XQ

5‘
I

 

LIBRARY
Michigan State
University

 

 

 

This is to certify that the
dissertation entitled

ENVIRONMENTAL ACOUSTICS AS AN ECOLOGICAL
VARIABLE TO UNDERSTAND THE DYNAMICS OF
ECOSYSTEMS

presented by

WOOYEONG JOO

has been accepted towards fulfillment
of the requirements for the

Doctoral degree in Zoology

 

 

SWIM

Major Pro iessoifi Signature
M / é’; We
0 ' /

Date

 

MSU is an Affirmative Action/Equal Opportunity Employer

 

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

 

DATE DUE DATE DUE DATE DUE

 

JUNE 8 30m

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5108 KlPro;/Acc8.Pres/CIRC/DaIeDue Indd

 

 

 

ENVIRONMENTAL ACOUSTICS AS AN ECOLOGICAL VARIABLE TO
UNDERSTAND THE DYNAMICS OF ECOSYSTEMS

By

Wooyeong 100

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

ZOOLOGY

2009

ABSTRACT

ENVIRONMENTAL ACOUSTICS AS AN ECOLOGICAL VARIABLE TO
UNDERSTAND THE DYNAMICS OF ECOSYSTEMS

By

Wooyeong 100
Although acoustic variables play a key role in understanding the ecology and behavior of
vocal organisms, little work has been done to investigate whether acoustic signals can
serve as an ecological variable to assess the current state of ecosystems. Our research
was guided by two overarching questions. The first question is can environmental
acoustics be used as ecological attributes that reflect ecosystem structure and processes?
The second question is can environmental acoustics provide a key means to measure and
monitor the biodiversity and distribution of vocal species? The study first developed
analytical methods to understand acoustic properties including: 1) development and
refinement of an Acoustic Habitat Quality Index using the distribution of acoustic power
across different frequency spectrum bands; and 2) measurement and analysis of
vocalizing species diversity using multiple methods of recording acoustic signals. The
second part of the study investigated a new approach to surveying avian species using
acoustic recordings. This analysis revealed that automated acoustic recordings facilitated
simultaneous breeding bird surveys at multiple locations with minimal variability and
high accuracy of bird community measures. Third, the study characterized the urban-
niral variability using environmental sounds based on quantification of environmental
acoustic properties across a gradient of ecosystems and landscapes. Finally, the study
illustrated that using wireless sensor networks as a new sampling tool in ecology and

environmental science provides tremendous opportunities to measure and monitor

complex ecological variables at relevant spatial and temporal scales. The integration of
acoustic research with the multi-science communities and advances in wireless sensor
networks will potentially enable and enhance our understanding of ecological change and
our ability to forecast changes in complex, interconnected ecosystems at scales ranging

from the ecosystem to global level

ACKNOWLEDGEMENTS

I would like to give special thanks to my advisor, Dr. Stuart Gage, for his
tremendous support, encouragement, and guidance to complete my dissertation. He
always listens to my questions and difficulties. He is my great mentor during my doctoral
program. I would also like to thank all my other graduate committee members, Drs.

J iaguo Qi, Jianguo Liu, R. Jan Stevenson, and Stephen K. Harniton for their helpful
suggestions and positive comments. Although Dr. Brian Maurer is not my committee, he
helped me to better understand acoustic data using multivariate statistic analyses. 1 also
give my gratitude to Dr. Fred Dyer for his care and efforts in establishing my program
when I had a crisis to find my advisor and resource.

This study would not have been possible without help from my laboratory
colleagues, Jordan Fox, Kurt Anderson, and Whitney Knollenberg for collecting and
processing acoustic samples during the field works. Erik Enbody helped me to survey and
identify bird species from the acoustic recordings. Eric Kasten and R. Duncan Selby

reviewed my dissertation and provided me with great comments and suggestions.

Much of the research was funded and supported by the NSF LTER Program at
KBS, the Michigan Agricultural Experiment Station, and the Sustainable Michigan
Endowed Project (SMEP), Michigan State University. The Park Department in Ingham
County, Land Management in Michigan State University, and 11 Lansing residents

provided the sites where to conduct the field surveys.

My family and friends also helped and encouraged me to carry out this research

especially when I was frustrated or encountered some issues. 1 am indebted to my wife

iv

Jungah Lee for her support and input. She was taking care of many family works so that 1
can focus on writing my dissertation. She also was my great reviewer with great insight
and different perspective. My parents also helped my research by spending much of time
with my kid, Micah Joo. They stood by me by praying for my health and research. My
son Micah gave me new energy and power to keep carrying out my research by smiling
and hugging. Finally, I would also like to thank my long time friends, Taejin Kim,
Yunkyung, Pae, Father Ilya, Ilyoung Han, Mikyung Jang, Rosa Yun, Yunsoon Kim,

Susan Vogt, and Jim Vogt who lent a listening ear when I went through rough time.

TABLE OF CONTENTS

LIST OF TABLES ................................................................................................ viii
LIST OF FIGURES ............................................................................................... ix
Chapter One Introduction to Soundscape and Acoustic Ecology .......................... 1
Introduction ..................................................................................................... 1
Concept of Soundscape .................................................................................. 2
Interrelationships between the landscape and the soundscape ..................... 6

Chapter Two Development of Analytical Framework to Understand the Structure

and Components of Environmental Acoustic Signals .......................................... 10
Background of acoustics ............................................................................... 10
Analytical tools to quantify and analyze environmental sounds .................... 13

Generalized Sound Classification Analysis ............................................ 14

Chapter Three Use of acoustic recordings for surveying avian species richness

and distribution .................................................................................................... 23
Introduction ................................................................................................... 23
Methods ........................................................................................................ 26

Study area .............................................................................................. 26
Survey methods ..................................................................................... 26
Data Analysis ......................................................................................... 28
Results .......................................................................................................... 30
Avian community similarities among the survey types ........................... 30
Relationship between the survey types and species richness ............... 32
Species abundance estimate analysis ................................................... 32
Temporal pattern of bird species richness ............................................. 34
Discussion ..................................................................................................... 37

Chapter Four Analysis and Interpretation of the “Heartbeat of the City” using

Acoustic Signatures along an Urban-rural Gradient ............................................ 41
Introduction ................................................................................................... 41
Methods ........................................................................................................ 45

Study area .............................................................................................. 45

vi

Survey methods ..................................................................................... 46

Data Analysis ......................................................................................... 47
Results .......................................................................................................... 51
Discussion ..................................................................................................... 61

Chapter Five Development of Automated Acoustic Sensor Observation SYstem

via Wireless Networks ......................................................................................... 68
Introduction ................................................................................................... 68
Design of an ecological wireless sensor network system ............................. 71
Case study .................................................................................................... 75

Deployment of Habitat Sensor/Server System ....................................... 75
Analyses and interpretation of sensor observations .............................. 78
Conclusions .................................................................................................. 82

Chapter Six Summary and Conclusions .............................................................. 84

Appendices .......................................................................................................... 87

Bibliography ......................................................................................................... 98

vii

LIST OF TABLES

Table 3. 1. The number of bird species identified by point count survey (PCS),
manual acoustic survey (MAS), and automated acoustic survey (AAS) in Essex
and Westphalia Township. .................................................................................. 31

Table 3. 2. Bird detection accuracy of all three survey types, based on the total
number of species identified in Essex and Westphalia Township. ...................... 31

Table 3. 3. Community similarity measures (Jaccard’s coefficient / Sorensen
coefficient) to investigate similarity of avifauna community between Point Count
Survey (PCS) and counts from two acoustic recordings: Manual Acoustic
Surveys (MAS) & Automated Acoustic Surveys (AAS). ...................................... 32

Table 4. 1. Temporal pattern of mean Acoustic Habitat Quality Indices (AHQIs)
over months and at six times of day in all land use types. .................................. 63

viii

LIST OF FIGURES

Figure 1.1 Diagram of the classification of a soundscape classification (modified
from Gage et al. (2001) ......................................................................................... 4

Figure 2. 1. The acoustic frequency slicing procedure. a) Each sound wave file is
divided into 11 frequency bands, and b) the relative mean intensity is calculated
for each band. Note that 5 kHz band has the highest mean intensity among 11
frequency bands. ................................................................................................. 15

Figure 2.2. Three main classes of environmental sounds: 1) biophony, 2)
anthrophony, 3) geophony, based on the location of the spectrum frequency
bands. .................................................................................................................. 17

Figure 2. 3. Principal Component loadings of different frequency bands on the
first three principal components from the acoustic data collected at the Long
Term Ecological Research site in W. K. Kellogg Biological Station. The labels on
X axis represent the acoustic frequency bands (e.g., L2 refers to the frequency
band ranges from 1 to 2 kHz). ............................................................................. 20

Figure 2. 5. Avian species were identified by listening to the recordings. The
number of bird species identified is shown on the left and the number of calls
identified is shown on the right. ........................................................................... 22

Figure 2. 7. a) The relationship between avian species richness and the number
of calls, and b) the relationship between acoustic species diversity and PC3
scores (indicating the biological acoustic energy). .............................................. 22

Figure 3. 1. Field bird survyes using a point count method and acoustic
recordings in Clinton County, Michigan during June 2007: a) field configuration
of the Manual Acoustic Recording (MAR) method using an Omni-directional
microphone and a digital acoustic recorder; b) simultaneous bird surveys by
human observation and MAR; c) a digital acoustic recorder (Tascam HD-P2)
deployed in a point; and d) field deployment of an acoustic recording unit
including an audio cassette-tape recorder with a timer (SanGeanVersaCorder)
and a omni-directional microphone (330-3020) for Automated Acoustic Recording
(AAR). .................................................................................................................. 28

Figure 3. 2. Mean (:1: SE) of number of bird species at 25 survey points where all
three surveys were conducted in Essex Township and Westphalia Township, MI.

ix

PCS, AAS, and MAS refer to three bird survey methods: Point Count Survey,
Automated Acoustic Survey, and Manual Acoustic Survey, respectively. ........... 33

Figure 3. 3. The relationship between cumulative species richness and time of
observation using automated recorders .............................................................. 35

Figure 3. 4. Mean (1: SE) of number of bird species at 25 survey points where
AASs were conducted in Essex Township and Westphalia Township from 500 to
1000 hours ........................................................................................................... 36

Figure 4. 1. Conceptual relationship between biological and anthropogenic
acoustic attributes including acoustic energy at different acoustic frequency
range, the number of species and vocalizations identified along urban-rural
gradients. (Modified from Stevenson et al. (2004)) ............................................ 44

Figure 4. 2. Map of the study area in the Greater Lansing area, Ml. Each named
symbol in the map represents the location where an acoustic monitoring unit was
deployed along an urban- rural gradient. The green symbols represent data that
were collected from February to December 2006, and the red symbols are the
locations where the recording devices were lost during the study period. .......... 46

Figure 4. 3. Average percent of land covered by a) forest, b) lawn, c)
pasture/crops, and d) buildings/pavement estimated by land cover maps and
aerial photo imagery (mean + standard error) ..................................................... 51

Figure 4. 4. 3) Average sound levels, b) Acoustic Habitat Quality Index (AHQI),
c) proportion of anthropogenic sounds (Anthrophony), and d) proportion of
biological sounds (biophony) of 5 different land use types where the acoustic
samples were recorded from February to December, 2006. The land use types
was listed on x-axis along an urban-rural gradient, defined by ranking the land
use types in order from the highest (left) to lowest (right) degree of urban
development. ....................................................................................................... 52

The bars represent mean t SE (n = 8475). Lowercase letters refer to means
contrasts among different land use types using Tukey’s HSD tests. ................... 53

Figure 4. 5. Temporal changes in Acoustic Habitat Quality Index (AHQI) at 5
different land use types: (a) commercial; (b) agricultural; (c) urban residential; (d)
urban park; and (e) rural residential sites. Data were collected 6 times a day for
two consecutive days at each month from February 14 to December 12, 2006.
Bars indicate average values of AHQI at each time of day during the month. ....54

Figure 4. 6. Avian community measures along the urban-rural gradient from May
and June 2006: a) mean number of bird species, b) mean number of encounters,
and 0) mean value of Shannon’s diversity index at each landscape. Note that the
bird species and their vocalizations were identified from 4 acoustic samples at

each site during the period. ................................................................................. 55

Figure 4. 7. Relative occurrence probability of avian species across all sites,
listed in order from highest occurrence probability in commercial sites to greatest
occurrence probability in rural residential sites (n = number of encounters/ total
acoustic surveys) ................................................................................................. 57

Figure 4. 8. Ordination diagrams of canonical correspondence analysis (CCA)
using bird survey data with environmental variables. (a) CCA ordination of 28 bird
species (points) against 6 environmental variables (arrows; the percent areas of
different land cover types: forests, lawn, pasture/crops, and buildings/pavement).
The contribution of each environmental variable to the ordination axes is
represented as the length of the arrow. The direction and distance of species
scores to the arrows indicate how well the abundance of each species is related
to each environmental variable (ter Braak 1986, Palmer 1993, Blair 1996). The
angle between the arrows represents correlation between environmental
variables. The smaller angle the arrows have, the greater the environmental
variables are correlated. For example, the Forest and Lawn variables are highly
correlated to one another. Four letter abbreviations represent species alpha
code(Sharpe 1886). (b) CCA ordination of 18 sampling sites and linear
combinations of environmental variables. The orientations of the site scores to
the arrows represent how strongly the sites are related to environmental
variables. See Appendix I for bird names and AOU codes. ............................... 59

Figure 4. 9. Distribution of Acoustic Habitat Quality Index (the normalized ratio of
biological sounds to anthropogenic sounds) based on National Land Cover Data
map 2001. The index ranges from -1 to 1; positive values indicate that the
intensity of biological sounds is higher than one of anthropogenic sounds, and
vice versa ............................................................................................................. 65

Figure 5. 1. Habitat sensor platform (HSP) hardware configuration. .................. 72

Figure 5. 2. Screen shots of the sensor management application developed with
a web-based program (left) and near real-time sensor observations from an
acoustic and image sensor from the KBS-LTER site (right). ............................... 74

Figure 5. 3. a) Map of the distribution of Habitat Sensor Platforms and Habitat
Server at KBS-LTER site (left). The Habitat Sensor Platform b) hardware

xi

components including a Crossbow Stargate processor, 12 to 5v power converter,
an acoustic sensor and a web camera, wireless network card (802.11b), and a 1

GB Compact Flash storage device; and c) field configuration consisting of Habitat
Sensor Platform, solar panel, and 12v deep cycle battery. ................................. 77

Figure 5. 4. Diagram of the advanced wireless Habitat Sensor Network System
using a wireless bridge system. ........................................................................... 78

Figure 5. 5. The mean Acoustic Habitat Quality Indices at 4 different habitat
types in the KBS-LTER site. The bars represent mean + SE (n=11,901). ......... 79

Figure 5. 6. The diurnal patterns of environmental acoustics based on
computation of AHQI in 4 different habitats including a) poplar, b) successional,
c) wheat with no-till treatment, and d) wheat with till treatment. .......................... 81

Figure 6. 1. A relationship between Acoustic Habitat Quality Index and Habitat
Quality Index at 14 grass-woodland sites in Australia on November 30, 2006
(permitted by Gage et al.). ................................................................................... 85

xii

Chapter One
Introduction to Soundscape and Acoustic Ecology

Introduction

It has long been recognized that research on acoustic signals has enabled us to
understand behavior and communication of vocal species. For instance, the North
American Breeding Bird Survey (BBS), one of the largest long-term, national-scale avian
monitoring programs, has been conducted for more than 30 years, based on auditory and
visual cues observed by humans (Bystrak I981, Robbins et al. 1986, Ralph et al. 1993,
Sauer et al. 1994, Ralph et al. 1995). Weir and Mossman (2005) noted in their discussion
of the North American Amphibian Monitoring Program (NAAMP) that identifying
amphibian species has been done primarily by listening to their calls. Thus, identifying
vocal organisms by auditory cues enabled the BBS and the NAAMP to monitor the
abundance and distribution of bird and amphibian species and their phenological patterns.
Acoustic signaling of vocal organisms can enable us to further investigate the effects of

habitat transformation and climate variability on biodiversity.

Kroodsma et al. (1996) established a theoretical framework and clear
demonstrations describing the diversity of vocal development in avian species and the
extent of spatial variation of bird songs within and among populations. Other studies
provided clear evidence that vocalizations in birds and frogs were highly influenced by

and adapted to their habitat structures (Wiley and Richards 1978, Ryan et al. 1990).

Although acoustic signals have provided a great deal of biological information in many
ecological studies, acoustic research has focused primarily on the significance of species-
specific sounds in nature with a few attempts to interpret the implications and

mechanisms of songs and calls at population or community levels (Warren et al. 2006).

It has been proposed that acoustic signals could contain some dimensions of
ecological information about the environment (Truax 1984b) and therefore could be used
to measure the dynamics and patterns of ecosystems (Gage et al. 2001 , Napoletano 2004).
Little work, however, has been done to investigate whether acoustic signals can serve as
an ecological attribute for describing the characteristics of ecosystems and the changes in
animal biodiversity, and to demonstrate whether soundscapes can reflect the current state

of various ecosystems.

Concept of Soundscape

A dictionary defines the term soundscape as ‘an acoustic environment or an
environment created by or with sound’ (The American Heritage Dictionary 2000).
According to this definition, the soundscape is related to many areas such as music and
acoustics, social science, and physics. However, from an ecological perspective, the term
soundscape refers to the collection of sounds emanated by acoustic forces in a given
place including abiotic, anthropogenic, and biological acoustics including human voices

(Schafer 1994, Napoletano 2004).

Schafer (1977) first introduced the idea of the soundscape, defined as the
biophysical environment or place where sounds are emitted, and stated that measuring

and understanding the structure and function of soundscapes could provide a key means

in environmental assessment. The dominant soundscapes in pristine ecosystems include
the sounds of flows of abiotic components (e. g., wind, water, thunder, rain, etc.), and
acoustic activities of vocal organisms (e.g., birds, frogs, insects, and mammals). In
contrast, highly urbanized areas are dominated by the industrial and mechanical acoustic
forces produced by anthropogenic activity. The soundscape in agricultural areas will
result in an intermediate intensity of sounds both from anthropogenic and biological
activity. The structures and components of the soundscape have thus far been
qualitatively described, but the quantitative measure and analysis of the soundscape is
necessary to understand ecological interactions between biological and human activities

within heterogeneous landscapes.

According to Schafer (1994), the soundscape consists of three different elements:
‘keynote sounds, ’ ‘sound signals, ' and ‘soundmarks '. Keynote sounds are originally
borrowed from a musical term that identifies the key of a piece. The keynote sounds may
not always be heard consciously, they are mostly produced by physical and biological
events such as movement of air (wind), flow of water, signaling of birds and insects.
Traffic and other mechanic sounds are also the keynote sounds in urban areas. On the
other hand, Sound signals are the acoustic signatures that can be heard consciously (such
as bells, whistles, horns, sirens, etc.). Lastly, soundmark originated from the term

landmark, and refers to a unique acoustic signature produced from a particular area.

The soundscape can also be classified into three primary components by an
acoustic frequency spectrum: biophony, anthrophony, and geophony (Gage et a1. 2001,
Napoletano 2004) (see Figure 1.1). The term “biophony”, originally coined by Krause

(1998), describes the complex chorus of ambient biological sounds. This category

encompasses the natural sounds produced mostly by vocalizing birds, amphibians, insects
and mammals. The acoustic frequency spectrum of most biophony generally ranges from
2.5 kHz to 8 kHz. The biophony also include two characteristic signals: 1) intentional
signaling, which is transmitted to exchange information about mating, territory defense,
etc., and 2) incidental, referring to the signals that propagate but do not include the
explicit purpose of communication. In addition, because humans have had a significant
impact on natural environment, the term, Anthrophony, was devised (Gage et al. 2001).
Anthrophony refers to any acoustic signals created by human activities such as musical
performance and oral conversation, or mechanical sounds resulted from operations of
machinery and automobiles. The frequency spectrum of anthrophony ranges from
approximately 40 Hz to 2 kHz. The last category, geophony, refers to the set of sound
generated by physical processes including wind, rain, river flow, and so forth. Geophony
generally includes the frequency spectrum of interest from approximately 40 Hz to 11

kHz, depending on intensity of the physical sound.

Soundscape

  

 

Biophony Geophony Anthrophony
(2 - 8 kHz) (0 - | | kHz) (0 - 2 kHz)

    
 
  

Intentional Musical Mechanical

   

Incidental

Figure 1.1 Diagram of the classification of a soundscape classification (modified from Gage et al.

(2001)

These acoustic components have less distinct boundaries than landscape elements
and often overlapping ranges of sound frequency. The spatial scale of the soundscape can
vary, depending on the number and the detection capacity of acoustic sensors and
recording devices. The frequency spectrum of the soundscape considered in this study
ranges from 20 to 11,000 kHz, because this spectrum range is audible to the human ear;

and many vocal organisms utilize this spectrum range for their communication.

The functions of the soundscape are to provide organisms with a sonic
environment to communicate intra- and interspecific interactions, and to indicate the
current states of acoustic sources. It is unlikely that changes in soundscape structures
have a significant influence on transformation or alteration of landscape configurations.
Rather, the properties of soundscapes will respond to alteration of spatial patterns in
landscapes and the associated movements of landscape objects. The soundscape thus can
function to provide critical information about landscape change. However, few studies
have attempted to identify the potential functions of soundscapes and thus they still

remain little known (Napoletano 2004).

Soundscape changes may occur as the consequences of movements of acoustic
resources as well as any disturbance to an acoustic environment. Soundscape dynamics
can substantially vary among different temporal scales from day to season to year.
However, the repeated dynamic patterns of the soundscape can be observed over a long
time period. Overall, anthropogenic disturbances will commonly amplify changes both

in landscapes and soundscapes.

Interrelationships between the landscape and the soundscape

The relationships between soundscapes and landscapes have been little studied
and rarely documented. In order to investigate the relationships between the landscape

and the soundscape, several conceptual hypotheses are posed.

The first hypothesis states that spatial patterns and structures of a landscape do
not have a direct relation to the characteristics of the soundscape (null hypothesis).
Rather, potential processes of acoustic sources are more likely to have a critical influence
on changes in soundscape structures. However, this hypothesis is unlikely to be accepted
because acoustic sources are likely to be closely related to the spatial configuration of the

landscape.

The second hypothesis states that greater heterogeneity of a landscape will
produce a more diverse soundscape. Acoustic diversity refers to the patterns of
frequency and temporal variability of the acoustic spectrum, indicating the degree that
different vocalizing organisms will utilize different frequency niches to propagate
information within a soundscape. It was shown that birds and amphibians make selective
use of an acoustic frequency when attempting to communicate information such as
mating potential, territory size, and potential predation (Catchpole and Slater 1995,
Narins 1995, Kroodsma and Miller 1996). This hypothesis assumes that more diverse
features of a landscape will result in diversity of acoustic environments where unique
soundscapes are established. Furthermore, when the communication of vocalizing

species is disrupted by human activities, the organism contributing to the soundscape can

be adversely influenced. In contrast, habitats with less human activities tend to exhibit

the more complex biological sounds in terms of acoustic frequency and periodicity.

In connection with the concept of the acoustic diversity, Krause (1997) proposed
the “Acoustic Niche Hypothesis” in which each vocal species can develop a dynamic
niche by adjusting the temporal and frequency properties of its respective signals to
unfilled portions of the soundscape in order to avoid competition for spectral or temporal

I'CSOUI‘CCS.

Although variability and diversity of soundscapes occur, dependent on spatial
configuration of landscapes and the extent of disturbances, these suggested hypotheses
need to be validated so that robust relationships between soundscape and landscape

attributes can be established.

Given that environmental acoustics has played a key role, not only in inferring the
ecology and behaviors of vocal organisms, but in understanding the dynamics of
ecosystem structure and processes, it was proposed that environmental acoustics can be
used as an ecological indicator of the state of the ecosystems (Napoletano 2004). To meet
the requirements of ecological indicators, the acoustic variable has several advantages.
First, sounds can reflect the degree of stress and perturbation in an ecosystem, and can be
routinely and easily monitored; Second, current technology enables us to deploy acoustic
monitoring systems with relative ease and minimum cost, leading to less interruption of
the sites by human activities. Lastly, measuring environmental sounds provides
integrated ecological information by interpreting their complex structures and identifying

sound sources.

The overall objectives of this research are to investigate whether environmental
acoustics can be used as an ecological attribute to indicate the current ecosystem status
and to determine if sound can be used to measure and monitor the biodiversity and
distribution of vocal species. To understand the acoustic characteristics in an ecosystem,
numerous acoustical data were collected at many locations in different ecosystems over

varying lengths of time. The specific goals of the study are:

Develop an analytical framework to characterize the structures of environmental

acoustics.

Understand spatial distribution and temporal variability of environmental

acoustics.

° Compare accuracy and efficiency of methods used to conduct acoustic surveys.

Assess the state of varying ecosystems and landscapes using metrics of

biological acoustic activities and anthropogenic acoustical disturbance.

° Establish and develop long-term ecological monitoring and observation systems

based on environmental acoustics.

To accomplish the specific goals, the study is organized into five chapters.
Chapter One introduces the concept of the soundscape, and provides a overview of
acoustic ecology. Chapter Two describes the framework to enable understanding of the
structures and components of environmental acoustic signals proposed by Gage et al.
(2001). This chapter also includes the conceptual process of quantification and extraction
of ecological information from acoustic signals. Chapter Three investigates whether

acoustic recordings have the potential to survey abundance and distribution of breeding

birds. In this chapter breeding birds were surveyed using automated and manual acoustic
recordings to compare the accuracy of this method with the traditional point count survey
method using the human ear as the sensor. In Chapter Four, I investigate whether a
quantitative acoustic index could be used to describe the characteristics of different
landscapes along an urban-rural gradient. This part of the research investigates and
describes the relationships between anthropogenic noise and biological acoustics. In
addition, I investigate how the acoustic index can be associated with social and economic
variables. Chapter Five examines the concept of environmental cyberinfrastructure and
the application of an automated sensor system in terrestrial ecosystems. Here I describe
the design, development and application of an automated wireless sensor system to
monitor and transmit environmental acoustics to a remote location for subsequent
analysis and interpretation. Chapter Five provides a summary and conclusions about the
findings of this research. In addition, I describe the accomplishments and challenges

associated with this study as well as suggest future directions and potential research.

Chapter Two
Development of Analytical Framework to Understand the
Structure and Components of Environmental Acoustic
Signals

Sound has played a critical role in understanding the behavior of vocal aquatic
and terrestrial animals, because vocal communication is a fundamental means of
interaction within and between species. Sound has been used in ecology to census
organisms (i.e. birds, amphibians, and mammals), and sound signatures of complex
ecological communications has been identified and interpreted(Kroodsma and Miller
1996). This chapter will address the background of acoustics, and describe the data
process and the analytical framework to understand the structure and components of
environmental acoustic signals. Lastly, case studies are described to provide evidence of

how acoustic signals can be applied to describe changes in different ecosystems.

Background of acoustics

Sound refers to vibration or compression waves transmitted through a solid, liquid,
or air, and encompasses three main types determined by the acoustic frequency spectrum:
1) infrasound, 2) ultrasound, and 3) audio/sound (Rossing 2007). Infrasound is located in
the frequency below 20 Hz, and is not audible to the human ear. In general, infrasound is
generated from the natural events such as avalanches, earthquakes, and volcanoes
(Tempest 1976). Ultrasound is sound with the frequency spectrum above 20 kHz, and is
also beyond the upper limit of human auditory sensitivity (Larkin et al. 1996, Hopp et al.
1998). Ultrasound has been used in many fields, especially in medical science using

ultrasonography. Ultrasound is also produced by some animals for communication,

10

navigation, and predation, including marine mammals, bats, some rodents, and even a
few amphibians (F eng et al. 2006). Lastly, audio refers to sound that can be heard by
human cars, including frequencies from 20 Hz to 20 kHz. Vocalizations of many song
birds, terrestrial mammals, amphibians, and insects fall into the audio-frequency

spectrum similar to those of humans (Dooling 1982, Pay and Wilber 1989).

Sounds have physical properties of waves including frequency, wavelength,
period, amplitude, intensity, speed, and direction (Rossing 2007). ‘High pitched' or
‘lower pitched' thus refer to sounds with high frequency (the number of cycle per a given
time) or low frequency. Wavelength or period represents the distance or time of one
cycle unit of sound wave. Because wavelength has an inverse relationship with
frequency, longer wavelengths have lower frequencies and vice versa. Using these
acoustic properties, sounds can be sampled, measured, and quantified. A common

process in acoustic measurement is to record sound samples using a microphone and to

analyze the acoustic signals based on frequencies and amplitude in a given time.

To analyze and interpret the characteristics of acoustic signals, there are two
techniques in measuring sounds: I) sound pressure level (SPL), and 2) sound power level
(SWL). Both methods are computed with the same 81 unit, “dB” but use different
acoustic properties. Firstly, SPL is measured using the sound pressure (defined as the
amplitude of an actual pressure of the sound wave in atmosphere, Pa). SPL is expressed

as follows:

11

P
LP (SPL) = 2010g10 ;— dB
ref

where p is the average sound pressure in a given period and pref is a reference

sound pressure, commonly 20 uPa (threshold of human hearing) in air. This

measurement represents the degree of sound loudness or amplitude, also called “sound
volume'. Because this equation is a logarithmic measure of the ratio between the root-

mean-square sound pressure and the reference sound pressure, increasing 10 dB of a
given sound means a 10 fold increase in the sound. In general, the SPL of traffic noise
produced 10 m away from a highway is in the range of 80 to 90 dB and the SPL of jet
engine sound 100 m distant is in the range of l 10 to 140 dB. Sound pressure has been
typically used to investigate the effects of noise on humans and wildlife because they are
responsive to sound waves as pressure (Larkin et al. 1996). The perception of sound
pressure in animals varies by different frequencies. Thus, it is important to estimate a
sound pressure level (dB) with filtering or weighting of sound amplitude at different

frequencies. The well-known method is called ‘A-weighting (dBA)’, approximating the
threshold of human ear’s response to sounds at different frequencies. Another common
method of estimating SPL is ‘C-weighting', cutting off the entire sound amplitude below

low frequencies (about 50 Hz).

Sound power level (SWL) refers to a measure of sound power or sonic energy

(watt), and is computed by the following equation:

12

LW(SWL) -_-101og10 3—1- dB
pref

where p1 is the sound power, and pref is the reference sound power,

commonly 10-12 watt in air. Sound power is not equivalent to the sound pressure but is
proportional to the square of the sound pressure (Pw 0: p2). Because the sound power is
independent from the distance of the sound source, SWL can be employed to measure the
sound power without knowing the distance from the source. Because both methods were
developed to measure the noise of mechanical sound sources generated by human
activities, they are not likely to be applied to measure the characteristics of entire

soundscape.

Analytical tools to quantify and analyze environmental sounds

Although there has been pioneering research on the potential of soundscape to
understand characteristics of environmental acoustics based on descriptive and qualitative
analysis of sounds (Schafer 1977, Krause 1998), little work has been done to quantify and
analyse the properties of environmental sounds and to extract ecological variables to
indicate the various states of habitats and ecosystems. Analytical tools to compute and
interpret acoustic data sets have been developed (Gage et a1. 2001, Napoletano 2004, Qi
et al. 2008), and consist of two main categories: Generalized Sound Classification

Analysis (GSCA) and Acoustic Identification Analysis (AIA).

13

Generalized Sound Classification An_alvsis

This acoustical analysis was developed based on the hypothesis that every
soundscape would have its unique acoustic signature, and the soundscape would have a
common pattern of acoustic signatures in ecosystems with the same physical and
biological components. Thus, this method characterizes the physical structure of
environmental acoustic signatures from the data set, based on sound intensity with a

particular combination of acoustic spectrum ranges.
Acoustic signal processing

There are two techniques used to process and quantify the values of sound
intensity in a soundscape: the relative intensity value from a spectrogram and power

spectrum density from digital sound samples.

Quantification of Spectrograms: Sound samples collected from a study site can
be processed to extract acoustical information. To analyze and interpret the
characteristics of acoustic signatures, a new analytical means was developed by
quantifying the acoustic properties of the spectrogram (Gage et al. 2001, Napoletano
2004). Spectrograms are generated by digital signal processing based on the short-time
Fourier transform (STFT)(Truax 1984a). In a spectrogram, time and spectrum frequency
are plotted on the horizontal and vertical axis, respectively. The amplitude of the sound
is represented by the intensity of each point in an image. This new acoustic processing
technique enabled quantification of the spectrogram (Figure 2. 1). To accomplish this,
the spectrogram image is divided into 1 kHz frequencies and then the intensity of sounds

in each of three frequency ranges is computed. In this process, Spectrograms are

14

converted to 8 bit images with possible intensity values in the range from 0 to 255. For
example, if a spectrogram visualizes acoustic signatures of the entire acoustic frequency
filled with the maximum sound amplitude, the average intensity value is 255. As
quantification and analysis of Spectrograms is based on image processing techniques
(Gage et al. 2001, Napoletano 2004), the method requires signal and image processing
computation tools, a complex procedure, and a long duration for completing the entire

process.

I
. 1.0
I Mean Acoustic spectrum power (w/Hz)

.I/IIIII..._

Frequency6 bins (kHz)

l
l

 

Figure 2. 1. The acoustic frequency slicing procedure. a) Each sound wave file is divided into 1 1
frequency bands, and b) the relative mean intensity is calculated for each band. Note that 5 kHz
band has the highest mean intensity among 11 frequency bands.

Power spectrum density: In contrast to the spectrogram analysis, an approach was
developed to directly process and quantify acoustic signals from sound data stored into a
digital format (Gage et al. in submission). The sound samples were divided into 1 1
frequency bands (0-1 kHz, 1-2 kHz, ...... , 10-1 1 kHz), and the power spectrum density
(PSD) for each frequency band was calculated using MATLAB’s signal processing tool

based on Welch’s algorithm (Welch 1967). PSD is computed with the unit, ‘watt/Hz’,

and the values of PSD at each frequency band vary by the amount of the sound power in
the sound sample. This method requires less computation time and eliminates the need to
produce a spectrogram and subsequently apply image analysis techniques to quantify the

mean number of pixels in each frequency interval.

Acoustic Habitat Quality Index

An Acoustic Habitat Quality Index (AHQI) was developed to interpret acoustic
data from soundscape characterized by biological sounds (biophony) and mechanical
sounds (anthrophony), based on the amount of acoustic energy in different frequency
intervals. By analyzing a large number of environmental sound recordings, Napoletano
(2004) found that the frequency spectrum of environmental sounds is concentrated into
two main soundscape categories: Anthrophony and Biophony. Anthrophony consists of
mechanical sounds mostly occurring at a frequency range from 1 to 2 kHz, whereas
biophony (biological sounds) are prevalent between 3 to 8 kHz (Figure 2.2). Based on
the method used to compute the Normalized Differential Vegetation Index (Myneni et al.
1995), normalized ratios of frequency levels were then employed to estimate the relative

amounts of biophony or anthrophony in an acoustic sample. The equation for AHQI is:

+

m

Q

 

AHQI =

Th

9

where a, [3 represent the total amount of acoustic energy in biophony and
anthrophony respectively. The value of AHQI ranges from -1 to 1. If the value of AHQI
is l, the sound sample is all biophony, while anthrophony is totally dominant in a sound

sample when the value is -1.

16

O)

Biophony (2-8 kHz)

N Frequency (kHz)
Geophony (0—11 kHz)

. Anthrophony (0.5 -2 kHz)
0 —
TIme (30 sec.)

 

Figure 2.2. Three main classes of environmental sounds: I) biophony, 2) anthrophony, 3)
geophony, based on the location of the spectrum frequency bands.

This index provides an indication of the relative amount of biological and
anthropogenic sounds occurring at a place, thus indicating the degree of human
disturbance in an ecosystem. Patterns of the occurrence of these acoustic elements can be
analyzed and trends assessed to develop estimates of acoustic signal types. The

computation of these indices requires moderate processing capacity.

This method was employed to characterize and compare the relative contributions
of biophony and anthrophony in soundscapes, and the results and applications of this
index are addressed using specific examples provided in Chapter Three, ‘Analysis and
interpretation of the “Heartbeat of the City” using acoustic signatures along an urban-

rural gradient’.

Principal component analysis (PCA)

To establish a statistical protocol to analyze acoustic signatures and characterize
the combination of frequency bands for biological acoustics, a feasible statistical method
to describe patterns of covariation in each acoustic signature is to calculate the
eigenvalues and eigenvectors of the temporal covariance matrix among the different
frequency bands using Principal Component Analysis (PCA). PCA has been widely used
in many areas of ecology and biosystematics (James and McCulloch, 1990), and reduces
the dimensions of observed variables by producing a smaller number of abstract variables.
PCA is based on maximization of the variance of linear combinations of variables
(Legendre and Legendre 1998). This statistical method is able to summarize patterns in
the correlation matrix among the different frequency bands across the temporal extent at
which the acoustic data was recorded. In the correlation matrix from the acoustic data set,
eigenvectors are weightings of individual frequency bands contributing to sources of
variability across the entire data set, and eigenvalues represent the variances of these
sources. Thus, it is anticipated that each principal component will correspond to a set of

contiguous frequency bands which characterize different classes of environmental sounds.

Our research shows that PCA can be used to separate the intensity of biological
sounds from other types of acoustic signals (See Figure 2. 3). A large acoustic data set
was collected from the Long Term Ecological Research (LTER) site at the W. K. Kellogg
Biological Station, Hickory Corners, MI. from May 18 to July 15, 2005. Automated
acoustic recording systems were deployed in 13 sites classified into two major land use
types: agricultural (4 sites) and forested (9 sites). Note that when PCA was performed on

the acoustic intensity data set, the first three principal components were determined as the

18

key components because those components explained nearly 80 % of the variance in the
acoustic signature data. The first Principal Component (PC) has negative loadings for all
frequencies, implying that it indicates the amount of sound power across the entire
frequency spectrum. The second PC has negative loadings at low to mid frequency bands
(anthrophony) and is dominated with positive loadings in the sounds with high frequency
bands (biophony). These patterns suggest that the distribution of high frequency sounds
is relatively independent of low frequency sounds. In contrast, the third PC has high
positive loadings for low frequencies (1 to 3 kHz), and moderately negative loadings for
the mid to high fi'equency spectrum. The results show that when low frequency sounds
are inversely correlated with high frequency sounds, there is a negative relationship

between anthrophony and biophony.

The PCA method was further used to compare sounds recorded both within sites
and between them, providing the capacity to separate the temporal and spatial dimensions
of biological and mechanical sounds. The third Principal Component was selected
because this PC includes the abstract information indicating which acoustic frequency
band was highly associated with the intensity of biophony and anthrophony. Thus, PC3
scores on the Y-axis represent integrating acoustic variables from frequency 1 to 11 kHz.
The positive PC3 scores contain biological signals (3 to 6 kHz) while the negative scores
contain anthropogenic sounds (I to 2 kHz). Note that the PC3 scores varied at different
times and locations (Figure 2. 4), indicating the temporal and spatial variability of
biophony and anthrophony. Note that high positive scores occurred in the morning and
low negative PC3 scores occurred in the aftemoon, providing evidence of the diurnal

pattern of environmental sounds in the study site.

19

0.8

0.6

0.4

0.2

PC loadings
O

 

Frequency bands (kHz)

Figure 2. 3. Principal Component loadings of different frequency bands on the first three principal
components from the acoustic data collected at the Long Term Ecological Research site in W. K.
Kellogg Biological Station. The labels on X axis represent the acoustic frequency bands (e.g., L2
refers to the frequency band ranges from 1 to 2 kHz).

 

0.5 ‘
0.2 I
.o.. r I” I- ,

-0.4

PC scores

 

 

 

 

0 630 830 143018302030 A1 C1 C3 D2 Pl SI 53
Time Location

Figure 2. 4. Using principal component analysis, the temporal patterns of ecological
sound intensity are shown on the lefl: and the variability of biological or anthropogenic acoustic
intensity at each site is shown on the right. The first letter of the location labels refers to one of
five different habitats (A=Alfalfa fields, C=Coniferous forests, D=Deciduous forests, P=Poplar

stands, S=Succession). The second letter of the location labels refers to one of three replicates.

Means (+/- s.e).

20

Acoustic identification analysis

One advantage of quantifying acoustic samples to identify and census vocal
organisms from recordings is that it provides a measure of the dynamics and patterns of
vocal organisms in an ecosystem (Gage et al. 2001) as well as providing a permanent
record of the observation. An analysis of the acoustic data set recorded at each site
includes computing the PSD for each frequency level, visualizing the acoustic signatures
in the Spectrograms, listening to the songs and calls of vocal organisms, and then

identifying species in the sound recordings.

The number of species present and the frequency of songs and calls made by each
species were compiled resulting in an estimate of the relative species richness and
abundance of vocalizations for each site. In addition to community measures including
the number of species and vocalizations, bird species diversity was calculated using a
modified Shannon-Weaver Index (Shannon and Weaver 1963, Magurran 1988, Blair

1996), thus providing an indication of biological acoustic diversity.

To illustrate the power of quantifying acoustic recordings of bird species, their
acoustic signatures were identified from the sound recordings made at the W. K. Kellogg
Biological Station, Hickory Comers, MI from May 18 to July 15, 2005. The number of

species and their vocalizations varied in the different habitat types (

Figure 2. 5). The results show that more complex habitat types (e. g., early
succession) had a higher avian species richness value and more frequent song
occurrences. We were then able to characterize each habitat sampled in terms of the

dominant bird species (Apendix 2.1).

21

 

'59 No. of Bird Calls

  

25tNo. of Bird Species

  

20
IO
15
50 IO1
5
0 0
AI Cl DI Pl SI Al C1 D1 Pl SI

Figure 2. 5. Avian species were identified by listening to the recordings. The number of bird
species identified is shown on the lefl and the number of calls identified is shown on the right.
Furthermore, the relationship between species richness and the
abundance/intensity of biological sounds was examined. The results indicated that the
number of avian species determined from the sound samples was positively related to the
number of bird vocalizations in these recordings (Figure 5.a); Acoustic biodiversity was

also positively related to the PC3 scores which indicate the intensity of biological sounds

(Figure 5.b)

 

 

 

 

 

 

 

. y=0.|4x+6.00 ' . y=0-|4x+|-06
5—— R2=0.79 - —— R2—0.28
20 60 I00 I40 ‘ 0.5 1.0 1.5
No. of calls ln(diversity index)

Figure 2. 6. a) The relationship between avian species richness and the number of calls, and b)
the relationship between acoustic species diversity and PC3 scores (indicating the biological

acoustic energy).

22

Chapter Three
Use of acoustic recordings for surveying avian species
richness and distribution

Introduction

Surveying birds has provided omithologists and wildlife biologists with a great
opportunity to quantify species richness, abundance and spatial distribution of species, to
illustrate the avian-habitat relationship, and to monitor changes in avian populations and
communities (Rosenstock et al. 2002, Thompson 2002, Williams et al. 2002, Bart 2005).
For example, the North America Breeding Bird Survey (BBS), one of the largest regional
biological data sets across North America, has been conducted more than 30 years to
monitor populations and communities of breeding birds at regional and national levels
(Robbins et al. 1986). The BBS was established in 1966, in response to the concern over
pesticides resulting in less reproductive success and increasing death of animals (Carson
1962, Patuxent Wildlife Research Center 2001), and so far has contributed data for more
than 270 scientific publications up until 2002 (Peterjohn and Pardieck 2002). Moreover,
many studies in support of the BBS data revealed a significant decline in populations of
North American breeding birds, particularly species that inhabit forest-interior and
grassland areas (Robbins 1979, Whitcomb et al. 1981, Herkert 1995). The declining
trends of some avifauna populations might be caused by lack of breeding, wintering, and
migration stopover habitats due to the extensive reduction and fragmentation (isolation)

of their habitats (Robinson and Wilcove 1994, Herkert 1995, Robinson et al. 1995).

23

Although the BBS has played a key role in monitoring and understanding trends
of breeding bird populations and communities across North America, some concerns
about the quality, validity and variability of the BBS data have been raised. The BBS has
used the point-count method; one of the most commonly used survey methods in
identifying and counting birds (Ralph et al. 1995, Rosenstock et al. 2002, Thompson
2002). The point-count method consists of an observer listening to, seeing and
identifying all birds within a fixed distance at a certain points in an area block or along
line transect during the breeding season (Hutto et al. 1986, Ralph et al. 1995). Several
studies have addressed concerns about using the point count method to survey birds,
because of the inconsistency of detection probability among species and habitats or
across time (Thompson 2001, Rosenstock et al. 2002), Others have raised issues due to
different levels of observer ability (i.e., experience and skills of survey, age, and hearing
loss) (O'Connor et al. 2000). It has been suggested that the method should be improved
(O'Connor et al. 2000) or supported by other methods (Parker III 1991, Haselrnayer and
Quinn 2000, Zimmerling and Ankney 2000, Hobson et al. 2002, Rosenstock et al. 2002,
Thompson 2002, Conway and Gibbs 2005, Acevedo and Villanueva-Rivera 2006).
Given that animal vocalizations serve as a fundamental means for many vertebrates and
invertebrates to communicate with each other, using acoustic recording technology for
breeding bird surveys has been suggested to increase detection and reduce variability
(Parker III 1991, Hobson et al. 2002, Conway and Gibbs 2005, Acevedo and Villanueva—
Rivera 2006). In addition, when using only the point count method, there is no validation

of species identified by an observer. However little work has been done to investigate the

24

potential utility and applications of automated sound recordings and comparing species

counts using these methods to those obtained by the point count method.

There are several advantages in the use of automated recording of biological
acoustic signals in support of the BBS point-count method including: 1) reduced
variability of detection probability within and among observers; 2) calibration and cross-
validation of data; 3) higher detection rate for identifying secretive birds in grasslands
and marsh areas (Parker III 1991, Allen et al. 2004, Conway and Gibbs 2005), 4)
discovering new taxa in remote regions (Haselmayer and Quinn 2000), and 5)
establishment of an archive and inventory of avian species thus building a birdsong

digital library (Parker III 1991).

Several studies have attempted to take advantage of acoustic signals for
measuring the abundance and distribution of bird species. Hobson et al. (2002) used
arrays of omni-directional microphones to survey breeding bird species in mature
deciduous stands areas in Canada. Allen et al. (2004) and Conway and Gibbs (2005)
compared detection rates of breeding marsh birds in passive and sound-playback surveys
in South Dakota. Recently, the effectiveness of using automated digital recording
systems was investigated for avian and amphibian surveys in the north coast of Puerto
Rico (Acevedo and Villanueva-Rivera 2006). However, there are limitations regarding
the accuracy of acoustic recordings for detection of breeding birds across various habitats

in temperate regions.

To address some of the issues, we investigated the detection probability of

breeding birds using acoustic recording technology, in addition to following the protocol

25

for the conventional point-count method to survey breeding birds. While recording the
soundscape on digital media during the time in which the observer conducted the point
count method, we set up automated recording units at each stopping position along the

transect to monitor the temporal variability of surveying birds.

Methods

Study area

This investigation was conducted in two locations in Clinton County, Michigan.
The study sites were selected based on the Michigan Breeding Bird Atlas mapping
system which consists of township, range, and section to identify location (Kalamazoo
Natural Center 2002). Two blocks were identified for the survey sites: one in Essex
Township (T08N R03W, Block 2) and the other in Westphalia Township (T06N R04W,
Block 4) (Kalamazoo Nature Center 2002). According to the Michigan Breeding Bird
Atlas protocol, 25 points per block were selected fi'om the point map which the MiBBA
provides to surveyors. The sites were classified into five habitats: wet and dry forests,
row crop fields, grasslands, and shrubs, representing typical habitat types in this

temperate region.

Survey methods

The Essex Township block and the Westphalia Township block were surveyed on
June 12 and on June 22, 2007, respectively (Figure 3. 1). Each study consisted of three
different breeding bird survey methods: 1) Point Count Surveys (PCS), 2) Manual

Acoustic Surveys (MAS), and 3) Automated Acoustic Survey (AAS). PCS was

26

performed according to the survey protocols of the Michigan Breeding Bird Atlas 11
program (similar to the Breeding Bird Survey-style point counts) (Kalamazoo Nature
Center). While conducting PCS for 5 min at each station, a field observer used a standard
point count data fonrr to record the following observations: geographic and
meteorological observations, the name and the abundance of all avian species, and the
time and distance of when and where each bird was first detected. The observer also
noted whether the bird observed was within or beyond 50 m. This survey also recorded
whether the bird identified was heard or seen (Haselmayer and Quinn 2000). The survey

was conducted for 5 minutes at each of the 25 locations along the transect.

As PCS was being conducted at each location, a digital recording was made
simultaneously for 5 minutes. In Essex Township a Sony® Mini Disk with a Sony
MS957 microphone was used and in Westphalia Township, we used a Tascam digital
recorder (HD-P2) with an Omni-directional microphone (ATRSS, Audio—Technica Inc.).
Recordings were captured in Waveform Audio (.wav) with a 22 kHz frequency sampling
rate and a monaural channel. In general, a sampling rate is determined to double the
highest frequency spectrum (Hartmann 1998), because most avian songs and calls are

generated in frequency spectrum ranging from 2.5 to 8 kHz.

To capture the temporal variability of the avian survey we placed an automated
recording unit (Sangean VersaCorder®, C. Crane Co with an Omni-directional boundary
microphone, RadioShaek Co. Model 330-3020) at each point count location. Each of the
25 recorders was programmed to automatically record ambient sounds for 5 minutes each
hour from 0500 to 1000 hrs. The recordings were digitized from audio tapes employing

the same protocol used for ”manual acoustic recording (22 kHz with monaural channel in

27

waveform format). Once sounds from acoustic surveys were all digitized and archived in
a digital acoustic library, bird species were identified by listening to bird songs and calls

in the recordings.

 

Figure 3. 1. Field bird survyes using a point count method and acoustic recordings in Clinton
County, Michigan during June 2007: a) field configuration of the Manual Acoustic Recording
(MAR) method using an Omni-directional microphone and a digital acoustic recorder; b)
simultaneous bird surveys by human observation and MAR; c) a digital acoustic recorder
(Tascam HD-P2) deployed in a point; and d) field deployment of an acoustic recording unit
including an audio cassette-tape recorder with a timer (SanGeanVersaCorder) and a omni-

directional microphone (330-3020) for Automated Acoustic Recording (AAR).

Data Analysis

The first objective of the study was to test the similarity of the three survey

methods based on the number of bird species identified by point counts (human ear) and

28

by digital audio recording including manual recordings and the automated recordings at
each location. Prior to comparing the observations using the three different methods, the
bird counts observed as “flying over” by point count method were excluded from the
analysis because it was difficult to compare this type of observation with the other

methods.

The primary objectives were to test the similarity of bird species detection using
the three survey methods and to examine the total number of bird species identified by
each method. Community similarity measures including the Jaccard’s coefficient and the
Sorensen’s quotient of similarity were used to compare composition and number of
species among three surveys is in the study sites (Hobson et al. 2002). The similarity
indices were initially developed for comparing the communities with number of species
(Krebs 1998), but the indices were modified to compare the species data among the

survey methods instead of communities.

We investigated the results of the three survey types in terms of avian species
richness. The number of bird species identified by each survey method was estimated at
each location. As the species richness data set was normally distributed, a one-way
analysis of variance (ANOVA) was performed. The Tukey’s pairwise comparison
method was conducted to examine mean difference of species richness among the bird

detection methods.

The acoustic recordings cannot be used to identify the number of individuals of
each species observed at each point unless microphone arrays were established at each
location and this was beyond the scope of the study. Instead, it was assumed that one

individual of a species identified at each location by all methods was considered as an

29

encounter regardless of how many individuals of the species were seen or heard at each
point. A paired t-test was performed to test whether there was a significant difference of
avifauna species occurrence among the three survey methods used to determine avian

species presence.

The number of bird species identified in the automated acoustic recordings was
determined for each hour (e.g., 5am, 6am... 10 am). We thus investigated the temporal

pattern of avian species richness during the survey time.

Results

The number of avian species detected in the acoustic recordings was almost the
same as those identified using the point count survey. All survey methods combined
identified 64 and 60 bird species in Maple Rapids and Westphalia respectively (Table 3.
l). The highest number of bird species was identified using the point count method in
both Essex Township and Westphalia Township. Automated acoustic recordings failed
to identify 7 and 6 species in Essex Township and Westphalia Township respectively.
There were 14 and 9 species identified by human observation, but not by manual acoustic

recordings in Essex Township and Westphalia Township, respectively.

Avian communityI similgities among the survev types

Bird community estimates between point count surveys and the two recording
surveys had high similarity indices ranging from 0.71 to 0.89 (Table 3. 3). Similarity
measures between point count surveys (PCS) and automated acoustic surveys (AAS) had

consistent similarity indices in both sites (0.75 based on Jaccard’s coefficient and 0.85

30

based on Sorenson coefficient in Essex and Westphalia sites), whereas similarity
measures between PCS and manual acoustic recordings (MAS) differed in two study sites
(0.70 and 0.80 of Jaccards’ coefficient and 0.83 and 0.89 of Sorenson’s coefficient in

Essex and Westphalia sites respectively).

Table 3. 1. The number of bird species identified by point count survey (PCS), manual acoustic

survey (MAS), and automated acoustic survey (AAS) in Essex and Westphalia Township.

 

 

Number of . . Number of species
. . . Number of specres Number of specres . o
Srte specres rn all in PCS in MAS (ty ) in AAS (/o) (500 —
surveys o 1000 hrs)
Essex 64 59 45 52
Westphalia 60 54 45 48

 

Table 3. 2. Bird detection accuracy of all three survey types, based on the total number of species

identified in Essex and Westphalia Township.

 

. . . Detection accurac
Detection accuracy Detection accuracy In y

 

Site . in AAS (%)(500 —
In PCS (%) MAS (%) 1000 hrs)

Essex 92.19 70.31 81.25

Westphalia 90.00 75.00 80.00

 

31

Table 3. 3. Community similarity measures (Jaccard’s coefficient / Sorensen coefficient) to
investigate similarity of avifauna community between Point Count Survey (PCS) and counts from
two acoustic recordings: Manual Acoustic Surveys (MAS) & Automated Acoustic Surveys
(AAS).

 

Jaccard's coefficient/ Sorensen coefficient

 

 

 

Essex Westphalia
AAS MAS AAS MAS
PCS 075/085 070/083 075/085 0.80/0.89
AAS 0.75/0.85 0.71/0.83

 

Relationship between the survey types and species richness

The bird survey methods had a significant effect on bird species richness in Essex
Township (F=9.10; df = 2; p < 0.01) and Westphalia Township (F=3.66; df = 2; p < 0.05).
The average number of species per location between Point Count Survey (PCS) and
Automated Acoustic Recording (AAS) was not significantly different (p > 0.05), whereas
there were significantly fewer species identified in the Manual Acoustic Recordings
(MAS) than in the PCS and the AAS in Essex Township (p < 0.05) and Westphalia

Township (p < 0.05) (Figure 3. 2 1).

Species abundance estimgte analysis

The number of encounters, defined as the number of locations where each species
was identified, was used to determine the relative abundance of bird species. For instance,
if the number of encounters for the American Robin is 23, this means that this species

was identified in 23 out of 25 locations in the Township. In Essex Township, the point

32

count survey methods (PCS) identified Indigo Bunting and Song Sparrow as the most
abundant species, whereas the Song Sparrow and the American Robin were the most
abundant species according to manual acoustic recordings (MAS) and automated acoustic
recordings (AAS), respectively. In Westphalia Township, Song Sparrow was the most

abundant species based on all methods (See Appendix 3.1 and 3. 2).

 

 

       

 

 

 

 

 

 

 

 

| 2.; t
: {AAS PCS
E ‘ T."
'6
3 8_ MAS
C
-- -I
§ 1
U I
a 4%
“6
4* 1
GI I... ...... ,
E West- I West-
ssex . Essex .
phalIa phalIa

 

Figure 3. 2. Mean (2+: SE) of number of bird species at 25 survey points where all three surveys
were conducted in Essex Township and Westphalia Township, MI. PCS, AAS, and MAS refer to
three bird survey methods: Point Count Survey, Automated Acoustic Survey, and Manual
Acoustic Survey, respectively.

The comparative analysis of bird species occurrence measures indicated that there
was no significant difference between the number of encounters by point count surveys
(PCS) and automated acoustic recording method (AAS) in Essex Township (t-value =
1.67, n = 50, P = 0.101) and Westphalia Township (t-value = -0.08, n = 50, p = 0.935).

However, fewer encounters were measured using the manual acoustic recording methods

33

(MAS) compared to the point count survey method (t-value = 4.42, n = 40, p < 0.001 in
Essex Township and t-value = 3.43, n = 48, p = 0.001 in Westphalia Township) and
automated acoustic recordings (t-value = -2.21, n = 40, p < 0.05 in Essex and t-value = -

2.30, n = 48, p < 0.05 in Westphalia).

There were 13 species identified by the point count survey but not by the two
acoustic recording methods in the two study sites. In Essex Township, Acadian
Flycatcher, Alder Flycatcher, Blue-winged Warbler, Common Grackle, Eastern Phoebe,
Green Heron, Pileated Woodpecker, and Wild Turkey were not detected with the acoustic
recordings. In Westphalia Township, the birds identified by the point count methods
only were the Eastern Phoebe, Mallard, Pileated Woodpecker, Red-tailed Hawk, and

Wild Turkey.

In contrast to the use of digital acoustic recordings to detect avian species, we also
investigated how many species were overlooked by the point count method. Five species,
including the Brown-headed Cowbird, Eastern Bluebird, House Sparrow, Ruby-throated
Hummingbird, and Sandhill Crane were identified using the two acoustic recording
methods, but not by the point count method in Essex Township. In Westphalia Township,
7 species, including the Brown-headed Cowbird, Eastern Towhee, House Finch, Hairy
Woodpecker, Scarlet Tanager, Yellow-billed Cuckoo, and Yellow-throated Vireo, were

detected only by the recording methods.

Temporal pattern of bird species richness

The relationship between sampling time and cumulative species richness from

automated acoustic recordings is shown in (Figure 3. 3). The cumulative number of

34

species recorded from 5 am to 8 am accounted for 91 and 92 % of the total species
richness in Essex Township and Westphalia Township, MI respectively. There were only
4 additional species identified after 8 am in the automated acoustic recordings in both

Township surveys (See Appendix 3. & 3.4).

60
a ..
° 5° 2.73 ---- g
C .’ ,-" ’
.: ,J' ,,,,
.2 n ../.‘—""'
b ’ / 'o”
03 40 I / Ln"
.2 / ."'l
i _’/L§"' o Westphalia
a 30 _.//' I Essex
2 /
:3 / - I", ----Log. (Westphalia)
'2 20 {’1' ' —- Log. (Essex)
8 e
'5 10
g .

0
0 5:00 am 6:00 am 7:00 am 8:00 am 9:00 am 10:00 am
Time (Hour)

Figure 3. 3. The relationship between cumulative species richness and time of observation using

automated recorders

35

DEssex

  

 

 

 

 

IWest-
g phalia
8
.c 4 " r
U
N
U
.5 ~
VI
.2
8
0.2 '
U)
u—
o
#1:
0

soo 600 700 7800 790071000”
Time (hours)

Figure 3. 4. Mean (i SE) of number of bird species at 25 survey points where AASs were
conducted in Essex Township and Westphalia Township from 500 to 1000 hours.

36

Discussion

The study showed that overall there was no significant difference between point
count surveys and automated acoustic surveys as methods to estimate species richness,
and abundance of the avian community. Between the two acoustic recording surveys,
however, automated acoustic recordings were more effective at identifying bird species
than were the manual acoustic recordings. Automated acoustic recordings were made 6
times at hourly intervals at each location between 5 and 10 am, compared to manual
acoustic recordings which were made for 5 minutes while the birds were being identified
by the observer. Thus, as expected, automated acoustic recordings were more likely to

capture vocal activities of bird species than manual acoustic recordings.

Although automated acoustic surveys indicated no statistical difference in species
richness per location from point counts, some species were not identified among the
survey methods. This lack of species detection by the acoustic recordings occurred in
part due to the distance of the vocalization from the recording source and the fact that
human observation can see birds that are infrequently vocalizing. Of the 13 species
identified by human observation, 10 of these species were identified beyond 50 m, and 4
species were seen but not heard. When the species identified by human observation
within 50 m were compared with those species identified in the acoustic recordings, the
detectability of acoustic recordings accounted for 100 % and 95 % of all the species by
point counts in Essex and Westphalia Township, respectively. These results show that all
vocalizations by avian species can be identified using acoustic recordings if the detection
distance is set to 50 m. The results also supported the suggestion by Schick (1997) that

50 m detection distance is appropriate for identifying bird species based on the

37

vocalizations among forest habitats in temperate region. He compared the ability to
detect bird vocalizations among varied forest habitats and investigated the appropriate
detection distance of bird songs, using the broadcast vocalization experiment. He found
that all of the vocalizations were heard at 50 m from the broadcast whereas 27 % of the
bird songs were not identified at 100 m. We found that the ability to detect species using
acoustic recordings can be improved in multiple ways and future studies should
investigate the variability of bird species detection in a diversity of physical environments
and at different times (Schieck 1997). In addition, improvement in sensor technology,

types of sensors, and sensor arrays all can improve detection range.

There were 5 and 7 species detected in the acoustic recordings but not in point
count surveys in Essex Township and Westphalia Township respectively. One of the
plausible reasons is that bird observers can overlook some bird species at locations with
high species richness during the dawn chorus in the breeding season when peak of avian
vocal activities occur (Bystrak 1981). The species only detected by acoustic recordings
from 5 to 7 am (sunrise time is 6 am in this study area) accounted for 64 % of all 11
species above (Brown-headed Cowbird commonly found in both sites). Thus, acoustic
recordings are less affected by observer confusion because recordists can repeatedly
listen to the recordings. We can then use visual and auditory cues to detect bird species
from the recordings by the aid of a spectrogram, a visualization of an acoustic .signal with
frequency, time, and intensity domains. Even so, the estimates can be verified by other
field experts or omithologists (Haselmayer and Quinn 2000). Moreover, in the study area,
there are three species whose population status is of special concern in Michigan

(Kalamazoo Nature Center 2002): the Dickcissel, Grasshopper Sparrow, and the

38

Prothonotary Warbler due to their rare status. In our study, these three species were

detected by all three methods.

The results of the study support the “dawn chorus” hypothesis which states that
the peak of avian vocal activity occurs at or near sunrise. Our results found that the
number of species detected by the automated acoustic survey from 5 to 8 am accounted
for more than 90 °/o of all species in the study area. There were only 2 additional species
identified after 9 am including Easter Bluebird and Ruby-throated Hummingbird. Easter
Bluebird is known as a “late riser” or sings sofily so only a nearby female hears songs
(Kroodsma 2007). Indeed, these two species were only detected by automated acoustic

recordings but not by point counts.

Although point counts have played a key role in understanding avian species
richness, abundance, and distribution across North America, the validity and variability
of point count surveys have been questioned over years (Bystrak 1981, Thompson 2002).
This study showed that by using an automated acoustic survey there was improved
detection of species and reduced biases in observer errors. Moreover, automated acoustic
recordings decreased the temporal variations by simultaneous recordings at multi-
locations. Automation of acoustic recordings facilitated monitoring bird vocal activities

without human interruption.

Despite the advantages of using automated acoustic recordings, some limitations
are evident. First, acoustic recordings cannot easily estimate abundance of avian
communities at locations. It is not feasible to distinguish different individuals of each

species only by auditory cues. However, the development of new computer tools such as

39

pattern matching and speech recognition to detect the unique acoustic signatures of
individuals is progressing (Mills 1995, Chesmore and Nellenbach 1997, Chesmore 2004,
2007, Chesmore and Ohya 2007, Kasten et al. 2007). In addition, deploying a multi-
array of microphones at a location can help to estimate the distance and direction of the
sound sources, leading to an estimation of abundance and distribution of bird

communities (Asano et al. 2001, Otsuki et al. 2007).

Since the automated acoustic surveys were conducted using cassette tape
recorders in the study, it required many hours to digitize the tapes and to archive the
recordings into an acoustic digital library. Moreover, it takes significant time for
researchers to repeatedly hear and identify bird species. Thus, while the current study
provides proof-of-concert, the development of automated acoustic sensor systems is

necessary for wider application of these techniques (Gage et al. in submission).

In conclusion, the results provided evidence that acoustic recordings can be used
as an alternative means to survey avian communities. Our study shows that automated
acoustic recordings facilitate breeding bird surveys at multiple locations with minimal
variability and high detectability of bird community measures, leading to correct
interpretation of a long term pattern of avian species composition and distribution in

regional scale (Rosenstock et al. 2002, Thompson 2002).

40

Chapter Four
Analysis and Interpretation of the “Heartbeat of the City”
using Acoustic Signatures along an Urban-rural
Gradient

Introduction

Urbanization causes substantial alteration of ecosystem structure and function
(Vitousek et a1. 1997). Noise, defined as "unwanted or detrimental sound or consistent
level of background sound" (Stutz 1986, F orkenbrock and Schweitzer 1999), can be
considered one of the main pollutants that causes degradation of human health and has
negative impacts on animal communication. The impact of noise can vary depending on
individuals; noise causes a variety of physical and psychological stresses from
disturbance of sleep and communication to noise annoyance and even hearing loss
(Forkenbrock and Schweitzer 1999, Ouis 2001, Warren et al. 2006). With this
recognition of noise pollution, the European Union (EU) passed legislation that required
member countries to provide the public with noise maps of all major or industrial cities,

highways, and airports by 2007 (Butler 2004).

Recently, several studies have attempted to investigate the effects of urban noise
on the reproductive success and distribution of animals. The breeding success of many
avian and amphibian species appears to be impaired near roads with high traffic such as
highways (Brumm and Todt 2002, Slabbekoom and Peet 2003, Sun and Narins 2005).
Extensively growing urbanization not only causes alteration of habitat structure and
function, but also provides a novel acoustic environment where animals must either adapt

or emigrate to communicate and reproduce. Warren et al. (2006) provided a conceptual

41

overview of how some avian species have adapted to the acoustic environment of urban
systems. Their study classified the modifications of bird calling behaviors by human
noise into three categories: I) "Amplitude shifts": sound amplitude of animal
vocalizations increases when human noise occurs; 2) "Frequency shifts": many species
produce their songs at higher acoustic fi'equencies than normal since most anthropogenic
sounds have lower frequencies; 3) "Temporal shifts": birds shifi the timing of their songs
in order to avoid traffic noise. These adaptations of birds to the urban acoustic
environment can make their reproduction successful. However the reproductive rate of
most amphibians was drastically reduced near roads with heavy traffic, because
amphibian vocalizations are masked by traffic noise (Sun and Narins 2005). Although
some studies have addressed urban noise as a critical impact on the communication of
vocal organisms and their reproductive success (Reijnen and Foppen 1994, Rundus and
Hart 2002, Rabin et al. 2003, Katti and Warren 2004, Warren et al. 2006), little work has
been done to understand the acoustic characteristics and the interaction between

biological communication and anthropogenic sounds across various landscapes.

The structure of urbanized systems generally includes heterogeneous landscape
patterns, ranging from rural to suburban to highly developed areas across the ‘urban-rural
gradient’, depending on the density of human population. The urban-rural gradient can
be characterized by several factors including: distance, land use types, vegetation
structure, and landscape attributes (Blair 1996). The characteristics of environmental
sounds can also vary depending on habitat type, the mosaic of habitats within the
landscape, the time of day, and season of the year. Patterns of acoustic signals therefore

reflect the dynamics of biological, social, and physical systems within each landscape.

42

Many groups of animals use acoustic signals to communicate information such as
breeding condition, territory size, and predator alerts. Typically, ecosystems with a lesser
degree of anthropogenic interference exhibit greater complexity in terms of sound
frequency and periodicity. When anthropogenic noise disrupts this communication,
critical information is not relayed and may result in population declines (Krause 1997).
With the growth of human populations and their subsequent expansion away from urban
centers, dramatic changes in the acoustic environment may occur. Therefore,
investigation of how changing acoustic patterns along urban-rural gradients influence
habitat quality and reflect that habitat’s capacity to sustain its array of organisms is of

critical importance.

Since human-induced noise has a critical impact on animal communication,
reproductive rate, and the abundance and distribution of species, there should be an
inverse relationship between Anthrophony (human-induced sounds) and Biophony
(biological sounds) (Figure 4. 1). Biophony is expected to be greater in rural areas
whereas Anthrophony is expected to be greater in the city where there is more human
activity. Given the hypothetical relationship between anthropogenic and biological
sounds, the various soundscape patterns will be dominant in different land types at
different times of the day and vary depending on the season. Highly urbanized areas (e.g.,
commercial sites) are expected to have human-induced sounds outcompeting biological
sounds over all seasons. In contrast, places with low development (e.g., rural sites) should
retain higher biological sounds in breeding seasons than anthropogenic sounds (Figure 4.
1). Assuming that the theoretical relationship is applicable to any urban-rural gradient,

we can investigate the urban-rural gradient by measuring several acoustic attributes

43

including acoustic frequency, acoustic energy, and diversity (acoustic patterns) in an

urban-nual landscape at different times of day over seasons.

The main objective of the investigation of acoustics in urban systems is to

quantify the acoustic patterns in different landscapes across an urban-rural gradient. To

meet the objective, we investigated whether: 1) characteristics of acoustic signals vary

along urban-rural gradients and over different seasons; 2) urbanization adversely affects

animal vocalizations; and 3) acoustic properties of urban systems are correlated with the

structure and composition of landscapes.

Anthropogenic acoustic attribute

 

 

Urban

Rural

Urban-rural gradient

Blologlcal acoustic attribute

 

 

Urban

Rural

Figure 4. 1. Conceptual relationship between biological and anthropogenic acoustic attributes

including acoustic energy at different acoustic frequency range, the number of species and

vocalizations identified along urban-rural gradients. (Modified from Stevenson et al. (2004))

44

Methods

Study area

This study was conducted in the Greater Lansing area where Lansing is the capital
of Michigan, and the sixth largest city in the state (Census 2006). Nineteen sampling
sites were selected along two transects: one crossing from northeast Lansing (Bath
Township) to southwest Lansing (Delta Township), and the other crossing from
northwest Lansing (DeWitt Township) to southeast Lansing (Mason Township). The two
transects cross at the Capitol Building in downtown Lansing. Most sites were spaced at
1.6 km (one mile) intervals. The distance from the most central urban site to the furthest
nrral site at each transact was about 10 to 11 km. The 19 sites were classified into five
land use types: (1) rural areas, (2) urban parks, (3) urban residential areas, (4) agricultural
areas, and (5) commercial areas (Figure 4. 2). An urban-rural gradient was defined with
the five land use types ranked from the highest to lowest degree of urban development.
The amount of land cover types was calculated at each site in a 100 m-radius circle (as
determined from the maximum distance of microphone capacity to capture acoustic
signals) using 2001 National Land Cover Data provided by the U. S. Geological Survey
and 2005 aerial imagery by the Michigan Center for Geographic Information. The area

of land cover areas at each site was converted to percentage of total site area (Blair 1996).

45

I cum-u z ~~\
\.\ // 9 'I i M.“ “.hf @ die \‘ 1 ,I
If”N i I y I ' ~:
(L, . Y“ I f""" l r f .23..
I Aha—I.- ) u, ‘ VI I . " A.
$1 I on” n~~ e 1‘ , u'
, w i r , .«~ I . ..
[I fi-nv-~ «wow-Q Lsth . I I .l‘
l ‘ m “w. I 2 a: I
w 4- mm en”... .~. . ”j“. 1
.1“-.. ”t"- . v1 . I I _..
no...“ ,,.. N ergo-m I‘ a..." ~..,. ‘
§ ‘ , ff”: 1 l I :’ Q 3‘33: i :C‘ums
I .r ? ___‘_ l ‘ : ..... I
I , i t ' “I? i i
// 7 ? win». , I . - Pf 3 Q .,.I.
.. ' ,9- l . )/./P’\‘:§V ‘ reams:
[I i I “\“‘-<_,.«' ' M j (y VMIsslng
/' ‘ g I

Figure 4. 2. Map of the study area in the Greater Lansing area, MI. Each named symbol in the
map represents the location where an acoustic monitoring unit was deployed along an urban- rural
gradient. The green symbols represent data that were collected from February to December 2006,
and the red symbols are the locations where the recording devices were lost during the study
period.

Survey methods

Environmental acoustic samples were collected once a month from February to
December in 2006. At each location an acoustic recording unit was deployed. The
recording technology used was an audio cassette-recording device (Sangean
VersaCorder®, C. Crane Co.) and one powered Omni-directional boundary microphone
(Model 330-3020, RadioShack Co.). The recording device was equipped with a timer
which enabled automated collection of acoustic samples six times per day without human
interaction. The microphone selected was able to capture the acoustic frequency range

from 40 to 14,000 Hz within which most songs and calls of vocal organisms are produced.

46

Each acoustic unit was set to record for 3 minute duration, three times in the morning at
2:00, 6:00, and 10:00, and three times during the day and evening at 14:00, 18:00, and

22:00 for two consecutive days.

The recordings were digitized and transferred to a wave file with a 22 kHz
frequency sampling rate and a monaural channel for quantification and analysis of
environmental acoustic signals. The size of the data files was large and required a high-
performance computer and extensive analysis. Each recording was divided into six 30-
second sound clips. Only the 3 middle sound clips were analyzed because the first and
last part of the raw sound files had recording variability such as recording duration and

technical sounds due to the recorder start-stop cycle.

Data Analysis

Two analytical methods were used to extract the acoustical information from
sound samples (Gage et al. 2001): (1) Acoustic Intensity Analysis, used to classify
environmental sounds into three main categories (biological, anthropogenic, and
geophysical sounds) by acoustic intensity and frequency range; and (2) Acoustic
Identification Analysis, which identifies sound sources or vocal organisms from sound

recordings.

(1) Acoustic Intensity Analysis

Each sound sample was analyzed to determine the spectral energy in l-kHz
frequency intervals by developing a program based on the Welsh algorithm (Welch 1967),

using MATLAB software (Mathworks 2005). The numerical values resulting from this

47

process provided a quantitative measure of how the energy is distributed across the
acoustic spectrum. Based on analyzing a large number of acoustic recordings
(Napoletano 2004), it was determined that the mechanical sounds (Anthrophony) were
most prevalent between 1-2 kHz, and biological sounds (Biophony) were most prevalent
between 3-8 kHz. Biophony was computed by the sum of acoustic power values at
frequency range from 3 to 8 kHz. Similarly, Anthrophony is the acoustic power values
between 1 and 2 kHz.

Based on the formula of Normalized Differential Vegetation Index (Myneni et al.
1995), an Acoustic Habitat Quality Index (AHQI) was developed to provide a
classification of a place relative to its biological composition and human disturbance
based on the amount of acoustic energy in different frequency intervals. Mechanical
sounds produced by most machines (cars, airplanes, trains) occur at lower frequencies
(average 1.5 KHz) whereas most biological sounds occur at higher frequencies (average
4.5 KHz). The distribution of sound frequencies in acoustic samples was computed to
determine what types of sounds occurred (e.g., mechanical, biological or physical).
Normalized ratios of frequency levels were then calculated to evaluate the relative
amounts of biological or mechanical sounds in a sample or set of samples. The equation

for AHQI is:

3+

9

”Q

 

AHQI:

Q

where a, [1 represent the total amount of acoustic energy of mechanical sounds

and biological sounds, respectively. The value of the index ranges from -1 to 1. If the

48

value is negative, mechanical sounds are dominant at a sampling place. If the value

approaches 1, biological sounds dominate at the site.

An Analysis of Variance and Tukey’s multiple comparison method were
performed to test whether acoustic variables, particularly the total sound level, the
percent of both biological and anthropogenic sound intensities, and the AHQI, exhibit
significant difference across different land use types. In addition, we investigated

whether temporal pattern of AHQI varied across the land use types.

(2) Acoustic Identification Analysis

The recording process enabled the identification of species specific vocalizations.
Bird species were identified from the acoustic recordings at each site, and the relative
number and the calling frequency of species were quantified. Four acoustic samples were
selected per site, recorded at 6 am on May 17-18, and June 14 -15 2006. In general, peak
vocalization of bird species occurs about sunrise during their breeding season (Staicer et
al. 1996). Regardless of how many calls and songs by the same species repeatedly
occurred in one sample, these vocalizations were considered as one "encounter" in the
sample (see Chapter 3). The maximum number of encounters per species was four
because the total number of samples analyzed per site was four. By estimating the
number of encounters per species, the relative occurrence probability of each species was
calculated based on the ratio of the number of encounters in four acoustic recordings per
site. For example, the House Finch was identified from all acoustic samples at site E05
(commercial area), so the number of encounters of the House Finch at that site was four.

In contrast, because the American Robin was recorded two times in the samples, the

49

number of the encounters was two. Accordingly, the probability of the House Finch and

the American Robin was 1 and 0.5, respectively.

Bird species diversity was calculated by using a modified Shannon-Weaver Index
(Shannon and Weaver 1963, Magurran 1988, Blair 1996). The equation used to compute

biological diversity is:

H': Zpi10g(pi)
i=1

where H ' is the Index of biodiversity and c is the number of bird species. Pi is the
proportion of encounters of the ith species to total encounters of all species at each site.
The mean values of species diversity for different land use types were calculated and
compared. In this study, species richness was defined as the total number of bird species

identified in the recordings.

The effects of urbanization on the composition of communities and distribution of
individual species were analyzed in basis of Canonical Correspondence Analysis using R
statistic package software (Blair 1996, R Development Core Team 2007). This analytical
model has been applied to investigate how species respond to environmental gradients by
separating the distribution of species in an ordination plot (ter Braak and Prentice 1988,
Blair 1996). For this analysis, the number of encounters of each species at a site was
used, and the environmental factOrs consisted of percentage area covered by forest,

pasture/crop, lawn, and buildings/paved roads at each site (Blair 1996).

50

Results

The percent of area covered by forest, lawn, pasture/crops, and
buildings/pavement varied with land use type (Figure 4. 3). Rural residential sites had
the highest percent area covered by forest (49.2 i 16.6) and the least amount of area
covered by buildings/pavement. In contrast, commercial sites had the highest percent area
covered by buildings/pavement (99.3 i 0.7) and the least percent area covered by forest.
The percent of land covered by lawn and pasture/crops was highest in the urban park and

agricultural sites respectively (60.3 i 14.4; 53.8 i 7.8).

'0 a ‘ b
2 8O ), )
cu
>
O
U40
a .
2
rd
o\° 0 - ah
C) d)
C -H’ ba L I "Aid-2‘ ibi-
ergiiilm Hgsid'ential Eggiailential wife-if u arrkan

Figure 4. 3. Average percent of land covered by a) forest, b) lawn, c) pasture/crops, and d)
buildings/pavement estimated by land cover maps and aerial photo imagery (mean + standard

error)

51

Average total sound levels (dB) varied across all land use types (Figure 4. 4a).
Land use type was ranked from the highest (left) to lowest (right) based on the degree of
urban development, the urban-rural gradient. Commercial land use had significantly
higher sound levels among sites (65.31 :t 0.104), and lower sound levels were found in
agricultural, urban park, and rural residential land uses (One-way ANOVA, F = 506.1, p
s 0.05). Urban residential land use had moderate sound levels and sound level was

significantly different from other land uses (62.78 i 0.12).

66 Sound level (dB) a) 0-4 AHQI b)
58 ‘ j i o - 4 A.
50 ___ _ ._'_ . _...i __ '0-4
.. 0° ‘ I l

60 . Anggrophony 60 % of Biophony C I
40 ‘ ‘ ’ 4o

20 20

0Comm- Urban ural 0 ' -

ercial Residential Residential ggr'gfu' gem?"

Figure 4. 4. a) Average sound levels, b) Acoustic Habitat Quality Index (AHQI), c) proportion of
anthropogenic sounds (Anthrophony), and d) proportion of biological sounds (biophony) of 5
different land use types where the acoustic samples were recorded from February to December,
2006. The land use types was listed on x-axis along an urban—rural gradient, defined by ranking
the land use types in order from the highest (left) to lowest (right) degree of urban development.

52

The bars represent mean i SE (n = 8475). Lowercase letters refer to means contrasts among
different land use types using Tukey’s HSD tests.

The proportions of biological and anthropogenic acoustic intensity varied along
the urban-rural gradient. Based on computation of Anthrophony and Biophony from
February to December, 2006 (Figure 4. 4b and 4.4c), the average percent of Anthrophony
gradually decreased along the urban-rural gradient, whereas the proportion of Biophony
increased along the gradient. The proportion of Anthrophony was highest in commercial
land use and lowest in urban park and rural residential land use; whereas the proportion
of Biophony was highest in rural residential, urban residential, urban park land use and

lowest in commercial land use (p < 0.05).

The Acoustic Habitat Quality Index (AHQI), a normalized ratio of Biophony and
Anthrophony, showed the pattern similar to the Biophony measure (Figure 4. 4d).
Commercial land use had the lowest AHQI value (-0.34 :l: 0.01), and rural residential and
urban park land use had the highest values of AHle across all land use types (0.03 :t
0.007; 0.01 i 0.009). Although the AHQI values in rural residential and urban park land
use were the highest in the study area, the values were small positive numbers, indicating

that these land uses had only slightly more Biophony than Anthrophony.

Environmental sounds exhibited temporal variability across the different land use
types examined. Because acoustic samples were made multiple times per day we were
able to examine the diurnal and seasonal changes. Temporal patterns of AHQI in
commercial and agricultural land uses were clearly different fi'om those in rural
residential land use types (See Figure 4. 5). The negative trend of AHQI in commercial

and agricultural land uses occurred during most of the sampling period, whereas AHQI

53

values were positive in rural residential land use except during winter. of AHQI showed
high variability in urban residential and urban park land uses depending on time of day or

season of the year.

I .0
0.5

-|.0

Feb. May Aug. Nov.

Figure 4. 5. Temporal changes in Acoustic Habitat Quality Index (AHQI) at 5 different land use
types: (a) commercial; (b) agricultural; (c) urban residential; (d) urban park; and (e) rural
residential sites. Data were collected 6 times a day for two consecutive days at each month from
February 14 to December 12, 2006. Bars indicate average values of AHQI at each time of day

during the month.

54

Twenty eight species of birds were identified from the recordings. Avian
community species richness, abundance, and diversity were not significantly different
between the sites (Figure 4. 6). However, commercial sites had the lowest values of
species richness and diversity and rural residential exhibited the greatest richness and

diversity.
a) Species richness b) Bird abundance

8 l6
4 8

0 0

c) Avian diversity Index

2.0

Figure 4. 6. Avian community measures along the urban-rural gradient from May and June 2006:
a) mean number of bird species, b) mean number of encounters, and c) mean value of Shannon’s
diversity index at each landscape. Note that the bird species and their vocalizations were

identified fi'om 4 acoustic samples at each site during the period.

The relative occrurence probability of all avian species varied along the urban-

rural gradient based on bird species identified by listening to the acoustic recordings

55

(Figure 4. 7). Four "urban adaptable" species were found in highly to moderately
urbanized land uses: Cedar Waxwing (only detected in the commercial area); House
Sparrow, House Finch, and European Starling (> 0.75 probability in the commercial area).
Fourteen species were "suburban adaptable" (Blair 1996). These included the Song
Sparrow, Northern Cardinal, American Robin (found in all types of land uses); Red-
winged Blackbird, Chipping Sparrow, Gray Catbird, Yellow Warbler, Baltimore Oriole,
American Goldfinch, Brown-headed Cowbird, Blue Jay, Common Yellowthroat, Horned
Lark, Willow Flycatcher (the maximum occurrence probability of all species in urban
residential land use). Seven species were "urban avoiders" with the maximum occurrence
probability in rural land use (Tufted Titrnouse, Black-capped Chickadee, House Wren,
and American Crow), or only found in rural land use (Downy Woodpecker, Great
Crested Flycatcher, and Red-eyed Vireo). The most abundant birds in the agriculture

land use were the Canada Goose and Mourning Dove.

56

C\ “3?» MO

co 6‘ rs 0" ages“? 0633“
Cedar Waxwin g
House Sparrow
House Finch
European Starling
Red- winged Blackbird _
Song Sparrow
Northern Cardinal
Chipping Sparrow
American Robin
Gray Catbird
Tufted Titmouse
Canadian Goose §
Black-ca ped Chickadee 3
House ren E
Killdeer
Mourning Dove
Yellow Warbler
American Goldfinch
Baltimore Oriole
Brown-headed Cowbird
Blue Jay
Common Yellowthroat
Horned Lark

Willow Flycatcher
American Crow
Downy Woodpecker .
Great Crested Flycatcher §
Red-eyed Vireo 3

me

o conq-o-o-otelee|.--. eulllfl~‘~l ~.-laeuo. ll. so no a ~o one on .. e o
. . o...-oqneeI- a e. on to e- u 4 4|... .. .. ..... . .. .
. IIAactsrcOCO II I o- O. In 0. I - lII'I. on a GI II I. - .u n. .-
II- -. |~ one one...-n.-a-.-u-eelee cu-nu- tree- . .euoe. .-. .- .. . a to I
.. . e o -.nsI->-.a.-- -I-u.‘. ............ nylon. . . .. .I ~ eel. .

 

———:0<n<0.25 —-—':0.25<n<0.5 — :0.5<n<0.75

—I:0.75<n<0.99 - :n=1.00

Figure 4. 7. Relative occurrence probability of avian species across all sites, listed in order from
highest occurrence probability in commercial sites to greatest occurrence probability in rural

residential sites (n = number of encounters/ total acoustic surveys)

A Canonical Correspondence Analysis (CCA) was performed to quantify the
distribution of bird species in relation to the percent of land cover type. There was a
significant relationship between the number of encounters of bird species and land cover
types (Figure 4. 8). The direction and length of the arrows in the ordination diagram
indicate the changes of environmental variables and the extent of correlation between the
variables and the ordination axes (Blair 1996). The distribution of bird species in relation
to the arrows (enviromnental variables) represents the preference of each species for a

land cover type.

The canonical axes (the first two ordination axes) accounted for 26% of the
variation in the bird species presence/absence data, which explained the percentage
variance of the weighted average scores distributed in the ordination plot. The first two
axes had species-environment correlations of 0.90 and 0.79. The correlations were
calculated in combination with linear combination scores of environmental variables
(here, the percent area of four different land cover types) and site scores that measured
the weighted average of species scores (ter Braak 1986, Blair 1996, Oksanen et al. 2007).
The first canonical axis in the plot separated bird species into groups by the extent of
urbanization. The second canonical axis partitioned species into groups by open field

(grassland and agricultural land uses) or non-open field (forests and building land uses).

The overall analysis and the first axis were marginally significant with more than
50 permutations (CCA: p = 0.08, Axis-1: p=0.06). This analysis demonstrates that the
most important variables for relating bird species and land uses were the amount of area

covered by building/pavement, forests, and pasture/crops.

58

Figure 4. 8. Ordination diagrams of canonical correspondence analysis (CCA) using bird survey
data with environmental variables. (a) CCA ordination of 28 bird species (points) against 6
environmental variables (arrows; the percent areas of different land cover types: forests, lawn,
pasture/crops, and buildings/pavement). The contribution of each environmental variable to the
ordination axes is represented as the length of the arrow. The direction and distance of species
scores to the arrows indicate how well the abundance of each species is related to each
environmental variable (ter Braak 1986, Palmer 1993, Blair 1996). The angle between the arrows
represents correlation between environmental variables. The smaller angle the arrows have, the
greater the environmental variables are correlated. For example, the Forest and Lawn variables
are highly correlated to one another. Four letter abbreviations represent species alpha
code(Sharpe 1886). (b) CCA ordination of 18 sampling sites and linear combinations of
environmental variables. The orientations of the site scores to the arrows represent how strongly

the sites are related to environmental variables. See Appendix I for bird names and AOU codes

59

CCA2

 

 

 

 

a) mo

Fg rest

 
  
    

 

 

 

 

 
   
   

 

 

HOWR
Building CEWA EUST BHCO Lawn
, 6. HOSP YEWA BLA
0‘ a m n HOFI F— AMRO NOCA J
GRCA
Pasture MODO
@919; 505?
RWBL GANG AMCR
. KILL
23 HOLA
cows
WAVI
'4 '2 0 2 4
2 A b) Rural resid.
Commercial '
Commercial ELTQSI
- . Commer ial. Lawn
MM Urban reSid. rlCUltural U b k
—— I r an par
0 Pavement 1::— Rural resid.
z Pasture
i Q l .
i r s ; Rural Resid.
, Urban resid.;
i Agricultural é Agricultural
I a f
2 Commercual;
i
l
e
-4 -2 o 4

60

Discussion

The results show that acoustic properties are associated with the structure and
composition of urban landscapes in several distinct ways. In general, biological acoustic
energy increased along a gradient from highly urbanized to rural land uses, whereas
anthropogenic acoustic energy was inversely correlated with the urban-rural gradient (See
Figure 4. 4). Warren et al. (2006) showed that ambient noise levels, mostly consisting of
motor vehicle traffic volume, were inversely correlated with the distance from an urban
center to a suburban area in Phoenix, Arizona, and suggested fiarther study to show how
ambient noise or anthropogenic sounds affect animal vocalizations. However, that study
did not reveal the source of ambient sounds. The environmental sound levels in our study
did not show the same pattern as the measure of Anthrophony or biophony. Rather,
simultaneous measures of biophony and Anthrophony strongly indicated that human-

induced sounds were negatively associated with biological sounds.

On closer examination of the temporal patterns of AHQI values (Table 4. 1),
commercial and agricultural land uses had negative AHQI values from February to
December in 2006, whereas rural residential land uses had positive AHQI values,
especially during the breeding season of vocal organisms. These results illustrate that
high level of human activities and urban development can be negatively associated with
animal vocalizations. Although agricultural land use includes large amount of vegetation,
negative values of AHQI were caused by frequent mechanized cultivation activities
(plowing, fertilizing, and harvesting) that overlapped with biological activities. In
addition, agriculture land use did not provide suitable habitat for vocal animals because

most agriculture land use in the study area planted in monoculture (low plant diversity).

61

In land uses with moderate development, including urban residential and urban park land
uses, variability of AHQI values was high over all seasons, whereas AHQI values were
positive in rural residential land use and remained consistent except in winter. The
results imply that vocal species can respond differently to various levels of anthropogenic
sounds. In general, vocal activities of bird species are most intense in habitats with the
least human disturbance. However, high variability of biophony and Anthrophony
occurring in land uses with moderate development suggests that some bird species in
urban land use may modify their vocalization levels. Given that ambient noise influences
communication of breeding birds, some studies have shown that male birds had higher
song amplitude in a noisier environment than birds in natural habitats (known as the
Lombard effect) (Brumm and Todt 2002, Brumm 2004, Katti and Warren 2004, Warren

et al. 2006).

62

Table 4. I. Temporal pattern of mean Acoustic Habitat Quality Indices (AHle) over months

and at six times of day in all land use types.

 

 

 

112:0!“ C°mmmial Agficu'm’a' Exildlential U'ba“ Park gzgdential AN?“
Feb —0.37a -0.09 a 0.02 a -0.17 a -0.24 a 0.01
Mar 039 a -O.27 b -0.10 c -0.03 c -0.06c < 0.001
Apr 057 a -0.17 b -0.07 b 0.17 c -0.08 b < 0.001
May -0.22 a -0.12 a -0.05 a 0.18 b 0.06 b < 0.001
Jun -0.20 a -0.11 ab 0.19 c -0.04 b 0.11 c < 0.001
Jul 026 a -0.26 a 0.14 c -0.01 b 0.26 d < 0.001
Aug -0.46 a 0.09 b -0.01 b 0.29 c 0.30 c < 0.001
Sep 031 a -0.15 b 0.05 c -0.07 b 0.13 c < 0.00]
Oct -0.32 a -0.17 ab -0.03 b 0.02 be 0.11 c < 0.001
Nov -0.49 a -0.32 b -0.17 be -0.04 c -0.17 b < 0.001
Dec -0.15 a -0.49 a -0.22 a -0.08 a 0.00 a < 0.001
Time

of day

02:00 -0.30 a -0.05 b 0.15 c 0.09 c 0.10 c < 0.001
06:00 -0.28 a -0.24 a 0.07 b 0.03 b 0.00 b < 0.001
10:00 -0.34 a -0.14 b -0.04 b 0.02 be 0.08 c < 0.001
14:00 -0.38 a -0.28 b -0.10 c -0.02 be 0.03 c < 0.001
18:00 -0.40 a -0.36 a -0.17 b -0.05 c -0.02 c < 0.001
22:00 -0.37 a -0.16 b 0.01 c 0.03 c 0.00 c < 0.001

 

* Mean AHQIs with the same superscript letter in each row are not signifinicantly different from

one another (Tukey’s multiple comparison)

63

The results suggest that vocal animals might overcome ambient noise in order to
effectively exchange information with their conspecifics (Brumm 2004). However, not
all animals show this type of behavior. Some birds shift either the song frequency range
from low pitch to high pitch, or the timing of their songs, to counteract intense acoustic
disturbance generated by human activities. Some species are not affected by ambient

noise for their communication (Warren et al. 2006).

Given the relationship between Biophony and Anthrophony along the urban-rural
gradient, it is possible to characterize urban landscapes using environmental sounds. A
“Soundscape map” was produced based on the values of the Acoustic Habitat Quality
Index. The average values of AHQI at the sampling sites were joined to attributes of
each land cover class in the National Land Cover Data map (NLCD) 2001, and a map of
the AHQI values were overlaid on the land cover classes in the study area. The spatial
variability of the mean AHQI values is shown in Figure 4. 9. Negative values of AHQI
occur in the center and main roads of the city, whereas the areas with less development
have positive AHQI values, indicating higher levels of Biophony. This soundscape map
showed how a cityscape can be characterized using environmental sounds. The accuracy
of the map of AHQI values can be enhanced with an increase in the number and
distribution of sampling sites and with additional ground verification (Gage et al. 2004).
However, such soundscape maps can also be used to visually assess temporal dynamics
of urban acoustic environment because our acoustic samples were made multiple times of

day and over several seasons.

64

    
  
 
 
 
  

1 £J5<x<£5
-o.5<x<-o.25 .
.-0.25<x<0

_ 0<x<o.25

025<x<0.5
Water

Figure 4. 9. Distribution of Acoustic Habita Quality Index (the normalized ratio of biological
sounds to anthropogenic sounds) based on National Land Cover Data map 2001. The index
ranges from -I to 1; positive values indicate that the intensity of biological sounds is higher than
one of anthropogenic sounds, and vice versa.

The pattern of relative bird occurrence probability (encounters in a land use
type/total encounters) demonstrates that the response of bird species to land use type can
vary depending on the extent of urbanization (Blair 1996). Five species appeared to be
highly adapted to urban areas. Three species, House Sparrow, House Finch, and

European Starling, accounted for more than 75 % of bird occurrence probability in land

use with high or moderate development. These species were introduced to the United

65

States and have successfully utilized resources in human-dominated ecosystems (Long
1981). The fourth, Cedar Waxwing, is known to be more common in open mral or
suburban areas (Stokes and Stokes 1996), but in this study this species was detected only
in commercial land use. This species, which generally feeds on a variety of fruits, berries,
and insects, has expanded its territory to urbanized areas because serviceberries or

weeping cherries are abundant as ornamental trees.

The urban residential land use had the highest species richness and bird
occurrence probability along the urban-rural gradient. These suburban adaptable species
probably utilize resources available in both urban and rural sites. This type of habitat
structure can provide edge habitats with high primary productivity and diversity of
resources (Cody 1985). Most classic studies in bird ecology conclude that urban sites
resulted in lower bird diversity and abundance than natural sites (Graber and Graber 1963,
Batten 1972, Guthrie I974, Beissinger and Osborne 1982, Green 1984, Degraaf and
Wentworth 1986). In contrast, Blair (1996) demonstrated that moderate levels of
urbanization had the maximum species richness, bird occurrence probability and biomass,
and that the Shannon diversity index peaks across various land use types in Palo Alto,
California. He argued that the contradictory conclusions were drawn because previous
works focused on comparisons of bird community measures in only two locations rather

than in sites with various degrees of urbanization.

However, we conclude that moderate levels of urbanization provide a suitable
habitat structure and abundant resources for bird species. The study did not investigate
whether high occurrence probability and diversity of birds resulted in high reproductive

success, and thus whether the population of each species can be sustained within the site

66

(Blair 1996). In addition, the edge habitats provided by urbanization can result in higher
cowbird parasitism and nest predation than, for example, the interior of forest habitats

(Gates and Gysel 1978, Johnson et a1. 1990).

This study revealed that it is more critical to investigate the relationship between
the biological and anthropogenic sounds rather than only the measure ambient noise
levels. The study also provides clear evidence that biological sounds (biophony) are
inversely correlated with anthropogenic sounds (Anthrophony) along an urban-rural
gradient. This study also shows that recording and interpreting environmental sounds in
human-dominated systems can offer enormous opportunities to not only measure ambient
noise levels, but to also understand the components of the acoustic environment and how
each component is related to another. It is suggested that urbanization affects not only
the transformation of habitat structure for vocal organisms but also changes the

soundscape, which can cause impairment of animal communication and breeding success.

67

Chapter Five .
Development of Automated Acoustic Sensor
Observation System via Wireless Networks

Introduction

Humanity has profoundly altered and degraded the biosphere in many ways
(Vitousek et al. 1997). Pervasive human activities have resulted in the perturbation of the
main biogeochemical processes and the transformation of about 40 to 50 % of the land
surface on Earth. In addition, human domination and modification of earth systems have
led to the massive loss of biological diversity (Vitousek et al. 1997, National Research

Council 2001).

Given the drastic modification and degradation of ecosystems by human activities,
the National Research Council (2001) proposed the “grand challenges of the
environmental sciences”: biogeochemical cycles, biological diversity and ecosystem
functioning, climate variability, hydrologic forecasting, infectious disease and the
environment, institutions and resource use, land-use dynamics, and reinventing the use of
materials. Although those challenges in environmental science require measuring and
monitoring complex environmental variables at various spatial-temporal scales, most
ecological research has been conducted at small spatial scales and over relatively short-

time periods (Porter et al. 2005).

Recently, several researchers argued that new advances in wireless sensor

technology will help scientists to better understand the changes in ecosystem structures

68

and processes at relevant spatial and temporal scales, providing a great deal of the
spatially dense, near-real time observations in ways that were previously impractical and
inaccessible to ecologists and environmental scientists (Estrin et al. 2003, Martinez et al.
2004, Porter et al. 2005, Butler et al. 2006, Collins et al. 2006, Hamilton et al. 2007,

Kasten et al. 2007).

A wireless sensor network is a system with spatially distributed automated sensor
arrays that monitor environmental conditions using wireless communication (Romer and
Mattem 2004). Wireless sensor networks were initially developed for military
applications and now have been used in many areas including agriculture, traffic controls,
environmental monitoring, etc. (Estrin et al. 2003, Romer and Mattem 2004, Porter et al.
2005). In general, the sensor arrays used in environmental and ecological science can be
classified into three main categories: 1) physical, 2) chemical, and 3) biological sensor
arrays (Estrin et al. 2003). To date, physical sensors have been well developed and are
relatively inexpensive. These sensors measure meteorological variables including
temperature, moisture, and wind speed and direction, and photosynthetic active radiation
(the spectral range of the sun light from 400 to 700 nanometers, which drives the process
of photosynthesis in plants). Chemical sensors are used to monitor biogeochemical
cycles including direct measurement of nitrate, carbon dioxide, and phosphorous. These
types of sensors are expensive and require high power consumption. While the
development of physical and chemical sensors has been relatively advanced and widely
used in environmental science and ecology, animal studies posed challenges to scientific
and engineering communities because most biological investigations require field surveys

to directly measure community characteristics such as species richness, abundance, and

69

population distribution. Today, advances in sensor technology and the widespread
network infrastructure have facilitated animal scientists to exploit the advantages of
wireless sensor networks such as tracking the movements of birds, small animals, and
insect monitoring with infrared sensors (Mainwaring et al. 2002) or equipped with GPS

sensor nodes to estimate animal location and speed of the movement (Juang et al. 2002). .

Acoustical sensor arrays are used to investigate the relevance and feasibility of the
microphone as the sensor integrated with wireless sensor networks to conduct ecological
research. Acoustic signals can provide fundamental information about communication of
animals, potentially providing presence/absence data about vocal organisms (Kroodsma
and Miller 1996, Gage et a1. 2001, Gage 2004, Qi et al. 2008, Gage et al. in submission).
Biological sounds have been used for many animal surveys (e. g., Breeding bird surveys
and Breeding Amphibian surveys) (Bystrak 1981, Robbins et al. 1986, Bridges and
Dorcas 2000, Haselmayer and Quinn 2000, O'Connor et al. 2000, Richter and Azous
2001, Hobson et al. 2002, Kendell et al. 2002). Furthermore, acoustic sensor nodes can
not only record biological communications and environmental conditions, but also

simultaneously sense anthropogenic activities (Gage et al. 2001).

To implement stationary and simultaneous measures of environment sounds at
different locations, a recording system was developed, consisting of an audio cassette-
recording device which could be programmed to record 6 time intervals (Sangean
VersaCorder®, C.Crane Co.) and a powered Omni-directional boundary microphone
(Model 330-3020, RadioShack Co.) (See chapter 3 & 4). The recording unit enabled 6
samples of environmental acoustics to measure temporal changes in a place. These initial

acoustic surveys were conducted using the cassette tape recorder system, but the

70

disadvantage was that it required many hours to manually digitize the tapes and then
archive these recordings into an acoustic digital library. Moreover, it required enormous

time to hear and identify bird species in the recordings.

Simultaneous observations using wireless acoustic sensor networks can
revolutionize the ability to investigate the acoustic variability and diversity in complex
ecosystems over a long-term period with minimal human effort and reduced processing
duration. The "Clickable Ecosystem" developed by Stuart Gage and his research team in
the Remote Environmental Assessment Laboratory (REAL) at Michigan State University,
provided an online system to enable the public to hear sounds, visualize acoustic patterns,
and interpret acoustic information for a specific ecosystem selected by users in near-real

time (http://www.real.msu.edu).

This chapter describes the structure, deployment and fimctions of a wireless
acoustic sensor network system. It also provides a case study of how the data can be
remotely transferred from remote locations so observations can be analyzed and

visualized within the remote research laboratory.

Design of an ecological wireless sensor network system

The overall goal of developing ecological wireless sensor networks is not only to
monitor and collect ecological observations in near-real time, but to remotely manage and
control the sensor system to minimize the loss of data. This development has coevolved
with new advances in sensor technology and wireless network systems, and led to the

development of an Ecological Wireless Sensor Network System (EWSNS) (Gage et al. in

71

submission). EWSNS comprises of three main components: a Habitat Sensor Platform

(HSP), a Habitat Server System (H88), and a Remote Server System (RSS).

Habitat Sensor Platform (HSP): The acoustic habitat sensor platform was
designed and developed by Gage, Biswas and 100 and their research teams, based on the
Crossbow Stargate processing board (Gage et al. 2006). The Stargate processing board
contains a 400 MHz Intel PXA225 processor and 64MB of RAM, and operates using
TinyOS (a Linux-based operating system) (Crossbow 2006). The Stargate processing
board requires relatively low power to operate (about 2.5 W). The main components of
HSP are the Stargate processing board, an acoustic sensor and a web camera (Logitech
pro 4000), a 2 GB flash card for local storage, a wireless communication card (802.11b),
a power converter from 12 V input to 5 V output, and a weatherproof case (Figure 5. 1).

Power is supplied via a 12 V deep cycle battery charged using a 18 W solar panel.

Habitat Sensor Platform

PmSuppty .' "

 

Figure 5. 1. Habitat sensor platform (HSP) hardware configuration.

72

Habitat Sensor Server (HSS): The main functions of the habitat sensor server are
to monitor and manage an array of HSPs deployed in the field, and to transmit the sensor
observations collected from HSPs into the regional server system. The HSS consists of a
laptop operated by a Linux system. Access to habitat sensor platforms in the field is a
critical function of the HSS. To achieve this, a habitat sensor management tool was
developed to enable communication and manipulation of all HSPs. Once the sensor
management application using a web-based program (a Perl CGI script) was embedded
into each HSP, the HSP can be remotely invoked from the habitat server for sensor
specific system configuration and programming applications (Gage et al. 2005) (See
Figure 5. 2). This web based management application regulates all configurations of the
habitat sensor platform including user login, setting time of day, sensor sample frequency,
sensor parameters, sensor addition, file location on the habitat server, and process
activities log. Note that since the HSP management application is developed using a
web-based program, the service cannot only be accessed from the habitat server itself, but

also from a regional server connected to the habitat server via the lntemet.

73

 

 

 

 

 

. ' " K88 Main Site Gu Shed
~ .. K38 “an Sac (On Short) ' - :- ' auxin “a l
. . , » .5 -- “’-
._ . 1‘ “mum ,‘ f 5;.“
‘ I
fl 1 ' 'A __,,. “‘ ’ ‘ w“ — " — ‘——* '
r" ‘ ’ . "' ‘ ‘
in. 1 Stargate 5004 Setting: H... ‘ Sensor Platform. $604
' ""7 l H ,. -- _ __._, -_
r ‘ _ ' ‘ in” 1'" run. 0'. ”*0.- Dou a? Acoustics [.mv)
Hm...- WQM mm ‘ ' -. .. .- _ - ._. Dmd I“ JWJ‘ 2007
mummm-umuwguau -- 7w . Gem «2:.
fl‘- r ~\ 79'» Ut"1m»F~'lfl‘.~ ,- ,-.r ' ,’*.I i "' tr -. 1.1-
- . be". an ‘ ' ’ - r ’
_t _ .1 ‘ _-..,_-. ,1.
13-"- ‘Al‘ ,_. . .- ‘ M, . if
_ T1”. ._ ‘ m ., .., , . .. .'
- '\ Intern-aunts '. _ _g -- .r v w an .1 .«
I
_W a. lull-- r , - . . s t» f -
m .(' '5'
‘r' ‘. 'V- bro-la. ~-~ -- -- i i ‘( ~~ L‘ -: '1 -‘ ~
HI. _ _‘ . 1
__ r u.. ‘5 m : ‘ , ' 1‘ .'
, Inuit-uh. m . ‘ , T!
I .. - , ..
‘ Inna—00mm .‘. .' fl
1 blah.- "'u "
{ . '/' .i
. ' cwa-u-hi-~nvouuuv .4 “.4” «via
I : what—Hmrmwo-w '.' "" ‘ "9' “V ‘ 7""‘ "-‘""J
. ‘ h”.- j it , '.- i.‘ 'w.‘ 9
. 'J. 7‘1, r 3 9
m . :1 , i ‘ u
______ _._______ _ ‘_-_. - I .I l 1 . \J
~-v '1‘ - ~' *1 .- r x i', i W- ' c
.. -. . 7:7a , - .213: ___..__ .. - " inn: - rec-c
"' ——“.—*——‘——'" ' " - .. " 3; hf.) \‘ ‘1: A ”Jr'q
. v). 1- -V u-u nn.4.n- .' 1--~.\,».-I ~-~ ,I. a" ‘
‘ . HH‘.___TW 7 -_ '~ ‘ . . I ' _ i _. . _. ___—J : n ‘3‘.) »r a]; flu 154‘; 'J
.1“;.~a . -.-‘n A p --. I-a r-.a.<l’)‘ ~.a.“.. 2h .7. . ‘p. ." ..
. _. 4..- F. . . . . _ . ___._._.___., _I ulr . _ - 4n)” .. \ r-vn. .3
‘ '1. -‘~V '4 u 'u o . v..-_¢- n. ,n 0.“. ' .I. '. I idl q I")
f I-OI" V" "I ‘l‘,|uts"‘ o-p- v'J-ru .n _I. l ‘ I)” --,J
_ >l' __A—‘r'vuni I-o- '4." >a- .‘ r _M—- _ 7‘ ' " ' “1 ”‘0‘: ‘ 'I'u}
‘ Hr- ———‘ -I\ win ri-jr-wu x. ' a“ "I ”'0’; 3"“ ‘9‘.th
1 . ___ ,______—_4L .‘_ I _____ _ -w _d " u" v 3. .1 -.’9
| ,_:_, " " -___ __ __ ___— _‘ 'ul . . ’- “"W'I‘U ,‘-\- Mm) y)
it "Qr'u‘u ‘- u.

 

 

 

 

 

Figure 5. 2. Screen shots of the sensor management application developed with a web-based
program (left) and near real-time sensor observations from an acoustic and image sensor from the
KBS-LTER site (right).

Remote Sensor Server (RSS): Each HSS transmits the sensor data set to a digital
archive of observations which reside on the RSS via wireless, broadband, satellite, or
other means of communication. The main goal of the RSS is to store, manage, access,
analyze, integrate, and distribute sensor observations from numerous arrays of the HSPs.
The RSS contains the three main fimctions including: 1) Digital sensor data library, 2)
Scientific query interface, and 3) Analytical processing tools (Gage et al. in submission).
The sensor observations and the associated metadata transmitted from the arrays of the
HPPs are deposited into the digital library. In addition, the digital library has a capacity
to store the archive of the various types of ancillary observations (satellite imagery,
meteorological and biogeochemical measures) from the ecosystems where arrays of the
HPSs are deployed. The scientific query interface provides tools to access the database
of the sensor observations and the associated metadata, and to select the sensor data set

based on the researcher’s interests, such as sensor types, locations and the time of the

74

observation. Lastly, the analytical processing tool is employed, based on the researcher’s
queries, and provides results which can be downloaded as data tables for statistical
analyses or can be visualized using graphic tools. The integration of the three main
functions in the RSS thus provides researchers with near-real time sensor observations

and their analytical products at appropriate spatial and temporal scales.

Case study

Deployment of Habitfiensor/Server System

To understand the spatial and temporal patterns of environmental acoustics in
various ecosystems based on the advanced habitat sensor system, 12 Habitat Sensor
Platforms and one Habitat Server System were deployed within the Long Term
Ecological Research (LTER) sites at W.K. Kellogg Biological Station (KBS) during June
2007. In addition, the study investigated the stability and efficiency of the Habitat

Sensor/Server System, and the integrity and quality of the sensor observations.

The 12 locations in the agricultural ecosystems at the LTER site were classified
into four different treatments: successional (T7, #4), poplar (T5, #3), and two wheat plots
with different treatments (no tillage (T2, #2) and conventional tillage(Tl, #1)). Each
treatment consists of three replicates and one Habitat Sensor Platform was deployed in
each replicate (Figure 5. 3). The acoustic sensor platform was programmed to collect
environmental acoustic signals for 30 seconds every 30 minutes during the study period.
The size of each sound file is 1.38 MB and 48 sound samples were recorded per day.
Since 12 habitat sensor platforms collected and transmitted sound files for 30 days in

June, the number of possible samples is 17,280 sound files which amounts to 22.46 GB.

75

Despite the integrity of habitat sensor system, the data set was incomplete due to severe
weather and a power failure. However, the Habitat Sensor System operated at 83 % of
capacity and successfully completed the data collection and transmission to a remote

server during the study period.

Deployment of the Habitat Sensor System (HSS) at KBS revealed two main
constraints: providing sustained power to the Habitat Sensor Platform (HSP) and the
distance limitation of the sensor platforms due to wireless transmission. The Habitat
Sensor Platform used at the KBS-LTER site was able to run for 14 days using one 12 V
deep-cycle battery charged by one solar panel (18 watts), in typical Michigan summer
weather. An improved Habitat Sensor Platform by Gage and his research team is able to
continuously collect and transmit sensor observations for more than 30 days using a 12 V

battery charged by two solar panels (36 watts) (Gage et al. in submission).

The HSS deployed at the KBS—LTER site supported wireless communication
from the sensor platforms to a habitat server within the radius of 90 meters, causing us to
increase the number of sensor platforms to cover the intended sampling area. To enhance
the sensor observation communication system, we developed a “Wireless Bridge System”
which relays sensor data from wireless hotspot to the wireless clouds of HSS (Figure 5.
4). The Wireless Bridge System consists of two antennas and one wireless access point. It
was noted that when connected with omni-directional antennas, the Wireless Bridge
System extended the communication to 235 meters with about 90% of wireless signal

strength.

76

1. Habitat types
" l1Wheat with till
.. 2.Wheat with no-till '
‘~ 3. Poplar
4. Successional
~ ‘ Habitat server

,T Habitat sensor
I Wireless bridge

 

Figure 5. 3. a) Map of the distribution of Habitat Sensor Platforms and Habitat Server at KBS-
LTER site (left). The Habitat Sensor Platform b) hardware components including a Crossbow
Stargate processor, 12 to 5v power converter, an acoustic sensor and a web camera, wireless
network card (802.1 lb), and a 1 GB Compact Flash storage device; and 0) field configuration

consisting of Habitat Sensor Platform, solar panel, and 12v deep cycle battery.

77

_¢_

HabitatServer
I Long distance I l

5 ------.LJ

At,
\« - ___,
:- , Wireless Bridge System Accesspomt

Habitat SensorPlatforms

Habitat Sensor Platforms

Figure 5. 4. Diagram of the advanced wireless Habitat Sensor Network System using a wireless

bridge system.

Analyses and interpretzflion of sensor obserwions

The acoustic data collected by the sensor arrays in field were transmitted into a
digital library embedded in the remote server system via the Internet. The data were
processed by quantifying the spectral energy in one-KHZ frequency intervals (See chapter
2). The numerical values resulting from this process provides a quantitative measure of

how the acoustic energy is distributed across the acoustic spectrum.

To quantify environmental acoustics, the Acoustic Habitat Quality Index (AHQI)
was computed to describe the temporal and spatial patterns of environmental acoustic
signatures. The Acoustic Habitat Quality Index (AHQI) provides a method of

classification of a place relative to biological sounds (biophony) and mechanical sounds

78

(anthrophony), based on the amount of acoustic energy in different spectrum frequency

intervals (See Chapter Two & Three for more details).

Acoustic Habitat Quality Index (AHQI) values varied with different habitat types
(Figure 5. 5). The poplar plots had significantly higher values of AHQI (0.78 i 0.006),
and the no—till wheat plots had the lower values of AHQI (0.45 i 0.008) (p < 0.001) in
June 2007. There was no significant difference between the successional plots and the
conventional-till wheat treatment. All the habitats measured in the KBS-LTER during
the sample period had the positive AHQI values, indicating that every habitat measured
in the KBS-LTER site exhibited dominance of biological sounds (biophony) during the

acoustic sampling period.

“:81

 

 

 

 

 

           

  

 

 

 

 

 

 

 

 

 

 

H i l' ’ ——.—
2' 044: ”'
< I ; 1
03 ‘1 ‘-¢-(.
0.21
l
0.1 —I
00 I ."-A t‘» .~ '- é' ‘ u‘ .u I ' all in:
Poplar Succession Wheat_noTIll Wheat_til|
Habitat

Figure 5. 5. The mean Acoustic Habitat Quality Indices at 4 difi'erent habitat types in the KBS-
LTER site. The bars represent mean + SE (n=1 1,901).

79

The diurnal patterns of AHQI at each sensor location in the 4 habitat types were
also investigated (See Figure 5. 6). The highest AHQI values commonly occurred at all
habitats between 0530 and 0600 hrs in June, 2007. The Poplar habitats had high positive
values of AHQI during most of the day and decreased between 2200 and 0430 hrs. On
the other hand, positive values of AHQI in successional habitats and in two different
wheat plots occurred during early morning (530 to 700 hrs) and in the early evening
(1900 to 2100 hrs) during June. The high positive AHQI values in Poplar habitats
accounted for more than 70 % of the time, whereas the positive AHQI values in
successional habitats and two different wheat plots occurred less than 30 % of the time.
The patterns show that frequent events of human acoustic disturbance are associated with
the negative AHQI values in the wheat plots and the successional plots at mid-day,

whereas biological vocalizations are dominant in the poplar plots from dawn to dusk.

80

 

 

.6
U'l

.0
en

 

 

 

 

 

 

Mean AHQI
o

 

 

 

 

 

 

 

 

 

 

 

 

 

T2r3

   

 

 

 

 

 

 

 

 

 

 

 

T1rS

d) T1r2

     

 

 

 

 

 

 

0 |000 2000
Time (hrs)
Figure 5. 6. The diurnal patterns of environmental acoustics based on computation of AHQI in 4

different habitats including a) poplar, b) successional, c) wheat with no-till treatment, and d)

wheat with till treatment.

81

Conclusions

This component of the research has presented the concept, design, and
development of wireless sensor networks for measuring and monitoring environmental
acoustics. In addition, the model of the sensor network system, Ecological Wireless
Sensor Network System (EWSNS) was introduced, and the specific components and
functions of EWSNS were described. One representative case study, conducted at the
KBS-LTER site, demonstrated the capabilities of wireless network systems for
automatically collecting, transmitting, and analyzing a large number of sound samples
from multiple points with a predetermined time schedule. Moreover, a simple analysis of
data interpreted from the sensor system clearly exhibited the temporal and spatial

variability of Acoustic Habitat Quality Indices in different ecosystems.

Although this study showed how EWSNS was able to be deployed to monitor the
dynamics of environmental sounds, we identified ways to enhance the sensor system by
solving some technical challenges including: power management, sensor network
topology and scalability, and wireless network capacity (Estrin et al. 2003, Martinez et al.
2004, Porter et al. 2005). Electric power management for habitat sensor platforms can be
one of the main constraints to sustain the operation of habitat sensor platforms over a
long term period (Estrin et al. 2003, Porter et al. 2005). In our study, the habitat sensor
platforms took advantage of solar energy and rechargeable batteries to extend power to
the system. We still need to explore using a variety of new energy sources available in
the environment such as wind, water, hydrogen, etc. (Biagioni and Bridges 2002). The

efficiency of the sensor power consumption can be also increased by using low-power

82

sensor processors or by programming to keep the operation of processors minimized

using a ‘hibernation’ mode (Kasten, personal communication).

Although there are some challenges to be overcome in developing sensor network
systems, using wireless sensor networks as a new sampling tool in ecology and
environmental science will provide tremendous opportunities to measure and monitor
complex ecological variables at relevant spatial and temporal scales, thus leading to
forecasting changes in ecosystems and even to addressing some of the “Grand

challenges” at a global scale (National Research Council 2001, Porter et al. 2005).

83

Chapter Six
Summary and Conclusions

The main goals of this research were to investigate whether environmental
acoustics can be used as an ecological attribute to indicate the current state of varying
ecosystems and to use acoustics as a key means to measure and monitor the biodiversity
and distribution of vocal species. The study focused on four main research projects to
accomplish the objectives: 1) development of analytical methods to understand acoustic
properties; 2) investigation of a new method to survey avian species using acoustic
recordings; 3) characterization of urban-rural gradient using environmental sounds; and

4) development of an automated acoustic sensor observation system.

Overall, this research has provided clear evidence that interpretation of
environmental acoustics has tremendous potential to measure and interpret changes in
ecosystems in space and time. By measuring and analyzing environmental sounds, key
ecological information could be extracted from the measurement of the soundscape
including; Acoustic Habitat Quality Index (Chapter 2 and 4); quantifying vocalizing
species diversity measures (Chapter 2, 3, and 4); and analyzing the degree of human
acoustical activities in various ecosystems (Chapter 2 and 4). Moreover, a wireless
sensor network system enabled automated monitor of patterns and changes in

environmental sounds at great spatial and temporal grain and extent (Chapter 5).

Based on a large number of acoustic observations, analyses of acoustic variables
from environmental sounds may indicate the current states of ecosystem structure and

processes. For instance, 14 grass-woodland sites in Australia were assessed using the

84

habitat quality index developed by the Victoria Department of Sustainability and
Environment (Straker and Lowe 2004). In addition, simultaneous acoustic recordings
were made at the same 14 sites in the morning and evening on November 30, 2006.
Preliminary results from the habitat assessment and acoustic recordings in Australia
showed a positive relationship between Acoustic Habitat Quality Index and the
ecological attributes determined by a habitat quality assessment across the sites (Gage et
al. in submission) (Figure 6. 1). The results motivated me to test a hypothesis that there
would be a relationship between acoustic attributes and ecological variables in various
ecosystems in a future study. Moreover, the relationship .will enable the development of
an ‘lndex of Acoustic lntegrity’ (IAI) to assess the current quality of terrestrial

ecosystems by measuring environmental acoustics.

I95

9)

Habitat Quality Index

43.5 7 " oIo 0T5 ”Wife
Acoustic Habitat Quality Index

Figure 6. l. A relationship between Acoustic Habitat Quality Index and Habitat Quality Index at
14 grass-woodland sites in Australia on November 30, 2006 (permitted by Gage et al.).

85

Although environmental acoustics is still a new research area to ecologists and
environmental scientists, I suggest that environmental acoustics will offer exciting
research opportunities in conjunction with development of wireless sensor networks in
ecological sensing (Porter et al. 2005, Gage et al. in submission). In addition, to better
understand the characteristics of environmental acoustics and their associated ecosystem
structure and processes including human systems, acoustic research will require the
multi-disciplinary science integration, including ecologists, acoustic engineers, computer
scientists, statisticians, and sociologists. The integration of acoustic research in the multi
science communities will enable an enhancement of our understanding and our ability to
forecast changes in complex, interconnected ecosystems at scales ranging from the

ecosystem to global level.

86

Appendices

Appendix 2. l. The diversity of bird species identified from the automated recordings from five

 

 

 

 

 

 

ecosystems.
LTER Sites List of avian Species identified
Agriculture
Song Sparrow *, Field Sparrow, Savanna Sparrow, American Crow,
Alfalfa 1 Northern Flicker, American Robin, Killdeer, Cedar Waxwing, American
Goldfinch
Alfalfa 3 Song Sparrow *, American Robin, Savanna Sparrow, American Crow,
Red-winged Blackbird
Indigo Bunting *, American Goldfinch, Northern Cardinal, American
Po lar l Crow, Yellow-billed Cuckoo, Killdeer, Common Yellowthroat, Tufted
p Titmouse, European Starling, Red—winged Blackbird, Song Sparrow,
American Robin
Song Sparrow *, Cedar Waxwing, Black-capped Chickadee, Northern
Poplar 2 Cardinal, Red-winged Blackbird, Chipping Sparrow, Indigo Bunting,

American Robin, American Goldfinch

 

Coniferous forests

 

Coniferous l

Red-wing Blackbird *, Northern Cardinal, Warbling Vireo, American
Robin, Rose-breasted Grosbeak, Baltimore Oriole, Black-capped
Chickadee, Tree Swallow, Tufted Titmouse, Mourning Dove, Eastern
Wood-pewee, Canada Goose, Gray Catbird, European Starling

 

Coniferous 2

Tufted Titmouse *, Hairy Woodpecker, Eastern Kingbird, Chipping
Sparrow, Black-capped Chickadee, Indigo Bunting, Blue Jay, American
Crow, Eastern Wood-pewee, Downy Woodpecker, Rose-breasted
Grosbeak, Red-winged Blackbird, American Robin, Northern Cardinal

 

Coniferous 3

Northern Cardinal *, Chipping Sparrow, Black—capped Chickadee,
Indigo Bunting, American Crow, Downy Woodpecker, American Robin

 

87

Appendix 2.1. Continued

 

Deciduous forests

 

Deciduous l

Scarlet Tanager *, Red-eyed Vireo, White-breasted Nuthatch, Red-bellied
Woodpecker, Northern Cardinal, Hairy Woodpecker, Blue Jay, American
Crow, Tufted Titmouse, Eastern Wood-pewee, American Robin

 

Deciduous 2

Baltimore Oriole *, Great Crested Flycatcher, Common Yellowthroat,
Red-bellied Woodpecker, Scarlet Tanager, Northern Cardinal, Black-
capped Chickadee, Blue Jay, American Crow, Tufted Titmouse, Eastern
Wood-pewee, Canada Goose, European Starling, Rose-breasted Grosbeak,
Red-winged Blackbird, Song Sparrow, Northern Flicker, American Robin,
Eastern Towhee,

 

Deciduous 3

Eastern Wood-pewee "‘, Brown Thrasher, Field Sparrow, Baltimore
Oriole, Great Crested Flycatcher, Red-eyed Vireo, White-breasted
Nuthatch, Red-bellied Woodpecker, Scarlet Tanager, Northern Cardinal,
Black-capped Chickadee, Indigo Bunting, Blue Jay, American Crow,
Tufted Titmouse, Eastern Wood-pewee, Canada Goose, European Starling,
Downy Woodpecker, Red-breasted Grosbeak, Red-winged Blackbird,
Northern Flicker, American Robin, American Goldfinch, Eastern Towhee

 

Succession

 

Succession I

Song Sparrow *, American Crow, Black-capped Chickadee, European
Starling, Baltimore Oriole, Eastern Wood—pewee, Northern Flicker, Indigo
Bunting, Scarlet Tanager, Red-winged Blackbird, Field Sparrow, House
Wren, American Crow, Chipping Sparrow, Northern Cardinal, Eastern
Towhee, Canada Goose, American Robin, Yellow Warbler, Rose-breasted
Grosbeak, Song Sparrow, American Goldfinch, Tufted Titmouse, Eastern
Bluebird

 

Succession 2

Brown Thrasher *, Indigo Bunting, Northern Flicker, Eastern Wood-
pewee, Gray Catbird, Eastern Towhee, Downy Woodpecker, Rose-breasted
Grosbeak, Northern Cardinal, Black-capped Chickadee, Blue Jay,
American Robin, American Goldfinch, Cedar Waxwing, Great Crested
Flycatcher, American Crow

 

Succession 3

Northern Cardinal *, Northern Flicker, Downy Woodpecker, Scarlet
Tanager, Yellow-billed Cuckoo, Baltimore Oriole, Song Sparrow, Black-
capped Chickadee, Field Sparrow, Eastern Towhee, Veery, Northern
Cardinal, American Robin, Rose-breasted Grosbeak, American Crow, Blue
Jay, Tufted Titmouse, Brown Thrasher

 

: the dominant bird species determined by abundance of bird vocalizations

88

Appendix 3. 1. List of bird species identified by all survey types in Essex Township. Note that
each species include its scientific name, AOU code, and the number of encounters in each survey

type.

 

Common name Scientific name Code PCS MAS AAS
Acadian Flycatcher Empidonax virescens ACFL 1 0 0
Alder Flycatcher Empidonax alnorum ALFL 1 0 0
American Crow Corvus brachyrhynchos AMCR 6 7 9
American Goldfinch Carduelis tristis AMGO 4 6 9
American Redstart Selophaga ruticilla AM RE 4 3 2
American Robin Turdus migratorius AMRO 14 16 16
Baltimore Oriole Icterus galbula B AOR 7 5 5
Bank Swallow Riparia riparia BANS 1 0 1
Black-capped Chickadee Poecile arrr‘capilla BCCH 3 2 2
Blue Jay Cyanocitta cristata BL J A 3 3 7
Blue-gray Gnatcatcher Polioptila caerulea ‘ IBch 2 0 3
Blue-winged Warbler Vermivora pinus BWWA | 0 0
Bobolink Dolichonyx oryzivorus 130130 1 1 2
Brown Thrasher Toxostoma rufum BRTH 4 2 2
Brown-headed Cowbird Molothrus ater BHCO 0 1 2
Cedar Waxwing Bombycilla cedrorum CEWA 2 1 2
Chipping Sparrow Spizella passerina CH SP 5 2 6
Common Grackle Quiscalus quiscula COGR 4 o 0
Common Yellowthroat Georhlypis trichas COYE 4 5 2
Dickcissel Spiza americana DICK 3 3 3
Downy Woodpecker Picoides pubescens DOWO 2 1 2
Eastern Bluebird Sialia sialis E ABL 0 0 1
Eastern Kingbird Tyrannus tyrannus E AKI 1 4 1
Eastern Meadlowlark Stumella magna E AME 7 0 4
Eastern Phoebe Sayornis phoebe E APH 1 0 0
Eastern Towhee Pipilo erythrophthalmus E ATO 2 1 2
Eastern Wood-Pewee Comopus virens [3wa 8 6 6
European Starling Sturnus vulgaris EUST 4 0 3
Field Sparrow Spizella pusilla 1:151) 10 6 7
Grasshopper Sparrow Ammodramus savannarum GRSP 2 1 1
Gray Catbird Dumetella carolinensis GRC A 6 1 3
Great crested Flycatcher Myr'archus crinitus GCFL 2 1 10
Green Heron Butorides virescens GRHE 1 0 0

 

89

Appendix 3.]. Continued

 

 

Common name Scientific name Code PCS MAS AAS
Homed Lark Eremophila alpestris HOL A 1 1 2
House Finch Carpodacus mexicanus HOFI 1 1 0
House Sparrow Passer domesticus lnosp 0 0 4
House Wren Troglodytes aedon HOWR 5 2 5
Indigo Bunting Passerina cyanea [NBU 15 8 5
Killdeer Charadrius vociferus KILL 3 5 2
Mourning Dove Zenaida macroura MODO 0 2
Northern Flicker Colaptes aurarus NOFL 2 1 1
Nothem Cardinal Cardinalis cardinalr's NOC A 11 10 10
Pileated Woodpecker Dryocopus pileatus PIWO 2 0 0
Prothonotary Warbler Protonotaria citrea pr A 1 1 1
Red-eyed Vireo Vireo olivaceus REV] 3 0 2
Red-winged Blackbird Agelaius phoeniceus RWBL 1 1 10 13
Ring-necked Pheasant Phasianus colehicus RNEP 5 3 2
Rose-breasted Grosbeak Pheucticus ludovicianus RBGR 4 2 2
Ruby-throated Hummingbird Archilochus colubris RTHU 0 0 1
Sandhill Crane Grus canadensis SACR 0 1 0
Savannah Sparrow Passerculus sandwichensis SAVS 4 2 2
Scarlet Tanager Piranga olivacea SCTA 3 5 5
Song Sparrow Melospiza melodia SOSP 15 18 15
Tree Swallow Tachycineta bicolor TRSW 1 0 1
Tufted Titmouse Baeolophus bicolor TUTI 1 l 4
Vesper Sparrow Pooecetes gramineus V ESP 2 1 O
Warbling Vireo Vireo gilvus WAVI 6 6 6
White-breasted Nuthatch Silta carolinensis WBNU 3 3 4
Wild Turkey Meleagris gallopavo wrru 1 0 0
Willow Flycatcher Empidonax trail/ii WIFL 4 2 2
Wood Thrush Hylocichla mustelina WOTH 1 0 1
Yellow Warbler Dendror'ca petechia yw AR 12 1 1 7
Yellow-billed Cuckoo Coccyzus americanus YBCU 5 2 0
Total No. of species 64 59 45 52
Total No. of points 25 25 20

 

90

Appendix 3. 2. List of bird species identified by all survey types in Westphalia Township. Note
that each species include its scientific name, AOU code, and the number of encounters in each

survey type.

 

Common name Scientific name Code PCS MAS AAS
American Crow Corvus brachyrhynchos AMCR 5 7 6
American Goldfinch Carduelis trislis AMGO 3 5 16
American Redstart Selaphaga ruticilla AMRE l I 0
American Robin Turdus migratorius AMRO I7 12 19
Bank Swallow Riparia riparia BANS 3 l 0
Baltimore Oriole [clerus galbula BAOR 2 3 I
Black-capped Chickadee Poecile atricapilla BCCH I I 3
Brown-headed Cowbird Molorhrus ater BHCO I 0 2
Blue Jay C yanocilta cristata BLJA I I 2
Bobolink Dolichonyx oryzivorus 8030 2 l I
Brown Thrasher Toxostoma rufum BRTH 3 3 4
Canada Goose Branta canadensis CANG I 0 I
Cedar Waxwing Bombycilla cedrorum CEWA l l 0
Chipping Sparrow Spizella passerina CHSP 12 8 3
Common Grackle Quiscalus quiscula COGR l l 0
Common Yellowthroat Geothlypis lrr'chas COYE l3 9 9
Downy Woodpecker Picoides pubescens DOWO 3 I 0
Eastern Kingbird Tyrannus tyrannus EAKI 3 2 3
Eastern Meadowlark Sturnella magna EAME 2 2 2
Eastern Phoebe Sayorm's phoebe EAPH l 0 0
Eastern Towhee Pipilo erythrophlhalmus EATO 0 0 2
European Starling Stumus vulgaris EUST 4 3 2
Eastern Wood-Pewee Contopus virens EWPE 8 5 4
Field Sparrow Spizella pusilla FISP 1 2 4
Great crested Flycatcher Myiarchus crim'ms GCFL S 2 2
Gray Catbird Dumetella carolinensis GRCA 4 3 4
House Finch Carpodacus mexicanus HOF I 0 l 0
Hairy Woodpecker Picoides villosus HAWO 0 0 l
Horned Lark Eremophila alpestris HOLA 3 2 8
House Sparrow Passer domesticus HOSP 4 4 I 1
House Wren Troglodytes aedon HOWR 5 2 2
Indigo Bunting Passerina cyanea INBU 6 3 5
Killdeer C haradrius vociferus KILL 4 5 9

91

Appendix 3.2. Continued

 

 

Common name Scientific name Code PCS MAS AAS
Mourning Dove Zenaida macroura MODO IO 5 5
Nothem Cardinal Cardinalis cardinalis NOCA 9 5 I 1
Northern Flicker Colaptes auratus NOF L l 2
Pileated Woodpecker Dryocopus pileatus PIWO l 0 0
Rose-breasted Grosbeak Pheucticus ludovicianus RBGR l l 3
Red-bellied Woodpecker Melanerpes carolinus RBWO l 0 l
Red-eyed Vireo Vireo olivaceus REVI l l 4
Ring-necked Pheasant Phasianus colchicus RNEP 10 7 6
Red-tailed Hawk Buteojamaicensis RTHA l 0 0
Red-winged Blackbird Agelaius phoem'ceus RWBL I8 19 18
Sandhill Crane Grus canadensis SACR l 5 3
Savannah Sparrow Passerculus sandwichensis SAVS 6 7 l
Scarlet Tanager Piranga olivacea SCTA 0 0 I
Sedge Wren C islothorus platensis SEWR l l 1
Song Sparrow Melospiza melodia SOSP 20 22 21
Tree Swallow Tachycineta bicolor TRSW l 0 I
Tufted Titmouse Baeolophus bicolor TUTI I l 2
Vesper Sparrow Pooecetes gramineus VESP 3 2 2
Warbling Vireo Vireo gilvus WAVI 3 3 6
White-breasted Nuthatch Sitta carolinensis WBNU l 0 4
Willow Flycatcher Empidonax traillii WIFL 4 3 2
Wild Turkey Meleagris gallopavo WITU l 0 0
Wood Thrush Hylocichla mustelina WOTH 4 l 3
Yellow-billed Cuckoo C occyzus americanus YBCU 0 O I
Yellow Warbler Dendroica petechia YEWA 6 6 4
Yellow-throated Vireo Vireo flavijrons YTVI 0 0 1
Total No. of species 60 55 45 48
Total No. of points 25 25 24

 

92

Appendix 3.3. List of bird species with the number of encounters identified by Automated
Acoustic Surveys (AAS) in Essex Township. Note that each species includes the number of
encounters at each recording sample from 5:00 am to 10 am every hour. Of 6 total recordings at
each point, the number of species occurrence was calculated.

Species 5:00 AM 6:00 AM 7:00 AM 8:00 AM 9:00 AM 10:00 AM Frequency

 

AMGO 1 3 4 5 7 7 6
AMRO 9 9 5 6 7 6 6
BLJA 1 1 5 l 2 1 6
EWPE I 4 2 3 4 1 6
FISP 2 3 4 5 2 4 6
HOWR 1 3 4 3 2 1 6
NOCA 4 5 5 5 4 3 6
RWBL 1 9 5 7 7 5 6
YEW A l 3 4 2 4 3 6
AMRE l I 1 1 2 5
BAOR l 2 2 2 2 5
8080 1 2 I 1 l 5
DICK 2 l 2 l 2 5
GCFL 1 6 3 1 5
INBU 4 2 3 1 3 5
PROW 1 1 1 I 1 5
RBGR 1 l 2 1 1 5
SCTA 3 4 2 2 1 5
SOSP 10 9 l 1 8 7 5
TUTI 2 1 I 2 2 5
WAVI 4 3 1 2 2 5
AMCR 6 2 3 2 4
CHSP 3 2 3 2 4
HOSP 1 1 1 1 4
REV! 2 1 1 l 4
WBNU 1 2 2 1 4
BCCH 1 l l 3

 

93

Appendix 3.3. Continued

 

Species 5:00 AM 6:00 AM 7:00 AM 8:00 AM 9:00 AM 10:00 AM Frequency

 

BRTH 1 2 1 3
COYE 1 1 1
EAME l 2 3

BUST 1 1 1
GRCA 2

RNGP 2 1 1
SAVS 2

“an. 1 1 2
‘WOTH 1 1 1
BHCO 1 1

CEWA l 1
DOWO 1 1
EATO 1 1
HOLA 1 1

NNNNNNMWMWWMWW

MODO 1 1
BANS 1 1
EABL l 1
GRSP 1 1
HAWO 1 1
Knt 2 1
NOFL 1 1
RTHU 1 1
TRSW 1 1

 

94

Appendix 3. 3. List of bird species with the number of encounters identified by Automated
Acoustic Surveys (AAS) in Westphalia Township. Note that each species includes the number of
encounters at each recording sample from 5:00 am to 10 am every hour. Of 6 total recordings at
each point, the number of species occurrence was calculated.

Species 5:00 AM 6:00 AM 7:00 AM 8:00 AM 9:00 AM 10:00 AM Frequency

 

 

EAKI 1 2 2 l l l 6
HOSP l 5 7 5 5 3 6
AMGO 2 1 5 6 8 5 6
NOCA 2 3 6 l 2 3 6
COYE 2 7 4 1 1 3 6
KILL 4 l 3 1 1 2 6
sosp 4 I7 1 0 1 3 12 l 3 6
RWBL 5 14 1 l 12 l 1 13 6
AMRO 10 1 1 10 12 8 6 6
CHSP I l 2 1 1 5
B080 1 1 l l l 5
INBU l 3 2 2 l 5
REV! 2 2 l 1 2 5
EWPE 2 3 1 1 1 5
YEWA 2 3 2 2 2 5
FISP 3 2 l l 2 5
WAVI 3 5 4 2 5 5
HOLA 3 2 2 2 4
GCFL 1 1 l 1 4
MODO 2 1 1 1 4
SAVS 1 I 1 l 4
BLJA l l 1 3
TRSW 1 1 1 3
VESP 1 1 1 3
BCCH l 1 2 3
EAME 1 2 3

95

Appendix 3.4. Continued
Species 5:00 AM 6:00 AM 7:00 AM 8:00 AM 9:00 AM 10:00 AM Frequency

 

worn 2 2 3 3
RBGR 2 1 I 3
RNEP 3 4 2 3
AMCR 4 2 l 3
BRTH 1 2 1 3
GRCA 3 2 1 3
BHCO 1 1 2
TUTI l 1 2
WBNU 2 2 2
EUST 1 I 1 2
YTVl l 1 2
NOFL l l 2
EATO 1 l 2
van 2 1
CANG l l
BAOR 1 1
SCTA 1 1
SEWR 1 1
SACR 3 1
HAWO 1 1
RBWK) 1 1
YBCU 1 1

 

96

Appendix 4. 1. Common and scientific name and AOU code of bird species identified from
acoustic recordings (see Figure 8).

 

 

Species Code Common Name Scientific Name
AMCR American Crow Corvus brachyrhynchos
AMGO American Goldfinch Carduelis tristis
AMRO American Robin Turdus migratorius
BAOR Baltimore Oriole Icterus galbula
BCCH Black-capped Chickadee Poecile atricapillus
BHCO Brown-headed Cowbird Molothrus ater
BLJA Blue Jay C yanocitta cristata
C ANG Canadian Goose Branta canadensis
CEWA Cedar Waxwing Vombycilla cedrorum
CHSP Chipping Sparrow Spizella passerina
COYE Common Yellowthroat Geothlypis trichas
DOWO Downy Woodpecker Picoides pubescens
EUST European Starling Sturnus vulgaris
GCF L Great Crested Flycatcher Myiarchus crim'tus
GRCA Gray Catbird Dumetella carolinensis
HOF I House Finch Carpodacus mexicarms
HOLA Horned Lark Eremophila alpestris
HOSP House Sparrow Passer domestic-us
HOWR House Wren Troglodytes aedon
KILL Killdeer Charadrius vociferus
MODO Mourning Dove Zenaida macroura
NOCA Northern Cardinal Cardinalis cardinalis
REVI Red-eyed Vireo Vireo olivaceus
RWBL Red-winged Blackbird Agelaius phoeniceus
SOSP Song Sparrow Melospiza melodia
TUTI Tufted Titmouse Baeolophus bicolor
WIF L Willow Flycatcher Empidonax traillii
YEWA Yellow Warbler Dendroica petechia

 

97

Bibliography

Acevedo, M. A. and L. J. Villanueva-Rivera. 2006. Using Automated Digital Recording
Systems as Effective Tools for the Monitoring of Birds and Amphibians. Wildlife
Society Bulletin 34:21 1—214.

Allen, T., S. L. Finkbeiner, and D. H. Johnson. 2004. Comparison of Detection Rates of
Breeding Marsh Birds in Passive and Playback Surveys at Lacreek National
Wildlife Refuge, South Dakota. Waterbirds 27:277-281.

Asano, F., M. Goto, K. Itou, and H. Asoh. 2001. Real-time Sound Source Localization

and Separation System and Its Application to Automatic Speech Recognition.
ISCA.

Bart, J. 2005. Monitoring the Abundance of Bird Populations. The Auk 122: l 5-25p.

Batten, L. A. 1972. Breeding bird species diversity in relation to increasing urbanisation.
Bird Study 19:157-166.

Beissinger, S. R. and D. R. Osborne. 1982. Effects of Urbanization on Avian Community
Organization. Condor 84:75-83.

Biagioni, E. and K. Bridges. 2002. The Application of Remote Sensor Technology to
Assist The Recovery of Rare And Endangered Species. International Journal of
High Performance Computing Applications 16:315.

Blair, R. B. 1996. Land use and Avian species diversity along an urban gradient.
Ecological Applications 6:506-519.

Bridges, A. S. and M. E. Dorcas. 2000. Temporal Variation in Anuran Calling Behavior:
Implications for Surveys and Monitoring Programs. Copeia 2000:587-592.

Brumm, H. 2004. The impact of environmental noise on song amplitude in a territorial
bird. Journal of Animal Ecology 73zpp. 434-440.

Brumm, H. and D. Todt. 2002. Noise-dependent song amplitude regualtion in a territorial
songbird. Animal Behaviour 63:891-897.

Butler, D. 2004. Sound and vision. NATURE 427:480-481.

Butler, R., M. Servilla, S. Gage, J. Basney, V. Welch, B. Baker, T. F Ieury, P. Duda, D.
Gehrig, M. Bletzinger, J. Tao, and D. M. Freemon. 2006. Cyberinfrastructure for
the analysis of ecological acoustic sensor data: a use case study in grid

development. Challenges in Large Applications in Distributed Environments.
IEEE:25-33.

98

Bystrak, D. 1981. The North American breeding bird survey. Studies in Avian biology
6:34-41.

Carson, R. 1962. Silent Spring. Houghton Miffiin, New York.

Catchpole, C. and P. J. B. Slater. 1995. Bird Song: Biological Themes and Variations.
Cambridge University Press.

Chesmore, D. 2004. Automated bioacoustic identification of species. Anais da Academia
Brasileira de Ci?cias 76:436-440.

Chesmore, D. 2007. The Automated Identification of Taxa: Concepts and Applications.
CRC press, Boca Raton, FL.

Chesmore, E. D. and C. Nellenbach. 1997. ACOUSTIC METHODS FOR THE
AUTOMATED DETECTION AND IDENTIFICATION OF INSECTS. Pages
223-231. ISHS.

Chesmore, E. D. and E. Ohya. 2007. Automated identification of field-recorded songs of
four British grasshoppers using bioacoustic signal recognition. Bulletin of
Entomological Research 94:319-330.

Cody, M. 1985. Habitat selection in grassland and open-country birds. PHYSIOL. ECOL.
SER.: 191-226.

Collins, S. L., L. M. A. Bettencourt, A. Hagberg, R. F. Brown, D. 1. Moore, G. Bonito, K.
A. Delin, S. P. Jackson, D. W. Johnson, and S. C. Burleigh. 2006. New
opportunities in ecological sensing using wireless sensor networks. Frontiers in
Ecology and the Environment 4:402-407.

Conway, C. J. and J. P. Gibbs. 2005. EFFECTIVENESS OF CALL-BROADCAST
SURVEYS FOR MONITORING MARSH BIRDS. The Auk 122:26-35.

Crossbow. 2006. Stargate developer's guide. Crossbow Technology, Inc., San Jose, CA.

Degraaf, R. M. and J. M. Wentworth. 1986. Avian guild structure and habitat association
in suburban bird communities. Urban Ecology 9:399-412.

Dooling, R. J. 1982. Auditory perception in birds. Pages 95-130 in D. E. Kroodsma and E.
H. Miller, editors. Acoustic Communication in Birds.

Estrin, D., W. Michener, and G. Bonito. 2003. Environmental Cyberinfrasturcutre Needs
for Distributed Sensor Networks. Scrippts Institute of Oceanography.

Fay, R. R. and L. A. Wilber. 1989. Hearing in Vertebrates: A Psychophysics Databook.
The Journal of the Acoustical Society of America 86:2044.

99

Feng, A. S., P. M. Narins, C. H. Xu, W. Y. Lin, Z. L. Yu, Q. Qiu, Z. M. Xu, and .1. X.
Shen. 2006. Ultrasonic communication in frogs. NATURE 440:333-336.

Forkenbrock, D. J. and L. A. Schweitzer. 1999. Environmental justice in transportation
planning. Journal of the American Planning Association 65:96-111.

Gage, S., W. Joo, E. Kasten, F. Yu, T. Wu, and S. Biswas. in submission. Design,
deployment and applications of a wireless sensor network for monitoring
environmental acoustics. BioScience.

Gage, S., P. Ummadi, A. Shortridge, J. Qi, and P. K. Jella. 2004. Using GIS to develop a
network of acoustic environmental sensors.in ESRI International User Conference
2004, San Diego, CA.

Gage, S. H., S. Biswas, W. Joo, T. Wu, and F. Yu. 2006. A sensor platform and local
sensor server for capture andtransmission of acoustics and ancillary sensor
observations toremote servers using wireless technology.in Long Term Ecological
Research All Scientist Meeting at W. K. Kellogg Biological Station East Lansing,
Michigan.

Gage, S. H., B. M. Napoletano, and M. C. Cooper. 2001. Assessment of ecosystem
biodiversity by acoustic diversity indices. Page 2430 in 14lst Meeting Acoustical
Society of America. American Institute of Physics, Palmer House Hilton Hotel;
Chicago, Illinois.

Gage, S. H., B. Subir, W. Joo, and F. Foust. 2005. Operation Manual: Habitat Sensor
System. Computational Ecology and Visualization Laboratory, Michigan State
University, East Lansing, Michigan.

Gage, S. H., Umadi, P.,Shortridge, A., QI, J., Kumar, J. 2004. Using GIS to Develop a
Network of Acoustic Environmental Sensors. ESRI International Users
Conference 1688.

Gates, J. E. and L. W. Gysel. I978. Avian nest dispersion and fledging success in field-
forest ecotones. Ecology 59:871-883.

Graber, R. R. and J. W. Graber. 1963. A comparative study of bird populations in Illinois,
1906-1909 and 1956-1958. Illinois Natural History Survey Bulletin 28:377-527.

Green, R. J. 1984. Native and exotic birds in a suburban habitat. Australian Wildlife
Research 11:181-190.

Guthrie, D. A. 1974. Suburban bird opulations in southern California. American Mildland
Naturalist 92:461-466.

Hamilton, M. P., E. A. Graham, P. W. Rundel, M. F. Allen, W. Kaiser, M. H. Hansen,
and D. L. Estrin. 2007. New Approaches in Embedded Networked Sensing for

100

Terrestrial Ecological Observatories. Environmental Engineering Science 24:192-
204.

Hartmann, W. M. 1998. Signals, Sound, and Sensation. Springer-Verlag, New York.

Haselmayer, J. and J. S. Quinn. 2000. A Comparison of Point Counts and Sound
Recording As Bird Survey Methods in Amazonian Southeast Peru. The Condor
102:887-893p.

Herkert, J. R. 1995. An Analysis of Midwestern Breeding Bird Population Trends: 1966-
1993. The American Midland Natualist 134:41-50p.

Hobson, K. A., R. S. Rempel, H. Greenwood, B. Turnbull, and S. L. V. Wilgenburg.
2002. Acoustic surveys of birds using electronic recordings: new potential from
an omnidirectional microphone system. Wildlife Society Bulletin 30:709-720.

Hopp, S. L., M. J. Owren, and C. S. Evans. 1998. Animal Acoustic Communication:
Sound Analysis and Research Methods. Springer-Verlag Berlin Heidelberg,
Germany.

Hutto, R. L., S. M. Pletschet, and P. Hendricks. 1986. A fixed-radius point count method
for nonbreeding and breeding season use. Auk 103:593-602.

Johnson, E. A., G. I. Fryer, and M. J. Heathcott. 1990. The influence of man and climate
on the fire frequency of the Interior Wet Belt forest, British Columbia. Journal of
Ecology 78:403-412.

Juang, R, H. Oki, Y. Wang, M. Martonosi, L. S. Peh, and D. Rubenstein. 2002. Energy-
Efficient Computing for Wildlife Tracking: Design Tradeoffs and Early
Experiences with ZebraNet. Pages 96-107 in Proceedings of the 10 th
international conference on Architectural support for programming languages and
operating systems, San Jose, USA.

Kalamazoo Nature Center. 2002. Handbook of Michigan breeding bird atlas II.
Kalamazoo Nature Center, Kalamazoo, MI.

Kasten, E. P., P. K. Mckinley, and S. H. Gage. 2007. Automated ensemble extraction and
analysis of acoustic data stream.in Proceedings of the 131 International Workshop
on Distributed Event processing, Systems and Applications (DEPSA), Toronto,
Ontario.

Katti, M. and P. S. Warren. 2004. Tits, noise, and urban bioacoustics. Trends in Ecology
and Evolution 19:109-1 10.

Kendell, K., Fish, D. Wildlife, Alberta, R. Status, A. Branch, and A. Alberta
Conservation. 2002. Survey Protocol for the Northern Leopard Frog. Fish &
Wildlife Division, Resource Status and Assessment Branch, Alberta Sustainable
Resource Development.

101

Krause, B. 1998. Into a Wild Sanctuary: A Life in Music and Natural Sound. Heyday
Books, Berkeley, California.

Krause, B. L. 1997. The Niche Hypothesis: Creature Vocalizations and the Relationship
between Natural Sound and Music. Nature Sounds Society: 1-5.

Krebs, C. J. 1998. Ecological methodology. Addison-Welsey Longman, Inc., New York,
New York.

Kroodsma, D. E. and E. H. Miller, editors. 1996. Ecology and Evolution of Acoustic
Communication in Birds. Comstock Publishing Associates, Ithaca, New York.

Larkin, R. P., L. L. Pater, D. J. Tazik, and L. Construction Engineering Research. 1996.
Effects of Military Noise on Wildlife: A Literature Review. US Army Corps of
Engineers, Construction Engineering Research Laboratories; National Technical
Information Service, distributor.

Magurran, A. E. 1988. Ecological diversity and its measurement. Princeton University
Press, Princeton, New Jersey.

Mainwaring, A., D. Culler, J. Polastre, R. Szewczyk, and J. Anderson. 2002. Wireless
sensor networks for habitat monitoring. Pages 88-97 in Proceedings of the lst

ACM international workshop on Wireless sensor networks and applications.
ACM New York, NY, USA.

Martinez, K., J. K. Hart, and R. Ong. 2004. Environmental sensor networks. Computer
37:50-56.

Mathworks. 2005. Power spectral density (PSD) using Welch's method.in Mathworks,
editor.

Mills, H. 1995. Automatic detection and classification of nocturnal migrant bird calls.
The Journal of the Acoustical Society of America 97 :3370.

Myneni, R. B., F. G. Hall, P. J. Sellers, and A. L. Marshak. 1995. The interpretation of
spectral vegation indexes. IEEE Transactions on Geosicence and Remote Sensing
33:481-486.

Napoletano, B. M. 2004. Measurement, Quantification and Interpretation of Acoustic
Signals within an Ecological Context. Michigan State University, East Lansing.

Narins, P. M. 1995. Frog Communication Scientific American 273:6 Page (3).

National Research Council. 2001. Grand Challenges in Environmental Sciences.
Committee on Grand Challenges in Environmental Sciences

Oversight Commission for the Committee on Grand Challenges in Environmental
Sciences, Washington, DC.

102

O'Connor, R. J ., E. Dunn, D. H. Johnson, S. L. Jones, D. Petit, C. R. Smith, and E.
Welling. 2000. A Programmatic Review of the North American Breeding Bird
Survey: Report of a Peer Review Panel. [Online] Available at w“ “.1an-
ureusgsgiLov/bbs/intlexlitml.

 

Otsuki, M., A. Kimura, T. Nishiura, F. Shibata, and H. Tamura. 2007. RealSound
Interaction: A Novel Interaction Method with Mixed Reality Space by Localizing
Sound Events in Real World. LECTURE NOTES IN COMPUTER SCIENCE
4551:653.

Ouis, D. 2001. Annoyance from road traffic noise: a review. Journal of Environmental
Psychology 21:101-120.

Palmer, M. W. 1993. Putting things in even better order: the advantages of canonical
correspondence analysis in ecology. Journal of Ecology 65:85-94.

Parker III, T. A. 1991. On the Use of Tape Recorders in avifauna] surveys. Auk 108:443-
444p.

Patuxent Wildlife Research Center. 2001. North American Breeding Bird Survey. U. S.
Geological Survey and Canadian Wildlife Service, [Online] Available at
http:waxxvarc.usgsgoflbbsfl

 

Peterjohn, B. and K. Pardieck. 2002. A bibliography for the North American breeding
bird survey.in USGS Patuxent Wildlife Research Center, editor. US. Geological
Survey, Laurel, Maryland.

Porter, J ., P. Arzberger, P. Hanson, T. Kratz, S. Gage, T. Williams, S. Shapiro, P. Bryant,
F.-P. Lin, H.-b. King, T. Hansen, H.-W. Braun, and W. Michener. 2005. Wireless
Sensor Networks for Ecology. BioScience 55:p.561-572.

Qi, J., S. H. Gage, W. 100, B. Napoletano, and S. Biswas. 2008. Soundscape
characteristics of an environment: A new ecological indicator of ecosystem health.
Pages pp.201—211 in J. W, editor. Wetland and water resource modeling and
assessment. CRC press, New York, New York.

R Development Core Team. 2007. R: A language and environment for statistical
computing. , Vienna, Austria. ISBN 3-900051-07-0, URL http://wuwR-
projectorg. R Foundation for Statistical Computing, Vienna, Austria.

Rabin, L. A., B. McCowan, S. L. Hooper, and D. H. Owings. 2003. Anthropogenic Noise
and its Effect on Animal Communication: An Interface Between Comparative

Psychology and Conservation Biology. International Journal of Comparative
Psychology 16:172-192.

Ralph, C. J ., G. R. Geupel, P. Pyle, T. E. Martin, and D. F. DeSante. 1993. Handbook of
field methods for monitoring landbirds. General Technical Report PSW-GTR-l44,

103

Pacific Southwest Research Station, Forest Service, US. Dept. of Agriculture,
Albany, CA.

Ralph, C. J ., J. R. Sauer, and S. Droege. 1995. Monitoring bird populations by points
counts. USDA, Albany.

Reijnen, R. and R. Foppen. 1994. The effects of car traffic on breeding bird populations
in woodland. 1. Evidence of reduced habitat quality for willow warblers
(Phylloscopus trochilus) breeding close to a highway. Journal of Applied Ecology
31 :85-94.

Richter, K. O. and A. L. Azous. 2001. C Amphibian Distribution, Abundance, and
Habitat. Wetlands and Urbanization Implications for the Future.

Robbins, C. S., D. Bystrak, and P. H. Geissler. I986. The Breeding Bird Survey: Its first
fifteen years, 1965-1979.

Robbins, D. S. 1979. Effects of forest fragmentation on bird populationsin R. M. Degraaf
and N. Tilghman, editors. Proceedings of the workshop on management of North-
Central and Northeastern forests for nongame birds. US. For. Serv. Gen. Tech.

Rep. NOS 1.

Robinson, S. K., F. R. T. III, T. M. Donovan, D. R. Whitehead, and H. Faaborg. 1995.
Regional Forest Fragmentation and the Nesting Success of Migratory Birds.
Science 267:1987—1990p.

Robinson, S. K. and D. S. Wilcove. 1994. Forest fragmentation in the temperate zone and
its effects on migratory songbirds. Bird Conservation International 42233-249.

Romer, K. and F. Mattem. 2004. The design space of wireless sensor networks. Wireless
Communications, IEEE [see also IEEE Personal Communications] 11:54-61.

Rosenstock, S. S., D. R. Anderson, K. M. Giesen, T. Leukering, and M. F. Carter. 2002.
Landbird counting techniques: current practices and an alternative. The Auk
119:46-53.

Rossing, T. D. 2007. Springer handbook of acoustics springer, New York, NY.

Rundus, A. S. and L. A. Hart. 2002. Overview: Animal Acoustic Communication and the
Role of the Physical Environment. Journal of Comparative Psychology 116:120-
122.

Ryan, M. J ., R. B. Cocroft, and W. Wilczynski. 1990. The role of environmental selection
in intraspecific divergence of mate recognition signals in the circket frog, Acirs
Crepitans. Evolution 44: 1 869-1 872.

Sauer, J. R., B. G. Peterjohn, and W. A. Link. 1994. Observer differences in the North
America breeding bird survey. The Auk 111:50-62. '

104

Schafer, R. M. 1977. The Soundscape: Our Sonic Environment and the Tuning of the
World. Destiny Books, Rochester, Vermont.

Schafer, R. M. 1994. The Soundscape: Our Sonic Environment and the Tuning of the
World. Destiny Books, Rochester, Vermont.

Schieck, J. 1997. Biased detection of bird vocalizations affects comparisons of bird
abundance among forested habitats. The Condor 99zpp. 179-190.

Shannon, C. E. and W. Weaver. 1963. The mathematical theory of communication.
University of Illinois Press, Urbana, Illinoise, USA.

Sharpe, R. 1886. The AOU Code and Check-List of American Birds. NATURE 34: 1 68-
170.

Slabbekoom, H. and M. Peet. 2003. Birds sing at a higher pitch in urban nise. NATURE
424:267.

Stevenson, R. J., B. C. Bailey, M. C. Harass, C. P. Hawkins, J. Alba-Tercedor, C. Couch,
S. Dyer, F. A. Fulk, J. M. Harrington, C. T. Hunsaker, and R. K. Johnson. 2004.
Designing data collection for ecological assessments. Pages Pgs 55-84 in M. T.
Barbour, S. B. Norton, H. R. Preston, and K. W. Thornton, editors. Ecological
Assessment of Aquatic Resources: Linking Science to Decision-Making. Society
of Environmental Toxicology and Chemistry, Pensacola, Florida.

Stokes, D. and L. Stokes. I996. Stokes filed guide to birds: eastern region. Little, Brown
and Compony, Boston, MA.

Straker, A. and K. W. Lowe. 2004. Native Biodiversity Resource Kit: Environmental
management in agriculture. Department of sustainablility and Environment,
Melbourne, Victoria, Australia.

Stutz, F. 1986. Environmental impact. Pages 329-343 in S. Hanson, editor. The
geography of urban transportation. The Guilford press, New York.

Sun, J. W. C. and P. M. Narins. 2005. Anthropogenic sounds differentially affect
amphibian call rate. Biological Conservation 121 :419-427.

Tempest, W. 1976. Infrasound and low frequency vibration. Academic Press, London.

ter Braak, C. J. F. 1986. Canonical Correspondence Analysis: a new eigenvector
technique for multivariate direct gradient analysis. Ecology 67zpp. 1167-1179.

ter Braak, C. J. F. and I. C. Prentice. 1988. A theory of gradient analysis. Advances in
theological Research 18:271-317.

The American Heritage Dictionary. 2000. The American heritage dictionary of the
English language. Houghton Mifflin, Boston.

105

Thompson, J. N. 2001. Frontiers of ecology. BioScience 51:15-24.

Thompson, W. L. 2002. Towards reliable bird surveys: Accounting for individuals
present but not detected. The Auk ll9:pp. 18-25.

Truax, B. 1984a. Acoustic communication. Ablex Publishing, New Jersey.
Truax, B. 1984b. Handbook for Acoustic Ecology. Cambridge Street Publishers.

Vitousek, P. M., H. A. Mooney, J. Lubchenco, and J. M. Melillo. 1997. Human
Domination of Earth's Ecosystems. Science 277:pp. 494 - 499.

Warren, P. S., M. Katti, M. Ermann, and A. Brazel. 2006. Urban bioacoustics: it's not just
noise. Animal Behaviour 71:491-502.

Weir, L. A. and M. J. Mossman. 2005. North American Amphibian Monitoring Program
(NAAMP). Pages pp. 307-313 in M. J. Lannoo, editor. Amphibian Declines:
Conservation Status of United States Species. Univ. of California Press, Berkeley.

Welch, P. 1967. The use of fast Fourier transform for the estimation of power spectra: A
method based on time averaging over short, modified periodograms. Audio and
Electroacoustics, IEEE Transactions on 15:70-73.

Whitcomb, R. F ., C. S. Robbins, J. F. Lynch, B. L. Whitcomb, M. K. Kliemkewicz, and
D. Bystrak. 1981. Effects for forest fragmentation on avifauna of the eastern
deciduous forest. Pages 125-205p in R. L. Burgess and D. M. Sharpe, editors.

Forest island dynamics in man-dominated landscapes. Springer-Verlag, New
York.

Wiley, R. H. and D. G. Richards. 1978. Physical constraints on acoustic communication
in the atmosphere: implications for the evolution of animal vocalizations.
Behavioral Ecology and Sociobiology 3:69-94.

Williams, B. K., J. D. Nichols, and J. C. Michael. 2002. Analysis and management of
animal populations. Academic Press.

Zimmerling, J. R. and C. D. Ankney. 2000. A technique that increases detectability of
passerine species during point counts. Journal of Field Ornithology 71:638-649.

106

G

 

“11111