4
Modeling and Databases of Noise in the Marine Environment

INTRODUCTION

The task statement for this committee states: “The study will review and identify gaps in existing marine noise databases and recommend research needed to develop a model of ocean noise that incorporates temporal, spatial, and frequency-dependent variables.” This chapter describes current acoustic models and extant databases of underwater noise and discusses efforts to model noise effects in marine mammals. High-quality, well-documented databases are essential for model validation and further model development and should contain information on the various environmental and biological factors that control the impact of noise on marine mammals. Gaps that must be filled to model the impact on marine mammals are identified for both models and databases. However, as with all models of the physical world, uncertainties in parameters and approximations in the modeling techniques are inevitable and must be accounted for using statistically valid means when interpreting the model predictions.

ACOUSTIC MODELING OF THE MARINE ENVIRONMENT

Noise in the ocean is usually broken into two broad categories based on the type of source. The first type of noise is generated by a single, identifiable, and usually close source of noise, such as an air-gun array or one or more marine mammals or other biological sources. The second type is generated by multiple indistinguishable sources of noise, such as vessels in a shipping lane and whitecaps. Some important parameters for characteriz-



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 109
4 Modeling and Databases of Noise in the Marine Environment INTRODUCTION The task statement for this committee states: “The study will review and identify gaps in existing marine noise databases and recommend research needed to develop a model of ocean noise that incorporates temporal, spatial, and frequency-dependent variables.” This chapter describes current acoustic models and extant databases of underwater noise and discusses efforts to model noise effects in marine mammals. High-quality, well-documented databases are essential for model validation and further model development and should contain information on the various environmental and biological factors that control the impact of noise on marine mammals. Gaps that must be filled to model the impact on marine mammals are identified for both models and databases. However, as with all models of the physical world, uncertainties in parameters and approximations in the modeling techniques are inevitable and must be accounted for using statistically valid means when interpreting the model predictions. ACOUSTIC MODELING OF THE MARINE ENVIRONMENT Noise in the ocean is usually broken into two broad categories based on the type of source. The first type of noise is generated by a single, identifiable, and usually close source of noise, such as an air-gun array or one or more marine mammals or other biological sources. The second type is generated by multiple indistinguishable sources of noise, such as vessels in a shipping lane and whitecaps. Some important parameters for characteriz-

OCR for page 109
ing the effects from single sources are frequency, source level, pattern of amplitude versus time (time series), directionality of radiation or beam pattern, and distance from the source. Effects from multiple unidentified sources are primarily characterized by frequency, directionality, and level at the receiver. To underwater acousticians, the term “ambient noise” refers to the second type of noise from multiple and unidentifiable sources as stated in Chapter 1. Models are used to assess the interactions of sound fields created by multiple sources, propagation through space and time, and interactions with marine mammals. The term “models” refers to a variety of tools, including empirical fits to measured data, such as the Wenz curves, computer simulation models, and numerical models, which can be either physics or empirical based. Physics models rely on known relations such as those expressed in Equations 1-1 to 1-5. Empirical models are based on observed data rather than underlying physics. In many cases the dominant mechanisms of natural sources of ocean ambient noise, for example, those associated with wind-generated noise, have not yet been conclusively identified. Therefore, physics-based approaches that incorporate actual source mechanisms are still in their infancy in underwater acoustics. In contrast, empirical models such as the Knudsen curves (Knudsen et al., 1948) and the Wenz curves (Wenz, 1962) have been extremely successful; they remain the basis of standardized noise spectra used by the U.S. and British navies. The first part of this chapter describes current acoustic models and efforts to model underwater noise effects on marine mammals. Gaps that must be filled to model the effects of noise on marine mammals are identified in modeling efforts and current databases. Modeling Single Sources of Noise Some ocean noise can be traced to a single identifiable source. High-quality models exist to predict the time series of the received signal from a source of specified directivity and given transmitted signal time series. Propagation models utilize bathymetric databases, geoacoustic information, oceanographic parameters, and boundary roughness models to produce estimates of the acoustic field at any point far from the source (see Glossary for definitions). The quality of the estimate is directly related to the quality of the environmental information used in the model. For example, in continental shelf waters, geoacoustic parameters such as compressional sound speed, attenuation, and sediment density can significantly affect the acoustic propagation. Variability introduced in these parameters can substantially affect model predictions; propagation loss can be incorrect by as much as 20 dB as a result of inaccurate geoacoustic parameters. There are four main categories of acoustic propagation models prima-

OCR for page 109
TABLE 4-1 Propagation Models and Other Information Available from the Current Contents of the Ocean Acoustics Library at SAIC Category Models Parabolic equation FOR3D, MMPE, PDPE, RAM, UMPE Normal modes AW, COUPLE, KRAKEN, MOATL, NLAYER, WKBZ Wavenumber integration OASES, RPRESS, SCOOTER, SPARC Rays BELLHOP, HARPO, RAY, TRIMAIN Other Related modeling software and data sets to support oceanographic and acoustical analyses   SOURCE: http://oalib.saic.com; Etter (2001). Reproduced courtesy of Academic Press/ Elsevier Ltd. rily used in underwater acoustics: parabolic equation (PE), normal mode, wavenumber integration, and ray models. Each of these different categories represents a different approach to simplifying either the acoustic wave equation (the fundamental mathematical equation that contains all the basic physics of sound propagation) or the model of the environment, or both. Simplification is required in order to allow computer codes to be constructed and to make them computationally efficient. Accuracy of all four model types is dependent on the frequency of sound being modeled and the environmental characteristics. In general, the PE is used for range-dependent environments at frequencies below 1,000 Hz. Normal mode models can be significantly more efficient for modeling in some environments at frequencies below 1,000 Hz. The accuracy of most normal mode models is limited in strongly range-dependent environments such as the continental shelf and slope. Wavenumber integration is usually limited to frequencies below 1,000 Hz and typically is limited to range-independent environments, although this approach recently has been extended to range-dependent environments. Ray codes are accurate and efficient for most environments but are limited to frequencies usually above 1,000 Hz. For all the models mentioned, azimuthal coupling resulting from three-dimensional medium variability (i.e., the transfer of acoustic energy propagating in one azimuthal direction into energy propagating in a different azimuthal direction) is not modeled and is considered less important than the effects of environmental uncertainty. Many propagation models are available to the public (Table 4-1). Examples of transmission loss computed using the MMPE model show the complexities of the propagation process, as well as the substantially reduced sound level at 3,000 Hz, when compared to those of 200 Hz, for longer ranges (Figure 4-1). The latter behavior is due to the effect of increased absorption at higher frequencies (cf. Figure 1-2 and Table 4-2).

OCR for page 109
FIGURE 4-1 Transmission loss in different oceanic regimes as predicted by the MMPE model at both 200 Hz and 3 kHz. (a) Arctic, with a point source depth of 50 m;

OCR for page 109
FIGURE 4-1 (b) the SOFAR channel at mid-latitude and a point source depth of 1,000 m;

OCR for page 109
FIGURE 4-1 (c) a surface duct in the mid-latitude ocean.

OCR for page 109
TABLE 4-2 Absorption by Seawater for Two Frequencies for a Range of 1,000 kma Noise Source Typical Frequency Absorption Loss at 1,000 km Absorption Loss at 10 km Shipping 100 Hz 2 dB 0.002 dB Wind 1,000 Hz 60 dB 0.6 dB aNote that these losses are in addition to geometrical spreading and scattering losses. Modeling Distributed Sources of Noise The Wenz curves are used to predict or model the noise level from unidentifiable sources (Plate 1; Wenz, 1962). These curves provide the noise spectrum level that a theoretical ideal receiver receives, given in decibels referenced to 1 µPa2/Hz. An ideal receiver has an omnidirectional reception sensitivity—in other words, its sensitivity does not vary with direction. Ambient noise is a random quantity, meaning that a given realization of the noise time series is unpredictable. However, statistical characteristics of the time series such as its variance are predictable (see Glossary). Low-frequency noise is usually much higher level than high-frequency noise due because of the character of the noise sources themselves and also as a result of the frequency dependence of sound absorption in the ocean, as described below. Typically, the property of the noise that is modeled is its pressure spectral density level. A spectrum and spectral density are frequency catalogues of a time-varying signal. The pressure spectral density of ambient noise, modeled as a random process, is the variance per hertz of the pressure time series (µPa2/Hz). For a deterministic process, the pressure spectral density is the mean squared pressure per hertz (see Glossary). Below 10 Hz, microseisms caused by the nonlinear interaction of ocean surface waves are the dominant source of ocean noise. Earthquakes also contribute intermittently. Between 10 and 200 Hz distant shipping is the largest contributor to the noise spectrum level (Wenz, 1962). From 200 Hz to 80 kHz, wind-generated breaking waves are the primary contributor to ambient noise. These levels are dependent on wind speed, and data validate the model (Felizardo and Melville, 1995). These ambient noise spectra use 1-Hz bands, while studies of noise masking in mammalian ears has typically found that one-third-octave bands are good models for these ears. For example, the one-third-octave band centered at 50 Hz runs from approximately 45 to 56 Hz. To convert a 1-Hz band level to a one-third-octave band, 10 times the logarithm of the bandwidth is added to the 1-Hz band level. For the one-third-octave band centered at 50 Hz, this translates to

OCR for page 109
about a 10-dB increase. At the one-third-octave band centered at 3,000 Hz, the difference between it and the 1-Hz band is approximately 28 dB. Ambient noise from distant sources is affected by the environment. Noise absorption by seawater is strongly dependent on frequency, effectively limiting the distance high-frequency sounds propagate (Figure 1-2). Absorption causes a decrease in received signal levels (i.e., an increase in transmission loss), which occurs in addition to the decrease produced by geometrical spreading effects, as discussed in Chapter 1. The absorption of shipping noise in the 1-Hz band, centered at 100 Hz, is approximately 0.002 dB per km. In other words, 1,000 km from a source of 100 Hz, the attenuation loss is about 2 dB in addition to the geometrical spreading losses. For higher-frequency sound, such as that generated by wind at 1,000 Hz, the absorption factor increases to approximately 0.06 dB per km. At a distance of 1,000 km from a 1,000-Hz source, the attenuation loss is about 60 dB (Table 4-2). For distant sources of ambient noise, frequency largely determines the region over which these sources can be important. Ships contribute to ambient noise at ranges of hundreds of kilometers, while wind noise contributes to ambient noise for distances of kilometers. It should be noted that the majority, if not all, the models for oceanic ambient noise have been developed for and supported by Navy sponsors. Appendix C provides a summary of underwater acoustic noise models. This summary is not meant to be all-inclusive but rather to indicate some of the better-known and more heavily used examples. Over the decades since World War II, naval sonar systems, starting from simple transducer units, have increased in complexity. Initially, the sonars operated in an omnidirectional mode and required only a knowledge of ambient noise as seen with that sensor. As the systems acquired more and more directional discrimination to help localize targets, knowledge of ambient noise directionality was also required. The initial attempts to define and measure noise directionality were confined to studies of the variations either in the vertical direction only or the azimuthal direction only. Later, as the sonar arrays became even more spatially discriminating, beam noise estimates were required where both horizontal and vertical limits were used. From the perspective of an omnidirectional system, the Wenz curves would be the model required. The summary in Appendix C, then, goes on from that point, with the ambient noise models where there are listed a number of directional models with respect to either the horizontal or vertical plane(s). The beam noise statistics category provides those models that describe beam noise properties. For more details the reader is directed to excellent texts devoted to underwater ambient noise, modeling, and mechanisms (Etter, 1996; Applied Acoustics, 1997). Models such as ANDES, CNOISE, and RANDI provide predictions of the geographic, seasonal, frequency, and directional dependence of ambient

OCR for page 109
FIGURE 4-2 Output of Comprehensive Acoustic System Simulation/Gaussian Ray Bundles (CASS/GRAB) Model. The thick solid curve shows the base level with no shipping noise, a sea state of 0, and no rain. The seven parallel dashed curves from 10 Hz to 100 kHz show the surface agitation component only for sea states 0 through 6 (in ascending level). The three dashed curves from 550 Hz to 15.5 kHz represent the rain component for intermittent (lower curve), moderate (middle curve), and heavy (upper curve) rain. SOURCE: Naval Undersea Warfare Center Division. noise from multiple unidentified sources such as distant shipping and wind. These models include shipping density statistics, wind-speed databases based on meteorological models, state-of-the-art propagation models, and oceanographic databases. The models are usually used for sonar performance prediction and maintained by the world’s navies. No existing model is capable of predicting the effects of distributed noise sources on marine mammals. An omnidirectional ambient noise model is included as part of the sonar simulation model CASS/GRAB (Comprehensive Acoustic System Simulation/Gaussian Ray Bundles; Weinberg and Keenan, 1996; Weinberg et al., 2001). The CASS/GRAB model, approved by the Oceanographic and Atmospheric Master Library (P.C. Etter, 1996, 2001) was developed at the Naval Undersea Warfare Center (formerly NUSC) using empirical fits to ambient noise measurements (Figure 4-2). The model accounts for “ocean

OCR for page 109
turbulence,” dominant in the 1-10-Hz band, shipping noise prevalent from 10 to 500 Hz, “surface agitation” from 500 Hz to 100 kHz, and thermal noise at frequencies greater than 100 kHz. Noise from rain also is included in the 550-Hz to 15.5-kHz band. Dynamic Ambient Noise Prediction System The Dynamic Ambient Noise Prediction System (DAPS) is the most recent development in the succession of U.S. Navy ocean ambient noise models. It is composed of three modules: Historical Vessel Module, an updated Historical Temporal Shipping (HTS) database containing information on commercial ships and fishing vessels with a simulated vessel movement module; Dynamic Ambient Noise Module (DANM), successor to the ANDES program; and Reported Vessel Module. DAPS was designed to predict the azimuthal dependence of ocean noise in the 25-5,000-Hz frequency band, including surface shipping and wind-generated noise. Lloyds of London records were used for initial shipping spatial distributions, and ship tracks were inferred from shipping lanes as input to a propagation model. Fishing vessel activity used historical vessel distributions and fishery statistics collected by the Food and Agricultural Organization of the United Nations to incorporate fishing vessel densities. The wind-generated component was obtained from the Surface Marine Gridded Climatology and empirical relations between the ocean ambient noise levels and wind speed. The DANM module presently is being reviewed by the U.S. Navy’s Oceanographic and Atmospheric Master Library (OAML), which is responsible for maintaining and distributing standardized databases and models to the U.S. Navy fleet. If successful, DANM will be the first ambient noise model to obtain OAML approval. MODELING THE EFFECTS OF NOISE ON MARINE MAMMALS A conceptual model can assist in describing the interactions necessary to assess the impact of ocean noise on marine mammals and other marine animals. The ocean noise input to the system of marine mammals consists of all types of ocean noise, including those generated naturally by physical and biological means and those generated from human activities (Chapter 2). The system being evaluated consists of marine mammals and in the simplest terms can be treated as multiple environmental and physical factors on which the ocean acts to produce the output. The output consists of metrics that can be used to assess the impact of the ocean noise on the

OCR for page 109
FIGURE 4-3 Components of the ocean noise input to the overall conceptual model. system. Such measures may be physiologically based, such as noise levels that produce temporary or permanent threshold shifts in given animals, or behaviorally based, such as sound levels that cause cessation of mating calls. Ocean noise can be dispersed (Figure 4-3) and is capable of incorporating available (sub)models. These existing models can be used to determine the appropriate input to a model and evaluate a given scenario. A model of effects that predicts the impact of acoustic signals on marine mammals should consist of six main components: (1) a description of

OCR for page 109
the source time function or source spectrum and level and a model of the source distribution, (2) physical oceanographic and geoacoustic databases, (3) models of marine mammals distribution in three dimensions to determine exposure, (4) models to predict the sound signal at an animal, (5) biological databases and models for marine mammal hearing and movement, and (6) population-based models to look for effects at these levels. Recent breakthroughs in the understanding of the effects of noise on animal hearing along with developments in the understanding of acoustic propagation have enabled the combination of hearing models with acoustic models, referred to as integrative models. These integrative models include a physical oceanographic component that controls the propagation of sound. Integrated models include a library of common sound sources (biological and man-made), environmental features that affect sound propagation such as bathymetry and ocean dynamics, and algorithms for modeling sound propagation, such as PE models. Biological components are divided into the following: animal distribution databases; animal behavior data and models, including migration, diving patterns, and behavioral responses to sounds; and models for the mechanical and neural responses to sound by the organism. Systems architecture of integrative models can be designed to include data synthesis display and communications tools that enable investigators to work as a distributed network and databases and modeling algorithms that are shared among widely distributed universities, labs, and data centers. The goals of these models or their successors are to predict the outcome of a given sound exposure regimen and to represent that information in dynamic graphical displays and probabilistic functions. In other words, model predictions will be quantitative with quantified limits of uncertainty. One example of these new and integrative animal effects models is the ESME (Effects of Sound on the Marine Environment) model, sponsored by ONR. At present ESME is halfway through the four-year development plan. ESME identifies the elements necessary for a predictive risk assessment model and develops an architecture for fitting the pieces together. Not all of the necessary databases are full, and gaps in understanding still exist. To date, a basic structure has been developed and applied to two simple but realistic test problems. The first scenario dealt with the effects of noise on dolphins in the Southern California Bight. The second test problem focused on dolphins in the Middle Atlantic Bight south of Rhode Island. These test problems assess intermodule communications and test different databases and modeling algorithms. The tests also examine different configurations for ESME and its successors. Several alternative models for further development will likely result from ESME, ranging from the simple to the most complex.

OCR for page 109
A relatively simple integrative model, such as the Acoustic Integration Model (AIM), could be PC based to enable a wider range of users to experiment with underwater sound scenarios, providing educational, scientific, and environmental management functions. AIM was designed to model the movements and behaviors of acoustic sources and receivers. These receivers are virtual animals and have been dubbed animats. The AIM model interfaces with another acoustic propagation model that simulates the acoustic field produced by the acoustic source(s). The animats can be programmed to simulate natural responses, including reactions to the sound field. The acoustic history of each animat is recorded, a valuable and important output. The model allows multiple Monte Carlo model simulations to estimate the impact of various scenarios. At the other end of the scale, ESME or an equivalent tool might be integrated with state-of-the-art complex multidimensional physical ocean models running on supercomputers. A complex version of the integrative tools would clearly limit accessibility to only the most sophisticated users but would offer the greatest possible flexibility and accuracy. At present, the integrative models are concentrating on modeling effects of individual sound sources on individual animals or individuals within pods. The effects of distributed sources are not, at this time, being investigated. Hearing at the individual level is being modeled at several levels, from the micromechanical activity of the inner ear through whole head resonance. Inner ear models are based on basilar membrane response data for well-studied ears, especially in vivo measurements in mice, cats, and gerbils. Inner ear structural data on these ears are being compared with parallel data from representative marine mammal ears (mysticete, odontocete, and pinniped) in order to modify the inner ear response models to accurately represent stiffness and mass variations in marine mammals compared to smaller land mammals. This will affect both sensitivity and frequency responses in marine ears. Middle ear and whole head responses, particularly head transfer functions, are two areas for which no adequate land analog exists. Models for these elements of hearing are being formed based on direct measures of marine mammal tissue mechanical characteristics, acoustic impedances, and complex tissue resonance. No attempts have been made to model the effects of noise on the habitat and ecosystem of marine mammals. Fish and other marine organisms respond to noise in both experimental systems and marine environments. Because they are prey items for some marine mammals and are important components of the marine ecosystem, it is also necessary to examine the effects of noise on these organisms to incorporate all of the effects of noise on marine mammals.

OCR for page 109
DATABASES Ancillary Data for Effects Modeling The data necessary to allow modeling of the overall effect of ocean noise on marine mammals are quite varied and in general do not yet exist in the volume and completeness needed. There are three major categories of data that are required: (1) data that characterize sources, (2) data that characterize how acoustic energy propagates from the source to the animal, and (3) data that characterize the effects that sounds have on marine mammals and fishes, both physiological and behavioral. The information needed about the sources is the characterization of the source itself, such as its output level, its frequency band, and so on, as discussed at the beginning of this chapter, and the activity level of the sources, where they are operating, and when. One needs to know the velocities, densities, and attenuation factors in the water column and in the upper strata below the seafloor to describe accurately the propagation of sound waves from the source to the animal some distance away. The information is needed to characterize the effects that sounds have on marine mammals and the specific research topic being examined. Data do exist that fall into all three of these categories, but they are incomplete, scattered, and, in many cases, inaccessible for national security reasons. However, two programs are specifically addressing the potential impact of ocean acoustic noise on the marine environment by developing comprehensive databases. These databases presently will be used for rapid data retrieval, mapping, and statistical correlation studies, but they also could be used as inputs to future physics-based, numerical modeling efforts. One is the Sound, Oceanography, and Living Marine Resources (SOLMAR) program at the NATO Supreme Allied Commander Atlantic Undersea Research Centre in La Spezia, Italy. Data, primarily from the Mediterranean Sea, are being assembled on the occurrence of cetacean strandings, results of visual surveys, and underwater acoustic recordings of vocalizations. Standard oceanographic and geophysical measurements such as water mineral and chlorophyll content, conductivity/temperature/depth profiles, and bathymetry data are also being collected. In addition, satellite-based measurements such as altimetry, sea color, and sea surface temperature are also being collected. SOLMAR databases are applied to a Geographical Information System (GIS) framework. Another program, the Living Marine Resources Information System, is developing databases of global distributions of marine animal species with no acoustic data. The primary sources of information presently are the National Marine Fisheries Service visual survey reports, as well as other publications in the open literature; however these data are confined largely to coastal areas. The Census of Marine Life is a new international effort to determine

OCR for page 109
numbers and types of marine organisms and their geographic and depth distribution worldwide. One part of this program is an open-access, Internet-based collection of databases and associated processing tools called the Ocean Biogeographic Information System (OBIS). The databases will initially include components such as a history of marine animal populations, biogeoinformatics of Hexacorallia for corals and sea anemones, and data on the chemosynthetic ecosystems in the Arctic and northern Atlantic Oceans. Also included are oceanographic and environmental databases, all in a GIS framework to permit full ecological system assessment. The computer and communications-based setting is expected to permit computational functionality among internationally distributed systems. Organized information about ports and shipping lanes is maintained by the U.S. Navy’s Space and Naval Warfare Systems Command, which defines 521 ports and 3,762 traffic lanes. Lloyds of London maintains information about the merchant fleets of the world, the number of ships in each ship-type category, and gross tonnage. Oil industry activity that contributes most to ocean noise can be monitored by subscribing to any of a number of commercial information services. For example, IHS Energy provides relatively comprehensive information dating back to 1994 about individual marine seismic crews and where they are, and have been, working. The location data may be no more specific than “North Sea,” and no information is given about the specifications of the air-gun arrays being used. IHS Energy will research its database and generate reports for a fee. A similar type of service is provided by ODS-Petrodata with regard to offshore mobile and platform drilling rigs. Together, these two services can supply an overall picture of where these noise sources are in time and space, but neither provides information about the noise generated by these operations. Measurements need to be recorded of different drilling techniques in different environments to determine if they make enough noise to cause concern. If so, then a catalogue of the noise output of the different techniques should be maintained and used to calculate the contribution to the noise budget from drilling rigs. Because the information is considered proprietary, it is unlikely that the details of airgun arrays will be included in the seismic crew databases. Using published values of air-gun array source levels of 260 dB re 1 µPa-m, peak-to-peak (Richardson et al., 1995), will produce estimates that err safely on the high side. This level is best used for the output oriented vertically, and for the horizontally oriented output, a number around 235 dB re 1 µPa-m, peak-to-peak, is more suitable. There appears to be no suitable, all-inclusive source of information about offshore construction activities outside of the oil and gas industry. These activities include cable laying, dredging and reclamation projects, tunnel boring and bridge building, dock building, and port construction. Petrodata’s Marine and Coastal Construction System is a database and

OCR for page 109
newswire service that provides information on planned worldwide marine and coastal construction projects. It seems most adept at capturing projects located in Europe, Asia, and the Middle East. Again, measurements of the noise created by these activities are not numerous, so this is an area where much work needs to be done if an assessment is to be made as to the importance of these activities to the ocean noise budget. Data on the physical properties of the ocean waters and the near-seafloor sediments exist in detail in some places and are nonexistent in others. It is beyond the scope of this report to discuss this topic here other than to say that in some instances it is crucial to know the details of the seafloor topography, the details of the water column sound speed and absorption properties, and the details of the seismic velocities, densities, and absorption properties of the strata below the seafloor. Although there is an extensive literature on the effects of sound on marine mammals, it is patchy and inconclusive. A tremendous amount of work remains to be done to determine the effects of sound on marine mammals. In particular there have been few studies to relate specific dosage of sound to effects likely to be of biological significance. One of the recommendations of this report is that a single federal agency or organization be charged with the responsibility of overseeing all of these activities, including all data collection. As more and more locations around the world place restrictions on activities that create noise in ocean waters, and such restrictions cause data to be collected with regard to this issue, it seems prudent to establish an official body that catalogues these different data sets, if it does not actually oversee the storage and archiving of them. Ocean Noise Databases Currently there is no coordinated program to organize, support, and execute an ongoing data collection effort to supplement the general ambient noise data sets that were the basis of empirical curves such as those of Wenz and Knudsen. There are ongoing individual efforts, but they are incomplete, scattered, and in some cases may not be available because of national security reasons within the United States and other nations. Typically, these efforts are focused on averaged values of the acoustic pressure spectrum and transients are excluded. One significant collection is the archived information of the U.S. Navy, held by the Naval Oceanographic Office (NAVOCEANO). Nearly 50,000 omnidirectional measurements of ambient marine noise are held within the NAVOCEANO Data Warehouse. Data collection began in the 1950s and is organized by season, frequency, location, and time. NAVOCEANO also maintains wind noise estimates based on the model projections using adaptations of the Wenz curves (Plate 6a-d). Representative samples of NAVOCEANO archives for two seasons

OCR for page 109
(summer and winter) and two frequencies (50 and 3,500 Hz) highlight the potential usefulness of such a dataset (Plate 7a-d). The data collected are oriented to geographic regions of past, current, and future naval operations interests. Data measurements vary in duration of collection, from very short (<1 hour) to drifting buoys that gather data for weeks to months. Through careful analysis data collected in the presence of known contaminants (seismic sources, nearby passing ships) were discarded. Perhaps the most striking feature of these figures is the lack of data in most of the world’s oceans. Additional noise databases can be found in Etter (1996). Etter’s Table 10.3 lists the Advanced Environmental Acoustic Support data bank as well as the NAVOCEANO database, and Table 10.5 contains noise databases that reside in the OAML. These OAML-approved databases include three shipping noise databases that cover all of the northern hemisphere as a function of season, Arctic noise near the marginal ice zone on a monthly basis, and the wind and residual noise database, which provides monthly variations in noise levels not containing shipping for the northern hemisphere. Access to the databases listed is restricted, making it difficult to review them and use them for scientific purposes. All were gathered to meet U.S. Navy sonar system needs. Much of the data probably were not collected in a systematic way using fixed procedures. A clear bias toward the horthern hemisphere exists. Other ambient noise data sets can be found in various places outside the operational navy community. As one example, the National Oceanographic and Atmospheric Administration has been collecting SOSUS (Sound Surveillance System) data off the coast of the State of Washington since 1991 (Chris Fox, personal communication to committee, 2001). However, at present, a major gap in existing noise databases is that no long-term (greater than a decade), systematically collected, ocean acoustic data set exists for any frequency band. Additional gaps in marine noise databases include the facts that no noise database is known to exist for the southern hemisphere except the set of measurements made around the continent of Australia by the Defence Science and Technology Office, and possibly those in the waters off New Zealand. In addition, no systematic noise monitoring data set has been collected in biologically sensitive areas for specific species. Finally, if the whole frequency band from 1 Hz to 200 kHz is taken as the band of interest, a gap exists in databases at frequencies above several kilohertz. Additional planning is required in collecting data at high frequencies because of the large data rates involved; continuous sampling is not practical unless some type of real-time processing is implemented. A well-recognized issue with ambient noise measurements, particularly in shallow water, is the effect of the propagation characteristics on the

OCR for page 109
received field. Therefore, a gap in existing shallow-water noise databases is lack of knowledge of the ocean-bottom geoacoustic properties in the regions where the measurements were made. More generally, the quality of ambient noise databases is directly related to the quality and variety of ancillary information (e.g., near-surface winds, shipping traffic, visual observations of marine animals) collected simultaneously at the same location. Development of a long-term ocean noise monitoring system requires careful consideration of which types and in what ways this supporting information will be collected. SUMMARY Sound sources in the ocean can be categorized and modeled as two main types: unknown distributed sources (that is, unknown location, source level, and spectral content) referred to as ambient noise and best modeled as statistical in nature, and identified single sources best modeled deterministically. Noise from the collection from all sources is referred to as “ocean noise” in this report. The dominant source of ambient noise is associated with ocean surface wave activity. In the frequency band from 5 to 200 Hz, shipping may be dominant, at least in the northern hemisphere. The time-averaged received levels of shipping noise in some locations can be fairly well modeled. Above 200 Hz, noise levels from breaking waves are roughly modeled through the use of empirical relations between noise level and wind speed. Limitations exist in ambient noise models not just from lack of knowledge of the source characteristics and distributions but also resulting from uncertainties in the environment. The sounds from single sources, such as sonar and air-guns, are usually well modeled by propagation codes. The accuracy of these models is limited by environmental uncertainty. The effects of sound from single sources on marine mammals are beginning to be modeled by integrative tools such as AIM and ESME. The effects of distributed sources, such as shipping and wind, on marine mammals are not yet well modeled. From field observations and threshold experiments on captive animals (see Chapter 3), it is clear that sound can disturb marine mammals both behaviorally and physiologically. Noise from shipping may be affecting marine mammals adversely. Similarly, high-intensity transient sources at short ranges may have significant effects on marine mammal physiology or behavior. Modeling these effects is possible and prudent. While modeling the physiologic effects is relatively straightforward, modeling behavioral effects is difficult and needs more effort. In all cases, field data must be collected to validate the model predictions.