THE EFFECT OF HUMAN INFLUENCE ON NORTH CENTRAL FLORIDA SOUNDSCAPES

By

JENET M. DOOLEY

A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2016

© 2016 Jenet M. Dooley

To my parents

ACKNOWLEDGMENTS

The completion of this dissertation would not have been possible without the support of many people. I thank my parents for their unwavering faith in my abilities. I thank David for his encouragement, patience, love, and enthusiasm for statistics. I also thank Erica, Amy, Torren, and Sean for field help, thoughtful advice, and friendship. I thank Bobby for being the best field friend I will ever have; his influence will be felt by many for a long time to come. I would also like to thank Midnight, my oldest and worst friend, for providing comical relief.

I thank my advisor, Mark Brown, and the members of my dissertation committee

Gary Siebein, Peter Frederick, and Barron Henderson for their thoughtful comments, time, and willingness to participate in a project that did not fall into a single traditional discipline.

4

TABLE OF CONTENTS

page

ACKNOWLEDGMENTS ...... 4

LIST OF TABLES ...... 8

LIST OF FIGURES ...... 11

LIST OF ABBREVIATIONS ...... 12

LIST OF ACRONYMS ...... 15

ABSTRACT ...... 16

CHAPTER

1 BACKGROUND ...... 18

Research Motivation ...... 18 Research Objectives ...... 20 Organization ...... 20 Plan of Study ...... 21 Chapter 2 ...... 21 Chapter 3 ...... 21 Chapter 4 ...... 22

2 NOISE AND LANDSCAPE DEVELOPMENT INTENSITY ...... 24

Introduction ...... 24 Soundscape Areal Dimensions ...... 25 Anthrophony vs. Biophony ...... 25 Landscape Development Intensity ...... 26 Problem Statement and Plan of Study ...... 26 Methods ...... 27 Description of Study Sites ...... 27 Data Collection ...... 27 Acoustic Analysis ...... 28 Acoustic metrics ...... 28 Analysis of total frequency spectrum ...... 28 Analysis of 1 kHz frequency bands ...... 29 Analysis of anthrophony vs. biophony ...... 29 Landscape Measures ...... 30 Landscape development intensity (LDI) ...... 30 Wetland indicator and road proximity ...... 31 Physical and Temporal Measures ...... 32 Statistical Analysis ...... 32

5

Results ...... 34 Variable Selection ...... 34 LDI index scale ...... 34 Physical and temporal control variables ...... 34 Wetland indicator and road proximity ...... 34 Models ...... 35 1000 Hertz Frequency Bands ...... 36 Anthrophony and Biophony ...... 37 Discussion ...... 38

3 LAND USE SOUNDSCAPE CHARACTERISTICS ...... 52

Introduction ...... 52 Habitat Specific Soundscapes ...... 52 Temporal Variation in Soundscapes ...... 53 Noise Thresholds ...... 53 Problem Statement ...... 54 Plan of Study ...... 54 Methods ...... 55 Site Descriptions ...... 55 30-minute sites ...... 55 48-hour sites ...... 56 Acoustic Sampling ...... 56 30-minute site data ...... 56 48-hour site data ...... 57 Data Analysis ...... 57 30-minute sites ...... 57 48-hour sites ...... 58 Results and Discussion...... 59 30-minute sites ...... 59 48-Hour Sites ...... 63 Comparison of 30-Minute and 48-Hour Sampling Timeframes ...... 65 Conclusions ...... 65

4 DETERMINING ZONES OF ACOUSTIC DISTURBANCE ADJACENT TO ROADWAYS ...... 77

Introduction ...... 77 Road Noise Response Thresholds ...... 78 Problem Statement and Plan of Study ...... 79 Methods ...... 80 Road Traffic Volumes ...... 80 Road Noise Sound Levels ...... 80 Theoretical roadways ...... 80 Case study ...... 81 Road Noise Dissipation ...... 81 Theoretical roadways ...... 81

6

Case study ...... 82 Zones Of Disturbance Calculation ...... 83 Average songbird communication zones of disturbance ...... 83 Recorded bird call zones of disturbance ...... 84 Grey Tree Frog communication zones of disturbance ...... 84 Results and Discussion...... 85 Road Noise Dissipation ...... 85 Bird Communication Zones Of Disturbance ...... 86 Grey Tree Frog Communication Zone Of Disturbance ...... 88 Case Study ...... 88 Observed sound levels ...... 88 SPreAD-GIS results ...... 89 Conclusion ...... 90

5 SUMMARY AND CONCLUSIONS ...... 99

Key Findings ...... 99 Implications and Future Research ...... 100

APPENDIX

A REGRESSION RESULTS ...... 104

B SPreAD-GIS RESULTS ...... 117

C SITE LOCATIONS ...... 118

LIST OF REFERENCES ...... 121

BIOGRAPHICAL SKETCH ...... 134

7

LIST OF TABLES

Table page

2-1 Acoustic metrics used to describe the soundscapes...... 43

2-2 Non-renewable areal empower intensities for LU/LC classes used to calculate the LDI index values ...... 44

2-3 Regression results comparing LDI scales to the inband power soundscape metric ...... 45

2-4 Regression results comparing LDI scales to the delta power soundscape metric ...... 45

2-5 Regression results comparing LDI scales to the aggregate entropy soundscape metric ...... 46

2-6 Regression results comparing LDI scales to the center frequency soundscape metric ...... 46

2-7 Regression results for total frequency spectrum models ...... 47

2-8 Summary statistics for LDI and soundscape metrics ...... 47

2-9 OLS regression results for 1 kHz band average PSD with site fixed effects ...... 48

2-10 Difference in means between the biophony and anthrophony average PSD and entropy values ...... 49

2-11 Regression results for anthrophony and biophony models ...... 50

3-1 Definitions of soundscape and acoustic metrics ...... 68

3-2 Difference in mean average PSD (dB) for each frequency band between CON and each of the other LU/LC classes ...... 71

3-3 OLS regression results for linear models that use LDI and LU/LC dependent variables to predict soundscape metrics ...... 73

3-4 P-values representing statistical significance of Wald Tests comparing inband power model LU/LC indicator coefficients ...... 73

3-5 Acoustic metrics for long term samples by site and season ...... 75

3-6 P-values for Mann Whitney U tests comparing the Leq sample distributions during the nine am hour to the other hourly Leq distributions during the day ...... 76

8

4-1 Traffic volume data used to model traffic noise for the three simulated roadways and observed traffic volume data for the case study interstate in 2015 ...... 93

4-2 The decrease in octave band sound levels that result from a 200-meter buffer of dense vegetation bordering a roadway ...... 95

4-3 The buffer distance requirements for an average bird to hear another average bird call with a dominant frequency within three octave bands for each of the simulated roadways ...... 95

4-4 Buffer distances converted from noise thresholds found in other studies for the three simulated roads ...... 95

4-5 The buffer distance requirements that ensure the Cope’s grey tree frog does not experience phonotaxis disruption as a result of traffic noise from each simulated roadway for the 1,000 and 2,000 Hz octave bands ...... 96

A-1 Regression results for models with inband power dependent variables and LDI100 and supplemental landscape independent variables ...... 104

A-2 Regression results for models with inband power dependent variables and LDI500 and supplemental landscape independent variables ...... 105

A-3 Regression results for models with inband power dependent variables and LDI1500 and supplemental landscape independent variables ...... 106

A-4 Regression results for models with delta power dependent variables and LDI100 and supplemental landscape independent variables ...... 107

A-5 Regression results for models with delta power dependent variables and LDI500 and supplemental landscape independent variables ...... 108

A-6 Regression results for models with delta power dependent variables and LDI1500 and supplemental landscape independent variables ...... 109

A-7 Regression results for models with center frequency dependent variables and LDI100 and supplemental landscape independent variables ...... 110

A-8 Regression results for models with center frequency dependent variables and LDI500 and supplemental landscape independent variables ...... 111

A-9 Regression results for models with center frequency dependent variables and LDI1500 and supplemental landscape independent variables ...... 112

A-10 Regression results for models with aggregate entropy dependent variables and LDI100 and supplemental landscape independent variables ...... 113

9

A-11 Regression results for models with aggregate entropy dependent variables and LDI500 and supplemental landscape independent variables ...... 114

A-12 Regression results for models with aggregate entropy dependent variables and LDI1500 and supplemental landscape independent variables ...... 115

A-13 Partial and Adjusted R Squared values for supplemental proxy variables in inband power, aggregate entropy, and delta power models estimated with LDI100 and LDI500...... 116

B-1 The 500 Hz band sound pressure levels (db) at six sites modeled by SPreAD-GIS for varying weather conditions ...... 117

C-1 Site type and GPS location for Chapter 2 and 3 data ...... 118

C-2 GPS coordinates for Chapter 3 transect points ...... 120

10

LIST OF FIGURES

Figure page

1-1 Systems diagram depicting the impact of noise from human land use on wildlife ...... 23

2-1 Location of the 67 sites where data was collected for this study...... 42

2-2 Visual representation of the compartmentalization of sound signals using Fourier Transforms...... 43

2-3 Marginal effect of human influence on average PSD across the frequency spectrum ...... 49

2-4 Examples of spectrograms from study sample covering ten minutes from four different sites...... 51

3-1 Ten-minute-long examples of spectrograms from each LU/LC class. From left to right, top to bottom ...... 69

3-2 Power spectrums for each LU/LC consisting of mean and standard deviation PSD values for each 1 kHz band between 20 and 15,000 Hz...... 70

3-3 Soundscape measure distributions grouped by site type...... 72

3-4 Time series of one minute LAeq sound levels over entire sampling period for each season and site. 55 dbA level is indicated with a black line ...... 74

4-1 Road noise sample locations for case study ...... 93

4-2 Octave band sound levels for 200,000 AADT simulated roadway at distances up to 5,000 meters from the road ...... 94

4-3 A-weighted sound levels for all three simulated roadways at distances up to 5,000 meters from the road ...... 94

4-4 Sound pressure levels measured within the case study area averaged for each octave band (500-8,000 Hz) and distance from the road ...... 96

4-5 1,000 Hz octave band sound levels in grey scale with the Cope’s grey tree frog (in blue) and average bird (in green) buffers depicting where the road noise levels reach the effect thresholds ...... 97

4-6 2,000 Hz octave band sound levels in grey scale with the Cope’s grey tree frog (in blue), Tufted Titmouse (in pink), and average bird (in green) buffers depicting where the road noise levels reach the effect thresholds...... 98

11

LIST OF ABBREVIATIONS

Avg Average

Ag Entropy Aggregate entropy

C Celsius

CA California

CBD Central business district

Cntr Center

Coeff Coefficient

CON Conservation

CSC Commercial shopping center dB Decibel dBA A-weighted decibel dmc Distance of maximum communication

DNL Day night level-equivalent sound level over 24 hours with 10 dB synthetic increase in sound level during night time hours of 10 pm to 7 am

EA Excess attenuation

EMPD Empower density

Freq Frequency

Ha Hectare

Hr Hour

Hz Hertz

Inbnd pwr Inband power

IND Industry kHz Kilohertz

12

L10 Sound level just exceeded for 10 percent of the time in the sampling time frame

L90 Sound level just exceeded for 90 percent of the time in the sampling time frame

LAeq A-weighted equivalent sound level

LAeq-day A-weighted equivalent sound level for the hours of 7am to 10pm

LAeq_n A-weighted equivalent sound level for n hours

LAeq_night A-weighted equivalent sound level for the hours of 10pm to 7am

LCEQ C-weighted equivalent sound level

LDI100 Landscape Development Intensity index for a 100-meter radius circular area

LDI500 Landscape Development Intensity index for a 500-meter radius circular area

LDI1500 Landscape Development Intensity index for a 1500-meter radius circular area

Leq Equivalent sound level

Leq_n Equivalent sound level for the n time frame

Leq_night Equivalent sound level for the hours of 10pm to 7am

Log Logarithm

LU/LC Land use/land cover

M Meter

Min Minimum

MU Mixed use

Prox Proximity

Pwr Power

Rd Road

Ref Reference

13

R-sqrd R-squared

RP Recreational park

Sel Selection

Sej Solar emjoules

SFR Single family residential

Std dev Standard deviation

Std Err Standard error

WAV Waveform audio file format

14

LIST OF ACRONYMS

AADT Average annual daily traffic

AIC Akaike information criterion

DEM Digital elevation model

EPA Environmental Protection Agency

FDOT Florida Department of Transportation

FGDL Florida Geographic Data Library

FHWA Federal Highway Administration

GIS Geographical Information System

GPS Global Positioning System

LDI Landscape development intensity

MPH Miles per hour

NPS National Park Service

OLS Ordinary Least Squares

PSD Power spectral density

REZ Road effect zone

SPL Sound pressure level

TNM Traffic Noise Model

USA United States of America

US EPA United States Environmental Protection Agency

WHO World Health Organization

15

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

THE EFFECT OF HUMAN INFLUENCE ON NORTH CENTRAL FLORIDA SOUNDSCAPES

By

Jenet M. Dooley

December 2016

Chair: Mark T. Brown Major: Environmental Engineering Sciences

The increasing quantity and intensity of human generated noise within landscapes has stimulated interest, in recent years, from the scientific community and the general public in the conflicts between humans and the natural soundscapes.

Although, it is wildly accepted that wildlife is subjected to detrimental human noise within urban landscapes, little is known about how intensity and the category of land use affect sounds. In addition, the most recent knowledge that has been gained about wildlife thresholds for anthropogenic noise have yet to be employed in conservation management applications. Accordingly, the objectives of this research were to advance the understanding of anthropogenic sound in the landscape and how it affects wildlife by: (1) describing soundscapes across a spectrum of human influence utilizing a metric that described the intensity of energy use in the surrounding landscape; (2) determining how specific land use/land cover (LU/LC) classes contribute to soundscapes; and (3) calculating zones of disturbance that encompass the area not acoustically suitable for wildlife communication through the application of known noise thresholds in the literature, field data, and modeled traffic noise. The results of objective 1 showed that soundscape characteristics could be predicted by the landscape development intensity

16

within a 100 and 500-meter radius surrounding the point of recording. The results also confirmed that human influence results in noise below 2 kHz but indicated that it should not be assumed that higher frequencies are not occupied by anthropogenic noise. The results for objective 2 established that human dominated LU/LC classes have unique soundscape patterns and sound sources that implied different impacts on surrounding natural soundscapes. Particularly, the industrial and central business district sites had significantly more sound power that was unique from each other, the other human land use classes, and the effect of general land use development intensity. The calculation of the acoustic zones of disturbance surrounding multiple simulated roadways indicated that birds were susceptible to detrimental noise at further distances than the anuran species analyzed, up to 288 meters from an interstate highway.

17

CHAPTER 1 BACKGROUND

Research Motivation

Sound is produced by a diverse array of sources within the human domain (WHO

1999). It spreads outward from energy intensive land uses, introducing background noise disturbance into neighboring areas (Miller 2002). In 1981, the US Environmental

Protection Agency (US EPA) estimated that almost half of the population was exposed to harmful levels of noise from traffic alone (Simpson and Bruce 1981). Like humans, the wellbeing of animals is compromised by anthropogenic noise (Slabbekoorn and

Ripmeester 2008; Barber et al. 2010; Kight and Swaddle 2011). Figure 1-1 depicts the detrimental interactions of noise from the human landscape with animal function. For example, noise disrupts acoustic communication that is crucial for animal reproductive success, survivorship, and recruitment for many species (Brumm and Slabbekoorn

2005). In addition, quiet soundscapes have social value for humans (Jensen and

Thompson 2004; NPS 2006) and ecological value for wildlife (Dumyahn and Pijanowski

2011). Currently, wildlife is facing huge losses of biodiversity in the face of human development (Vitousek et al. 1997; Chapin et al. 2000; McShane et al. 2010) and noise is a potent disturbance that propagates vast distances from its original source.

There is substantial and growing evidence that animals are detrimentally affected by human noise. Amphibians and birds are known to alter the timing, intensity and frequency of their acoustic signals in response to human noise stimulus in a natural environment (Brumm 2004; Sun and Narins 2005; Halfwerk and Slabbekoorn 2009;

Parris et al. 2009; McLaughlin and Kunc 2013). Further, adjusted calls could impact mate selection by females and alter population dynamics (Slabbekoorn and Ripmeester

18

2008; Parris et al. 2009). Animals that cannot modify their calls or are sensitive to increased stress from noise (Ruiz et al. 2002; Campo et al. 2005; Bonier et al. 2007;

Chloupeck et al. 2009) might abandon noisy areas (Hu and Cardoso 2009). Recently, experimental studies that controlled for confounding variables that usually accompany noise disturbance (visual stimulus, edge effects, etc.) have shown significant effects on bird population structure, reduced nest species richness, reduced density in bird communities, and reduced pairing success (Habib et al. 2007; Bayne et al. 2008;

Francis et al. 2009; Blickley et al. 2012). Ware et al. (2015) created road noise conditions in a migratory bird area that resulted in a decreased capture rate of 31% and reduced body condition of the birds. This abundant research indicates that animals are detrimentally affected by human noise and a better understanding of the acoustic conditions that human activity creates is warranted.

The conservation of natural or quiet soundscapes is motivated by the adverse effects of noise on humans and animals but practical efforts to mitigate noise at a landscape scale are lacking. In addition, soundscapes themselves have been recognized as important sensory experiences that should be conserved for the enjoyment of the public (NPS 2006, Dumyahn and Pijanowski 2011). Noise mitigation has largely been motivated by adverse effects of noise on humans (Miller 2002, Blickley and Patricelli 2010) and usually target one specific sound source such as a road or industrial plant (Dumyahn and Pijanowski 2011). At the landscape scale, natural soundscapes are surrounded by a heterogeneous patchwork of human land uses, each with unique acoustic characteristics. The resulting noise disturbance on nearby biologic communities has not been adequately characterized. There are calls for research

19

linking the landscape and soundscape (Pijanowski et al. 2011a) but few attempts thus far (Bormpoudakis et al. 2013; Fuller et al. 2015). Therefore, documenting the human soundscape is fundamental to acoustic-related conservation efforts.

Soundscape is the term for all of the sounds in a location. Schafer (1977) created this originally human-centric theory as a holistic alternative to piece-meal analysis of individual sound sources. Recently, work in many fields including landscape ecology

(Pijanowski et al. 2011a), conservation biology (Farina et al. 2011b), and bioacoustics

(Farina et al. 2011a) have used soundscapes to analyze acoustic signals from other perspectives. This dissertation utilizes the top down approach of soundscape theory to describe the relation between the landscape and soundscape and analyze acoustic confrontations that arise from the heterogeneous human and natural landscape.

Research Objectives

The overall objective of this research is to advance the understanding of anthropogenic sound in the landscape and how it effects wildlife by: (1) describing soundscapes across a spectrum of human influence utilizing a metric that describes the intensity of energy use in the landscape; (2) determining if specific land use/land cover

(LU/LC) classes contribute unique sounds to soundscapes; and (3) estimating the zone bordering busy roads that encompasses the area not acoustically suitable for wildlife communication through the application of known noise thresholds in the literature, field data, and modeled traffic noise.

Organization

This dissertation is a compilation of five chapters. Chapters 2-4 are papers that each address a goal of the overall dissertation research objectives. Chapter 1

20

introduces and motivates the research. Chapter 5 is the conclusion that synthesizes the results of the other chapters.

Plan of Study

Chapter 2

This chapter addresses how anthropogenic soundscapes change with overall human land use intensity. It uses the Landscape Development Intensity (LDI) index to establish a gradient of human influence on 67 soundscapes in north central Florida.

From this gradient, quantitative relations between LDI and the sound power of several regions of the frequency spectrum are established through multiple soundscape metrics. These relations address two contested values in the literature directly: the size of the landscape footprint that affects a soundscape and the frequency characteristics of anthropogenic noise.

Chapter 3

Chapter 3 builds on the relation between landscape development intensity and sound power established in Chapter 2. Land use/land cover (LU/LC) directly surrounding the place of recording is characterized for each soundscape and soundscapes are described and compared across LU/LC classes. LU/LC characterizations are added to the quantitative relations between LDI and the soundscape descriptions from Chapter 2. The implication of the temporal nature of soundscapes is addressed with time series sound level data from four sites. The timeframe of the LU/LC soundscapes is compared to the rest of the day and the samples are compared to noise levels known to cause harm to humans and animals.

21

Chapter 4

Chapter 4 calculates zones of ecological disturbance that result from road noise using current knowledge about the disruptions of noise on animal communication. Noise is modeled from three theoretical roadways. In addition, the noise from an interstate is modeled as it propagates into an adjacent conservation area. The modeled sound levels are compared to modeled bird calls, known traffic noise thresholds for an anuran, and bird calls recorded in the area to determine the point where the signals could be successfully received.

22

Figure 1-1. Systems diagram depicting the impact of noise from human land use on wildlife

23

CHAPTER 2 NOISE AND LANDSCAPE DEVELOPMENT INTENSITY

Introduction

The increasing quantity and intensity of human generated noise within landscapes has stimulated interest, in recent years, from the scientific community and the general public (Flatow 2011; Harris 2011; Krause 2013). This has led to a significant expansion in the measurement and characterization of what has been termed the soundscape, or the composite of sounds of an environment (Pijanowski et al. 2011b).

The analysis of soundscapes has been fast evolving, resulting in new perceptions of the interface between human and the natural soundscape and new tools for its study.

Based primarily on findings of acoustic signal alteration in birds and amphibians

(Brumm and Slabbekoorn 2005; Laiolo 2010) and evidence of altered reproductive behavior (Habib et al. 2007; Bayne et al. 2008; Ware et al. 2015), the anthropogenic portion of soundscapes is increasingly considered a disturbance. And while it may be intuitively obvious that noise intensity is related to land use intensity, few if any studies have documented the relation in a quantitative way.

Research documenting the relation between soundscape patterns and gradients of human development intensity has been limited. Two recent studies used the percent cover of specific human dominated land uses to measure human influence and found negative correlations with soundscape complexity (n=7) (Pijanowski et al. (2011a) and positive correlations with sound power below 2,000 hertz (Joo et al. 2011). Two earlier studies in western Greece found correlations between the human perception of the acoustic environment and quantitative descriptions of the surrounding human landscape

(Matsinos et al. 2008; Mazaris et al. 2009). Tucker et al. (2014) found that the

24

soundscape characteristics of forest patches in Queensland, Australia correlated with patch size and connectivity. In another study, Fuller et al. (2015) suggested that research linking the soundscape and landscape has been inhibited by the lack of landscape evaluation methods.

Soundscape Areal Dimensions

The physical area that contributes sound to a recorded soundscape is not easily defined and not standardized in the literature (Farina and Pieretti 2014). The physical area of influence for a recording is dependent on the location and loudness of transient sound signals and changing environmental conditions (Qi et al. 2008; Egan 2007).

Previous soundscape ecology studies have used circular areas with radii of 100 meters

(Pijanowksi et al. 2011a), 175 meters (Mazaris et al. 2009), 300 meters (Joo et al. 2011;

Depraetere et al. 2012), and 500 meters (Krause et al. 2011). In the literature, the size of the areas of influence used have not been compared or substantially validated

Anthrophony vs. Biophony

Studies of the soundscape commonly identify sources of sound as either anthropogenic or biologic using a sound frequency cutoff of 2 kilohertz (kHz), where sounds below 2 kHz are classified as anthropogenic (anthrophony) and above as biologic (biophony) (Joo 2009; Krauss et al. 2011; Joo et al. 2011). However, there is evidence that this boundary is somewhat arbitrary (Sueur et al. 2014). The 2 kHz boundary was based on observations of the frequency ranges of biological and anthropogenic sounds (Kasten 2012; Napoletano 2004; Joo et al. 2011). However, human and biological sounds have been documented crossing the 2 kHz point in the frequency spectrum (Napoletano 2004; Bee and Swanson 2007; Can et al. 2010;

Makarewicz and Sato 1996). It is important that the boundary is accurate, as it is used

25

in studies to infer the relative contribution of biophony and anthrophony to soundscapes

(Qi et al. 2008; Fuller et al. 2015; Gage and Axel 2014).

Landscape Development Intensity

This study uses a metric of land use intensity, called the Landscape

Development Intensity (LDI) index (Brown and Vivas 2005) to link the sound by- products of land uses to soundscapes. The LDI index is a quantitative measure of land use intensity based on non-renewable resource use per unit area per unit time. Further, it uses the concept of emergy (Odum 1996) to express all resource flows supporting land uses in common units of solar emergy. Expressed in this way, resource use per unit area per unit time (units = sej area-1 time-1) becomes aerial empower intensity

(where empower is emergy per time). The LDI index of a point or area takes into account the aerial empower intensity of surrounding land uses. The LDI index has been used as a human disturbance gradient to develop bio-indicators of ecosystem, stream and lake condition (Fore 2004; Fore 2005; Lane and Brown 2007; Reiss et al. 2010).

The LDI index correlated with wetland rapid assessment evaluation methods in Florida

(Brown and Vivas 2005), South Dakota (Bouchard 2009), Ohio (Mack 2006), and

Hawaii (Margriter et al. 2014). It also correlated with coral reef condition in St. Croix and wetland condition in Taiwan (Oliver et al. 2011; Chen and Lin 2011).

Problem Statement and Plan of Study

In this study, we evaluate several characteristics of soundscapes generated by landscapes of varying degrees of development intensity to produce quantitative associations between noise and land use intensity and spatial area. In addition, we explore the 2 kHz demarcation between anthrophony and biophony with statistical correlations between land use intensity and the sound power spectrum.

26

Methods

Description of Study Sites

This study collected data from 67 sites in north central Florida (Figure 2-1) selected to capture a variety of land use/land cover (LU/LC) classes across a spectrum of anthropogenic influence. Six different LU/LC classes were targeted as follows, conservation (CON; n=10), recreational park (RP; n=10), single family residential (SFR; n=16), industrial (IND; n=8), commercial shopping center (CSC; n=13), central business districts (CBD; n=4), and mixed uses (including roadsides and agriculture, MU; n=6).

Data Collection

The soundscapes of each site were sampled using a Fostex fr-2le field recorder and a Seinheisser ME 62 omni-directional microphone. The recordings were 30 minutes long and used a waveform format with a 24-bit depth and 48 kHz sampling rate. The recordings took place between 0900 and 1000 on week days over two years from April

2013 until March 2015. This time of day was selected for its ambient nature, avoiding known times of increased acoustic activity like early morning rush hour traffic. A tripod was used so that the recording equipment was always 1 meter off the ground.

Recordings were not taken if the wind in the area exceeded 10 mph or if it was raining.

At each site, data was gathered at the time of the recordings to document climatic conditions. Temperature and humidity is known to affect the attenuation of sound through the air (Embleton 1996). This information was taken from real time data available online from the closest weather station. Temperature and relative humidity were confirmed on site with a psychrometer. The location of the recording was documented with a handheld GPS unit.

27

Acoustic Analysis

The analysis of the acoustic data utilized different divisions of the frequency spectrum to describe the soundscapes. First, the total sampled frequency spectrum (20-

20,000 Hz) was analyzed as a whole. Second, each site’s spectrogram was divided into twenty, 1 kHz wide frequency bands. The first band ranges from 20-1000 Hz to accommodate for the frequency limits of the microphone used. Third, the two frequency regions attributed to anthropogenic and biologic sourced sounds (Anthrophony, 20-2000

Hz and biophony, 2000-8000 Hz) were analyzed.

Acoustic metrics

The metrics used in this study to describe the soundscapes were calculated from spectrograms made with Raven Pro 1.4 (Cornell University, Ithaca, New York) software.

The spectrograms used a Hann window with a discrete Fourier Transform size of 256 samples and 188 Hz grid size. The metrics were based on the power spectral density

(PSD), the amount of sound power per unit frequency (dB/Hz), of the frequency-time bins in the spectrogram, which was calculated internally within Raven using Fourier

Transforms (see Figure 2-2). Five different metrics were calculated using the PSD values from each recording: inband power, average PSD, delta power, aggregate entropy, and center frequency. A brief description of each measure is given in Table 2-

1. Aggregate Entropy is calculated as follows:

푓2 퐸푏푎푛푑 퐸푏푎푛푑 퐻푠푒푙푒푐푡푖표푛 = ∑ ( ×푙표푔2 ( )) (2-1) 퐸푠푒푙푒푐푡푖표푛 퐸푠푒푙푒푐푡푖표푛 푓=푓1

Analysis of total frequency spectrum

The analysis of the total frequency spectrum (20-20,000 Hz) applied the inband power, aggregate entropy, delta power, and center frequency measures. These

28

measures were selected because they were expected to indicate the presence of anthropogenic noise in the soundscape. Human noise was known to be powerful and concentrated in low bands (Warren et al. 2006; Slabbekoorn and Ripmeester 2008;

Barber et al. 2010; Pieretti and Farina 2013). When present, it was anticipated to overpower biologic noise and dominate the soundscape resulting in an increase in overall power, a spectrum with sound energy concentrated in low frequencies, and increased disparity between high and low frequencies. Consequently, human influence was expected to correlate with increased sound power and decreased aggregate entropy, delta power, and center frequency.

Analysis of 1 kHz frequency bands

The 1 kHz band analysis looked at the relation between human influence and sound in each frequency band. The average PSD measure was used instead of inband power to avoid bias because microphone specifications restricted the first band to 20-

1,000 Hz. The aggregate entropy, delta power, and center frequency measures were not used because, intuitively, they did not provide useful information when applied to the

1 kHz frequency range.

Analysis of anthrophony vs. biophony

The anthrophony (20-2,000 Hz) and biophony (2-8 kHz) analysis compared the two regions directly and looked at how human influence affected them separately. Each frequency region was described with the aggregate entropy and inband power measures. The anthrophony power was expected to be higher and more concentrated than the biophony. Human influence was expected to decrease the biologic presence in the soundscapes.

29

Landscape Measures

Landscape development intensity (LDI)

LDI measures human influence based on the amount of nonrenewable resource use in a determined area of influence surrounding the system (point) of interest. The surrounding areas of influence for this study were circular areas centered at the point where the microphone was placed during site recordings. Three different sized circular areas were used with 100, 500, and 1,500 meter radii to compute the LDI. Within each area of influence, the area of each unique LU/LC was calculated in ArcMap 10.3 (ESRI

2014), using the most recent land use and cover layers from the Florida Geographic

Data Library (FGDL 2015). LU/LC classes were confirmed from observations on site and discrepancies were updated before analysis. Aerial empower intensities for the

LU/LC classes in Table 2-2 were multiplied by the area of each type and summed to compute land use areal empower intensity. The LDI for each area of influence was computed using the equation presented in Reiss et al. (2010), an updated version of the original LDI equation from Brown and Vivas (2005) as follows:

푒푚푃퐷 퐿퐷퐼 = 10log⁡( 푇표푡푎푙) (2-2) 푒푚푃퐷푅푒푓

Where:

푒푚푃퐷푇표푡푎푙 = ∑푒푚푃퐷퐿푈/퐿퐶 + 푒푚푃퐷푅푒푓

−1⁡ −1 푒푚푃퐷푅푒푓 = 1.97⁡퐸 + 15⁡푠푒푗⁡ℎ푎 푦

The total areal empower density was calculated from the area weighted areal empower intensity of the LU/LC classes within the analyses area. In addition, a distance weighted LDI index was calculated for each landscape scale. For these index values, the aerial empower intensity of each LU/LC was area weighted and divided by the

30

distance (meter) from the center. This distance weighting reflected the inverse distance law for sound power decay (Harris 1991).

Wetland indicator and road proximity

Two additional measures were included in the analysis to capture the landscape variability. The LDI index accounts for all human and natural landscape features in its calculation. In addition to LDI, wetlands and road proximity were included separately to capture these sources of sound.

Wetland habitats support a large abundance of wildlife (Mitsch and Gosselink

2007), many of which use acoustic communication. A wetland indicator variable was used to specify sites near wetlands where this influence may have an effect on the soundscape. If there was a wetland system within 100 meters of the sampling location, the wetland indicator variable was given a value of one, otherwise the variable was zero. This determination was done post-sampling in ArcMap 10.3 (ESRI 2014) using the latest National Wetland Inventory shapefile from US Fish and Wildlife, accessed through the Florida Geographic Data Library (FGDL 2015).

Road noise is loud and can be heard over large distances. The literature indicated that it was an important part of human influenced soundscapes (Dooling and

Popper 2007; Kociolek et al. 2011; Nega et al. 2013). Roads were included in LDI calculations but the physical area they cover is relatively small. A separate variable that captured the proximity and intensity of roads was created to identify the exaggerated acoustic effect of roads. This measure was computed as the distance (meters) from each site to the nearest large scale road (collector or arterial) and the annual average daily traffic (AADT) on the road. These values were calculated from the AADT shapefile accessed through the Florida Geographic Digital Library (FGDL 2015). The variable was

31

calculated as the product of the AADT and the inverse distance (meter) of the closest road to each site.

1 푅표푎푑⁡푃푟표푥 = (퐴퐴퐷푇 ∗ ( )) (2-3) 푑푖푠푡푎푛푐푒⁡푡표⁡푟표푎푑⁡(푚)

Physical and Temporal Measures

Soundscapes have seasonal patterns (Schafer 1977; Truax 2001; Krause et al.

2011; Gage and Axel 2014). They are driven by temporal biologic trends such as foliage abscission and bird migration. Variables were included to capture the temporal variation present in the soundscapes. The season indicator included four categories defined by the month the site was sampled during. Winter sites were defined as sites sampled in

December, January, and February, spring as March through April, summer as June through August and fall as September through November. The temperature variable was the observed temperature in Celsius rounded to the nearest integer at the beginning of the sample. Relative humidity was rounded to two decimal places normalized between 0 and 1.

Statistical Analysis

The relations between the variables describing the landscape and the soundscape metrics were described with multiple ordinary least squares (OLS) linear regressions with robust standard errors using Stata SE13 software (Stata Corp, College

Station, Texas). Robust standard errors were used to control for heteroscedasticity and avoid the possibility of invalid statistical inferences (Kutner et al. 2004). Many executions of these models were compared to determine the best quantitative association between the landscape and the soundscapes. First, the soundscape metrics that measured over the total frequency spectrum (20-20,000 Hz) were used for variable

32

selection. The three LDI scales, the road proximity, wetland indicator, and climatic and temporal control variables were all considered as explanatory variables. Once the most appropriate set of explanatory variables were selected with multiple goodness-of-fit measures, they were included in regressions with the soundscape metrics measuring the total frequency range, the anthropogenic and biologic regions, and the 1 kHz bands.

Indicator variables were used to specify the spectrogram region that the metrics described. The 1 kHz band analysis regressions used interaction terms to capture the multiplicative effects of the LDI index and the sound power within each frequency band.

The 1-2 kHz frequency band was chosen as the baseline to compare the other frequency bands against because it is known to contain the majority of anthropogenic mechanical noise (Kasten et al. 2012). This analysis also used site fixed effects to control for any time-invariant site characteristics.

Mean-comparison tests were performed between the acoustic properties of the anthropogenic and biologic portions of soundscapes using paired T-tests in Stata SE13

Software.

Multiple measures of fit were used in the model selection. The model selection analysis favored models with larger R Squared measures, indicating higher explanatory power of models (Kutner et al. 2004). If the models being compared had unequal numbers of explanatory parameters, the Adjusted R Squared measure was used to avoid inflated explanatory power (Kutner et al. 2004). The Partial R Squared measure was used to analyze the supplementary explanatory power that the road proximity and wetland indicator measures added to the LDI index. In addition, the Akaike’s Information

Criterion (AIC) was used to measure goodness-of-fit for all of the model comparisons.

33

Multiple measures of fit were utilized to ensure that the best model specifications were selected.

Results

Variable Selection

LDI index scale

Tables 2-3, 2-4, 2-5, and 2-6 show that for regressions between LDI scores and each soundscape metric, either the flat weighted LDI100 or LDI500 resulted in the highest

R Squared values and lowest AIC values, indicating highest explanatory power. Each table presents the regression results for one dependent soundscape metric variable with each column representing a model for each independent LDI variable. There was little difference in the performance of the flat and distance weighted LDI100. The LDI500 flat weighted scores consistently slightly outperformed the distance weighted LDI500 scores. Further, the distance weighted LDI1500 scores had higher explanatory power than the flat weighted LDI1500 but was always outperformed by LDI100 or LDI500.

Physical and temporal control variables

The Partial and Adjusted R Squared values suggest that neither the season nor weather data variables were better control variables in the regressions (Appendix A).

Climatic conditions were intrinsically included in time of year variables. The temporal nature of the season variables also provided control for any sampling bias that was present. For these reasons, the season indicator variables were chosen to be included in the final models.

Wetland indicator and road proximity

The supplementary landscape variables did not add substantial explanatory power to the models, conditional on the inclusion of the LDI index (Table A-13,

34

Appendix A). The road proximity variable had low Partial R Squared values and did not increase Adjusted R Square for the inband power models. Neither the wetland indicator nor the road proximity variable provided additional explanatory power of more than one percent for the aggregate entropy models. Further, the Adjusted R Squared values decreased with the addition of the wetland or road proximity variables. Neither addition of the road proximity or wetland indicator variables resulted in an increase in Adjusted R

Squared of more than 0.02 for the delta power models. The regression results (Tables

A-1 through A-12; Appendix A) were also considered and the coefficients for these variables lacked statistical significance. These results indicate that the LDI index captured the landscape variability provided by the road or wetland proximity variables.

Models

After considering the different landscape scales for the LDI measure, supplementary landscape descriptors, and two sets of temporal proxy variables, the optimal model was picked for each soundscape measure based upon the goodness-of- fit measures. The results for these regressions are given in Table 2-7 which shows the relation between the landscape and the soundscape. The inband power regression model results in the first column of table 2-7 indicate that a one-unit increase in LDI100 for a site was associated with a 0.46 decibel (dB) increase in inband power. This relation, scaled to the sample standard deviation of the LDI100 scores (Table 2-8), provides a reference for the impact of change. The model predicts that a one standard deviation change in a site’s LDI100 score (15.27) results in an increase in inband power of 7.02 dB (72% of inband power’s standard deviation). Following the same logic, if the

LDI500 score for a site was increased by one standard deviation, the site’s delta power,

35

aggregate entropy, and center frequency values would be expected to decrease by 74,

79, and 54 percent of a standard deviation change, respectively.

1000 Hertz Frequency Bands

Table 2-9 illustrates the results of the OLS linear regression that predicted the average PSD values of the 1 kHz bands using the LDI100 scores, band indicator variables, interaction terms between LDI and each band indicator, season controls, and site fixed effects.

The coefficients for the frequency band indicators (Band2-Band20) express the average PSD of the frequency bands. The pattern of the coefficients indicates that band average PSD decreased in the first four bands, plateaued until 8 kHz, dropped drastically in the 8-9 kHz band and steadily declined until 20 kHz. Visual and audial inspection of the spectrograms revealed that wildlife sourced acoustic activity was mostly absent above 9 kHz. This mirrors the relatively high PSD values in the 5-9 kHz region and subsequent decline. The band indicator terms generally lack significance in the model, indicating that the average PSD of the excluded frequency band (1-2 kHz) was not significantly different than the average PSD of frequency bands 3 through 8.

All coefficients for the interaction terms between LDI and the frequency bands in the model were statistically significant (p<0.01) except for the interaction terms between

LDI100 and the second and third frequency bands (LDI_B2 and LDI_B3). The high p- value (p=0.50) for the Band1 interaction term indicated that the relation between LDI and this band was not significantly different from the relation between LDI and the 1-2 kHz band, the excluded indicator variable. There was a shift in the relation between LDI and average PSD beginning at 2 kHz. This was indicated by the marginal significance

(p=0.01) of the LDI_B3 interaction term and the strong statistical significance of all

36

subsequent frequency band interactions. These results illustrate that at this point in the frequency spectrum, the relation between average power and human influence was significantly different from the lowest frequencies. This result confirms 2 kHz as the best division between anthrophony and biophony 1 kHz frequency bands.

For clarity, the marginal impact of increasing the LDI100 index by one unit on the average PSD of each frequency band (y), controlled across sites and frequency bands, is displayed in Figure 2-31. The marginal effect between frequency band and LDI indicated that the impact of LDI on average PSD varied substantially by frequency band.

The maximum positive relation occurred in the 1-2 kHz band. The impact of LDI on average PSD decreased drastically after the 2 kHz band and continued to decrease in the higher frequency bands. However, this effect was not constant. Possibly, because of the presence of high-frequency, powerful anthropogenic noises (e.g. brakes squealing) that occurred in some of the high LDI sites (see example in Figure 2-4A).

Anthrophony and Biophony

The following results describe the anthrophony and biophony portions of the spectrogram for each site. This division at 2 kHz represents the anthropogenic and biologic sourced portions of each soundscape as supported by the strong correlation between LDI and average PSD up to 2 kHz (Table 2-9, Figure 2-3). Paired T-tests indicated that, on average, the PSD for the biologic portion of the spectrogram was

18.65 dB less than for the anthropogenic portion (p<0.01) and the aggregate entropy values for the biologic portion were larger by 2.12 (p<0.01) (Table 2-10).

1 For example, the marginal impact for Band 2 is calculated from the derivative of the regression model 푑(퐴푣푒푟푎푔푒⁡푃푆퐷퐵푎푛푑2) equation: = 0.68 + 0.04⁡, Where 0.68 is the LDI100 variable coefficient and 0.04 is the 푑(퐿퐷퐼100) LDI_B2 coefficient.

37

Table 2-11 presents the results from OLS regressions for the inband power and aggregate entropy of the anthrophony and biophony portions of the soundscape predicted by the LDI index and the season control variables. The LDI100 coefficients for the inband power models were higher for the anthrophony portion of the soundscape, indicating that the human sourced portion of the soundscapes responded more dramatically to LDI. Interestingly, the coefficients for the biophony inband power were positive, indicating that they correlated positively to LDI. The aggregate entropy models indicate that neither portion of the soundscape had a significant relation with LDI.

Discussion

The results of this study indicated that the LDI index is a good predictor of anthropogenic influence on soundscapes. The LDI index explained a substantial amount of variance in the sound metrics describing the total frequency spectrum. This was bolstered by the result that adding the wetland and road proximity variables did not add sufficient explanatory power to the regressions. These results are notable because the study sample spanned considerable variation in site LU/LC class, geographic location, and time of year. The trends between the LDI index and the soundscape measures reflect the generally accepted concept that human generated noises are loud, continuous, and occupy low frequencies (Warren et al. 2006; Slabbekoorn and

Ripmeester 2008; Barber et al. 2010; Pieretti and Farina 2013). Specifically, the results confirmed the results of Pijanowski et al. (2011a) showing that soundscape complexity

(as measured by aggregate entropy) decreased as human influence increased (as measured by the LDI index). The LDI correlated especially well with the inband power measures of the anthropogenic region of the soundscapes (R Squared = 0.65). The LDI

38

index provides a promising framework for predicting anthropogenic noise remotely from easily available LU/LC data.

The lack of methods to describe the landscape has hampered research that related the landscape to soundscapes (Fuller et al. 2015) and there were few studies to compare results with. Joo et al. (2011) used over ten variables to capture site variability with the surrounding landscape and found marginal significance (p<0.05) in their anthrophony model for a limited number of their measures. Alternatively, the LDI index was a highly significant (p<0.01) control variable that explained a high portion of the variance in our data set. The LDI index considers the presence and footprint of all

LU/LC classes using one continuous variable versus a multiple variable LU/LC indicator.

Additionally, the LDI index weights the land use configuration by the areal intensity of resource consumption, indirectly capturing noise production.

The results confirmed that 2 kHz is the most appropriate point to distinguish between anthrophony and biophony sourced 1 kHz bands. The 1 kHz band results showed that the influence of LDI, an indicator of human impact, on band average PSD became statistically significantly smaller at bands higher than 2 kHz. This study sampled soundscapes across a human influence gradient and captured a variety of human sourced sounds from many LU/LC classes. By statistically modeling the relation between LDI and the power of each 1 kHz band across the entire sample, it was determined that the LDI index was the most correlated with the sound power within the

20-2000 Hz soundscape region.

Statistically, the 1-2 kHz frequency band best captured the sound power associated with human influence but there was evidence of anthropogenic noise at

39

higher frequencies. The 1 kHz results indicated that increases in the LDI index was associated with increases in average power for the entire soundscape (Figure 2-3). The greatest effect was from 1-2 kHz but there was a sizeable effect up to 5 kHz and in the

9-10 kHz band. For example, the marginal effect of an increase in LDI for the 2-3 kHz band was 75% of the effect for the 1-2 kHz band and the 9-10 kHz band had 57% of the effect of the 1-2 kHz band. Figure 2-4A, C and D show examples of anthropogenic noises present at frequencies higher than 2 kHz within the study sample. The presence of human noise above 2 kHz explains why LDI100 had a positive significant effect on the inband power of the biophony region from 2-8 kHz (Table 2-11) versus a negative influence that was expected.

In this study we compared the impact of different sized contributing areas on soundscape characteristics. Inband power was most closely related to the surrounding area within a 100-meter radius. The metrics that considered the distribution of sound power across frequency (delta power, aggregate entropy, and center frequency) were more strongly correlated with LDI500. This indicated that overall power was more influenced by the closer sound sources but sound sources up to 500 meters away were more influential on the distribution of that sound power. In addition, the distance weighted LDI index measures did not perform as well as the flat weighted.

Anthropogenic noise is a result of energy transformation within human controlled land uses, as captured in the LDI index. It is well known that the decay of sound is proportional to distance from the source (Harris 1991). The results indicated that the effect of distance on LDI did not mimic the decay of sound as it was thought it would.

40

Possibly, because the sound sources that were found to be most influential were not far enough away for the impact of the decay pattern on LDI to emerge.

The regions of the soundscape associated with LDI indicate regions with anthropogenic noise and therefore, areas of potential conflict with biology.

Anthropogenic noise masks ecologically important acoustic cues that it overlaps in the frequency realm (Brumm and Slabbekoorn 2005; Dooling and Popper 2007). Animals use a broad range of frequencies (1-9 kHz) for communication (Napoletano 2004).

Acoustic signals by amphibians, birds, and insects captured in this study’s samples (see

Figure 2-4B and C for examples) occupied frequencies from 300 to 20,000 Hz but were most concentrated from 4-8 kHz. The frequency overlap of these sounds with human noise suggests that a conflict exists that could inhibit the transmission of ecologically valuable information.

Soundscape ecology research is currently trying to determine the extent and intensity of anthropogenic noise disturbance on wildlife. Tools that accurately predict soundscape characteristics from remote sensing data like the LDI index have high potential value in this pursuit. The results from this study have shown that soundscape characteristics can be predicted by the landscape features within a 100 and 500-meter radius surrounding a point. The results also confirmed that human influence results in noise below 2 kHz but indicated that it should not be assumed that higher frequencies are not occupied by anthropogenic noise. The LDI index could be used to target areas in the landscape with the greatest need for mitigation of conflict between anthropogenic noise and wildlife.

41

Figure 2-1. Location of the 67 sites where data was collected for this study.

42

Figure 2-2. Visual representation of the compartmentalization of sound signals using Fourier Transforms.

Table 2-1. Acoustic metrics used to describe the soundscapes. Measure Definition Average Power Spectral The total PSD within the selection of spectrogram Density (db/hz) divided by the number of time-frequency bins.

Average Inband Power This measure quantifies the amount of sound power (dB/Hz) within a selection. The PSD within the selection averaged over time, summed over frequency and then multiplied by the size (Hz) of the frequency bins to convert to total time-averaged spectral power. This value is then divided by the sampling rate of the recording.

Aggregate Entropy This measure quantifies sound disorder. The measure is high if the sound energy is evenly spread throughout the frequency spectrum. Selection of spectrogram is broken up into frequency bands.The sound power value is summed over time for each frequency band. The diversity of the band power values is calculated.

Delta Power (dB) The difference in time-averaged PSD at the upper and lower frequency limits of the selection.

Center Frequency (Hz) The point in the frequency spectrum that splits the selection into two equaled energy parts.

Metric descriptions adapted from Charif et al. (2010) and Joo et al. (2011)

43

Table 2-2. Non-renewable areal empower intensities for LU/LC classes used to calculate the LDI index values Non-renewable areal empower intensity (E15 sej Land use ha-1 year-1) Natural land / open water 0.00 Pine plantation 0.51 Low intensity open space / recreational 0.52 Unimproved pasture (with livestock) 0.53 Low intensity pasture (with livestock) 3.38 High intensity pasture (with livestock) 5.93 Medium intensity open space / recreational 6.06 Citrus 7.76 General agriculture 15.10 Row crops 20.30 High Intensity agriculture (dairy farm) 50.40 Recreational / Open space (high-intensity) 123.00 Single family residential (low-density) 197.50 Transportation - 2 lane highway 308.00 Single family residential (med-density) 658.33 Single family residential (high-density) 921.67 Institutional 4042.20 Multi-family residential (low density) 4213.33 Trasnportation 4 lane highway - high intensity 5020.00 Low Intensity commercial (comm strip) 5175.40 Industrial 5210.60 High intensity commercial (mall) 8372.40 Multi-family residential (high rise) 12771.67 Central business district (avg 2 stories) 16150.30

Adapted from Reiss et al. 2010

44

Table 2-3. Regression results comparing LDI scales to the inband power soundscape metric Inband Power Flat Weighted Distance Weighted

LDI100 0.48*** 0.48*** (0.05) (0.05) LDI500 0.52*** 0.49*** (0.05) (0.05) LDI1500 0.50*** 0.54*** (0.06) (0.06) Constant 101.29*** 99.18*** 99.92*** 101.42*** 99.83*** 98.25*** (1.25) (1.48) (1.80) (1.25) (1.56) (1.68) R-sqrd 0.58 0.53 0.36 0.58 0.51 0.48 AIC 439.42 446.54 467.36 440.00 449.77 453.35 Note: Asterisks indicate statistical significance: *p<0.1, **p<0.05, ***p<0.01. Parentheses indicate the heteroskedastic robust standard errors

Table 2-4. Regression results comparing LDI scales to the delta power soundscape metric Delta Power Flat Weighted Distance Weighted

LDI100 -0.43*** -0.44***

(0.06) (0.05)

LDI500 -0.49*** -0.47***

(0.05) (0.05)

LDI1500 -0.52*** -0.52***

(0.06) (0.06) Constant -42.25*** -39.65*** -39.33*** -42.24*** -40.35*** -38.51*** (1.49) (1.52) (1.75) (1.47) (1.53) (1.67) R-sqrd 0.44 0.45 0.36 0.45 0.43 0.43 AIC 462.08 460.78 471.29 461.04 463.80 464.04 Note: Asterisks indicate statistical significance: *p<0.1, **p<0.05, ***p<0.01. Parentheses indicate the heteroskedastic robust standard errors

45

Table 2-5. Regression results comparing LDI scales to the aggregate entropy soundscape metric Aggregate Entropy Flat Weighted Distance Weighted

LDI100 -0.03*** -0.03*** (0.01) (0.01)

LDI500 -0.04*** -0.03*** (0.01) (0.01)

LDI1500 -0.04*** -0.04*** (0.01) (0.01) Constant 3.73*** 4.00*** 4.03*** 3.72*** 3.94*** 4.06*** (0.18) (0.18) (0.21) (0.18) (0.18) (0.20) R-sqrd 0.27 0.34 0.27 0.27 0.31 0.30 AIC 154.98 147.82 154.17 154.34 150.06 151.68 Note: Asterisks indicate statistical significance: *p<0.1, **p<0.05, ***p<0.01. Parentheses indicate the heteroskedastic robust standard errors

Table 2-6. Regression results comparing LDI scales to the center frequency soundscape metric Center Frequency Flat Weighted Distance Weighted

LDI100 -49.72*** -48.87*** (14.72) (14.54)

LDI500 -66.16*** -61.57*** (18.09) (17.36)

LDI1500 -71.44*** -67.83*** (22.10) (20.35) Constant 2176.25*** 2729.84*** 2833.09*** 2152.31*** 2617.92*** 2832.88*** (508.65) (609.83) (707.44) (502.62) (592.93) (690.66) R-sqrd 0.23 0.32 0.27 0.23 0.30 0.29 AIC 1162.63 1154.07 1158.71 1163.11 1156.46 1157.44 Note: Asterisks indicate statistical significance: *p<0.1, **p<0.05, ***p<0.01. Parentheses indicate the heteroskedastic robust standard errors

46

Table 2-7. Regression results for total frequency spectrum models Inband Pwr Delta Pwr Entropy Cntr Freq

LDI100 0.46***

(0.06) LDI500 -0.54*** -0.05*** -62.54*** (0.08) (0.01) (17.39)

Winter

Spring -2.3 -2.65 0.59* 187.5 (2.63) (3.39) (0.35) (352.53) Summer -4.33 1.86 -0.05 -506.83** (5.53) (2.25) (0.18) (212.20) Fall -5.21*** 0.28 0.12 -58.85 (1.74) (2.42) (0.20) (312.63) Constant 104.44*** -37.82*** 4.42*** 2626.59*** (2.38) (3.48) (0.40) (635.87) R-sqrd 0.63 0.47 0.42 0.33 Note: Asterisks indicate statistical significance: *p<0.1, **p<0.05, ***p<0.01. Parentheses indicate the heteroskedastic robust errors.

Table 2-8. Summary statistics for LDI and soundscape metrics Variable Mean Std Dev Minimum Maximum LDI 100 24.51 15.27 00.00 41.70 LDI 500 26.71 13.58 00.00 41.00 LDI 1500 26.18 11.59 00.10 38.50 Inbnd Pwr 113.07 09.69 89.60 136.20 Avg PSD 117.21 09.65 93.50 140.10 Delta Pwr 52.86 09.97 -73.10 -30.90 Entropy 03.00 00.86 00.96 05.33 Cntr Freq 962.69 1581.55 187.50 7125.00

47

Table 2-9. OLS regression results for 1 kHz band average PSD with site fixed effects Variable Coeff. Robust Std. Err. P>t

LDI100 0.72 0.06 0.00 Band1 10.83 2.10 0.00 Band3 -0.37 2.37 0.88 Band4 -1.40 2.36 0.55 Band5 -3.62 2.31 0.12 Band6 -3.30 2.26 0.14 Band7 -3.54 2.02 0.08 Band8 -2.85 2.16 0.19 Band9 -10.49 2.14 0.00 Band10 -15.38 1.90 0.00 Band11 -18.45 1.94 0.00 Band12 -20.03 1.96 0.00 Band13 -20.52 2.03 0.00 Band14 -20.77 2.04 0.00 Band15 -21.99 1.99 0.00 Band16 -23.36 1.87 0.00 Band17 -24.01 1.89 0.00 Band18 -25.14 1.88 0.00 Band19 -24.65 2.25 0.00 Band20 -26.51 1.87 0.00 LDI_B1 -0.04 0.07 0.50 LDI_B3 -0.18 0.07 0.01 LDI_B4 -0.25 0.07 0.00 LDI_B5 -0.30 0.07 0.00 LDI_B6 -0.37 0.07 0.00 LDI_B7 -0.40 0.06 0.00 LDI_B8 -0.47 0.07 0.00 LDI_B9 -0.35 0.06 0.00 LDI_B10 -0.31 0.06 0.00 LDI_B11 -0.33 0.06 0.00 LDI_B12 -0.36 0.06 0.00 LDI_B13 -0.40 0.06 0.00 LDI_B14 -0.44 0.06 0.00 LDI_B15 -0.41 0.06 0.00 LDI_B16 -0.41 0.06 0.00 LDI_B17 -0.44 0.06 0.00 LDI_B18 -0.46 0.06 0.00 LDI_B19 -0.51 0.07 0.00 LDI_B20 -0.50 0.06 0.00 spring -7.77 1.91 0.00 summer -7.59 2.75 0.01 fall -8.02 1.55 0.00 Constant 106.33 2.18 0.00 Notes: Each row of the table has the coefficient value (column two), robust standard error (column three) and p-value (column four) for one dependent variable included in the model. The included variables are the LDI100 scores (row one), band indicator variables (Band2-Band20, rows 2-20), interaction terms between LDI100 and each band indicator (ldi_B2-ldi_B20, rows 21-39), season controls (rows 31-33), and site fixed effects (not included in table). Frequency band labels represent the highest extent of the frequency band in kHz. For example, Band2 ranged from1-2kHz.

48

Figure 2-3. Marginal effect of human influence on average PSD across the frequency spectrum

Table 2-10. Difference in means between the biophony and anthrophony average PSD and entropy values Avg. PSD Entropy Difference in -18.65*** 2.12*** Means Note: Asterisks indicate statistical significance: *p<0.1, **p<0.05, ***p<0.01.

49

Table 2-11. Regression results for anthrophony and biophony models Average Power Entropy

Anthrophony Biophony Anthrophony Biophony LDI100 0.52*** 0.24*** (0.07) (0.06) LDI500 0.00 -0.01 (0.00) (0.00) Spring -2.37 -0.88 -0.04 0.01 (2.71) (2.52) (0.15) (0.16) Summer -3.63 -4.81 0.28* 0.70** (5.67) (5.08) (0.14) (0.31) Fall -5.94*** -4.45* 0.03 0.54*** (1.86) (2.24) (0.12) (0.13) Constant 102.26*** 98.24*** 2.26*** 4.35*** (2.53) (2.59) (0.16) (0.17) R-sqrd 0.65 0.29 0.03 0.32

AIC 445.48 452.21 72.53 79.65 Note: Asterisks indicate statistical significance: *p<0.1, **p<0.05, ***p<0.01. Parentheses indicate the heteroskedastic robust standard errors.

50

Figure 2-4. Examples of spectrograms from study sample covering ten minutes from four different sites. A) The black outlines highlight vehicular brake noise, B) the black outlines highlight sounds sourced from insects, birds, and amphibians, C) the black outlines highlight sounds sourced from birds in lower frequencies and exceptionally broad band insect noise up to 20 kHz, and D) a spectrogram showing acoustic calls from wildlife overlapping with noise from a truck and boat.

51

CHAPTER 3 LAND USE SOUNDSCAPE CHARACTERISTICS

Introduction

The relation between land use and land cover (LU/LC) and the soundscapes they produce has only recently become a focus of study. A limited number of studies have found associations between specific LU/LC classes and soundscape measures. For example, Pijanowski et al. (2011a) qualitatively described the relation between the acoustic activity, diversity, and evenness of eight soundscapes with the proportion of agriculture and urban land uses of the surrounding areas. Also, Joo et al. (2011) found statistically significant correlations between sound power and the presence of surrounding agriculture and commercial land uses. Dooley (2016) studied the quantitative relation between the intensity of human land uses of landscapes, as measured by the Landscape Development Intensity (LDI) index (Brown and Vivas 2005;

Reiss et al. 2010), and soundscape measures in north central Florida.

Habitat Specific Soundscapes

Unique habitat-dependent soundscapes exist in the natural environment that are the product of physical properties like the quantity of vegetation structure (Farina and

Pieretti 2014) and the different biotic sources of sound (Slabbekoorn 2004; Radford et al. 2010; Bormpoudakis et al. 2013). Similarly, human land uses are areas of unique activity and architecture. The acoustic characteristics of urban soundscapes have been well studied in the context of their effect on human well-being since the introduction of the soundscape concept by Schafer (1977). Urban sites create unique sound events and environments that result in distinct human perceptions (Raimbault et al. 2003; Kang and Zhang 2010). The realization that noise has detrimental effects on birds and other

52

animals (Barber et al. 2010; Blickly and Patricelli 2010; Francis and Barber 2013) encourages the study of urban soundscapes from the perception of wildlife.

Temporal Variation in Soundscapes

Temporal variation is a key characteristic of soundscapes (Botteldooren et al.

2006). For example, vehicular traffic is a sound source with diurnal fluctuations (Warren et al. 2006) that is known to dominate the soundscapes of substantial geographical area

(Barber et al. 2011; Nega et al. 2013). Monitoring urban soundscapes over daily and seasonal patterns ensures that the full range of potential impact is captured and could inform noise management decisions (Joo et al. 2011; Gage and Axel 2014).

Noise Thresholds

Agencies concerned with human health have established noise exposure limits but formal noise thresholds for animals do not exist because there is not adequate evidence for where they should be set (Dooling and Popper 2007). The most comprehensive noise threshold for terrestrial animals was given by Dooling and Popper in their 2007 report. They determined that 50-60 dBA was the threshold for traffic noise interference with effective bird communication in a quiet suburban or rural area based on the hearing abilities of 14 birds and the characteristics of bird vocalizations. Similarly, the EPA suggested a day night level (DNL) of 55 dBA in outdoor communities to ensure human health and public well fare based on effective communication, community reaction and the reduction of human annoyance and complaints (US EPA 1974a). The

World Health Organization’s (1999) recommended threshold was 55 dB (LAeq-day) in order to avoid serious annoyance during the day or evening. The threshold of 55 dB is a good point of comparison for observed environmental sound levels because it has

53

strong significance of ensuring human wellbeing and some evidence that it protects birds.

Problem Statement

Recently, interest in noise impacts expanded from a human centric perspective to including the impacts of anthropogenic noise on animals. As the influence of humans within landscapes has increased, the reach and intensity of noise disturbance has grown, and efforts to mitigate it have been inhibited by a lack of understanding of where these conflicts exist. As a descriptor of anthropogenic noise sources, LU/LC patterns have the potential to inform noise disturbance mitigation regimes. Thus far, LU/LC- specific sound sources and noise patterns that have potential to disturb wildlife have not been well established.

Plan of Study

In this study we characterize the soundscapes that result from sixty-one landscapes having differing combinations of human land uses in north central Florida,

USA. The soundscapes were classified by the LU/LC immediately surrounding the point of recording. Thirty-minute soundscape recordings were captured from each site across the study area over a two-year time period and were augmented by 48-hour sound level data recorded for each of four seasons on a sub sample of the 30-minute sites. Overall, we describe the major soundscape characteristics for conservation areas, recreational parks, single family residential neighborhoods, commercial shopping centers, industrial areas, and central business districts during ambient morning conditions that avoid times of high acoustic activity from road traffic (0900-1000 hours). In addition, to capture diurnal characteristics, we characterize noise generation over a 48-hour period at four

54

sites and then use these data to test the appropriateness of using 30-minute recordings to characterize the sound signature of LU/LC classes.

Methods

Site Descriptions

30-minute sites

Sound recordings were captured from sixty-one sites in north central Florida between April 2013 and July 2015. The sites spanned a spectrum of human influence and captured six different LU/LC classes as follows, conservation (CON; n=10), recreational park (RP; n=10), single family residential (SFR; n=16), industrial (IND; n=8), commercial shopping center (CSC; n=13), central business districts (CBD; n=4). The site of recording was selected to be close to the center of the targeted LU/LC to minimize capturing sounds from other land uses.

The CON sites were large tracts of publicly owned land, protected from human development. Placement of the recording equipment avoided being near roads. These sites were included to provide a background, or in a sense “controls” for the rest of the

LU/LC classes that were human dominated areas. The RP sites had smaller land areas that included recreational facilities like sports fields, playgrounds, and picnic pavilions.

Many of these sites were situated close to residential areas. The SFR sites were suburban areas of detached single family homes. The IND sites were industrial process facilities including, waste water treatment plants, coal fired power plants, and cement manufacturing. The location of these recordings was subject to safety concerns by the plant operators and as a result, were not always taken at the actual center of the facilities. The recordings at the CSC sites were taken in parking lots associated with

55

large shopping plazas and malls. The CBD sites were taken within the city centers of four cities in Florida: Orlando, Tampa, Jacksonville, and Gainesville.

48-hour sites

Four sites that spanned the diversity of human influence and LU/LC classes of the short term sample were selected for long term data collection. These sites were at a remote, forested wetland in a conservation area (CONX), the central business district of a city with a population of approximately 150,000 people (CBDX), an 8.5-hectare forested, urban, recreational park including a creek and bordered by residential areas

(RPX), and a cement plant distribution center (INDX). A construction material recycling facility was selected as the initial INDX site. Limited access and topography resulted in low sound levels. The cement facility was utilized for subsequent samples and references to INDX data do not include the fall INDX data unless specified.

Acoustic Sampling

30-minute site data

The soundscape of each site was captured for 30 minutes between the 900 and

1000 hour. This time was targeted for its ambient nature to better understand the continual background sound conditions, as these have the most potential for disturbance to the biological community. The thirty-minute soundscape samples were recorded at each of the 61 sites with a Fostex fr-2le field recorder and a Seinheisser ME

62 omni-directional microphone secured to a tripod at a height of 1 meter. The recordings were waveform format of 24-bit depth and sampled at a rate of 48 kHz.

Periods of rain or days with wind exceeding 10 miles per hour were avoided. Sound sources were identified and documented during the recording of the samples from a distance.

56

48-hour site data

The long-term sampling regime was a series of 48-hour samples gathered four times a year, once in each season. The seasonal sampling for all sites was collected within a two-week period. These samples were A-weighted equivalent continuous sound pressure levels (LAeq) measures taken every minute with a Rion NL-32 sound level meter. The meter was calibrated prior to deployment. The summer CBDX data only spans 12 hours from approximately 0800 to 2000 due to the confiscation of the recording equipment by the local police department.

Data Analysis

30-minute sites

The 30-minute recordings were converted into time, frequency and intensity components in Raven Pro 1.4 (Cornell University, Ithica, New York) using Fourier

Transforms with a Hann window size of 256 samples and 188 Hz grid size. The sound intensity measure was power spectral density (PSD), which is a measure of power per unit frequency (dB/Hz). Spectrograms and power spectrums were created to visually represent the soundscape PSD measures. The mean power spectrum for each LU/LC class was created from the average PSD of each 1 kHz band from 0-20 kHz. The audio recordings, spectrograms, power spectrums and on-site observations were examined for each soundscape and similarities within the LU/LC classes were distinguished.

The LU/LC classes were also compared statistically using Stata SE13 software

(Stata Corp, College Station, Texas). Five metrics that describe the intensity and/or distribution of the PSD over 20-20,000 Hz were calculated in Raven Pro for each spectrogram. Table 3-1 gives a brief description of each measure. Additionally, the LDI index (Brown and Vivas 2005; Reiss et al. 2010) was calculated for the landscape within

57

a 100 and 500-meter radius surrounding each soundscape sampling location in ArcMap

10.3 (ESRI, 2014) following the methodology of Dooley (2016). The differences between the acoustic metrics for each LU/LC class was investigated in two ways. Mann

Whitney-U tests were used to identify significantly different distributions of acoustic metrics between LU/LC classes. In addition, ordinary least squares (OLS) robust regressions were performed with each acoustic metric as the dependent variable. The independent variables were indicator variables for each LU/LC class, the LDI score for the site calculated over either 100m or 500m following Dooley (2016), and an indicator variable for the season the recording was taken during. High LDI index scores are associated with high levels of non-descript, low frequency noise (Dooley 2016) that could be common across LU/LC classes. The LDI index was included as a control variable to separate the effect of overall landscape development intensity and LU/LC class soundscape characteristics on the acoustic metrics.

48-hour sites

The long term, one-minute LAeq sound levels were used to calculate common acoustic metrics and compared over the time of day, site and season. A brief description of these acoustic metrics that averaged sound levels over longer time frames is presented in Table 3-1.

The appropriateness of using 30-minute recordings to characterize the sound signature of LU/LC classes was tested with Mann Whitney U tests using Stata SE13 software (Stata Corp, College Station, Texas). The statistically significant differences between the distribution of sound levels collected during the 0900-1000 hour and each of the other daylight hours was determined. Day light hours were 0700-2200 as defined in the DNL measure by the EPA (US EPA 1974a).

58

Results and Discussion

30-minute sites

An example of a site spectrogram for each LU/LC class is shown in Figure 3-1.

These sites exemplified many characteristics common with the other sites in their LU/LC class but there was site specific variation. Qualitative inspection of the site recordings and their spectrograms revealed characteristic sounds and patterns for the LU/LC classes. The CON sites (Figure 3-1A) commonly exhibited insect noise as a relatively constant band of varying width between 5 and 9 kHz and many lower frequency events at about 1-5 kHz from amphibians and birds. The RP sites (Figure 3-1B) like the SFR sites (Figure 3-1C) had a mix of biological and anthropogenic noise (indistinct noise between 20 Hz and 2 kHz) that consisted of individual cars (frequency peaks around 10 kHz), insects (varying widths between 3-20 kHz), and birds (single events at 2-8 kHz).

Generally, the industrial sites (Figure 3-1D) had relatively few sound events, with a more diffuse, continuous noise than the other human land uses. In addition, the diffuse noise was most intense in low frequencies and occupied bands as high as 12 kHz. They generally lacked biological noises or it was masked by the relatively high frequency industrial noises. Generally, the CBD (Figure 3-1E) and the CSC (Figure 3-1F) sites were located near busy roads and their soundscapes exhibited the intense traffic noise that was evident as a constant intense band of sound below 3 kHz. In addition, there were noise events from nearby cars, large trucks or motorcycles that occupied frequencies up to 12 kHz and occasional intense, high-pitched brake noise (12-20 kHz).

Some CSC sites also had a small amount of bird and insect sounds between 3 and 10 kHz.

59

Figure 3-2 displays the distribution of the mean spectral power and standard deviation for each site type. Consistently, in all site types, the first band (0-1kHz) had the highest mean PSD. This frequency region contained nondescript noise in almost every site that was more intense in heavily developed areas and lightest or nonexistent in the CON sites. Although all of the power spectrums follow a general pattern of diminishing sound power with increasing frequency, there were some patterns unique to

LU/LC class. Unlike the other LU/LC classes, the CON spectrum had similar power values from 1-8 kHz. These sites had the most acoustic activity from wildlife of any of the classes and it was in this frequency range. The IND power spectrums were distinctive. While the band power decreased with increasing frequency in a relatively linear pattern, the variance increased with increasing frequency. The CBD and CSC power spectrums had abnormally high 15 kHz power bands. This was most likely the result of the high pitched brake noise that was unique to these two LU/LC classes.

Table 3-2 displays the difference between the CON power spectrum and each of the other LU/LC classes. The values are the average PSD of each band for the CON sites subtracted from the equivalent value for each of the other LU/LC classes. The RP mean power levels were most similar to the CON group. The IND class had the highest mean PSD and largest deviation from the CON class for all bands. Across all human

LU/LCs, the 7-8 kHz band had the most similar mean power values to the CON class.

This band had especially intense insect noise in the CON sites and elevated the PSD in this area to levels found in the other LU/LC classes.

Figure 3-3 illustrates the distribution of soundscape measures calculated for the

20-20,000 Hz frequency range by LU/LC class. The inband power measure

60

distinguished between LU/LC classes the most with four statistically significant different groups. The CON sites were the least influenced by human activity and have relatively low levels of sound energy in the low frequency bands (below 2 kHz) and more energy between 2 and 9 kHz from wildlife. This explains the CON sites’ significantly smaller delta power values, larger values and variation in center frequency, high entropy values, and low inband power values. The RP/SFR and the CSC/CBD combinations do not have statistically different distributions for any of the measures. Both of these LU/LC combinations shared similar observed soundscape characteristics. In addition, the RP and SFR LU/LC classes were closely associated on the landscape; eight out of ten of the parks sites were bordered by residential LU/LCs. The IND group had relatively low delta power values and significantly higher inband power values from the other human dominated LU/LC classes. The IND sites had high variance in entropy values. The industrial sites with the four highest entropy values were dominated by the broadband industrial noise in their spectrograms. Although the industrial sites with smaller entropy values had broadband noise, it was not as overwhelming as at the sites with higher values.

In addition, site specific sound events had large impacts on the soundscape metrics. The sites with soundscape metrics that differed significantly from other sites within their class (i.e. were in the 10th and 90th percentiles) tended to have powerful sound events in their 30-minute recordings (Figure 3-3). For example, the IND class had a very large range in delta power and entropy partially driven from one event at one site where hard material fell into a metal silo causing intense noise that spanned the entire frequency range for the recording (20-20,000 Hertz). In addition, insect noise was

61

associated with high entropy values and low delta power values. Sites with significantly higher center frequency values tended to have higher amounts of acoustic activity from wildlife and lower amounts of sound energy below 2 kHz compared to their LU/LC class.

The results in Figures 3-1, 3-2, 3-3 suggest that the differences in soundscape characteristics of each site was driven by LU/LC class differences. In a recent study,

Dooley (2016) illustrated that soundscape characteristics were driven by human development intensity as measured by the LDI index. The LDI index accounts for the intensity of nonrenewable resource use, but not unique characteristics of specific LU/LC classes. For example, a shopping center and central business district could have similar

LDI scores but parking lot sounds like car doors shutting and rolling shopping carts would be unique to the shopping center. In an attempt to disentangle the individual effects of landscape intensity and immediately surrounding LU/LC class on soundscapes, regression results for models that incorporated both LDI and LU/LC class as descriptor variables for soundscape metrics are shown in Table 3-2.

The regression results in Table 3-2 indicated that the LU/LC indicators were only a useful addition to the inband power models already controlled for land use intensity.

The LU/LC indicator coefficients were not statistically significant additions to the delta power, aggregate entropy, or center frequency models controlled for LDI. This result was not expected because these measures account for the distribution of sound power across the frequency spectrum and not just the sum of power as with inband power. It was thought that the characteristic sound events observed for the LU/LC classes would drive statistically significant coefficients for the LU/LC indicators in the regression models. Possibly, the site specific sound events that introduced sample variance could

62

have contributed to the LU/LC class indicators’ statistical insignificance in the regression models. Further, the sample size and 30-minute recording time may not have been large enough to reveal statistically significant frequency spectrum similarities between

LU/LC classes in the entropy, delta power, and center frequency regression models.

The inband power regression results revealed that LDI and LU/LC class had separate and significant impacts on soundscape inband power. In general, as was seen in Figure 3-3D, the IND and CON LU/LC classes had the highest and lowest impact on inband power, respectively. The intense, constant, broadband noise common in the IND sites likely contributed to the high industrial inband power measures. These noise properties and the tendency for these sites to be located in remote areas indicated that they create soundscapes with a high potential for wildlife disturbance. The CBD LU/LC also had a relatively high marginal impact on inband power. When controlled for overall human influence with LDI, the impact of the SFR, CSC, and RP LU/LC classes on inband power were similar. Table 3-3 presents the statistically significant difference between the LU/LC indicator coefficient estimations for the inband power model that included LDI. These results verify that the impact that the CBD and IND LU/LC classes had on inband power was significantly different from each other and the other LU/LC classes.

48-Hour Sites

Figure 3-4 displays the variation of the A-weighted sound level over a 48-hour time period for the CONX, PRX, INDX, and CBDX sites for each of the four seasons.

The CBDX and INDX site sound levels were above the 55 dBA benchmark 51% and

48% of the time sampled. This finding suggested that these two environments were not suitable for effective bird communication (Dooling and Popper 2007). Nevertheless,

63

RPX and CONX only surpassed 55 dBA 1.5% and 1% of the time sampled, respectively. This implied that recreational park areas could constantly provide an adequate acoustic habitat for birds in urban environments. However, sound levels lower than 55 dBA have been associated with reduced bird presence and anuran relative abundance near roads (Barber et al. 2011). Further, this urban park site had lower sound levels that others surveyed (Kuehne et al. 2013; Brambilla et al. 2013).

Common acoustic metrics were exhibited in Table 3-4 for each site by season.

Generally, INDX had the highest sound levels, followed by CBDX, RPX and CONX, respectively. The CBDX and INDX sites’ sound levels consistently far exceeded the 55 dBA DNL and Leq day levels recommended by the EPA and WHO. INDX had the largest difference in L90 and L10 levels, which indicated high variance in sound levels. The INDX recording location was near the cement loading area and was subject to loud and varied noises from the filling procedure, truck engines, and back-up alarms. Personal communication with the plant manager revealed that the frequency and timing of the trucks was different every day depending on demand. The resulting sound level spikes can also be seen in the time series data (Figure 3-4). The A-weighted sound levels

(Figure 3-4) indicate that they start as early as 0500.

The daily and seasonal time series patterns (Figure 3-4) at the RPX and CONX sites were similar but the RPX trends were more subdued. This could have been driven by the differences in the biologic communities between CONX, a minimally impacted wetland system and RPX, a park situated in an urban matrix. In addition, RPX had a higher ambient sound level that dampened the impact of sound events. The CONX spring, summer, fall and RPX summer and spring time series sound levels increased in

64

the evenings. The night Leq for fall and spring at CONX were higher than the corresponding day measures and the summer Leq-night measure was even higher. This was most likely the result of frogs and insect noise, a dominant part of the CON and RP

30-minute samples. The summer CONX and RPX data violated the 55 DNL and LAeq-day thresholds, defined as unhealthy for human use. This result suggests that these thresholds are not appropriate for analyzing the suitability of acoustic environments for wildlife.

Comparison of 30-Minute and 48-Hour Sampling Timeframes

Table 3-5 displays p-value results for Mann-Whitney U tests comparing the distribution of Leq values for each hour in the day to 0900-1000 for each 48-hour sample period. The number of hours during the day where the distribution of sound level failed to be significantly different from the nine o clock hour ranged from 0-12 with an average of 5.25. There were no consistent patterns across site or season for failed significance.

However, across site and season, the morning to late afternoon was more similar to the

0900 hour than the evening. This supported the assumption that 30-minute results could be generalized to a longer portion of the day time soundscape. Sampling sound for long periods of time creates large data sets that are difficult to analyze (Kasten et al. 2012).

The informed timing of recording could efficiently capture temporal variation within a soundscape without over sampling.

Conclusions

This study established that human dominated LU/LC classes have unique soundscape patterns and sound sources. The CON and IND classes stood out at opposite ends of the human influence spectrum. These two LU/LC classes had observable qualitative unique attributes that contributed to statistically different

65

measures of their soundscapes from the other classes. Two sets of the LU/LC classes had statistically and qualitatively similar soundscapes, the SFR and RP pair and the

CSC and CBD pair. Inband power distinguished between the LU/LC classes the most of the soundscape metrics. Landscape development intensity and LU/LC class had separate and statistically significant effects on inband power. The effect of IND and

CBD on inband power were significantly different than all other human dominated

LU/LC classes. The temporal data revealed that the distribution of sound level data taken during 0900 hour failed to be significantly different from an average of 5.25 other daylight hours. The CBDX and INDX sites surpassed the noise thresholds to protect human health and surpassed the 55 dBA threshold aimed to protect birds half the sampled time. The RPX and CONX sites did not surpass the bird threshold but did surpass the human health thresholds during the summer.

The differences between the acoustic conditions of the LU/LC classes imply different impacts on surrounding natural soundscapes. For instance, the relatively loud, constant, and broad band noise at the industrial 30-minute sites suggest masking of many sounds in natural soundscapes. The absence of this type of noise and large sound level spikes in the central business districts suggested that some wildlife might be able to adapt to the acoustic conditions of these areas. This is bolstered by the presence of some bird calls in the CBD sites. However, these CBD sites were relatively modest compared to some urban areas in the world. The 48-hour RPX and CONX sites surpassed human health thresholds in the summer sampling but not the bird health threshold suggesting that guidelines based on human health are not appropriate for

66

wildlife. These results also suggest that parks in modest urban areas can provide acoustically appropriate habitat for birds.

67

Table 3-1. Definitions of soundscape and acoustic metrics Measure Definition Average PSD (db/hz) The total PSD within the selection of spectrogram divided by the number of time-frequency bins. Average Inband Power This measure quantifies the amount of sound power within a (dB/Hz) selection. The PSD within the selection averaged over time, summed over frequency and then multiplied by the size (Hz) of the frequency bins to convert to total time-averaged Aggregate Entropy Thisspectral measure power. quantifies This value sound is then disorder. divided The by the measure sampling is high if the sound energy is evenly spread throughout the frequency spectrum. Selection of spectrogram is broken up into frequency bands.The PSD value is summed over time for each frequency band. The diversity of the band energy values is calculated.

Delta Power (dB) The difference in time-averaged PSD at the upper and lower frequency limits of the selection. Center Frequency (Hz) The point in the frequency spectrum that splits the selection into two equaled energy parts as measured by PSD summed over time. Leq_n (dB), LAeq_n Equivelant sound level for the n time frame. Calculated as the (dBA) average of sound pressure levels sampled during the n time frame. LAeq_n is for A weighted sound levels. Ln The sound level just exceeded for nth percent of the time in the sampling time frame. Day Night Level (DNL) Equivelant sound level over 24 hours with a 10 dB synthetic increase in sound level during night time hours of 10pm to 7am. Metric descriptions adapted from Charif et al. (2010) for PSD measures and EPA (1974) for day night level

68

A B C

D E F

Figure 3-1. Ten-minute-long examples of spectrograms from each LU/LC class. From left to right, top to bottom: A) CON, B) RP, C) SFR, D) IND, E) CBD, F) CSC

69

Con RP SFR

160 160 160

140 140

140

120

120

120

100

100

Average Power Average Power Average Power Average

100

80

80 80

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 Frequency Band (kHz) Frequency Band (kHz) Frequency Band (kHz)

Ind CBD CSC

160 160 160

140

140

140

120

120

120

100

Average Power Average Power Average Power Average

100

100

80

80 80

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 Frequency Band (kHz) Frequency Band (kHz) Frequency Band (kHz)

Figure 3-2. Power spectrums for each LU/LC consisting of mean and standard deviation PSD values for each 1 kHz band between 20 and 15,000 Hz.

70

Table 3-2. Difference in mean average PSD (dB) for each frequency band between CON and each of the other LU/LC classes Freq Band RP SFR IND CBD CSC 0.02-1kHz 12.9 12.8 31.1 22.6 20.7 1-2 kHz 14.4 15.2 33.4 25.4 22.8 2-3 kHz 7.2 7.9 26.6 16.3 14.2 3-4 kHz 3.8 6.8 24.4 12.9 11.2 4-5 kHz 1.3 5.4 22.9 11.0 7.8 5-6 kHz 0.3 3.0 20.4 8.5 5.3 6-7 kHz -0.3 3.0 18.1 8.1 4.5 7-8 kHz -2.9 1.6 14.7 5.1 1.6 8-9 kHz 0.7 5.3 19.1 12.2 6.8 9-10 kHz 2.3 7.2 21.3 14.6 9.4 10-11 kHz 4.1 7.1 21.8 15.7 9.1 11-12 kHz 3.6 6.0 20.5 14.6 7.6 12-13 kHz 3.0 4.8 18.0 13.7 6.8 13-14 kHz 2.3 4.1 15.7 11.3 4.6 14-15 kHz 1.4 3.6 14.3 15.1 5.7

71

A B

-30

8000

-40

6000

-50

4000

-60

2000

Delta PowerDelta(dB)

-70

CenterFrequency (Hz) 0

Con RP SFR Ind CBD CSC Con RP SFR Ind CBD CSC I II II III II,III III I II II,III II,III II,III III LU/LC LU/LC

C D

140

5

130

4

120

3

Entropy

110

2

100

InbandPower(dB)

1 90

Con RP SFR Ind CBD CSC Con RP SFR Ind CBD CSC I II II II,III II,III III I II II IV III III LU/LC LU/LC

Figure 3-3. Soundscape measure distributions grouped by site type. A) Distribution of delta power measures. B) Distribution of center frequency. C) Distributions of aggregate entropy measures. D) Distribution of inband power measures. Box plots indicate the 25th, 50th, and 75th percentiles. Whiskers extend to 10th and 90th percentiles. Data points outside the 10th and 90th percentiles are circles. The roman numerals indicate groups within each plot that have significantly different (p<0.05) distributions as determined with Mann-Whitney U tests.

72

Table 3-3. OLS regression results for linear models that use LDI and LU/LC dependent variables to predict soundscape metrics Dependent Inband Var Power Delta Power Entropy Center Freq 100mLDI 0.21*** (0.07) 500mLDI -0.53** -0.05** -61.46** (0.24) (0.02) (24.10) RP 6.42*** -1.76 -0.05 -407.10 (2.24) (6.47) (0.53) (1079.09) CSC 7.69** -2.72 -0.06 -112.15 (3.14) (8.16) (0.71) (1187.47) SFR 7.96*** 1.32 0.03 -2.66 (2.68) (7.31) (0.70) (1167.13) Ind 20.48*** -5.62 0.26 48.11 (3.51) (9.82) (0.83) (1284.57) CBD 12.43*** 5.10 0.25 307.94 (3.49) (9.50) (0.81) (1269.91) Spring -1.97 -3.90 -0.47 286.82 (2.04) (3.78) (0.39) (392.88) Summer -0.91 -1.30 -0.01 -685.78 (6.30) (3.40) (0.35) (472.66) Fall -4.15*** 1.45 0.11 -218.17 (1.21) (2.63) (0.23) (276.96) Constant 102.27*** -35.65*** 4.33*** 2721.35*** (2.25) (3.78) (0.44) (846.72) Robust R Sqrd 0.84 0.54 0.40 0.34

Table 3-4. P-values representing statistical significance of Wald Tests comparing inband power model LU/LC indicator coefficients Comparison P-Value RP/CSC 0.41 RP/SFR 0.32 RP/Ind 0 RP/CBD 0.01 CSC/SFR 0.87 CSC/Ind 0 CSC/CBD 0.01 SFR/Ind 0 SFR/CBD 0.04 Ind/CBD 0.01

73

Figure 3-4. Time series of one minute LAeq sound levels over entire sampling period for each season and site. 55 dbA level is indicated with a black line

74

Table 3-5. Acoustic metrics for long term samples by site and season Leq_48hr DNL Leq_Day Leq_Night L10 L90 L10-L90 CONX Fall 42.5 50.1 42.3 42.8 47.8 29.8 18.0 Winter 40.6 40.4 42.5 26.1 44.0 19.3 24.7 Spring 43.9 53.0 42.1 45.8 47.3 30.7 16.6 Summer 52.0 56.2* 53.4 47.8 51.5 35.5 16.0

RPX Fall 44.9 48.9 46.4 40.6 47.9 37.0 10.9 Winter 45.3 50.2 46.5 42.2 48.1 37.8 10.3 Spring 48.0 53.1 49.2 45.1 48.4 42.1 6.3 Summer 49.6 56.1* 50.1 48.5 51.3 41.1 10.2

INDX Fall** 49.7 50.3 51.7 38.2 47.9 33.8 14.1 Winter 73.2* 81.0* 72.8 73.6 78.1 44.3 33.8 Spring 72.4* 74.8* 74.1 65.3 76.8 39.6 37.2 Summer 69.0* 76.6* 68.8 69.2 74.8 48.0 26.8

CBDX Fall 59.7* 64.4* 61.0 56.3 61.8 48.9 12.9 Winter 62.6* 65.6* 64.2 56.7 64.1 50.2 13.9 Spring 62.5* 64.4* 64.3 54.4 63.8 48.4 15.4 Summer** 59.4* N/A N/A N/A 61.4 54.4 7.0 Note: * Indicates sound levels that exceed OSHA or US EPA recommendations for human health. ** Indicates sample with limited data.

75

Table 3-6. P-values for Mann Whitney U tests comparing the Leq sample distributions during the nine am hour to the other hourly Leq distributions during the day Hour ConXF ConXW ConXSp ConXSu RPXF RPXW RPXSp RPXSu IndXF IndXW IndXSp IndXSu CBDXF CBDXW CBDXSp CBDXSu

7 0.10* 0.00 0.00 0.59* 0.00 0.00 0.78* 0.00 0.00 0.02* 0.54* 0.01* 0.00 0.00 0.00 N/A 8 0.04* 0.68* 0.00 0.36* 0.00 0.65* 0.07* 0.00 0.00 0.29* 0.39* 0.00 0.03* 0.22* 0.00 N/A 9 ------10 0.00 0.07* 0.00 0.44* 0.00 0.00 0.72* 0.00 0.00 0.16* 0.00 0.41* 0.02* 0.00 0.00 0.07* 11 0.11* 0.91* 0.01 0.00 0.00 0.00 0.01 0.07* 0.00 0.76* 0.00 0.00 0.32* 0.04* 0.00 0.03* 12 0.98* 0.19* 0.00 0.00 0.00 0.00 0.28* 0.00 0.63* 0.00 0.00 0.00 0.79* 0.00 0.00 0.73* 13 0.20* 0.07* 0.00 0.00 0.00 0.00 0.80* 0.15* 0.00 0.00 0.00 0.00 0.24* 0.00 0.00 0.39* 14 0.62* 0.45* 0.00 0.00 0.00 0.77* 0.25* 0.37* 0.00 0.68* 0.00 0.00 0.01 0.03* 0.00 0.82* 15 0.12* 0.85* 0.00 0.00 0.11* 0.00 0.89* 0.05* 0.09* 0.00 0.00 0.00 0.61* 0.00 0.00 0.27* 16 0.02* 0.00 0.00 0.09* 0.00 0.18* 0.25* 0.01 0.02* 0.00 0.00 0.00 0.04* 0.00 0.00 0.01* 17 0.09* 0.00 0.00 0.02* 0.00 0.78* 0.94* 0.00 0.31* 0.00 0.35* 0.00 0.00 0.41* 0.00 0.55* 18 0.07* 0.00 0.00 0.15* 0.00 0.00 0.91* 0.00 0.05* 0.00 0.00 0.00 0.00 0.01 0.00 0.00 19 0.00 0.00 0.00 0.03* 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.62* 0.00 0.00 20 0.00 0.00 0.00 0.05* 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.88* 21 0.00 0.00 0.26* 0.00 0.00 0.00 0.03* 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 N/A 22 0.00 0.00 0.15* 0.00 0.00 0.00 0.06* 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 N/A Notes: * Indicates the null hypothesis that the hour and the 0900 nine am hour Leq measures were from the same distribution failed to be rejected.

76

CHAPTER 4 DETERMINING ZONES OF ACOUSTIC DISTURBANCE ADJACENT TO ROADWAYS

Introduction

The network of roads in landscapes is extensive (Forman 2000; Miller 2002;

Nega et al. 2013) and a well-known source of ecological disturbance. Roads reduce the abundance of nearby wildlife (Forman and Alexander 1998; Fahrig and Rytwinski 2009;

Benitez-Lopez et al. 2010) through many mediums including direct road mortality, barriers to movement, and chemical pollution (Spellerberg 1998; Trombulak and Frissell

2000; Coffin 2007; Andrews et al. 2008). The area surrounding roads that suffers ecologically has been termed the road effect zone (REZ) (Forman and Alexander 1998).

Noise is thought to be one of the most far reaching and significant sources of road disturbance for wildlife (R. Reijnen et al. 1995; M. Reijnen et al. 1995). Research that disentangles road noise from the associated road disturbances has been limited

(Warren et al. 2006; Halfwerk et al. 2011). However, a recent study by Ware et al.

(2015) found that isolated road noise significantly decreased bird abundance and reduced body condition for those that endured the noise exposure. The known effects of anthropogenic noise on wildlife are physiological responses (ear damage, stress, immune response), behavioral responses (changes to foraging/vigilance trade-offs, changes in temporal or physical patterns to avoid noise), and signal masking (mate evaluation, territorial defense, parent/offspring communication, predator/prey cues) that result in fitness costs (Barber et al. 2011; Francis and Barber 2013).

Observed effects of road disturbance, such as areas of decreased biodiversity or abundance, and noise thresholds determined in the laboratory can be used to project the extent of REZs in other areas (Barber et al. 2011). Forman (2000) and Forman and

77

Deblinger (2000), using the data from R. Reijnen (1995), M. Reijnen et al. (1995) and R.

Reijnen (1996), estimated the extent of ecologically impacted areas based on bird community composition near roads. Nega et al. (2013) predicted the impact of roads on natural areas in Minnesota based on modeled traffic noise and an assumed 50 dBA threshold effect on bird communities. While these studies used either measured changes in community structure or assumed impact, an understanding of the actual mechanisms of noise impact may allow the demarcation of detailed REZs for specific groups of animals or even individual species.

Road Noise Response Thresholds

Noise thresholds for extreme effects on wildlife (e.g., hearing loss and ear damage) have been documented for many species (Reinis 1976; Bondello 1977; Ryals et al. 1999) but the thresholds are at noise levels not commonly found in the environment (Dooling and Popper 2007). Effects that result from traffic noise at lower sound levels are not as comprehensively understood but create disturbance over greater areas. Birds are the most studied group of animals in relation to noise thresholds, although anurans have also received some attention (Sun and Narins 2005;

Lengagne 2008; Eigenbrod et al. 2009).

Dooling and Popper (2007) provided a detailed review of the current understanding of the mechanisms of impacts on birds from road noise. The road noise thresholds and impacts discussed ranged from behavioral responses resulting from any audible road noise to physical ear damage at 110 dBA continuous noise. There is evidence that chaffinches reduce feeding and increase vigilance (Quinn et al. 2006) and tree swallow nestlings fail to respond to parent alarm calls (McIntyre et al. 2014) under white noise treatments. White-crowned sparrows also alter their foraging/vigilance trade

78

off in simulated road noise (Ware et al. 2015). The mechanisms for call masking in birds is better understood because the ability of birds to hear signals in noise has been well documented (Dooling et al. 2000). In addition, call signal source levels (Brackenbury

1979) and the dynamics of bird call propagation (Marten and Marler 1977; Dooling

1982) are known. Dooling and Popper (2007) considered the results from these studies and made a conservative traffic noise threshold of 50-60 dBA for birds in quiet residential areas.

Eigenbrod et al. (2009) quantified the REZ for anurans to be between 198 and

1417 meters based on differences in community structure resulting from the cumulative effect of road disturbance. Although altered anuran call activity resulting from traffic noise has been documented (Sun and Narins 2005; Lengagne 2008), the specific noise thresholds that induce these effects are not well studied. However, Bee and Swanson

(2007) tested the ability of female Cope’s Grey Tree Frogs (Hyla chrysoscelis) to orient towards male calls at varying levels of traffic noise. They found that the females could successfully orient towards a 67 dB male call at a threshold of 70 dB LCEQ (C-weighted equivalent sound level) traffic noise.

Problem Statement and Plan of Study

Laboratory and field studies have shown that isolated traffic noise has ill effects on wildlife (Bee and Swanson 2007; Lengagne 2008; Ware et al. 2015). An awareness of the mechanisms underlying these effects on wildlife and the thresholds at which they occur is increasing. However, this knowledge has yet to be applied to the creation of conservation buffers. Currently, estimates in the literature, of the appropriate widths of

REZs are based off roughly estimated noise thresholds or distance thresholds assumed from observations in other areas (Forman 2000; Nega et al. 2013). The aim of this study

79

was to calculate zones of ecological disturbance based on the current understanding of the mechanisms and thresholds of specific road noise impacts. To determine these zones of ecological disturbance, exposure limits from the prior literature were applied to both a case study of an interstate highway bordering conservation land, and modeled traffic noise for three theoretical traffic volumes.

Methods

Road Traffic Volumes

Three traffic volumes that represented realistic sources of constant road noise disturbance in Florida were generated based on 2015 average annual daily traffic

(AADT) reports (Florida Department of Transportation (FDOT) 2016). The three AADT volumes selected for analysis were 25,000, 80,000, and 200,000 vehicles per day. The case study analyzed a conservation area in north central Florida bordered by an interstate highway with an AADT of 56,500 (FDOT 2016). The traffic volumes used to calculate noise levels for each road were the 30th highest traffic hour of the year (Table

4-1).

Road Noise Sound Levels

Theoretical roadways

The noise emissions for the three theoretical roadways were based on the

Federal Highway Administration’s (FHWA) Traffic Noise Model (TNM). This model predicts noise emissions from roads based on an extensive collection of observed nationwide traffic noise collected from 1993-1995 (FHWA 2014). The user interface for the model was not utilized as it is limited to overall dBA noise levels. The model equations, as presented in the technical manual (Anderson et al. 1998) and subsequent

80

update sheet (Lau et al. 2004), were used to predict the octave band sound level at a

15-meter distance for each of the simulated roads.

Case study

A conservation area in north central Florida bordered by an interstate highway was selected for study. The traffic noise in the study area was measured along nine transects at distances of 40 (n=9), 80 (n=7), 160 (n=7), 320 (n=7), 640 (n=9), and 1280

(n=2) meters from the roadway (See Figure 4-1). The location of the start of each transect along a 1000-meter length bordering the interstate was randomly selected. The ambient noise environment was recorded at each distance for 20 minutes between

0900 and 1200 in 2013 and 2014. The recordings were collected with a Fostex fr-2le field recorder and a Seinheisser ME 62 omni-directional microphone at a 24-bit depth and 48 kHz sampling rate. The recording instrument chain was calibrated with a Cesva

SC310 sound level meter and the recorded amplitude measures were converted to sound pressure levels. The sound level meter was used to collect sound pressure measurements in addition to the WAV recordings during three transects to confirm the accuracy of the calibration procedures.

Road Noise Dissipation

Theoretical roadways

The dissipation of sound from three theoretical roadways was modeled using the following formula for the attenuation of traffic noise in an open area (Harris 1991):

푑푖 푑푖 (4-1) 퐴푡푒푛푢푎푡푖표푛 = 10 log ( ⁄ ) + 5 log ( ⁄ ) + 훼(푑푖 − 푑푟푒푓) + 훾(푑푖 − 푑푟푒푓) 푑푟푒푓 푑푟푒푓

81

Where 푑푖 is the distance from the road for predicted sound level I; 푑푟푒푓 is the distance from the road for the known noise levels (15 meters); 훼 is the air absorption coefficient;

훾 is the vegetation absorption coefficient.

The first term in Equation 4-1 accounts for attenuation from geometric spreading, the second term accounts for attenuation from ground interaction, and the third term accounts for attenuation from air absorption. The air absorption coefficient (훼) is dependent on frequency (Hz), temperature, and relative humidity. The fourth term accounts for vegetation attenuation. The vegetation coefficient (훾) is unique to each octave band and it was only applied for distances up to 200 meters to account for sound paths that travel above the vegetation (Harris 1991). The road noise sound levels over distance were modeled under realistic climatic conditions (10C, 90% relative humidity) for north central Florida that optimized the propagation of 1000 Hz, the dominant frequency of road noise (Anderson et al.1998). This ensured that the resulting ecologically unsuitable areas were large enough to dissipate harmful road noise during all weather conditions.

Case study

The attenuation of the road noise for the case study was modeled using SPreAD-

GIS (The Center for Landscape Analysis, Seattle, WA). SPreAD-GIS is a free Arc GIS toolbox that models noise propagation and allows for detailed, map-based consideration of vegetation cover, atmospheric temperature profile, wind, and topography (Reed et al.

2012). SPreAD-GIS is limited to frequencies from 355 to 2,239 Hz. Therefore, the analysis for the case study did not include the 4 kHz octave band. The model was run with ArcMap 10.3 (ESRI, Redland, CA) and calibrated with the observed sound levels

82

and climatic conditions during the transect data collection. The most recent Digital

Elevation Model (DEM) and land use shapefiles accessed through the Florida

Geographic Data Library (FGDL 2015) were used as inputs for the model.

Zones Of Disturbance Calculation

The zones of ecological disturbance were defined as the distance between the road and the points where road noise dissipated to sound levels that satisfied thresholds for negative effects of road noise on wildlife.

Average songbird communication zones of disturbance

Birds communicate to attract mates and defend territory (Marler and Slabbekoorn

2004). To be effective, the signal must be heard and understood by other birds after it has degraded over the communication distance. The bird communication disturbance zoens were calculated as the distance required to dissipate road noise to sound levels that would not mask communication. The zones were calculated separately for the 1 kHz, 2 kHz, and 4 kHz octave bands to span the frequency range of bird communication

(Dooling and Popper 2007). First, the call signal sound level at the source was determined. It was assumed to be 89.4 dB SPL at one meter, the average peak sound pressure level of seventeen song birds (Brackenbury 1979). The dissipation of the bird call was calculated for each octave band following the attenuation model for bird calls by Marten and Marler (1977) and Dooling (1982).

푑 퐸퐴×푑 (4-2) 푑푟표푝 = 20 log ( 푚푐 + 푚푐) 푑0 100

The distance of maximum communication (푑푚푐) was assumed to be 40 meters, an estimate of average territory distance for a song sparrow (Dooling et al. 2000). One meter was used for measured source intensity distance (d0) from Brakenbury (1979).

83

퐸퐴, excess attenuation, was calculated as the sum of molecular absorption, anonymous excess attenuation, and vegetation effects (Harris 1991).

The maximum road noise level was calculated by subtracting the critical ratio, the spectrum sound level difference between a signal and maximum background noise levels where the signal can be heard, from the bird call octave band sound levels at 40 meters. The median critical ratio for fourteen bird species (Dooling et al. 2000) was used. The zone of disturbance was defined as the distance from the road to the point where the maximum noise level was reached.

Recorded bird call zones of disturbance

The songs of four Tufted Titmice (Baeolophus bicolor) were recorded with a

Fostex fr-2le field recorder and a Seinheisser ME 62 omni-directional microphone within the case study area. The distance between the bird and microphone during the recording was measured with a Bushnell Legend 1200 ARC laser rangefinder. The average peak sound power and dominant frequency of each song was measured using

Raven Pro 64 1.4. The sound level of the dominant frequency for each song was estimated at 40 meters using Equation (4-2). The same procedure for determining maximum road noise level was completed as with the average songbird communication zones of disturbance for the octave band that included the dominant frequency.

Grey Tree Frog communication zones of disturbance

The Grey Tree Frog communication disturbance zone is based on the traffic noise threshold found to interrupt phonotaxis by female Cope’s Grey Tree Frogs (Hyla chrysoscelis) (Bee and Swanson 2007). This species is found throughout the southern

United States, including north Florida (Means and Simberloff 1987). Bee and Swanson

84

(2007) found that 67 dB LCEQ male calls could not be adequately received by females when 70 dB LCEQ traffic noise was present. Gerhardt (1975) found that the amplitude of the call of the Grey Tree Frog is 85 dB at one meter. Choruses are crowded places

(Gerhardt 2001) and a one-meter active call distance was assumed to be a reasonable requirement (Beckers and Schul 2004). The Grey Tree Frog’s call has two peak frequency harmonics, one between 1,000 and 1,500 Hz and the other between 2,000 and 2,800 Hz (Gerhardt 2001). The zone of disturbance requirements for Grey Tree

Frog communication were calculated within the 1,000 and 2,000 Hz octave bands to span the range of both spectral peaks in the call. The noise threshold was adjusted for a call of 85 dB and produced an allowable traffic limit of 68.9 dB for the 1,000 Hz band and 66.4 dB for the 2,000 Hz band.

Results and Discussion

Road Noise Dissipation

Figure 4-2 shows the dissipation of road noise over distance for the 200,000

AADT simulated road. The results of the simulation show that road noise is more intense in low frequencies and the higher frequency components have lower sound levels and higher rates of attenuation. These characteristics of road noise and anthropogenic noise in general, support the focus on low frequency noise in the literature. However, it also indicates that high frequency noise is present at close distances to the noise source, highlighting the importance of frequency range when analyzing interference with wildlife. Descriptions of noise and thresholds that use overall sound levels hide relevant information on the spectral shape of noise.

Figure 4-3 displays A-weighted sound levels for the three simulated roadways over linear perpendicular distance away from the road. The 200,000 AADT road ranged

85

from 2.72 to 2.88 dBA louder than the 80,000 AADT road at each distance. The differences between the 200,000 and 25,000 AADT roads at each distance ranged from

8.23 to 8.85 dBA. For reference, Lohr et al. (2003) found that the signal to noise threshold for budgerigars and zebra finches to distinguish between signals versus just detect a signal was an average of 3.29 dB higher. This infers that at the same distance, the noise difference between the 200,000 and 80,000 AADT roads is almost the same as the difference in noise that masks signal discrimination verses detection.

The road noise dissipation model (Equation 4-1) was also run without the vegetation term for the three simulated roads to reflect conditions where vegetation does not visually block the road (Harris 1991). The presence of vegetation exaggerated the attenuation of noise at all frequencies for the first 200 meters. Table 4-2 displays the difference in sound level with and without the vegetation term at all distances after 200 meters, by frequency. This result can be generalized for many conditions as the vegetation attenuation term is not affected by climate or road noise levels. It shows that a 200-meter buffer of dense vegetation can mitigate significant levels of road noise, especially in the high frequencies.

Bird Communication Zones Of Disturbance

Distances from the roads required for an average bird to hear another average bird call from 40 meters away is shown in Table 4-3. The results for each simulated road are shown in each column. Each row represents the distance from the simulated road, in meters, required to attenuate the road noise so that no interference occurs. The

1,000 Hz band required the largest zone. These results show that the effect of roads was dependent on which frequencies the bird of concern uses for communication. For

86

example, birds that communicate with a dominant frequency in the 4,000 Hz octave band would lose the least amount of useful communication area near roads.

Several studies have calculated traffic noise thresholds for birds. These results offer comparisons to the disturbance zone sizes that this study calculated based on call masking. The distance required to dissipate road noise from the three AADT volume roads (Figure 4-3) to the threshold levels from the literature are given in Table 4-4.

Reijnen et al. (1997) calculated threshold traffic noise values based on bird densities within woodlands bordering roads to be between 42 and 52 dBA. Ware et al.

(2015) found that white crown sparrows altered the tradeoff between vigilance and foraging when subject to 55 dBA of traffic noise in the laboratory. These traffic levels incite larger zones of disturbance (Table 4-4) than those calculated to protect against call masking (Table 4-3). These results indicate that behavioral effects of road noise on birds might persist at further distances from roads than call masking effects.

Lohr et al. (2003) utilized their results on Budgeriger and Zebra Finches hearing abilities to estimate maximum call distances for these species in traffic noise. The traffic noise threshold for detection of a call 40 meters away was 82 dBA for Budgerigers and

80 dBA for Zebra Finches (Lohr et al. 2003). These traffic noise levels exceed the modeled noise levels of the 25,000 and 80,000 AADT roads (Figure 4-3). In our study, the modeled 200,000 AADT road reached 82 and 80 dBA at a distance of 16 and 21 meters, respectively (Figure 4-3). These distances are much smaller than our call masking zone of disturbance estimates in Table 4-3. The call characteristics used by

Lohr et al. (2003) to model the call propagation reflected maximum communication distances. Our study used a call source sound level of 89.4 dB, the average peak sound

87

pressure level found by Brakenbury (1979), which is smaller than 95 dB used by Lohr et al (2003). In addition, we used higher rates of excess attenuation to reflect a vegetated area, and modeled the road noise for an average day in a warm climate, conditions favorable for road noise propagation. The differences in the results of our two studies reflect the variation in communication ability that results from individual, climate, and habitat characteristics.

Grey Tree Frog Communication Zone Of Disturbance

The tree frog communication disturbance zones (Table 4-5) are much smaller than those required for bird communication. Frog choruses can be dense and loud with intense competition between males (Gerhardt and Huber 2002). It is inherent that frogs can communicate in road noise more effectively than birds. However, less is known about anuran communication in traffic noise than birds. These results only reflect one specific effect on one species. For example, Eigenbrod et al. (2009) quantified the REZ for seven anuran species to be between 198 and 1,417 meters based on species relative abundance controlling for habitat effects. This indicates that other effects of roads and road noise might reach further than phonotaxis interruption.

Case Study

Observed sound levels

Figure 4-4 displays the sound pressure levels for the 500, 1,000, 2,000, 4,000, and 8,000 Hz octave bands at the case study sample sites. The 500, 1,000, and 2,000

Hz bands’ sound levels decreased with distance from the road. The 500 and 1,000 Hz octave band noise, assumed to be attributable to traffic noise as seen in Figure 4-2, dominated the soundscape at closer distances but as its intensity decreased further away from the road, noises in the 4,000 and 8,000 Hz octave bands are more

88

influential. The 4,000 and 8,000 Hz bands initially increased in sound level and then fell after 160 meters. This region had higher variance than the lower frequencies. The frequency region of the 4,000 and 8,000 octave bands is occupied most frequently by wildlife vocalizations (Joo 2009; Krauss et al. 2011; Pijanowski et al. 2011a, b) which are known to be more complex than road noise (Pieretti and Farina 2013).

The sound levels at 1,280 meters did not follow the pattern established at the other distances. The sound levels of the 500 and 1,000 Hz bands were higher than expected while the 4,000 and 8,000 Hz bands had lower than expected sound levels.

Sound levels 1,280 meters away from the interstate were only measured twice and both recordings were completed in December when foliage density and wildlife acoustic activity would be expected to be lower than during other seasons. This could have contributed to high traffic noise propagation (high noise in low frequencies) and low levels of acoustic activity in and above the 4,000 Hz region, the frequency range used by most birds (Slabbekoorn and Ripmeester 2008).

In general, the variance of the observed sound levels at each distance was high.

Weather, elevation, and vegetation structural differences across transects likely contributed to the inconsistencies. Halfwerk et al. (2011) also found inconsistent sound levels with distance in a road side forested area in the Netherlands.

SPreAD-GIS results

The extent of the zones of disturbance for the average bird, tree frog, and recorded Tufted Titmouse calls are superimposed on the road traffic noise model results for the 1,000 and 2,000 Hertz (Hz) octave bands in Figure 4-5 and 4-6, respectively.

The background of the figures shows the sound levels in a black to white color scheme, black being the most intense (80 dB). The 1,000 Hz noise was more intense at further

89

distances than the 2,000 Hz noise as was seen in the simulated road results in Figure

4-2.

Figures 4-5 and Figure 4-6 show that the modeled noise levels did not vary consistently with distance from the road as was seen in the observed sound level data in Figure 4-4. As the model was run under consistent weather conditions, the differences in sound levels across the length of the road were due to elevation and vegetation differences. The 1,000 Hz average bird disturbance zone ranged from 184 to

288 meters in perpendicular distance from the interstate. The 2,000 Hz average bird zone of disturbance was smaller. It ranged from 77 to 196 meters. The Hyla

Chrysoscelis zone of disturbance was also larger for the 1,000 Hz noise, ranging from

63 to 119 meters, compared to the 2,000 Hz zone (29-75m). The Tufted Titmouse song’s peak frequency was in the 2,000 Hz octave band and therefore, the communication disturbance zone was calculated for traffic noise in this frequency region

(Figure 4-6). The Titmouse zone of disturbance ranged from 41 to 92 meters from the road, which was smaller than the average bird zone. The zones of disturbance for the case study were larger than the zones calculated from traffic noise modeled with

Equation 4-1 because the SPreAD-GIS model considered wind and temperature inversion that can aid in the spread of noise. The case study results reflect reasonable but maximized traffic noise levels expected for the area. The results in this study show that the area up to 288 meters from the roadways was acoustically unsuitable habitat for the wildlife assessed.

Conclusion

This study explored the role of traffic noise on wildlife conservation by estimating the distance required to dissipate road noise from four traffic volumes to levels required

90

for aspects of successful communication. The spectral shape of traffic noise and animal communication signals were an important consideration in the analysis. In addition, differences in vegetation and climate had dramatic effects on the propagation of traffic noise. The average bird required a distance of 128 meters from the 200,000 AADT simulated road to communicate in the 1,000 Hz band. The Grey Tree Frog only required a 55-meter distance for females to successfully orient towards males under the same noise stimulus. The results of the case study indicated that even wildlife in a protected area were subject to communication noise effects up to 288 meters from a bordering interstate. These disturbance zones were smaller than distances estimated to protect against other effects, including behavioral changes. This study exemplified the application of noise thresholds from the literature to delineate ecologically unsuitable areas for specific species and animal groups subject to generalized traffic volumes and a specific traffic noise source.

The results in this study are an application of the best knowledge available on road noise effects on wildlife. Unfortunately, significant research remains to be done to increase understanding of road noise thresholds. Evidence suggests that behavioral effects might persists at even further distances from roads. To address the gaps in knowledge, studies documenting the response of wildlife to road noise should report the spectrum level of the stimulus so that the results can be applied to other circumstances.

In addition, testing multiple stepped sound levels would identify noise thresholds more accurately than a single stimulus sound level.

Conservation plans should consider the acoustic environment and apply appropriate buffer distances between roads and conservation areas or ecologically

91

sensitive areas to insure needs of wildlife are met. The results from this study indicate that environmental areas sensitive to road noise should be buffered from interstate roads for at least 288 meters, larger if the area is not forested. In addition, forest fragmentation resulting from road edge effects could allow more sound penetration. increasing the ecologically degraded area. Any application of buffers should be regularly updated to reflect the most current understanding of noise levels that harm wildlife.

The methods and results of this study have the potential to inform the design of future conservation buffers that aim to protect sensitive environmental areas from road noise. Conservation buffers are protected areas that control destructive external influences (e.g., pollutants or erosion) before they reach a targeted, sensitive environmental area (Castelle et al. 1994; Wenger 1999). The results from this case study indicated that a zone up to 288 meters away from the interstate was inappropriate for wildlife use, essentially reducing the size of the adjacent conservation area. In this case, the conservation area closest to the road acted as a buffer for the rest of the conservation area, not as habitat that could be used effectively by wildlife. Further, the methods of this study can be used proactively to inform the location of a future road.

From the predicted traffic volumes for the new road, the resulting road noise and zones of disturbance can be modeled for the alternative road locations. The ideal location of the road would minimize conflicts between the zones of disturbance and sensitive environmental areas. The results of this study highlight the importance of considering the effect of road noise on wildlife for conservation management and the methodologies presented are valuable tools that allow this consideration to come to fruition.

92

Table 4-1. Traffic volume data used to model traffic noise for the three simulated roadways and observed traffic volume data for the case study interstate in 2015 Road Case Study AADT 25000 80000 200000 56500 Cars/hr 2183 6290 11928 4764 Med Truck/hr 25 186 350 427 Heavy Truck/hr 40 303 570 693 Buses/hr 2 13 24 29 Motorcycle/hr 3 26 48 58

Figure 4-1. Road noise sample locations for case study

93

Figure 4-2. Octave band sound levels for 200,000 AADT simulated roadway at distances up to 5,000 meters from the road

Figure 4-3. A-weighted sound levels for all three simulated roadways at distances up to 5,000 meters from the road

94

Table 4-2. The decrease in octave band sound levels that result from a 200-meter buffer of dense vegetation bordering a roadway Octave Band Sound Level Difference (dB) 63 Hz 3.70 125 Hz 5.55 250 Hz 6.48 500 Hz 7.40 1000 Hz 9.25 2000 Hz 11.10 4000 Hz 14.80 8000 Hz 22.20

Table 4-3. The buffer distance requirements for an average bird to hear another average bird call with a dominant frequency within three octave bands for each of the simulated roadways AADT Octave Band 1000 Hz 2000 Hz 4000 Hz 25,000 61 m 36 m 15 m 80,000 103 m 63 m 29 m 200,000 128 m 80 m 38 m

Table 4-4. Buffer distances converted from noise thresholds found in other studies for the three simulated roads Threshold Effect AADT 25k 80k 200k 42-52 dBA Reduced bird density in wooded area near road1 146- 226- 327- 395m 789m 1065m 55 dBA White crown sparrow foraging/vigilence behavior change2 116m 176m 219m 80 dBA Zebra finch max communication of 40m3 N/A N/A 21m 82 dBA Budgeriger max communication of 40m3 N/A N/A 16m

1 Reijnen et al. (1997) 2 Ware et al. (2015) 3 Lohr et al. (2003)

95

Table 4-5. The buffer distance requirements that ensure the Cope’s grey tree frog does not experience phonotaxis disruption as a result of traffic noise from each simulated roadway for the 1,000 and 2,000 Hz octave bands AADT Octave Band 1000 Hz 2000 Hz 4000 Hz 25,000 61 m 36 m 15 m 80,000 103 m 63 m 29 m 200,000 128 m 80 m 38 m

Figure 4-4. Sound pressure levels measured within the case study area averaged for each octave band (500-8,000 Hz) and distance from the road

96

Figure 4-5. 1,000 Hz octave band sound levels in grey scale with the Cope’s grey tree frog (in blue) and average bird (in green) buffers depicting where the road noise levels reach the effect thresholds

97

Figure 4-6. 2,000 Hz octave band sound levels in grey scale with the Cope’s grey tree frog (in blue), Tufted Titmouse (in pink), and average bird (in green) buffers depicting where the road noise levels reach the effect thresholds

98

CHAPTER 5 SUMMARY AND CONCLUSIONS

Key Findings

The overall objective of this dissertation was to understand the role of anthropogenic noise in the landscape. This goal was addressed through the following key findings of this study:

 The LDI Index is a good and efficient predictor of anthropogenic noise in soundscapes. The LDI index correlated well with soundscape characteristics indicative of anthropogenic noise and additional road and wetland proximity descriptors did not improve model performance. Compared to a previous study, the LDI index outperformed a suite of landscape descriptors as predictors of soundscape characteristics.

 Confirmation that the 2,000 Hz is the best 1,000 Hz incremented distinction between the anthropogenic and animal sourced portions of the soundscape. However, there was indication that the soundscape reflects anthropogenic influence up to 5,000 Hertz and in the 9,000-10,000 Hz band.

 Frequency overlap between anthropogenic noise and wildlife noise was identified

 Landscape development intensity within 100 meters of the recording point most strongly correlated with overall sound power. Alternatively, measured landscape development intensity within 500 meters was more strongly correlated with soundscape metrics that consider the distribution of sound power across the frequency spectrum.

 30-minute recordings during the 0900 hour revealed similarities and differences between the sound sources present in the soundscapes of different LU/LC classes. The conservation and industrial LU/LC classes were different qualitatively and statistically from the other LU/LC classes. Two pairs of LU/LC classes had statistically and qualitatively similar soundscapes: single family residential/recreational park and commercial shopping center/central business district.

 Overall sound power, or inband power, distinguished between the LU/LC classes the most of the soundscape metrics measured. Landscape development intensity and LU/LC class had separate and significant effects on inband power. The IND and CBD LU/LC classes had significantly different impacts on inband power than the other LU/LC classes. This indicates that anthropogenic influence has a general effect on soundscapes as well as a LU/LC class specific effect.

99

 The sound pressure levels of the 0900 hour are significantly similar to an average of 5.25 other hours within the morning and afternoon hours at the 48- hour sites.

 The CBDX and INDX temporal sound pressure levels exceeded the 55 dBA threshold suggested to protect birds from traffic noise roughly half of the sampling time. They also exceeded the EPA and WHO guidelines for human health.

 RPX and CONX sites only exceeded the 55 dBA guideline 1% of the sampling time. However, these sites violated the EPA and WHO guidelines during the summer samples. There were similar temporal patterns between the RPX and CONX sites but the RPX site had higher ambient sound levels that likely contributed to the overall patterns being subtler than at CONX.

 Traffic noise varies substantially by frequency, and differences in vegetation climate have dramatic effects on its propagation.

 The average bird was predicted to require a 63m, 103m, and 128m from a 25,000 AADT, 80,000 AADT, and 200,000 AADT road, respectively before it could successfully communicate with another bird of average hearing.

 The Cope’s Grey Tree Frog was predicted to require 20m, 40m, 55m from a 25,000 AADT, 80,000 AADT, and 200,000 AADT road, respectively before a female could successfully orient towards male signals.

 The zones of disturbance calculated for communication disruption were smaller than buffers estimated from noise thresholds found to effect bird behavior and reduce bird and anuran densities in other studies.

 The case study showed that the conservation area of concern had substantial area (up to 288 meters from the road) that was acoustically unsuitable for the wildlife tested. The size of the unsuitable area varied along the length of the road due to road noise differences resulting from elevation and vegetation structure.

Implications and Future Research

The role of anthropogenic noise in landscapes is only just starting to be recognized and understood. This dissertation observed anthropogenic noise within soundscapes across a spectrum of intensity, compared the soundscapes of specific

LU/LC classes, and estimated buffer distances to protect birds and a frog from communication disruptions caused by road traffic noise.

100

Through this work, the LDI index was identified as having high potential in the soundscape ecology field. It is a single measure that encompasses a lot of information about landscape configuration and the intensity of surrounding energy use. It was a better, more efficient predictor of soundscape patterns than suites of landscape descriptors in other studies. Also, it only requires site information that can be easily obtained remotely from aerial photographs and land use map files. This indicates that the LDI index is an accurate, cheap, and efficient way to identify noise in the landscape.

The LDI has a lot of promise as a tool to be used in further soundscape ecology work and noise mitigation.

The identification of conflicts between anthrophony and wildlife throughout this dissertation presents opportunities for further exploration. The 2,000 hertz distinction between human and wildlife sourced sounds often cited in the literature was confirmed.

However, the identification of frequency ranges above this (2,000-5,000 Hz, 9,000-

10,000 Hz) that were highly influenced by anthropogenic influence are of concern.

Animals that communicate below 2,000 Hz as well as the higher frequencies identified should be targeted as subjects of noise effect studies. In addition, industrial sites and central business districts were identified as LU/LC classes that had unique sounds that have high potential for conflict with wildlife.

Urban landscapes are heterogeneous conglomerates of noise. Results from

Chapter Two supported this characteristic and encouraged characterizing adjacent landscapes as one, aggregate source of noise measured through the LDI index.

However, the results also indicated that the area immediately surrounding the soundscape (100 and 500 meters) was the most strongly correlated with soundscape

101

characteristics. Chapter Three supported this finding by showing that individual surrounding LU/LC classes have additional, significant impact on soundscape power, independent of surrounding overall landscape development intensity. The differences in spectral distribution of power in LU/LC soundscapes were observed in the site recordings, but they were not statistically significant when overall land use intensity was accounted for. It is my opinion that these differences would be stronger statistically with a larger sample of sites and time recorded. The 48-hour results in Chapter Three found that samples at key times of day can represent larger timeframes. In addition, there are automatic sampling recorders available that regularly sample soundscapes over extended periods of time. This technology was unavailable for this study but the sound power distribution differences between LU/LC classes should be explored with a larger dataset that spans a full 24-hour period.

The results of Chapter 4 show the promising application of a comprehensive understanding of a noise effect. The methodology calculates zones of disturbance for each noise effect, octave band, and species. Additionally, the properties of road noise and how it propagates through the environment is well known. Road noise can be modeled with only minimal field measures to confirm estimated sound levels. As a result, calculating noise buffers is an accessible and sophisticated tool that has value for conservation managers. The methods and results in Chapter 4 show how conservation buffers can be calculated for general animal groups (average bird) or specific species

(Tufted Titmouse, Cope’s Gray Tree Frog). These procedures can be tailored for a specific species of concern or to protect as much of the community as possible depending on conservation objectives. Wildlife noise effects is a popular area of

102

research and the literature is constantly expanding. As the science advances, conservation buffers will be able to be calculated for more noise effects and wildlife.

103

APPENDIX A REGRESSION RESULTS

Multiple OLS regressions for each dependent soundscape variable were used to determine if the supplementary landscape descriptor variables added explanatory power for the soundscape metrics. The regression results are depicted for models with each dependent soundscape metric and the LDI variable selected for best performance in Tables A-1 through

Table A-1. Regression results for models with inband power dependent variables and LDI100 and supplemental landscape independent variables Inbnd Pwr Inbnd Pwr Inbnd Pwr Inbnd Pwr Inbnd Pwr Inbnd Pwr Inbnd Pwr LDI100 0.48*** 0.46*** 0.50*** 0.46*** 0.43*** 0.48*** 0.43*** (0.05) (0.04) (0.07) (0.06) (0.05) (0.05) (0.05) Wetland -2.44 (1.51) Road Prox 0.00 (0.01) Winter

Spring -2.30 (2.63) Summer -4.33 (5.53) Fall -5.21*** (1.74) Temp -0.20*** -0.20*** (0.07) (0.07) Humid 0.99 -0.64 (4.47) (4.64)

Constant 101.29*** 102.70*** 101.27*** 104.44*** 115.09*** 100.47*** 115.69*** (1.25) (1.36) (1.28) (2.38) (5.07) (3.58) (6.38) R-sqrd 0.58 0.59 0.58 0.63 0.62 0.58 0.62 AIC 439.42 439.47 441.19 436.74 433.74 441.39 435.72

104

Table A-2. Regression results for models with inband power dependent variables and LDI500 and supplemental landscape independent variables Inbnd Pwr Inbnd Pwr Inbnd Pwr Inbnd Pwr Inbnd Pwr Inbnd Pwr Inbnd Pwr LDI500 0.52*** 0.48*** 0.51*** 0.54*** 0.46*** 0.52*** 0.46*** (0.05) (0.05) (0.07) (0.08) (0.05) (0.05) (0.05) Wetland -3.28** (1.62) Road Prox 0.00 (0.00) Winter

Spring 0.67 (3.38) Summer -4.6 (5.35) Fall -4.93*** (1.84) Temp -0.38*** -0.36*** (0.13) (0.13) Humid 7.03 4.71 (5.24) (5.27)

Constant 99.18*** 101.16*** 99.21*** 180.50*** 107.50*** 93.15*** 103.11*** (1.48) (1.52) (1.50) (3.24) (5.41) (4.77) (5.66) R-sqrd 0.53 0.55 0.53 0.59 0.58 0.54 0.59 AIC 446.54 445.22 448.45 442.85 440.67 447.15 441.99

105

Table A-3. Regression results for models with inband power dependent variables and LDI1500 and supplemental landscape independent variables Inbnd Pwr Inbnd Pwr Inbnd Pwr Inbnd Pwr Inbnd Pwr Inbnd Pwr Inbnd Pwr LDI1500 0.50*** 0.45*** 0.46*** 0.48*** 0.43*** 0.50*** 0.43*** (0.06) (0.06) (0.07) (0.10) (0.07) (0.06) (0.07) Wetland -5.09*** (1.83) Road Prox 0.01 (0.00) Winter

Spring -3.76 (3.56) Summer -8.4 (5.60) Fall -6.86*** (2.21) Temp -0.53*** -0.53*** (0.16) (0.16) Humid 4.78 1.73 (5.98) (6.41)

Constant 99.92*** 102.77*** 100.17*** 104.27*** 111.24*** 95.89*** 109.67*** (1.80) (2.10) (1.76) (3.87) (3.99) (5.17) (6.91) R-sqrd 0.36 0.42 0.37 0.46 0.47 0.37 0.47 AIC 467.36 463.08 468.22 462.28 456.49 468.89 458.42

106

Table A-4. Regression results for models with delta power dependent variables and LDI100 and supplemental landscape independent variables Delta Pwr Delta Pwr Delta Pwr Delta Pwr Delta Pwr Delta Pwr Delta Pwr LDI100 -0.43*** -0.44*** -0.39*** -0.44*** -0.41*** -0.44*** -0.41*** (0.06) (0.06) (0.07) (0.07) (0.06) (0.06) (0.06) Wetland -0.89 (1.86) Road Prox -0.01 (0.01) Winter 0.44 (3.00) Spring

Summer 2.56 (2.24) Fall 1.22 (2.43) Temp 0.19 0.21 (0.17) (0.17) Hum 5.64 6.60 (5.63) (5.97)

Constant -42.25*** -41.74*** -42.33*** -42.76*** -46.37*** -46.95*** -52.21*** (1.49) (1.83) (1.49) (1.64) (3.71) (4.67) (6.22) R-sqrd 0.44 0.44 0.46 0.45 0.46 0.45 0.46 AIC 462.08 463.89 462.41 467.60 462.51 463.37 463.54

107

Table A-5. Regression results for models with delta power dependent variables and LDI500 and supplemental landscape independent variables Delta Pwr Delta Pwr Delta Pwr Delta Pwr Delta Pwr Delta Pwr Delta Pwr LDI500 -0.49*** -0.50*** -0.44*** -0.54*** -0.46*** -0.49*** -0.46*** (0.05) (0.05) (0.07) (0.08) (0.06) (0.05) (0.06) Wetland -0.31 (2.04) Road Prox -0.01 (0.01) Winter 2.65 (3.39) Spring

Summer 4.52* (2.57) Fall 2.93 (2.85) Temp 0.19 0.19 (0.17) (0.17) Humid 0.10 1.35 (5.18) (5.60)

Constant -39.65*** -39.47*** -39.88*** -40.46*** -43.87*** -39.73*** -45.13*** (1.52) (1.74) (1.52) (1.56) (3.89) (4.56) (6.11) R-sqrd 0.45 0.45 0.48 0.47 0.47 0.45 0.47 AIC 460.78 462.76 460.02 465.06 461.22 462.78 463.18

108

Table A-6. Regression results for models with delta power dependent variables and LDI1500 and supplemental landscape independent variables Delta Pwr Delta Pwr Delta Pwr Delta Pwr Delta Pwr Delta Pwr Delta Pwr LDI1500 -0.52*** -0.50*** -0.44*** -0.55*** -0.47*** -0.52*** -0.47*** (0.06) (0.06) (0.07) (0.10) (0.07) (0.06) (0.07) Wetland 1.34 (2.14) Road Prox -0.01** (0.01) Winter 0.51 (3.53) Spring

Summer 6.49** (3.17) Fall 2.61 (3.01) Temp 0.33* 0.34* (0.19) (0.19) Humid 2.32 4.30 (5.96) (6.72)

Constant -39.33*** -40.08*** -39.81*** -39.79*** -46.38*** -41.29*** -50.27*** (1.75) 2.16206 (1.74) (1.76) (4.29) (5.21) (7.55) R-sqrd 0.36 0.36 0.40 0.38 0.40 0.36 0.41 AIC 471.29 472.90 469.04 474.77 468.87 473.18 470.49

109

Table A-7. Regression results for models with center frequency dependent variables and LDI100 and supplemental landscape independent variables Cntr Freq Cntr Freq Cntr Freq Cntr Freq Cntr Freq Cntr Freq Cntr Freq LDI100 -49.72*** -49.73*** -48.59*** -41.27*** -48.69*** -51.27*** -49.55*** (14.72) (15.97) (15.50) (12.72) (15.43) (14.71) (15.35) Wetland -0.93 (468.47) Road Prox -0.19 (0.45) Winter 392.20* (232.23) Spring 1023.74** (424.08) Summer

Fall 450.88 (277.27) Temp 6.95 11.82 (26.69) (26.95) Hum 1961.39** 2015.77** (928.77) (959.79)

Constant 2176.25*** 2176.79*** 2174.38*** 1404.55*** 2027.96* 542.75 245.02 508.65 584.26 512.76 (303.00) (733.04) (660.02) (996.13) R-sqrd 0.23 0.23 0.23 0.26 0.23 0.26 0.26 AIC 1162.63 1164.63 1164.60 1166.15 1164.57 1162.16 1163.99

110

Table A-8. Regression results for models with center frequency dependent variables and LDI500 and supplemental landscape independent variables Cntr Freq Cntr Freq Cntr Freq Cntr Freq Cntr Freq Cntr Freq Cntr Freq LDI500 -66.16*** -66.52*** -66.05*** -62.54*** -66.47*** -65.87*** -65.43*** (18.09) (19.73) (19.26) (17.39) (19.65) (17.82) (19.46) Wetland -33.80 (465.94) Road Prox -0.02 (0.46) Winter 506.83** (212.02) Spring 694.33** (333.84) Summer

Fall 447.98 (286.10) Temp -1.87 2.61 (28.80) (29.94) Humid 1295.41* 1312.25 (713.09) (791.01)

Constant 2729.84*** 2750.21*** 2729.40*** 2119.76*** 2771.25*** 1618.33** 1546.03 (609.83) (696.38) (616.67) (471.47) (903.64) (722.43) (1227.55) R-sqrd 0.32 0.32 0.32 0.33 0.32 0.34 0.34 AIC 1154.07 1156.07 1156.07 1159.28 1156.07 1154.85 1156.84

111

Table A-9. Regression results for models with center frequency dependent variables and LDI1500 and supplemental landscape independent variables Cntr Freq Cntr Freq Cntr Freq Cntr Freq Cntr Freq Cntr Freq Cntr Freq LDI1500 -71.44*** -69.83*** -69.03*** -62.76*** -69.01*** -71.91*** -68.84*** (22.10) (23.39) (23.19) (21.29) (22.75) (21.79) (22.47) Wetland 171.46 (469.43) Road Prox -0.36 (0.47) Winter 32.52 (161.74) Spring 479.96 (393.32) Summer

Fall 185.55 (280.91) Temp 17.69 22.60 (26.86) (27.63) Humid 1598.07** 1728.74* (801.59) (885.91)

Constant 2833.09*** 2737.24*** 2818.87*** 2399.74*** 2456.53*** 1483.58* 892.01 (707.44) (767.21) (714.33) (672.24) (863.55) (823.03) (1199.54) R-sqrd 0.27 0.28 0.28 0.28 0.28 0.29 0.30 AIC 1158.71 1160.49 1160.59 1163.88 1160.28 1158.97 1160.26

112

Table A-10. Regression results for models with aggregate entropy dependent variables and LDI100 and supplemental landscape independent variables Ag Entropy Ag Entropy Ag Entropy Ag Entropy Ag Entropy Ag Entropy Ag Entropy LDI100 -0.03*** -0.03*** -0.03*** -0.03*** -0.03*** -0.03*** -0.03*** (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Wetland -0.17 (0.23) Road Prox 0.00 (0.00) Winter 0.30 (0.32) Summer 0.31 (0.22) Fall 0.50* (0.28) Spring

Temp 0.01 0.01 (0.01) (0.01) Hum 0.95 1.01 (0.67) (0.68)

Constant 3.73*** 3.83*** 3.73*** 3.56*** 3.54*** 2.94*** 2.64*** (0.18) (0.21) (0.18) (0.21) (0.34) (0.61) (0.63) R-sqrd 0.27 0.28 0.27 0.32 0.28 0.29 0.30 AIC 153.98 155.30 155.77 155.82 155.63 153.97 155.39

113

Table A-11. Regression results for models with aggregate entropy dependent variables and LDI500 and supplemental landscape independent variables Ag Entropy Ag Entropy Ag Entropy Ag Entropy Ag Entropy Ag Entropy Ag Entropy LDI500 -0.04*** -0.04*** -0.04*** -0.05*** -0.04*** -0.04*** -0.04*** (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Wetland -0.16 (0.24) Road Prox -0.00 (0.00) Winter 0.59* (0.35) Spring 0.54** (0.25) Summer 0.72** (0.32) Fall

Temp 0.01 0.01 (0.02) (0.01) Humid 0.56 0.61 (0.61) (0.61)

Constant 4.00*** 4.10*** 4.00*** 3.82*** 3.87*** 3.52*** 3.30*** (0.18) (0.22) (0.18) (0.18) (0.37) (0.55) (0.60) R-sqrd 0.34 0.34 0.34 0.42 0.34 0.34 0.35 AIC 147.82 149.14 155.43 144.71 149.65 149.06 150.76

114

Table A-12. Regression results for models with aggregate entropy dependent variables and LDI1500 and supplemental landscape independent variables Ag Entropy Ag Entropy Ag Entropy Ag Entropy Ag Entropy Ag Entropy Ag Entropy LDI500 -0.04*** -0.04*** -0.04*** -0.05*** -0.04*** -0.04*** -0.04*** (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Wetland -0.03 (0.25) Road Prox -0.00 (0.00) Winter 0.43 (0.36) Spring 0.74** (0.33) Summer 0.71* (0.36) Fall

Temp 0.02 0.02 (0.01) (0.01) Humid 0.73 0.84 (0.65) (0.69)

Constant 4.03*** 4.05*** 4.01*** 3.91*** 3.67*** 3.42*** 2.90*** (0.21) (0.24) (0.21) (0.23) (0.35) (0.61) (0.65) R-sqrd 0.27 0.27 0.28 0.36 0.28 0.28 0.30 AIC 154.17 156.15 155.43 151.92 154.85 155.00 155.28

Partial and Adjusted R Squared values were used to compare the additional explanatory power that is provided by adding supplemental landscape variable(s), conditional on the inclusion of LDI index measures. Table A-13 demonstrates these results from the variable selection for the linear regression models for inband power, aggregate entropy, and delta power. Partial and Adjusted R Squared values could not be calculated for the center frequency models because the data is heteroscedastic. The model selection for this variable relied solely on AIC and R Squared values.

115

Table A-13. Partial and Adjusted R Squared values for supplemental proxy variables in inband power, aggregate entropy, and delta power models estimated with LDI100 and LDI500. Inband Power Entropy Delta Power LDI Scale Proxy Variable(s) Partial R Sqrd Adjusted R Sqrd Partial R Sqrd Adjusted R Sqrd Partial R Sqrd Adjusted R Sqrd 100m None 0.57 0.26 0.43 100m Wetland 0.03 0.58 0.01 0.26 0.00 0.43 100m Road 0.00 0.57 0.00 0.25 0.02 0.44 100m Wetland/Rd 0.03 0.57 0.01 0.25 0.03 0.43 100m Temp/Hum 0.11 0.61 0.04 0.27 0.04 0.44 100m Seasons 0.12 0.61 0.06 0.27 0.01 0.41 500m None 0.52 0.33 0.45 500m Wetland 0.05 0.54 0.01 0.32 0.00 0.44 500m Road 0.00 0.52 0.00 0.32 0.04 0.46 500m Wetland/Rd 0.05 0.53 0.01 0.31 0.04 0.45 500m Temp/Hum 0.12 0.57 0.02 0.32 0.02 0.44 500m Seasons 0.12 0.56 0.13 0.38 0.03 0.43

116

APPENDIX B SPreAD-GIS RESULTS

The case study SPreAD-GIS results for the interstate noise propagation in the conservation area were run under multiple weather conditions. The resulting sound levels at six of the road noise sampling locations (Figure 4-1) are given in Table B-1 for the 500 Hz band.

Table B-1. The 500 Hz band sound pressure levels (db) at six sites modeled by SPreAD-GIS for varying weather conditions 500 Hertz Band Traffic Sound Pressure Level (dB) Temp Relative Hum. Conditions Site1 Site2 Site3 Site4 Site5 Site6 92F 100% Clear, windy Day 58.37 28.88 24.00 57.04 27.34 23.99 92F 100% Clear, windy Night 58.37 28.88 24.00 57.04 27.34 24.00 92F 100% Clear,calm Night 60.12 28.88 24.00 58.14 27.34 24.00 92F 100% Clear, calm Day 46.85 25.24 0 51.04 24.27 0 25F 24% Clear, windy Night 57.65 26.89 0 56.34 26.00 0 56F 81% Clear, windy Night 58.10 29.18 24.00 56.85 27.56 24.00 68F 81% Clear, windy Night 58.29 28.45 24.00 56.95 27.04 23.98 80F 87% Clear, windy Night 58.33 28.66 24.02 56.99 27.19 23.99 71F 89% Clear, windy Night 58.35 28.77 24.03 57.02 27.27 23.99 50F 90% Clear, calm Night 60.19 29.36 24 58.23 27.69 24.03 50F 90% Clear, windy Night 58.45 29.36 24 57.14 27.69 24.03 50F 100% Clear, calm Night 60.1919 29.35 24.07 58.22 27.68 24.01 50F 100% Clear, calm Day 46.94 25.43 24 51.14 24.34 0 50F 90% Clear, calm Night 46.94 25.43 24.01 51.14 24.34 0

117

APPENDIX C SITE LOCATIONS

The GPS location in decimal degrees for each site included in this dissertation are provided in Table C-1 and C-2. Table C-1 includes the locations where data was collected for Chapter 2 and 3. Table C-2 includes the locations where data was collected for Chapter 3.

Table C-1. Site type and GPS location for Chapter 2 and 3 data Site Name LU/LC Type GPS Decimal Degrees Jennings CON 30.13820,-81.96320 Osceola CON 30.39594,-82.40459 Long Leaf CON 29.56616,-82.19370 Ocala CON 29.42845,-81.89617 Goethe CON 29.25419,-82.62381 Upper WaccasassaCON 29.36948,-82.73343 Steinhatchee CON 29.84634,-83.27067 Belmore CON 29.86435,-81.85891 TM Econ CON 28.37856,-81.11436 FortWhite CON 29.90193,-82.77529 TownofTioga CSC 29.65580,-82.48093 Oaks Mall CSC 29.65838,-82.41182 Walmart CSC 29.66309,-82.29970 Publix CSC 29.65095,-82.37182 Butler CSC 29.62237,-82.37970 Lowes_13th CSC 29.67587,-82.33994 HomeDepot_13th CSC 29.70206,-82.34348 Publix_Hail CSC 29.60106,-82.42052 Publix_53rd CSC 29.70205,-82.38995 Publix_16th CSC 29.67276,-82.38722 FreshMarket CSC 29.67417,-82.38760 PeachTree_34th CSC 29.62416,-82.37109 IndianCuisine CSC 29.62124,-82.37325 DT_Jax CBD 30.32902,-81.65912 DT_Orlando CBD 28.54068,-81.37889 DT_Tampa CBD 27.94824,-82.45832 DT_Gainesville CBD 29.65136,-82.32467 ALBuck Mixed 29.72551,-82.35377 AL_Beef Mixed_Ag 29.74675,-82.26424 Horse_Peak Mixed_Ag 29.56616,-82.19370

118

Table C-1. Continued. Site Name LU/LC Type GPS Decimal Degrees IFAS Mixed_Ag 29.79968,-82.41453 Flaggler Mixed_roadside 29.47624,-81.26756 I-75 Mixed_roadside 28.52287,-82.23969 MorningSide RP 29.66143,-82.27504 34_8_Park RP 29.66105,-82.37086 Northeast_Park RP 29.66518,-82.32093 Bivens Arm RP 29.62010,-82.33311 Lake Alice RP 29.64361,-82.36263 Northside_Park RP 29.70815,-82.35159 Squirrel_Ridge RP 29.61275,-82.34405 Boulware RP 29.62120,-82.30745 Alfred Ring Park RP 29.67293,-82.34755 PalmPointPark RP 29.63662,-82.23755 AL43RD SFR 29.73219,-82.38866 Hom_Golf SFR 28.73882,-82.54039 FancyNeighborhoodSFR 30.11105,-81.62466 ALJoel SFR 29.67130,-82.32560 Jax Church SFR 30.20316,-81.76371 DuckPond SFR 29.65687,-82.31981 LakeCityHood SFR 30.17192,-82.65423 PalakaHood SFR 29.63394,-81.69824 Newberry Hood SFR 29.63434,-82.61090 WillistonHood SFR 29.38300,-82.44333 keystone Heights SFR 29.77557,-82.03928 High Springs SFR 29.83269,-82.60483 WillistonHood_B SFR 29.39066,-82.44960 Micanopy SFR 29.50435,-82.28597 Chiefland SFR 29.47731,-82.85763 Starke SFR 29.94604,-82.11781

119

Table C-2. GPS coordinates for Chapter 3 transect points Transect Latitude Longitude 1 29.71860 82.46650 1 29.71855 -82.46598 1 29.71882 -82.46548 1 29.71952 -82.46417 1 29.72103 -82.46156 2 29.72423 -82.46953 2 29.72440 -82.46912 2 29.72468 -82.46837 2 29.72523 -82.46681 2 29.72626 -82.46380 3 29.72264 -82.46869 3 29.72279 -82.46828 3 29.72306 -82.46753 3 29.72362 -82.46600 3 29.72470 -82.46293 4 29.71869 -82.46654 4 29.71882 -82.46616 4 29.71906 -82.46539 4 29.71960 -82.46387 4 29.72073 -82.46078 5 29.72079 -82.46771 5 29.72292 -82.46196 5 29.72517 -82.45590 6 29.72454 -82.46970 6 29.72652 -82.46389 6 29.72875 -82.45778 7 29.72594 -82.47042 7 29.72614 -82.46995 7 29.72647 -82.46917 7 29.72703 -82.46771 7 29.72829 -82.46471 8 29.72570 -82.47033 8 29.72588 -82.46990 8 29.72618 -82.46912 8 29.72679 -82.46767 8 29.72800 -82.46480 9 29.72497 -82.46994 9 29.72515 -82.46947 9 29.72543 -82.46874 9 29.72605 -82.46723 9 29.72724 -82.46421

120

LIST OF REFERENCES

Anderson, GS, Lee, CSY, Fleming, GG, Menge, CW (1998) FHWA Traffic Noise Model User Guide. Version 1.0

Andrews, K, Gibbons, W, & Jochimsen, D (2008) Ecological effects of roads on amphibians and reptiles: a literature review. Herpetological Conservation 3:121– 143

Babisch, W (2011) Cardiovascular Effects of Noise. Noise and Health 13(52): 201

Barber, JR, Crooks, KR, & Fristrup, KM (2010) The costs of chronic noise exposure for terrestrial organisms. Trends in Ecology & Evolution 25(3)

Barber, JR, Burdett, CL, Reed, SE, Warner,K, Formichella, C, Crooks, KR, … Fristrup, KM (2011) Anthropogenic noise exposure in protected natural areas: estimating the scale of ecological consequences. Landscape Ecology 26(9):1281–1295

Bayne, EM, Habib, L, & Boutin, S (2008) Impacts of chronic anthropogenic noise from energy-sector activity on abundance of songbirds in the boreal forest. Conservation Biology 22(5):1186–93

Beckers, OM, & Schul, J (2004) Phonotaxis in Hyla versicolor (Anura, Hylidae): The effect of absolute call amplitude. Journal of Comparative Physiology A: Neuroethology, Sensory, Neural, and Behavioral Physiology 190(11):869–876

Bee, M, & Swanson, EM (2007) Auditory masking of anuran advertisement calls by road traffic noise. Animal Behaviour 74(6):1765–1776

Benítez-López, A, Alkemade, R, & Verweij, PA (2010) The impacts of roads and other infrastructure on mammal and bird populations: A meta-analysis. Biological Conservation 143(6):1307–1316

Bioacoustics Research Program (2011) Raven Pro: Interactive Sound Analysis Software (Version 1.4) [Computer software] Ithica, NY: The Cornell Lab of Ornithology. Available from http://www.birds.cornell.edu/raven

Blickley, J, & Patricelli, G (2010) Impacts of Anthropogenic Noise on Wildlife: Research Priorities for the Development of Standards and Mitigation. Journal of International Wildlife Law & Policy 13(4): 274–292

Blickley, JL, Blackwood, D, & Patricelli, GL (2012) Experimental evidence for the effects of chronic anthropogenic noise on abundance of Greater Sage-Grouse at leks. Conservation Biology : The Journal of the Society for Conservation Biology 26(3):461–71

121

Boelman, NT, Asner, GP., Hart, PJ, & Martin, RE (2007) Multi-trophic invasion resistance in Hawaii: Bioacoustics, field surveys, and airborne remote sensing. Ecological Applications 17(8): 2137–2144

Bondello, M (1977) The effects of high-intensity motorcycle sounds on the acoustical sensitivity of the desert iguana, dipsosaurus dorsalis. Doctoral Dissertation. California State University, Fullerton

Bondello, M, Huntley, AC, Cohen, HB, & Brattstrom, B (1979) The effects of dune buggy sounds on the telencephalic auditory evoked response in the Mojave fringe-toed lizards (Urna scorpano). Bureau of Land Management, Riverside, California

Bonier, F, Martin, PR, Sheldon, KS (2007) Sex-specific consequences of life in the city. Behavioural Ecology 18:121–129

Bormpoudakis, D, Sueur, J, & Pantis, JD (2013) Spatial heterogeneity of ambient sound at the habitat type level: ecological implications and applications. Landscape Ecology 28(3): 495–506

Botteldooren, D, De Coensel, B, & De Muer, T (2006) The temporal structure of urban soundscapes. Journal of Sound and Vibration 292(1-2):105–123

Bouchard, M (2009). Wetland Resources of Eastern South Dakota; Drainage Patterns, Assessment Techniques, and Predicting Future Risks. South Dakota State University

Brackenbury, J (1979) Power Capabilities of the Avian Sound-Producing System. Journal of Experimental Biology 78:163–166

Brambilla, G, Gallo, V, & Zambon, G (2013) The soundscape quality in some urban parks in Milan, Italy. International Journal of Environmental Research and Public Health 10(6): 2348–2369

Brown, MT, & Vivas, MB (2005) Landscape development intensity index. Environmental Monitoring and Assessment 101(1-3): 289–309

Brumm, H (2004) The impact of environmental noise on song amplitude in a territorial bird. Journal of Animal Ecology 73(3): 434–440

Brumm, H (2006) Animal communication: city birds have changed their tune. Current Biology 16(23): R1003–R1004

Brumm, H & Slabbekoorn, H (2005). Acoustic communication in noise. Advances in the Study of Behavior 35:151–209

Campo, JL, Gil, MG, & Dávila, SG (2005) Effects of specific noise and music stimuli on stress and fear levels of laying hens of several breeds. Applied Animal Behaviour Science 91(1-2):75–84

122

Can, A, Leclercq, L, Lelong, J, & Botteldooren, D (2010) Traffic noise spectrum analysis: Dynamic modeling vs. experimental observations. Applied Acoustics 71(8):764–770

Castelle, AJ, Johnson, AW, & Conolly, C (1994) Wetland and stream buffer size requirements-A Review. Journal of Environment Quality 23(5):878-882

The Center for Landscape Analysis (2014) SPreAD-GIS: Version 2.0. Seattle, WA: The Wilderness Society

Chan, AAY-H, Giraldo-Perez, P, Smith,S, & Blumstein, DT (2010) Anthropogenic noise affects risk assessment and attention: the distracted prey hypothesis. Biology Letters 6(4): 458–461

Chapin, FS, Zavaleta, ES, Eviner, VT, Naylor, RL, Vitousek, PM, Reynolds, HL, … Díaz, S. (2000) Consequences of changing biodiversity. Nature 405(6783):234–42

Charif, RA, AM Waack, and LM Strickman (2010) Raven Pro 1.4 User’s Manual. Cornell Lab of Ornithology, Ithaca, NY

Chen, TS, & Lin, HJ (2011) Application of a Landscape Development Intensity Index for Assessing Wetlands in Taiwan. Wetlands 31(4):745–756

Chloupek, P, Voslářová, E, Chloupek, J, Bedáňová, I, Pištěková, V, & Večerek, V (2009) Stress in Broiler Chickens Due to Acute Noise Exposure. Acta Veterinaria Brno 78(1): 93–98

Coffin, AW (2007) From roadkill to road ecology: A review of the ecological effects of roads. Journal of Transport Geography 15(5):396–406

Depraetere, M, Pavoine, S, Jiguet, F, Gasc, A, Duvail, S, & Sueur, J (2012) Monitoring animal diversity using acoustic indices: Implementation in a temperate woodland. Ecological Indicators 13(1): 46–54

Dooley, JM (2016) Noise and Landscape Development Intensity. Landscape Ecology. Submitted paper

Dooling, RJ (1982) Auditory perception in birds. In: Acoustic Communication in Birds, Vol. 1, edited by DE Kroodsma and EH Miller, (Academic Press, New York), pp. 95-130

Dooling, RJ, Lohr, B, and Dent, ML (2000). Hearing in birds and reptiles, In: Comparative Hearing: Birds and Reptiles, edited by R.J. Dooling, A.N. Popper, and R.R. Fay (Springer-Verlag, New York eds), pp. 308-359

Dooling, RJ & Popper, A (2007) The Effects of Highway Noise on Birds. Rockville, Maryland: Environmental BioAcoustics LLC

123

Dumyahn, SL & Pijanowski, BC (2011) Soundscape conservation. Landscape Ecology 26(9):1327–1344

Egan, DM (2007) Architectural Acoustics (Reprint.) USA: J. Ross

Egnor, SER & Hauser, MD (2006) Noise-Induced Vocal Modulation in Cotton-Top Tamarins ( Saguinus oedipus ) American Journal of Primatology (68):1183–1190

Eigenbrod, F, Hecnar, SJ, & Fahrig, L (2009) Quantifying the road-effect zone: Threshold effects of a motorway on anuran populations in Ontario, Canada. Ecology and Society 14(1)

Embleton, TFW (1996) Tutorial on sound propagation outdoors. The Journal of the Acoustical Society of America 100:31

ESRI (2014) ArcGIS Desktop: Release 10.3. Redlands, CA: Environmental Systems Research Institute

Fahrig, L, & Rytwinski, T (2009) Effects of roads on animal abundance: an emperical review and synthesis. Ecology and Society 14(1):21–41

Farina, A, Lattanzi, E, Malavasi, R, Pieretti, N, & Piccioli, L (2011a) Avian soundscapes and cognitive landscapes: Theory, application and ecological perspectives. Landscape Ecology 26(9):1257–1267

Farina, A, Pieretti, N, & Piccioli, L (2011b) The soundscape methodology for long-term bird monitoring: A Mediterranean Europe case-study. Ecological Informatics 6(6): 354–363

Farina, A, & Pieretti, N (2014) Sonic environment and vegetation structure: A methodological approach for a soundscape analysis of a Mediterranean maqui. Ecological Informatics 21:120–132

Farina, A, James, P, Bobryk, C, Pieretti, N, Lattanzi, E, & McWilliam, J (2014) Low cost (audio) recording (LCR) for advancing soundscape ecology towards the conservation of sonic complexity and biodiversity in natural and urban landscapes. Urban Ecosystems 17(4):923-944

Felton, A, Alford, RA, Felton, AM, & Schwarzkopf, L (2006) Multiple mate choice criteria and the importance of age for male mating success in the microhylid frog, Cophixalus ornatus. Behavioral Ecology and Sociobiology 59(6):786–795

FGDL Metadata Explorer (2015) Gainesville, FL: University of Florida GeoPlan Center. Available: http://www.fgdl.org/metadataexplorer/explorer.jsp

FHWA (2014) Traffic Noise Model. Federal Highway Administration, Washington, DC, USA. Available from http://www.fhwa.dot.gov/environment/noise/traffic_noise_model/

124

Flatow, I (Host) (2011, April 22) Listening to Wild Soundscapes. Science Friday. National Public Radio Retrieved from http://www.npr.org/2011/04/22/135634388/listening-to-wild-soundscapes

Fore, LS (2004) Development and testing of biomonitoring tools for macroinvertebrates in Florida streams. Statistical Design, Seattle, Washington. A report for the Florida Department of Environmental Protection, Tallahassee, Florida, USA. 62 p

Fore, LS (2005) Assessing the biological condition of Florida lakes: development of the lake vegetation index (LDV). Statistical Design, Seattle, Washington. A report for the Florida Department of Environmental Protection, Tallahassee, Florida, USA. 29 pp. & Appendixes

Forman, R (2000) Estimate of the Area Affected Ecologically by the Road System in the United States. Conservation Biology 14(1):31–35

Forman, R, & Alexander, L (1998) Roads and their major ecological effects. Annual Review of Ecology and Systematics 29:207–231

Forman, R, & Deblinger, R (2000) The ecological road-effect zone of a Massachusetts (USA) suburban highway. Conservation Biology 14(1):36–46

Forman, R, Reineking, B, & Hersperger, AM (2002) Road traffic and nearby grassland bird patterns in a suburbanizing landscape. Environmental Management 29(6): 782–800

Francis, CD, & Barber, JR (2013) A framework for understanding noise impacts on wildlife: An urgent conservation priority. Frontiers in Ecology and the Environment 11(6): 305–313

Francis, CD & Blickley, JL (2012) Research and perspectives on the study of anthropogenic noise and birds. Ornithological Monographs 74(1):1–5.

Francis, CD, Ortega, CP, & Cruz, A (2009) Noise pollution changes avian communities and species interactions. Current Biology 19(16):1415–9

Fuller, S, Axel, AC, Tucker, D, & Gage, SH (2015) Connecting soundscape to landscape: Which acoustic index best describes landscape configuration? Ecological Indicators 58:207–215

Gage, SH, & Axel, AC (2014) Visualization of temporal change in soundscape power of a Michigan lake habitat over a 4-year period. Ecological Informatics 21:100–109

Gage, SH, Napoletano, BM, & Cooper, MC (2001) Assessment of ecosystem biodiversity by acoustic diversity indices. Journal of the Acoustical Society of America 109(2430)

125

Gerhardt, HC (1975) Sound pressure levels and radiation patterns of the vocalizations of some North American frogs and toads. Journal of Comparative Physiology 102(1):1–12

Gerhardt, HC (2001) Acoustic communication in two groups of closely related treefrogs. Advances in the Study of Behavior 30:99–167

Gerhardt, HC and Huber, F (2002). Acoustic Communication in Insects and Anurans. Chicago, London: University of Chicago Press

Goerlitz, HR, & Siemers, BM (2007) Sensory ecology of prey rustling sounds: acoustical features and their classification by wild Grey Mouse Lemurs Functional Ecology 21(1):143–153

Gibbons, JW, Winne, CT, Scott, DE, Willson, JD, Glaudus, X, Andrews, KM, Todd, BD, Fedewa, LA, Wilkinson, L, Tsaliogas, RN, Harper, SJ, Greenr, JL, Tuberville, TD, Metts, BS, Dorcas, ME, Nestor, JP, Young, CA, Akre, T, Reed, RN, Buhlmann, KA, Norman, J, Croshow, DA, Hagen, C and Rothermel, BB (2006), Remarkable Amphibian Biomass and Abundance in an Isolated Wetland: Implications for Wetland Conservation. Conservation Biology 20: 1457–1465

HR, Greif, S, & Siemers, BM (2008) Cues for acoustic detection of prey: insect rustling sounds and the influence of walking substrate The Journal of Experimental Biology 211:2799–806

Habib, L, Bayne, EM, & Boutin, S (2007) Chronic industrial noise affects pairing success and age structure of ovenbirds Seiurus aurocapilla. Journal of Applied Ecology 44(1): 176–184

Halfwerk, W, & Slabbekoorn, H (2009) A behavioural mechanism explaining noise- dependent frequency use in urban birdsong. Animal Behaviour 78(6):1301–1307

Halfwerk, W., Holleman, L. J. M., Lessells, Ck. M., & Slabbekoorn, H. (2011). Negative impact of traffic noise on avian reproductive success. Journal of Applied Ecology, 48(1), 210–219. http://doi.org/10.1111/j.1365-2664.2010.01914.x

Harris, CM (1991) Handbook of acoustical measurements and noise control (3rd ed.). New York, NY: McGraw-Hill Inc.

Harris, R (2011, March 26) Scientists Tune in to the “Voices of the Landscape”. Weekend Edition Saturday. National Public Radio

Hu, Y, & Cardoso, GC (2009) Are bird species that vocalize at higher frequencies preadapted to inhabit noisy urban areas? Behavioral Ecology 20(6):1268–1273

Huet des Aunay, G, Slabbekoorn, H, Nagle, L, Passas, F, Nicolas, P, & Draganoiu, TI (2014) Urban noise undermines female sexual preferences for low-frequency songs in domestic canaries. Animal Behaviour 87: 67–75

126

Jensen M, Thompson H (2004) Natural sounds: an endangered species. George Wright Forum 21:10–13

Joo, W (2009) Environmental Acoustics As an Ecological Variable To Understand the Dynamics of Ecosystems. Dissertation. Michigan State University.

Joo, W, Gage, SH, & Kasten, EP (2011) Analysis and interpretation of variability in soundscapes along an urban–rural gradient. Landscape and Urban Planning 103(3-4):259–276

Kang, J, & Zhang, M (2010) Semantic differential analysis of the soundscape in urban open public spaces. Building and Environment 45(1):150–157

Kasten, EP, Gage, SH, Fox, J, & Joo, W (2012) The remote environmental assessment laboratory’s acoustic library: An archive for studying soundscape ecology. Ecological Informatics 12:50–67

Kawada, T (2011) Noise and health--sleep disturbance in adults. Journal of Occupational Health 53(6):413–6

Kight, CR, & Swaddle, J (2011) How and why environmental noise impacts animals: an integrative, mechanistic review. Ecology Letters 14(10):1052–61

Kociolek, AV, Clevenger, AP, St Clair, CC, & Proppe, DS (2011) Effects of road networks on bird populations. Conservation Biology : The Journal of the Society for Conservation Biology 25(2):241–9

Koskinen, H, & Toppila, E (2011) Hearing loss among classical-orchestra musicians. Noise and Health 13(50)

Krause, B (1987) Bioacoustics, habitat ambience in ecological balance. Whole Earth Review 57:14–18

Krause, B (2013) The Voice of the Natural World [Video File]. Retrieved from https://www.ted.com/talks/bernie_krause_the_voice_of_the_natural_world?langu age=en

Krause, BL, Gage, SH, & Joo, W (2011) Measuring and interpreting the temporal variability in the soundscape at four places in Sequoia National Park. Landscape Ecology 26(9):1247–1256

Kuehne, LM, Padgham, BL, & Olden, JD (2013) The Soundscapes of Lakes across an Urbanization Gradient. PLoS ONE 8(2)

Kutner, MH, Nachtscheim, CJ, & Neter, J (2004) Applied Linear Regression Models (Fourth Ed.) New York, NY: McGraw-Hill/Irwin

127

Laiolo, P (2010) The emerging significance of bioacoustics in animal species conservation. Biological Conservation 143(7):1635–1645

Lane, CR, Brown, MT, (2007) Diatoms as indicators of wetland condition. Ecological Indicators 7:521–540

Langguth, B, Kreuzer, PM, Kleinjung, T, & De Ridder, D (2013) Tinnitus: causes and clinical management. Lancet Neurology 12(9):920–30

Lau, MC, Lee, CSY, Rochat, JL, Boeker, ER, Fleming, GG (2004) FHWA Traffic Noise Model User Guide. Version 2.5 Addendum

Lengagne, T (2008) Traffic noise affects communication behaviour in a breeding anuran, Hyla arborea. Biological Conservation 141(8):2023–2031

Lohr, B, Wright, TF, & Dooling, RJ (2003) Detection and discrimination of natural calls in masking noise by birds: estimating the active space of a signal. Animal Behaviour 65(4):763–777

Mack, JJ (2006) Landscape as a predictor of wetland condition: An evaluation of the Landscape Development Index (LDI) with a large reference wetland dataset from Ohio. Environmental Monitoring and Assessment (120):221–241

Magrath, RD, Pitcher, BJ, & Dalziell, AH (2007) How to be fed but not eaten: nestling responses to parental food calls and the sound of a predator’s footsteps. Animal Behaviour 74(5):1117–1129

Margriter, SC, Bruland, GL, Kudray, GM, & Lepczyk, CA (2014) Using indicators of land-use development intensity to assess the condition of coastal wetlands in Hawaii. Landscape Ecology 29(3):517–528

Marler, P, & Slabbekoorn, HW (Eds.) (2004) Nature ’s Music. The Science of Birdsong (Vol. 1)

Marten, K & Marler, P (1977) Sound Transmission and Its Significance for Animal Vocalization: I. Temperate Habitats. Behavioral Ecology and Sociobiology 2(3):271–290.

Matsinos, YG, Mazaris, AD, Papadimitriou, KD, Mniestris, A, Hatzigiannidis, G, Maioglou, D, & Pantis, JD (2008) Spatio-temporal variability in human and natural sounds in a rural landscape. Landscape Ecology 23:945–959

Mazaris, AD, Kallimanis, AS, Chatzigianidis, G, Papadimitriou, K, & Pantis, JD (2009) Spatiotemporal analysis of an acoustic environment: interactions between landscape features and sounds. Landscape Ecology 24(6):817–831

Means, DB, & Simberloff, D (1987) The peninsula effect: habitat-correlated species decline in Florida’ s herpetofauna. Journal of Biogeography 14(6):551–568

128

McIntyre, E, Leonard, ML, & Horn, A (2014) Ambient noise and parental communication of predation risk in tree swallows, Tachycineta bicolor. Animal Behaviour 87:85– 89

Mclaughlin, KE, & Kunc, HP (2013) Experimentally increased noise levels change spatial and singing behavior. Biology Letters 9:21–25

McShane, TO, Hirsch, PD, Trung, TC, Songorwa, AN, Kinzig, A, Monteferri, B, … O’Connor, S (2011) Hard choices: Making trade-offs between biodiversity conservation and human well-being. Biological Conservation 144(3): 966–972

Miller, NP (2002). Transportation noise and recreational lands. Noise/News International 11(1):9-21

Mills, JH, Gilbert, RM, & Adkins, WY (1979) Temporary threshold shifts in humans exposed to octave bands of noise for 16 to 24 hours. The Journal of the Acoustical Society of America 65(5): 1238–48

Mitsch, WJ, & Gosselink, JG (2007) Wetlands (Fourth.) Hoboken, New Jersey: John Wiley and Sons

Muzet, A (2007) Environmental noise, sleep and health Sleep Medicine Reviews 11(2): 135–42

Napoletano, BM (2004) Measurement, quantification and interpretation of acoustic signals within an ecological context. Dissertation. Michigan State University.

Nega, T, Yaffe, N, Stewart, N, & Fu, WH (2013) The impact of road traffic noise on urban protected areas: A landscape modeling approach. Transportation Research Part D: Transport and Environment 23:98–104

National Park Service, & U.S. Department of the Interior (2006) Management Policies 2006

Odum, HT (1996) Environmental Accounting: Emergy and Environmental Decision Making. New York, NY: John Wiley and Sons

Oliver, L, Lehrter, J, & Fisher, W (2011) Relating landscape development intensity to coral reef condition in the watersheds of St. Croix, US Virgin Islands. Marine Ecology Progress Series 427:293–302

Parris, K, Velik-Lord, M, & North, J (2009) Frogs call at a higher pitch in traffic noise. Ecology and Society 14(1)

Passchier-Vermeer, W, & Passchier, WF (2000) Noise exposure and public health. Environmental Health Perspectives 108 Suppl (6):123–31

129

Patricelli, GL, & Blickley, J (2006) Avian communication in urban noise: causes and consequences of vocal adjustment. The Auk 123(3):639–649

Penna, M, & Zúñiga, D (2013) Strong responsiveness to noise interference in an anuran from the southern temperate forest. Behavioral Ecology and Sociobiology 68(1):85–97

Pieretti, N, & Farina, A (2013) Application of a recently introduced index for acoustic complexity to an avian soundscape with traffic noise. The Journal of the Acoustical Society of America 134(1):891–900

Pieretti, N, Farina, A, & Morri, D (2011) A new methodology to infer the singing activity of an avian community: The Acoustic Complexity Index (ACI). Ecological Indicators 11(3): 868–873

Pijanowski, BC, Villanueva-Rivera, LJ, Dumyahn, SL, Farina, A, Krause, BL, Napoletano, BM, … Pieretti, N (2011a) Soundscape Ecology: The Science of Sound in the Landscape BioScience, 61(3): 203–216

Pijanowski, BC, Farina, A, Gage, SH, Dumyahn, SL, & Krause, BL (2011b) What is soundscape ecology? An introduction and overview of an emerging new science. Landscape Ecology 26:1213–1232

Qi, J, Gage, SH., Joo, W, Napoletano, BM, & Biswas, S (2008) Soundscape Characteristics of an Environment: A New Ecological Indicator of Ecosystem Health. Wetland and Water Resource Modeling and Assessment: A Watershed Perspective (pp. 201–214). CRC Press, Taylor and Francis Group

Quinn, J, Whittingham, M, Butler, S, & Cresswell, W (2006) Noise, predation risk compensation and vigilance in the chaffinch Fringilla coelebs. Journal of Avian Biology 37(6):601–608

Radford, C, Stanley, J, Tindle, C, Montgomery, J, & Jeffs, A (2010) Localised coastal habitats have distinct underwater sound signatures. Marine Ecology Progress Series 401: 21–29

Raimbault, M, Lavandier, C, & Bérengier, M (2003) Ambient sound assessment of urban environments: Field studies in two French cities. Applied Acoustics 64(12):1241–1256

Reed, SE, JL Boggs, and JP Mann. (2012) SPreAD-GIS: a GIS tool for modeling anthropogenic noise propagation in natural ecosystems. Environmental Modelling & Software 37: 1-5

Reinis, S (1976) Acute changes in animal inner ears due to simulated sonic booms. Journal of the Acoustical Society of America (60):133–138.

130

Reijnen, M, Veenbaas, G, & Foppen, R (1995) Predicting the effects of motorway traffic on breeding bird populations

Reijnen, R, Foppen, R, & Veenbaas, G (1997) Disturbance by traffic of breeding birds: evaluation of the effect and considerations in planning and managing road corridors. Biodiversity and Conservation 581(4): 567–581

Reijnen, R, Foppen, R, Braak, C, & Thissen, J (1995) The effects of car traffic on breeding bird populations in woodland. III. Reduction of density in relation to the proximity of main roads. Journal of Applied Ecology 32(1):187–202

Reijnen, R, Foppen, R, & Meeuwsen, H. (1996) The effects of traffic on the density of breeding birds in Dutch agricultural grasslands. Biological Conservation 75(3):255–260

Reiss, KC, Brown, MT, & Lane, CR (2010) Characteristic community structure of Florida’s subtropical wetlands: The Florida wetland condition index for depressional marshes, depressional forested, and flowing water forested wetlands. Wetlands Ecology and Management 18(5): 543–556

Ríos-Chelén, AA, Quirós-Guerrero, E, Gil, D, & Macías Garcia, C (2013) Dealing with urban noise: vermilion flycatchers sing longer songs in noisier territories. Behavioral Ecology and Sociobiology 67(1):145–152

Ruiz G, Rosenmann M, Novoa FF, Sabat P (2002) Hematological parameters and stress index in rufous-collared sparrows dwelling in urban environments. Condor 104:162–166.

Ryals, BM, Dooling, RJ, Westbrook, E, Dent, ML., MacKenzie, A, & Larsen, ON (1999) Avian species differences in susceptibility to noise exposure. Hearing Research 131(1-2):71–88

Saunders, JC, Mills, JH, & Miller, JD (1977) Threshold shift in the chinchilla from daily exposure to noise for six hours. The Journal of the Acoustical Society of America 61(2):558–70

U.S. EPA (1974a) Information on Levels of Environmental Noise Requisite to Protect Public Health and Welfare with an Adequate Margin of Safety

U.S. EPA (1974b) Population Distribution of the United States as a Function of Outdoor Noise Level

Makarewicz, R, & Sato, Y (1996) Representative spectrum of road traffic noise. Journal of the Acoustical Society of Japan, 5:249–254

Schafer, MR (1977) The tuning of the world. New York, NY: Knopf

131

Schafer, MR (1994) Our Sonic Environment and the Soundscape: The Tuning of the World. Rochester, Vermont: Destiny Books

Schwartz, JJ, & Wells, KD (1983) The influence of background noise on the behavior of a neotropical treefrog, Hyla ebraccata Herpetologica 39(2):121–129

Siemers, BM, & Schaub, A (2011) Hunting at the highway: traffic noise reduces foraging efficiency in acoustic predators. Proceedings. Biological Sciences / The Royal Society 278(1712): 1646–52

Simpson, M, & Bruce, R (1981) Noise in America: Extent of the Noise Problem. US Environmental Protection Agency Washington D.C.

Slabbekoorn, H (2004) Habitat-dependent ambient noise: Consistent spectral profiles in two African forest types. The Journal of the Acoustical Society of America 116(6):3727

Slabbekoorn, H, & Peet, M (2003) Birds sing at a higher pitch in urban noise. Nature 424(July):267

Slabbekoorn, H, & Ripmeester, E (2008) Birdsong and anthropogenic noise: implications and applications for conservation. Molecular Ecology 17(1):72–83

Smith, M, Roberts, J, Hammond, T, & Davis, R (2003) Intraspecific variation in the advertisement call of the sunset frog Spicospina flammocaerulea (Anura: Myobatrachidae): a frog with a limited geographic distribution. Journal of Herpetology 37(2):285-291

Speares, P, Holt, D, & Johnston, C (2011) The relationship between ambient noise and dominant frequency of vocalizations in two species of darters (Percidae: Etheostoma). Environmental Biology of Fishes 90(1):103–110

Spellerberg, IF (1998) Ecological effects of roads and traffic: a literature review. Global Ecology and Biogeography 7(5):317–333

Sun, J. WC, & Narins, PM (2005) Anthropogenic sounds differentially affect amphibian call rate. Biological Conservation 121(3): 419–427

Sueur, J, Pavoine, S, Hamerlynck, O, & Duvail, S (2008) Rapid acoustic survey for biodiversity appraisal. PloS One 3(12)

Sueur, J & Farina, A (2015) Ecoacoustics: the ecological investigation and interpretation of environmental sound. Biosemiotics 8(3):493-502

Sueur, J, Farina, A, Gasc, A, Pieretti, N, & Pavoine, S (2014) Acoustic Indices for Biodiversity Assessment and Landscape Investigation. Acta Acustica United with Acustica 100(4):772–781

132

Sun, J, & Narins, PM (2005) Anthropogenic sounds differentially affect amphibian call rate. Biological Conservation 121(3):419–427

Swaddle, J, & Page, L (2007) High levels of environmental noise erode pair preferences in zebra finches: implications for noise pollution. Animal Behaviour 74(3):363– 368

Szalma, JL, & Hancock, PA (2011) Noise effects on human performance: a meta- analytic synthesis. Psychological Bulletin 137(4):682–707

Trombulak, SC, & Frissell, CA (2000) Review of ecological effects of roads on terrestrial and aquatic communities. Conservation Biology 14(1):18–30

Truax, B (2001) Acoustic Communication (Second.) Westport, Connecticut: Ablex.

Tucker, D, Gage, SH, Williamson, I, & Fuller, S (2014) Linking ecological condition and the soundscape in fragmented Australian forests. Landscape Ecology 29(4):745– 758

Turner, MG (2005) LANDSCAPE ECOLOGY: What Is the State of the Science? Annual Review of Ecology, Evolution, and Systematics 36(1):319–344

Vitousek, PM, Mooney, HA, Lubchenco, J, & Melillo, JM (1997) Human domination of Earth’s ecosystems. Urban Ecology: An International Perspective on the Interaction Between Humans and Nature 277(July):3–13

Ware, HE, McClure, CJW, Carlisle, JD, & Barber, JR (2015) A phantom road experiment reveals traffic noise is an invisible source of habitat degradation. Proceedings of the National Academy of Sciences 112(39):12105–12109

Warren, PS, Katti, M, Ermann, M, & Brazel, A (2006) Urban bioacoustics: it’s not just noise. Animal Behaviour 71(3):491–502

Welch, PD (1967) The Use of Fast Fourier Transform for the Estimation of Power Spectra: A Method Based on Time Averaging Over Short, Modified Periodograms. IEEE Transactions on Audio and Electroacoustics AU-15(2)

Wenger, S (1999) A review of the scientific literature on riparian buffer width, extent and vegetation. Athens, Georgia

World Health Organization (WHO). Guidelines for community noise. (1999). Geneva

Wollerman, L (1999) Acoustic interference limits call detection in a Neotropical frog Hyla ebraccata. Animal Behaviour 57(3):529–536

Wood, WE, & Yezerinac, SM (2006) Song Sparrow (Melospiza Melodia) Song Varies with Urban Noise. The Auk 123(3): 650

133

BIOGRAPHICAL SKETCH

Jenet M. Dooley was raised in Houston, Texas. Jenet received a Bachelor of

Science degree in environmental engineering management from Miami University in

Oxford, Ohio in 2010. She received a Master of Science in environmental engineering sciences from the University of Florida in 2014. Jenet completed her Doctor of

Philosophy in environmental engineering sciences from the University of Florida in

2016. Jenet is continuing the pursuit of her research interests in wetland and soundscape ecology as a consultant in Alberta, Canada.

134