17.2 ACOUSTIC SIGNAL CLASSIFICATION Viewed in terms of information theory, the acoustic frequency spectrum is primar-ily an information-carrying medium.. Early observations led to the co
Trang 1Characteristics
of an Environment
A New Ecological Indicator
of Ecosystem Health
Jiaguo Qi, Stuart H Gage, Wooyeong Joo,
Brian Napoletano, and S Biswas
17.1 INTRODUCTION
Landscape characteristics are important measures of an ecosystem’s environmental health, as they depict spatial patterns of physical attributes of the landscape that many organisms rely on The visual features of a landscape, such as forest type, density, patch size and shape, affect habitat properties that are specific to differ-ent organisms Change or disruption of the spatial patterns of a landscape has been shown to impact biodiversity (Crist et al., 2004, Jeanneret et al., 2003, Sala et al.,
2000, Foley et al., 2005), ecological function (Allan, 2004, Alberti, 2005, Grigulis
et al., 2005, Battin, 2004), and ecosystem services (Tscharntke et al., 2005, Fischer and Lindenmayer, 2007)
A suite of landscape matrices has been developed based on land use and land cover maps derived from satellite images as a measure of landscape fragmentation They include, for example, patch density, Shannon diversity index, as proxies of landscape characteristics These matrices have been found to be important indica-tors of an ecosystem’s biodiversity and integrity (Sala et al., 2000, Foley et al., 2005, Fischer and Lindenmayer, 2007)
Although these landscape characteristics, often derived from analysis of remotely sensed imagery, are important indicators of ecosystem health, they are temporally static and do not provide a sufficient spatial resolution to observe the responses of individual organisms to anthropogenic disturbances The audio characteristics emit-ted from an ecosystem, such as sounds from birds, mechanical movements, or wind (Truax, 1999, Schafer, 1977), provide unique insight into spatial and temporal pat-terns of ecosystem responses to human disturbances While soundscape characteris-tics provide complementary information to landscape characterischaracteris-tics, little research has been done to fully explore the usefulness of coupling these two complementary indicators of ecological dynamics
Trang 2We define an ecosystem’s soundscape as the physical extent of acoustic signals and the spectral range of signal frequencies associated with an ecosystem’s biophysi-cal processes Truax (1999) and Schafer (1977) introduced the idea of a soundscape
in their early studies of acoustic ecology Environmental soundscape analysis as a complementary measure of ecosystem dynamics uniquely addresses some of the key criteria for the establishment of ecological indices as articulated by Dale and Beyler (2001) Soundscape analysis is a predictable measure of ecosystem stress, is antici-patory, is integrative, and can measure disturbance Because an ecosystem’s sound-scape is a function of a variety of ecological variables, assessment of the soundsound-scape serves to integrate several variables in the measure of integrity and biocomplexity (Thompson 2001, Holling 2001, Mueller and Kuc 2000, Porter et al., 2005) This chapter demonstrates the capability of acoustic sensing techniques to characterize
an ecosystem’s soundscape
17.2 ACOUSTIC SIGNAL CLASSIFICATION
Viewed in terms of information theory, the acoustic frequency spectrum is primar-ily an information-carrying medium An organism or force generating the acous-tic signal acts as the encoder and transmitter, and the acousacous-tic spectrum acts as the medium through which the encoded signal travels The receiver then registers and decodes the signal (as in human conversation, for instance) The various signals within the acoustic spectrum are commonly classified as either natural or human-induced sounds (Schafer 1977)
Krause (1998), in his studies of natural soundscapes, devised the term biophony
to describe the complex chorus of ambient biological sounds (biophony = biologic
and symphony), and geophony for a region’s ambient geological sounds (Figure 17.1)
Similarly, the term anthrophony refers to the human-imposed sounds (0.2–2.0 kHz)
The two primary categories, biophony and anthrophony, can be further subdivided conceptually Early observations led to the conclusion that signals within the bio-phony range (2.0–11.0 kHz) can be characterized as intentional, meaning the trans-mitter of the signal wishes to communicate information, such as mating or distress calls, through the acoustic spectrum, or incidental, in which signals transmitted may contain relevant information but are not dispatched for the explicit purpose of communication
Anthropogenic sounds can be further divided into mechanistic and oral classes Oral sounds are those produced by human beings themselves (i.e., talking, shout-ing, or singing) Conversely, mechanistic signals involve sounds produced by vari-ous forms of human-made machinery and technology Within this class, two further subcategories exist: stationary and temporal Stationary refers to those signals that impose themselves on the ambient soundscape permanently (i.e., turbulence from air-conditioner fans), and temporal signals include the noises that move through the soundscape over a given temporal scale (i.e., automobile or train traffic) While this schema does not provide an absolute standard of acoustic classification, it does provide the framework to begin characterization of acoustic signaling (see Figure 17.1)
Trang 317.3 SOUNDSCAPE ANALYSIS
17.3.1 ECOLOGICAL SOUNDSCAPES
Acoustic diversity refers to the patterns of frequency and temporal use of the acous-tic spectrum Biophonic complexity thereby indicates the degree to which different vocalizing organisms utilize different niches to relay information within the spec-trum Specifically, ecosystems with lesser degrees of human interference tend to exhibit greater biophonic complexity in terms of frequency and periodicity utiliza-tion Moreover, anthropogenic interference, and more particularly temporal interfer-ence, within a soundscape will tend to hinder organism populations by lowering reproduction rates and increasing predation rates Organisms make careful use of the acoustic frequency when attempting to communicate information such as mating potential, territory size, and potential predation When anthropogenic interference disrupts this communication, critical information is not relayed and the organism’s population experiences a decline (Krause 1998) Therefore, acoustic characteristics may serve as an ecological indicator of ecosystems
17.3.2 DEVELOPMENT OF SOUNDSCAPE INDICATORS
An acoustic signal is characterized by multiple physical attributes including timing, frequency, and intensity The data set produced by acoustic recordings and quanti-fication is an array of acoustic intensity of contiguous, nonoverlapping frequency bands (Figure 17.2) These data form a data matrix where the rows represent record-ing intervals and the columns are frequency bands A wide frequency band summa-rizes the intensity of sound waves across a relatively wide set of frequencies, while
a narrow band restricts the range of frequency summarized The analytical role is to summarize patterns in covariation among the different frequency bands across the temporal period during which the acoustic data were recorded The most convincing and feasible statistical method for describing such patterns of covariation in each acoustic signature is to calculate the dominance in each frequency band and compute their statistical distributions
Intentional
Signaling
Incidental Signaling
Biophony Geophony
Oral Communications
Stationary Temporary
Mechanistic Sounds
Anthrophony
Sound Spectrum
FIGURE 17.1 View diagram of acoustic taxonomy
Trang 4In impacted ecosystems the spectral properties of acoustic signals in the environ-ment sometimes aggregate within two primary regions of a spectrogram The first region occurs at the lower frequencies of the sound spectrum This band typically extends from 0.2 to 1.5 kHz and consists primarily of mechanical signals (e.g., trains,
cars, air conditioners, etc.), and is therefore referred to as the anthrophonic region.
The second band of concentration begins in the range of 2 kHz and is prevalent up to
8 kHz, but may reach a higher spectral range especially when organisms
communi-cate using wider signal bandwidths (e.g., Molothrus ater) or ultrasound (e.g., bats) We
currently restrict our range to human detection to match with human auditory survey techniques This realm of acoustic activity consists primarily of signals generated by
biological organisms, and is therefore referred to as the biophonic region We have
delineated this frequency band as the biological band based on our observations and the frequency ranges referred to in the literature These two bands correspond to two
of the three taxonomic categories of the soundscape described above, but do not cover acoustics emanating from the physical (i.e., wind, rain, etc.) or geophonic component This is because the geophony, when present, occurs as a signal that is diffuse through-out the entire spectrum The geophony is a diffuse signal that is strongest at the lowest frequencies, but continues with a relatively high intensity into the higher frequencies, and its individual components are difficult to isolate and identify
Using this structure we compute the acoustic intensity for anthrophony (F), bio-phony (G), and geobio-phony (L) These three acoustic ranges are then compared to the
Divided into 11 frequency bands, each 1 kHz wide
Relative mean intensity of sound
in each 1 kHz band
Frequency Band
Frequency Class
11 Classes, Each Class~ = 1kHz
Paris Park; July 7, 2002, 0530
Acoustic Signature Map (Spectrogram)
Time (30 sec)
60
40
20
0
FIGURE 17.2 The acoustic frequency slicing procedure Each sound wave file is divided
into 11 frequency bands and the relative mean intensity is calculated for each band (See color
frequency bands
Trang 517.4 A SAMPLE APPLICATION
To demonstrate the usefulness of the acoustic signals as an environmental indica-tor, sounds were recorded in Nanchang city, China (Figure 17.3) and another one
in Michigan Nanchang Park was once a plant nursery but was transformed into a
FIGURE 17.3 A photograph of the China study site where acoustic data were collected and analyzed in this paper
value of the entire signal (s) A value > 1 indicates that the concentration of acous-tic activity in the analyzed region was greater than the value for the entire signal Therefore, the region with the highest value was the predominant source of acoustic activity in the signal For example, if the br had the highest value, then biological activity was predominant, while a larger ar value indicated dominant anthropogenic activity To emphasize the comparison of biological and anthropogenic activity, we divided the b value by the a value to calculate r (=b/a), the ratio of biological to anthropogenic activity
In addition to computing the ratios of activity from our spectrumgram, we also determined the percentage of total activity a single band contributes to the total sig-nal A gp value near 100% coincident with a bp value of approximately the same value
indicated that the primary signal source in the sound sample was biophony
(geophysi-cal) activity When the ap value was greater than 50%, it indicated that the primary
signal source was anthrophony (anthropogenic) activity, whereas a value of bp greater
than 50% indicated that biophony (biological) activity was the dominant source.
Trang 6natural reserve after it changed owners in 1996 Soon after that, the park became one of the primary nesting and mating areas for summer migratory birds Sound recording ecosystems were developed and calibrated, and the sounds were recorded between July 7 and 15, 2005 at 30-minute time intervals The Michigan site was located in a backyard of a private house in a rural residential area in Okemos, Michi-gan, surrounded by forests woodlots Acoustic recorders were placed about 40 yards away from the house for a multiple year data collection However, in this study, we only used a short period of time data in July 7, 2005 that are coincident with the data from China
As a demonstration of the soundscape characteristics, Figure 17.4, depicts the sound spectra of selected acoustic signals from data collected on July 5, 2005 at 7:30 p.m local time in Nanchang (top) and on July 9, 2005 at 5:30 a.m in Michigan (bottom) The horizontal axis is the time (30 seconds in this case) while the y-axis is the frequency The brightness of the image represents the vocal strength or intensity The brighter the image, the intense or loud the sound is The two spectra from Michi-gan and Nanchang showed different acoustic patterns suggesting different biological activities at the two sites
The two sites also showed different proportions of biological and anthropogenic activities Analysis of the acoustic signals in the frequency domain (Figure 17.5) suggest that Michigan site had more biological signals than anthropogenic activities while the Nanchang site has almost equal biological and anthropogenic activities, as indicated in the histograms of the frequency Although qualitative, the Nanchang site indeed had more human related acoustic signals as it is in the Center of the big city, Nanchang, China, while the site in Michigan is a residential area at the outskirts of
a small city, Okemos, Michigan The ratios of biological to anthropogenic signals (W = G/F) of the two sites are compared in Figure 17.6 and they suggest the same results as in Figure 17.5 that the biological activities are dominant at the Michigan site while the anthropogenic activities were dominant at the Nanchang site
Another type of application of the acoustic sensing technology is monitoring
of bird species through pattern recognition Once an acoustic image is generated, a signature of a specific bird, for example, can be identified (Figure 17.7) This identi-fied acoustic signature (training signature) can then be used in image processing to search for similar patterns in other acoustic data, thus detecting the presence of such bird Once expanded in time series, one can detect and monitor bird species and possibly population
17.5 DISCUSSION AND CONCLUSIONS
The research results presented in this paper represent a frontier work in expanding traditional remote sensing to acoustic sensing The fundamental difference between traditional remote sensing and acoustic remote sensing is that the former utilizes electromagnetic fields while the latter relies on air for signal transmission There-fore, a series of questions arises that needs to be addressed The first one is related
to the transmission of acoustic signals—how far does the acoustic signal travel, that
is, what is the distance between the recording device and the sound of origin? This may well depend on the location of the sensor (in forested lands, grasslands, open
Trang 7urban lands) and its surrounding physical environment One may record the acoustic signal of a bird, for example, but may also realize that the bird was just flying over rather than inhabiting the landscape where the sensor is placed Unlike traditional remote sensing where each pixel is associated with a fixed physical dimension of
a landscape (e.g., pixel size), acoustic signals do not have a fixed range of physical dimension, as the recorded signals will vary depending on the sensor’s sensitivity, distance of sound of origin, and physical characteristics of the environment (windy days, or densely forested environment, for example) Therefore, interpretation of
10000
8000
6000
4000
2000
10000
8000
6000
4000
2000
FIGURE 17.4 Sound spectra of selected acoustic signals from data collected on July 5,
2005 at 7:30 p.m local time in Nanchang (top) and on July 9, 2005 at 5:30 a.m in Michigan
(bottom) (See color insert after p 162.)
Trang 8acoustic signals is best achieved when considering the physical environment or land-scape properties
The use of acoustic signals as an ecological indicator is only feasible for infer-ring ecological information of those species that generate vocal signals Amphib-ians and mammals, for example, do not generate sounds that can be recorded with traditional recording devices Thus, at this time, we can only infer information about vocal species
The temporal characteristics of acoustic signals are critical components of any interpretation Unlike the physical environment of a landscape, the soundscape is a very dynamic field that varies considerably within a short period of time Diurnal behavior of many bird species would result in a strong biological frequency in a soundscape in the early morning, while crickets are active in the evening These
0
5
10
15
20
Acoustic Frequency Bands (kHz)
Acoustic Frequency Bands (kHz)
0
5
10
15
20
25
FIGURE 17.5 Frequency distributions of the acoustic spectra fromFigure 17.4
Trang 9temporal characteristics need to be considered when attempting to capture the bio-logical soundscape of these species
The analytical methods used in this paper are only examples in analyzing tic signals and there are other ecological indicators that can be derived from acous-tic signals However, this paper represents the first involving remote sensing that utilizes frequencies or wavelengths that can only be transmitted through a physical medium such as air Nevertheless, the expansion of the remote sensing concept to acoustic signal analysis has provides complementary and useful information about the ecological characteristics of an environment When applied spatially and tempo-rally across a landscape, much more comprehensive information can be inferred For example, a network of sensors in a city with simultaneous measurements of acoustic signals may provide not only information on ecological characteristics, but also a
0
5
10
15
20
25
China U.S.
FIGURE 17.6 Calculated alpha ( ), beta ( ), and their ratios using the data from
Figure 17.5
Sonogram
Chipping Sparrow
Time 0
11
FIGURE 17.7 Demonstration using acoustic signals in time series analysis to identify bird
species and population (See color insert after p 162.)
Trang 10quantitative measure of human-induced noise levels across the entire city, which is
a very valuable indicator of the environmental quality of the city With long-term measurements of such acoustic signals, one may further understand environmental degradation processes
Finally, this technology is relatively inexpensive compared with traditional remote sensing devices, and therefore can be deployed to obtain long-term and spa-tially distributed data Furthermore, the operation of recording devices is relatively simple and inexpensive in comparison with optical remote sensing devices, thus pro-viding a convenient technology for broader applications
ACKNOWLEDGMENTS
The Great Lakes Fisheries Trust provided support for investigating acoustic signals
as part of a grant entitled Ecological Assessment of the Muskegon River
Water-shed awarded to a consortium of investigators This work was also supported by the
NASA grant (NNG05GD49G) and by a grant at IGSNRR of Chinese Academy of Sciences (Human Activities and Ecosystem Changes) We want to thank Nathan Tor-bick for installation of the recording devices and data recording, Liu Ying at Jiangxi Normal University for his assistance in data acquisition, and Weitao Ji at the Poyang Lake Station for allowing the authors to use their facilities at Tiangxing Yuan Park and Poyang Lake
REFERENCES
Alberti, M., 2005, The effects of urban patterns on ecosystem function International
Regional Science Review Vol 28, No 2, 168–192
Allan, J David, 2004, Landscapes and riverscapes: the influence of land use on stream
eco-systems Annual Review of Ecology, Evolution, and Systematics Vol 35:257–284
Battin, J., 2004, When good animals love bad habitats: Ecological traps and the conservation
of animal populations Conservation Biology 18, 1482–1491.
Crist, P J , T W Kohley and J Oakleaf, 2004 Assessing land-use impacts on biodiversity
using an expert systems tool Landscape Ecology Vol 15, no 1, pp 1–84.
Dale, V H., and S C Beyler 2001 Challenges in the development and use of ecological
indicators Ecological Indicators 1:3–10.
Fischer, Joern and David B Lindenmayer, 2007 Landscape modification and habitat
frag-mentation: A synthesis Global Ecology and Biogeography 16 (3), 265–280.
Foley, J.A., R DeFries, G.P Asner, C Barford, G Bonan, S.R Carpenter, F.S Chapin, M.T Coe, G.C Daily, H.K Gibbs, J.H Helkowski, T Holloway, E.A Howard, C.J Kucharik, C Monfreda, J.A Patz, I.C Prentice, N Ramankutty, and P.K Snyder,
2005 Global consequences of land use Science 309, 570–574.
Grigulis, Karl, Sandra Lavorel, Ian D Davies, Anabelle Dossantos, Francisco Lloret, Mont-serrat Vilà, 2005 Landscape-scale positive feedbacks between fire and expansion of
the large tussock grass, Ampelodesmos mauritanica in Catalan shrublands Global
Change Biology 11(7), 1042–1053.
Holling, C S 2001 Understanding the complexity of economic, ecological, and social
sys-tems Ecosystems 4:390–405.
Jeanneret P., B Schüpbach, H Luka, and W Büchs 2003 Quantifying the impact of
land-scape and habitat features on biodiversity in cultivated landland-scapes Biotic indicators for
biodiversity and sustainable agriculture 2003, Vol 98, no 1–3, pp 311–320.