANOVA analyses were conducted on the Necker Cube data
Stroop Test.
Unlike the Necker Cube Pattern Control test that measured a single variable, the
MouseTracker program that was used to present the Stroop Test measures multiple variables, including the following ten:
Response time
o Congruent color matches
o Incongruent color matches
Maximum deviation from an ideal straight line time
o Congruent color matches
o Incongruent color matches
Maximum deviation from an ideal straight line distance
o Congruent color matches
o Incongruent color matches
Area under the curve
o Congruent color matches
o Incongruent color matches
The number of incorrect answers
o Congruent color matches
o Incongruent color matches
95
ANOVA was conducted on each of the ten Stroop Test variable scores for the four groups of participants as defined by the sound treatments. These groups were also divided into subgroups determined by three criteria:
Gender (Male, Female)
Age (born before 1980, born during or after 1980)
Community type by population density (Rural, Suburban, Urban)
Therefore including the primary grouping by sound treatment and the sub-groups by sound treatment, 80 ANOVA analyses were conducted on the Stroop Test data.
Summary
The foundation of the research design for this dissertation study was begun many years ago as I came to understand my own experience of living near the city of Detroit, including the observation of thousands of trees being cut down as they were infected with
Dutch Elm Disease. Experiences lived at a young age tend to set a trajectory for future directions, and professionally I came to build a career in urban forestry. When yet another wave of thousands of trees were killed in the Detroit area, this time by Emerald
Ash Borer, the repeated ecological tragedy led me to wonder “How does this affect the people who live in these neighborhoods?”
As much as any other reason, these observations and questions led me to enroll at the
Levin College of Urban Affairs at Cleveland State University to seek a better understanding of how communities are designed, built and maintained, and to wonder about yet another set of questions: “How are built systems like cities connected to natural
96 ecosystems?” and “How can we build better cities by improving their connection to the natural ecosystems in which they exist?”
The opportunity to study at Cleveland State led me to an extensive review of the published literature that focused on understanding the environmental, economic, and social benefits that urban forests provide to humans living in cities. It was the surprising array of social benefits related to human health, including aspects of physical, mental and emotional health, that inspired further inquiry.
The questions kept coming: “Since we humans experience our world through our five senses, how do the sights, sounds and other sensory stimuli of cities support (or not support) our health?” which in turn led to “Can we improve our health by increasing what our senses find to be pleasing and reassuring while decreasing what our senses find to be irritating and distracting?” Many studies and experiential reports in the literature point to improving access to nature as an effective tool for achieving these health-improving objectives.
Many of the studies that explored the benefits of access to nature in cities focused primarily on what humans see in their parks, along their streets, in the spaces around their homes and even what they see outside of their windows. At the same time that I was learning about the benefits to humans of access to nature in cities I also became aware of the emerging field of study of soundscape ecology. Since most studies related to sounds in cities considered only the quantity of unwanted sound, otherwise called “noise,” I realized that there was a gap in our understanding of the potential benefits of increasing access to supportive sounds while decreasing our exposure to detrimental sounds.
97
And still the questions kept coming: “Can our experience of cities be improved by decreasing human exposure to sounds that cause fatigue and increasing opportunities to experience sounds that relieve fatigue?” To find the answer to this question it became apparent that urban soundscapes should be investigated not only from the perspective of unwanted sound intensity in the form of noise, but also from the perspective of sound quality. “What are the quantitative and qualitative components of an urban soundscape and how do these components individually and collectively affect humans?”
Clearly the components of urban soundscapes include human-sourced sounds and sounds associated with nature. While I wrestled with these questions of cities, human health, and sounds I realized that a significant change to urban forests was underway as the Emerald Ash Borer population advanced and killed trees. Here was an opportunity to test some of these questions in a relatively controlled before-and-after setting.
As will be discussed in the section titled “Recommendations for future research” in
Chapter X of this dissertation, the predictable and on-going advance of Emerald Ash
Borer provides a rare opportunity to measure a wide range of attributes within communities both before and after the destruction caused by the invasive insect takes place. As with the introduction and spread of Chestnut Blight and Dutch Elm disease before current times, and with additional threats including Asian Longhorned Beetle looming in the future, the opportunities to study these natural experiments frequently are lost in the research that is quickly and appropriately focused on eradicating the pathogen.
Never-the-less, it is advisable to recognize the research value that exists within the problem, and pursue the hypotheses that develop from “I wonder” questions.
98
It was through this lengthy journey of observations followed by questions followed by purposeful inquiry of the literature followed by more questions followed by more observations that led to the eventual questions that drove the research design of this dissertation.
99
CHAPTER V
METHODS PART 1A – SOUNDSCAPE DATA COLLECTION
Locating a Host Community
The first task of the research project was to identify potential communities in the
Chicago metropolitan region that could serve as suitable data collection sites. The criteria included:
Neighborhoods or individual streets where ash trees represent greater than 20% of
the total tree canopy cover. It is not known if there is a critical level of change in
tree canopy that will result in meaningful changes to the soundscape, but it is
preferable to start the study on the high end of tree canopy change to increase the
likelihood of detecting changes in the soundscape, and work down rather than to
begin with small changes and work up.
o Relatively small residential lots are preferable because the street trees
(those that are controlled by the municipal government and therefore more
likely to be removed as a group) represent a higher proportion of the tree
canopy (because there is less room on the lots for trees in yards).
100
The ash trees may be infested with emerald ash borer but preferably have at least
50% of their foliage remaining (because of potential sound attenuation and
reflecting, and for wildlife habitat).
The ash trees are likely to be removed within the next 2 years.
The first two criteria can be determined by observation but the last item requires information from local authorities who are charged with the removal of infested ash trees.
Therefore it was essential to get assistance from city foresters/municipal arborists to identify the schedule with which ash street trees were likely to be removed. With this information neighborhoods could be targeted in the order that removals would take place.
The Davey Tree Expert Company has many municipal clients in the metropolitan
Chicago area and local managers provided the contact information of those people responsible for public tree maintenance. One of the project managers from the Davey
Resource Group, Josh Behounek, went the extra step of directly contacting his clients and asking for their cooperation in providing tree location and removal information.
Responses were received from the city foresters / municipal arborists of the following northeast Illinois communities:
Tinley Park Schaumburg Round Lake
Westmont South Elgin Grays Lake
Riverside Elgin
Lincolnwood Mount Prospect
Skokie Glenview
Park Ridge Evanston
Lake Forest Lincolnshire
101
Each of these communities was visited and evaluated as potential sites for the study.
Some communities were found to have relatively low populations of ash trees. Others had suffered from Emerald Ash Borer infestation for more than two years and the majority of ash trees were already dead. Some had already begun to remove ash trees and others reported no plans to remove trees within the time frame of the study.
While driving through the suburbs northwest of Chicago a startling discovery was made in the Village of Arlington Heights. Many of the street trees in residential neighborhoods had large tags tied to them. Each tagged tree was an ash tree, and the tag
(Figure 4) provided residents with instructions on how they could find information on the village’s battle against Emerald Ash Borer. The sign suggested that residents go to the
Village of Arlington Heights web site and access the Emerald Ash Borer management plan. A detailed map was available on the website that included the locations of all of the publicly maintained ash trees within Arlington Heights. It was immediately apparent that the Village was populated with thousands of ash street trees, which comprised the vast majority of sites along residential streets (Figure 5). Along many streets, an ash tree was planted in front of virtually every Figure 4: The City of Arlington house. Figure 6 presents an aerial image of a Heights, Illinois uses signs tied to neighborhood in Arlington Heights and Figure 7 ash street trees to provide residents with information on Emerald Ash shows the relatively large ash trees that line both Borer and the city’s plan for dealing with the insect pest. sides of Canterbury Street.
102
Figure 5: Portion of Village of Arlington Heights ash trees map (2011). Red dots show locations of ash trees. The red arrow shows the location of Canterbury Street (pictured below). Map obtained from http://www.vahgis.com/AshTrees/ on 7/16/13.
Figure 6: Aerial photo of portion of Arlington Heights. Even though residential lot sizes are adequate to support many yard trees, the nearly exclusive use of ash for street trees represents at least 20% of the total canopy. The red arrow shows location of Canterbury Street (pictured in Figure 7). Image obtained from Mapquest.com on 9/23/13.
103
Figure 7: Canterbury Street in Arlington Heights, Illinois. Virtually every street tree visible is an ash tree. Photo taken on July 16, 2013 looking west from Prindle Ave.
Meeting with local officials in the Chicago metropolitan area.
Following the initial scouting trip, contact was made with local officials in the communities that appeared to have the greatest concentration of living ash trees. A meeting was arranged on September 3, 2013 with Ms. Sandy Clark, the City Forester for
Mount Prospect, Illinois and her assistant Mr. David Hull. At this meeting Ms. Clark and
Mr. Hull explained that ash trees are removed systematically in regions of the city in the order in which the 5-year pruning cycle is scheduled. Ash trees had been being removed for several years, and there were few areas remaining with high concentrations of infested trees. They provided a map of areas where they expected to remove trees during the
104 winter of 2013-14. A visit to these areas revealed them to have ash concentrations lower than the targeted 20% canopy, but still worthy of consideration.
Contact was established with Mr. Dru Sabatello, the City Forester for the Village of
Arlington Heights and a meeting was arranged with him and his assistant, Ms. Ashley
Karr on September 4, 2013. During the meeting Mr. Sabatello explained that the Village had embarked on a public awareness campaign that invites residents to take advantage of a cost-sharing program for treating ash street trees with an insecticide (TREE-Age) that may protect the trees from Emerald Ash Borer. In this program residents can split the cost with the Village to treat street trees adjacent to their property. In some areas of the village many residents had chosen to participate, and in other areas few residents hd chosen to join the cost share treatment program.
Ms. Karr is responsible for identifying areas within the village where damage from
EAB is most evident. These areas are then targeted over winter months for tree removal.
She provided a map (Figure 8) that identified three neighborhoods that were targeted for the winter of 2013-14, and stated that as many as 1,000 ash trees may be removed. A visit to one of the neighborhoods the following day revealed a large population of ash street trees, and the first two sound recordings were made.
105
Figure 8: Locations of three neighborhoods in Arlington Heights where ash trees were scheduled to be removed during the winter of 2013-14.
Sound Recording Equipment
The sound recordings collected for this research were gathered on two Zoom H4n digital sound recorders. Data were recorded on removeable SD memory cards. All of the original recordings were copied to several backup devices inclucing two laptop computers and two Seagate 3 TB external hard drives. As each SD memory card approached the capacity for data storage it was removed and archived, and replaced with a new SD memeory card. No original data were copied over. In the field the sound recorders were mounted on Manfrotto MKC3-H01 tripods. Figure 9 shows one of the
Zoom recorders mounted on a tripod.
106
Figure 9a (left) - A Zoom H4n digital sound recorder; and 9b (above) deployed at the intersection of Shirra Ct. and Raleigh St. in Arlington Heights, Illinois.
Sound recorder options and settings.
The Zoom H4n digital sound recorder has several settings that affect the quality and data storage requirements of the recordings. For the recordings made in this study the following settings were used:
Mode: Stereo
Lo Cut: Microphone 98 Hz; Input 98 Hz. (The low cut filter eliminates the
distorted sound that occurs when wind blows directly on the microphone.)
Recording Format: WAV 48 KHz/16 Bit (several early recordings were made at
WAV 48 KHz /24 Bit.) The recording format setting defines the sampling rate of
107
the analog-to-digital conversion, and the bit depth of the analog to digital
conversion with which sound data are collected. The combinations of 48 KHz
sampling rate with either the 24 bit or 16 bit depth provide recorded sounds that
are consistent with human hearing in the original soundscape. The 16 bit format
was utilized for most of the recordings to allow for longer recording time data
storage on the SD memory cards without discernable degradation of the recording
quality.
Record Volume: 90 (several early recordings were made at 80)
Windscreen on
Calibration of sound recording equipment.
Before the two digital sound recorders were deployed in the field, they were set up side-by-side to gather simultaneous recordings. These recordings were then analyzed using the Raven sound analysis program using the “Slice and Dice” method, as will be described in the section titled “Slice and Dice Analysis” within Chapter VI. The two recordings were found to show no statistically significant differences in any of the wavelength bands. Toward the end of the collection of sound recordings the two recorders were once again set up side-by-side for another calibration recording. Once again analysis using the Raven program showed no statistically significant differences in the respective sound recordings. This demonstrates that any difference between the soundscape recordings was due to the soundscapes themselves and not as an internal function of the sound recorders.
108
Selection of Sound Recording Sites
The street tree inventory data for Arlington Heights showed that more than 60% of the publicly owned trees were ash trees, and most of those had been infested with Emerald
Ash Borer but were still alive. Some ash trees had been treated with a systemic insecticide in an effort to protect the trees from the insect, and those trees that received treatment before being attacked were generally in good condition. Those trees that were not treated varied in condition from showing the initial signs of damage (less than 10% of the canopy affected) to completely dead. It became apparent that because of the large proportion of ash trees in the community, the amount of available tree inventory data, and the village’s aggressive plan for tree removals in the coming 2 years, that Arlington
Heights was a desirable location in which to focus the research.
At the September 4, 2013 meeting with Dru Sabatello, the Village Forester of
Arlington Heights, and his assistant Ashley Karr, plans for tree removals for the coming year were reviewed. During the summer of 2013, most of the street trees had been inspected. Those ash trees that were determined to have less than 50% of living foliage in the canopy were identified in the tree inventory and marked for removal. Mr.
Sabatello and Ms. Karr later provided a map of areas within the village that were scheduled to have trees removed between December 2013 and May 2014. Most of the tree removals along single-family residential streets would occur in the Greenbrier neighborhood of Arlington Heights located north of Palatine Road and east of State
Route 53 (Figure 10). This area had the most advanced population of Emerald Ash Borer beetles and the lowest proportion of ash trees that had been treated with the insecticide.
109
A series of sound recordings began at two locations within the Greenbrier neighborhood: one at the corner of Shirra Court and Raleigh Street, and the second at a small traffic island on Huron Street (Figure 11). Both of these locations would experience the removal of more than half of the street trees in the immediate area
(Figures 12a & 12b and 13a, 13b, and 13c).
The removal of trees in the Greenbrier neighborhood occurred in a relatively short amount of time over several weeks during March through May 2014. Trees in the area of
Shirra Court were removed on Monday May 5. The first post tree-removal sound recordings were collected on Wednesday May 7. Trees in the area of Huron Street were removed on Friday May 16, and the first post tree-removal sound recordings were collected on Monday May 19.
On most days during which sound recordings were collected only one researcher was available to tend to the sound recorders. In order to prevent unauthorized tampering or theft of the sound recorders, one of the sound recorders was concealed in a group of shrubs in the traffic island of the Huron Street location. The sound recorder was mounted on a tripod at approximately 1.5 feet (0.3 m) above ground level and surrounded by dense juniper shrubs (Figure 14). The sound recorder at the Shirra Ct. location was mounted on a tripod at a level of approximately 4 feet (1.0 m) above ground level, and located on a grassy strip between the curb and sidewalk on the southwest corner of the intersection of
Shirra Ct. and Raleigh St (Figure 9). Care was always taken to set the recorders in the same location and oriented in the same direction each time subsequent recordings were made.
110
Figure 10: Location of Greenbrier neighborhood in Arlington Heights, Illinois (maps obtained from MapQuest.com on July 25, 2014).
111
Figure 11: Street trees in the Greenbrier neighborhood of Arlington Heights, Illinois are represented by green dots. Red dots indicate locations of ash trees removed between December 2013 and May 2014. Additional trees were removed throughout the neighborhood to the south during the months that followed. The blue arrow indicates the location where sound recordings were gathered on Shirra Court; the yellow arrow indicates the location where sound recordings were gathered on Huron Street. Map provided by Ashley Karr, Village of Arlington Heights.
112
Figure 12a (above) – Shirra Court as photographed on September 4, 2013 before ash trees were removed, and 12b (below) as photographed on May 17, 2014 after ash trees were removed.
113
Figure 13a (above) – Huron Street as photographed on September 4, 2013 before ash trees were removed, and 13b (below) as photographed on July 16, 2014 after some ash trees were removed.
114
Figure 13c (above) – Huron Street as photographed on April 9, 2016 after additional ash trees were removed.
The method of tree removal is not important to the research other than contributing to the very short time interval between streets with many trees and streets with few trees.
However it was very interesting to observe the contract tree removal crews accomplish their task. A whole-tree harvester was employed to cut and limb the trees (Figure 15).
The same machine inserted the limbs of the tree into a large chipper. Those tree trunks that were too big to be consumed by the chipper (larger than 10 inches (25 cm) in diameter) were stacked along the street side and a self-loading log truck transported the logs off site.
115
Figure 14 a (above – July 15, 2014) & 14b (below – April 9, 2016): The traffic island at the Huron St. location includes a group of juniper shrubs in which the sound recorder was concealed to prevent tampering or theft.
116
Figure 15: A whole-tree harvester was used in the Greenbrier neighborhood to remove ash trees. An entire tree could be cut, reduced to chips, and blown into a chip van in less than 10 minutes. This efficiency led to entire neighborhood blocks losing most of their tree canopy in a single day.
In early 2014 while trees were being removed in the Greenbrier neighborhood, fewer trees were removed approximately 1.5 miles east in the Carousel Park neighborhood because a higher proportion of ash trees had been injected with a systemic insecticide.
Three streets, East Fleming Drive, Stratford Road, and Suffield Drive served as control locations for sound recordings in areas that experienced few tree removals to be compared to the sites that experienced many tree removals (the treatment areas). Few ash trees along Prindle Avenue in the Carousel Park neighborhood had received the insecticide and these tree would be removed by early 2016. These locations are illustrated in Figure 16.
117
Figure 16: The red arrows show the locations within the Greenbrier neighborhood where sound recordings were gathered in areas of many tree removals. The blue arrows indicate the locations within the Carousel Park neighborhood where sound recordings were gathered in areas of few tree removals. The yellow arrow indicates the Prindle Ave. location within the Carousel Park neighborhood where sound recordings were gathered with many tree removals.
Figure 17a and 17b show Prindle Avenue in July, 2013 and again in July, 2014.
During this period a few trees were removed due to EAB damage but most of the street trees remained. However, those trees that remained were in noticeably worse condition and exhibited a much higher proportion of dead canopy than they showed during the summer of 2013. Figure 17c shows the same location on Prindle Avenue in August,
2015. By this date most ash trees had been removed.
118
Figure 17a (above) – Prindle Avenue as photographed on July 17, 2013, and 17b (below) as photographed on July 16, 2014. Few ash trees were removed during this period but the condition of the remaining ash trees had deteriorated considerably.
119
Figure 17c (above) – Prindle Avenue as photographed on August 13, 2015. Most ash trees in the immediate vicinity had been removed by this date. East Fleming Drive (Figure 18), Stratford Road (Figure 19), and Suffield Drive
(Figure 20) were three streets that lost few trees and served as control sites where sound recordings were gathered. The majority of the publicly-maintained trees along these three streets were treated with TREE-age (active ingredient Emamectin Benzoate) to protect the trees from attack by EAB. The treated trees were evident by the tags attached to the trees at the time of treatment (Figure 21). The presence of these tags was helpful in knowing which areas would not experience tree removal during the following year and those areas where tree removal was more likely.
120
Figure 18a (above) – East Fleming Drive as photographed on October 8, 2013, and 18b (below) as photographed on July 15, 2014. No ash trees were removed during this period due primarily to the systemic insecticide treatment the trees had received.
121
Figure 19a (above) – Stratford Road as photographed on October 19, 2014, and 19b (below) as photographed on May 18, 2015. No ash trees were removed during this period due primarily to the systemic insecticide treatment the trees had received.
122
Figure 20a (above) – Suffield Dr. as photographed on May 9, 2014, and 20b (below) as photographed on September 3, 2014. Few ash trees were removed during this period due primarily to the systemic insecticide treatment the trees had received.
123
Figure 21: Publicly- owned trees that were treated to protect from
Emerald Ash Borer attack are marked with tags.
Why were some trees treated and other trees left without protection? If a tree exhibited advanced signs of EAB infestation it was not included in the population for possible treatment. Those trees that exhibited early signs of infestation, or appeared to be free of Emerald Ash Borers, were included in the population of those that may be treated.
The Village of Arlington Heights offered homeowners an opportunity to split the cost of treatment for street trees in front of residential properties. The cost of treatment was dependent on the size of the tree but generally fell within the range of $100 to $200 for one treatment. Therefore, property owners could elect to spend $50 to $100 to attempt to save the tree in front of their home. In some cases homeowner associations or block clubs organized to arrange for all of the trees on their street to receive treatment. In other areas treatment was sporadic. This provides circumstantial evidence that some residents place at least $50 to $100 value on a street tree in front of their home. It also provides evidence that some residents place little or no value on the street trees.
124
Measurement of Tree Canopy Cover
As the damage from Emerald Ash Borer intensified in Arlington Heights, most of the ash trees showed increasing visual evidence of decline. The Village Forester and his staff used this visual evidence, and the records of trees that had been treated with the insecticide, to determine which trees would be removed and which trees would remain.
As can be seen in the preceding figures, in some cases the ground-level visual change to individual streets was dramatic. Figure 22 presents 180 degree panoramic photos of the recording sites on Prindle Avenue and Stratford Road, which are helpful in visualizing the difference in tree canopy cover over a larger area than is shown in the previous photo pairs, but for the purpose of this study it would be helpful to have a method to quantify tree canopy loss within individual neighborhoods.
Figure 22a (top): May 18, 2015 180-degree panoramic photo of the recording site at Stratford Rd. and 22b (above) of the recording site at Prindle Ave.
125
Shortly after Arlington Heights was selected as the community in which to locate the study, aerial images of the individual test neighborhoods were obtained from three
Internet sources: Google Maps, Mapquest, and Bing. Some of the available imagery was taken during leaf-off season for deciduous trees, and some of the imagery was taken during leaf-on season, which was preferred for identifying tree crowns. Fortunately, toward the end of 2015, new leaf-on aerial images became available through Google
Maps that clearly showed where trees had been removed up until the spring of 2015.
Additional trees have been removed since this imagery was captured, but by the time these images were flown, the majority of trees removed during the study period of July
2013 to August 2015 are shown as absent on the Google Maps imagery.
Leaf-on aerial imagery taken before July 2013 (obtained from www.Mapquest.com) and leaf-on aerial imagery taken after May 2015 (obtained from www.Google.com/Maps) were printed at the same scale. An area of approximately one-quarter mile radius around the intersection of Shirra Ct. and Raleigh St. in the Greenbrier neighborhood was included in a pre-tree removal and post tree removal pair of aerial images (Figure 23). A similar sized area was included in a pair of images centered on the Huron St. traffic island in the Greenbrier neighborhood (Figure 24). A third pair of aerial images was produced for the Carousel Park neighborhood that includes the Prindle Avenue, Suffield
Drive, and Stratford Rd. sites (Figure 25).
Using a ¼ inch dot grid, tree canopy cover was measured on the pre-tree removal and post-tree removal aerial images for the Shirra Ct. and Huron St. areas of the Greenbrier neighborhood and the Prindle Ave./Stratford Rd./Suffield Dr. area of the Carousel Park neighborhood.
126
Figure 23a (above right) Pre-tree removal aerial image from Mapquest and 23b (below right) post-tree removal aerial image from Google Maps of the Shirra Ct. area of the Greenbrier neighborhood. Pre-tree removal tree canopy cover is 51% and post- tree removal tree canopy cover is 38%. This represents a loss of 25% of the tree canopy.
127
Figure 24a (above right) Pre-tree removal aerial image from Mapquest and 24b (below right) post- tree removal aerial image from Google Maps of the Huron St. area of the Greenbrier neighborhood. Pre-tree removal tree canopy cover is 45% and post- tree removal tree canopy cover is 37%. This represents a loss of 19% of the tree canopy.
128
Figure 25a (above) Pre-tree removal aerial image from Mapquest and 25b (below) post-tree removal aerial image from Google Maps of the Prindle Ave. – Stratford Rd. – Suffield Ave. area of the Carousel Park neighborhood. . Pre-tree removal tree canopy cover is 46% and post-tree removal tree canopy cover is 33%. This represents a loss of 28% of the tree canopy.
129
Sound Recording Data Collection
Between July 2013 and August 2015 sound recordings were collected in Arlington
Heights, Illinois on 35 separate days. Table 1 summarizes the sound recording dates and recording times. Appendix A lists all of the individual recordings made including locations and time durations. More than 326 hours of sound recordings were collected.
The majority of time, two sound recorders were operating simultaneously, and since on most days only one researcher was present, one recorder was tended while the other recorder was left unattended. Several of the early recordings at Huron St. were tended by a researcher, but the majority of recordings at this location were gathered by concealing the sound recorder in the dense shrubs in the center of the traffic island (Figure 14). This kept the sound recording equipment out of sight and prevented tampering or theft. The sound recordings made on Stratford Rd. were unattended. This was accomplished by approaching two of the resident homeowners and asking permission to set the sound recording equipment up on their property near the front corner of the house (Figure 26).
The sound recording equipment was in clear view of passers-by, but the fact that the equipment was on private property deterred tampering and theft. Leaving unattended equipment for extended periods of time always produced a sense of unease in the morning and relief in the afternoon when the equipment was found to still be in place.
130
Table I: Summary of sound recording dates and times
Date Start Time Total Recording Time 7/17/2013 7:00 AM 1:02 9/4/2013 10:29 AM 2:05 9/5/2013 7:16 AM 2:18 10/8/2013 10:04 AM 2:49 12/28/2013 9:09 AM 4:18 4/11/2014 3:27 PM 2:07 4/12/2014 2:30 PM 10:39 4/13/2014 7:00 AM 11:43 5/7/2014 5:42 AM 15:28 5/8/2014 5:35 AM 14:18 5/19/2014 1:30 PM 2:08 5/20/2014 5:22 AM 17:42 5/21/2014 5:27 AM 13:59 7/15/2014 2:00 PM 2:09 7/16/2014 5:45 AM 17:42 9/3/2014 1:30 PM 7:33 9/4/2014 5:58 AM 9:51 9/5/2014 5:48 AM 16:11 9/6/2014 6:00 AM 3:10 10/7/2014 12:53 PM 6:47 10/8/2014 6:00 AM 12:26 10/9/2014 6:45 AM 6:30 10/19/2014 6:47 AM 6:22 12/27/2014 9:11 AM 4:24 12/28/2014 11:14 AM 4:12 12/29/2014 9:56 AM 4:14 4/10/2015 5:55 AM 15:05 4/11/2015 5:41 AM 20:10 4/12/2015 6:18 AM 11:48 5/5/2015 1:44 PM 6:14 5/6/2015 5:37 AM 19:50 5/7/2015 5:29 AM 6:04 5/18/2015 5:24 AM 19:50 5/19/2015 5:27 AM 12:26 8/13/2015 5:56 AM 12:42 TOTAL TIME 326:16
131
Figure 26: Unattended sound recording equipment was placed on private property with the homeowner’s permission. Tampering and theft proved not to be problems.
Field notes.
At the beginning of each sound recording the researcher announced the date, time and location of the recording. Weather conditions including approximate temperature, wind conditions, cloud cover and precipitation activity was also announced into the recording and written on a field note sheet. The sound recorder settings, including the recorder number, recording format, lo-cut filter, and recorder microphone volume, were also announced on the recording and written on the field note sheet. When two recordings were collected simultaneously the locations of both recorders were noted on both recordings and the matching field sheets.
At the researcher-attended sound recording locations additional field notes were written. As the recordings were collected notes were made on the field sheet including the recording time of each distinctive sound, the source of the sound (species of bird,
132 human voice, passing car, etc.), and the approximate duration of the sound. These field sheets were helpful when listening to the recording and conducting the sound-by-sound analysis. Appendix B contains an example of a completed field sheet.
As I was taking these field notes it occurred to me that there are several ranges of sound intensity (loudness) that are relevant to human perception of the environment. I imagined myself sitting in my backyard reading a book, which requires directed attention.
It seems to me that there are some sounds that our brains treat as background sounds that do not require our attention and there are other sounds that are loud enough or different enough to draw our attention away from our task.
The separation between background sounds and attention-grabbing sounds seems to be important since my objective involves testing directed attention. Sound intensity or loudness is one variable that determines how a sound will be perceived, but it is not absolute. For example, the sound of a passing car may be fairly intense in terms of decibel level, but if it is a common occurrence for cars to pass on the street as you are reading your book your brain is likely to filter these “car-passing” sounds into the pile of those to be ignored. The sound of the Goodyear blimp passing overhead may actually be quieter than the sound reaching you from the passing car, but it is different enough to possibly draw your attention away from your book and cause you to look skyward to determine the source of the sound.
In an effort to categorize sounds into more meaningful groupings than decibel level or
“loudness”, a four-class sound rating system was devised for use while taking notes during sound recordings. The first two classes fall into the category of background sounds that are normally ignored due to the filters our brains create, and the last two
133 classes fall into the category of “attention-grabbing” sounds that are likely to interrupt directed attention. These classes are:
1. Faint – Sounds that are quiet enough to be mostly masked by normal background
sounds. In some cases, the almost imperceptible sound of a rare bird may actually
cause a person to strain to listen, in which case these faint sounds could actually
be “attention-grabbers”. But normally they are weak components of the
background.
2. Background – Sounds that are common and within the expected range of intensity
or loudness. These background sounds may be very different from location to
location. To some people the nearly constant drone of traffic noise becomes
commonplace and ignorable. Likewise, airplane noise may be so common around
airports that even intense sounds are filtered out as background noise. For any
given location, background sounds are those that are so frequent within “normal”
intensity ranges that they are considered common elements of the local
soundscape.
3. Attention – Sounds that are different enough in their source, quality, frequency, or
intensity they command attention. These sounds need not be irritating or
annoying (although they can be) – they just need to be different enough from
background sounds to draw your attention. This usually means turning toward
and/or looking at the source of the sound.
4. Loud – Loud or irritating sounds are those that command so much of your
attention that you want them to stop. These are sounds that make it difficult to
concentrate on anything else.
134
Sound intensity class using these four categories and as perceived by the researcher is listed on the field data form for each sound heard. Additional observations such as “the cicadas were only actively buzzing when the sun was shining and the temperature was warm” are also spoken into the end of each recording and noted on the field sheets.
Unfortunately, since almost half of the sound recordings were captured by unattended sound recorders, there is no sound-by-sound detailed notes for these recordings.
Spatially matched sound recordings.
Pairs of sound recordings for use in comparison can be accomplished using two methods. The first method is to control for space and vary time. This method was used to gather sound recordings at a specific location before trees were removed and return to that exact location to gather sound recordings after trees are removed.
To create the before-treatment and after-treatment conditions for comparison of data sets, recordings were made in at least one of two possible scenarios:
1. Record sounds within a short time interval (several days) before and after
trees are removed.
2. Record sounds as close to 1 year apart during which time trees have been
removed.
Recording sounds within a short time interval, immediately before and then again immediately after trees are removed, controls variables (as much as possible) that might change over time, such as the people that live in the neighborhood, and possibly some weather conditions.
135
Because soundscapes vary seasonally and diurnally, annual post tree removal recordings were scheduled as closely as possible to the calendar day, day of the week, and time of the day of the original (pre tree removal) recording. Day of the week was assumed to produce greater soundscape variability than calendar date. For example, if the pre tree removal recording was gathered on Sunday April 13, 2014, the post tree removal recording was scheduled for the following year for Sunday April 12, 2015 instead of Monday April 13, recognizing that vehicle traffic (including school buses) and air traffic patterns from nearby O’Hare airport are likely to be different on Sundays when compared to Mondays. This method controls for as much seasonal and diurnal variability as possible, although weather conditions between pre-tree removal and post-tree removal recording dates in some cases were quite different. At times weather conditions became an important consideration particularly when the spatially matched pairs of recordings differed in wind or temperature conditions. Temperature condition became relevant when it affected the calls of animals. Frogs and insect calls are particularly sensitive to temperature, and this became apparent on several dates in spring and late summer.
Temporally matched sound recordings.
The second method for matching sound recordings is to control for time and vary space. This method was used to gather sound recordings in one area that had many trees removed while making a simultaneous recording in an area, such as along a street where ash trees had been treated with the systemic insecticide, where no trees had been removed. The advantage of this method over the spatially matched sound recordings is
136 that weather conditions are essentially identical during both recordings. The disadvantage of this method is that locations of simultaneous recordings may experience different levels of background sounds, especially if one location is closer to a busy road than the other location. For this reason, sites of temporally matched sound recordings where selected for their similarity in distance from sources of constant noise, such as busy roadways.
One pair of temporally matched recordings was collected to specifically address the question of different levels of background sounds. A one-hour pair of recordings was collected at mid-day on May 6, 2015 with a sound recorder concealed in the traffic island on Huron Street, which is approximately 1,500 feet east of a busy freeway (Illinois Route
53), and the second recorder was located at the intersection of Raleigh St. and Wilke Rd., which is within 100 feet of the edge of the Route 53 freeway (Figure 27). This provided a comparison of traffic noise at distinctly different distances from the freeway.
Figure 27: A one-hour recording of the soundscape adjacent to the Illinois Route 53 freeway was collected on May 6, 2015. This is paired with a simultaneous recording located at the Huron Street location.
137
Change in Tree Canopy Not Attributable to Emerald Ash Borer
The majority of loss of tree canopy during the study period was caused by the removal of publicly-maintained ash trees. There were some additional tree removals, primarily on private property that contributed to canopy loss. The single most important event relative to tree loss occurred on September 5, 2014 when a remarkably powerful wind and rain storm occurred. During a period of approximately 30 minutes (during which sound recordings were taking place) heavy wind and rain damaged or destroyed approximately
3% to 5% of the trees in Arlington Heights (Figure 28).
To the credit of the Arlington Heights urban forestry department, new trees were quickly planted on the streets where ash trees were removed. These new trees are obviously much smaller than the trees that were removed, and their foliage does not yet contribute much to the overall tree canopy, but in a relatively short number of years the loss of approximately one-quarter of the tree canopy will be replaced by these new trees.
Figure 28: Hundreds of trees in Arlington Heights were damaged or destroyed during a wind storm on September 5, 2014, further contributing to the loss of tree canopy in the community.
138
Soundscape Data Collection Limitations
Several factors influenced or limited the collection of soundscape recordings:
1. The proportion of ash streets trees was remarkably large in the Arlington Heights
neighborhoods where soundscape recordings were gathered, and the time period
during which trees were removed was remarkably short. The combination of a
high proportion of ash trees of known location in a community with an aggressive
ash tree removal policy managed by a Village Arborist willing to share
information of the scheduling of tree removals created the almost ideal conditions
in which to conduct this type of research. The location, timing and conditions I
stumbled in to were astoundingly advantageous. Despite this good fortune, the
reality is that the street tree population comprises between 15% to 25% of the
total urban forest population in the area. Trees on private property make up the
majority of the tree canopy, and the proportion of ash trees is much lower than in
the street tree population. This means that even with the aggressive tree removal
activity during the study period, only between 19% and 28% of the total canopy
was removed. While this is a relatively high number of trees to lose in a short
period of time it remained to be seen if this change in urban forest canopy
produces a meaningful change in soundscape composition or in human cognition.
2. Purposely scheduled collection of sound recordings began in September 2013 and
aggressive removal of trees in the test neighborhoods began in May 2014.
Therefore the window of opportunity to collect pre-removal sound recordings was
less than one year, and no pre-tree sound recordings were made during spring and
139
summer months. In retrospect it would have been favorable to have begun the
collection of sound recordings in the early spring of 2013.
3. There was a minor learning curve with understanding the full range of sound
recording options on the Zoom digital recorders. This led to a few of the early
recordings gathered with inconsistent options. In some cases correcting for the
inconsistencies is relatively easy, such as for the format difference of 48 KHz / 24
bits versus 48 KHz / 16 bits, in which case a simple mathematical conversion is
possible. In other cases, such as selecting a recording volume of 80 for the initial
recording and 90 for subsequent recordings, is not as easy or precise. In this case
a side-by-side calibration recording was made and a comparison of the sound
attributes associated with power and energy are compared was adjusted. Minor
differences in recording volume level does not affect the sound attributes
associated with time, frequency, or entropy.
4. The presence of only one researcher on most of the sound recording collection
trips made it necessary to have one sound recorder unattended. The unattended
sound recorder at the Huron Street location was concealed in a thicket of bushes.
Upon listening to the sound recordings during analysis it was discovered that bird
activity in this shrub thicket was relatively high as one or more species may have
used the thicket as a nesting site. This is not a problem when comparing spatially
matched pairs of sound recordings that are both from this location, but
comparison of Huron St. recordings to other locations yielded a relatively higher
level of biophony activity at Huron St. for this reason. In retrospect it may have
been preferable to seek out a homeowner who was willing to allow the sound
140
recorder to be placed in their yard, similar to leaving the sound recorder
unattended at the Stratford Rd. location.
5. The presence of only one researcher during simultaneous recordings limited the
information gathered on the field sheets at the unattended location. This is a
minor concern since the information from the field forms was used sparingly
(mostly when visual confirmation of a bird species was needed) during the sound
analysis task.
6. Weather conditions (as expected) were not identical during the spatially matched
sound recordings. In most cases these differences were minor – mostly when one
year was noticeable warmer or colder than the comparison year. Temperature
differences at times affected the calling activity of amphibians, insects, and to a
lesser extent birds. Since sound recordings were not made during periods of high
winds or heavy precipitation the geophony associated with these conditions may
be under-represented. In retrospect it was fortunate that very little potential
recording time was lost due to inclement weather.
7. Travel distance to Arlington Heights from Ohio limited the amount of occasions
and total time that sound recordings were gathered, however the seasonal
distribution of sound recordings that were gathered is acceptable and the total
time of sound recordings greatly exceeds the capacity of time needed to analyze
the recordings.
8. In retrospect the limitations for the collection of sound recordings are remarkably
few. Overall the weather cooperated surprisingly well. The digital sound
recorders performed without failure. The citizens who came to ask about the
141 nature of my business were friendly and helpful. (The police were called on three occasions to investigate “suspicious behavior” but they too were friendly.) The most regrettable limitation is that the collection of sound recordings would have been more comprehensive had the process began at least a full year before tree removal commenced.
142
CHAPTER VI
METHODS PART 1B – SOUNDSCAPE ANALYSIS
Sound Recording Interpretation and Spectrogram Analysis
The collection of sound recordings in Arlington Heights, Illinois began in August
2013 and concluded in August 2015. Almost 100 recordings events were gathered totaling over 326 hours of recording time. In March, 2014 preparation began for analyzing a subset of the sound recordings.
Sound analysis using the Raven program.
There are several computer software programs available that are designed to analyze sounds. Some of these focus primarily on the music industry, others are for use in sound engineering and development of noise-cancelling technology, and a few are well-suited for studying the characteristics of sounds in nature. After investigating the functions and intended use of these programs the Raven Pro Interactive Sound Analysis Software
(Version 1.5) built and supported by the Cornell Lab of Ornithology was chosen for use in this project. The Cornell Lab is the repository for perhaps the world’s most extensive
143 collection of recorded bird songs and calls, and is home to a group of researchers dedicated to the study of sounds in nature. An additional benefit for Raven users is the opportunity to learn sound analysis methods from these experienced researchers through an annual Raven sound analysis workshop. I participated in the workshop held from
March 31 to April 4, 2014 at Cornell University in Ithaca, New York. Each workshop participant was encouraged to bring their own sound recordings to use during the lessons.
This was a great help in beginning to build a custom method for separating sounds into types (anthrophony, geophony and biophony), and measuring an extensive range of sound attributes.
Choosing sound presentation options within Raven.
The Raven sound analysis program offers a variety of options to assist in the interpretation of sound recordings. After initial trial-and-error attempts, I settled on a consistent collection of settings that worked well in the analysis of the sound recordings.
To begin the sound analysis process within Raven, go to File: Open Sound File and select a sound recording to open. When opening a sound recording in Raven the analyst is given the option of displaying the entire recording on a single screen (compressing time to fit the dimensions of the computer screen) or presenting sequential segments of uniform time called paging. During this research analysis, sound recordings were paged in 10-second increments. This view of very short segments of time presents a detailed look at the patterns within a sound that are not visible in more compressed time segments.
144
I also chose a 95% page increment and a 10% step increment. These define how the sound pages are advanced.
The intensity or loudness of a sound is referred to as power spectral density and is expressed in decibels (dB). Raven represents sounds as graphs in both waveforms (that graph a sound’s intensity or loudness over time), and spectrograms. Spectrograms provide more information that do waveforms; time is plotted on the X-axis, frequency is plotted on the Y-axis, and intensity (loudness) is presented in colors (or shades of grey).
For the purpose of this research, sound recordings were analyzed while viewing spectrograms, not waveforms. There are several options of “color maps” for presenting sound intensity on the spectrograms. The default option uses shades of gray to represent intensity with loud sounds appearing darker (toward black) and faint sounds appearing lighter (toward white). There are also several options using different combinations of colors. For this research all spectrograms were presented in the “cool” color map option that presents low-intensity (quiet) sounds as shades of blue, intermediately powered sounds in shades of green and high-intensity (loud sounds) in shades of yellow to white
(Figure 29).
145
Figure 29: Raven Pro Sound Analysis Software spectrogram of stereo channels of a 10-second sound sample. Time is plotted on the X-axis; frequency is plotted on the Y-axis; Intensity (loudness) is represented by color (dark colors are quiet sounds and bright colors are loud sounds).
All of the recordings used in this research were recorded in stereo. Raven allows the analyst to view both channels simultaneously (two spectrograms are displayed) or a single channel at a time. For this research, the two stereo channels were viewed simultaneously. This allows the analyst to see differences in the channels that occur when a sound source is located to one side of the recorder, and also to see the progression of sound intensity as a moving sound source passes by the recorder.
When interpreting soundscapes the analyst uses the computer mouse or other pointing device to draw boxes around a single sound signature or a group of similar signatures.
The sounds enclosed by the box are referred to as Selections. When viewing stereo spectrograms, Raven provides the option of selecting sounds independently from either channel or selecting sounds simultaneously on both channels. To the right of each
146 kilohertz scale of the two-channel spectrograms is a vertical bar shaded in blue. By holding down the computer control key and clicking on each of the two bars, their color changes to yellow, and the channel numbers become highlighted in yellow. With these settings in place, a selection made in one spectrogram appear in both spectrograms, and the sound attribute data for each channel are generated individually.
Choosing sound measurement options within Raven.
A tab labelled “Table 1” is visible in the lower left corner of a Raven workspace. By clicking and holding on the spotted bar above the Table 1 label the analyst can pull the bar upward, revealing the Selection Table with columns labelled Selection, View,
Channel, Begin Time, End Time, Low Frequency and High Frequency. These are the default data fields whose values are automatically generated when a sound signature is selected. There are many other attributes (measurements) that can be added to the table.
By right-clicking on any of the column headings, a pop-up menu appears. Select
“Choose Measurements” and a new window appears. The current measurement column headings appear on the left, indicating that they are operational. The measurement titles that appear on the right can be made operational by selecting them (one-by-one) and then clicking on the left-facing arrow. This moves the measurement title to the left list, and a new column for that measurement will appear in the selection table. The following measurements can be calculated by Raven in the spectrogram view (the condensed measurement descriptions have been taken from the Raven user’s manual) (Charif,
Waack, & Strickman, 2010).
147
Begin Time: The time at which the selection begins. Units: seconds.
End Time: The time at which the selection ends. Units: seconds.
Delta Time: The difference between Begin Time and End Time for the selection.
Units: seconds.
Center Time: The point in time at which the selection is divided into two time
intervals of equal energy. Units: seconds.
1st Quartile Time: The point in time that divides the selection into two time
intervals containing 25% and 75% of the energy in the selection. The computation
of this measurement is similar to that of Center Time, except that the summed
energy has to exceed 25% of the total energy instead of 50%. Units: seconds.
3rd Quartile Time: The point in time that divides the selection into two time
intervals containing 75% and 25% of the energy in the selection. The computation
of this measurement is similar to that of Center Time, except that the summed
energy has to exceed 75% of the total energy instead of 50%. Units: seconds.
IQR (Inter-quartile Range) Duration: The difference between the 1st and 3rd
Quartile Times. Units: seconds.
Time 5%: The point in time that divides the selection into two time intervals
containing 5% and 95% of the energy in the selection. The computation of this
measurement is similar to that of Center Time, except that the summed energy has
to exceed 5% of the total energy instead of 50%. Units: seconds.
Time 95%: The point in time that divides the selection into two time intervals
containing 95% and 5% of the energy in the selection. The computation of this
148
measurement is similar to that of Center Time, except that the summed energy has
to exceed 95% of the total energy instead of 50%. Units: seconds.
Duration 90%: The difference between the 5% and 95% times. Units: seconds
Max Time: For a spectrogram view, the first time in the selection at which a
spectrogram point with power equal to Max Power/ Peak Power occurs. Units:
seconds.
Low Frequency: The lower frequency bound of the selection. Units: Hz.
High Frequency: The upper frequency bound of the selection. Units: Hz.
Delta Frequency: The difference between the upper and lower frequency limits of
the selection. Units: Hz.
Max Frequency/ Peak Frequency: The frequency at which Max Power/ Peak
Power occurs within the selection. If Max Power/ Peak Power occurs at more than
one time and/or frequency, the lowest frequency at Max Time at which Max
Power/ Peak Power occurs. Units: Hz.
Center Frequency: The frequency that divides the selection into two frequency
intervals of equal energy. Units: Hz.
1st Quartile Frequency: The frequency that divides the selection into two
frequency intervals containing 25% and 75% of the energy in the selection. The
computation of this measurement is similar to that of Center Frequency, except
that the summed energy has to exceed 25% of the total energy instead of 50%.
Units: Hz.
3rd Quartile Frequency: The frequency that divides the selection into two
frequency intervals containing 75% and 25% of the energy in the selection. The
149
computation of this measurement is similar to that of Center Frequency, except
that the summed energy has to exceed 75% of the total energy instead of 50%.
Units: Hz.
IQR (Inter-quartile Range) Bandwidth: The difference between the 1st and 3rd
Quartile Frequencies. Units: Hz.
Frequency 5%: The frequency that divides the selection into two frequency
intervals containing 5% and 95% of the energy in the selection. The computation
of this measurement is similar to that of Center Frequency, except that the
summed energy has to exceed 5% of the total energy instead of 50%. Units: Hz.
Frequency 95%: The frequency that divides the selection into two frequency
intervals containing 95% and 5% of the energy in the selection. The computation
of this measurement is similar to that of Center Frequency, except that the
summed energy has to exceed 95% of the total energy instead of 50%. Units: Hz.
Bandwidth 90%: The difference between the 5% and 95% frequencies. Units: Hz.
Average Power: The value of the spectrogram’s power spectral density, as it
appears in each pixel, or bin, of the spectrogram, averaged over the selection (that
is, the frequency-time rectangle that forms a selection in Raven). The values of
the spectrogram’s power spectral density are summed, and the result is then
divided by the number of time-frequency bins in the selection. Units: dB.
Max Power/ Peak Power: The maximum power in the selection. In a grayscale
spectrogram, the maximum power in a selection is the power at the darkest point
in the selection. Units: dB.
Energy: The total energy within the selection bounds. Units: dB.
150
Aggregate Entropy: The aggregate entropy measures the disorder in a sound by
analyzing the energy distribution within a selection. Higher entropy values
correspond to greater disorder in the sound whereas a pure tone with energy in
only one frequency bin would have zero entropy. This is calculated by treating the
fraction of energy in a selection present in a given frequency bin as a probability.
Average Entropy: The average entropy in a selection is calculated by finding the
entropy for each frame in the selection and then taking the average of these
values. Unlike the aggregate entropy which uses the total energy in a frequency
bin over the full time span, the average entropy calculates an entropy value for
each slice in time and then averages. As a result, the average entropy
measurement describes the amount of disorder for a typical spectrum within the
selection, whereas the aggregate entropy corresponds to the overall disorder in the
sound.
Minimum Entropy: The lowest value of entropy found within the frames of the
selection.
Maximum Entropy: The highest value of entropy found within the frames of the
selection.
It is not necessary to have all of the desired measurements added to the selection table at the beginning of the analysis. Measurement options may be added to the selection table at any time and the program will calculate the measurement values for the selections that have already been drawn. The following measurements were found to be most useful to view while conducting the analysis:
151
Begin Time Average power
End time Peak power
Delta time Energy
Low frequency Average entropy
High frequency Aggregate entropy
Delta frequency Minimum entropy
Center frequency Maximum entropy
Peak frequency
Testing differences between calibration recordings.
As previously discussed, there are several variables associated with sounds or collections of sounds that can be used to measure the similarity or difference of those sounds, including the frequency range of a sound, the duration of a sound, the intensity
(loudness) of a sound, the entropy or disorder of a sound, and the source of the sound.
The first step for sound analysis was to determine if the two digital sound recorders were recording sounds in the same way or if there are measurable differences between two sound recordings collected at the same place and time.
To explore this question two sound recordings were made at precisely the same time using two sound recorders set up about 2 feet (0.5 m) apart. The sound recorders were both the same model of recorder (Zoom H4n) using the same settings (recording format of WAV 48 kHz/16 bit, low cut microphone = 98 Hz & low cut input = 98 Hz, recording volume = 90, wind screen on). The sound recordings are 31 minutes in duration.
152
The process for the identification of individual sounds and their sources is described in a later section titled “Sound signature analysis of initial sound recording pair.” For this calibration exercise the first step is to simply test the similarity or difference of these two side-by-side recordings. Rather than attempt to compare the recordings as one block of time and frequencies, the two recordings are divided into a series of slices and dices.
First, an Excel spreadsheet was built that defined slices of frequency ranges along boundaries of 2,000 Hz (0-2000 Hz, 2000-4000, 4000-6000….22,000-24,000 Hz). Also included was one broad range of 0-20,000 Hz. Then dices of time ranges are defined along 30 second intervals (0-30 seconds, 30-60 seconds…..1830-1860 seconds). This
“slice and dice” table was built for each of the two channels (stereo), and for each of the two sound recorders. The table was imported into the Raven program and the program reported the characteristics for each of the two sound recordings. A sample of the boundaries of the sound clips is illustrated in Figure 30.
For each of the twelve 2,000 Hz slices of the frequency spectrum there were 62 30- second sound samples, plus 62 30-second samples within the broad 0-20,000 Hz frequency range, yielding (13 x 62) 806 sound clips, which then is doubled (1,612) to account for the two-channel stereo recording.
153
Figure 30: Sixty-second spectrogram showing 2,000 Hz slices and 30 second dices used in the sound recorder calibration analysis. The Average Power attribute (one measure of sound loudness) was reported for each sound clip within a 2000 Hz frequency range and these data were prepared to conduct a t-
Test within Excel in which the 62 30-second dices within each frequency range are compared for each of the 13 frequency slices. Therefore there are 13 null hypotheses:
1) The mean average power value for frequency range 0-2000 Hz from recorder #1
is equal to that of recorder #2.
2) The mean average power value for frequency range 2000-4000 Hz from recorder
#1 is equal to that of recorder #2.
3) The mean average power value….
12) The mean average power value for frequency range 22,000-24,000 Hz from
recorder #1 is equal to that of recorder #2.
13) The mean average power value for frequency range 0-20,000 Hz from recorder #1
is equal to that of recorder #2.
154
Statistical analysis of data generated by Raven was performed using Microsoft Excel
2013. The data analysis tools are not accessible in Excel until the data analysis toolpak is loaded into the program. The process is as follows:
1. Click the File tab, click Options, and then click the Add-Ins category.
2. In the Manage box, select Excel Add-ins and then click Go.
3. In the Add-Ins box, check the Analysis ToolPak check box, and then click OK.
Within Excel there are several options for conducting a t-Test. Two of those options are “t-Test assuming equal variances” and “t-Test assuming unequal variances.” In order to know which of these options is most appropriate for the data an “F-test two sample for variances” was run within Excel to test if the variances from the two recordings were statistically equal or different.
There is an important detail worth noting when using the F-test and t-Test functions within Excel. Excel requires the data set with the larger variance to be entered into the
Variable 1 field, and the data set with the smaller variance to be entered into the Variable
2 field. Table IIa illustrates an F-test result that shows a higher value of variance under the Variable 2 column and should therefore be re-run after swapping the data entered in the Variable 1 and Variable 2 columns. Table IIb illustrates the corrected F-test with the data set with the greater variance entered into the Variable 1 field.
An alpha level of 0.05 was selected for each of the F-tests (providing a 95% confidence interval). The null hypothesis is that the variance of variable 1 is equal to the variance of variable 2. The variances are considered to be statistically equal when the F
Critical one-tail value is smaller than the calculated F value.
155
Table IIa (top) & IIb
(bottom): Examples of positioning variables within Excel according to
Variance value. When preparing to run an F-test Two-sample for Variances within Excel, the data set with the greater variance should be entered into the Variable 1 field. When the result shows the greater variance under Variable 2, the test should be re-run after
swapping the data sets in the variable fields. In the case shown in Table IIb the F Critical one-tail value is larger than the calculated
F value. Therefore the null hypothesis is rejected and the variances are considered to be unequal. This result indicates the proper selection of t-Test with unequal variances within Excel. Note that had the data sets not been swapped so that the variable with the highest variance appeared as Variable 1, the results would have been different as is illustrated in Table IIa, and the null hypothesis would have incorrectly been accepted.
The F-test revealed statistically equal variances for all of the slices from 0-2000 Hz to
14,000 – 16,000 Hz plus the thick slice of 0-20,000 Hz. The F-test revealed statistically unequal variances for the slices from 16,000 - 18,000 Hz up to 22,000 – 24,000 Hz.
These higher frequency slices record very few sounds and the variances in these slices are quite small, meaning that even small differences in the variances produce a relatively large ratio.
156
Based on the results of the F-tests, appropriate t-Tests (for equal or unequal variances) were conducted for each 2,000 Hz slice of Average Power data (plus the 0 – 20,000 Hz thick slice) expressed as decibels. Tables IIIa & IIIb illustrate the results of two of the t-
Tests; one for the 4,000 – 6,000 Hz slice and one for the 6,000 – 8,000 Hz slice. The t-
Test results are interpreted by comparing the T Critical two-tail value to the T stat value.
If the absolute value of the T stat value is less than the T Critical two-tail value, (and the
P value is greater than 0.05) do not reject the null hypothesis that the means of the two samples are considered equal. If the absolute value of the T stat value is greater than the
T Critical two-tail value (and the P value is less than 0.05) reject the null hypothesis that the means of the two samples are equal. In the example presented in Tables 3a & 3b, the frequency range slice from 4,000 - 6,000 Hz reveals that the mean values for Average
Power (dB) are statistically different, while the frequency range slice from 6,000 – 8,000
Hz reveals that the mean values for Average Power (dB) are not statistically different.
These side-by-side recordings were intended to test the calibration of the two digital recorders. Since they were recording the same soundscape at the same time it was expected the t-Tests would show no statistically significant differences for each of the frequency slices. The results from the t-Tests indicate no statistical differences between average power means (expressed in decibels) for all of the frequency slices except 4,000
– 6,000 Hz, 20,000 – 22,000 Hz and 22,000 – 24,000 Hz. Since the two highest frequency slices are above the normal hearing range of humans, and very little energy is collected in these frequency ranges, the differences in recorders is of little consequence.
The significant difference in the 4,000 – 6,000 frequency range is surprising. This is the frequency range that contains the constant drone of insect call plus a portion of the
157 energy from the many bird calls. One possible explanation is that while the recorders were in close proximity, one was closer to a cricket that was calling throughout the 30 minute recording.
Tables IIIa (top) & IIIb (bottom): Examples of two t- Tests for 2,000 Hz slices of
side-by-side sound recordings. The absolute value of the t Stat for the 4,000 – 6,000 Hz slice is greater than the t Critical two-tail value, and the P value is less than 0.05, therefore the means of the Average Power (dB) measurements are considered statistically
different. The absolute value of the t Stat for the 6,000 – 8,000 Hz slice is less than the
t Critical two-tail value, and the P value is greater than 0.05, therefore the means of the Average Power (dB) measurements are not considered statistically different. Testing differences between simultaneous field recordings.
Slice-and-Dice analysis of initial sound recording pair: 2,000 Hz Slices of
Average Power (dB).
After developing the slice-and-dice analysis method for comparing side-by-side calibration recordings, and using the same methods for F-Tests and t-Tests, analyses for
158 recordings made simultaneously at two Arlington Heights locations with different levels of tree canopy cover were conducted.
Prindle Avenue and Stratford Road are parallel roads separated by two blocks. Figure
31 shows an aerial photo of the Carousel Park neighborhood including Prindle Ave. and
Stratford Rd. with the locations of the two sound recorders.
Figure 31: Aerial photo of the Carousel Park neighborhood in Arlington Heights, IL. The red star indicates the position of a sound recorder on Stratford Rd. and the yellow star indicates the position of a sound recorder on Prindle Ave. on the morning of September 5, 2014. Aerial photo acquired from Mapquest.com As discussed in the previous methods section “Selection of sound recording sites” the
Carousel Park neighborhood was heavily stocked with ash trees. The street trees along
Stratford Rd. were treated with a systemic insecticide, protecting them from attack by the
Emerald Ash Borer. Most of the trees along Prindle Ave. were not treated. Figures 17a through 17c show the change of tree canopy over time on Prindle Ave; Figures 19a and
19b show tree canopy on Stratford Road.
159
Many of the trees on Prindle Ave. were removed over the summer of 2014 while during the same period few trees were removed on Stratford Road. This provides an opportunity to collect sound recordings simultaneously at locations with light tree canopy and heavy tree canopy. Both sound recorder locations are approximately equal distance south from Hintz Rd., the closest street with moderate traffic volume.
Several recordings (July 2013, September 2013, December 2013 , April 2014) were collected at the Prindle Ave. location before trees were removed and again after trees were removed (May 2014, July 2014, September 2014, October 2014, and more). The first simultaneous recording between Prindle Ave. and Stratford Rd. was made on
September 5, 2014. The recording began at 5:48 AM and concluded about 3 hours later.
There are several sound signatures including sirens of emergency vehicles on Hintz Rd. to the north and of chain saws and chippers working to the south that are audible on both recordings. This makes it possible to compare the two recordings precisely in time.
The “slice and dice” analysis of 2,000 Hz frequency slices and 30 second time interval slices was conducted using the t-Test data analysis tool in Excel. F-Tests for each 2,000
Hz frequency slice were conducted to determine if the variances of 30-second time interval measurements of Average Power (dB) are statistically different at the 95% confidence interval. The results of each F-Test were used to select an appropriate t-Test option, namely assuming equal or unequal variances.
The results of the t-Tests for Average Power (dB) reveal that the mean values in all twelve of the 2,000 Hz slices and in the 0 - 20,000 Hz thick slice are significantly different at the 95% confidence interval, even in instances when the F-Test indicates that
160 the variances are not significantly different. Table IV presents an example of F-Test and t-Test results for average power (dB) data from two of the 2,000 Hz slices.
Table IV: F-Test and t-Test results for two 2,000 Hz “slices” that test for significant differences in variances (F-Test) and means (t-Test) of average power (dB) of simultaneous sound recordings at Prindle Ave. (few trees) and Stratford Rd. (many trees) on September 5, 2014. The F value in yellow indicates no significant difference in variance in the 4,000 – 6,000 Hz frequency range. F and t Stat values in green indicate significant difference in mean value for the 4,000 – 6,000 Hz frequency range, and significant difference in both variance and mean values in the 6,000 – 8,000 Hz frequency range.
Conversion of measures of sound intensity from decibels (dB) to millivolts
(mV).
Measures of sound intensity (or loudness) including Average Power, Peak Power and
Energy are reported in decibels (dB), which is expressed on a logarithmic scale. An increase of 10 dB indicates a ten-fold increase in the power of a particular sound.
161
Therefore to compare the sums of measures of sound intensity between sound clips it is useful to first convert the dB readings to a straight scale.
Dr. Dean Hawthorne (2014) at Cornell University’s Ornithology Lab provided the formula to convert data expressed in decibels from the logarithmic scale to a straight line scale:
Milliwatts (mW) =10^(value of sound intensity in dB)/10)
So, for example, an Average Power reading of 50 dB is converted to milliwatts (mW) by:
10^(50 dB / 10) = 100,000 mW
To convert from milliwatts back to decibels the formula is:
dB =LOG(value in mW)*10
Using this formula the Average Power, Peak Power, and Energy measurements reported by Raven in decibels were converted to milliwatt units before conducting additional F-
Tests and t-Tests.
Slice-and-Dice analysis of initial sound recording pair: 2,000 Hz Slices of
Average Power (mV).
The preliminary “slice and dice” analyses were performed using the Average Power value expressed in dB. A second slice and dice analysis was performed after the Average
Power values expressed in decibels were converted to the non-logarithmic value of milliwatts.
The “slice and dice” analysis of these converted milliwatt units of 2,000 Hz frequency slices and 30 second time interval slices was conducted using the F-test and t-Test data
162 analysis tools in Excel. The results of each F-Test were used to select an appropriate t-
Test option, namely assuming equal or unequal variances.
The results of the t-Tests suggest that the Average Power (mW) means in the 0 –
2,000 Hz range, and the 2,000 – 4,000 Hz range, as well as the thick slice 0 – 20,000 Hz range are not significantly different at the 95% confidence interval (P = 0.09, 0.09, and
0.17 respectively). The ten slices between 4,000 Hz and 24,000 Hz however are all significantly different and indicate that the Stratford Rd. (many trees) recording has greater Average Power (mV) mean values than does the Prindle Ave. recording. Table V presents an example of F-Test and t-Test results for data converted from Average Power
(dB) to Average Power (mW) from two of the 2,000 Hz slices.
Table V: F-Test and t-Test results for two 2,000 Hz “slices” that test for significant differences in variances (F-Test) and means (t-Test) of average power (mV) of simultaneous sound recordings at Prindle Ave. (few trees) and Stratford Rd. (many trees) on September 5, 2014. F and t Stat values in yellow indicate no significant difference in variance and mean values in the 2,000 – 4,000 Hz frequency range. F and t Stat values in green indicate significant difference in variance and mean values in the 4,000 – 6,000 Hz frequency range.
163
Does the conversion of Average Power measurements using the logarithmic scale of decibels to the straight scale of milliwatts affect the results of the t-Tests? In two of the twelve 2,000 Hz frequency range slices and in the one thick 20,000 Hz slice the t-Test results differed between Average Power expressed in decibels and milliwatts. In ten of the twelve 2,000 Hz frequency range slices the results were the same (Figure 32).
Figure 32: Comparison of slice- and-dice t-Test results for Average Power reported in decibels and milliwatts for the sound recordings at Stratford Rd. and Prindle Ave. on
September 5, 2014. Green cells indicate statistically significant greater mean values for Stratford Rd. (many trees). Blue cells indicate statistically significant greater mean values for Prindle Ave. (few trees). Tan
cells indicate no significant difference between mean values of Average Power. Alpha = 0.05
These slice and dice analyses hold the frequency range and the time interval for each sound selection constant while considering one attribute (Average Power) of sound intensity when comparing two recordings made in areas of different tree canopy cover.
The advantage to the slice and dice approach is that all frequency ranges are considered over the entire time of the recording, and there is no overlap between sound selection cells. The method is comprehensive and objective – no sounds are left out and there can be no interpreter bias. The question this analysis addresses is “Are the two recordings different enough at this preliminary level to warrant deeper investigation?” It is apparent that for this one attribute of sound (Average Power) that these two simultaneous
164 recordings do indeed seem to be different. The next task was to consider potential differences between recordings using a larger array of sound attribute measurements.
Sound signature analysis of initial sound recording pair.
The “slice and dice” method of sound analysis creates uniform and arbitrary boundaries to sound samples across frequency range and time. The sound signature analysis takes a much different approach. This method involves a human analyst listening closely to each recording in its entirety. I was the only sound analyst for this research. While this proved to be slow work for only one analyst, the advantage is that consistency is gained, and bias from different interpretations from multiple analysts is avoided. All sound analysis was conducted using stereo headphones, either Sony model
MDR-CD160 or JVC model HA-RX900 connected to Lenovo laptop computers mounted in docking stations and equipped with full-sized monitors.
To identify individual sounds within each sound recording the appropriate recording is opened in Raven. Once a sound recording is opened in Raven and analysis options are chosen, the settings and all sound signature selections with their associated measurements can be saved as a Raven Workspace. The values for all measurements and annotation fields are saved by Raven as a Selection Table in the Excel text format (.txt). These selection tables can then be opened in Excel and converted to an Excel spreadsheet format (.xlsx). A sound recording is grouped with a Workspace (the analyst setting instructions for Raven), and a Selection Table (the sound selection time and frequency boundaries, the sound attribute measurements, and the annotation notes).
165
For each detailed analysis of sound signatures the following Raven Workspace options were selected:
Waveform turned off
Stereo channels of the spectrogram turned on
Simultaneous drawing of selections in both stereo channels turned on
Automatic commitment of selections turned on
Paging
o Page size = 10 seconds
o Page increment = 95%
o Step increment = 10%
Spectrogram display parameters
o Color display = blue/green/yellow (“cool “ display)
o Brightness = 50
o Contrast = 50
o Spectrogram display size = 910
Measurements recorded in selection table (These are the preliminary
measurements that were calculated at the time of the original sound signature
analysis. Additional measurements were added once the statistical analysis
began.)
o Start time
o End time
o Delta time
o Low frequency
166
o High frequency
o Delta frequency
o Average power
o Peak power
o Energy
o Aggregate Entropy
o Average Entropy
Analyst-derived annotation recorded in selection table
o Sound type
o Intensity class
o Notes
For each of the two matched sound recordings the analysis starting point was identified (the moment in time when each of the two recorders began simultaneous uninterrupted recording of the soundscapes) and the analyst commenced to listening to each 10-second recording interval. Initially, when conducting simultaneous recordings, the sound recorders were started individually when they were set up at a recording site, and then sounds such as sirens that could be heard on both recordings were later identified and used to synchronize the recordings. It was eventually discovered to be more efficient to start the recorders side-by-side at one location in order to synchronize the recorders from the beginning of the recording, and then move one of the recorders to the second site. The running recorders were then reunited at the initial site so that the recordings were synchronized at the beginning and end of each recording. Once the synchronized starting point was identified within Raven, the analyst selected individual
167 sounds by using the computer cursor to draw a rectangle around a sound signature. The vertical boundaries of a selection rectangle represent the start and end time of a sound, and the horizontal boundaries of a selection rectangle represent the low and high frequency range. Some selections, such as an individual chirp of a single bird may be very short in duration. Some selections, such as the trill of a toad may be lengthy in time but very narrow in frequency range. Other sounds, such as the morning chorus or constant drone of background traffic, last for extended periods of time. The time duration of a sound selection is not limited to a ten-second page, but can be extended over multiple pages. As each selection rectangle is drawn, Raven calculates the values for the measurements displayed in the selection table.
By listening to the sound recording, the analyst also identifies each sound by type.
Each sound source is placed into one of three general types (Anthrophony, Biophony, and
Geophony) and within one of several sub-types. The definitions of sound types and sub- types is as follows:
1. Anthrophony – Sounds from human sources
a. Ax - Type 1: Mechanical sounds
b. Av - Type 2: Human voices and music
2. Biophony – Sounds from non-human living organisms
a. Ba – Amphibians
b. Bb – Birds
c. Bi – Insects
d. Bm - Mammals
3. Geophony – Sounds from non-living sources (water, wind, fire, etc.)
168
a. Gp – Precipitation (Rain, snow, hail)
b. Gw – Wind
When possible, in addition to placing the sound into a sound type, the source of the sound is identified in a Notes data field. For example, when listening to the morning chorus of birdsong, the sound analyst identifies and notes individual species of birds singing within the chorus. Identification of individual sound sources within the biophony provides useful information on species diversity. Spectrograms provide a fascinating look at the energy we normally experience as sound. Figures 33a through 33e show examples of spectrograms representing geophony, biophony and anthrophony.
Figure 33a (right): Spectrogram of tree leaves
rustling in the wind – an example of geophony (Gw). Figure 33b (below): Spectrogram of a cicada buzzing (horizontal lines) and a house wren calling (wavy lines), both examples of
biophony (Bi & Bb).
169
Figure 33c: Spectrogram of a cedar waxwing (upper green clouds), a northern cardinal (middle green sweeps) and a mourning dove (lower yellow horizontal lines), examples of biophony
(Bb), and background traffic noise (lower green cloud), an e xample of anthrophony (Ax).
Figure 33d: Spectrogram of human voices, an example of anthrophony (Av).
Figure 33e: Spectrogram of overlapping sound signatures. Selection 494 represents anthrophony type 1 (Ax - jet airplane); Selection 497 represents anthrophony type 2 (Av - human voices); Selections 498 and 499 represent biophony (Bb - chipping sparrow & house sparrow); Selections 500 and 501 represent Geophony (Gp - falling rain).
170
In addition to identifying the sound source, a subjective qualitative evaluation of sound “loudness” or uniqueness is interpreted. As previously discussed, the Raven program measures the average power, peak power, and total energy of each sound sample and reports those quantifiable variables in units of decibels. However, the perceived loudness of a sound is not a definitive measure of the listener’s response to the sound, nor if the listener’s attention is drawn to the sound. Therefore a more qualitative evaluation of the loudness or distractive nature of a sound may be useful. In the same way that sound intensity classes were entered on the field notes sheets during the collection of sound recordings, I listened to each sound selection and placed them into one of four sound intensity classes:
1. Faint sounds – almost imperceptible sounds
2. Background sounds – common elements of the soundscape that do not draw
attention
3. Attention-grabbing sounds – sound of sufficient loudness or of unusual or
unexpected nature that command the attention of a listener
4. Loud and/or irritating sounds – sounds that many (but not necessarily all) listeners
find objectionable
These are subjective classes. A person who is used to the sounds of a near-by airport may find the frequent jets overhead to be background sounds that are seldom noticed, while a person used to the sounds of the rural countryside may find the same airplane sounds objectionable or at least attention-grabbing. For the purpose of this research that focuses on the effect of sounds on directed attention it may be important to distinguish
171 between background sounds that are normally ignored (class 2) and sounds that trigger attention (class 3).
With these preliminary protocols established I decided to select one pair of sound recordings and follow the process of sound analysis from beginning to end, including the detailed sound signature analysis using Raven followed by the statistical analysis of the resultant data. Based on the findings from the “slice-and-dice” analysis that indicate significant differences in the Average Power sound attribute, the September 5, 2014 paired simultaneous recordings from Prindle Ave. and Stratford Rd. were selected for this detailed sound signature analysis. The strength that I brought to manually analyzing the sound recordings is experience in identifying bird song and amphibian calls, that proved useful in properly classifying sounds into the general sound types as well as developing a sense of species diversity and abundance. The weakness I brought to the task was inexperience with using a sound analysis program and the need to develop some consistent and reasonable ground rules to the process.
As previously discussed, the Raven sound analysis program provides measures of a sound signature for loudness, time, frequency and entropy. Raven can be “trained” to recognize sound patterns and automatically identify some sound sources in much the same way that image interpretation programs can be programmed to recognize patterns within a digital image. For this research no automation of identifying sound signatures was used. All interpretation of sound signatures was performed manually by a single analyst.
When manually interpreting a sound recording the sound analyst identifies individual sounds or groups of sounds by using the computer cursor to draw a box or rectangle
172 around the sound signature. Many sound signatures overlap, and some sounds gradually fade into the soundscape and then gradually fade from perception. The question for the analyst becomes “Where is the best place to locate the box in the spectrogram that most accurately captures the sound?” Figures 34a and 34b provide examples of two interpretations of a single sound, that of a passing car. In Figure 34a the selection boundaries extend horizontally along the time scale from the first audible sound of the approaching car, through the increasing loudness of the sound and concludes at the last audible sound of the moving car. The selection boundary extends vertically along the frequency scale from the lowest frequency where energy is portrayed as a bright white color up through the full range of higher frequencies where energy is portrayed in shades of green. This liberal or inclusive interpretation of the sound captures the vast majority of the energy associated with the sound in terms of time and frequency, but a substantial area within the rectangle is dark, indicating that for much of the time duration of the sound little or no energy is present in the mid to upper frequencies. It is only when the car is close to the recorder that these higher frequencies are detected.
Figure 34b presents the same spectrogram but in this example the selection boundaries around the sound signature have been drawn more conservatively or restrictive. In this case the sound analyst has chosen to capture only the frequency range that is dominated by the sound over the duration of the sound. Note that the peak power reading in both interpretations are the same (93.7 dB), but the average power is quite different (64.1 dB vs. 72.5 dB).
A limitation of the sound analysis program is that time and frequency boundaries can only be used to form rectangles around what are often complexly shaped sound
173 signatures. Because the placement of a rectangle around the sound signature greatly influences the calculation of average power, the average power value may not be a reliable variable with which to determine the uniqueness or complexity of a soundscape unless a consistent method of establishing sound selection boundaries is employed.
In the above example, the peak power calculation is consistent in both sound signature interpretations. It may seem preferable to use peak power as a more reliable indicator, but as can be seen in both examples, the peak power is only apparent for a short duration of time while the car is very near, and the peak power is only apparent in the low end of the frequency range. Therefore a restrictive interpretation that only identifies the area of maximum energy ignores much of the energy associated with the sound signature in both time and frequency range. The most accurate method of analyzing a sound signature may be to create a series of small rectangles around parts of the sound signature to capture the majority of the energy without including regions of little or no energy. This approach is somewhat impractical given the time required to draw multiple selections over many hours of recordings.
For the same reasons that the subjective placement of rectangle boundaries around a sound signature affect the calculation of average power, the placement of boundaries affect the calculation of delta time (the duration of a sound signature) and delta frequency
(the frequency range of a sound signature).
174
Figures 34a (above) and 34b (below): Spectrograms of the sound of a passing car. The red line indicates the selection boundaries determined by the analyst to be the duration of the sound (left and right vertical lines) and the frequency range (top and bottom horizontal lines). Sound energy is represented by color – bright colors indicate sounds with much energy (loud); dark colors indicate sounds with little energy (quiet). The position of the selection boundaries influence the calculation of average power. An inclusive interpretation of the sound (above) results in a relatively low average power calculation; a restrictive interpretation of the sound (below) results in a relatively high average power calculation. Peak power is consistent in both selections.
175
Another important question involves the interpretation of distinct but repetitive sounds, such as bird song. Figures 35a and 35b present a spectrogram for a series of bird songs. The choice of whether to identify the sound signature as a single repetitive sound with multiple elements, or as four distinct sounds, influences the average power calculation, time duration, and frequency range. Perhaps an important question is, do most humans perceive a string of birdsongs as individual unique sounds, or as a single complex sound with multiple elements? It may be similar to listening to a song played by an orchestra. Do most humans hear the song as a symphony or as a collection of coordinated individual instruments? For consistency of interpretation of sound signatures
I have adopted the convention that if two similar sounds are within 2 seconds of each other they are identified as a single sound signature. If there is more than a two second separation of quiet between two similar sounds they are identified as individual sound signatures.
176
Figures 35a (above) and 35b (below): Spectrogram of several bird songs including a house wren, cardinal and robin. During the interpretation process they could be gathered into a single selection, as above, or separated into individual selections, as below. The method chosen for gathering or separating sound signatures affects the calculation of average power, time duration, frequency range, and possibly peak power. The convention chosen for this research suggests that if two similar sounds are within 2 seconds of each other they are identified as a single selection. If there is more than a two second separation of quiet between two similar sounds they are identified as individual selections.
177
The exercise of manually interpreting the two simultaneously captured 3-hour recordings provided a sense that the recordings were distinctly different. Perhaps this suspicion was fostered by the slice and dice analysis exercise that revealed statistically significant differences in the average power data throughout most of the frequency ranges. Figures 36a and 36b present spectrograms of the same 10 second time interval as recorded at the Stratford Rd. and Prindle Ave. locations. Within the Stratford Rd. segment are sound signatures of two mechanical sounds (selections 698 & 700), an
American toad (701), wind rustling tree leaves (703) and an overhead airplane (695).
Throughout most of the Stratford Rd. spectrogram there is a relatively bright band in the
7,000 to 9,000 Hz range with a fainter band in the 5,000 to 6,000 Hz range. This represents the nearly constant drone of crickets. The Prindle Ave. spectrogram also has a band in the 7,000 to 9,000 Hz range, but it is much fainter than on the Stratford Rd. recording. This is verified by the Average Power measurements of 39.8 dB and Peak
Power of 85.5 dB from Prindle Ave. compared to the Average Power measurements of
49.6 dB and Peak Power of 85.8 dB from Stratford Rd. It seems that the insects are louder (from a perspective of listening, from visually viewing the spectrograms, and from considering the quantitative data of average and peak power) on Stratford Rd. than on
Prindle Ave. Sound signatures from Prindle Ave. also include a wind chime (selection
575), an American goldfinch (576), background traffic sounds (2), and the same overhead airplane heard in the Stratford Rd. recording.
178
Figure 36a (above) and Figure 36b (below): A spectrogram of stereo channels with associated data table for a ten-second sound segment simultaneously recorded on September 5, 2014 on Stratford Road (many trees) (above) and Prindle Ave. (few trees) (below).
179
Measuring silence.
A soundscape is the collection of individual sounds that occur in a landscape, but periods of quiet are also important elements of a soundscape. The proportion of relatively quiet times and the properties of the background “silence” may be as important to human quality of life as are the sounds themselves. While completing the detailed sound-by-sound analysis of the sound recordings the question arose: “How should periods of silence be accounted for?” Unlike a sound signature that is visible on a spectrogram, a period of silence has no distinct boundaries around which to draw a box.
Therefore the following method was adopted.
Within each sound recording there is usually a constant and fairly distinct band of low frequency energy that is apparent in a spectrogram. This band of energy represents the background sounds that are present that usually originate from distant mechanical sounds such as traffic. While conducting the sound-by-sound analysis, periods of relative silence were also identified. A period of relative silence is defined as at least one second in time where no other discernable sound other than the background sound is heard on the recording or is apparent on the spectrogram. Samples of periods of relative silence were gathered by creating a spectrogram selection within the frequency range of human hearing (20 Hz – 20,000 Hz), and duration of one second. Figures 37 a, b & c illustrate three examples of relative silence. Figure 37a presents a one-second period of relative silence at the Shirra Ct. sound recording site. Note that the “loudness” of the silence as measured by Energy averages 97.2 dB for the two stereo channels. Figure 37b presents a one-second period of relative silence at a location adjacent to the Route 59 freeway in
180
Arlington Heights. Note that the average “loudness” of the silence as measured by
Energy is 106.3 dB. As a point of reference, Figure 37c presents a one-second period of silence as measured at the Seney National Wildlife Refuge in Michigan’s Upper
Peninsula, a location that is relatively isolated from human background sounds. Note that the average “loudness” of the silence as measured by Energy is 77.2 dB.
Figure 37a: A one-second sample of relative silence at the Shirra Ct. recording site in Arlington Heights (Selection 327). Note that the “loudness” of the silence as measured by Energy averages 97.2 dB as indicated by the green color in the low frequency bands.
181
Figure 37b (above): Spectrogram of a one-second sample of relative silence adjacent to the Route 59 freeway in Arlington Heights (Selection 71). Note that the “loudness” of the silence as measured by Energy averages 106.3 dB as indicated by the green and yellow color in the low frequency bands. Figure 37c (below): Relative silence as measured at the Seney National Wildlife Refuge. Energy averages 77.2 dB. Note the absence of energy in the low frequency bands.
182
Selection of sound recording pairs to be analyzed.
The September 5, 2014 simultaneously recorded pair of soundscapes from Stratford
Rd. and Prindle Ave. had a total overlap time of approximately 2 hours and 53 minutes, meaning that a total of 5 hours and 46 minutes for the two sound recordings were analyzed. The process of conducting the detailed sound-by-sound analysis for this sound recording pair using the Raven program took approximately 60 hours to complete. As my interpretation experience using Raven grew I was eventually able to complete a single hour of sound analysis in approximately 6 to 8 hours of time, depending on the complexity of the recording. At this pace, it was estimated that it would take an analyst over a year of full-time work to interpret the approximately 325 hours of sound recordings gathered in Arlington Heights. Future research should explore the options available for automating some of the sound analysis, but for this research it was decided that careful and consistent analysis by a single analyst was important to the objective of controlling as many soundscape analysis variables as possible.
To narrow the collection of sound recordings down to a reasonable and manageable number of hours, a subset of carefully matched sound recording pairs was selected for full analysis. The sound recordings were chosen to provide a balance of temporally matched pairs and spatially matched pairs:
• 14 sound recording pairs.
• 5 spatially-matched tree-removal treatment pairs (many trees then few trees) –
Same exact location, 4 recorded one year apart (less one day to consistently
183
record on the same day of week, and the same hour of day); 1 recorded 13 days
apart. Tree removal took place between recording dates.
• 1 spatially-matched control pair (many trees then many trees) – Same exact
location, recorded one year apart (less one day to consistently record on the same
day of week, and the same hour of day). There was little tree removal in the
immediate area between recording dates, but substantial tree removal within a
quarter mile radius of the site.
• 8 temporally-matched pairs – Simultaneous recording at two locations; one with
many trees and the other with few trees.
Tables VIa and VIb summarize the sound recording pairs selected for analysis. The total recording time for the spatially matched sound recording pairs is approximately 17 hours and 50 minutes; the total recording time for the temporally matched sound recording pairs is approximately 19 hours and 35 minutes. Approximately 280 hours of time was required to conduct the sound-by-sound analysis of these pairs of recordings.
Throughout this dissertation tables are presented that list pairings of sound recordings.
To facilitate the interpretation of these tables green highlighting is used to identify sound recordings collected in areas of many trees and blue highlighting is used to identify sound recordings collected in areas with few trees. An easy way to remember this is if a person looks up in an area of many trees they will see the green canopy overhead, and if a person looks up in an area of few trees they will see the blue of the sky.
184
Table VIa: Spatially matched sound recording pairs (same location; before-and-after dates). Green highlighted dates represent pre-tree removal (light green represents no tree removal in immediate area); blue highlighted dates represent post-tree removal.
Table VIb: Temporally matched sound recording pairs (Same time; different locations for many trees and few trees). Green highlighted dates represent sites with many trees; blue highlighted dates represent sites with few trees. Dark blue represents recording collected adjacent to freeway.
185
Slice-and-Dice analysis of sound recording pairs.
As with the analysis of the first sound recording pair (September 5, 2014 Prindle Ave. and Stratford Rd.), it was useful to conduct an initial slice-and-dice analysis to address the question “Are the sound recordings significantly different” using an objective and comprehensive method. As previously discussed, the initial Table VII: Frequency ranges (Hz) of one slice-and-dice method divided the frequency range into uniform third octaves used in slice-and-dice 2,000 Hz slices and divided time into 30 second dices. analysis.
Communication with Dr. Dean Hawthorne (2014) of the Cornell
Lab of Ornithology revealed that an alternate method could be used that would more closely match sound analysis industry standards (American National Standards Institute, 2004). Instead of using uniform 2,000 Hz frequency slices, the frequency bands are determined with boundaries that define one-third octave intervals. A sound octave is the familiar difference in sound between the middle C note on a piano and the C note 8 white piano keys to the left or right of middle C. Sound octave frequencies progress along a logarithmic scale, and therefore the third-octave frequency ranges are not uniform in width as measured in Hz. The advantage to using the third-octave slices in the slice-and-dice analysis is that these frequency ranges more closely resemble the pattern in which the human brain processes sound energy than a series of uniformly spaced frequency ranges.
186
Table VII lists the frequency ranges used in the third-octave slice-and-dice analysis.
Figure 38 shows a spectrogram with the third octave slice intervals displayed.
Characteristics of the third octave slice-and-dice analysis include:
• 32 one-third octave frequency slices
• One hour recordings diced into 120 30-second intervals
• 2-channels (stereo recording)
• 120 intervals x 32 frequency bands x 2 channels = 7,680 cells
• Advantage: Complete coverage with no overlap of cells
• Disadvantage: No information on sound types
Figure 38: Spectrogram of stereo channels including the one third octave frequency slices.
As will be discussed in the Results section of this dissertation, the one third octave slice-and-dice analysis of each of the fourteen pairs of sound recordings revealed varying
187 degrees of statistically significant differences in measures of sound intensity within each of the recording pairs.
The Raven program calculates three measures of sound intensity for each slice-and- dice cell: Average Power (dB), Peak Power (dB), and Energy (dB). Energy is a sum of all sound energy within a sound selection, Peak Power is the measure of maximum intensity (the moment when the sound is loudest), and Average Power is the sum of all energy divided by time. Energy is the variable that is compared in the one third octave slice-and-dice analysis.
Statistical Analysis
One-third octave slice-and-dice t-Tests.
Following the completion of the one third octave slice-and-dice analysis in Raven, each Raven selection table is imported into Excel. The data are sorted by third octave frequency slices. For each of the 32 frequency slices, a t-Test is conducted that compares the Energy data for a matched sound recording pair. Alpha for the t-Tests is 0.05. As will be discussed in the Results section of this dissertation, the 32 t-Tests that compare
Energy over the one third octave frequency ranges provide a compelling picture of the differences, and in some cases the similarities between the matched sound recording pairs. Figure 39 displays a summary graphic of the third octave slice-and-dice analysis for the sound recording pair of Prindle Ave. and Stratford Rd. recorded on September 5,
2014.
188
Figure 39: Summary graphic of the third octave slice-and-dice analysis for the sound recording pair of Prindle Ave. and Stratford Rd. recorded on September 5, 2014. Cells shaded in green represent frequency ranges where Energy is significantly higher at Stratford Rd. where there are many trees. Cells shaded in blue represent frequency ranges where Energy is significantly higher at Prindle Ave. where there are few trees. Cells shaded in tan represent frequency ranges where differences in Energy between Stratford Rd. and Prindle Ave. are not statistically significant. The grey bar between slices 21 and 22 represents the approximate boundary below which are frequently heard background sounds originating from vehicle traffic (anthrophony) and above which are frequently heard bird songs and other elements of biophony. In this example the t-Test results suggest that there is greater sound energy in the frequency ranges associated with human background sounds in the location with few trees and greater sound energy in the frequency ranges associated with biophony in the location with many trees.
The slice-and-dice analysis is a useful first step in understanding how soundscapes are different is some aspects and similar in others. The advantage of the slice-and-dice method is that it creates uniform segments in time and frequency range, of which there is no overlap, and compares one measure of sound intensity. This method is objective and is not influenced by analyst bias. The disadvantage of the method is that it reduces a soundscape to the equivalent of pixels in a photo but reveals little about the sources of sounds or of the properties of individual sound signatures.
189
Sound signature t-Tests.
As previously discussed, the manual sound-by-sound analysis performed by a human analyst also has some disadvantages. The manner in which an analyst chooses to draw a sound selection box around a sound signature affects some (but not all) of the sound attribute measures. There is frequent overlap among sound signatures, for example the sound of a bird singing at the same time a lawn mower is being used creates a greater measure of sound intensity in both selections than would be measured without overlap.
There are however many advantages to using a thorough sound-by-sound identification method, particularly if the interpretation protocols are applied consistently. Since measures of time and frequency are not held constant, as in the slice-and-dice method, these two variables become additional descriptors of individual sound signatures.
Perhaps the greatest value of carefully listening to and comparing pairs of soundscapes is the ability to classify individual sounds into groups of anthrophony, biophony, and geophony. Understanding the proportion and intensity of human-sourced sounds compared to those sounds associated with nature is crucial to understanding the effect of sounds on human directed attention.
As previously discussed, the Raven program provides a wide range of measures for sound selections. These sound attributes can be organized into five groups:
1. Loudness (Intensity)
2. Time
3. Frequency
4. Complexity / Entropy
190
5. Sound source
Within each of these groups there are multiple measures. For example measures of sound intensity include average power, peak power, and energy (see section “Choosing sound measurement options within Raven” for a description of all of the sound measures provided). Each of these measures can be compared between the sound signatures of a pair of recordings. It is important to select the appropriate statistical tool when making comparisons. In this case, a single individual sound attribute measure was always compared between two paired recordings. For this application, the Student’s t-Test is the appropriate tool for statistical comparison. Sound attribute measures were not compared among more than two recordings, in which case a different statistical analysis tool such as Analysis of Variance (ANOVA) would have been a more appropriate statistical analysis tool.
When repeatedly using the t-Test method, there is concern about diluting the confidence interval by repeatedly re-introducing the chance for a type 1 error; that is the incorrect rejection of a null hypothesis (accepting a false hypothesis as correct). This is a problem when repeatedly using a datum to compare to against multiple data from other sets. In this case, the data from one sound recording were compared to only one other recording, and the data were independent.
The completion matrix for t-Tests between sound recordings is illustrated in Table
VIII. Figure 40 presents the summarized results from the set of t-Tests comparing the mean values for sound attributes for the paired sound recording from Stratford Rd. and
Prindle Ave. on September 5, 2014. The results from each of the t-Test sets for paired sound recordings are presented and discussed in the Results section.
191
Table VIII: t-Test matrix for combinations of sound types and sound attributes. For each matched pair of sound recordings the means for each sound attribute (delta time, delta frequency, center frequency, etc.) were compared using t-Tests for each sound type (anthrophony, biophony and geophony) and each sub-type (human mechanical, human voices, birds, wind, etc.). Since samples of silence were collected using fixed time and frequency ranges, t-Tests involving the attributes of delta time and delta frequency were not applicable.
Figure 40: t-Test summary for sound recording pair of Stratford Rd. and Prindle Ave. on September 5, 2014. Blue cells represent sound attributes with statistically significant greater mean values at Prindle Ave. (few trees). Green cells represent sound attributes with statistically significant greater mean values at Stratford Rd. (many trees). Tan cells are silence measures that are not applicable. White cells represent measures that show no significant difference in means.
192
Proportional statistics and graphs.
F-Tests provide a comparison of variance values between two sets of sound attribute data, and t-Tests provide a comparison of mean values. There is also useful insight to be gained by considering the proportion of total values for sound attributes. For example comparing the total time during which bird song was observed during each of two paired sound recordings provides insight on the amount of birds in each location. Therefore the total values for the following sound attributes were calculated and presented in summary graphs:
Delta Time
Delta Frequency
Energy (dB)
Energy (mV)
Aggregate Entropy
Figure 41 presents the sum of Delta Time values for each of the sound categories and sub-categories for the September 5, 2014 simultaneous sound recordings gathered at
Prindle Ave. (few trees) and Stratford Rd. (many trees). These proportional graphics are useful in understanding the relative strength of sound attributes when comparing a matched pair of sound recordings. The complete set of proportional statistics and graphs for each of the paired sets of sound recordings is discussed in the Results section.
193
Figure 41: Sum of the Delta Time sound attribute values for Stratford Rd. (green - many trees) and Prindle Ave (blue - few trees) recorded on September 5, 2014.
What Was Learned During the Sound Analysis Process?
A large amount of quantitative data was generated during the sound analysis process that was used to document changes in the soundscapes as tree canopy changed. The process of completing the sound analysis data collection also provided valuable experience and insight for the analyst.
The sound attributes reported by the Raven sound analysis program are all useful in understanding the components of sound energy, however some of these attributes are
194 more relevant to understanding how sound affects humans than others. Since this study ultimately explores how sounds affect directed attention, a subjective ordinal score was recorded for each sound’s potential for attracting a listener’s attention (see the section titled “Field notes” in Chapter 5). The sound analysis exercise shed additional light (or more appropriately sound) on which sound attributes are most important when considering their potential to disrupt directed attention.
The loudness of sounds is obviously important, as is their duration. Loud and sudden sounds may grab a person’s attention more than loud and lengthy sounds. The attributes that are relevant to loudness and duration include measures of power and energy. The pitch of a sound may be important, but not all humans can hear the full range of sounds between 20 Hz and 20,000 Hz, so sounds at the low and high ends of this frequency range may affect people differently. Frequency (pitch) and entropy (related to complexity) may be important when considered in tandem. The varying pitch and complexity of an ambulance siren is obviously designed to grab a person’s attention.
Another important aspect of sounds is the proximity of the sound source to the listener. Consider a train whistle. If the brain perceives the sound as distant, and therefore not dangerous, the sound may be pleasing and calming. But up close the sound may cause a sense of alarm and danger. Consider also the quiet whine of a mosquito.
Even such a faint sound next to your ear will certainly draw immediate attention. The attributes related to power and energy, frequency range, and intensity class are relevant.
Uniqueness, and conversely the commonness of sounds may be very important to human perception. The human brain can build filters for sounds and quickly come to ignore non-threatening sounds that are very common. Consider how some people who
195 live near airports report hardly noticing the loud sounds of airplanes overhead. On the other hand, a sound that is unique to a listener will draw immediate attention. Consider how a person interested in birds will notice a bird call that they are not familiar with, even if it occurs in a chorus of familiar bird song.
Lastly, it seems that the quality of a soundscape and whether that soundscape is beneficial or detrimental to human directed attention is related to the soundscape’s ability to evoke wonder or worry. A soundscape full of attention-grabbing sounds that warn of potential danger, such as walking next to a street full of fast-moving vehicles will, for many people, cause a rise in directed attention. Over time directed attention may become fatigued. Conversely, a soundscape without frequent alarming sounds, but rich in pleasing sounds – the type of sounds that reassure one that danger is not near – may refresh a fatigued person.
In summary, the experience of conducting sound analysis on a sound-by-sound basis leads to the recommendation of conducting additional research on the composition of soundscapes relative to the proportion of sounds thought to be disruptive to attention, and to use recorded sounds with disparate characteristics of disruptiveness during tests for directed attention.
196
CHAPTER VII
RESULTS PART 1 – ANALYSIS OF SOUND RECORDNGS
Introduction
The question “Are these two sounds similar or different?” seems easy enough. We can ask the same question about two apples, and the task would also seem simple. But the answer to the apple question might depend on what particular attribute of the apples you are comparing. Both apples may be red and be the same weight, but one apple may be sweet while the other is tart. One may be juicy while the other is dry. One may be oblong-shaped while the other is a somewhat flattened sphere. One may be a Red
Delicious variety and the other a McIntosh. And when you begin to consider Yellow
Delicious and Gala and Granny Smith varieties, the answer to the question “Are they similar or different?’ becomes “It depends on what attribute you compare.”
So it is with comparing sounds. Sounds have many attributes including loudness
(intensity), pitch (frequency), duration and complexity (entropy) with which they can be described, and within each of these attributes there are multiple methods of measure and description.
197
The objective of this research is to compare collections of sounds within two frames of reference. The first frame is that of changing landscapes, and in particular changing amounts of tree canopy within urban landscapes. The second frame is that of human behavior, and in particular how sounds affect our ability to pay attention.
There are four broad potential outcomes of the research:
1. Tree canopy does not affect sound and sound does not affect directed attention
2. Tree canopy does affect sound but sound does not affect directed attention
3. Tree canopy does not affect sound but sound does affect directed attention
4. Tree canopy does affect sound and sound does affect directed attention
Given what previous research has revealed about soundscapes and their relation to landscapes, it is unlikely that a substantial change in tree canopy would not result in undetectable changes to the collection of sounds. This of course leads to additional questions including “How much change in tree canopy must take place before changes in the soundscape are detectable?” and “Which attributes of sound are likely to change and which are less likely to change?” If we suspect that there will be some measurable change to the soundscape as tree canopy changes, then the task becomes one of measuring which sound attributes change, and by how much.
If some aspects of the soundscape do indeed change as tree canopy changes, that leaves us with potential outcomes 2 and 4 – either these changes in the soundscape do affect human directed attention or they do not. The results discussed in this chapter of the dissertation address differences in soundscapes associated with changing tree canopy cover. Results concerning the relationship between sounds and directed attention are discussed in part 2 of this dissertation.
198
Comparison of Sound Recording Pairs
As discussed in Methods Part 1b a subset of the 99 recording events totaling 37.5 hours of sound recordings in Arlington Heights, Illinois were selected for detailed analysis using the Raven program from the Cornell Ornithology Lab. These recordings were chosen in pairs so as to control as much as possible the variables such as season, day of the week, time of the day, and weather conditions that may affect soundscapes while varying the level of tree canopy cover. Multiple measures, both quantitative and qualitative, were gathered and compared to document the extent and type of change to soundscapes, if any, are attributable to loss of tree canopy.
As described in section Methods Part 1b, the sound recording pair collected on
September 5, 2014 at the Prindle Ave. (few trees) and Stratford Rd. (many trees) locations was the first to be analyzed and the first to undergo statistical analysis. This initial process led to refinement in analysis methods for the remaining sound recording pairs as well as revisiting the initial recording pair for additional analysis.
Part 2 of this dissertation focuses on the effects of sounds on measures of directed attention. As will be described in the Methods section of Part 2, the sound recording pair that was simultaneously recorded from 5:33 AM to 8:33 AM on Tuesday May 19, 2014 at the Prindle Ave. (few trees) and Stratford Rd. (many trees) locations are two of the recordings played for participants during the tests for directed attention. The results from each step in the analysis of this pair of recordings are discussed in detail in this section.
The results from the statistical analysis from the other 13 pairs of sound recordings are discussed collectively and summarized in this section.
199
Third octave slice-and-dice analysis.
As discussed in the Methods Part 1b section, the sound recordings were first analyzed using an arbitrary (non-sound source specific) method of slicing each sound recording along one-third octave frequency boundaries and 30 second time intervals for a one-hour sample of each recording. This method creates a comprehensive grid of non-overlapping selections. The 120 selections within each of the 32 third octave frequency ranges for each of the two sound recording pairs are first analyzed using the Excel F-Test tool to determine if the variances within each frequency range are statistically different (alpha =
0.05). The results of the F-Test then determine the appropriate t-Test (for equal or unequal variances) to use for testing mean values within each frequency range.
The slice-and-dice method holds the time range and frequency range constant for each selection cell. Therefore it is unnecessary to conduct tests for variance or mean values on any of the sound attributes associated with time or frequency. This leaves the variables associated with sound intensity and entropy as candidates for comparison. For the F-
Tests and t-Tests to compare basic differences between recordings, the Energy measure of sound intensity was used. The Energy measure of sound intensity reports on all energy detected within a selection as opposed to the Peak Power and Average Power measurements. Since the frequency slices and time dices cover the complete ranges of those variables, the Energy attribute is consistent with an inclusive approach.
Figure 42a and 42b present the one-third octave slice-and-dice analysis summary for temporally matched and spatially matched recording pairs. Green cells indicate frequency bands where mean Energy (dB) is significantly greater at sites with many
200 trees; blue cells indicate greater Energy (dB) at sites with few trees; tan cells indicate no significant difference in mean Energy (dB).
The grey bar between slices 21 and 22 in Figure 42a and 42b represents the approximate boundary below which are frequently heard background sounds originating from vehicle traffic (anthrophony) and above which are frequently heard bird songs and other elements of biophony.
Two special cases are worthy of note in the one-third octave summaries. One spatially matched pair was recorded at Suffield Drive on April 13, 2014 and again on April 12,
2015. Few trees were removed in the immediate area of the recording site between these dates, however many trees were removed within a radius of a quarter mile of the site.
Therefore in Figure 42a two shades of green are presented for this recording pair to indicate many trees are still present in the immediate area but fewer trees are present on a larger scale. One temporally matched pair was recorded on May 6, 2015. One location was the often-recorded site on Huron St. The second location was on Wilke Rd. adjacent to the Route 53 freeway. This pairing was not chosen as an example of a site with many trees compared to a site with few trees, but rather as a comparison of one site close to a source of many traffic sounds and another site more distant from the freeway. The recording date was after many trees had been removed from the neighborhood, therefore the Huron St. site is represented in blue in Figure 42b. Since the Wilke Rd. site has much less overall tree canopy than does Huron St. it is shown in a shade of grey.
201
Figure 42a (above) and 42b (below): One-third octave slice-and-dice analysis summary for spatially matched (above) and temporally matched (below) recording pairs. Green cells indicate frequency bands where mean Energy (dB) is significantly greater at sites with many trees; blue cells indicate greater Energy (dB) at sites with few trees; grey cells indicant greater energy at the site adjacent to a freeway; tan cells indicate no significant difference in mean Energy (dB).
202
The results of the one-third octave slice-and- dice analysis clearly show differences in measured Energy (dB) between recordings in the matched pairs. But the differences are not always consistent. How do these actual results compare with might be predicted based on my field observations? Before conducting the slice- and-dice analysis it was anticipated that a loss of tree canopy and the associated decrease in noise attenuation would result in more sound energy present in the lower frequency ranges (100 Hz –
2,000 Hz) – those frequencies most closely associated with human mechanical sounds such as traffic noise. It was also anticipated that in Figure 43: Predicted (not actual) results for slice-and-dice analysis for some, but not necessarily all seasons, there may Energy (dB) in areas with many trees be an increase in sound energy present in the mid and few trees. frequency ranges (2,000 Hz – 10,000 Hz) – those frequencies most closely associated with bird song and insect calls. The highest frequency ranges (> 10,000 Hz) could exhibit higher energy in areas with many trees since there would be more foliage to rustle in the wind (greater geophony), or conversely there may be more energy in areas with few trees since noise attenuation associated with trees is mostly in higher frequency ranges (resulting in greater detection of anthrophony). There would be few differences expected at the very lowest frequency ranges (< 100 Hz) due to the low-cut filters in the
203 sound recorders that are designed to eliminate wind-on-microphone distortions. Figure
43 presents one prediction of possible slice-and-dice analysis results based on analyst expectations, not on actual data from sound recordings.
The temporally matched pair of sound recordings from September 5, 2014 recorded at
Prindle Ave. (few trees) and Stratford Rd. (many trees) closely resembles the predicted results with more energy in lower frequency bands from the site with few trees and more energy in mid to high frequency bands from the site with many trees. The spatially matched pair recorded at Prindle Ave, on April 12, 2014 (many trees) and April 11, 2015
(few trees) unexpectedly reverses the predicted results with more mid to high-frequency energy after trees were removed. One explanation for this might be that while both days were sunny and calm, the warmer temperature for the 2015 recording day (low 60°’s F
(low teens oC) as compared to low 50°’s F (mid-teens oC) for 2014) may have been more favorable for birds to be singing than one year earlier. However this is only speculation since the slice-and-dice analysis provides no information about sound sources.
The trend from the spatially matched pairs suggests higher mean values for the single sound attribute of energy after trees were removed, particularly in the mid to high- frequency ranges. The trend for temporally matched pairs suggests something different relative to tree canopy – higher mean values for energy in areas with more trees.
What plausible explanations might there be for this apparent contradiction? One explanation might be that the spatially matched recording pairs are more susceptible to weather-related differences that could affect energy measured in frequency ranges typically associated with biophony and geophony. Since the temporally matched pairs were recorded simultaneously, weather variables are controlled.
204
Another explanation might be that the spatially matched pairs of recordings are better at detecting changes in noise attenuation due to the change in tree canopy at that exact location, while the temporally matched pairs of recordings are better at detecting relative concentrations of biophony attributable to tree canopy and not subject to differences in weather conditions. It is possible that there is a reduction of bird numbers at a particular location as tree cover is removed, therefore a decrease in bird song is detected in a spatially matched pair of recordings. It is possible that birds from one location migrate to an adjoining area with greater tree cover after trees are removed resulting in greater birdsong detected in the area of greater tree cover during temporally matched recordings.
The danger of speculating the cause and effect while considering the slice-and-dice analysis is that there is no information gathered on sound source (and therefore the identification of anthrophony, biophony, and geophony) in this exercise.
The slice-and-dice analyses for the spatially matched pair of recordings at Suffield Dr. are interesting to consider. The analyses indicate that there is more energy in the low frequency range for the initial 2014 recording, and more energy in the mid to high frequency range for the 2015 recording. Why might this be? There was virtually no change in tree canopy in the immediate area but considerable loss of tree canopy in the larger neighborhood. If noise attenuation was primarily a function of the vegetation in the immediate area (perhaps less than 100 meter radius) then no difference in the recordings would be expected. If noise attenuation was primarily a function of the vegetation present over a larger scale (perhaps greater than 100 meter radius) then an increase in energy would be expected, which is what was observed. Did, weather conditions play a role in the soundscape for the two dates? It was 10° F to 15° F (5° C to
205
8° C) colder during the Sunday April 12, 2015 recording than on the Sunday April 13,
2014 date. Colder temperature would be expected to lead to less sound from singing birds, however more sound energy was detected in all but the lowest frequency ranges in
2015. Understanding the possible changes in sound energy relative to sound source will have to wait until the sound-by-sound analysis is conducted.
There was no surprise to the results from the slice-and-dice analysis for the simultaneous recordings on May 6, 2015 from Huron St. and Wilke Rd. It was fully expected that sound energy would be significantly higher next to the freeway and it certainly was.
The results from the slice-and-dice analysis are intriguing to consider, but they reveal little about the actual sound sources. The objective of determining if there are significant differences in recording pairs was fulfilled. The analysis of individual sounds for a much greater number of sound attributes is a better tool for understanding changes in anthrophony, biophony and geophony.
Evaluating the method of analysis: The process of adding and subtracting
decibels.
There are two methods of conducting t-Tests to compare the means of measures of sound intensity reported in decibels. The first method is to use the decibel values for
Average Power, Peak Power, or Energy to calculate the mean and variance for each group of data (such as slices of frequency bands). This method was used during the initial slice-and-dice analyses but is technically incorrect since the mean values are
206 calculated by adding the individual decibel values. As previously discussed in Methods
Part 1b (see the section titled “Conversion of measures of sound intensity from decibels
(dB) to millivolts (mV)”) the decibel scale is logarithmic so adding and subtracting decibel values first requires converting the values to a non-logarithmic scale, performing the arithmetic and then converting the sum or the difference back to the logarithmic decibel scale.
Table IX presents the mean Energy values for each of the 32 third octave frequency bands calculated from logarithmic decibel values and from logarithmic values first converted to non-logarithmic values before conversion back to decibels. What does this added step of conversion to non-logarithmic values gain? From a purely mathematical perspective the conversion is necessary because it is incorrect to simply add or subtract logarithmic values such as decibels. At first glance the mean values between the two methods seem similar, however the fact that an increase of 3 decibels represents a doubling of power demonstrates that a small difference in decibel values can translate into a large difference in sound energy.
The most striking difference in the results of the two methods of dealing with decibel values is apparent when subtracting decibels. The statement “a decrease of 3 decibels” is not the same as the statement “a difference of 3 decibels.” For example in Table IX, in the frequency range of 14.1 Hz to 17.8 Hz the decibel value for mean Energy at Wilke
Rd. is 103.70, and at Huron St. the decibel value is 93.70. While there is a decrease of
10 decibels in sound energy when moving from Wilke Rd. to Huron St., it is not correct to say that the difference of energy values is 10 dB. By first converting the logarithmic decibel values to non-logarithmic millivolt values prior to subtraction and then re-
207 converting the result back to decibels reveals that there is actually a difference of 104.40 decibels. From a mathematical perspective of subtracting values, this is the correct answer.
Table IX: Mean Energy values for each of the 32 third octave frequency bands calculated from logarithmic decibel (dB) values and from logarithmic values first converted to non-logarithmic millivolt (mV) values before conversion back to decibels.
208
However from the perspective of human hearing, the difference value calculated by first converting decibels to millivolts prior to subtraction followed by converting the difference in millivolts back to decibels may seem less intuitive than directly comparing the logarithmic values. The decibel scale is useful when measuring sound energy because it more closely resembles human perception of sound. Most humans perceive a
10 fold increase of sound pressure level (an increase of 10 dB) to result in a doubling of a sound’s “loudness.” Therefore a decrease of 10 decibels moving from Wilke Rd. near the freeway to Huron St. (approximately 1,500 feet east of the freeway) would be perceived by most humans as moving to an area that is half as noisy (or twice as quiet).
Since the method of directly adding and subtracting decibel levels is incorrect from a mathematical perspective, the preferred method of first converting all of the individual decibel values to non-logarithmic values before calculating the means was adopted for use in the t-Tests going forward.
The slice-and-dice analysis of the sound recording pairs revealed significant differences in most of the frequency ranges when considering sound energy. The fact that the differences were not always consistent when considering tree canopy cover led to a series of interesting questions that would best be answered through a different method of analysis, namely the identification and measurement of individual sounds.
Manual identification of individual sounds and statistical analysis.
As described in the methods section, the slice-and-dice analysis was followed by the more detailed manual analysis of the sound recordings in which the analyst identifies the
209 source of each sound and the Raven program calculates multiple sound attribute measures for each sound. After completing the manual analysis of each of the two recordings in a matched pair, the Raven selection tables were saved as Excel spreadsheets. Then the statistical analysis tools in Excel were used to perform a series of analyses.
As is illustrated in Figure 29, each sound that is represented in a spectrogram has several variables that are graphically depicted:
1. The duration of time is determined by the vertical sides of the sound selection
rectangle (time is represented on the horizontal x axis).
2. The frequency range is determined by the horizontal sides of the sound selection
rectangle (frequency is represented on the vertical y axis).
3. The sound intensity (related to the perception of “loudness”) is represented by the
range of colors in the sound selection rectangle.
An additional variable that is apparent within a sound signature is the pattern of energy within the frequency range and over time. Some sounds like the rustling of leaves in a tree are non-descript and appear as a cloud or fog on the spectrogram over a broad range of frequencies. Other sounds like the call of a cardinal are very distinct and appear as an easily recognizable pattern of rising and falling lines within a relatively narrow range of frequencies. Entropy is the sound attribute within Raven that reports on the amount of disorder of energy within a sound signature, which is related to (but not synonymous with) the complexity of patterns within a sound signature.
Measures of each of these sound attributes were analyzed to explore the differences and similarities of two sound recordings in each of the matched pairs. Analysis of sound
210 recordings was conducted first by organizing sounds according to the following sound source types and sub-types:
All sounds (anthrophony, biophony & geophony)
o All anthrophony
. Anthrophony human voices & music (Av)
. Anthrophony mechanical (Ax)
o All Biophony
. Biophony amphibians (Ba)
. Biophony birds (Bb)
. Biophony insects (Bi)
. Biophony mammals (Bm)
o All Geophony
. Geophony wind (Gw)
. Geophony precipitation (Gp)
Then measures for multiple sound attributes related to time, frequency, intensity, and entropy are analyzed for each sound type and sub-type. These measures of sound attributes were considered in sum and as means.
One disadvantage of the sound-by-sound analysis method is that many (not all) sound signatures overlap, and a sound signature for a bird song for example may overlap with a sound signature for falling rain. Careful placement of selection boundaries is the best method of minimizing this problem.
211
Plotting the data.
The first attempt to compare the data for the sound recordings made in locations with many trees with the data for sound recordings made in locations with few trees was simply to sum the sound attribute data for Delta Time, Delta Frequency, Energy reported in decibels, Energy converted to millivolts, and Aggregate Entropy by sound type and sub-type. This was done for the six spatially matched sound recordings and also for the eight temporally matched sound recordings. Figures 44a1 through 44e2 present graphic comparisons for those data.
Figure 44a1 through 44e2: Sums of the sound attribute data for Delta Time, Delta Frequency, Energy reported in decibels, Energy converted to millivolts, and Aggregate Entropy by sound type and sub-type for spatially matched and temporally matched sound recordings. Green bars represent sound recordings made at sites with many trees; Blue bars represent sound recordings made at sites with few trees.
Figure 44a1 Figure 44a2
Figure 44b1 Figure 44b2
212
Figure 44a1 through 44e2 (continued): Sums of the sound attribute data for Delta Time, Delta Frequency, Energy reported in decibels, Energy converted to millivolts, an Aggregate Entropy by sound type and sub-type for spatially matched and temporally matched sound recordings. Green bars represent sound recordings made at sites with many trees; Blue bars represent sound recordings made at sites with few trees.
Figure 44c1 Figure 44c2
Figure 44d1 Figure 44d2
Figure 44e1 Figure 44e2
The data plots reveal some interesting trends but they are simply plots of sums; they do not reveal any statistically significant differences between recording sites, sound attributes or sound types. Never-the-less they provide a first glimpse at similarities and differences between groups of data. When considering the sound attribute of time
(Figures 44a1 and 44a2) it is apparent for the spatially matched recording pairs (the recordings were made at the same location at two different times) and for the temporally
213 matched recording pairs (the recordings were made at the same time in two different locations) that human sourced sounds (Anthrophony), and in particular mechanical sounds (Ax) occupy more time in the soundscapes than any other sound type (biophony or geophony). Mechanical sounds are present for more time in sites with few tees than at sites with many trees. This is true for both the spatially and temporally matched recording pairs. Within the biophony sub-types, the amount of time for birdsong was fairly close for both sets of recordings. Insects were detected for more time in the sites with more trees for the temporally matched pairs of recordings while the amphibians were detected for more time in the sites with few trees. This may be due to the fact that the Prindle Ave. site with few trees was slightly closer to a vernal pond with loudly calling Western Chorus Frogs (Pseudacris triseriata) during some of the spring recordings. It is not surprising that the small amount of time detected for geophony sounds (primarily wind rustling tree leaves) is greater at the sites with many trees.
When considering the sound attribute of frequency (Figures 44b1 and 44b2) it is apparent for the sound recording pairs matched by space that human sourced mechanical sounds (Ax) occupy a greater range of frequencies at the sites with few trees, but this trend is reversed when considering the temporally matched pairs of recordings. Within the biophony types, and in particular for birds (Bb), the total (remember, this is not an average value – mean values will be compared later) range of frequencies is greater at sites with few trees. This is true for both the spatially and temporally matched recording pairs. It is not surprising that the frequency range summed for geophony sounds
(primarily wind rustling tree leaves) is greater at the sites with many trees. This is
214 probably due to a greater abundance of trees near the recorder than an actual difference in the frequency of sound of rustling leaves being different.
The sound attribute of energy is reported here in two units. As previously discussed sound intensity (also called spectral power density or energy) is reported in decibels, which is a logarithmic scale. In order to add units of energy it is appropriate to first convert decibels to a non-logarithmic scale such as millivolts. When considering energy reported in millivolts it is apparent that the majority of sound energy in the soundscapes is associated with human mechanical sounds (Ax). It is interesting however to compare the summed energy from the spatially matched recording pairs with that from the temporally matched pairs. There seems to be greater sound energy from mechanisms at the sites with few trees for the spatially matched recording pairs (Figure 44c1) and greater sound energy from mechanisms at the sites with many trees for the temporally matched recording pairs (Figure 44c2).
There are also some confusing findings when considering measures of energy for animal-sourced sounds (biophony). When considering energy reported in decibels for the spatially matched pairs of recordings it appears that there is more sound energy from animals at sites with few trees (Figure 44d1), but when considering the temporally matched pairs of recordings (Figure 44d2) the opposite may be true. Once again, no statistical significance is implied in these graphs – this exercise simply tallies data.
Figures 44e1 and 44e2 present the sums of the interesting sound attribute aggregate entropy. In many (but not all) of the sound sub-types for anthrophony and biophony, entropy appears to be greater at the sites with few trees.
215
Proportional analysis.
The plots of the summed values of several sound attributes are interesting to compare but they do not provide information on the statistical significance of differences. Does the proportional mix of anthrophony, biophony, and geophony within a soundscape change as trees are removed? To answer that question a statistical comparison of sound type and sub-type proportions using Z scores was performed that compares sound recordings at sites with many trees to sound recordings at sites with few trees.
The proportional summaries using Z scores as well as the numerous t-Tests for multiple sound attributes were completed for all 14 pairs of sound recordings that were analyzed. This results in a considerable number of individual graphs and tables
(approximately 1,800) that would require over 200 printed pages to present in a readable format. Rather than attempt to present all of the results of these analyses within the covers of this dissertation it is more reasonable to present specific examples of the results for one of the recording pairs, and accompany this representative sample of the results with summaries of the findings for the complete set of sound recording pairs. The pair of recordings chosen as a representative example for the larger group of recordings is the temporally matched pair recorded on May 19, 2015 at Prindle Ave. (few trees) and
Stratford Rd. (many trees). This pair of recordings will ultimately be used in the tests for directed attention in Part 2 of this dissertation and therefore is the most relevant pair of recordings to illustrate here. Figures 45a through 45e present proportional summaries for this pair of recordings. The complete set of data from the slice-and-dice analysis, from the sound-by-sound analysis, and from the statistical analysis of these data are available in digital format upon request.
216
Figures 45a through 45e: Proportional summaries for the sound attributes Delta Time, Delta Frequency, Energy (dB), Energy (mV), and Aggregate Entropy for the temporally matched sound recording pair gathered on May 19, 2015 at Stratford Rd. and Prindle Ave. Green cells and stars indicate statistically significant (alpha = 0.05) greater proportions of sound types at Stratford Rd. (many trees); Blue cells and stars indicate statistically significant (alpha = 0.05) greater proportions of sound types at Prindle Ave. (few trees).
Fig 45a
Fig 45b
Fig 45c
217
Figures 45a through 45e (continued): Proportional summaries for the sound attributes Delta Time, Delta Frequency, Energy (dB), Energy (mV), and Aggregate Entropy for the temporally matched sound recording pair gathered on May 19, 2015 at Stratford Rd. and Prindle Ave. Green stars indicate statistically significant greater proportions of sound types at Stratford Rd. (many trees); Blue stars indicate statistically significant greater proportions of sound types at Prindle Ave. (few trees). (alpha = 0.05)
Fig 45d
Fig 45e
Figures 45a through 45e present some interesting findings relative to the proportions of several sound attributes for the pair of recordings gathered on May 19, 2015 at Stratford
Rd. and Prindle Ave. The proportion of time that sound is detected for human mechanical sounds (Ax) is significantly greater (alpha = 0.05) at Prindle Ave., (the site with fewer trees) (Figure 45a). Since there were very few human voices (Av) detected at either site, the difference in mechanical sounds is largely responsible for the significant difference in total time when considering all human-sourced sounds together (A all). No other sound types or sub-types had significantly different proportions for delta time.
218
Figure 45b shows that the proportion of frequency range that sound is detected for bird songs and calls (Bb) is significantly greater (alpha = 0.05) at Prindle Ave., (the site with fewer trees) than at Stratford Rd. (the site with more trees). Since there were very few mammals (Bm) detected at either site, the difference in bird sounds is largely responsible for the significant difference in frequency range when considering all animal- sourced sounds together (B all). No other sound types or sub-types had significantly different proportions for delta time.
Why might there be a greater proportion of the sound frequency range occupied by birds at the site with fewer trees than at the site with many trees? Since a sound recorder that is close to a sound source tends to detect a wider range of sound frequencies than when the recorder is at greater distance (higher frequency sounds are dispersed and attenuated by foliage and other objects over shorter distances than longer frequency sounds) it may be that on average the birds were closer to the sound recorder at Prindle
Ave. than at Stratford Rd. This would be somewhat unexpected (but not impossible) since with more trees present, there are more locations for birds to perch and sing from at the Stratford Rd. location. (A bird actually twice landed on the sound recorder’s foam windscreen at the Stratford Rd. site.) Another possible explanation might be that the diversity of bird species at the Prindle Ave. site includes more bird species that sing in the higher frequency ranges, such as cedar waxwings, cowbirds, and finches in addition to the birds such as robins and mourning doves that typically sing in lower frequency ranges. A third possible explanation might be that the birds detect more human mechanical sounds at the Prindle Ave. site and therefore must shift their songs slightly into higher frequency ranges in order to avoid the sound competition from human
219 mechanical sounds. This may be less likely if birds shift both the upper and lower frequency boundaries of their songs rather than shifting only the upper frequency boundary.
Figure 45c shows that the proportion of energy (reported as decibels) that sound is detected for bird songs and calls (Bb) is significantly greater (alpha = 0.05) at Prindle
Ave., (the site with fewer trees) than at Stratford Rd. (the site with more trees). The total energy reported results from adding decibels values for each sound selection, which is inadvisable since the decibel scale is logarithmic.
Figure 45d corrects the problem of adding decibels by converting the energy values out of the logarithmic scale and into the straight scale of millivolts. When this conversion is done the proportional results change. When sound energy is reported as millivolts, the proportion of sound energy associated with human mechanical sounds (Ax) and all human-sourced sounds combined (A all) is greater at the Prindle Ave. site (few trees).
This may seem to be counterintuitive when looking at the graph since the total amount of human-sourced sound energy is much greater at Stratford Rd. than at Prindle Ave.
However the proportion of human-sourced sounds is greater at Prindle Ave. (98.9%) than at Stratford Rd. (95.2%).
The proportion of sound energy associated with bird song (Bb) and all animal-sourced sounds combined (B all) is significantly greater at the Stratford Rd. site (many trees) than at Prindle Ave. (few trees). It is interesting to note that even in the relatively “quiet” neighborhood where both the Prindle Ave. and Stratford Rd. recording sites are located that the vast majority of sound energy (mV) is associated with human mechanical sounds
(98.8% at Prindle Ave.; 95.2% at Stratford Rd.) versus animal sounds (1.0% at Prindle
220
Ave.; 3.2% at Stratford Rd.). Since there was very little breeze and no precipitation during this 3-hour recording session, geophony sounds occupy a very small proportion of total sound energy.
Figure 45e shows that the proportion of aggregate entropy for bird songs and calls
(Bb) is significantly greater (alpha = 0.05) at Prindle Ave., (few trees) than at Stratford
Rd. (many trees). Since there were very few mammals (Bm) detected at either site, the difference in bird sounds is largely responsible for the significant difference in aggregate entropy when considering all animal-sourced sounds together (B all). No other sound types or sub-types had significantly different proportions for aggregate entropy.
Entropy is a measure of disorder that is a function of sound energy and the distribution of that sound energy within frequencies and within time. In general entropy (disorder) is great when sound energy is evenly distributed over a wide range of frequencies and throughout time, such as the case with “white” noise, or the sound of rain falling. In general, entropy is low when sound energy is focused in a narrow frequency band throughout time, such as the case of a monotonal bell ringing.
Figures 46 (a through f) and 47 (a through h) provide summaries of proportions of sound attributes for each of the six spatially matched pairs and each of the eight temporally matched pairs of sound recordings.
221
Figure 46a through 46f: Proportional summaries for the sound attributes Delta Time, Delta Frequency, Energy (dB), Energy (mV), and Aggregate Entropy for the six spatially Figure 46a matched sound recording pairs. Green cells indicate statistically significant (alpha = 0.05) greater proportions of sound types at sites with many trees; Blue cells indicate statistically significant greater proportions Figure 46b of sound types at sites with few trees; Light green cells at the Suffield Rd. site (45f) indicate statistically significant greater proportions of sound types in 2015 where little loss of tree canopy had occurred in the immediate area but substantial Figure 46c tree loss had occurred within the larger neighborhood as compared to the same site in 2014 prior to tree removal in the larger neighborhood.
Figure 46d
Figure 46e
Figure 46f
222
Figure 47a through 47h: Proportional summaries for the sound attributes Delta Time, Delta Frequency, Energy (dB), Energy (mV), and Aggregate Entropy for the eight Figure 47a temporally matched sound recording pairs. Green cells indicate statistically significant (alpha = 0.05) greater proportions of sound types at sites with many trees; Blue cells indicate statistically significant greater proportions of sound types at sites with few trees. Figure 47b
Figure 47c
Figure 47d
Figure 47e
Figure 47f
223
Figure 47a through 47h (continued): Green cells indicate statistically significant (alpha = 0.05) greater proportions of sound types at sites with many trees; Blue cells indicate statistically significant Figure 47g greater proportions of sound types at sites with few trees; Gray cells at the Wilke Rd. site (47h) indicate statistically significant greater proportions of sound types adjacent to freeway compared to the Huron St. site after many trees Figure 47h had been removed.
When considering the measures of sound attributes for the various pairs of sound recordings are there meaningful patterns relative to the types and sub-types of sounds?
Do the proportions of sounds associated with humans, animals and weather change as trees are removed? It is useful to consider each of the sound attributes individually when exploring these questions.
Proportional changes to soundscape types relative to time.
When considering the sound attribute of delta time for anthrophony, biophony, and geophony there is little evidence of consistent change as tree canopy decreases. For the six spatially matched pairs of recordings there is only a single incidence of a significant difference in delta time, and that occurs in the Prindle Ave. 4-12-14 / 4-11-15 pair of recordings as human mechanical sounds (Ax) occupy a greater proportion of time after trees are removed.
224
For the seven temporally matched pairs of recordings (excluding the recording pair that includes Wilke Rd. adjacent to the freeway) there are many examples of significant proportional differences relative to time, however the results are mixed. There is no definite trend that suggests that components of anthrophony, biophony, or geophony consistently change relative to the proportion of time occupied in a soundscape as tree cover decreases. The one exception occurs on the temporally matched pair recorded on
September 5, 2014 at the Stratford Rd. and Prindle Ave. sites. For this recording there was proportionally more time for insect sounds at Stratford Rd. where there were more trees. This is as expected since the greater number of trees at Stratford Rd. provides more sites for cicadas to call from.
Proportional changes to soundscape types relative to frequency.
When considering the sound attribute of delta frequency for anthrophony, biophony, and geophony there is relatively strong evidence of consistent change as tree canopy decreases. When considering the six spatially matched pairs of recordings, four pairs of recordings show significant increase in the range of sound frequencies of human mechanical sounds (Ax) as tree cover decreases while only one pair shows a proportionally narrower range of frequencies as tree cover decreases. When considering the seven temporally matched pairs of recordings, four pairs of recordings show significant increase in the range of sound frequencies of human mechanical sounds (Ax) as tree cover decreases while two pairs show a proportionally narrower range of frequencies as tree cover decreases. What might be the reason for this change? One
225 explanation may be that by chance more human mechanical sources of sound such as passing cars or lawn mowers were closer to the sound recorder at the sites with fewer trees. This does not seem to be the case. Another explanation may be that sound attenuation by vegetation decreases as tree canopy is lost, and since noise attenuation by vegetation is most effective in the higher frequency ranges the frequency range of sounds detected would be expected to be broader in areas with fewer trees. If this were true the trend would perhaps be evident across all sound types and not just anthrophony, and indeed the frequency range of biophony sounds considered collectively shows a significant increase of frequency range as tree cover decreases. When considering the six spatially matched pairs of recordings, three pairs of recordings show significant increase in the range of sound frequencies of biophony sounds considered collectively (B all) as tree cover decreases, while only one pair shows a proportionally narrower range of frequencies as tree cover decreases. When considering the seven temporally matched pairs of recordings, five pairs of recordings show significant increase in the range of sound frequencies of biophony sounds considered collectively (B all) as tree cover decreases, while two pairs show a proportionally narrower range of frequencies as tree cover decreases.
The trend as tree canopy decreases is that a broader range of sound frequencies is detected both in human produced sounds and in animal produced sounds. One plausible reason is that as the density of vegetation decreases the noise attenuation properties of the landscape decreases and sound energy, particularly in the higher frequency ranges, would travel farther before being dispersed. This leads to the question, are animals shifting the frequency of their vocalizations (or stridulation in the case of insects) to avoid increasing
226 sound competition from humans? In three of the seven spatially matched pairs of recordings and in three of the seven temporally matched pairs of recordings the proportion of sound frequency range for birds increased as trees were removed. This may be that on average birds were closer to the sound recorder, or that the species diversity of birds changed as trees were removed with a higher proportion of higher- pitched bird species present in areas of fewer trees. This may be a response by birds to avoid sound competition as the frequency range of human mechanical sounds broadens, or it may be a decrease in sound energy attenuation as tree cover is lost.
Proportional changes to soundscape types relative to energy.
Changes to the amount of energy carried by sound waves is interpreted by human hearing as a change in loudness. As previously discussed, sound energy is usually reported in the unit decibels (dB), which is built on a logarithmic scale. To compare the proportion of sound energy between sound recordings I used both the decibel scale as well as converting logarithmic decibels to the non-logarithmic millivolt scale (mV).
Figure 48a through 48f present the total sound Energy (mV) by sound type (anthrophony, biophony, and geophony) and sub-type for the six spatially matched sound recording pairs. Figure 49a through 49h present the total sound Energy (mV) by sound type
(anthrophony, biophony, and geophony) and sub-type for the eight temporally matched sound recording pairs.
When considering the sound energy reported in millivolts for anthrophony, there is evidence of change as tree canopy decreases. When considering the six spatially matched
227 pairs of recordings (Figures 46a through f), three pairs of recordings (Figures 46b, c, & d) show a significant proportional increase in sound energy of human mechanical sounds
(Ax) as tree cover decreases, while two pairs (Figures 46a & f) show a proportional decrease in sound energy as tree cover decreases. When considering the eight temporally matched pairs of recordings (Figures 47a through h), five pairs of recordings (Figures
47b, c, e, f, & g) show significant increase in sound energy (mV) of human mechanical sounds (Ax) as tree cover decreases, while zero pairs show a proportional decrease in sound energy as tree cover decreases.
What may be the reason for this apparent increase in sound energy for human mechanical sounds? One reason may be the random occurrence of more vehicles, aircraft, lawn maintenance equipment or other mechanisms in the immediate area. In some cases lawn mowers and leaf blowers passed within feet of the sound recorder releasing stunning amounts of energy. But that seems not to be the dominant source of sound energy in most recordings. For example, in the Stratford Rd. and Prindle Ave. simultaneous recordings from May 19, 2015, more than three times as much sound energy from human mechanical sources was detected at the Stratford Rd. site (with many trees) than at the Prindle Ave. site (with few trees). This is not due to the localized presence of loud machines but rather greater sound energy detected in background sounds, overhead planes, and passing cars.
The vast difference in total sound energy attributable to human mechanisms measured between Stratford Rd. and Prindle Ave. on May 19, 2015 does not translate into the same significant difference in proportional sound energy, and in fact the direction of these relationships is surprisingly opposite.
228
Figure 48a through 48f: Total Fig 48a sound Energy (mV) by sound type (anthrophony, biophony, and geophony) and sub-type for the six spatially matched sound recording pairs. Green bars indicate total sound energy (mV) at sites with many trees; Blue bars indicate total sound energy at sites with few trees Fig 48b
Fig 48c
Fig 48d
229
Figure 48a through 48f Fig 48e (continued): Light green cells at the Suffield Rd. site (47f) indicate total sound energy (mV) in 2015 where little loss of tree canopy had occurred in the immediate area but substantial tree loss had occurred within the larger neighborhood as compared to the same site in 2014 prior to tree removal in the larger Fig 48f neighborhood (dark green bars).
Figure 49f shows that the Stratford Rd. site (many trees) had more than 3 times the total sound energy (mV) associated with human mechanical sounds as did the Prindle Ave. site
(few trees) and yet the Prindle Ave. site has a significantly higher proportion of the soundscape represented by human mechanical sounds than at Stratford Rd. This is possible because while Stratford Rd. had more than 3 times the sound energy attributable to human mechanical sounds as did Prindle Ave., Stratford Rd. had more than 15 times the total sound energy attributable to animal sounds as did Prindle Ave. (Figure 45d).
Therefore, proportionally the Stratford Rd. site (many trees) had a significantly lower proportion of human mechanical sounds and a significantly greater proportion of animal sounds (Figure 47f).
230
Figure 49a through 49h: Total Fig 49a sound Energy (mV) by sound type (anthrophony, biophony, and geophony) and sub-type for the eight temporally matched sound recording pairs. Green bars indicate total sound energy (mV) at sites with many trees; Blue bars indicate total sound energy at sites with few trees.
Fig 49b
Fig 49c
Fig 49d
231
Figure 49a through 49h (continued): Total sound Energy (mV) by sound type (anthrophony, biophony, and geophony) and sub-type for the eight temporally matched sound recording pairs. Green bars indicate total sound energy (mV) at sites with many trees; Blue bars indicate total sound energy at sites with few Fig 49f trees. Gray bars at the Wilke Rd. site (47h) indicate total sound energy adjacent to freeway compared to the Huron St. site (blue bars) after many trees had been removed.
Fig 49g
Fig 49h
232
Is there evidence that sound energy associated with animals (biophony) changes as tree canopy changes? When considering the sound energy reported in millivolts for biophony, there is evidence of change as tree canopy decreases. When considering the six spatially matched pairs of recordings (Figures 46a through f), two pairs of recordings
(Figures 45b & c) show a significant proportional decrease in sound energy of all animal sounds considered collectively as tree cover decreases while three pairs (Figures 46a, e & f) show a proportional increase in sound energy as tree cover decreases. When considering the seven temporally matched pairs of recordings (Figures 47a through g), six pairs of recordings (Figures 47a, b, c, e, f, & g) show significant proportional decrease in sound energy (mV) of all animal sounds considered collectively (B all) as tree cover decreases while zero pairs show an increase in sound energy as tree cover decreases.
What may be a plausible explanation for this proportional decrease in sound energy associated with animal sounds as tree canopy decreases? It may be that there are fewer places for birds, insects (such as cicadas) and even amphibians (in the case of tree frogs) to perch and call. It may be that the decrease in trees results in fewer places to mate and lay eggs. This may be the case for cicadas and katydids, but less likely for most of the songbirds detected (since few of these nest in large trees preferring to nest in shrubs and smaller trees), and certainly not the case for amphibians that lay their eggs in water.
At least one bird species certainly has lost nesting habitat as large ash trees were removed. Many nests occupied by Cooper’s hawks (Accipiter cooperii) were seen in large ash trees (Figure 50). Cooper’s hawks are relatively quiet and rarely are detected in the soundscape recordings, but this species may be the most severely affected by the destruction and necessary removal of ash trees due to loss of nesting habitat.
233
Figures 50a (top left), b (bottom left), & c (right): Cooper’s hawks and occupied nests were frequently observed in large ash trees. Red arrow indicates location of nest.
Perhaps the most surprising result from the proportional analysis for sound energy of the soundscapes is that of the six spatially matched pairs of recordings. Only one shows a significance difference in the proportional measure of geophony, which occurs during the early morning Huron St. recordings of May 7, 2014 and May 6, 2015. The significant difference shows that for the measure of total energy reported as decibels (but not when converted to millivolts) there is proportionally greater sound energy associated with geophony (in this case wind) when more trees are present (Figure 46b). For the analysis of the soundscapes of the seven temporally matched pairs of recordings only one shows a significance difference in the proportional measure of geophony, which occurs during the simultaneous recordings at Stratford Rd. and Prindle Ave. on September 5, 2014. The
234 significant difference shows that for the measure of total energy reported as millivolts
(but not when reported as decibels) there is proportionally greater sound energy associated with geophony (in this case wind) when more trees are present (Figure 47b).
These two results are as expected since the movement of air as wind does not result in sound unless the wind is intercepted by an object such as trees. The presence of more trees creates more obstacles for wind to impact and therefore more sound energy to be created. But why do so few of the pairs of recordings demonstrate significant proportional differences in sound energy associated with wind or other weather-related sounds? One reason is that most of the pairs of sound recordings begin before dawn when wind is typically calm. Another perhaps more influential reason is that in order to protect the digital sound recorders, recording was terminated when precipitation was greater than a light drizzle. Sound recording was also avoided during periods of moderate to heavy wind. This was due to the fact that when wind impacts a microphone directly (even when the wind screen is in place and when the digital low cut filters are activated) it causes severe distortion of the sound recording resulting in an artificially high reading of sound energy and frequency range. Therefore there was a bias against recording soundscapes during conditions when sounds associated with geophony where active.
Proportional changes to soundscape types relative to entropy.
The attributes of time, frequency range, and energy (perceived as loudness) are easily understood as characteristics of sound, and easy to recognize in a spectrogram. The
235 sound attribute of entropy is more difficult to comprehend. Entropy, by definition, is a measure of disorder, not only in sound, but in other types of energy (such as in thermodynamics) and in the organization of matter itself. A simple example of entropy might be illustrated by dumping the pieces of a jigsaw puzzle onto a table. This situation is one of little order (great disorder) and entropy is high. When the puzzle is completely assembled the situation is one of great order (little disorder) and entropy is low.
Entropy is related to but not synonymous with the patterns of sound energy (the sound signature) seen in a spectrogram. Figures 51a through 51e provide examples of spectrograms with a range of entropy values. The highest level of disorder and therefore entropy is measured in sounds that exist over a wide range of frequencies and have no distinct pattern to the sound energy, such as in the case of wind blowing through tree leaves or rain striking vegetation (Figure 51f). On a spectrogram these sounds appear as a non-descript fog over the entire range of frequencies. The lowest level of disorder and therefore entropy is measured in sounds that exist in a very narrow frequency range such as the monotonal sound of a single bell. On a spectrogram this note appears as a narrow horizontal line. Some birds such as the black-capped chickadee have call notes that are simple and close to monotonal (Figure 51a) while their song is complex with a higher degree of entropy (Figure 51c).
In general, sounds associated with human mechanisms tend to have greater entropy than sounds associated with human voices or calling animals. There are of course sounds from human mechanisms such as the whistle of a distant train or a jet plane passing overhead that have relatively low entropy. Figure 51b presents the relatively high
236 entropy spectrogram of a truck engine with the relatively low entropy spectrogram of the truck’s squealing brakes.
Bird song frequently occurs in repeated complex patterns that are easily recognizable as sound signatures. But a highly complex and easily recognizable song does not equate to either low or high entropy. Certainly these songs are highly organized, and the pitch of a song may occur within a very narrow frequency range at any given moment (both of which lead to low entropy), but the songs can vary greatly over a wide range of frequencies over time (both of which lead to high entropy). Calls of amphibians such as spring peepers and tree frogs generally occur within a relatively narrow range of frequencies (leading to low entropy values). Buzzing and grinding insects such as cicadas create poorly focused sound signatures over relatively broad frequency ranges
(tending toward high entropy values) (Figure 51e).
When sorting the aggregate entropy sound attribute data from greatest to least values there is no definitive breaking points between sound types, however geophony sounds associated with wind and precipitation tend to have higher values for aggregate entropy than anthrophony and biophony sounds. Biophony sounds (birds and other animals) and human voices tend to have mid-range aggregate entropy values. Anthrophony from mechanical sounds tends to span the entire range of aggregate entropy values, but many tend to be in the mid to low range of values for aggregate entropy. The lowest values for aggregate entropy are almost exclusively associated with the distorted sounds from wind directly striking the recorder microphone.
237
Fig 51a: Black-capped chickadee call note. Aggregate entropy = 2.8 bits Fig 51b (center): Selection 11 – Truck engine. Aggregate entropy = 7.8 bits; Selection 12 – Truck brakes. Aggregate Fig 51c: Black-capped chickadee song. Aggregate entropy = 4.8 bits entropy = 7.2 bits
Fig 51d: House wren song. Aggregate entropy = 6.0 bits
Fig 51e: Scissors-grinder cicada. Aggregate entropy = 7.2 bits
Fig 51f: Rain. Aggregate entropy = 7.9 bits
Figure 51a through 51f: Spectrogram examples of aggregate entropy values of various sound signatures.
238
Figure 45e shows that there are significantly greater proportional values in aggregate entropy for bird sounds (Bb) and for all biophony considered together (B all) at Prindle
Ave. (few trees) when compared with Stratford Rd. (many trees) for the May 19, 2015 simultaneous recordings.
When considering aggregate entropy for anthrophony, there is little evidence of proportional change as tree canopy decreases. When considering the six spatially matched pairs of recordings (Figures 46a through f), three pairs of recordings (Figures
46c, d & f) show a significant proportional increase in aggregate entropy of human mechanical sounds (Ax) as tree cover decreases while only one pair (Figure 46a) shows a proportional decrease in entropy as tree cover decreases. When considering the seven temporally matched pairs of recordings (Figures 47a through g), two pairs of recordings
(Figures 47a & e) show significant increase in aggregate entropy of human mechanical sounds (Ax) as tree cover decreases while two pairs (Figures 47d & g) show a proportional decrease in entropy as tree cover decreases.
When considering aggregate entropy for biophony, there is stronger evidence of proportional change as tree canopy decreases. When considering the six spatially matched pairs of recordings (Figures 46a through f), two pairs of recordings (Figures 46a
& d) show a significant proportional increase in aggregate entropy of combined biophony sounds (B all) as tree cover decreases while only two pairs (Figures 46c & f) show a proportional decrease in entropy as tree cover decreases. When considering the seven temporally matched pairs of recordings (Figures 47a through g), five pairs of recordings
(Figures 47b, c, d, f & g) show significant increase in aggregate entropy of combined biophony sounds (B all) as tree cover decreases while only two pair (Figures 47a & e)
239 shows a proportional decrease in aggregate entropy as tree cover decreases. But somewhat surprisingly, when considering only the birdsong component of biophony (the component that has the greatest potential for variation and change relative to entropy) the results are divided. When considering the eight temporally matched pairs of recordings
(Figures 47a through h), three pairs of recordings (Figures 47b, c & f) show significant proportional increase in aggregate entropy of bird sounds (Bb) as tree cover decreases while three pairs (Figures 47a, d & g) show a significant proportional decrease in aggregate entropy as tree cover decreases.
Proportional changes to soundscape types summary.
What might be the expected pattern of change, if any, relative to time, frequency, energy and entropy as tree canopy decreases? Part of that answer is driven by the role trees play in attenuating sound. Trees affect sound waves primarily through scattering of the sound waves. The foliage, branches and trunks of trees reflect the sound wave energy in various directions which reduces the sound energy that reaches the receiver from the sound transmitter. The proportion of sound energy that is scattered is dependent on the wavelength (and therefore the frequency) of the sound energy as it relates to the size of the foliage, branches and trunks of the trees. Because of this relationship between sound wavelength and tree parts, sound energy in the wavelengths between 1,000 Hz and beyond 8,000 Hz is most susceptible to disruption by trees. Therefore as tree canopy cover diminishes it might be expected that recorded sound energy within these wavelengths would increase, regardless of the source of the sound (human mechanical,
240 birdsong, wind, etc.) The frequency range might also broaden into the frequencies that are most susceptible to scattering by vegetation, but only with the sound sources that transmit in those frequencies. Since entropy is a function of the range of frequencies detected within a sound, measures of entropy may increase as tree cover decreases.
Another reason that there may be a proportional change in sound attributes as tree cover diminishes is the response to the loss of trees by living sources of sound. Human mechanical sources of sounds will not change the noise that they make in response to tree cover, but some animals may change their transmitted sounds. This is because unlike noise that carries no purposeful information, biophony is sound produced with the purpose of carrying information from the source to the intended receivers, and therefore the source of the sound (birds, mammals, amphibians and insects) may adjust their songs and calls to match the conditions of the environment in which they exist. Because of their vocalization structure, birds have the greatest ability to adjust their songs and calls, and therefore may have the ability to purposefully adjust one or more attributes of the sounds they produce to maximize the transmission of the sound. This might be accomplished by singing louder (transmitting their sounds with greater energy) and/or by shifting the frequency of their songs and calls. The combination of increased energy and altered frequency range may also affect measures of entropy.
If birds and other animals alter their songs and calls according to the density of vegetation, what direction would these shifts be likely to take? The amount of time an animal transmits sound may or may not be affected. If as tree canopy is lost and therefore the sound refraction decreases and as a result more human mechanical sounds are present, birds and other animals may decrease their calling time during peaks of
241 anthrophony, such as traffic rush periods. Because of this same decrease in sound refraction, animals (particularly birds) may shift their songs and calls to a higher frequency range to avoid competition from noise. If there is more sound energy present in the form of mechanical noise due to decreased sound refraction, animals may need to essentially “shout” to be heard by increasing the energy that goes into their calls. Since increasing sound energy output requires more energy from the source, animals will expend more energy in a shorter amount of time, and therefore call less frequently.
Birdsong has been shown to be less complex in urban areas with high levels of background noise (Wood, Yezerinac, & Duffy, 2006) and therefore measures of entropy may decrease as tree canopy is removed.
Were any of these possible changes to sound attributes observed in the measures of proportions for the matched pairs of sound recordings?
There is little evidence of proportional changes in measures of time for
anthrophony, biophony and geophony in either the temporally matched or
spatially matched pairs of sound recordings.
When considering the sound attribute of delta frequency for anthrophony,
biophony, and geophony there is relatively strong evidence of consistent change
as tree canopy decreases. Four of the six spatially matched pairs of recordings,
and four of the seven temporally matched pairs of recordings show significant
proportional increase in the range of sound frequencies of human mechanical
sounds (Ax) as tree cover decreases. Three of the six spatially matched pairs of
recordings, and five of the seven temporally matched pairs of recordings show
242
significant proportional increase in the range of sound frequencies of biophony
sounds considered collectively (B all).
When considering proportional sound energy reported in millivolts for
anthrophony, there is evidence of change as tree canopy decreases. Three of the
six spatially matched pairs of recordings and five of the seven temporally matched
pairs of recordings show a significant proportional increase in sound energy of
human mechanical sounds (Ax) as tree cover decreases. When considering
proportional sound energy reported in millivolts for biophony, there is also
evidence of change as tree canopy decreases. Three of the six spatially matched
pairs of recordings show a significant proportional decrease in sound energy of all
animal sounds considered collectively as tree cover decreases while two pairs
show a proportional increase in sound energy as tree cover decreases. Six of the
seven temporally matched pairs of recordings show significant proportional
decrease in sound energy (mV) of all animal sounds considered collectively (B
all) as tree cover decreases while zero pairs show a proportional increase in sound
energy as tree cover decreases.
When considering aggregate entropy for anthrophony, there is little evidence of
consistent proportional change as tree canopy decreases. Three of the six
spatially matched pairs of recordings show a significant proportional increase in
aggregate entropy of human mechanical sounds (Ax) as tree cover decreases
while only one pair shows a proportional decrease in entropy as tree cover
decreases. Two of the seven temporally matched pairs of recordings show
significant increase in aggregate entropy of human mechanical sounds (Ax) as
243
tree cover decreases while two pairs show a proportional decrease in entropy as
tree cover decreases. There is stronger evidence of proportional change in
aggregate entropy within biophony as tree canopy decreases. While only two of
the six spatially matched pairs of recordings show a significant proportional
increase in aggregate entropy of combined biophony sounds (B all) as tree cover
decreases, six of the seven temporally matched pairs of recordings show
significant increase in aggregate entropy of combined biophony sounds (B all) as
tree cover decreases.
These results in some ways support the predicted outcome. As trees are removed a proportionally wider range of frequencies of human mechanical sounds are detected, and a wider range of frequencies of animal sounds are also detected. It is unclear if this is purely a function of less sound refraction due to fewer trees or if animals are shifting their sounds to higher frequencies to avoid sound competition from human mechanical sources. As trees are removed proportionally more sound energy is detected from human mechanical sounds and proportionally less sound energy is detected from animals. The result for measures of entropy are not as expected. As trees are removed proportional aggregate entropy increases for animal sounds.
Analysis of sound attribute mean values.
The most detailed analysis of sound attributes by sound source type was conducted by comparing mean values using t-Tests of a wide range of sound attributes collected during the sound-by-sound manual identification of individual sound selections. The detailed t-
244
Test results are presented for the temporally matched recording pair of Stratford Rd. and
Prindle Ave. on May 19, 2015. This pair was selected for discussing the detailed analysis in this chapter because the recording pair is used in part 2 of this research during the tests for directed attention. This recording pair is 3 hours and 3 minutes in length beginning at
5:30 am (local time). The sound-by-sound analysis of the Stratford Rd. (many trees) recording identified 2,130 distinct sounds or sound choruses. The analysis of the Prindle
Ave. (few trees) recording identified 2,061 sounds. Added to these were 33 silence samples from Stratford Rd. and 15 silence samples from Prindle Ave. Table X presents the summary of identified sounds by type and sub-type.
Table X: Count of identified sounds and sound choruses for the May 19, 2015 sound recordings at Stratford Rd. (many trees) and Prindle Ave. (few trees).
The following discussions present the findings for each sound attribute analysis of this recording pair.
t-Test for Delta Time by sound type.
A series of t-Tests for the Delta Time sound attribute were conducted to identify statistically significant differences in mean values of time duration of individual sound
245 selections. The alpha value for these tests is 0.05. Table XI presents the results for the temporally matched sound recording pair from Stratford Rd. and Prindle Ave. on May 19,
2015. Similar t-Tests were completed for each of the six spatially matched recording pairs and each of the seven additional temporally matched recording pairs. Each set of t-
Tests compared the anthrophony, biophony, and geophony type of each sound recording, and also each of the sub types (Av – anthrophony human voices & music, Ax – anthrophony mechanical, Ba – biophony amphibians (when present), Bb – biophony birds, Bi – biophony insects (when present), Bm – biophony mammals, and Gw – geophony wind).
The results for the Stratford Rd. and Prindle Ave. May 19, 2015 recordings show that there is a statistically significant difference in the mean Delta Time values for the anthrophony: voices and music sub class (Av), but not for the mechanical (Ax) sub type or for all anthrophony combined (A all). This indicates the sound selections identified as human voices on the recording with few trees had longer time intervals on average than at the site with many trees.
When considering the biophony subclasses together (Bb, & Bm) there was no statistically significant difference in the mean Delta Time values. When considered separately there was no statistically significant difference between bird (Bb) sounds.
There was however a statistically significant difference in mammal sounds (Bm) suggesting on average squirrels and dogs were calling longer at the Prindle Ave. location
(few trees) than at the Stratford Rd. location (many trees). This result is deceiving however because there were only four mammal sounds detected at Prindle Ave. while
Stratford Rd. heard 150 individual mammal sounds (mostly squirrels). Total time for
246 mammal sounds was much greater at Stratford Rd. but mean time for individual mammal sounds was significantly longer at Prindle Ave.
Table XI: Delta Time sound attribute t-Test results for anthrophony, biophony, and geophony sound types for the temporally matched sound recordings at Stratford Rd. and Prindle Ave. on May 19, 2015. Blue shading indicates statistically significant (alpha = 0.05) greater delta time mean values for Prindle Ave. (few trees); Green shading indicates statistically significant greater delta time mean values for Stratford Rd. (many trees).
247
There is another number that is interesting in the t-Test results table. The difference in the number of observations for mammal sounds between Stratford and Prindle results in a very low degrees of freedom for the t-Test (26). With so few observations of mammals at one of the sites, statistical analysis becomes more demanding to achieve significant results.
There was a statistically significant difference in the time duration of wind rustling trees sounds (geophony wind (Gw) sound types). As expected the site with many trees had on average longer periods of wind-related sounds.
t-Test for Delta Frequency by sound type.
A series of t-Tests for the Delta Frequency sound attribute were conducted to identify statistically significant differences in mean values of frequency range of individual sound selections. The alpha value for these tests is 0.05. Table XII presents the results for the temporally matched sound recording pair from Stratford Rd. and Prindle Ave. on May 19,
2015. Similar t-Tests were completed for each of the six spatially matched recording pairs and each of the seven additional temporally matched recording pairs. Each set of t-
Tests compared the anthrophony, biophony, and geophony types of each sound recording, and also each of the sub types (Av – anthrophony human voices & music, Ax – anthrophony mechanical, Ba – biophony amphibians (when present), Bb – biophony birds, Bi – biophony insects (when present), Bm – biophony mammals, and Gw – geophony wind.
The results for the Stratford Rd. (many trees) and Prindle Ave. (few trees) May 19,
2015 recordings show that there is a statistically significant difference in the mean Delta
248
Frequency values for the anthrophony: voices and music sub class (Av), but not for the mechanical (Ax) sub classes or for all anthrophony combined (A all). This indicates the sound selections identified as human voices on the recording with few trees had broader frequency ranges on average than at the site with many trees.
When considering the biophony sub types of birds (Bb) and mammals (Bm) individually as well as when considered collectively (B all) there was statistically significant difference in the mean Delta Frequency values. This suggests that the mean frequency ranges for bird song and mammal sounds was greater at the site with few trees.
What might be an explanation for this? One reason could simply be that both birds and mammals were closer on average to the sound recorder at the site with few trees resulting in less attenuation of higher frequency sounds due to distance between the sound source and the receiver. This seems unlikely since there were more locations and closer locations in which birds and squirrels could perch and call at the site with many trees.
Another reason could be that with fewer trees there is less sound refraction at Prindle
Ave. resulting in more high frequency sounds reaching the sound recorder.
There was a statistically significant difference in the frequency range of wind rustling trees sounds (geophony wind (Gw) sound types). The site with few trees had on average broader frequency ranges of wind-related sounds. On the one hand this is unexpected since on average trees are located a greater distance from the recorder at the site with few trees. The expected result would be that narrower frequency ranges on average would be recorded due to attenuation caused by distance between the source and the receiver. On the other hand with fewer trees to refract sounds, higher frequency sounds may actually reach the receiver even at greater distances.
249
Table XII: Delta Frequency sound attribute t-Test results for anthrophony, biophony, and geophony sound types for the temporally matched sound recordings at Stratford Rd. and Prindle Ave. on May 19, 2015. Blue shading indicates statistically significant (alpha = 0.05) greater delta frequency mean values for Prindle Ave. (few trees).
250
t-Test for Delta Time X Delta Frequency by sound type.
The Delta Time attribute is plotted on the x-axis of a spectrogram and the Delta
Frequency is plotted on the y-axis. The low and high values for each of these attributes creates the boundaries of a sound selection. The values of Delta Time and Delta
Frequency were multiplied for each sound selection to create a measure of sound selection area. These products were then compared using t-Tests as previously described.
Table XIII presents the results for the t-Tests for each of the sound types and sub-types.
The results for the Stratford Rd. and Prindle Ave. May 19, 2015 recordings show that there is a statistically significant difference in the mean Delta Time x Delta Frequency values for the anthrophony: voices and music sub class (Av), but not for the mechanical
(Ax) sub classes or for all anthrophony combined (A all). This indicates the sound selections identified as human voices on the recording with few trees had greater combined values for time and frequency on average than at the site with many trees. This is as expected since there are significant findings for Av in both the Delta Time and Delta
Frequency tests, but not in the Ax and A all sound types.
When considering the biophony sub types individually (Bb, & Bm) there was statistically significant difference in the mean Delta Time x Delta Frequency values for each, but the results suggest different findings for birds and mammals. The mean values of the Delta Time x Delta Frequency products are greater for birds at the site with many trees and greater for mammals at the site with few trees. When considered collectively
(B all) there was statistically significant difference between animal sounds at the two sites, indicating animal sounds at the Stratford Rd. location (many trees) had greater mean Delta Time x Delta Frequency values.
251
Table XIII: Delta Time X Delta Frequency sound attribute t-Test results for anthrophony, biophony, and geophony sound types for the temporally matched sound recordings at Stratford Rd. and Prindle Ave. on May 19, 2015. Blue shading indicates statistically significant (alpha = 0.05) greater delta time x delta frequency mean values for Prindle Ave. (few trees); Green shading indicates statistically significant greater delta time x delta frequency mean values for Stratford Rd. (many trees).
While the individual Delta Time and Delta Frequency t-Tests for Geophony sounds showed no significant difference between the two sites the Delta Time x Delta Frequency
252 product values show a strong significant difference. Not surprisingly the site with many trees has greater mean values than the site with few trees as wind has many more trees to interact with at Stratford Rd.
t-Test for Center Frequency by sound type.
The Delta Frequency sound attribute measures the total range of frequencies within a sound selection. The Center Frequency sound attribute calculates the frequency within a sound selection that divides the selection into two intervals of equal energy. The advantage of this measure is that unlike Delta Frequency which is completely dependent on the analyst’s placement of the sound selection boundaries, the Center Frequency attribute is less sensitive to analyst bias. Table XIV presents the results for the Center
Frequency t-Tests for each of the sound types and sub-types.
The results for the Stratford Rd. and Prindle Ave. May 19, 2015 recordings show that there is a statistically significant difference in the mean Center Frequency values for the anthrophony: voices and music sub-type (Av), for the mechanical (Ax) sub-type and for all anthrophony combined (A all). This indicates the human-sourced sound selections from the site with few trees (Prindle Ave.) had higher Center Frequencies on average than at the site with many trees (Stratford Rd.).
253
Table XIV: Center Frequency sound attribute t-Test results for anthrophony, biophony, and geophony sound types for the temporally matched sound recordings at Stratford Rd. and Prindle Ave. on May 19, 2015. Blue shading indicates statistically significant (alpha = 0.05) greater center frequency mean values for Prindle Ave. (few trees); Green shading indicates statistically significant greater center frequency mean values for Stratford Rd. (many trees)
When considering the biophony subclasses individually (Bb, & Bm) and collectively (B all) there was statistically significant difference in the mean Center
Frequency values. As with human-sourced sounds, this indicates the animal-sourced
254 sound selections from the site with few trees (Prindle Ave.) had higher Center
Frequencies on average than at the site with many trees (Stratford Rd.). The most plausible reason for these consistently higher Center Frequency values is that with fewer trees there is less sound refraction at Prindle Ave. (few trees) resulting in more high frequency sounds reaching the sound recorder.
If this were true then the same trend would be expected for geophony sounds, although in this case there is no statistically significant difference between the two sites.
The sound attributes of Delta Time and Delta Frequency are consistent for samples of silence. All samples of silence are 1 second long and have a frequency range of 20 Hz to
20,000 Hz, therefore t-Tests for silence were not conducted for these sound attributes.
Center Frequency varies among silence samples. With fewer trees at the Prindle Ave. site it was expected that sound refraction would be decreased and therefore a broader range of frequencies would be recorded (as was shown for the human and animal-sourced sounds). Unexpectedly the t-Test result for silence shows a significantly higher Center
Frequency mean value at Stratford Rd. (many trees). This may be explained by the mean values for Center Frequency in the silence samples. Both sites have mean center frequency values below 200 Hz. Trees are most effective at scattering sound energy with wavelengths greater than 1,000 Hz, so the density of tree cover is less likely to be a sound attenuation factor for these low-frequency background sounds.
255
t-Test for Peak Frequency by sound type.
The Peak Frequency sound attribute calculates the frequency within a sound selection at which maximum power takes place (for human listeners this is the frequency that sounds the loudest). The advantage of this measure is that unlike Delta Frequency which is completely dependent on the analyst’s placement of the sound selection boundaries, the
Peak Frequency attribute (as with Center Frequency) is less sensitive to analyst bias.
Table XV presents the results for the Peak Frequency t-Tests for each of the sound types and sub-types.
The results for the analysis of Peak Frequency are very similar to those for Center
Frequency. There is a statistically significant difference in the mean Peak Frequency values for the anthrophony: voices and music sub-type (Av), for the mechanical (Ax) sub-type and for all anthrophony combined (A all). This indicates the human-sourced sound selections from the site with few trees (Prindle Ave.) had higher Peak Frequencies on average than at the site with many trees (Stratford Rd.).
When considering the biophony sub types individually (Bb, & Bm) and collectively
(B all) there was statistically significant difference in the mean Peak Frequency values.
As with human-sourced sounds, this indicates the animal-sourced sound selections from the site with few trees (Prindle Ave.) had higher Peak Frequencies on average than at the site with many trees (Stratford Rd.).
As with the Center Frequency attribute, there is no significant difference for Peak
Frequency between geophony sounds at the two sites. Unlike the Center Frequency sound attribute there is also no significant difference for Peak Frequency between the
256 silence samples at the two sites. At both sites mean Peak Frequency values for the silence samples are less than 100 Hz indicating the very low frequency of background sounds.
Table XV: Peak Frequency sound attribute t-Test results for anthrophony, biophony, and geophony sound types for the temporally matched sound recordings at Stratford Rd. and Prindle Ave. on May 19, 2015. Blue shading indicates statistically significant (alpha = 0.05) greater peak frequency mean values for Prindle Ave. (few trees).
257
t-Test for Average Power (mV) by sound type.
As with a digital camera image, a spectrogram is composed of a series of pixels or bins that each represent a small bundle of energy. The Average Power sound attribute sums the power spectral density for each bin within a sound selection and the total is divided by the number of bins in the selection to provide an average value of power. One disadvantage of this measure is that it is somewhat sensitive to the analyst’s placement of the sound selection boundaries.
As previously discussed, sound power and energy is reported within the Raven program with the units of decibels (dB), which is a logarithmic scale. Before adding the individual Average Power values for each sound selection the logarithmic decibel values were converted to straight scale millivolt (mV) values. Table XVI presents the results for the Average Power (mV) t-Tests for each of the sound types and sub-types.
The results for the Stratford Rd. and Prindle Ave. May 19, 2015 recordings show that there is a statistically significant difference in the mean Average Power (mV) values for the anthrophony: voices and music sub-type (Av), for the mechanical (Ax) sub-type and for all anthrophony combined (A all). This indicates the human-sourced sound selections from the site with many trees (Stratford Rd.) had higher mean Average Power values than at the site with few trees (Prindle Ave.).
When considering the biophony sub types individually (Bb, & Bm) and collectively
(B all) there was statistically significant differences in the mean Average Power (mV) values. As with human-sourced sounds, this indicates the animal-sourced sound
258 selections from the site with many trees (Stratford Rd.) had higher Average Power values on average than at the site with few trees (Prindle Ave.).
Table XVI: Average Power (mV) sound attribute t-Test results for anthrophony, biophony, and geophony sound types for the temporally matched sound recordings at Stratford Rd. and Prindle Ave. on May 19, 2015. Green shading indicates statistically significant (alpha = 0.05) greater mean Average Power values for Stratford Rd. (many trees)
259
This trend continues with geophony sounds. There are significantly greater mean values for Average Power at Stratford Rd. than at Prindle Ave. When considering the silence samples the pattern ceases as no significant difference in Average Power (mV) is found.
What might explain these findings? If indeed trees provide perching opportunities from which birds and squirrels to call, then measuring higher values of average power
(which is perceived as loudness by human hearing) at the site with more trees is predictable. This is also true for geophony sounds as more tree provide more obstacles with which to intercept wind. However it was expected that if a difference in human- sourced sounds was detected that higher average power values would be found at the site with few trees due to the diminished scattering of sounds by trees. The fact that more average power from human-sourced sounds was found at Stratford Rd. seems to be due to the chance presence of more passing cars and overhead planes.
t-Test for Peak Power (mV) by sound type.
The Peak Power sound attribute reports on the point of maximum power spectral density within a sound selection. One advantage of this measure is that it is not particularly sensitive to the analyst’s placement of the sound selection boundaries.
As previously discussed, sound power and energy is reported within the Raven program with the units of decibels (dB), which is a logarithmic scale. Before adding the individual Peak Power values for each sound selection the logarithmic decibel values
260 were converted to straight scale millivolt (mV) values. Table XVII presents the results for the Peak Power (mV) t-Tests for each of the sound types and sub-types.
The results for the Stratford Rd. and Prindle Ave. May 19, 2015 recordings show that there is a statistically significant difference in the mean Peak Power (mV) values for the mechanical (Ax) sub-type and for all anthrophony combined (A all), but not for the human voices and music (Av) sub-type. This indicates the human mechanical sound selections from the site with many trees (Stratford Rd.) had higher mean Peak Power values than at the site with few trees (Prindle Ave.).
The results for biophony were consistent with the Average Power results. When considering the biophony sub types individually (Bb, & Bm) and collectively (B all) there were statistically significant differences in the mean Peak Power (mV) values. As with human mechanical sounds, this indicates the animal-sourced sound selections from the site with many trees (Stratford Rd.) had higher mean Peak Power values than at the site with few trees (Prindle Ave.).
No statistically significant differences in mean Peak Power values were found in geophony or in the silence samples.
Once again, the results here for biophony are as expected. The results for human mechanical sounds (Ax) are somewhat unexpected. It was also expected that sounds associated with wind (G) would be louder at the site with many trees, and this proved not to be significant, although if the one-tailed t-Test result is used (which is reasonable because of the expected results of the test) the difference would have barely achieved significance (p = 0.0491).
261
Table XVII: Peak Power (mV) sound attribute t-Test results for anthrophony, biophony, and geophony sound types for the temporally matched sound recordings at Stratford Rd. and Prindle Ave. on May 19, 2015. Green shading indicates statistically significant (alpha = 0.05) greater peak power mean values for Stratford Rd. (many trees)
262
t-Test for Energy (mV) by sound type.
The Energy sound attribute sums the power spectral density for all bins within a sound selection. As with Average Power, one disadvantage of this measure is that it is somewhat sensitive to the analyst’s placement of the sound selection boundaries.
As previously discussed, sound power and energy is reported within the Raven program with the units of decibels (dB), which is a logarithmic scale. Before adding the individual Average Power values for each sound selection the logarithmic decibel values were converted to straight scale millivolt (mV) values. Table XVIII presents the results for the Energy (mV) t-Tests for each of the sound types and sub-types.
The results for the Stratford Rd. and Prindle Ave. May 19, 2015 recordings show that there is no statistically significant difference in the mean Energy (mV) values for the anthrophony: voices and music sub-type (Av), for the mechanical (Ax) sub-type and for all anthrophony combined (A all).
When considering the biophony sub types individually (Bb, & Bm) and collectively
(B all) there were statistically significant differences in the mean Energy (mV) values.
Unlike the human-sourced sounds, this indicates the animal-sourced sound selections from the site with many trees (Stratford Rd.) had higher mean Energy values than at the site with few trees (Prindle Ave.).
No statistically significant differences in mean Energy values were found in geophony or in the silence samples.
263
Table XVIII: Energy (mV) sound attribute t-Test results for anthrophony, biophony, and geophony sound types for the temporally matched sound recordings at Stratford Rd. and Prindle Ave. on May 19, 2015. Green shading indicates statistically significant (alpha = 0.05) greater energy mean values for Stratford Rd. (many trees)
Once again, the results here for biophony are as expected. The non-significant results for human mechanical sounds (Ax) are also somewhat expected, although measuring greater sound energy with human mechanical sounds due to decreased sound scattering at
264 the site with few trees was also possible (but not demonstrated). It was also expected that sounds associated with wind (G) would be louder at the site with many trees, and this proved not to be significant.
t-Test for Average Entropy by sound type.
As previously discussed entropy is a measure of disorder. Highly ordered sounds such as the monotonal sound of a bell have very low entropy. Highly disordered sounds such as the sound of rain falling on the ground or wind rustling tree leaves have high entropy.
Entropy is related to but not synonymous with complexity. Sounds with very high entropy and very low entropy tend to have low complexity. Sounds that are highly complex such as human speech or the call of a house wren are toward the middle of the entropy scale. Complex sounds tend to carry more translatable information than do simple sounds.
Unlike the Average Power sound attribute than averages the values from every pixel or bin within a sound selection, the Average Entropy sound attribute sums the entropy values for each time frame within a sound selection (that is each dice of time over all frequencies within the selection without additionally being sliced by narrow frequency bands). As with Average Power, one disadvantage of the Average Entropy measure is that it is somewhat sensitive to the analyst’s placement of the sound selection boundaries.
Table XIX presents the results for the Average Entropy t-Tests for each of the sound types and sub-types.
265
Table XIX: Average Entropy sound attribute t-Test results for anthrophony, biophony, and geophony sound types for the temporally matched sound recordings at Stratford Rd. and Prindle Ave. on May 19, 2015. Blue shading indicates statistically significant (alpha = 0.05) greater Average Entropy mean values for Prindle Ave. (few trees).
The results for the Stratford Rd. and Prindle Ave. May 19, 2015 recordings show that there is a statistically significant difference in the mean Average Entropy values for the anthrophony: voices and music sub-type (Av), for the mechanical (Ax) sub-type and for
266 all anthrophony combined (A all). This indicates the human-sourced sound selections from the site with few trees (Prindle Ave.) had higher Average Entropy values on average than at the site with many trees (Stratford Rd.).
When considering the biophony sub types individually (Bb, & Bm) and collectively
(B all) there were statistically significant differences in the mean Average Entropy values. As with human-sourced sounds, this indicates the animal-sourced sound selections from the site with few trees (Prindle Ave.) had higher mean Average Entropy values than at the site with many trees (Stratford Rd.).
Statistically significant differences in mean Average Entropy values were also found in geophony and in the silence samples.
What might be an explanation for this strongly significant increase in Average
Entropy values at the site with fewer trees? It is not explained by living organisms changing the complexity or order of their calls in response to changing tree canopy. If this were the case there would not be a difference in mechanical sounds between the sites.
Since entropy is a function of the range of frequencies over which sound energy is measured it could be that with fewer trees at the Prindle Ave. site to scatter sound energy the higher frequency energy that tends to be scattered by trees is being detected by the recorder. This is somewhat consistent with the t-Tests results for Delta Frequency,
Center Frequency and Peak Frequency (with the exception of the Center Frequency result for the silence sample). The fact that the significance of the Average Entropy results are so strong is puzzling.
267
t-Test for Aggregate Entropy by sound type.
Unlike the Average Entropy sound attribute than averages the values from every time frame within a sound selection, the Aggregate Entropy sound attribute sums the entropy values for each time frame within a sound selection. As with Energy (the sound attribute that sums all of the power spectral density), one disadvantage of the Aggregate Entropy measure is that it is somewhat sensitive to the analyst’s placement of the sound selection boundaries.
Table XX presents the results for the Aggregate Entropy t-Tests for each of the sound types and sub-types.
The results for the Aggregate Entropy analysis are consistent with those for the
Average Entropy t-Tests. There is a statistically significant difference in the mean
Average Entropy values for the anthrophony: voices and music sub-type (Av), for the mechanical (Ax) sub-type and for all anthrophony combined (A all). This indicates the human-sourced sound selections from the site with few trees (Prindle Ave.) had higher
Aggregate Entropy values on average than at the site with many trees (Stratford Rd.).
When considering the biophony sub types individually (Bb, & Bm) and collectively
(B all) there were statistically significant differences in the mean Aggregate Entropy values. As with human-sourced sounds, this indicates the animal-sourced sound selections from the site with few trees (Prindle Ave.) had higher mean Aggregate Entropy values than at the site with many trees (Stratford Rd.).
Statistically significant differences in mean Aggregate Entropy values were also found in geophony and in the silence samples.
268
Table XX: Aggregate Entropy sound attribute t-Test results for anthrophony, biophony, and geophony sound types for the temporally matched sound recordings at Stratford Rd. and Prindle Ave. on May 19, 2015. Blue shading indicates statistically significant (alpha = 0.05) greater Aggregate Entropy mean values for Prindle Ave. (few trees).
Once again, the fact that the significance of the Average Entropy results are so strong is puzzling.
269
t-Test for Maximum Entropy by sound type.
The Maximum Entropy sound attribute reports the value for the time frame within a sound selection with the greatest entropy. One advantage of the maximum entropy measure is that it is not sensitive to the analyst’s placement of the sound selection boundaries.
Table XXI presents the results for the Maximum Entropy t-Tests for each of the sound types and sub-types.
The results for the Maximum Entropy analysis are consistent with those for the
Average Entropy and Aggregate Entropy t-Tests. There is a statistically significant difference in the mean Maximum Entropy values for the anthrophony: voices and music sub-type (Av), for the mechanical (Ax) sub-type and for all anthrophony combined (A all). This indicates the human-sourced sound selections from the site with few trees
(Prindle Ave.) had higher Maximum Entropy values on average than at the site with many trees (Stratford Rd.).
When considering the biophony subclasses individually (Bb, & Bm) and collectively
(B all) there were statistically significant differences in the mean Maximum Entropy values. As with human-sourced sounds, this indicates the animal-sourced sound selections from the site with few trees (Prindle Ave.) had higher mean Maximum Entropy values than at the site with many trees (Stratford Rd.).
Statistically significant differences in mean Maximum Entropy values were also found in geophony and in the silence samples.
270
Table XXI: Maximum Entropy sound attribute t-Test results for anthrophony, biophony, and geophony sound types for the temporally matched sound recordings at Stratford Rd. and Prindle Ave. on May 19, 2015. Blue shading indicates statistically significant (alpha = 0.05) greater Maximum Entropy mean values for Prindle Ave. (few trees).
Once again as with Average Entropy and Aggregate Entropy, the fact that the significance of the Maximum Entropy results are so strong is puzzling.
271
t-Test for Delta Entropy by sound type.
The Minimum Entropy sound attribute reports the value for the time frame within a sound selection with the least entropy. As with maximum entropy, one advantage of the minimum entropy measure is that it is not sensitive to the analyst’s placement of the sound selection boundaries.
Rather than complete a series of t-Tests for the Minimum Entropy values I decided to compare the difference between the Maximum Entropy and Minimum Entropy values for each sound selection. This Delta Entropy value is not directly calculated by the Raven program, rather it is calculated within the Excel program prior to statistical analysis. The
Delta Entropy value provides a measure of the range of entropy found within a single sound signature.
Table XXII presents the results for the Delta Entropy t-Tests for each of the sound types and sub-types.
The results for the Delta Entropy analysis are not completely consistent with those for the Average Entropy, Aggregate Entropy and Maximum Entropy t-Tests. There is a statistically significant difference in the mean Delta Entropy values for the mechanical
(Ax) sub-type and for all anthrophony combined (A all), but not for the anthrophony voices and music sub-type (Av). This indicates the human mechanical sound selections from the site with many trees (Stratford Rd.) had higher mean Delta Entropy values than at the site with few trees (Prindle Ave.). This is opposite of the significant results found in the other measures of entropy for Ax and A all.
272
Table XXII: Delta Entropy sound attribute t-Test results for anthrophony, biophony, and geophony sound types for the temporally matched sound recordings at Stratford Rd. and Prindle Ave. on May 19, 2015. Blue shading indicates statistically significant (alpha = 0.05) greater Maximum Entropy mean values for Prindle Ave. (few trees); Green shading indicates statistically significant (alpha = 0.05) greater energy mean values for Stratford Rd. (many trees).
When considering the biophony sub types individually (Bb, & Bm) and collectively
(B all) there were statistically significant differences in the mean Delta Entropy values.
273
Consistent with the results from the other measures of entropy, this indicates the animal- sourced sound selections from the site with few trees (Prindle Ave.) had higher mean
Delta Entropy values than at the site with many trees (Stratford Rd.).
Comparison of Geophony sounds found no significant difference for Delta Entropy. A statistically significant difference in mean Delta Entropy was found in the silence samples, indicating a greater range of entropy for the silence samples recorded at the site with many trees.
Summary of t-Test analysis for mean values of sound attributes.
Figure 52 presents a summary of the t-Test results for each of the sound attributes by each of the sound types and sub-types for the May 19, 2015 temporally matched pair of sound recordings at Stratford Rd. (many trees) and Prindle Ave. (few trees).
Figure 52: Summary of the sound attribute t-Test results for anthrophony, biophony, and geophony sound types for the temporally matched sound recordings at Stratford Rd. and Prindle Ave. on May 19, 2015. Blue shading indicates statistically significant (alpha = 0.05) greater mean values for Prindle Ave. (few trees); Green shading indicates statistically significant (alpha = 0.05) greater mean values for Stratford Rd. (many trees).
The trends for this pair of sound recordings are fairly clear. The measures of sound attributes related to frequency (Delta Frequency, Center Frequency and Peak Frequency)
274 indicate that for human-sourced sounds (Av, Ax, and A all), for animal-sourced sounds
(Bb, Bm and B all), and to a lesser extent weather-related sounds (G all), the average frequency of sounds is higher at the site with few trees. This may be due to less scattering of sounds by trees at the site with few trees, recognizing the fact that trees scatter sounds most efficiently above 1,000 Hz to about 8,000 Hz.
The measures of sound attributes related to power and energy (perceived as loudness)
(Average Power, Peak Power and Energy) indicate that for human-sourced sounds (Av,
Ax, and A all), for animal-sourced sounds (Bb, Bm and B all), and to a lesser extent weather-related sounds (G all), the average power spectral density of sounds is higher at the site with many trees. In other words the soundscape at the site with many trees sounds louder to many people than the site with few trees. This is unexpected for human mechanical sounds since there are more trees to scatter sounds at the Stratford Rd. site.
This is somewhat expected for the animal-sourced sounds since there are more trees in which birds and squirrels can perch and call from. It is expected for wind-related sounds since there are more trees to intercept the wind at Stratford Rd.
It is unexpected to find no difference in the measures for power spectral density for the silence samples. It was expected that periods of silence would be quieter at the site with many trees due to the scattering of background sounds. Perhaps this is due to the fact that the vast majority of sound energy during periods of relative silence are in the low frequency ranges that are not effectively scattered by trees.
Strongly significant differences for three of the four measures of entropy (Average
Entropy, Aggregate Entropy, and Maximum Entropy) were found between the site with many trees compared to the site with few trees. In all sound types and subtypes,
275 including the silence samples measures of entropy were greater at the site with few trees.
This is unexpected in general and for geophony sounds in particular since the sounds with typically the highest entropy values are those associated with wind and precipitation intercepting solid objects including trees. The relative strength of sound entropy as a predictor of tree canopy density is not yet understood.
Are the t-Test findings for this temporally matched pair of recordings consistent with the rest of the pairs of recordings? Figures 53a through 53f present the t-Test summaries for the six spatially matched pairs of sound recordings. Figures 54a through 54h present the t-Test summaries for the seven temporally matched pairs of sound recordings plus the pair of recordings that includes the site adjacent to the I-53 freeway.
276
Figure 53a through 53f: t-Test comparison of mean sound attribute values by sound type (anthrophony, biophony, and geophony) and sub-type for the six spatially matched sound recording pairs. Blue shading indicates statistically significant (alpha = 0.05) greater mean values for the site after trees were removed; Green shading indicates statistically significant (alpha = 0.05) greater mean values for the site before trees were removed.
Fig 53a
Fig 53b
Fig 53c
Fig 53d
277
Figure 53a through 53f (continued): Blue shading indicates statistically significant (alpha = 0.05) greater mean values for the site after trees were removed; Green shading indicates statistically significant (alpha = 0.05) greater mean values for the site before trees were removed. Light green cells at the Suffield Rd. site (65f) indicate statistically significant greater proportions of sound types in 2015 where little loss of tree canopy had occurred in the immediate area but substantial tree loss had occurred within the larger neighborhood as compared to the same site in 2014 prior to tree removal in the larger neighborhood.
Fig 53e
Fig 53f
278
Figure 54a through 54h: t-Test comparison of mean sound attribute values by sound type (anthrophony, biophony, and geophony) and sub-type for the eight temporally matched sound recording pairs. Blue shading indicates statistically significant (alpha = 0.05) greater mean values for sites with few trees); Green shading indicates statistically significant (alpha = 0.05) greater mean values for sites with many trees.
Fig 54a
Fig 54b
Fig 54c
Fig 54d
279
Figure 54a through 54h (continued): Blue shading indicates statistically significant (alpha = 0.05) greater mean values for sites with few trees; Green shading indicates statistically significant (alpha = 0.05) greater mean values for sites with many trees. Gray bars at the Wilke Rd. site (66h) indicate statistically significant greater mean values adjacent to freeway compared to the Huron St. site (blue cells) after many trees had been removed.
Fig 54e
Fig 54f
Fig 54g
Fig 54h
280
Soundscape Analysis Summary
Differences and similarities in the six spatially matched pairs of sound recordings and the eight temporally matched pairs of sound recordings were compared using four different methods:
1. One-third octave slice and dice method using t-Tests to compare means of
uniform sound selections without regard to individual sound sources. Statistically
significant differences (alpha = 0.5) were noted.
2. A detailed analysis of each sound recording was performed in which individual
sound selections were identified by sound type (anthrophony, biophony, and
geophony) and sub-type. The values from the sound attributes of Delta Time,
Delta Frequency, Energy (dB), Energy (mV) and Aggregate Entropy were
summed for the six pairs of spatially matched pairs of sound recordings and the
eight pairs of temporally matched sound recordings. The results were graphed.
This exercise did not provide a statistical analysis that identified significant
differences, it simply graphed the data.
3. Using the sound attribute data from the detailed identification of individual
sounds, a statistical analysis was performed for each pair of sound recordings to
test for changes in the proportions of 5 sound attributes within sound types and
sub-types as tree canopy density changed. The statistical tests for differences in
proportions were accomplished by calculating Z-scores and using the Z-scores to
estimate probability of statistically significance (alpha = 0.05) through the
identification of p-values. These tests revealed where the proportion of human-
281
sourced sounds (anthrophony) changed in relation to animal-sourced sounds
(biophony) and weather-related sounds (geophony).
4. Using the sound attribute data from the detailed identification of individual
sounds, a statistical analysis was performed for each pair of sound recordings to
test for changes in the mean values of 12 sound attributes within sound types and
sub types as tree canopy density changed. The statistical tests for differences in
mean values of sound attribute data were accomplished by conducting t-Tests
(alpha = 0.05). These tests revealed where the mean values for 12 sound
attributes of human-sourced sounds (anthrophony), animal-sourced sounds
(biophony), weather-related sounds (geophony) and samples of relative silence
changed as tree canopy changed.
What did these exercises and tests reveal? Do the anthrophony, biophony and geophony components of a soundscape change as tree cover changes? The shortest answer is “yes.” A better answer is “yes, but the successful detection of change depends on how you measure the sounds.”
Summary of one-third octave slice-and-dice analysis.
Sounds can be compared by measures that fall into the four basic attributes of time, frequency, energy or intensity (perceived as loudness), and entropy. The one-third octave slice-and-dice analysis considered only the sound attribute of Energy, and only one method of measuring energy, that being a measure of total energy within each sound selection. This analysis did not involve the identification of individual sounds sources
282 and so there is no breakdown of sound sources into anthrophony, biophony and geophony. However an advantage to this approach is that there is no overlap between the sound selections and the array of sound selections cover all time and all frequencies, therefore all sound energy is accounted for once and only once. This method clearly shows that there are statistically significant differences in total energy for most of the one third octave frequency slices for all of the pairs of sound recordings, both the spatially matched pairs (Figure 42a) and the temporally matched pairs (Figure 42b). However there is not a consistent increase (or decrease) in total energy as the density of tree cover changes. For the pairs of spatially matched recordings (collected at the same site at different times) the evidence points to a consistent increase in total sound energy as trees are removed over time. For the pairs of temporally matched recordings (collected at the same time at different locations) the evidence is mixed. In the low frequency bands below approximately 1,800 Hz some recording pairs exhibit greater sound energy at the sites with many trees and some sites exhibit greater sound energy at the sites with few trees. As the sound frequency increases the evidence points to more sound energy detected at the sites with many trees. It is interesting to observe that these frequency bands are those typically occupied by animals to communicate, so this analysis may be relative to biophony, but the other three methods of analysis provide clearer information relative to sound sources.
283
Summary of total values for four sound attributes.
Following the detailed analysis of the sound recordings in which individual sounds were identified by type (anthrophony, biophony, and geophony) and sub-types, the data for the sound attributes Delta Time, Delta Frequency, Energy (dB), energy (mV) and
Aggregate energy were summed and graphed for the six spatially matched recording pairs and for 7 temporally matched pairs. This exercise has the advantage of considering more sound attributes than just total energy, with the additional advantage of having sounds identified by sound type. The disadvantage to this exercise is that while it visually displays trends in the data, it provides no indication of statistical significance of differences. Figure 55 presents a basic summary of this tally. Blue cells indicate greater total values for sites with few trees; green cells indicate greater total values for sites with many trees.
Figure 55: Summary of total values for the sound attributes Delta Time, Delta Frequency, Energy (mV), and Aggregate Entropy for all individual sound selections in the spatially matched recording pairs and the temporally matched recording pairs. Blue cells indicate
greater total values for sites with few trees; green cells indicate greater total values for sites with
many trees. This represents sums of values and the results do not indicate statistically significant differences.
284
This most basic of summaries suggests that when considering the sound attribute of time, sounds associated with humans are heard more frequently in areas with few trees, and sounds associated with non-living sources such as wind are heard more frequently in areas with many trees. Sounds associated with animals seem to be heard more frequently after trees are removed for the spatially matched pairs, but at sites with more trees for the temporally matched pairs.
When considering the sound attribute of frequency, sounds associated with humans and sounds associated with animals are heard over a greater range of frequencies in areas with few trees, and sounds associated with non-living sources such as wind are heard over a greater range of frequencies in areas with many trees.
When considering the sound attribute of energy (which is perceived as loudness), sounds associated with humans and sounds associated with animals seem to be louder after trees are removed for the spatially matched pairs, but louder at sites with more trees for the temporally matched pairs. Sounds associated with non-living sources such as wind seem to be louder in areas with many trees.
When considering the sound attribute of entropy, sounds associated with humans and sounds associated with animals have greater disorder in areas with few trees, and sounds associated with non-living sources such as wind have greater disorder in areas with many trees.
Are any of these results as expected? Yes. As expected geophony, those non-living sourced sounds associated with wind show consistently greater values for time, frequency range, energy, and entropy in areas with many trees. With more trees that are on average
285 closer to the receiver it is expected that there will be more sounds of wind being intercepted by trees.
Because of the sound energy scattering properties of trees, and because the effectiveness of this scattering increases above 1,000 Hz, it was expected that measures for frequency, energy and possibly entropy would be greater for human-sourced sounds in areas with few trees. This appears to be the case except for energy at the temporally matched sites, where human-sourced sounds seem to be louder at the sites with many trees. Are these long-distance background sounds or close-distance cars and lawn mowers passing by? The analysis for differences in mean values will answer this.
Because of the greater number of sites for birds, insects, and tree frogs to perch on and call from it was expected that measures of time, frequency, and energy would be greater for animal-sourced sounds (biophony) at the sites with many trees. The results for biophony are mixed and somewhat confusing. Measures for biophony time and energy are conflicting between spatially and temporally matched pairs of recordings. Measures for frequency range are consistent but in the opposite direction as expected as biophony seems to occupy a greater range of frequencies at the sites with few trees. Could the diminished number of scattering obstacles be allowing a greater range of frequencies to reach the receiver? Is this function stronger than having calling animals closer to the receiver as would be expected with more trees to perch from? The answers to these questions are unknown.
If the sound scattering properties of trees is relatively strong then expectations might be different. Because of the sound energy scattering properties of trees, and because the effectiveness of this scattering is greatest in the frequency range occupied by many
286 calling animals, it might be expected that measures for frequency, energy and possibly entropy would be greater for animal-sourced sounds in areas with few trees. This appears to be the case except for energy at the temporally matched sites, where human- sourced sounds seem to be louder at the sites with many trees. Are these long-distance background sounds or close-distance cars and lawn mowers passing by? The analysis of mean values of sound attributes will answer this.
Since entropy is a function of frequency range, it was somewhat expected that the pattern for measures of entropy would follow the pattern for measures of frequency, which is indeed the case.
Summary of proportional values for four sound attributes.
One of the most important questions to be investigated is “Do the proportions of sounds within a soundscape change when tree cover changes?” When considering the potential effects of soundscapes on human directed attention this question may be modified by asking “Does the proportion of human-sourced sounds as compared to sounds associated with nature change as tree cover changes?” These questions are addressed by the detailed analysis of the sound recordings followed by the statistical analysis of proportions using Z-scores that lead to p values.
The proportional analysis considers four sound attributes (Delta Time, Delta
Frequency, Energy (mV), and Aggregate Entropy) for the spatially matched recording pairs and the temporally matched recording pairs. Figure 56 presents the summary of sound attribute proportions. Blue cells indicate more recording pairs with statistically
287 significant greater proportional values for sites with few trees; green cells indicate more recording pairs with statistically significant greater proportional values for sites with many trees; white cells indicate no statistical significance in proportional differences.
Figure 56: Summary of sound attribute proportions for Delta Time, Delta Frequency, Energy (mV), and Aggregate Entropy for the spatially matched recording pairs and the temporally matched recording pairs. Blue cells indicate a greater number of matched recording pairs with significantly higher proportional values for sites with few trees; green cells indicate a greater number of matched recording pairs with significantly higher proportional values for sites with many trees. White cells indicate no difference in the number of matched recording pairs with significantly higher proportional values.
The proportional analysis does not report on how the total values of sound attributes compare and instead reports on how the proportions of anthrophony, biophony, and geophony compare. A property of proportions is that if the proportion of one sound type increases then the proportion of one or both of the other sound types must decrease.
(Making one slice of a pie bigger means that one or more of the remaining slices gets smaller.)
This summary of proportions suggests that when considering the sound attribute of time, sounds associated with humans and with animals are heard less frequently in areas with few trees, and sounds associated with non-living sources such as wind are heard
288 more frequently in areas with many trees (at least for the temporally matched recording pairs).
When considering the sound attribute of frequency, sounds associated with humans and sounds associated with animals are heard in greater proportions of the audible frequency range in areas with few trees, and sounds associated with non-living sources such as wind are heard in greater proportions of the audible frequency range in areas with many trees.
When considering the sound attribute of energy (which is perceived as loudness), as expected the proportion of sound energy associated with non-living sources such as wind is proportionally greater in areas with many trees (at least for the temporally matched recording pairs). Sound energy associated with human sources is proportionally greater in areas with few trees (at least for the temporally matched recording pairs). The proportion of sound energy associated with animals seems to shift in different directions for the spatially matched pairs of recordings and for the temporally matched pairs. When considering the temporally matched pairs, the proportion of sound energy associated with animals and wind is greater at sites with more trees and the proportion of sound energy associated with human sounds is greater at sites with few trees. Another way of stating this in more general terms is that sounds associated with humans seem to be proportionally louder at the sites with few trees, and the sounds associated with animals and non-living sources such as wind seem to be proportionally louder in areas with many trees. This seems to be the case for the temporally matched pairs of recordings but not for the spatially matched pairs of recordings.
289
When considering the sound attribute of entropy, where there are statistically significant proportional differences, sounds associated with humans and sounds associated with animals have greater disorder in areas with few trees, and sounds associated with non-living sources such as wind have greater disorder in areas with many trees.
Are any of these results as expected? Yes. As expected geophony, those non-living sourced sounds associated with wind show consistently greater proportional values for time, frequency range, energy, and entropy in areas with many trees. With more trees that are on average closer to the receiver it is expected that there will be proportionally more sounds of wind being intercepted by trees.
As expected, when considering human-sourced sounds (anthrophony), where statistically significant results exist there are more instances of proportionally greater measures for time, frequency, energy and entropy at sites with few trees.
Somewhat unexpectedly, when considering animal-sourced sounds (biophony), where statistically significant results exist there are more instances of proportionally greater measures for time, frequency, and entropy at sites with few trees. For the sound attribute of energy the results are conflicting. There are more instances of proportionally greater energy after trees are removed for the spatially matched pairs of recordings. There are more instances of proportionally greater energy at sites with many trees for the temporally matched pairs of recordings.
The proportional mix of sounds is highly dependent on season and time of day. The proportion of biophony greatly increases (and therefore the proportion of anthrophony and geophony decreases) in the spring when birds are most actively vocalizing. This is
290 particularly noticeable during the morning and evening choruses. The soundscapes of summer have a much higher proportion of biophony of the insect variety due to the stridulation of cicadas during the day and of katydids and crickets at night. For a brief period in spring the amphibians are conspicuous in localized areas. This temporal variability in animal-sourced sounds is in constant proportional flux with human-sourced sounds. Spring and summer is accompanied by frequent noise from landscape maintenance machines, many of which are irritatingly loud. Even the background sounds are affected by diurnal fluctuations in vehicle and air traffic. There is evidence that the proportion of sounds within a soundscape is affected by the density of tree cover, but the magnitude of the effect may be subtle compared to the magnitude of seasonal and diurnal fluctuations.
Summary of mean values for four sound attributes.
The next summary considers mean values of the individual sound selections for four sound attributes (Delta Time, Delta Frequency, Energy (mV), and Aggregate Entropy) for the spatially matched recording pairs and the temporally matched recording pairs. Figure
57 presents the summary of mean sound attribute values. Blue cells indicate more recording pairs with statistically significant greater mean values for sites with few trees; green cells indicate more recording pairs with statistically significant greater mean values for sites with many trees; white cells indicate no statistical significance in mean value differences.
291
Figure 57: Summary of sound attribute mean values for Delta Time, Delta Frequency, Energy (mV), and Aggregate Entropy for the spatially matched recording pairs and the temporally matched recording pairs. Blue cells indicate a greater number of matched recording pairs with significantly higher mean values for sites with few trees; green cells indicate a greater number of matched recording pairs with significantly higher mean values for sites with many trees. White cells indicate no difference in the number of matched recording pairs with significantly higher mean values.
In the comparison of mean values there is considerable inconsistency between the results for the spatially matched recording pairs and for the temporally matched recording pairs. The results for the spatially matched recording pairs resemble the results for the first comparison of total values. The results for the temporally matched recording pairs show less resemblance to the first comparison of total values and suggests nearly the opposite results for mean values as does the spatially matched pair.
When considering the sound attribute of time, both spatial and temporal pairs suggest no significance difference in sounds associated with humans. For sounds associated with animals the results conflict. For sounds associated with non-living sources such as wind the results are consistent and suggest these sounds are heard more frequently on average in areas with many trees.
For the spatially matched pairs, when considering the sound attribute of frequency, sounds associated with humans and sounds associated with animals occur on average
292 over a greater range of frequencies in areas with few trees, and sounds associated with non-living sources such as wind are heard over a greater range of frequencies in areas with many trees. For the temporally matched pairs the results are just the opposite.
Sounds associated with humans and sounds associated with animals occur on average over a greater range of frequencies in areas with many trees, and sounds associated with non-living sources such as wind are heard over a greater range of frequencies in areas with few trees.
For the spatially matched recording pairs, when considering the sound attribute of energy, sounds associated with humans and sounds associated with animals occur on average with more energy (sound louder) in areas with few trees, and sounds associated with non-living sources such as wind occur on average with more energy (sound louder) in areas with many trees. For the temporally matched recording pairs sounds associated with humans are inconclusive. Sounds associated with animals and with non-living sources such as wind occur on average with more energy in areas with many trees.
For the spatially matched pairs mean values for entropy are greater for anthrophony, biophony and geophony before trees are removed. For the temporally matched pairs mean values for entropy show the opposite trend suggesting greater entropy at the sites with few trees.
293
Summary of mean values for eight sound attributes in silence samples.
It is also enlightening to consider the one-second samples of silence identified in each sound recording. Figure 58 presents a summary analysis of silence samples for the eight sound attributes for the spatially matched recording pairs and the temporally matched recording pairs. Blue cells indicate more instances of statistically significant greater mean values for sites with few trees; green cells indicate more instances of statistically significant greater mean values for sites with many trees; white cells indicate no statistical significance in mean value differences.
Figure 58: Summary of eight sound attribute mean values for silence samples for the spatially matched recording pairs and the temporally matched recording pairs. Blue cells indicate a greater number of matched recording pairs with significantly higher mean values for sites with few trees; green cells indicate a greater number of matched recording pairs with significantly higher mean values for sites with many trees. White cells indicate no difference in the number of matched recording pairs with significantly higher mean values.
For the silence samples there are no identifiable sources of sound other than the low frequency background sounds, which are primarily from distant vehicle traffic. There is evidence of increased levels of background noise in the samples of silence from sites with few trees. In the 6 spatially matched recordings pairs more sound energy was detected at
5 sites after trees were removed. One site showed no significance difference in sound energy. For the seven temporally matched recording pairs, 2 sites with few trees experienced significantly higher levels of sound energy and 2 sites with many trees had
294 higher levels of sound energy for the silence samples. Three pairs showed no significant difference. The evidence for both the spatially matched and temporally matched pairs of recordings is consistent for peak power, suggesting that the background noise at sites with few trees is more intense than the background noise and sites with many trees.
Other sound attributes showed similar trends. For the spatially matched pairs, 4 of the
6 sites had significantly greater center frequency values after trees were removed while the other two pairs of recordings show no significant difference relative to tree cover.
For the temporally matched pairs the trend is reversed, 6 of the 7 sites with many trees had significantly greater center frequency values than the sites with few trees. The other pair shows no significant difference. Why might this be? Since these silence samples have no other sounds present than the background sound, there are no animals calling that could potentially shift the frequency of their calls, and all sources for the background sounds are at a considerable distance from the source. Also the sound frequencies that trees are most efficient at scattering are above the frequencies present in the background noise. I have no plausible explanation for center frequency on average being higher at the sites with more trees.
For the spatially matched pairs, 4 of the 6 sites had significantly greater aggregate entropy values before trees were removed while the other 2 sites had significantly greater aggregate entropy values after trees were removed. For the temporally matched pairs the trend is reversed. Only 1 of the 7 sites with many trees had significantly greater aggregate entropy values while 5 of the 7 sites with few trees had greater entropy. One pair of sites had no significant difference in entropy.
295
Common patterns revealed by sound recording analysis.
Obviously the results of the tests are mixed, but patterns from the comparison of matched recording pairs emerge. The evidence can be summarized by sound type:
Geophony: The most consistent trend is with geophony – those sounds from non-
living sources usually associated with weather. These sounds are heard more
frequently, over a wider range of frequencies, with greater intensity, and with
greater entropy at sites with many trees. This is as expected. With more trees
present there is more opportunity for wind and precipitation to be intercepted.
Anthrophony: The trend with anthrophony is fairly consistent, and in particular
with the human-sourced mechanical sounds. In many, but not all soundscapes
these sounds are heard more frequently, over a wider range of frequencies, and
with greater intensity, at sites with few trees. This is as expected. With fewer
trees in a landscape there is less opportunity for sound energy to be scattered and
attenuated, particularly in frequencies greater than 1,000 Hz.
Biophony: The trends for animal-sourced sounds relative to tree density are not
consistent when considering all animal sounds collectively. The most important
variables that determine the presence of biophony are season and time of day.
Birds sing with greater duration and intensity in spring particularly in morning
and evening, but can be heard to varying degrees from dawn to dusk throughout
the year. Since most urban and suburban songbirds do not nest in large
landscape trees the presence and species diversity of songbirds did not seem to
be affected by the removal of ash trees.
296
Not all stridulating insects call from trees but some of the most noticeable
including the cicadas and katydids call almost exclusively from trees. These
insects are usually present from mid-summer to early fall but are completely
absent from the soundscape for most of the year. Few amphibians call from
trees, but tree frogs were a minor component to the recorded soundscapes during
spring to mid-summer. Mammals most frequently heard are dogs (that do not
rely on trees) and squirrels (that heavily rely on trees). Mammal vocalizations
are less seasonally dependent than other animals. Table XXIII presents a list of
the bird, amphibian, insect and mammal species identified in the sound
recordings.
Silence: The only sound present in the samples of silence analyzed in the t-Tests
for means is the background noise, therefore the vast majority of the sound
present is attributable to mechanical sounds such as distant vehicle and aircraft
traffic. The evidence is mixed for measures of frequency and entropy between
the spatially matched and temporally matched pairs of recordings, but the
evidence for sound attributes associated with intensity suggest that the sites with
few trees have greater sound energy in the background than the sites with many
trees. In other words, the silence in areas with few trees is louder than the
silence in areas with many trees.
297
Table XXIII: Bird, amphibian, insect and mammal species identified in the sound recordings. Code refers to identification codes used in the Notes data field during sound analysis.
Areas for Improvement
I learned a lot about soundscapes, sound recording, and sound analysis during this research and there is certainly room for improvement to the methods. My first regret is not starting my search for an appropriate community sooner. Arlington Heights, Illinois proved to be an ideal setting due to their abundance of ash trees, the infestation of
Emerald Ash Borer and the aggressive removal of infested trees taken by the village.
298
Unfortunately there was only a period of several months between the time I began the research and the time that large-scale tree removal began, preventing me from getting a full year of pre-tree removal recordings collected.
I was unable to use several of the valuable pre-tree removal recordings because I failed to have the sound recorders set on the same recording settings. In two cases I was able to fully correct the data and in two other cases the corrections would have been close but not exact, so I chose not to use those recording pairs. I quickly learned to calibrate both recorders to the same settings before each recording session.
I would have preferred to use more of the recordings that I gathered in the analysis process. This limitation was purely a matter of time available. It took approximately 300 hours to complete the detailed listening analysis of the 14 recording pairs and approximately 200 hours to complete the statistical analysis once the sound-by-sound analysis was complete. Utilizing more than one sound analyst would have sped up the process but some of the sound selection consistency would have been sacrificed, at least in this initial research. It may be possible to build an automated sound recognition program similar to training image analysis programs to classify digital imagery. This is an objective for future attempts.
In spite of these challenges I am pleased that the outcome of the analysis demonstrates as many statistically significant differences in multiple sound attributes between soundscapes with many trees and those with few trees.
299
What Matters?
Noise matters to both animals and humans. As previously noted, Wiley (2015) describes noise as “sound that has no interest for us yet it makes sounds of interest hard to hear.” Therefore one definition of noise is sound energy that carries no information and interferes with communication between individuals or groups. According to Wiley one appropriate measure of noise is the amount of mistakes made by a receiver due to the corruption of the information within an auditory communication caused by noise.
Noise can also have negative effects outside of communication. Many animals rely on sound for locating their prey. Most predators like owls listen passively, but others including bats use active echo-location sound energy that is subject to interference from human sources of energy. In an interview for National Public Radio, sound researcher
Kurt Fristrup of the National Park Service reports that an increase in human background noise of 3 dB effectively reduces the area in which an owl can detect the sounds of its prey by half (McQuay & Joyce, 2016). Therefore it is not just the communication of sounds between individuals of the same species but also the sound information within a soundscape that animals use to exist in their environment.
In humans, noise can disrupt our attention, our sleep, and if excessive can lead to psychological or physical problems. It is no surprise then that sensory deprivation and sensory overload including sound have both been used as forms of torture. For humans, many sounds are subjective in terms of their usefulness or irritation, and therefore noise to one person may be music to another. If noise affects our ability to concentrate then one measure of noise may be the number of errors we make while working (or writing a
300 disserertation). Noise, therefore interferes with communication, which affects animals that communicate through sound (including humans). However for humans, the greatest negative impact from noise may be on how it affects our ability to function due to irritation and distraction.
Do the changes in urban soundscapes associated with loss of trees matter to humans, and if so what aspects of sound matter? Once again the answer lies in noise, but perhaps also with the sounds associated with nature. Human vocalization obviously carries the greatest amount of information between individuals of any organisms, and therefore is subject to the greatest amount of distortion and scattering from the environment.
Fortunately humans communicate at much closer range than do songbirds, therefore the scattering of the sounds of human voices by trees is of little concern. Of much greater concern is the transmission of human mechanical sounds – the sounds that carry no information and interfere with the transmission of information (fitting the definition of noise). Counteracting the distractive and irritating properties of noise may be the beneficial properties of sounds associated with nature.
Part 2 of this study investigates the question of the effect of changes to the soundscape on humans. If human directed attention is found to be affected by changes to the soundscapes related to tree loss it could be a function of several sound attributes:
Increase in loudness and/or proportion of human mechanical sounds that function
as noise
Decrease in the loudness and/or proportion of sounds associated with nature
(animal sounds and weather-related sounds) that may function as refreshers of
directed attention
301
While shifts in sound frequencies made necessary by increasing background noise may play an important role in animal communication it is less likely that perceived or subconscious changes in peak sound frequencies will affect human directed attention. It is not known what role changes to sound entropy will play in affecting human directed attention.
302
CHAPTER VIII
METHODS PART 2– TESTS FOR DIRECTED ATTENTION
The Stroop Effect
In his doctoral dissertation J. Ridley Stroop (1935) devised a test that measures the human brain’s ability to complete a task that is at odds with a slightly less demanding task. For most individuals who are proficient at reading, the brain easily recognizes the pattern of letters within a printed word without having to first recognize the individual letters and assemble them into the proper order. For most individuals, this rapid recognition of a printed word occurs faster than the recognition of a color. For example, when presented with the word GREEN, most English language reading-proficient individuals instantly recognize the word without reading the individual letters G-R-E-E-
N, and understand the meaning of the word. When the word GREEN is printed in the same color ink as the color that the word represents, such as GREEN, the brain instantly recognizes the word and the associated color. But for most people the brain acts slightly quicker in the recognition of the printed word than for the printed color.
303
The Stroop test is designed to measure this apparent difference in word / color recognition. When given the task of reporting on the color that they see, most individuals quickly respond to the color that is displayed as long as the printed color is not in conflict with the context of the printed word. However if the color is presented in an incongruent pattern, such as presenting the word GREEN in a color of ink that does not match the meaning of the word, the brain in most individuals has a momentary problem to solve, that being to report on the color of the ink (in this case red), which takes a sliver of time longer than to report on the meaning of the word. The brain must suppress the initial impulse to report the printed word (in this case green) in order to correctly report on the color it perceives.
This suppression of impulse in order to sort out a color requires a bit of concentration.
It requires a person to direct their attention toward the task. Therefore using the Stroop
Effect to test the strength of a person’s directed attention has become a standard method in psychological testing (De Young, 2014).
Administering the Stroop Test
Many versions of the Stroop Test have been developed, some of which simply count the number of answers that a participant correctly reports the color seen versus the number of answers that the written word is reported. In addition to knowing “correct” versus “incorrect” answers, it is useful to know the amount of time needed for a participant to report their answer.
304
I originally enlisted a computer programmer at the Davey Tree Expert Company to build a computer-based prototype version of the Stroop Test. The result presented the needed mix of colors and words but did not record answer times with the necessary precision. I showed the prototype program to Dr. Conor McLennan of the Psychology
Department at Cleveland State University, who offered advice on how the program might be improved. Dr. McLennan introduced me to Sara Incera, a Ph.D. candidate in psychology who was using the Stroop Test to measure participant’s response time in a language experiment. Ms. Incera had built her version of the Stroop Test using the
MouseTracker program.
The MouseTracker program (Freeman & Ambady, 2010) is computer-based and is used to present participants with a series of screens that require a response. Each screen begins with a small rectangle that appears at the bottom of the screen that includes the word “Start”. The participant uses the computer mouse to click on the word start, which causes a colored pattern to appear at the center of the screen and two words to appear in the upper corners of the screen. One of the words describes the color seen in the pattern and the other word describes a different color. For example, the pattern may be of four yellow stars in the center of the screen and the word in the upper left corner may be
“yellow” and the word in the upper right corner may be “blue”. The task for the participant is to click on the word “Start,” recognize the color in the center of the screen and quickly click on the word that correctly matches the color. Figure 59 illustrates this
305 example as presented by the MouseTracker program during the instruction phase of the
Stroop Test.
Figure 59: Example of an instruction screen from the Stroop Test.
While some of the colored patterns are of shapes or random letters, most of the colored patterns are of words. Some of these word / color combinations are congruent matches as the printed word is presented in the same color that the word represents. For example the word RED is shown in red ink (Figure 60a). Other word / color combinations are incongruent matches as the printed word is presented in a color that is different that the meaning of the word, such as the word GREEN shown in the color red
(Figure 60b).
Figure 60a (left) and 60b (right): Examples of a congruent word / color match in which the color of the print matches the meaning of the word (left) and an incongruent word / color match in which the color of the print does not match the meaning of the word (right).
306
The MouseTracker program gathers data for several measurements. First it records the time interval between the participant clicking on start and the participant clicking on one of the two word choices. Next, the program tracks the trajectory of the path that the participant moves the mouse between the Start word and the color word. Since the pathway that a participant moves the mouse between the Start word and the color word is rarely a straight line, the MouseTracker program reports on the deflection between the potential straight-line path and the actual path that the mouse is moved (Figure 61).
Figure 61: Example of computer mouse trajectory for an incongruent word / color combination. The MouseTracker program records the trajectory that a participant moves the computer mouse between the START word and the selected color word (GREEN), and compares the actual trajectory (yellow dashed arrow) to an alternate straight-line trajectory (solid blue arrow). The maximum deviation is recorded (dotted red arrow).
307
The MouseTracker program also records whether the color word selected correctly or incorrectly matches the color that the word is displayed.
Necker Cube Pattern Control
Like the Stroop Effect test, the Necker Cube Pattern Control test is a frequently used tool in psychological testing including tests for directed attention (De Young, 2016). The
Necker Cube Pattern Control Test is named for Louis Necker, who in the 1880’s noticed that drawings that represent three-dimensional cubes involuntarily change perspective when viewed for several seconds. For many people, when staring at a drawing of a cube, one of the cube faces will appear to be closest to them and then the perspective will shift so that a different face appears to now be closer (Figure 62). The frequency with which this apparent change in perspective is thought to be a useful measure of the strength of a person’s ability to direct their attention at that time.
Figure 62: The optical illusion cube graphic used in the Necker Cube Pattern Control test. For many people while staring at the cube (top) one of the cube faces will appear to be closer to the viewer (bottom left). The perspective will then occasionally flip so that a different face now appears to be closer to the viewer (bottom right).
308
The program used to present the Necker Cube Pattern Control test in this research is
Microsoft PowerPoint. The test consists of several slides in which the participant is instructed to simply stare at the cube on the computer monitor for three minutes and press the Enter key on the computer keyboard whenever the cube seems to change perspective.
Pressing the enter key simply causes the program to advance to the next identical slide.
There is no perceivable interruption in the projection of the cube. The participant is told when to start and when to stop the experiment. Following the trial, the researcher simply exits the slide show mode of the PowerPoint program and records the number of slides that have been advanced.
Enlisting Test Participants
The most difficult aspect of this stage of the research is attracting individuals to participate in the tests for directed attention. Testing rooms were set up at several regional and national conferences and events including:
Davey Institute of Tree Sciences – Kent, Ohio
Ohio Tree Care Conference – Sandusky, Ohio
National Catholic Cemetery Conference – Notre Dame University
National Cemetery Administration Landscape Management Training – St. Louis
Chautauqua Institution – Chautauqua, New York
International Society of Arboriculture Annual Conference – Ft. Worth, Texas
In addition to these organized events, I set up the tests for directed attention at many locations where I expected groups of people with similar interests to congregate, such as
309 at the School of Forest Resources and Environment at Michigan Technological
University and at multiple training sessions for professional arborists. I found that the most successful opportunities to enlist volunteers was following a presentation or training session that I conducted. Volunteers seemed grateful for the information that was provided and willing to return the favor by spending time in the research test.
At some events an incentive of a chance to win a $50 Amazon gift card was offered, but few people seemed to be motivated purely by the possibility of winning a prize.
Approximately half of the test participants took the tests as part of a small group.
Seven laptop computers were available during testing, so at times groups of between two and seven people would begin and end the tasks and tests at the same time. At other events people could enter the testing room at their convenience and would proceed through the tasks and tests individually. At times there were as many as five individuals involved in different stages of the testing at the same time. The testing rooms were always large enough to provide adequate space between individuals so as not to be distracting to one another.
Selecting the Sound Recordings for Play During Tests for Directed Attention
The tests for directed attention were taken while one of the recordings from Arlington
Heights was played. The initial question to be addressed was which of the recordings would be selected for play. With 14 pairs of matched recordings analyzed to great detail there was plenty of sound recording to choose from. The advantage to using the full range of many tree versus few tree recordings is that both temporally matched and
310 spatially matched recordings are included, and a full range of seasons are represented (at least with the temporally matched recording pairs). The disadvantage of using the full range of recording pairs is that each pair would only be likely to be represented at most a few times during the tests for directed attention, making the sample size for individual pairs of recordings very low. I therefore decided to select one pair of matched recordings for use in a subset of the tests for directed attention. I selected the temporally matched pair recorded on May 19, 2015 at the Stratford Rd. and Prindle Ave. sites. I chose this pair of recordings because they represented two distinctly different levels of tree canopy cover. The Prindle Ave. site had experienced the vast majority of street trees removed in the immediate area with approximately 30% of the overall tree canopy removed within approximately 100 yards (91 meters). The Stratford Rd. site had lost very few trees within approximately 100 yards because the majority of ash trees had been injected with a systemic insecticide that protected them from Emerald Ash Borer. Another advantage of using a temporally matched pair of recordings is that the weather conditions were consistent at both sites, which controls for an important variable that affects soundscapes.
The statistical analysis of proportions and mean values of sound category selections also demonstrated that there were subtle but statistically significant differences in measures of sound frequency, intensity and entropy between these recordings.
In addition to the temporally matched set of recordings from May 19, 2015 two additional sound treatments were used during the tests for directed attention. While the matched pair of recordings are different from an analytical perspective they are quite similar when simply listening to them. In spite of my very close inspection of both
311 recordings during the sound-by-sound analysis, I would be challenged to correctly identify the individual recordings by simply listening to them.
In an effort to investigate the effect on directed attention between two distinctly different soundscape recordings, the sound recording gathered on May 6, 2015 at the
Wilke Rd. site immediately adjacent to the Route 53 freeway was selected as the third sound treatment played during a subset of tests for directed attention. This recording clearly has a much higher proportion of traffic sounds and a much lower proportion of animal and weather-related sounds than do any of the other recordings. The sounds in this recording are on average much louder than in the pair of temporally matched recordings. To the casual listener, there is no difficulty in correctly identifying this recording as distinctly different from the other recordings.
Lastly, to serve as a control treatment, a subset of the tests for directed attention were administered with no sound recordings being played. The detectable sounds during these trials were limited to the indoor background sounds of a typical office setting.
The selection of which of the four sound treatments would be delivered during each of the directed attention test sessions was random as decided by two successive coin flips.
Two heads selected the Stratford Rd. recording; a head followed by a tail selected the
Prindle Ave. recording; a tail followed by a head selected the Wilke Rd. recording; two tails selected the control treatment of no sounds.
The delivery of the recorded sounds during the tests for directed attention was variable depending on the setting. Some of the test rooms were equipped with ceiling-mounted speaker systems that projected the recorded sounds equally throughout the rooms. When a built-in sound system was not available the recorded sounds were played through a
312
Marshall Stockwell model portable stereo speaker that delivered high-quality playback of the recordings. The speaker was positioned centrally in the room to provide as even volume of the recorded sounds as possible. The volume of sound recording playback was set to resemble the intensity level a listener might typically encounter in an indoor room with an open window. In many cases test participants did not comment on the presence of sounds. Those participants that did comment about the sounds rarely did so upon entering the test room but were more likely to comment that they became aware of the
“outdoor” sounds gradually as the session progressed. Some participants asked if the sounds were coming through an open window, which seemed puzzling when the test room was one without windows.
Administering the Tests for Directed Attention
Setting and introduction.
The facilities in which the tests for directed attention were offered ranged from relatively small rooms of approximately 300 square feet (28 square meters) to large conference meeting rooms of approximately 1,000 square feet (93 square meters). Some rooms had no windows. Windows were closed in those rooms that had windows, and when possible shades or blinds drawn to limit the view to the outdoors.
The randomly selected sound recording was playing when participants entered the testing room so that there was not an abrupt and noticeable change in the sounds of the setting. When entering the room participants were greeted and told that the testing
313 required about an hour of time. They were then seated and given a release form that briefly described the objective of the research and the tasks involved in the testing. They were also given a single sheet that gathers limited demographic information including name, phone number, e-mail address, gender, and age range. This sheet also asks participants to list the communities where they have lived since childhood (including zip codes if they remember). The community information is needed for comparing test results for groups identified by the density of the community where they live in order to explore the question “Do people from large densely populated communities respond differently to the sound recordings than do people from smaller rural communities?”
Purpose of the testing design.
After completing the release form and demographic sheet participants were told that the testing involved three ten-minute timed tasks followed by two computer-based exercises, and that the testing would take no longer than 60 minutes. They were instructed both in writing on the release form and again verbally that they could ask questions at any time or stop at any time.
The sequence of the testing involved the following segments:
1. 10 minute timed arithmetic task (on paper)
2. 10 minute timed spell check task (on paper)
3. 10 minute timed rest period
4. Stroop test (computer based – not timed)
5. Necker Cube Pattern Control test (computer based – 3 minutes)
314
Each of the tasks and tests were introduced with written instructions, and these instructions were accompanied by verbal instructions from the researcher. Participant questions were answered without revealing the details of test scoring.
The purpose of presenting the tasks and tests in this order is to provide a short period
(20 minutes) during which all participants uniformly focus and direct their attention on tasks that result in minor fatigue. The accuracy of the participant’s arithmetic problem and spell-checking answers are not scored, and the results do not enter into the analysis of directed attention. The tasks are utilized to bring participants to a similar level of elevated and slightly fatigued directed attention. These tasks are followed by a brief period of rest. This is a critical period of time in the testing because the research is designed to measure the participant’s recovery of directed attention with the sound environment as the variable of interest. The tests that are employed to measure the strength of directed attention are then administered following this uniform pattern of mild directed attention fatigue followed by a short period of rest that may (or may not) be affected by the sounds in the setting.
Arithmetic mental fatigue task.
The first 10 minute timed task involved simple adding and subtraction. Table XXIV presents an example of a problem set. Participants were asked to add two columns of five numbers and record the sums. The next step is to subtract the sum of column B from column A and record the difference. Before moving on to the next problem set, participants were asked to subtract the five rows of numbers in column B from the five
315 rows of numbers in column A and to record the differences in column C. When adding the resulting numbers in column C they should arrive at the same number that they calculated by subtracting the total of column B from the total from column A.
The participants were instructed to complete as many of the four problem sets as possible within ten minutes. They were also instructed to complete an additional page of short problems in the event they completed the four problem sets. The participants were told when to begin and when 10 minutes had passed.
This task seemed to achieve the desired goal of elevating the level of directed attention for most participants, and indeed it caused mild frustration in some. Frequent groans were emitted when the 10-minute timer was heard. Comments included “What have I gotten myself into?” and “I didn’t realize how much I depend on my calculator!” and “I didn’t come CLOSE to finishing!” It seems that the task of rectifying the results from adding the columns of numbers and subtracting the rows of numbers proved challenging for some. Only one participant completed the task in less than 10 minutes.
Most finished between two and three of the problems sets. Fewer than 10 participants finished all four problem sets and moved on to the final page of short problems.
Table XXIV: Example of a problem set used for the arithmetic task for inducing mental fatigue in participants before the tests for directed attention.
316
Spell check mental fatigue task.
The second 10-minute timed exercise was introduced immediately following the first timed task. The participants were handed a highlighter pen and a paper copy of the first ten sections of the Constitution of the United States in which many spelling errors had been inserted. The participants were instructed to use the highlighter to identify as many spelling errors as possible within ten minutes.
This timed task seemed to cause much less mental fatigue in the participants than did the arithmetic task. Participants in general did not voice frustration at the end of the 10 minute period. No participants completed the task in less than 10 minutes.
10 minute rest period.
The third 10-minute timed exercise was introduced immediately following the second timed task. Participants were verbally informed that their objective was to rest for 10 minutes. They were allowed to read non-work related material that they had with them or to choose from a collection of magazines and newspapers I provided, or they could close their eyes and relax. The only rule to the exercise is that they were not allowed to engage in work related activities such as checking e-mail. Participants were instructed to stay in the room unless they needed to use the restroom. They could quietly talk among themselves if they chose to but most participants chose to remain quiet during the 10 minute period.
317
Stroop Test.
Immediately following the 10-minute rest period, the participants were assigned a position at a laptop computer equipped with a cordless mouse and a large mouse pad. All of the laptop computers used for the computer-based tests were Lenovo ThinkPads running Windows 7 or 8 operating systems. The participants were encouraged to adjust the angle of the computer screen, mouse and mouse pad to a comfortable position.
The Stroop Test as built using the MouseTracker program was started by pressing the computer Enter key and an instruction screen appeared. In addition to the written instructions the researcher verbally explained the test.
The first section of the test allows the participants to practice the movement and operation of the computer mouse as it will be used throughout the test. Participants are instructed to click on the word START that appears at the bottom of the screen and as quickly as possible find the word HERE at the top of the computer screen and to click on it. Upon clicking the word START two white boxes appear on the screen, one in each of the upper corners of the screen. One of the white boxes is empty while the other one includes the word HERE (Figure 63).
HERE
Figure 63: Practice screen in the Stroop Test. Participants must first click on the word START and then click on whichever corner displays the word HERE. The process is repeated 12 times.
START
318
Participants must first click on the word START and then click on whichever corner displays the word HERE, and then return to the word START and repeat the process 11 times. Each time, the word HERE appears randomly in one of the two upper corners.
This exercise familiarizes the participants with the process that is used throughout the rest of the test. After completing the practice session a second instruction screen appears. In addition to the written instructions the researcher also gives the instructions verbally.
Participants are instructed that in the following sequence they are to once again click on the word START, but now upon doing so a color will appear in the middle of the computer screen. Their job is to recognize the color, then find the word in either upper corner that matches the color and click on it. Each time one of the words that appears is the name of the color while the other word that appears is the name of a different color.
The first set of slides with colors that appears is a practice session and all of the colored patterns that appear are of four asterisks (Figure 64). The participants move through 12 of these practices slides.
YELLOW BLUE
Figure 64: Example of a Stroop Test practice slide in which the color appears as a pattern, not as letters.
START
Upon completing the set of practice slides another screen appears instructing the participants to press the Enter key to move to the first set of test slides. The process
319 remains the same of clicking first on the word START, recognizing the color in the center of the screen and then clicking on the matching word in one of the upper corners of the screen. Three sets of 20 test slides follow. Each group of 20 test slides is followed by an instruction screen advising the participants to press the Enter key to begin the next sequence of test slides.
The test slides contain three different types of color representations. The first type is four colored letters that do not form a recognizable word, such as four of the letters
“hhhh”. For most participants this version of the task is easy since there is no word to interfere with the quick recognition of a color. The second type is a word that is the printed name of a color that is displayed in the same color, such as the word RED appearing in red ink (Figure 59a). This is an example of a congruent match. The third type is a word that is the printed name of a color that is displayed in a different color, such as the word GREEN appearing in red ink (Figure 59b). This is an example of an incongruent match. When the participants finish the first set of 12 practice slides followed by the three sets of 20 test slides a screen appears that informs them the test is over and thanks them for their participation.
Necker Cube Pattern Control test.
After completing the Stroop Test participants are introduced to the final test of the session – the Necker Cube Pattern Control test. Participants are asked to press the Enter key and an instruction screen appears that advises the test is to be performed simply by pressing the Enter key on the keyboard and that unlike the previous test, the computer
320 mouse is not used for this test. Participants are told to press the Enter key once more and a line drawing of a three-dimensional cube appears (Figure 61). Participants are told to stare and the cube for a few seconds and then the researcher asks if anything interesting happens when they stare at the cube. If they respond “No” they are asked to stare at the cube for a few more seconds. When a “Yes” response is given the participant is asked to describe what they see. Most participants describe the cube as “flipping” or “shifting” or
“changing perspective”. The participants are then instructed that the following is to be a
3-minute timed exercise and that their only task is to simply press the Enter key each time the cube appears to change. The researcher says “Ready, set, go” upon which the participant begins to stare at the cube. After 3 minutes the researcher says “Stop” and the participants concludes the test.
Upon conclusion of the Necker Cube test participants are thanked for their time and assistance. At this point questions about the nature of the research were addressed for those participants who expressed interest.
Limitations to the Experimental Design
There are several limitations to this experimental design. The first is that the setting where the tests are administered is variable. The size, furniture arrangement and presence of windows was not uniform, which may have affected the sound quality or the amount of distractions present. These variables were controlled as much as possible.
The second challenge is the time requirement for the testing. The single greatest limiting factor to the number of participants who volunteered was the time commitment.
321
The objective was to adequately bring participants to a similar level of pre-test directed attention activity while being exposed to the soundscape recording for a reasonable amount of time, followed by the tests for directed attention. Packing the necessary variable treatment into an acceptable amount of time was essential. Had the testing period extended beyond one hour it is likely that far fewer participants would have volunteered.
The third challenge is dealing with the level of vigor or fatigue felt by each participant as they enter the testing. A well-rested and relaxed person entering the tests may score differently on the tests for directed attention than the same person at the end of a stressful day. One advantage of presenting the tests to groups of people who had participated in a day-long training event (as was frequently the case) is that participants come to the tests for directed attention already feeling some level of fatigue.
The fourth challenge is selecting an appropriate number and combination of soundscape recordings for use in the testing. Using the full range of 14 pairs of spatially and temporally matched pairs of sound recordings would have exposed test subjects to a wider variety of many tree versus few tree soundscapes but would have brought a higher level of variability to the testing.
Methods for Analysis of Data
For the purpose of data analysis, participant scores for directed attention are arranged by sound treatment group:
Stratford Rd. (many trees)
322
Prindle Ave. (few trees)
Wilke Rd. (freeway)
No sound recording
Information from the demographics survey is summarized in an Excel spreadsheet, which allows participants to be organized by sub-groups:
Gender (male & female)
Age
o 18 to 25 years
o 26 years and over
Population density
o Urban - Within city limits of community with a minimum population of
100,000 (Roughly the size of Flint, Michigan or Canton, Ohio)
o Suburban – Within 20 miles of a community with a minimum population
of 100,000
o Rural – More than 20 miles from a community with a minimum
population of 100,000
Groups and sub-groups are then used to compare scores from the tests for directed attention.
Stroop Test analysis.
Of the data gathered within the MouseTracker program, three variables are used in the scoring of the Stroop test:
323
Percent correct for incongruent color match
Response time
Mouse path maximum deviation from straight line
Scores for each of these three MouseTracker test variables are arranged by sound treatment group and compared using Analysis of Variance (ANOVA) (alpha = 0.05).
ANOVA is also used to compare these test variables for sub-groups of the participant population according to gender, age, and population density. Table XXV presents the matrix for ANOVA analysis comparing MouseTracker scores for groups and sub-groups of participants by sound treatment.
324
Table XXV: Matrix for ANOVA analysis comparing MouseTracker scores for groups and sub- groups of participants by sound treatment.
325
Necker Cube Pattern Control analysis.
The Necker Cube Pattern Control test measures only one participant variable, namely the number of times the cube seems to flip perspective within three minutes. As with the
Stroop Test, participants are arranged by sound treatment. The Necker Cube tests scores are analyzed using Analysis of Variance (ANOVA) first including all participants by sound treatment and then by sub-groups according to gender, age, and population density using the same grouping definitions as described for the Stroop Test. Table XXVI presents the matrix for ANOVA analysis comparing Necker Cube scores for groups and sub-groups of participants by sound treatment.
Table XXVI: Matrix for ANOVA analysis comparing Necker Cube scores for groups and sub-groups of participants by sound treatment.
326
CHAPTER IX
RESULTS PART 2 – TESTS FOR DIRECTED ATTENTION
Participant Demographics
Between January 2016 and September 2016, 213 people participated in the tests for directed attention while listening to one of the three sound treatments or the control of no sound recording. Prior to starting the test session each participant was asked to read and sign a release form and provide limited demographic information including their gender, year of birth (reported in 5-year increment classes), and the name and zip code (postal code for Canadian residents) of their current community of residence. Fifty-four of the participants are female; 159 are male.
Using the name and zip code of the current community residence, additional information was collected (Statistics Canada, 2016; United State Census Bureau, 2016;
USNaviguide, 2013) that was used to place each participant into one of three community types as determined by population density classes.
327
Figure 65 presents a summary of participant gender and residential community type as determined by population density. Figure 66 presents a summary of participant gender by birth year. Figure 67 presents a summary of participants by gender according to the sound treatment they experienced while taking tests for directed attention.
Figure 65: Residential community type by gender for participants of tests for directed attention.
Figure 66: Birth year by gender for participants of tests for directed attention.
328
Figure 67: Participant count by gender and sound treatment experienced while taking tests for directed attention.
Results of Necker Cube Pattern Control Test
The Necker Cube Pattern Control Test provided a single score for each participant, namely the number of times the image of a cube appeared to change perspective during a three minute period. For analysis participants were grouped by sound treatment
(Stratford Rd., Prindle Ave., Wilke Rd. or None (the control)). Single factor analysis of variance (ANOVA) was performed for the Necker Cube scores for four sound treatment groups. Three additional ANOVA analyses were then conducted for subgroups of the four sound treatment groups including residential community type, gender, and age group. Table XXVII presents the results of the Necker Cube ANOVA analyses including the mean values for each sound treatment group and subgroup and the resulting p-value for each ANOVA. Results were considered statistically significant if the p-value is less than 0.05 (alpha = 0.05).
329
The results of the ANOVA tests revealed that none of the trials, either by all participants grouped by sound treatment, or by considering subgroups of gender, age or community type resulted in statistically significant results. This indicates that at the 95% confidence interval, the sound treatments (including the control of no sounds) were not correlated with significant differences in Necker Cube Pattern Control test scores.
Table XXVII: Necker Cube Pattern Control Test - Mean values of perceived change in cube perspective during 3-minute timed trial. Color-coding of cells indicates lowest to highest mean values for each group and subgroup. ANOVA alpha = 0.05.
In addition to studying the ANOVA P-values to determine statistical significance, it is also useful to observe trends in the mean values of Necker Cube scores to find clues of emerging patterns. Are the lowest mean scores clustered around a single sound treatment? Are the highest mean scores clustered around a different sound treatment?
The color-coding of cells in Table XXVII indicates the order of mean values for each
ANOVA group and subgroup test, making it visually easier to answer these questions.
There is no strong ordering trend of Necker Cube scores that would suggest that
330 increasing the sample size is likely to reveal a consistent effect of sound treatments on
Necker Cube test scores.
Stroop Test Results
The MouseTracker program allows for the Stroop test to be evaluated in several ways.
Figure 68 illustrates the results of one of the 213 Stroop test trials. Purple lines trace the pathways of the computer mouse that presented color matches where the color of the word matched the printed word, known as congruent matches. Blue lines trace the pathways of the computer mouse that presented color matches where the color of the word did not match the printed word, known as incongruent matches. It is these incongruent matches that require a greater level of directed attention to correctly recognize a color while repressing the recognition of a printed word. The mean trajectories demonstrate slightly greater uncertainty for incongruent matches for this particular participant.
331
Figure 68: Sample of one MouseTracker analysis screen. The top boxes illustrate computer mouse pathways for congruent (purple) and incongruent (blue) color match screens. The lower left box illustrates the mean computer mouse pathways for congruent and incongruent color match screens. The lower center boxes illustrate mean computer mouse response times for congruent and incongruent color match screens along X and Y coordinates.
The computer mouse pathways are compared to ideally efficient straight-line pathways between the start square and the correct answer square, both from a spatial deviation from the ideal pathway and from a temporal deviation from the ideal pathway. Four measures of the computer mouse pathway can then be reported:
Reaction time (RT): The elapsed time between clicking on “Start” and clicking on
a color word after viewing a color on the computer screen.
332
Maximum deviation (MD): The largest perpendicular deviation between the
actual response trajectory and the idealized response trajectory (Figure 68).
Maximum deviation time: The difference in elapsed time between the estimated
idealized response trajectory time and the actual reaction time.
Area Under the Curve (AUC): The geometric area between the actual response
trajectory and the idealized response trajectory (Figure 69).
Figure 69: Maximum deviation (MD) and area under the curve (AUC) formed by an idealized response trajectory and an actual computer mouse trajectory within the MouseTracker program.
In addition to these measures that are based on computer mouse trajectories, the percentage of incorrect color selections for incongruent color matches is used to evaluate results of the Stroop test.
The mean values for response time, maximum deviation distance, maximum deviation time, area under curve, and percentage of incorrect color selection for incongruent color matches are used in analysis of variance (ANOVA) tests to evaluate the effect of the four sound treatments on these Stroop Test measures.
333
Stroop Test response time results.
Stroop Test results were analyzed by considering the data for MouseTracker variables grouped by congruent color matches and incongruent color matches. This is because the congruent color matches do not require elevated directed attention to reach a correct solution to the color match task. The incongruent color matches however do require the participant to closely pay attention to the color they see while ignoring the printed word that they see. Therefore if a sound treatment does indeed affect directed attention it is most likely that the difference in Stroop test results will be detected in the incongruent color matches.
Analysis of Variance (ANOVA) was conducted for the mean values of congruent color match response time and incongruent color match response time for all participants grouped by the four sound treatments. Alpha for all ANOVA tests was set at 0.05. Table
XXVIII presents the ANOVA results for the MouseTracker response time measure for congruent color matches for all participants. One of the Wilke Rd. trials was omitted due to a failure of the wireless computer mouse during the test.
An important number to consider it the reported P-value. In order to reject the null hypothesis that there is no difference in Stroop Test response time scores between sound treatments the P-value must be less than 0.05. Since the P-value of this ANOVA test is
0.34, the null hypothesis cannot be rejected meaning that there is no statistically significant difference in mean value response time scores among any of the sound treatments. This is also demonstrated by the calculated F value of 1.14 being less than the critical F value of 2.65.
334
As previously discussed, it is less likely that significant differences in Stroop Test scores will be detected in the congruent color match measures and more likely to be detected in the incongruent color match measures. Table XXIX presents the ANOVA results for the MouseTracker response time measure for incongruent color matches for all participants. Once again, the important number to consider it the reported P-value.
Since the P-value of this ANOVA test is 0.34, the null hypothesis cannot be rejected meaning that there is no statistically significant difference in mean value response time scores among any of the sound treatments. This is also demonstrated by the calculated F value of 1.14 being less than the critical F value of 2.65.
Are there other informative numbers within the ANOVA results? The values for variance calculated by the Excel ANOVA program were used to calculate standard deviation as shown in Tables XXVIII and XXIX. The standard deviation values demonstrate the overlap in mean response time values between the four sound treatments, which leads to the calculated P-value exceeding the target of 0.05. It is possible that increasing the sample size may result in tighter standard deviation ranges, but this is not certain.
Another important pattern to observe is the presence or absence of ordering among the mean values within the four sound treatment types. If soundscapes that have a greater proportion of nature sounds are believed to have a more restorative effect on directed attention than soundscapes with a greater proportion of human mechanical sounds then the sound recording treatments from Stratford Rd. and Prindle Ave. would be expected to result in lower mean values for Stroop test response time than the sound treatment for
Wilke Rd. or the control of no sounds. Therefore, the expected ordering of mean
335 response time values from low to high for the four sound treatments would be Stratford
Rd. (many trees in a residential neighborhood) followed by Prindle Ave. (few trees in a residential neighborhood) then Wilke Rd., (few trees next to a freeway) with the no sound treatment expected to result in either the third or fourth highest response times. Is there evidence of this if the mean response time values are ordered from low to high? In both of the congruent and incongruent color match trials, the Prindle Ave. sound treatment
(the neighborhood recording gathered after many trees were removed) has the lowest mean response time value, followed by the Wilke Rd. sound treatment (the recording gathered next to the freeway that includes a very high proportion of traffic noise).
Somewhat surprisingly the Stratford Rd. sound treatment led to the third highest mean values followed lastly by the no sound treatment. However these values were not shown to be significantly different so observing this trend is of limited diagnostic value.
Table XXVIII: Analysis of Variance (ANOVA) results for MouseTracker response time for congruent color matches for all participants.
336
Table XXIX: Analysis of Variance (ANOVA) results for MouseTracker response time for incongruent color matches for all participants.
The ANOVA test revealed no significant differences in the mean values for response times for all participants grouped by sound treatment. ANOVA tests were then performed for subgroups of the participant population including:
Rural, suburban, and urban residential community types
Gender
Age (born before 1980 and born in 1980 and after)
Table XXX presents the mean values of response time for congruent color matches and incongruent color matches for all participants grouped by sound treatment and for these sub-groups of the participant population.
337
Table XXX: Stroop Test - Mean values of response time for congruent color matches and incongruent color matches and ANOVA P-value results. Color-coding of cells indicated lowest to highest mean values for each group and subgroup. ANOVA alpha = 0.05.
None of the 8 ANOVA tests for mean Stroop Test response times for congruent color matches produced statistically significant results. The same is true for the incongruent color matches. High variability in the response time scores lead to high values of standard deviation and therefore relatively high P-values.
Is there evidence of trends among the sound treatments if not significant differences?
Perhaps, but they are not as predicted. In an effort to make potential trends easier to detect Table XXX includes a color scheme that identifies mean values from lowest (blue shading) to second (green), third (yellow) to highest (red). In 11 of the 16 ANOVA tests
338 the Prindle Ave. sound treatment is associated with the lowest mean Stroop Test response times. In 8 of the 16 ANOVA tests the no sound treatment control is associated with the highest mean Stroop test response times. Perhaps most surprisingly, the Wilke Rd. sound treatment is associated with the second lowest mean Stroop Test response times in 10 of the 16 trials, and the Stratford Rd. sound treatment (the treatment from the residential neighborhood with many trees) is associated with the second highest Stroop Test response times in 8 of the 16 ANOVA tests. These trends may be curious, but are not statistically significant.
Stroop Test maximum deviation time results.
The same grouping of all participants by sound treatment was used to conduct
ANOVA tests for the mean maximum deviation time values both for congruent and incongruent color matches in the Stroop Test. As with the response time values, the participant population was also divided into subgroups for community type, gender and age when conducting ANOVA tests for mean maximum deviation values. Table XXXI presents the mean values of maximum deviation time for congruent color matches and incongruent color matches for all participants grouped by sound treatment and for the community type, gender and age sub-groups of the participant population.
339
Table XXXI: Stroop Test - Mean values of maximum deviation from straight line time (milliseconds) for congruent color matches and incongruent color matches and ANOVA P-value results. Color-coding of cells indicated lowest to highest mean values for each group and subgroup. ANOVA alpha = 0.05.
None of the 8 ANOVA tests for mean Stroop Test maximum deviation from straight line time for congruent color matches produced statistically significant results. The same is true for the incongruent color matches. High variability in the maximum deviation from straight line time scores lead to high values of standard deviation and therefore relatively high P-values.
340
Is there evidence of trends among the sound treatments if not significant differences?
Perhaps, but once again they are not as predicted. In 11 of the 16 ANOVA tests the
Prindle Ave. sound treatment is associated with the lowest mean Stroop Test maximum deviation from straight line time. In 11 of the 16 ANOVA tests the no sound treatment control is associated with the highest mean Stroop Test maximum deviation from straight line time, and the remaining 5 mean values were the second highest scores. Again most surprisingly, the Wilke Rd. sound treatment is associated with the second lowest mean
Stroop Test response times in 13 of the 16 trials, including all 8 of the incongruent color match trials. The Stratford Rd. sound treatment (the treatment from the residential neighborhood with many trees) is associated with the second highest Stroop Test maximum deviation from straight line time in 7 of the 16 ANOVA tests. Once again, these trends may be curious, but are not statistically significant.
Stroop Test maximum deviation distance results.
The same grouping of all participants by sound treatment was used to conduct
ANOVA tests for the mean maximum deviation distance values both for congruent and incongruent color matches in the Stroop Test. As with the response time and maximum deviation time values, the participant population was also divided into subgroups for community type, gender and age when conducting ANOVA tests for mean maximum deviation distance values. Table XXXII presents the mean values of maximum deviation distance for congruent color matches and incongruent color matches for all
341 participants grouped by sound treatment and for the community type, gender and age sub- groups of the participant population.
Table XXXII: Stroop Test - Mean values of maximum deviation from straight line distance (bins) for congruent color matches and incongruent color matches and ANOVA P-value results. Color- coding of cells indicated lowest to highest mean values for each group and subgroup. Red P- values indicate statistically significant results ANOVA alpha = 0.05.
One of the 8 ANOVA tests for mean Stroop Test maximum deviation from straight line distance for congruent color matches produced statistically significant results, that being the suburban population density subgroup. This indicates that at least one of the four sound treatments is associated with significantly different scores on the Stroop Test
342 when considering the maximum deviation of the computer mouse pathway from the ideally efficient straight line between start and answer. None of the incongruent color matches produced statistically significant results.
The ANOVA test reveals a statistically significant difference among the mean values for maximum deviation from straight line distance among the suburban subgroup, but it does not reveal which of the sound treatment pairings are different from the others. To determine this a post hoc series of t-tests was performed in which each of the four sound treatment mean values was individually compared to each of the other three sound treatment mean values for a total of 6 potential combinations of sound treatments.
However repeated comparisons using t-tests effectively dilutes the confidence interval for the test. Therefore the post hoc t-tests employed the Bonferroni correction by dividing the alpha level (0.05) by the number of the pairings (6) to produce an adjusted critical P- value of 0.0083. Table XXXIII presents the two-tail P-values for each of the sound treatment pairings.
Table XXXIII: t-Test P-values for mean values of maximum
deviation distance of the suburban community type subgroup for the six sound treatment pairings. The yellow cell indicates statistically significant difference after the Bonferroni correction.
343
Only one of the six pairings, namely the Prindle Ave. and no sound (control) pairing shows a statistically significant difference in mean value for maximum deviation from straight line distance for the suburban subgroup of participants.
There is little evidence of trends when considering the ordering of maximum deviation from straight line distance values relative to the four sound treatments. In 8 of the 16
ANOVA tests the no sound treatment control is associated with the lowest mean Stroop
Test maximum deviation from straight line time. In 9 of the 16 ANOVA tests the Prindle
Ave. treatment is associated with the highest mean Stroop Test maximum deviation from straight line time, and 6 of the remaining 7 mean values were the second highest scores.
Not only is this surprising because it is not consistent with the predicted order of sound treatment scores, but it also is in contrast with the trend observed with the ordering of maximum deviation from straight line time mean values. Since these two measures are both based on the maximum deviation from an ideal straight-line pathway calculated within the Mouse-Tracker program I expected the results of these two methods of analysis to be similar. Once again, these trends may be curious, but are not statistically significant.
Stroop Test area under the curve results.
The same grouping of all participants by sound treatment was used to conduct
ANOVA tests for the area under the curve values both for congruent and incongruent color matches in the Stroop Test. As with the previously discussed MouseTracker analysis scores, the participant population was also divided into subgroups for
344 community type, gender and age when conducting ANOVA tests for area under the curve. Table XXXIV presents the mean values of area under the curve for congruent color matches and incongruent color matches for all participants grouped by sound treatment and for the community type, gender and age sub-groups of the participant population.
Table XXXIV: Stroop Test - Mean values of area under curve (AUC) for congruent color matches and incongruent color matches and ANOVA P-value results. Color-coding of cells indicated lowest to highest mean values for each group and subgroup. Red P-values indicate statistically significant results. ANOVA alpha = 0.05.
345
One of the 8 ANOVA tests for mean Stroop Test area under the curve for congruent color matches produced statistically significant results, that being the suburban population density subgroup. This indicates that at least one of the four sound treatments is associated with significantly different scores on the Stroop Test when considering the area under the curve of the computer mouse pathway couple with the ideally efficient straight line between start and answer. A post hoc series of six t-Tests that employed the
Bonferroni correction matching the individual sound treatments revealed that the
Stratford – no sound (control) treatment pairing is the only pairing with a statistically significant difference between mean values within the suburban community type subgroup. None of the incongruent color matches produced statistically significant results.
There is little evidence of trends when considering the ordering of area under the curve values relative to the four sound treatments. In 7 of the 16 ANOVA tests the no sound treatment control is associated with the lowest mean Stroop Test area under the curve and five of the remaining 9 tests associated with the second lowest mean scores. In
8 of the 16 ANOVA tests the Stratford Rd. treatment is associated with the second highest mean Stroop Test area under the curve scores. Once again, these trends may be curious, but they are weak trends and are not statistically significant.
Stroop Test percent incorrect answers results.
The previously discussed MouseTracker variables for the Stroop Test all consider the pathway of the computer mouse between the start and answer points for each color match
346 task. The number of incorrect answers variable does not consider the computer mouse pathway but rather scores the incorrect color match answers during the Stroop Test. It is expected that the number of incorrect answers will be higher for the incongruent color match tasks in general, and if the sound treatments associated with nature are more effective at replenishing directed attention the predicted order of number incorrect answers from low to high is Stratford Rd. followed by Prindle Ave., and the Wilke Rd. sound treatment associated with the most incorrect answers. Table XXXV presents the mean values of the number of incorrect answers for congruent color matches and incongruent color matches for all participants grouped by sound treatment and for the community type, gender and age sub-groups of the participant population.
Two of the 8 ANOVA tests for mean Stroop Test number of incorrect answers for congruent color matches produced statistically significant results, those being the suburban and rural population density subgroups. In both cases it is easy to see that there were no incorrect color match answers for three of the sound treatments for each of the suburban and rural subgroups and at least one incorrect color match answer for one sound treatment (four incorrect answers in the no sound control treatment in the suburban subgroup and two incorrect color match answers in the Wilke Rd. sound treatment in the rural subgroup). None of the incongruent color matches produced statistically significant results.
347
Table XXXV: Stroop Test - Mean values of number of incorrect answers for congruent color matches and incongruent color matches and ANOVA P-value results. Color-coding of cells indicated lowest to highest mean values for each group and subgroup. Red P-values indicate statistically significant results. ANOVA alpha = 0.05.
Are there any trends in the data that may provide clues relative to the sound treatments? The first noticeable trend is that for the congruent color matches those participants listening to the Stratford Rd. sound treatment had perfect scores on the color match tasks. The other three sound treatments were associated with one or more incorrect color match answers. The incongruent color matches are expected to produce more incorrect answers and this proved to be true. In 5 of the 8 tests for incongruent
348 color matches, the no sound control treatment was associated with the lowest percentage of incorrect color matches with 2 of the remaining 3 tests associated with the second lowest percentage of incorrect matches. In 3 of the 8 tests for incongruent color matches, the Stratford Rd. treatment was associated with the lowest percentage of incorrect color matches with 2 of the remaining 5 tests associated with the second lowest percentage of incorrect matches. Once again, these trends may be curious, but they are not statistically significant.
In his book “Noise Matters” R. Haven Wiley (2015) describes noise as sound that interferes with communication and causes the receiver to make mistakes. Based on this belief it was expected that the sound treatment with the most mechanical sounds, that being the Wilke Rd. recording that was made adjacent to the freeway would have caused the greatest number of mistakes during the Stroop Test. There was no statistically significant evidence of this. Observing the trends evident in Table XXXV the Wilke Rd. recording tends to have the highest or second highest ranking of number of mistakes.
And yet the Prindle Ave. recording, which is relatively quiet and contains a relatively high proportion of biophony seems to trend close behind the Wilke Rd. treatment in the number of incorrect answers during the Stroop Test. I had expected different and significant results.
Tests for Directed Attention Results by Demographics
If the group of 213 participants is divided into demographic subgroups but not considered according to sound treatment, are there differences in the scores for Necker
349
Cube Pattern Control or Stroop Test? In order to answer that question the participant population was first sorted by demographic subgroups without also sorting into sound treatment groups. Statistical analysis was conducted on test scores to determine if significant differences in directed attention test scores exist among demographic groups.
Directed Attention test scores by gender and omitting sound treatment.
The participant population was sorted by gender classes (Female & Male), and a series of t-Tests were conducted that compared Necker Cube scores and the Stroop Test scores for each of the scoring variables. Table XXXVI presents the results for the t-Tests.
Alpha for all t-Tests is 0.05.
Table XXXVI: t-Test results for mean scores of Necker Cube
Pattern Control and Stroop Test variables for participant
population sorted by gender. Alpha = 0.05. No t-Tests results indicate statistically significant differences.
350
The results of the t-Tests indicate that there are no significant differences in the mean test scores for directed attention when considering the participant population sorted by gender.
Directed Attention test scores by age and omitting sound treatment.
The participant population was sorted by age classes (Born before 1980, Born during or after 1980), and a series of t-Tests were conducted that compared Necker Cube scores and the Stroop Test scores for each of the scoring variables. Table XXXVII presents the results for the t-Tests. Alpha for all t-Tests is 0.05.
Table XXXVII: t-Test results for mean scores of Necker Cube
Pattern Control and Stroop Test variables for participant
population sorted by age. Alpha = 0.05. Yellow cells indicate t-Test results with statistically significant differences.
The results of the t-Tests indicate that there are significant differences in the mean test scores for the Stroop Tests variables of response time (RT) for the congruent and incongruent color matches, and maximum deviation (MD) time for the congruent and incongruent color matches when considering the participant population sorted by age.
351
Participants born before 1980 took more time on average in responding to both the congruent and incongruent color match tasks than did the younger participants.
Directed Attention test scores by community type and omitting sound
treatment.
The participant population was sorted by community type as defined by population density (Rural, Suburban, Urban), and because more than two sample groups were being compared, a series of ANOVA analyses were conducted that compared Necker Cube scores and the Stroop Test scores for each of the scoring variables. Table XXXVIII presents the results for the ANOVA analyses. Alpha for all ANOVA tests is 0.05.
Table XXXVIII: ANOVA results for mean scores of Necker Cube
Pattern Control and Stroop Test variables for participant
population sorted by community type. Alpha = 0.05. Yellow cells indicate ANOVA results with statistically significant differences.
The results of the ANOVA analyses indicate that there are significant differences in the mean test scores for the Stroop Tests variables of response time (RT) for the congruent and incongruent color matches, and maximum deviation (MD) time for the congruent and
352 incongruent color matches when considering the participant population sorted by community type. Because the ANOVA results detect statistically significant differences in several of the Stroop Test scores, but do not indicate between which of the three community type groups these differences exist, a series of post hoc t-Tests were conducted that compare mean test scores between rural and suburban community types, rural and urban community types, and between suburban and urban community types.
Because each community type was compared to more than one other community type the
Bonferroni correction was applied to the t-Test alpha (0.05), yielding an effective alpha =
0.0167. Table XXXIX presents the results for the post hoc t-Tests.
Table XXXIX: Post hoc t-Test results for mean scores of Stroop Test variables for
participant population sorted by community type. Alpha of 0.05 adjusted using the Bonferroni correction yielding an effective alpha = 0.0167. Yellow cells indicate t-Test results with statistically significant differences.
For the congruent and incongruent color match scores the urban community type subgroup had significantly shorter mean response time (RT) scores than both the suburban and rural subgroup scores. For the congruent color match scores the urban community type subgroup had significantly shorter mean maximum deviation (MD) time scores than both the suburban and rural subgroup scores. For the incongruent color
353 match scores the urban community type subgroup had significantly shorter mean maximum deviation (MD) time scores than the suburban subgroup scores but no significant difference with the rural subgroup scores.
Summary
The Necker Cube Pattern Control test and the Stroop Test were the two tests used to search for potential differences in measures of directed attention relative to the sound treatment experienced by participants during test sessions. When Necker Cube and
Stroop test scores were analyzed using the Analysis of Variance (ANOVA) statistical method, no statistically significant differences associated with sound treatment were found when considering the participant group in its entirety. Statistically significant differences associated with sound treatment were found only in three of five Stroop test measures when considering subgroups of participants living in the suburban community type based on population density, and in one of the five measures when considering the rural community type group.
Do trends among the Necker Cube and Stroop Test scores suggest that increasing the population size of test participants might lead to statistically significant differences relative to sound treatment? The few trends that may be apparent are weak and inconsistent between test measures. Potential reasons for these findings are discussed in the Results and Conclusions chapter.
When considering the participant population as a whole, without consideration of the sound treatments played during the test sessions several interesting results were noted.
354
No significant differences in test scores were detected when comparing the participant population sorted by gender. There were significant differences in four of the ten Stroop
Test variables when comparing the participant population sorted by age. Those participants born before 1980 had greater average response times during the Stroop test than did the younger participants. This may be because younger people have had access to computers for a greater proportion of their lives than older people. There were also significant differences in the same four Stroop Test variables when comparing the participant population sorted by community type. Those currently living in urban communities had greater average response times during the Stroop test than did the participants currently living in suburban and rural community types.
355
CHAPTER X
DISCUSSION AND CONCLUSIONS
Introduction
The dissertation Introduction (Chapter I), Literature Review (Chapter II) and
Commentary on the Literature Review (Chapter III) examined the foundation of previous observations, inquiry, and published research on which this dissertation research rests.
There are three general areas of study that influence the hypotheses of this dissertation and the resultant research design:
The benefits of urban forests relative to humans
The investigation of soundscapes by component sounds
The effect of urban soundscapes on human directed attention
These three areas of study are related:
1. The density and composition of urban forests may affect the quantity and quality
of sounds within a community;
2. The quantity and quality of sounds, and the proportion of sounds within a
community’s soundscape, may affect human cognition;
356
3. One component of human cognition that sounds may affect is directed attention;
4. An understanding of how urban forests affect soundscapes and how soundscapes
affect humans may lead to improvements in communities that are more supportive
of human functioning.
As discussed in Chapter IV – Research Design, this is the line of reasoning that influenced the design, data collection, and analysis of the dissertation research.
The published literature has demonstrated the wide range of benefits that urban forests provide to the inhabitants of cities. The benefits are frequently grouped into environmental benefits such as improved air quality, economic benefits such as reduced energy demand for space conditioning, and social benefits such as improved human health. These benefits are linked. For example the foliage of trees can absorb gaseous air pollutants (an environmental benefit) while the shade from trees can reduce the demand for electricity to power air conditioners (an economic benefit). Reducing electricity demand leads to a reduction in fossil fuel combustion to generate electricity and a reduction in combustion emissions. This, along with the reduction in airborne pollutants attributed to trees leads to improved air quality, which leads to a reduction in respiratory diseases (a social benefit) which in turn leads to a reduction in public health costs
(another economic benefit).
In much the same way, there are links between the structure and density of urban tree canopy, the quantity and quality of sounds within and urban soundscape, and the quality of life that human inhabitants of cities experience. Admittedly the link between trees and air quality followed by the link between air quality and human health may be easier to conceptualize and indeed to measure than the linkages between trees, sounds, and human
357 cognition, but understanding how the environment that surrounds us may enhance or detract from how we experience life in cities may prove to be as important.
Summary of Important Observations and Findings
Because this dissertation research considers aspects of urban forestry, soundscape ecology and human cognition, several important observations and findings emerge from the research some of which serve to bridge gaps in knowledge that fall between these disciplines.
Urban forestry.
The importance of tree species diversity within an urban forest has become apparent with successive attacks of invasive insects and pathogens. The demise of the American chestnut (Castanea dentata) by chestnut blight in the early 1900’s followed by the death of millions of American elm (Ulmus Americana) trees in the 1950’s and ‘60’s demonstrated the danger of relying too heavily on a single genus or species of tree to populate urban forests. Ironically, many of the same spaces once occupied by elm trees were re-planted with ash (Fraxinus sp.) trees in an effort to replenish tree canopy, and once again the unwise management decision to rely heavily on a single genus has backfired. The wisdom of increasing tree diversity in urban forests cannot be claimed as an important finding of this dissertation research, but the measurement of 19% to 28%
358 loss in tree canopy in the neighborhoods of Arlington Heights, Illinois reinforces this knowledge.
Soundscape ecology.
This dissertation research advances the knowledge of soundscape ecology in several important ways, including the methods of obtaining sound recordings and the methods used for analyzing the recordings.
Soundscape data collection.
The destruction of a large percentage of trees by Emerald Ash Borer provided an unusual experimental condition of studying an urban soundscape before the trees died and after the trees were removed, thereby using changing tree canopy as the independent variable to study the effects on a dependent variable, namely sounds. This same scenario could be used to study a wide range of other associations between urban trees and environmental, economic, and social variables (as will be discussed in the future research section of this chapter). Spatially matched pairs of recordings were made by gathering sound recordings at precisely the same location on two different dates one year apart.
The calendar day differed by one (for example May 7, 2014 and May 6, 2015) in order that the recordings be made on the same day of the week at the same time. By recording on the same day of the week, variations in traffic sounds including air traffic schedules and vehicles such as school buses were minimized. This method of taking advantage of
359 an invasive insect pest known to kill a particular species of tree proved to be an effective method of conducting a natural before-and-after experiment.
Soundscape data analysis.
The results of the soundscape analysis provided statistically significant results that demonstrated differences between multiple measures of soundscapes relative to different levels of tree canopy. While the results were not always consistent, the soundscape analysis did provide very interesting insight into the proportional elements of an urban soundscape related to tree canopy cover, and confirmed that soundscapes do indeed change in complex ways as tree cover changes.
The soundscape analysis demonstrated that urban soundscapes are highly variable by season, and throughout any given day. Obviously, there is high variability at small scales within a neighborhood, as passing vehicle traffic and the frequency and timing with which your neighbor mows his or her lawn will alter the quality and quantity of sounds.
The exercise of gathering one-second samples of relative silence in each of the sound recordings provided a broader scale perspective of the neighborhood soundscapes. The evidence from the silence samples suggests that as tree canopy decreases, fewer geophony sounds are heard. This confirms what is intuitive since with fewer trees to intercept wind and precipitation there will be fewer sounds associated with weather.
Some, but not all of the evidence suggests that low-frequency background sounds increase as tree canopy decreases. Even though trees are not particularly effective at attenuating low-frequency sound energy, the distance between interior residential
360 neighborhoods and high-volume traffic streets, which are the source of much of the background noise, can result in dozens, hundreds or perhaps thousands of trees between the source of the sound energy and the receiver. Removal of up to 30% of these trees may indeed result in measurable and significant differences in low frequency sounds detected within neighborhoods.
Trees are more effective at scattering higher frequency sound energy, therefore less distance between the sound source and the receiver, and therefore less change in tree canopy may be necessary for significant attenuation of these higher frequency sounds.
Since the silence samples were essentially void of any sounds other than the low frequency background noise, this higher frequency scattering of sound energy would not be detectable. If there is evidence of scattering of higher frequency close-source sounds it would expected to be found in the slice-and-dice sound analysis, and indeed it was.
The slice-and-dice sound analysis does show a significant and consistent increase in mean sound energy as tree canopy decreases, at least with the spatially matched pairs of recordings. The temporally matched pairs of recordings gathered at Prindle Ave (few trees) and Stratford Rd. (many trees) is less consistent. This may be because while the spatially matched pairs were recorded at exactly the same place at different times they may be more representative of change over time, while the temporally matched pairs may be more of a measure of locality differences than a reliable measure of tree canopy change. This may be due to the fact that Stratford Rd. (with many trees) is several hundred feet closer to Heinz Rd. than the Prindle Ave. (with few trees) location. So the evidence from the silence samples, which primarily measure differences in low frequency distant sounds, and from the slice-and-dice analysis, which measure differences in all
361 frequency bands for sound sources at all distances, suggests that the loss in tree canopy may result in greater sound energy throughout the frequency bands but the degree of sound energy increase relative to tree canopy loss is dependent on the distance between the sound source and the receiver. To generalize, if a sound source is relatively close to the receiver, more scattering of higher frequency sound energy will take place, and lower frequency sound energy reaching the receiver will be less effected. On the other hand, the farther the sound source is away from the receiver the greater will be the measurable reduction of sound energy in the low frequency range.
The reduction of geophony sounds as tree canopy density is reduced is evident and intuitive, but what of anthrophony and biophony? There is not much evidence relative to sounds associated with human voices and music because the proportion of sounds from these sources is very small within the soundscapes that were studied. On the other hand, sounds from human mechanical sources represent a large proportion of the soundscape.
The evidence from the detailed sound-by-sound analysis of anthrophony, biophony, and geophony is mixed but the stronger evidence indicates that sounds associated with mechanical sources increase as tree canopy decreases. There were exceptions to this depending on the recording pair and on which sound attribute is measured. This supports my expectation going in to the research, but I expected the results to be more consistent and compelling.
And what of the sounds associated with animals (biophony)? The results of the detailed sound-by-sound analysis are very mixed, with no obvious pattern detected. My expectation going in to the research was that there would be a measurable decrease in biophony, particularly bird song, as tree canopy decreased. This seems intuitive since
362 there would be less nesting and perching habitat for birds to occupy. This result was true in some of the recording pairs (both temporally and spatially matched pairs) and the opposite result where biophony actually increased with decreasing tree canopy was discovered in other cases. Why are the results mixed?
One reason is that few songbird species that were present in Arlington Heights actually build nests in large trees. Most of these birds, such as cardinals, robins, and sparrows, prefer to build nests in smaller trees or shrubs, therefore the nesting habitat of the birds most likely to be singing did not change substantially when the large ash trees were removed. Also, the species composition of songbirds did not substantially change as tree canopy density decreased. What did decrease was the number and variety of perching sites from which to call. Since songbirds are most vocal during the spring nesting season as they announce and defend their territories, a decrease in bird sounds attributable to diminished tree canopy is most apparent in spring, particularly during the morning and evening chorus. The difference in the amount of bird song attributable to changes in tree canopy are less apparent at other times of the year when bird song occupies a smaller proportion of the soundscape.
Perhaps the most noticeable decrease in biophony as tree canopy decreases is attributable to fewer perching and breeding locations available for cicadas and katydids.
These insects can produce surprising sound intensity during summer days and evenings.
There are also fewer locations for tree frogs to call from as trees are removed, but these sounds comprise a very small proportion of the soundscape and are detectable for a relatively short period during the summer. Most of the amphibian sounds detected are from frogs and toads that do not rely of trees for habitat. And lastly, it is the squirrels
363 among the mammals that rely on trees from which to call, and their sounds comprise a small (but never-the-less surprisingly frequent) component of the soundscape.
Therefore, the generalized results of the soundscape analysis can be summarized as follows:
Low frequency distant sounds attributable primarily to mechanical sources,
including traffic, increase as tree canopy density decreases. This is due to a
decrease in noise attenuation as the density of vegetation between distant sound
sources and the receiver decreases.
The intensity of mid to high frequency sounds from closer sound sources increases
as tree canopy density decreases. This is due to a decrease in sound energy
scattering by foliage and branches as the density of vegetation decreases.
The proportion of geophony sounds (those non-living sourced sounds associated
primarily with weather) decreases as tree canopy density decreases. This is due
to a decrease in the number of surfaces such as tree leaves that produce sound
energy as they intercept wind and rain.
The proportion of anthrophony sounds associated with human mechanical
sources increases as tree canopy density decreases. This is not necessarily due to
an increase in mechanical sounds but rather a decrease in the scattering of sound
energy attributable to tree foliage and branches between the sound sources and the
receiver.
The proportion of biophony sounds may or may not change as tree canopy density
decreases, depending on the season, the time of day, and the sound-producing
animals. For common suburban songbirds, the most important variable associated
364
with tree canopy is perching sites from which to call, and less importantly
associated with nesting habitat. A decrease in sounds associated with seasonal
stridulating insects that rely on trees for mating and calling sites occurs as tree
canopy density decreases.
Tests for directed attention.
The tests for directed attention resulted in few statistically significant results that suggest that the sound treatments played during the tests for directed attention had an effect on the test scores. Never-the-less, some important observations and results were found.
As will be discussed in greater depth in the area of research limitations later in this chapter, enlistment of volunteers to participate in the tests for directed attention was a challenge, and the primary limiting factor influencing a potential volunteer was the time required to complete the trial. The take-home observation for this point is the importance of balancing the desire for maximizing participant sample size with the critical necessity of conducting test trials with adequate time to gather useful data. To re-state this take- home message I stress the importance of first identifying the proper tests to measure your variable of interest (in this case directed attention), then realistically design the trial to adequately accommodate the experiment’s treatment and control variables (in this case the estimated length of sound treatments needed to elicit a response), and then conduct the trials accordingly. This is preferable to first allowing sample size to drive the
365 condensed trial time and then packing as much into (or conversely eliminating tests from) the trial as the time restriction allows.
The sound treatments did not prove to have a significant effect on the scores for tests of directed attention. But somewhat unexpectedly, another variable did. When ignoring the sound treatment variable and considering instead the demographic variable of participant age, it was found that older participants (those born before 1980) had significantly longer response times and greater maximum computer mouse path deviations during the Stroop Test as compared to younger participants (born during or after 1980). This is not too surprising given that younger people have had exposure to computers for a greater proportion of their lives than older people, and therefore may simply be more proficient at working a computer mouse than their more aged counterparts. Or it may indeed be that as people age their reflexes slow down, and this is reflected in the slower response times during the Stroop Tests.
Perhaps more surprising was that urban dwellers were found to have significantly faster response times and shorter maximum mouse path deviations during the Stroop Test than did their suburban and rural counterparts. This may be caused by urban-dwellers having an elevated sense of response to external stimuli caused by more frequent sources of alarms, or more likely it may be that younger people are more likely to live in urban areas than older people and this is really a reflection of age (as discussed above) rather than population density.
366
Interpretation of Findings
The significant findings that were revealed during the soundscape analysis are relatively easy to interpret, even if the findings were not always consistent. The lack of significant findings during the tests for directed attention could simply mean that sounds have no effect on human directed attention, or the findings indicate flaws in the research design.
Soundscape characteristics relative to tree canopy.
The interpretation of the analysis of the soundscape recordings can be summarized by sound type:
Geophony: The most consistent trend is with geophony – those sounds from non-
living sources usually associated with weather. These sounds are heard more
frequently, over a wider range of frequencies, with greater intensity, and with
greater entropy at sites with many trees. This is as expected. With more trees
present there is more opportunity for wind and precipitation to be intercepted.
Anthrophony: The trend with anthrophony is fairly consistent, and in particular
the human-sourced mechanical sounds. In many, but not all soundscapes, these
sounds are heard more frequently, over a wider range of frequencies, and with
greater intensity at sites with few trees. This is as expected. With fewer trees in
a landscape there is less opportunity for sound energy to be scattered and
attenuated, particularly in frequencies greater than 1,000 Hz.
367
Biophony: The trends for animal-sourced sounds relative to tree density are not
consistent when considering all animal sounds collectively. The most important
variables that determine the presence of biophony are season and time of day.
Birds sing with greater duration and intensity in spring, particularly in morning
and evening, but can be heard to varying degrees from dawn to dusk throughout
the year. Since most urban and suburban songbirds do not nest in large
landscape trees the presence and species diversity of songbirds did not seem to
be affected by the removal of large ash trees.
Not all stridulating insects call from trees but some of the most noticeable,
including the cicadas and katydids, call almost exclusively from trees. These
insects are usually present from mid-summer to early fall but are completely
absent from the soundscape for most of the year. Few amphibians call from
trees, but tree frogs were a minor component to the recorded soundscapes during
spring to mid-summer. Mammals most frequently heard are dogs (which do not
rely on trees) and squirrels (which heavily rely on trees). Mammal vocalizations
are less seasonally dependent than other animals.
Silence: The only sound present in the samples of silence analyzed in the t-Tests
for means is the background noise, therefore the vast majority of the sound
present is attributable to mechanical sounds such as distant vehicle and aircraft
traffic. The evidence is mixed for measures of frequency and entropy between
the spatially matched and temporally matched pairs of recordings, but the
evidence for sound attributes associated with intensity suggest that the sites with
few trees have greater sound energy in the background than the sites with many
368
trees. In other words, the silence in areas with few trees is louder than the
silence in areas with many trees.
Tests for directed attention.
No statistically significant differences at the 95% confidence interval were detected in any of the eight ANOVA analyses of Necker Cube scores. The ordering of Necker Cube test mean values for each of the ANOVA results did not reveal a consistent pattern among the four sound treatments to suggest that increasing the participant population size might lead to significant results.
Including the primary grouping by sound treatment and the sub-groups by sound treatment, 80 ANOVA analyses were conducted on the Stroop Test data. Of these, statistically significant differences at the 95% confidence interval were detected in 4 of the ANOVA analyses - 3 of which were detected when considering the suburban community type subgroup and 1 when considering the urban community type subgroup.
In each of the four cases in which ANOVA analysis indicates a statistically significant difference among the four sound treatments, post hoc analysis was conducted to determine which sound treatment(s) was significantly different from the others. In three of the four cases the no sound (control) treatment was found to be significantly different from one or more of the other three sound treatments.
The fact that significant differences in Stroop Test results were only found among the community type subgroups suggests that the community type as defined by population density may play a role in the type of soundscape that is most effective at restoring
369 directed attention for any given individual. However when considering the ordering of mean scores among the community type subgroups, there is no consistent pattern that would support this hypothesis.
The ordering of Stroop Test mean values for each of the ANOVA results did not reveal a consistent pattern among the four sound treatments to suggest that increasing the participant population size might lead to significant results.
Conducting ANOVA analysis on the Stroop Test results grouped by demographic characteristics and ignoring the sound treatments revealed that response time and maximum deviation scores were lower for younger participants (those born during or after 1980) as compared to older participants. Also, those participants living in urban areas had significantly faster response times and shorter maximum deviation scores than suburban or rural participants. This suggests that a person’s age or place of residence may have greater influence on computer-based Stroop Tests than does the sound treatment that was playing during the trial.
Does the lack of statistically significant differences in scores for directed attention indicate that sounds have no effect on directed attention? Perhaps, but not necessarily.
Potential reasons for no significant differences found in tests for directed attention scores in this research include:
Sounds have no effect on human directed attention. I cannot reject this null
hypothesis based on the dissertation research results, but I do not believe this is
true based on other published research and personal observation.
The test results showed few significant differences because the sample size of
participants is too small for the variance of the scores. I am skeptical that
370
reasonably increasing the sample size would have produced significant results.
There were no consistent trends in the order of test scores that suggested a
pattern would emerge with a larger sample size.
The Necker Cube Pattern Control and the Stroop Test were not appropriate tests
to measure differences in directed attention. Had time allowed, it would have
been advisable to have included additional tests for directed attention, such as
digit span forward and backward, the trail-making test, or other recognized tests
used in psychological evaluations.
The administration of the tests for directed attention was improperly designed. It
may be that the format of two 10-minute timed tasks intended to induce mild
mental fatigue, followed by a 10-minute rest period followed by the tests for
directed attention, failed to induce sufficient mental fatigue, failed to allow
adequate time for recovery, or failed to utilize the proper tests to evaluate
directed attention. Perhaps a longer period of fatigue inducing tasks followed by
a longer rest period would have created measurable differences in the subsequent
tests for directed attention.
The variability of test site conditions affected test scores more than the sound
treatments. There was considerable variability in the rooms where tests were
administered. Some rooms were small (approximately 250 square feet) while
others were spacious (over 1,200 square feet). The smaller rooms tended to have
windows and the larger rooms did not have windows. This variability is test
sites may have affected test scores.
371
The three sound treatments chosen for use during the trials may not have been
representative of differences in soundscapes relative to tree canopy. Two of the
sound treatments where similar in sound characteristics even though the amount
of tree canopy cover was markedly different. However, the Wilke Rd. sound
treatment (recorded next to the freeway) was distinctly different from the other
two and was characterized by sounds of heavy vehicle traffic and very few
sounds associated with nature. I expected that scores for directed attention
would be significantly different for participants exposed to this sound recording,
however they were not. This suggests that either sounds indeed have no effect
on directed attention or (more probably) that the design of the trials was faulty.
The sound recordings were out of context with the trial settings. As previously
mentioned, there was great variability in the rooms where the tests for directed
attention were administered. My observation was that people who participated
in the trials in rooms with windows (particularly during the morning or
afternoon) rarely mentioned the sounds within the room, while participants who
were in rooms with no windows, or who took the tests in the evening, often
commented about the sounds shortly after entering the room. Perhaps it is
important for the setting where trials are conducted to support the context of the
sound recordings. Bird song or heavy traffic noises in conference rooms with no
windows may prove to be more distracting than when presented in rooms where
such sounds might normally be heard.
Sounds may have effects on other human functions but not on directed attention.
Perhaps directed attention is not as affected by sounds as other physiological
372
functions. If this is true then measures such as heart rate, blood pressure, or
adrenaline levels in blood may be more appropriate variables to measure.
Access to nature and the composition of soundscapes affects the degree to which
a person feels alert or relaxed. If, as I suspect, there are components of
soundscapes that alert a person to danger or conversely reassure a person that it
is safe to relax, then measures of agitation or ease (worry versus wonder) may be
appropriate indicators of the effect of sounds on humans. I do not know if such
tests exist or if new methods of measuring worry versus wonder are needed.
Contributions to the Literature
Certainly the predicted spread of Emerald Ash Borer through the Chicago metropolitan area confirmed what was predicted in the published literature. The primary contribution to the literature is the use of an invasive insect pest as a vehicle for anticipating and measuring change in an urban forest. By using a serious problem as an opportunity to study change, opportunities for inquiry into a wide range of causes and effects, costs and benefits, and treatment and control studies become available. The key is recognizing when a situation is looming that will likely cause a significant change in some aspect of an urban forest.
The important dissertation research finding that background sound intensity increased as trees were removed confirmed published literature on the topic of noise attenuation by trees. It was known that trees are more effective at reflecting sound energy in the frequency ranges between 1,000 Hz and 8,000 Hz (Wiley, 2015). However, the samples
373 of relative silence that were gathered from the spatially matched pairs of Arlington
Heights sound recordings also showed an increase in sound intensity in wavelengths longer than 1,000 Hz, which is the frequency range occupied by the background sounds of distant vehicle traffic. This is believed to be a function of the relatively large distance between the sound source and the receiver and the number of trees removed within this distance, causing a decrease in sound attenuation and reflection even in the lower frequency range.
A specific contribution to the literature are the dissertation research methods used to evaluate characteristics of urban soundscapes. Gathering spatially and temporally matched pairs of recordings demonstrates two methods of comparing soundscapes relative to tree canopy cover. Carefully identifying individual sound signatures and grouping them by sound type (anthrophony, biophony and geophony) demonstrates an effective method of evaluating the components and proportions of sounds within a soundscape. These can be used as indicators of ecological health and measures of human encroachment into ecosystems.
Based on previously published studies that tested attentional selectivity in humans during exposure to varying levels of industrial noise (Szalma and Hancock, 2011) and studies that employed tests for directed attention with participants who had (or had not) been provided access to nature (Cimprich and Ronis, 2003) it was anticipated that participants in this research would show significant differences in scores on tests for directed attention based on sound treatment. This was not the case. As was previous discussed in the sections “Summary of important observations and findings” and
“Interpretations of important findings” there may be several explanations for this,
374 including improper test session design, the sound recordings seeming out of context to the participants, or simply that the differences in the sound recordings were not great enough to cause differences in performance on the tests for directed attention.
Although this research did not find significant differences in scores of tests for directed attention, the results of the research are still valid and contribute to the body of knowledge found in the literature. Findings in this study can be used to refine the method in which sound treatments are administered to participants, and the method in which testing sessions are designed.
But what about the larger question of “Can we use this information to build better cities?” The wide range of environmental, economic, and social benefits of urban forests has been well-documented, and knowledge continues to grow, particularly in the area of human health and well-being. This dissertation research advances the body of knowledge relevant to urban soundscapes by demonstrating a soundscape evaluation method that goes beyond simply measuring the intensity of noise. By studying the components of urban soundscapes, and in particular the proportions of human-sourced sounds versus sounds associated with nature this research promotes building cities that not only deplete human cognition less but also allow healing of fatigue by providing access to sounds that replenish and support more effective human functioning.
Contributions to Practice
The advancement of methods to evaluate both the quantity and quality of sounds within an urban soundscape will contribute to future studies that consider sound in cities.
375
While the research design for the tests of directed attention did not develop a reliable method for testing the effect of sounds on directed attention, the experience of conducting the tests leads to recommendations for refinements of the test methods. When computer- based testing is appropriate, the MouseTracker program was found to be an effective tool.
Sound data analysis methods.
The process of manually identifying individual sounds within spectrograms was time consuming but very informative. The Raven program provides detailed data on sound attributes including time, frequency, intensity, and entropy, but does not provide information of sound types such as anthrophony, biophony, and geophony. Identification of sound types is critical in understanding the human-sourced components of a soundscape compared to the sounds associated with nature.
Once the data on sound attributes and sound types was gathered the next task was to analyze these data and make sense of the findings. An important finding was the use of multiple methods of data analysis to reveal patterns and statistically significant differences between matched pairs of recordings.
The first method of sound data analysis used was the slice-and-dice approach of slicing the spectrograms into uniform frequency band widths and dicing the frequency slices into uniform time increments. This allowed for matched one-hour segments of two sound recordings to be compared in their entirety with no overlap between sound selections. This method provided no information on comparing sound types
(anthrophony, biophony, and geophony) but proved to be an effective method of
376 answering the simple question “Are these two sound recordings different, and if so in which frequency bands?” This method was also used to compare simultaneous side-by- side recordings for the purpose of calibrating the digital sound recorders.
The proportional method of sound data analysis did not use the uniform spectrogram segmenting of the slice-and-dice method but rather turned to the information rich data gathered during the manual interpretation of sound signatures including the identification of sound types. This method was valuable for comparing the proportion of human- sourced sounds to the sounds associated with nature as tree canopy cover varied.
Because proportional analysis considers percentages of sound types instead of summed values of sound attributes, this is a particularly effective method of comparing the relative composition of sound types between two soundscapes that may have very similar or very different quantitative measures of individual sound attributes.
The analysis of mean values method of sound data analysis does not consider the percentage of sound types as in the proportional method of analysis but rather considers the mean values of sound attributes grouped by sound types. One analogy that helps to distinguish the two methods is to consider two pies of different size. The large pie is cut into six pieces, each weighing three ounces. The small pie is cut into four pieces, each weighing two ounces. Proportionally each piece of the small pie is larger than each piece of the large pie (25% vs. 17%), but by weight each piece of the small pie is smaller than each piece of the large pie. Each method of measuring difference between the objects, whether sound recordings or pies, is useful in understanding the relationship between the two objects.
377
Soundscapes as a monitoring method.
For many decades continuous forest inventory (CFI) plots have been used to monitor changes in the structure and health of forest ecosystems over time. In the CFI system, a network of permanent fixed-radius random plots were established throughout North
American forest lands, and each tree within a plot was numbered and inventoried. Every
5 to 10 years the CFI plots are revisited and the trees re-measured to determine growth rates, mortality and harvesting. Monitoring of permanent locations over time is used in other aspects of natural resource management from agriculture to wetlands.
As this dissertation research demonstrates, recording and analysis of soundscape components can provide a measure of wildlife species diversity, intensity of human mechanical sounds, and proportions of anthrophony, biophony, and geophony.
Continuous Soundscape Inventory sites might work in much the same way as CFI plots through the recording and analysis of keynote sounds, signals and soundmarks within a landscape. In the same way that remotely sensed imagery (aerial photos, satellite imagery, LIDAR, etc.) gathered periodically is used to create visual evidence of land use
/ land cover change over time, periodic recording and analysis of soundscapes can provide a valuable measure of changes to landscapes over time. Landscape attributes that are not easily visible, such as songbird species diversity, and distance at which human encroachment is audible can be revealed by soundscape analysis. Including sounds as a measure of change over time adds another dimension to the inquiry of human impacts on ecosystems.
378
Components of this dissertation research can be employed to build a continuous soundscape inventory rubric of data to be gathered and analysis methods to be used including:
Proportions of sound types
o Anthrophony (Types 1 & 2)
o Biophony
o Geophony
Quantification of sound attributes
o Time
o Frequency
o Intensity
o Entropy
Analysis of sound sources
o Species diversity of calling birds, insects, amphibians & mammals
o Arrival and departure dates and duration of seasonal calls
o Intensity and duration of human mechanical sounds
o Frequency, duration and intensity of background “silence”
Design of tests for directed attention.
The Stroop Test and the Necker Cube Pattern Control test are both recognized as effective tests for measuring the strength of concentration and directed attention. As previously discussed in Chapter 2: Literature Review and the Chapter 6: Methods Part 2 –
379
Tests for Directed Attention, these two tests have been used in numerous published studies and therefore they were selected for use in this research. Other standardized tests such as Digit Span Forward, Digit Span Backward, and the Trail-Making Test have also been used in published studies, and were considered for this study. In order to keep test sessions to no more than one hour in duration it was decided to use the Necker Cube and
Stroop tests.
The design of the Necker Cube Pattern Control test was very simple. The Microsoft
PowerPoint program was used to present brief instructions for the test followed by a series of identical slides illustrating the test cube. The test was very easy to administer, and the score was very simple to retrieve by simply determining the number of slides that the participant had advanced during the three-minute test period.
The design for the Stroop Test is more complex, however the MouseTracker program proved to be a reliable platform on which to base the test. The advantages of using the
MouseTracker program over other presentation methods include:
Extremely precise measure of response time for each iteration of color match
tasks
Measure of deviation from the most efficient pathway to an answer
Records of correct and incorrect answers for each iteration
Random presentation patterns
Records for type of iteration including control, congruent and incongruent color
matches
The Stroop Test on the MouseTracker platform performed flawlessly without any glitches in over 200 tests.
380
Limitations
Analysis of sound recordings.
Because of the wide range of sound attributes that can be measured, the Raven Pro sound analysis program was a good choice for analyzing the sound recordings. Unlike some other sound analysis programs that are designed primarily for use in the music recording industry, the Raven program was specifically designed by the Cornell
Ornithology Lab for use in studying sounds recorded in nature. I was also very fortunate to be accepted into the Raven training workshop and gained valuable knowledge from the
Cornell Ornithology Lab sound engineers and researchers on the efficient use of this program. This experience improved my ability to conduct the detailed analysis of the research sound recordings.
While the Raven workshop experience improved my efficiency, the task of completing the detailed sound analysis was highly labor intensive. I found that it took between six and eight hours to draw the selection boundaries around individual sounds and sound choruses for each hour of sound recording. It is possible to automate some of the analysis by training the program to recognize specific sound signatures in much the same way that image processing programs classify patterns in digital imagery, but since analysis of complete soundscapes involves many overlapping sound signatures I believe that the laborious identification of each individual sound and sound chorus was worth the effort. Also, accomplishing this task with one analyst removed inconsistencies between multiple analysts. The cost of relying on a single analyst to manually analyze each
381 recording is that a much smaller sample of recordings can be analyzed in a given period of time. For this research project I believe achieving consistency and accuracy was worth the price of a smaller sample size in the statistical analysis. The experience of manually analyzing sound signatures in great detail provided insight in how programs for automated soundscape analysis could be improved. For future work, particularly if a network of Continuous Soundscape Inventory recording plots were to be established, a system of automated sound signature identification would be essential.
Enlisting volunteers for tests for directed attention.
Sample size is an important variable when determining statistical significance in tests that compare treatment and control populations. At the same time, gathering the correct data, and sufficient data that reflect true differences in the street tree populations (if there is indeed a difference) is equally important. A third important consideration is how the data are analyzed.
During this dissertation research an important observation was made. The longer a time period is required to participate in a research trial, the less likely a person is to volunteer to participate. This obviously has direct implications on sample size. I observed that when approached to participate in the tests for directed attention the first question I was asked was usually “How long does it take?” Other frequent questions asked about the nature of the study, or what activities would be involved during the testing, but by far the most frequent question, and most often the first question involved the time commitment. When told that the trial took approximately one hour, many
382 potential participants responded “I don’t have that much time” and declined to participate. Some volunteers who reluctantly agreed to participate asked for assurance that the trial would not exceed an hour. My observation is that the participation rate of volunteers for the tests for directed attention would have decreased substantially if the trial design required more than one hour to complete. I believe than many more people would have agreed to participate had the trial design required less than 30 minutes.
However, time constraints of potential volunteers and the potential effect on sample population size should not drive the research design. The critically important task for the research design is to identify what is the correct information to gather (and equally important what information is not relevant and wastes time). In retrospect, another important question to consider is if the treatment of sound recordings for one hour during the tests for directed attention is sufficient time to result in adequate recovery of mental fatigue. Perhaps a longer trial period at the cost of a smaller sample population would have been a superior method to test the effect of sound on directed attention.
Design of test for directed attention sessions.
The first few minutes of the test session was allocated to reading the release form and filling out the demographics survey. In retrospect it was useful to gather data on gender and age so that subgroups defined by these variables could be analyzed within the Necker
Cube and Stroop tests. It also seems worthy of the effort to gather information on the communities in which the participant currently lives and has lived in the past. This information was used to gather additional data on the population density for their current
383 residence community and used to create subgroups defined by urban, suburban, and rural population densities. It was anticipated that residents of densely urban communities may react differently to the sound recordings than those from smaller communities and rural areas. Although the question was not asked, it may have been useful to know if participants had a preference for indoor or outdoor recreational activities.
One of the challenges in designing the test sessions was to seek a balance between the need to pre-stress participants with providing sufficient time for completing tests for directed attention. A related challenge was the need to enlist sufficient participants, which was a limiting factor for many potential subjects when they were told of the time required to complete the testing session.
To keep these challenges in balance it was decided to achieve uniform pre-test stressing of participants by presenting two ten-minute timed exercises designed to mildly fatigue directed attention in the participants, followed by a ten-minute rest period. In retrospect, the timed exercise that involved simple arithmetic seemed to achieve the desired results as most participants expressed some level of frustration at the end of the exercise. The spell-check exercise seemed to be much easier for most participants and was perhaps not as effective at depleting directed attention.
I am unsure of the effectiveness in refreshing directed attention of the ten-minute rest period. In theory, this should be the most important segment of the test session since participants experience the treatment of the variable sound recordings during this period of recovery. Participants were not uniform, however, in how they utilized the ten-minute rest period. Some participants did indeed close their eyes or seemed thankful for the time to read for enjoyment. Other participants seemed to react to the rest period somewhat
384 agitated, apparently eager to return to meaningful activities and not eager to engage in what may have seemed to them as wasting time.
The time needed to read the release form, complete the demographic survey, and complete the two ten-minute tasks followed by the ten-minute rest period left about 20 to
25 minutes in which to complete the Stroop and Necker Cube tests. Most participants completed the Stroop Test in less than 15 minutes, depending on their proficiency with operating a computer. In general, older participants took longer to complete the tests while younger participants at times blazed through the test in less than 8 minutes. This observation was confirmed by significantly different response time scores when comparing older to younger age groups. Some participants struggled to begin moving the mouse quickly enough after clicking on each Start box to satisfy the parameters of the
MouseTracker program, which resulted in repeated pop-up messages reminding them to start more quickly. These pop-up messages seemed to achieve the opposite results and frequently slowed the process down. In retrospect I would adjust the parameters of the
MouseTracker program to allow for slower starts so as to avoid the frustrating pop-up warnings.
By the time the Stroop Test was completed the test session had usually consumed about 50 minutes. The instructions for the Necker Cube required about three minutes followed by the 3-minute timed exercise. Some participants were eager to be on their way while others lingered after the session to ask questions about the nature of the study.
I did not engage questions about the details of the study at the beginning of the test session, but provided answers to participant’s questions at the end for as long as they were willing to discuss details.
385
Demographic and geographic limitations.
The selection of Arlington Heights, Illinois as a suitable test community for this research was driven by the density of the ash tree population and the location of the advancing Emerald Ash Borer population in 2013. As chance would have it these conditions highlighted Arlington Heights as a favorable location, and as chance would have it Arlington Heights is a suburban community adjacent to Chicago, a large and densely populated urban city. The Arlington Heights community could be described as middle class with homes built primarily in the 1960’s to 1980’s on lots typically between approximately 5,000 to 15,000 square feet (450 to 1,400 square meters) in size. As can be seen in the many photographs of streets included in this dissertation, most properties in the residential neighborhoods where this research took place have adequate space on the lot for the house, and a front and back yard that is landscaped with turf, shrubs and trees of various sizes and density. All of the streets in this study were stocked with relatively large street trees, between 40 and 70 feet (12 to 20 meters) in height). The community is not as densely populated as many neighborhoods in Chicago and not as sparsely populated as many more recently developed suburban communities farther away from Chicago. Demographically, Arlington Heights is primarily populated by
Caucasians with minorities of Latinos, African Americans and Asians. In many ways
Arlington Heights seems somewhat close to what an average American suburban community looks like.
It is a reasonable assumption that the residents of Arlington Heights have largely chosen to live here as opposed to somewhere with more urban or rural characteristics,
386 and therefore it is a reasonable assumption that many residents (not all but perhaps most) place value on residing in a single-family dwelling situated on a lot that is large enough to provide moderate space between houses and provide space for landscaping. Given the generally well-kept appearance of most properties, including the quality of the landscaping, it is a reasonable assumption that many residents place value on their lawns, flower beds, shrubs and trees. While most residents assumingly did not specifically select their homes because of the presence of an ash street tree, it is reasonable to assume that the presence of large street trees in the neighborhoods may have had some (even if minor) influence on the decision to live in a particular neighborhood.
Given these considerations, conducting tests for directed attention or any other psychological or physiological variable for only residents of Arlington Heights while using tree canopy as the independent variable may provide valuable insight for the population of Arlington Heights, but such an experiment has the limitation of a demographic bias. The residents of Arlington Heights may be familiar with the keynote sounds of their community, including the level of background sounds from the freeway, and find nothing distracting about any of the sound recordings from Arlington Heights.
On the other hand, conducting tests for directed attention on a wide range of people from differing communities and population densities from across North America while using the sound recordings from Arlington Heights (as was done in this study) may present test participants with unfamiliar sounds that they find distracting, therefore adversely affecting their scores on tests of directed attention. The characteristics of the soundscape of a community where an individual has chosen to live may indeed reflect that person’s preferences, and may indeed reflect another person’s distaste. Since sound preference to
387 some degree is subjective, using sound as an independent variable to study human cognition has limitations that must be recognized.
Potential improvements to methods.
As discussed, there are challenges and limitations to this experiment. A greater diversity of many-tree and few-tree sound recording treatments would have contributed a more representative sample of soundscapes. A more intense set of stressing tasks may have brought test participants to a higher level of fatigue, and a more lengthy rest period that exposed participants to the treatment sound recordings for a greater period of time may have produced more pronounced differences in the scores of tests for directed attention. Also, utilizing more than two tests for directed attention may have provided more insight into which forms of fatigue and recovery are affected (if at all) by sound.
As with many experiments in research the challenge is to achieve an adequate sample size of test participants with which to allow appropriate resolution to the question, with which to recognize patterns through the fog of variability, and to reach reasonably confident answers in the form of statistically significant results (if indeed there are patterns to be recognized). The analysis of the data gathered by these sound treatments and the associated tests for directed attention provide a first step at understanding the role of soundscapes in human cognition.
388
Recommendations for Future Research
The methods employed in this dissertation research and to some extent the data collected may be of value for future research in the areas of urban forestry, soundscape ecology, human health and social concerns.
Urban forestry.
Emerald Ash Borer.
Emerald Ash Borer continues to spread throughout North America. Pheromone traps are used in areas where the insect has yet to arrive but are likely to be in the pathway of distribution. Since Emerald Ash Borer attacks only ash trees (and possibly white fringe tree) and affects all species of ash (Fraxinus), there exists a fortunate condition of predicting where the insect is about to arrive and knowing what the insect’s target will be, and therefore where and to what extent change in an urban forest will take place. It is not quite as pure an experiment as randomly assigning treatment and control conditions to human subjects, but knowing the extent of change about to take place in a community opens the door for a variety of before-and-after studies in which the treatment (loss of trees) is known in advance. Potential studies include:
Change in environmental conditions associated with tree canopy such as air
quality, storm water runoff, and temperature.
389
Change in economic conditions associated with tree canopy including energy
conservation and property values.
Change in social conditions associated with tree canopy including human health
(respiratory and heat-related illness), use of open spaces, and even crime rates.
The value of landscape trees.
Several studies have been conducted that attempt to determine the influence of landscape trees relative to residential property values. Most of these studies have approached the challenge through the use of hedonic modeling, a form of regression analysis that attempts to determine the strength of influence on property value of the various components of real estate, including the characteristics of the structure, the location of the property, and the quality of the landscaping. The arrival of the Emerald
Ash Borer to Arlington Heights and the subsequent cost sharing program for injecting a protective insecticide provides a unique opportunity to measure landscape tree value with a different approach.
As the arrival of Emerald Ash Borer became imminent, the Village of Arlington
Heights offered a cost-sharing program to residents who had an ash street tree in front of their house. The Village offered to pay half of the estimated $100 to $150 cost of having the street tree injected with a protective insecticide. Of the more than 13,000 street trees that qualified for this program, approximately 3,000 (approximately 22%) of residents took advantage of the cost-share program (Jonnes, 2016). This suggests that fewer than one quarter of the residents of Arlington Heights placed a value of at least $50 to $75 on
390 the mature ash tree adjacent to their property. This is simply looking at the superficial facts and not considering the variables including the fact that the subject trees were not on the resident’s property but were technically on village property, and the uncertainty of the effectiveness of the insecticide or of the frequency with which the treatment would need to be applied. Although this dissertation research did not consider the effect of the loss of tree canopy on residential property values, the presence of this cost share program provides a unique opportunity for further research.
Soundscape ecology.
The Ivory-billed woodpecker was long thought to be extinct until a glimpse of a single bird was reported in the swampland of Arkansas in 2004. This led to several expeditions of ornithologists to the region in hope of seeing the bird. The chances of seeing such a rare bird are limited by our range of vision within a dense forest compounded by limitations of available daylight. A much more efficient method of detecting the bird is through the use of autonomous sound recorders that can detect the bird’s call and characteristic “double knock” over a much wider area and for uninterrupted periods of time. This is one example of the use of sound analysis to better understand an organism.
Perhaps sound analysis can also be used to better understand humans and the environment of which we are a part.
The simple act of attentive listening can reveal valuable information about the environmental health of a community or region, and the comparison of soundscape recordings over time can provide and understanding of the trend of changing
391 environmental health. In the same way that regularly scheduled physical examinations can provide benchmarks for our bodies and alert our physicians when key markers change over time, analysis of soundscapes over time can alert us to ecological trends, such as changes in species diversity. These changes may be indicators of small-scale changes such as the draining of a small pond that previously provided habitat to spring peepers, or indicators of large-scale changes such as the northern movement of certain bird or insect species attributable to climate change. In the same way that there are valuable indicators of environmental health to be discovered in soundscape analysis, there may be valuable indicators of human physical and mental health to be discovered by studying properties and trends within community soundscapes. How can this be achieved?
A reasonable first step toward understanding soundscapes is to begin to form a historical record of sounds at designated locations over time. These sound recordings can be compared in the same way that aerial photos and satellite imagery of a location illustrate changes to land cover and land use patterns over time. In addition to providing insight to changes in species distribution of songbirds and certain insects and amphibians, understanding changes to soundscapes over time also provides insight to changing proportions of anthrophony, biophony, and geophony.
The results of this study demonstrate changes to neighborhood soundscapes in
Arlington Heights, Illinois over the period of time when substantial loss of trees took place. The changes are subtle and yet significant, suggesting a shift in the proportions of anthrophony and geophony during the relatively short period of time between July, 2013 and August 2015.
392
This dissertation research gathered more than 325 hours of sound recordings in
Arlington Heights between 2013 and 2015 during the height of the Emerald Ash Borer infestation. Approximately 39 hours of these recordings have been analyzed in detail, leaving approximately 286 hours of recordings yet to be explored. Future research could utilize either or both of the sets of analyzed or unexplored recordings to document change over time. Examples of potential questions to be explored include:
Is there a change in the proportion of bird species identified in recordings over
time?
Do sounds indicate the arrival or departure of bird, insect, or amphibian species
during the time period? Have seasonal arrivals and departures of birds shifted?
Will increases in solar gain do to a lack of shade affect calling insect or
amphibian populations?
Do birds shift the range of frequency of their calls in response to background
noise?
Do recordings gathered in the future (5 years, 10 years, 20 years) indicate
significant changes in species diversity?...to proportions of sound types?...to
levels of background sounds or silence?
Does the soundscape of Arlington Heights “recover” to pre-EAB conditions of
quantity and quality of sound types as new trees are planted and mature?
Unlike landscapes that can clearly be delineated by land use / land cover types and geographic features such as watersheds, soundscapes extend over geographic and land use boundaries. Soundscapes function at a wide range of scales. Depending on the time of day, atmospheric conditions, and the density and characteristics of objects in a
393 landscape, background sounds at times may be heard from miles away. This is particularly true of air vehicle traffic. On the other hand, a single tree may influence the soundscape on a scale of a portion of a residential lot when it intercepts breezes or provides perching sites for songbirds to call from. Therefore sounds can affect a person’s perception of the environment in which they exist on a very wide range of scales.
Additional research is needed to investigate how natural and built landforms and the vegetation characteristics of a landscape affect the transmission of sound energy over a range of scales. Additional insight is needed on the scale at which human mechanical sounds, which are typically in lower frequency ranges, travel through a soundscape as compared to sounds associated with nature, which are typically in mid frequency ranges that are more effectively reflected and attenuated by vegetation.
Human health and social concerns.
The existing sound recordings may be analyzed to investigate questions relating to sounds and the social health of the neighborhoods. Questions to explore may include:
Is there a difference in the frequency of human voices in areas of different tree
cover? Do shady streets attract more humans to use outdoor spaces?
Does the absence of tree shade lead to increased rate of turf growth and
subsequent increases in mowing frequency?
While the experiment of testing human directed attention during exposure to three soundscape recordings and a control of no recordings did not reveal significant differences in test scores, this does not necessarily indicate that sound does not affect
394 humans. This may rather indicate that better experimental design is needed to detect the effects of sounds on directed attention or other measures of human cognitive ability or physiology. The existing sound recordings could once again be used as treatments during tests for cognitive or physiological conditions in addition to tests for directed attention.
Questions to explore may include:
Does changing the trial session format for tests of directed attention lead to
different results? Does lengthening, shortening, or changing the fatigue-inducing
tasks or the period of rest affect results? Does the inclusion of additional tests for
directed attention affect results?
Are physiological functions such as heart rate, blood pressure, respiratory rate, or
blood enzymes affected by the sound recordings?
Does sound context affect directed attention or physiological functions? Does
what a test participant sees while being subjected to sound recording treatments
affect their response?
Upon listening to samples of the neighborhood recordings what soundscapes do
research participants find relaxing or irritating? What sound types (anthrophony,
biophony, geophony) are preferred and what sound types are avoided? How is the
feeling of relaxation measured?
No attempt was made during this research to question or interview the residents of
Arlington Heights to gather information on their insights or opinions on the dramatic loss of trees due to Emerald Ash Borer. Caution is advised when gathering qualitative information in surveys and interviews as it is very easy to bias the information through the structure of the questions or survey. But with the proper design of qualitative inquiry
395 the rapid and significant loss of trees in Arlington Heights and other communities presents a unique opportunity to understand urban resident’s value of trees. Surveys of residents, interviews of homeowners, and/or development of focus groups to discuss changes in the community may reveal important information on citizen values and preferences.
The change in tree canopy experienced in Arlington Heights, Illinois and the subsequent collection of sound recordings provided valuable insight into the changing conditions of a single community. Lessons learned from this experience can be employed in many other communities to broaden the geographic and demographic scale of the research. Based on the experience of designing this study, gathering and analyzing sound recordings, and gathering and analyzing the data on the tests for directed attention, the following recommendations are made for future studies in other communities:
1. Recognize events that lead to substantial change in urban forests such as the
geographic advance of destructive pests and pathogens or large-scale tree-planting
programs to conduct before-and-after research relative to benefits for humans.
2. Establish sound benchmark locations in communities where sound recordings are
gathered and analyzed on a regular basis to document ecological and
anthropogenic change over time (similar to how continuous forest inventory plots
are used to document change in forest ecosystems over time).
3. Improve sound recognition programs so that automated soundscape analysis can
be used to develop soundscape maps in the same way that digital image
interpretation is used to develop land cover maps.
396
4. Continue to use tests for directed attention as part of investigating the effects of
sounds on humans but also introduce additional tests to measure physiological
and emotional response to sound.
5. Gather soundscape examples for a much larger range of environments than a
single community for use in tests for directed attention and other responses.
6. Develop a web-based sound response testing system to reach a much larger and
more diverse population of participants.
7. Use soundscape recordings during tests for directed attention to document and
improve harmful sonic environments in cities that negatively affect human
cognition.
Conclusions
As stated in the introduction to this dissertation, the null hypotheses to this research considered both the soundscape characteristics of a community that experiences change in tree canopy cover and also in potential effects of the soundscape on human directed attention.
The null hypotheses that were tested in the dissertation research were stated as:
4. The percentage of tree canopy cover does not change in northeast Illinois.
5. The soundscapes do not change as tree canopy cover changes.
a. The quantity of sounds do not change as tree canopy cover changes.
b. The quality of sounds do not change as tree canopy cover changes.
397
6. Human directed attention does not change as the quantity and/or quality of sounds
change.
Change in the tree canopy cover.
The first null hypothesis can be rejected. On-the-ground observation and interpretation of before-and-after aerial imagery (Figures 23, 24 and 25 in Chapter V) clearly demonstrates the loss of between 19% and 28% of the tree canopy cover associated with Emerald Ash Borer in neighborhoods of Arlington Heights, Illinois where sound recordings were collected.
Change in the soundscapes.
The second null hypothesis can be rejected in the general terms in which it is stated.
Many, but not all of the measures used to detect changes in the soundscapes associated with the loss of trees yielded statistically significant results, both in terms of the quantity and quality of sounds. The quantity of sounds was considered in terms of the sound intensity (perceived as loudness), the range of frequencies of the sounds, the duration of the sounds, and the entropy (disorder) within the sound signatures. The quality of the sounds was considered in terms of the proportion of sounds associated with humans
(anthrophony), animals (biophony), and non-living sources (geophony). The characteristics of the relative silence in the soundscape recordings were also examined.
398
All of the recording pairs that were compared yielded statistically significant differences in many but not all sound measures for quantity and quality, and the differences that were identified were not always consistent in direction. Therefore it is important to not only find instances of statistical significance in differences between soundscapes but also to identify meaningful patterns in the differences. As discussed in
Chapter VII (Results Part 1 – Analysis of sound recordings) patterns from the comparison of matched recording pairs emerge. The evidence from the detailed analysis of sound recording pairs suggests the following changes to the soundscape occurred as tree were removed in Arlington Heights:
Geophony: The sounds from non-living sources usually associated with weather
are heard more frequently, over a wider range of frequencies, with greater
intensity, and with greater entropy at sites with many trees.
Anthrophony: Human-sourced mechanical sounds are heard more frequently,
over a wider range of frequencies, and with greater intensity, at sites with few
trees.
Biophony: Animal-sourced sounds vary to greater degrees with season and time
of day than with tree density. The removal of the large ash trees apparently did
not substantially reduce nesting habitat for the common songbirds but did
reduce perching locations from which to vocalize for birds, cicadas and
katydids. This may increase competition for preferred vocalization (or
stridulation) perches, which may in turn affect populations of songbirds and
insects.
399
Silence: The evidence for sound attributes associated with intensity (perceived as
“loudness”) suggests that the sites with few trees have greater sound energy in
the background than the sites with many trees. In other words, the silence in
areas with few trees is louder than the silence in areas with many trees.
Change in measures of directed attention.
The third null hypothesis cannot be rejected. Practically no statistically significant differences were found in the results from the tests for directed attention associated with the four sound treatments used during the testing sessions. Additionally, the ordering of mean values for the variables considered in the tests for directed attention revealed no consistent patterns associated with the sound treatments. Why is this? The results from the Necker Cube Pattern Control test and the Stroop Test can be interpreted from three general perspectives:
1. Changes to the soundscapes of Arlington Heights, Illinois associated with the loss
of tree canopy cover did not have an effect on human directed attention as
measured in the research trials.
2. Changes to soundscapes of Arlington Heights, Illinois associated with the loss of
tree canopy cover may have had an effect on human directed attention, but the
methods used in this research failed to detect those effects.
3. Changes to soundscapes of Arlington Heights, Illinois associated with the loss of
tree canopy cover may not affect human directed attention but may affect a
different human characteristic such as degree of stress or agitation.
400
If the first interpretation is correct the methods used in this research may have been appropriate, well-designed and properly executed yielding reliable results. If the second interpretation is correct, then the testing methods need to be improved if changes to measures of directed attention are to be detected. Additionally the participant sample size may need to be substantially larger. If the third interpretation is correct then more appropriate measures, perhaps physiological measures instead of cognitive measures are needed to detect those changes.
I was not successful in demonstrating that sounds affect directed attention. I conclude that the methods that I used for administering the tests for directed attention were flawed.
I still believe that the sounds that surround us somehow affect us (either positively or negatively), and that trees play a role in the characteristics of soundscapes, and that documenting changes in urban soundscapes is a valid method of monitoring ecological health, and possibly some aspects of human cognition.
This study does demonstrate that the soundscape in a single community changed in significant measures as tree canopy changed. The changes were subtle but measurable.
In some aspects the change to the soundscape was as expected and in other ways the changes were surprising in their direction or in their absence.
The tests for directed attention administered during different sound treatments did not reveal significant effects on humans. On the surface this suggests that the change to
Arlington Height’s soundscape as a result of tree loss is not associated with a change in human cognition. But I am not yet convinced that this is the correct take-away finding. I believe that what I learned is that the methods I used to detect the effects on humans associated with the changing soundscape must be improved.
401
Closing Thoughts
Something odd happened. I was flying to someplace that caused me to have a connection at Chicago O’Hare airport. I didn’t have much time between flights and I needed to speed walk through several terminals. I got to my next flight and took my window seat. As the plane rolled toward takeoff I had a strong and unexpected urge to go back to Arlington Heights, just a few miles to the northwest. As the plane lifted from the runway I was straining to see some familiar landmark near the neighborhoods where I had gathered the sound recordings. Afterwards it struck me as odd that I should have a desire to go and sit in a suburban neighborhood that most people (including myself) would describe as common and unremarkable. I have no connection to Arlington Heights other than the fact that I have now listened attentively to Arlington Heights – listening for the purpose of listening – and I enjoyed the experience, as unremarkable as it may seem.
For the past 18 months I have worked on the data analysis and writing of this dissertation, mostly during early morning hours. The majority of this work has been conducted in a small upstairs room with one window that is usually open to some degree
(sometimes meaning I pile on the sweatshirts). It’s not that I sit near the window and actively listen to the neighborhood sounds, but the sounds are present and mostly not consciously noticed. But are they subconsciously noticed? As I am typing now I can hear a dog barking, probably several blocks away – far enough so I am not concerned about what is disturbing him. If he was closer would I feel the need to investigate? Is there a spatial boundary at which sounds that are “background” become sounds that are
402 warnings? The late-night train whistle may seem comforting from a mile away but may inspire terror if I am driving across the tracks.
I have occasionally noticed how the songbirds begin their first alerts before 5 am in the spring and summer, and may not interrupt the crickets in October until well after 7 am
(a Carolina wren at 7:16 on this Sunday morning). I know when it is raining without turning my eyes from my work. If I have forgotten to take out the garbage on
Wednesday night I hear the truck from down the street around 5:30 on Thursday morning, giving me time to respond. There is a constant low hum from distant traffic during the night and very early morning when nothing else is stirring and the air is still, and as creatures and people come awake the background hum is overtaken and masked.
I may be odd in this preference for the presence of outdoor sounds. I may be an outlier variable that should be dealt with as I think about soundscapes and humans. This trait probably clouds my judgment during those “I wonder” moments that have led to this research. But something leads me to think that I am not the only person who benefits from peaceful sounds.
Perhaps the variable that is important is some measure of peacefulness in the soundscape, and that sound quality perhaps translates to signals reassuring the brain that it’s safe to relax. Maybe this signal recognition goes back to the days before humans lived in houses and night brought potential dangers. Perhaps we were much more attentive to the sounds beyond the campfire, and the composition of sounds from approaching weather or agitated birds or unusual silence served as the precursor to smoke detectors and carbon monoxide alarms. Maybe there are still vestiges of attentive listening that tell us when to be alert and when to rest. Maybe this is what I need to
403 figure out how to detect and measure going forward. There are many “perhaps” and
“maybes” and few “statistically significants” at this point.
We can be confident that to some degree the soundscape of Arlington Heights, Illinois changed during the time period when Emerald Ash Borer arrived in the community, killed a significant portion of the urban forest, and caused the removal of thousands of trees. The soundscape analysis conducted as part of this dissertation demonstrates that changes, even if they are subtle, did occur. Obviously the more noticeable changes in the community are visual. There is less shade, there are fewer streets with a green canopy overhead, and to many people (not all but perhaps most) the neighborhoods are not as attractive and inviting as they were before the trees were removed. I base this last belief on the many informal conversations I had with residents who came to know me as I stood next to the sound recorder in their neighborhoods. Admittedly those people who were most likely to stop and engage me in conversation were those people who found my study of trees, sounds, and people to be of interest, and therefore their comments were not representative of a random survey, but never-the-less I heard frequent opinions on how residents missed the large ash trees. One person, without prompting, mentioned that he thought that the traffic sounds from the nearby freeway were more noticeable since the trees were removed. Indeed, the background sounds in the periods of silence in his neighborhood were a little less silent. However, the perceived difference in traffic noise was probably more attributable to the fact that vehicles could now be seen on the freeway from this person’s property, where before the tree canopy provided a visual screen.
Several other people with interests in birds commented that they seemed to hear less birdsong. My sample of sound recordings could not confirm this in a universal sense.
404
There were seasonal variations in bird vocalizations, but I did not detect a distinct decrease in the volume, frequency, or diversity of birdsong at all times following the removal of trees. These observations of decreased bird song may have actually occurred on a small scale near the resident’s property, or the visual impact of diminished tree canopy may have produced an imagined decrease in birds. In either case, there was a difference in how these bird fanciers perceived their surroundings, and the perception was one of loss.
The range of opinions and preferences relevant to a person’s place of residence is vast.
In general given a choice, most people will choose to live in a community that they find is supportive of their physical, mental, and emotional needs, and conversely irritates them the least. On the other hand there are many people who have little choice on where they live. In general, those people with low incomes tend to have the least choice on where they live, and in general those people with low incomes tend to be clustered in densely populated urban neighborhoods. These may be the communities and neighborhoods that would benefit most from urban planning, design and implementation of landscapes that foster feelings of safety and allow periods of relaxation and wonder. These may be the communities and neighborhoods that would benefit most from a clear understanding of how the presence of urban trees and the composition of the urban soundscape affects how humans perceive their environment, and ultimately how neighbors interact with one another.
Lastly, it is appropriate for one last voice to be recognized in this research, that being my voice to be heard saying “Thank you” to the people who have encouraged me and supported me during this research including the residents of Arlington Heights, Illinois,
405 my Dissertation Committee at Cleveland State University, my employers at the Davey
Tree Expert Company, and my family. I hope that this study serves as a foundation for future inquiry by others who wonder about trees, sound and people.
406
REFERENCES
Alvarsson, J. J., Wiens, S., & Nilsson, M. E. (2010). Stress Recovery during Exposure to
Nature Sound and Environmental Noise. International Journal of Environmental
Research and Public Health, 7(3), 1036-1046.
Anderson, L. M., Mulligan, B. E., Goodman, L. S., & Regen, H. Z. (1983). Effects of
Sounds on Preferences for Outdoor Settings. Environment and Behavior, 15(5),
539-566. doi: 10.1177/0013916583155001
Austin, M. E., & Kaplan, R. (2003). Identity, involvement, and expertise in the inner city:
Some benefits of tree-planting projects. In S. Clayton & S. Opotow (Eds.),
Identity and the natural environment: The psychological significance of nature
(pp. 205 - 226). Cambridge, Massachusetts: The MIT Press.
Barber, J. R., Burdett, C. L., Reed, S. E., Warner, K. A., Formichella, C., Crooks, K. R.,
Theobold, D.M., & Fristrup, K. M. (2011). Anthropogenic noise exposure in
protected natural areas: estimating the scale of ecological consequences.
Landscape Ecology, 26(9), 1281-1295. doi: 10.1007/s10980-011-9646-7
Barton, J., & Pretty, J. (2010). What is the Best Dose of Nature and Green Exercise for
Improving Mental Health? A Multi-Study Analysis. Environmental Science &
Technology, 44(10), 3497 - 3955. doi: 10.1021/es903183r
Basu, A. (2015). Bringing Out the Best in Ourselves. In R. Kaplan & A. Basu (Eds.),
Fostering Reasonableness: Supportive Environments for Bringing Out Our Best
(pp. 87-104). Ann Arbor, MI: Michigan Publishing, University of Michigan
Library.
407
Basu, A., & Kaplan, R. (2015). The Reasonable Person Model: Introducing the
framework and the chapters. In R. Kaplan & A. Basu (Eds.), Fostering
Reasonableness: Supportive Environments for Bringing Out Our Best (pp. 1-19).
Ann Arbor, MI: Michigan Publishing, University of Michigan Library.
Beckett, K. P., Freer-Smith, P., & Taylor, G. (2000). Effective tree species for local air
quality management. Journal of Arboriculture, 26(1), 12 - 19.
Belojevic, G., Evans, G. W., Paunovic, K., & Jakovljevic, B. (2012). Traffic noise and
executive functioning in urban primary school children: The moderating role of
gender. Journal of Environmental Psychology, 32(4), 337-341. doi:
10.1016/j.jenvp.2012.05.005
Benfield, J. A., Bell, P. A., Troup, L. J., & Soderstrom, N. C. (2010). Aesthetic and
affective effects of vocal and traffic noise on natural landscape assessment.
Journal of Environmental Psychology, 30(1), 103-111. doi:
10.1016/j.jenvp.2009.10.002
Betts, K. S. (2005). Choosing the right trees to improve urban air. Environmental Science
& Technology, 39(17), 356A-357A.
Brambilla, G., Gallo, V., Asdrubali, F., & Alessandro, F. (2013). The perceived quality of
soundscape in three urban parks in Rome. The Journal of the Acoustical Society of
America, 134(1), 832-839. doi: doi:http://dx.doi.org/10.1121/1.4807811
Bratman, G. N., Hamilton, J. P., & Daily, G. C. (2012). The impacts of nature experience
on human cognitive function and mental health. Annals of the New York Academy
of Sciences, 1249(2012), 118 - 136. doi: 10.1111/j.1749-6632.2011.06400.x
408
Brumm, H. (2006). Animal Communication: City Birds Have Changed Their Tune.
Current Biology, 16(23), R1003-R1004. doi:
http://dx.doi.org/10.1016/j.cub.2006.10.043
Bucur, V. (2006). Urban Forest Acoustics. Berlin, Heidelberg, Germany: Springer -
Verlag.
Byoung-Suk, K., Ulrich, R. S., Walker, V. D., & Tassinary, L. G. (2008). Anger and
Stress: The Role of Landscape Posters in an Office Setting. [Article].
Environment & Behavior, 40(3), 355-381. doi: 10.1177/0013916506298797
Cain, R., Jennings, P., Adams, M., Bruce, N., Carlyle, A., Cusack, P., Davies, W., Hume,
K., Plack, C. J. (2008). SOUND-SCAPE: A framework for characterising positive
urban soundscapes. The Journal of the Acoustical Society of America, 123(5),
3394-3394. doi: 10.1121/1.2934071
Carson, R. (1962). Silent Spring (Twenty-fifth anniversary ed.). Boston: Houghton
Mifflin Company.
Cartwright, L. A., Taylor, D. R., Wilson, D. R., & Chow-Fraser, P. (2014). Urban noise
affects song structure and daily patterns of song production in Red-winged
Blackbirds (Agelaius phoeniceus). Urban Ecosystems, 17(2), 561-572. doi:
10.1007/s11252-013-0318-z
Charif, R. A., Waack, A. M., & Strickman, L. M. (2010). Raven Pro 1.4 User's Manual.
Ithaca NY: Cornell Lab or Orniyhology.
Cimprich, B., & Ronis, D. L. (2003). An environmental intervention to restore attention
in women with newly diagnosed breast cancer. Cancer Nursing, 26(4), 284 - 292.
409
Cohen, S., Glass, D. C., & Singer, J. E. (1973). Apartment noise, auditory discrimination,
and reading ability in children. Journal of Experimental Social Psychology, 9(5),
407-422. doi: 10.1016/s0022-1031(73)80005-8
Coley, R. L., Kuo, F. E., & Sullivan, W. C. (1997). Where does community grow? The
social context created by nature in urban public housing. Environment &
Behavior, 29(4), 468.
Coutts, C. J. (2010). Public Health Ecology. [Opinion]. Journal of Environmental Health,
72(6), 53-55.
De Coensel, B., Boes, M., Oldoni, D., & Botteldooren, D. (2013). Characterizing the
soundscape of tranquil urban spaces. The Journal of the Acoustical Society of
America, 133(5), 3371-3371. doi: 10.1121/1.4805779 de Vries, S., Verheij, R. A., Groenewegen, P. P., & Spreeuwenberg, P. (2003). Natural
environments…healthy environments? An exploratory analysis of the relationship
between greenspace and health. Environment & Planning, 35(10), 1717-1731.
Donovan, G. H., & Prestemon, J. P. (2012). The Effect of Trees on Crime in Portland,
Oregon. Environment and Behavior, 44(1), 3 - 30. doi:
10.1177/0013916510383238
Dumyahn, S. L., & Pijanowski, B. C. (2011). Soundscape conservation. Landscape
Ecology, 26(9), 1327-1344. doi: 10.1007/s10980-011-9635-x
Dunn, A. D. (2010). Siting Green Infrastructure: Legal and Policy Solutions to Alleviate
Urban Poverty and Promote Healthy Communities. Boston College
Environmental Affairs Law Review, 37(1), 41 - 66.
410
Dwyer, J. F., Schroeder, H. W., & Gobster, P. H. (1994). The Deep Significance of
Urban Trees and Forests. In R. H. Platt, R. A. Rowntree & P. C. Muick (Eds.),
The Ecological City: Preserving and restoring urban biodiversity (pp. 137 - 150).
Amherst, MA: The University of Massachusetts Press.
Dzhambov, A. M., & Dimitrova, D. D. (2015). Green spaces and environmental noise
perception. Urban Forestry & Urban Greening, 14(4), 1000-1008. doi:
http://dx.doi.org/10.1016/j.ufug.2015.09.006
Eiseley, L. (1946). The Immense Journey. New York: Vintage Books.
Emerson, R. W., Thoreau, H. D., & Elder, J. (1991). Nature Walking. Boston: Beacon
Press.
Faber Taylor, A., Kuo, F. E., & Sullivan, W. C. (2002). Views of nature and self
discipline: Evidence from inner city children. Journal of Environmental
Psychology, 22(1 - 2), 49 - 63.
Fang, C.-F., & Ling, D.-L. (2003). Investigation of the noise reduction provided by tree
belts. Landscape and Urban Planning, 63(2003), 187 - 195.
Farina, A., Lattanzi, E., Malavasi, R., Pieretti, N., & Piccioli, L. (2011). Avian
soundscapes and cognitive landscapes: theory, application and ecological
perspectives. Landscape Ecology, 26(9), 1257-1267. doi: 10.1007/s10980-011-
9617-z
Francis, C. D., Paritsis, J., Ortega, C. P., & Cruz, A. (2011). Landscape patterns of avian
habitat use and nest success are affected by chronic gas well compressor noise.
Landscape Ecology, 26(9), 1269-1280. doi: 10.1007/s10980-011-9609-z
411
Friedman, M. S., Powell, K. E., Hutwagner, L., Graham, L. M., & Teague, W. G. (2001).
Impact of Changes in Transportation and Commuting Behaviors During the 1996
Summer Olympic Games in Atlanta on Air Quality and Childhood Asthma.
JAMA, 285(7), 897-905. doi: 10.1001/jama.285.7.897
Frumkin, H. (2003). Healthy Places: Exploring the Evidence. [Abstract]. American
Journal of Public Health, 93(9), 1451-1456.
Gidlöf-Gunnarsson, A., & Öhrström, E. (2007). Noise and well-being in urban residential
environments: The potential role of perceived availability to nearby green areas.
Landscape and Urban Planning, 83(2-3), 115-126.
Gilbert, O. L. (1991). The Ecology of Urban Habitats. London, UK: Chapman and Hall.
Goodwin, S. E., & Shriver, W. G. (2011). Effects of Traffic Noise on Occupancy Patterns
of Forest Birds. [Article]. Efectos del Ruido de Tráfico sobre los Patrones de
Ocupación de Aves de Bosque., 25(2), 406-411. doi: 10.1111/j.1523-
1739.2010.01602.x
Hartig, T., Mang, M., & Evans, G. W. (1991). Restorative Effects of Natural
Environment Experiences. Environment and Behavior, 23(1), 3-26.
Hartig, T., & Marcus, C. C. (2006). Healing gardens-places for nature in health care (Vol.
368, pp. 36-37): Lancet.
Hawkins, D. (2012). 'Soundscape ecology': the new science helping identify ecosystems
at risk. Ecologist, 40(31), 5-7.
Hawthorne, D. L. (2014, 12/20). [Personal communication - Raven questions].
412
Herrera-Montes, M. I., & Aide, T. M. (2011). Impacts of traffic noise on anuran and bird
communities. Urban Ecosystems, 14(3), 415-427. doi: 10.1007/s11252-011-0158-
7
Horowitz, S. S. (2012). The Universal Sense: How Hearing Shapes the Mind. New York:
Bloomsbury.
Irvine, K. N., Devine-Wright, P., Payne, S. R., Fuller, R. A., Painter, B., & Gaston, K. J.
(2009). Green space, soundscape and urban sustainability: an interdisciplinary,
empirical study. Local Environment, 14(2), 155-172. doi:
10.1080/13549830802522061
Irwin, A., Hall, D. A., Peters, A., & Plack, C. J. (2011). Listening to urban soundscapes:
Physiological validity of perceptual dimensions. Psychophysiology, 48(2), 258-
268.
Jeng, H. A. C., Englande, A. J., Bakeer, R. M., & Bradford, H. B. (2005). Impact of
urban stormwater runoff on estuarine environmental quality. Estuarine, Coastal
and Shelf Science, 63(4), 513-526. doi: 10.1016/j.ecss.2004.11.024
Jonnes, J. (2016). Urban Forests: A natural history of trees and people in the American
cityscape. New York: Viking Press.
Joo, W., Gage, S. H., & Kasten, E. P. (2011). Analysis and interpretation of variability in
soundscapes along an urban-rural gradient. Landscape and Urban Planning,
103(3-4), 259-276. doi: 10.1016/j.landurbplan.2011.08.001
Kaplan, R. (2001). The Nature of the View From Home: Psychological Benefits.
Environment & Behavior, 33(4), 507.
413
Kaplan, R. (2007). Employees' reactions to nearby nature at their workplace: The wild
and the tame. Landscape and Urban Planning, 82(1-2), 17-24.
Kaplan, R., & Basu, A. (Eds.). (2015). Fostering Reasonableness: Supportive
Environments for Bringing Out Our Best. Ann Arbor, MI: Michigan Publishing,
University of Michigan Library.
Kaplan, R., & Kaplan, S. (1995). The Experience of Nature: A psychological perspective.
Ann Arbor, MI: Ulrich's Bookstore.
Kaplan, S. (1995). The restorative benefits of nature: Toward an integrative framework.
Journal of Environmental Psychology, 15(3), 169-182.
Kaplan, S., & Berman, M. G. (2010). Directed attention as a common resource for
executive functioning and self-regulation. Perspectives on Psychological Science,
5(1), 43-57. doi: 10.1177/1745691609356784
Kaplan, S., & Kaplan, R. (1982). Cognition and Environment: Functioning in an
Uncertain World. New York NY: Praeger.
Kaplan, S., & Kaplan, R. (2003). Health, Supportive Environments, and the Reasonable
Person Model. American Journal of Public Health, 93(9), 1484-1489.
Karlik, J. F., McKay, A. H., Welch, J. M., & Winer, A. M. (2002). A survey of California
plant species with a portable VOC analyzer for biogenic emission inventory
development. [Article]. Atmospheric Environment, 36(33), 5221.
Kellert, S. R. (2002). Values, Ethics, and Spiritual and Scientific Relations to Nature. In
S. R. Kellert & T. J. Farnham (Eds.), The Good in Nature and Humanity:
Connecting Science, Religion and Spirituality with the Natural World (pp. 49-64).
Washington DC: Island Press.
414
Kragh, J. (1981). Road traffic noise attenuation by belts of trees. Journal of Sound and
Vibration, 74(2), 235-241. doi: 10.1016/0022-460x(81)90506-x
Krause, B. (2015). Voices of the Wild: Animal Songs, Human Din, and the Call to Save
Natural Soundscapes. New Haven CT: Yale University Press.
Kuo, F. E., Bacaicoa, M., & Sullivan, W. C. (1998). Transforming Inner-City
Landscapes: Trees, Sense of Safety, and Preference. Environment & Behavior,
30(1), 28-59.
Kuo, F. E., & Sullivan, W. C. (2001a). Aggression and Violence in the Inner City:
Effects of Environment via Mental Fatigue. Environment and Behavior, 33(4),
543-571.
Kuo, F. E., & Sullivan, W. C. (2001b). Environment and Crime in the Inner City: Does
Vegetation Reduce Crime? Environment and Behavior, 33(3), 343-367.
Kuo, F. E., Sullivan, W. C., Coley, R. L., & Brunson, L. (1998). Fertile Ground for
Community: Inner-City Neighborhood Common Spaces. American Journal of
Community Psychology, 26(6), 823-851.
Kuppuswamy, H. (2009). Improving health in cities using green infrastructure: A review.
Forum Ejournal, 9(December 2009), 63 - 76. Retrieved from
http://research.ncl.ac.uk/forum/v9i1/Papers/Kuppuswamy%20(2009)%20Improvi
ng%20Heatlh%20in%20Cities.pdf
Lafortezza, R., Carrus, G., Sanesi, G., & Davies, C. (2009). Benefits and well-being
perceived by people visiting green spaces in periods of heat stress. Urban
Forestry & Urban Greening, 8(2), 97-108.
415
Laumann, K., Garling, T., & Stormark, K. M. (2003). Selective attention and heart rate
responses to natural and urban environments. Journal of Environmental
Psychology, 23(2), 125-134.
Laverne, R. J., & Lewis, G. M. (1996). The effect of vegetation on residential energy use
in Ann Arbor, Michigan. Journal of Arboriculture, 22(5), 234 - 243.
Laverne, R. J., & Lewis, G. M. (2000). Trees and building energy use. In K. K.
Abdollahi, Z. H. Ning & A. Appeaning (Eds.), Global Climate Change and the
Urban Forest (pp. 58 - 69). Baton Rouge, LA: Gulf Coast Regional Climate
Change Council.
Lee, J., Park, B.-J., Yuko, T., Takahide, K., & Yoshifumi, M. (2009). Restorative effects
of viewing real forest landscapes, based on a comparison with urban landscapes.
Scandinavian Journal of Forest Research, 24(3), 227-234. doi:
10.1080/02827580902903341
Maas, J., Verheij, R. A., Groenewegen, P. P., de Vries, S., & Spreeuwenberg, P. (2006).
Green space, urbanity, and health: how strong is the relation? Journal of
Epidemiology & Community Health, 60(7), 587-592. doi:
10.1136/jech.2005.043125
Maller, C., Townsend, M., Pryor, A., Brown, P., & St Leger, L. (2006). Healthy nature
healthy people: 'contact with nature' as an upstream health promotion intervention
for populations. Health Promotion International, 21(1), 45-54.
416
Marans, R. W. (2003). Understanding environmental quality through quality of life
studies: the 2001 DAS and its use of subjective and objective indicators.
Landscape and Urban Planning, 65(1–2), 73-83. doi:
http://dx.doi.org/10.1016/S0169-2046(02)00239-6
Martínez-Sala, R., Rubio, C., García-Raffi, L. M., Sánchez-Pérez, J. V., Sánchez-Pérez,
E. A., & Llinares, J. (2006). Control of noise by trees arranged like sonic crystals.
Journal of Sound and Vibration, 291(1–2), 100-106. doi:
http://dx.doi.org/10.1016/j.jsv.2005.05.030
Matteo, M., Randhir, T., & Bloniarz, D. (2006). Watershed-Scale Impacts of Forest
Buffers on Water Quality and Runoff in Urbanizing Environment. [Article].
Journal of Water Resources Planning & Management, 132(3), 144-152. doi:
10.1061/(asce)0733-9496(2006)132:3(144)
McQuay, B., & Joyce, C. (Producer). (2016). Beyond Sightseeing: You'll Love The
Sound Of America's Best Parks. National Public Radio: Morning Edition. [Radio
story] Retrieved from http://www.npr.org/2016/06/29/483241647/beyond-
sightseeing-youll-love-the-sound-of-americas-best-parksisten•7
Mitchell, R., & Popham, F. (2007). Greenspace, urbanity and health: relationships in
England. Journal of Epidemiology & Community Health, 61(8), 681-683. doi:
10.1136/jech.2006.053553
Mitchell, R., & Popham, F. (2008). Effect of exposure to natural environment on health
inequalities: an observational population study. Lancet, 372(9650), 1655-1660.
doi: 10.1016/s0140-6736(08)61689-x
417
Moser, G. (1988). Urban stress and helping behavior: Effects of environmental overload
and noise on behavior. Journal of Environmental Psychology, 8(4), 287-298. doi:
10.1016/s0272-4944(88)80035-5
Muir, J. (2014). A wind storm in the forests. Carlisle MA: Applewood Books.
Nassiri, P., Monazam, M., Dehaghi, F., Ghavam Abadi, L. I., Zakerian, S., & Azam, K.
(2013). The Effect of Noise on Human Performance: A Clinical Trial. [Article].
International Journal of Occupational & Environmental Medicine, 4(2), 87-95.
Nowak, D. J. (1993). Atmospheric Carbon Reduction by Urban Trees. Journal of
Environmental Management, 37(3), 207-217.
Nowak, D. J., & Crane, D. E. (2002). Carbon storage and sequestration by urban trees in
the USA. Environmental Pollution, 116(3), 381-389.
Nowak, D. J., Hoehn, R., & Crane, D. E. (2007). Oxygen Production by Urban Trees in
the United States. Arboriculture & Urban Forestry, 33(3), 220-226.
Olmsted, F. L. (1995). The Yosemite Valley and the Mariposa Grove of Big Trees: A
Preliminary Report, 1865. San Francisco CA: Yosemite Conservancy.
Payne, S. R. (2013). The production of a Perceived Restorativeness Soundscape Scale.
Applied Acoustics, 74(2), 255-263. doi: 10.1016/j.apacoust.2011.11.005
Pijanowski, B. C., Villanueva-Rivera, L. J., Dumyahn, S. L., Farina, A., Krause, B. L.,
Napoletano, B. M., Gage, S.H., & Pieretti, N. (2011). Soundscape Ecology: The
Science of Sound in the Landscape. BioScience, 61(3), 203-216.
Ratcliffe, E., Gatersleben, B., & Sowden, P. T. (2013). Bird sounds and their
contributions to perceived attention restoration and stress recovery. Journal of
Environmental Psychology, 36, 221-228. doi: 10.1016/j.jenvp.2013.08.004
418
Samara, T., & Tsitsoni, T. (2011). The effects of vegetation on reducing traffic noise
from a city ring road. [Article]. Noise Control Engineering Journal, 59(1), 68-74.
Schafer, R. M. (1994). The Soundscape: Our sonic environment and the tuning of the
world. Rochester, Vermont: Destiny Books.
Schapkin, S. A., Falkenstein, M., Marks, A., & Griefahn, B. (2006). Executive brain
functions after exposure to nocturnal traffic noise: effects of task difficulty and
sleep quality. European Journal of Applied Physiology, 96(6), 693-702. doi:
10.1007/s00421-005-0049-9
Schulte-Fortkamp, B., & Fiebig, A. (2006). Soundscape Analysis in a Residential Area:
An Evaluation of Noise and People's Mind. Acta Acustica united with Acustica,
92(6), 875-880.
Scott, K. I., Simpson, J. R., & McPherson, E. G. (1999). Effects of tree cover on parking
lot microclimate and vehicle emissions. Journal of Arboriculture, 25(3), 129 -
142.
Shepherd, D., Welch, D., Dirks, K. N., & Mathews, R. (2010). Exploring the
Relationship between Noise Sensitivity, Annoyance and Health-Related Quality
of Life in a Sample of Adults Exposed to Environmental Noise. International
Journal of Environmental Research and Public Health, 7(10), 3579-3594.
Sherman, S. A., Varni, J. W., Ulrich, R. S., & Malcarne, V. L. (2005). Post-occupancy
evaluation of healing gardens in a pediatric cancer center. Landscape and Urban
Planning, 73(2-3), 167-183.
Slabbekoorn, H., & den Boer-Visser, A. (2006). Cities Change the Songs of Birds.
Current Biology, 16(23), 2326-2331. doi: 10.1016/j.cub.2006.10.008
419
Slabbekoorn, H., & Ripmeester, E. A. P. (2008). Birdsong and anthropogenic noise:
implications and applications for conservation. Molecular Ecology, 17(1), 72-83.
doi: 10.1111/j.1365-294X.2007.03487.x
Sommer, R. (2003). Trees and Human Identity. In S. Clayton & S. Opotow (Eds.),
Identity and the Natural Environment: The Psychological Significance of Nature
(pp. 179 - 204). Cambridge, Massachusetts: The MIT Press.
Southworth, M. F. (1969). The sonic environment of cities. Environment & Behavior, 1,
49 - 70.
Statistics Canada. (2016). Population and Dwelling Count Highlight Tables, 2011
Census. Retrieved September 1, 2016, from https://www12.statcan.gc.ca/census-
recensement/2011/dp-pd/hlt-fst/pd-pl/Table-
Tableau.cfm?LANG=Eng&T=1201&SR=1001&S=22&O=A&RPP=100&PR=0
&CMA=0
Sternberg, E. M. (2009). Healing Spaces: The science of place and well-being.
Cambridge, Massachusetts: The Belknap Press of Harvard University Press.
Sullivan, W. C. (2015). In search of a clear head. In R. Kaplan & A. Basu (Eds.),
Fostering Reasonableness: Supportive Environments for Bringing Out Our Best
(pp. 54-69). Ann Arbor, MI: Michigan Publishing, University of Michigan
Library.
Sullivan, W. C., Kuo, F. E., & Depooter, S. F. (2004). The Fruit of Urban Nature: Vital
Neighborhood Spaces. Environment and Behavior, 36(5), 678-700. doi:
10.1177/0193841x04264945
420
Svendsen, E. S. (2009). Cultivating resilience: Urban stewardship as a means to
improving health and well-being. In L. Campbell & A. Wiesen (Eds.), Restorative
Commons: Creating Health and Well-being through Urban Landscapes Gen.
Tech. Rep. NRS-P-39 (pp. 278). Newtown Square, PA: U.S. Department of
Agriculture Forest Service.
Szalma, J. L., & Hancock, P. A. (2011). Noise Effects on Human Performance.
Psychological Bulletin, 137(4), 682-707. doi: 10.1037/a0023987
Takano, T., Nakamura, K., & Watanabe, M. (2002). Urban residential environments and
senior citizens’ longevity in megacity areas: the importance of walkable green
spaces. Journal of Epidemiology and Community Health, 56(12), 913-918. doi:
10.1136/jech.56.12.913
Tippett, K., & Sternberg, E. M. (2010). Knowing how to heal ourselves: Stress and the
balance within. In: Einstein's God (pp. 197 - 221). New York: Penguin Books.
Tse, M. S., Chau, C. K., Choy, Y. S., Tsui, W. K., Chan, C. N., & Tang, S. K. (2012).
Perception of urban park soundscape. The Journal of the Acoustical Society of
America, 131(4), 2762-2771. doi: 10.1121/1.3693644
Twombly, R. (Ed.). (2010). Frederick Law Olmsted: Essential Texts. New York: W. W.
Norton and Company.
U.S. Census, Bureau. (2000). United States -- Urban / Rural and Inside / Outside
Metropolitan Area Retrieved July 31, 2010, from
http://factfinder.census.gov/servlet/GCTTable?_bm=y&-geo_id=01000US&-
_box_head_nbr=GCT-P1&-ds_name=DEC_2000_SF1_U&-format=US-1
421
Ulrich, R. S. (1984). View through a window may influence recovery from surgery.
Science, 224(4647), 420 - 421. doi: 10.1126/science.6143402
Ulrich, R. S., Simons, R. F., Losito, B. D., Fiorito, E., Miles, M. A., & Zelson, M. (1991).
Stress recovery during exposure to natural and urban environments. Journal of
Environmental Psychology, 11, 201 - 230.
United States Census Bureau. (2016). American Fact Finder Retrieved September 1,
2016, from
http://factfinder.census.gov/faces/nav/jsf/pages/community_facts.xhtml#
USNaviguide. (2013). Zip Code Finder and Boundary Map. Retrieved September 1,
2016, from http://maps.huge.info/zip.htm
Van Buren, M. A., Watt, W. E., Marsalek, J., & Anderson, B. C. (2000). Thermal
enhancement of stormwater runoff by paved surfaces. Water Research, 34(4),
1359-1371. van den Berg, A. E., Hartig, T., & Staats, H. (2007). Preference for Nature in Urbanized
Societies: Stress, Restoration, and the Pursuit of Sustainability. [Article]. Journal
of Social Issues, 63(1), 79-96. doi: 10.1111/j.1540-4560.2007.00497.x
Van Renterghem, T., & Botteldooren, D. (2016). View on outdoor vegetation reduces
noise annoyance for dwellers near busy roads. Landscape and Urban Planning,
148, 203-215. doi: 10.1016/j.landurbplan.2015.12.018
Van Wieren, G. (2008). Ecological Restoration as Public Spiritual Practice. [Article].
Worldviews: Environment Culture Religion, 12(2/3), 237-254. doi:
10.1163/156853508x360000
422
Weinstein, N., Przybylski, A. K., & Ryan, R. M. (2009). Can Nature Make Us More
Caring? Effects of Immersion in Nature on Intrinsic Aspirations and Generosity.
Personality and Social Psychology Bulletin, 35(10), 1315-1329.
Westphal, L. M. (2003). Urban greening and social benefits: A study of empowerment
outcomes. Journal of Arboriculture, 29(3), 137 - 147.
Wiley, R. H. (2015). Noise Matters: The Evolution of Communication. Cambridge MA:
Harvard University Press.
Won Sop, S., Hon Gyo, K., Hammitt, W. E., & Bum Soo, K. (2005). Urban forest park
use and psychosocial outcomes: A case study in six cities across South Korea.
Scandinavian Journal of Forest Research, 20(5), 441-447. doi:
10.1080/02827580500339930
Wood, W. E., Yezerinac, S. M., & Dufty, J. A. M. (2006). Song sparrow (Melospiza
melodia) song varies with urban noise. The Auk, 123(3), 650-659. doi:
10.1642/0004-8038(2006)123[650:ssmmsv]2.0.co;2
Xiao, Q., & McPherson, E. G. (2002). Rainfall interception by Santa Monica’s municipal
urban forest. Urban Ecosystems, 6(4), 291-302.
423
APPENDICES
424
APPENDIX A Sound recording locations, dates and recording length
Address Date Start Time Length (Hr:Min:Sec) 2632 Prindle Ave. 7/17/2013 7:00 AM 1:02:00 Huron St. 9/4/2013 10:29 AM 1:04:00 2632 Prindle Ave. 9/4/2013 6:49 AM 1:01:00 Huron St. 9/5/2013 7:16 AM 1:04:00 1611 Shirra Ct. 9/5/2013 8:31 AM 1:14:00 1404 Fleming Dr. N 10/8/2013 10:04 AM 1:05:00 Huron St. 10/8/2013 11:43 AM 0:40:00 1611 Shirra Ct. 10/8/2013 8:22 AM 1:04:00 Huron St. 12/28/2013 9:09 AM 1:05:00 2632 Prindle Ave. 12/28/2013 11:50 AM 1:04:00 1611 Shirra Ct. 12/28/2013 9:09 AM 1:05:00 Suffield Dr. & Forest Ln. 12/28/2013 11:50 AM 1:04:00 1404 Fleming Dr. N 4/11/2014 3:27 PM 1:04:00 Huron St. 4/11/2014 1:42 PM 1:03:00 1404 Fleming Dr. N 4/12/2014 2:30 PM 2:05:00 Huron St. 4/12/2014 5:54 AM 4:21:00 2632 Prindle Ave. 4/12/2014 9:07 AM 2:07:00 1611 Shirra Ct. 4/12/2014 6:09 AM 2:06:00 Huron St. 4/13/2014 7:00 AM 6:34:00 1611 Shirra Ct. 4/13/2014 11:00 AM 2:04:00 Suffield Dr. & Forest Ln. 4/13/2014 7:00 AM 3:05:00 Huron St. 5/7/2014 5:42 AM 10:15:00 1611 Shirra Ct. 5/7/2014 5:52 AM 3:38:00 2632 Prindle Ave. 5/7/2014 2:10 PM 1:35:00 Huron St. 5/8/2014 5:35 AM 8:09:00 1611 Shirra Ct. 5/8/2014 5:44 AM 6:09:00 Huron St. 5/19/2014 1:30 PM 0:01:00 1611 Shirra Ct. 5/19/2014 1:46 PM 1:03:00 Suffield Dr. & Forest Ln. 5/19/2014 3:04 PM 1:04:00 Huron St. 5/20/2014 5:22 AM 10:21:00 1611 Shirra Ct. 5/20/2014 5:30 AM 7:21:00 Huron St. 5/21/2014 5:27 AM 7:53:00 2632 Prindle Ave. 5/21/2014 5:40 AM 6:06:00 1611 Shirra Ct. 7/15/2014 2:00 PM 1:04:00 2632 Prindle Ave. 7/15/2014 3:32 PM 1:05:00 Huron St. 7/16/2014 5:45 AM 9:45:00 1611 Shirra Ct. 7/16/2014 5:54 AM 4:08:00 2632 Prindle Ave. 7/16/2014 7:45 AM 1:33:00 1610 Peachtree St. 7/16/2014 9:15 AM 1:12:00 1404 Fleming Dr. N 7/16/2014 10:41 AM 1:04:00
425
Appendix A: Sound recording locations, dates and recording length (continued)
Address Date Start Time Length (Hr:Min:Sec) Huron St. 9/3/2014 1:30 PM 1:31:00 1611 Shirra Ct. 9/3/2014 1:47 PM 6:02:00 Huron St. 9/4/2014 5:58 AM 6:07:00 1611 Shirra Ct. 9/4/2014 6:03 AM 1:32:00 2632 Prindle Ave. 9/4/2014 2:10 PM 2:12:00 2622 Stratford Rd. 9/5/2014 5:48 AM 3:14:00 2632 Prindle Ave. 9/5/2014 5:54 AM 3:03:00 Huron St. 9/5/2014 9:29 AM 5:35:00 1611 Shirra Ct. 9/5/2014 9:35 AM 2:03:00 1404 Fleming Dr. N 9/5/2014 12:30 PM 2:16:00 1611 Shirra Ct. 9/6/2014 6:00 AM 3:10:00 Huron St. 10/7/2014 12:53 PM 3:41:00 1611 Shirra Ct. 10/7/2014 1:01 PM 1:33:00 1404 Fleming Dr. N 10/7/2014 2:47 PM 1:33:00 Huron St. 10/8/2014 6:00 AM 7:00:00 1611 Shirra Ct. 10/8/2014 6:00 AM 3:21:00 1404 Fleming Dr. N 10/8/2014 10:04 AM 2:05:00 2625 Stratford Rd. 10/9/2014 6:45 AM 3:15:00 2632 Prindle Ave. 10/9/2014 6:45 AM 3:15:00 2632 Prindle Ave. 10/19/2014 6:47 AM 3:11:00 2625 Stratford Rd. 10/19/2014 6:45 AM 3:11:00 Huron St. 12/27/2014 9:11 AM 1:03:00 1611 Shirra Ct. 12/27/2014 9:11 AM 1:03:00 Suffield Dr. & Forest Ln. 12/27/2014 11:53 AM 1:09:00 2632 Prindle Ave. 12/27/2014 11:53 AM 1:09:00 2625 Stratford Rd. 12/28/2014 11:14 AM 2:06:00 2632 Prindle Ave. 12/28/2014 11:14 AM 2:06:00 Huron St. 12/29/2014 9:56 AM 1:04:00 1611 Shirra Ct. 12/29/2014 9:56 AM 1:04:00 2625 Stratford Rd. 12/29/2014 12:17 PM 1:03:00 2632 Prindle Ave. 12/29/2014 12:17 PM 1:03:00 2632 Prindle Ave. 4/10/2015 5:55 AM 6:17:00 2625 Stratford Rd. 4/10/2015 5:55 AM 6:17:00 Huron St. 4/10/2015 1:03 PM 2:13:00 1404 Fleming Dr. N 4/10/2015 3:27 PM 0:18:00 Huron St. 4/11/2015 5:41 AM 5:06:00 1611 Shirra Ct. 4/11/2015 5:41 AM 2:40:00 2632 Prindle Ave. 4/11/2015 8:41 AM 5:09:00 2625 Stratford Rd. 4/11/2015 11:10 AM 3:07:00 1404 Fleming Dr. N 4/11/2015 2:27 PM 4:08:00 Huron St. 4/12/2015 6:18 AM 6:09:00 Suffield Dr. & Forest Ln. 4/12/2015 6:30 AM 3:32:00 1611 Shirra Ct. 4/12/2015 10:20 AM 2:07:00
426
Appendix A: Sound recording locations, dates and recording length (continued)
Address Date Start Time Length (Hr:Min:Sec) Huron St. 9/3/2014 1:30 PM 1:31:00 2632 Prindle Ave. 5/5/2015 1:44 PM 3:07:00 2625 Stratford Rd. 5/5/2015 1:44 PM 3:07:00 Huron St. 5/6/2015 5:37 AM 10:33:00 1611 Shirra Ct. 5/6/2015 5:45 AM 6:13:00 Raleigh St. at Wilke Rd. 5/6/2015 12:22 PM 1:02:00 2632 Prindle Ave. 5/6/2015 1:53 PM 2:02:00 2625 Stratford Rd. 5/7/2015 5:29 AM 3:05:00 2632 Prindle Ave. 5/7/2015 5:29 AM 2:59:00 Huron St. 5/18/2015 5:24 AM 7:28:00 1611 Shirra Ct. 5/18/2015 5:31 AM 6:12:00 2625 Stratford Rd. 5/18/2015 1:24 PM 3:05:00 2632 Prindle Ave. 5/18/2015 1:24 PM 3:05:00 2625 Stratford Rd. 5/19/2015 5:27 AM 6:13:00 2632 Prindle Ave. 5/19/2015 5:27 AM 6:13:00 2625 Stratford Rd. 8/13/2015 5:56 AM 6:21:00 2632 Prindle Ave. 8/13/2015 5:56 AM 6:21:00 TOTAL 326:16:00
427
APPENDIX B Example of a sound recording field data sheet
428