American National Standard on Towed Passive Acoustic Monitoring

and Mitigation Systems

Strawman Draft, July 2015

1. Scope and Organization

This American National Standard (ANSI) concerns the use of passive

acoustic monitoring (PAM) that uses towed arrays from surface vessels for

research or environmental regulatory compliance. A subset of this standard

covers operations that attempt to process data in real-time, either using on-

board or remote operators, with or without the aid of automated detection

software. The standard covers seven aspects of a complete PAM operation:

Initial planning, hardware, software, acoustic localization, operator

qualifications, operations, and performance validation. The planning and

performance validation are described here in some detail; the other aspects

are outlined in the accompanying spreadsheet.

A primary guiding principle of this standard is not to specify how PAM

operations are conducted, but to specify how an operator reports these

activities to contractors, regulators, and outside reviewers. By clarifying and

prescribing how to report PAM operations, communication between these

stakeholders can be enhanced and a virtuous feedback cycle enforced that

may gradually raise the quality and sophistication of this growing activity

without suppressing innovation. Thus the goal is that relatively few actions

in this standard will be required for compliance with this standard; however,

nearly every action discussed here will be required to be documented and

reported in a certain fashion in order to comply with the standard.

From this principle three kinds of actions are specified in this standard:

those that are required, those that must be documented and reported, and

those that are recommended. Required actions are specific actions or

restrictions a PAM operator must undertake (or not undertake) to comply

with this standard, and are indicated in this document by the use of the word

“shall”. Reported actions require an operator to document and communicate

the action in a manner described by this standard in order to achieve

compliance, but the standard does not place restrictions on how the action is

carried out. The phrase “shall report” will indicate this level of compliance.

Recommended actions are statements that reflect an attempt on consensus

opinion of “best practices” by individuals from academia, government, and

industry, and seek to serve as a reference or illustrative benchmark for PAM

activities. The phrase “should” will indicate whether an action is

recommended.

Another guiding principle behind this standard is that it requires a PAM

operator to predict and simulate system performance prior to a survey via a

planning document, and then to report to what degree the predictions and

simulations matched the reality encountered with a follow-on validation

document. The motivation behind this requirement is, once again, to

enhance clarity of communication between operator, contractor, and

regulator by allowing even a non-technical reviewer to understand whether

PAM operations achieved desired goals, and if not, to clearly understand

why. Such a formal procedure helps protect everyone’s interest by clarifying

whether issues with PAM operations arise from the operator, or from factors

outside the operator’s control. One particular scenario this standard seeks to

address is helping contractors and regulators easily understand whether

towing platform noise prevents a towed PAM operation from attaining its

required detection range.

2. Definitions and terms

PAM may be conducted for mitigation or monitoring. Mitigation is

defined here as a project whose only goal is to prevent direct physical injury

to marine life from industrial activities, while monitoring is a project with any

additional goal, including preventing behavioral disruption of marine line,

conducting an acoustic census, or characterizing the acoustic fields

generated by industrial, physical, or biological sources. These definitions are

delineated here because the working group may eventually decide to apply

different criteria to mitigation or monitoring operations. In this document

both operations are treated the same.

A PAM operator is defined as any organization that has responsibility

for the acquisition and interpretation of passive acoustic data from towed

array systems. A contractor is defined as any organization that employs a

PAM operator, and to whom the operator is expected to report. Either the

contractor or operator may be responsible for providing the towed array

hardware. A regulator is any government-based entity that has a legal

responsibility to enforce compliance with relevant laws applicable to the

region where PAM is being conducted.

Acoustical terms used here will generally follow the definitions of ANSI-

ASA_S1.1-2013, “Acoustic Terminology,” and the draft ISO-TC43-SC3-N0037

document on “Underwater Acoustics-Terminology”. In cases where

definitions may conflict between those standards, the definitions listed in this

section will apply:

Impulsive noise:

Signal-to-noise ratio:

Duty cycle:

Sonar equation:

Spherical spreading model:

Species cluster:

Industrial cluster:

3. Initial Planning

3.1 Planning and Validation documents

Before commencement of PAM operations a PAM operator shall

create a report, defined here as a planning document, that shall be provided

to contractors and regulators before operations commence. After a cruise

has been completed, a PAM operator shall create a second, validation,

document that complements the planning document. Only a single pair of

documents needs to be produced for a single operation, even if the operation

utilizes multiple towed arrays. A third document, a briefing document, shall

also be produced and given to PAM operators participating in a give project,

and will be discussed under “Operations”. See Section 9.0 for details on the

validation document.

The planning document shall contain the information listed below.

3.1.1: The planning document shall describe the time period and

geographical region to be covered by the PAM effort.

3.1.2: The planning document shall list marine mammal species

expected to be present over this time period and region, with citations of

specific databases or literature used to create this list. (An Appendix of

database resources or literature reviews can be appended to this standard,

e.g. Richardson and Greene, “Marine mammals and Noise,” or “Marine

Mammal Observer & Passive Acoustic Monitoring Handbook”). If an

operation does not plan to monitor or mitigate for certain species, the

planning document shall explicitly state which species shall not be covered

by operations.

3.1.3: The planning document shall list the acoustic bandwidth that will

need to be monitored for each species listed in 3.1.2, with appropriate

references. If a set of species is expected to share similar acoustic

characteristics, the report may then define “clusters” of species that will

streamline the rest of the planning document. For example, some clusters

might include “low-frequency baleen whales,” “beaked whales,” or “high-

frequency echolocators”. It is recommended that sperm whales and right

whales always be treated as their own species clusters.

3.1.4: The planning document shall list the industrial activities

expected during operations. This list may include not only different acoustic

sources, but also different modes of operation for each source. For example,

airgun array sources may have a “mitigation” mode and a “full power”

mode. Self-noise from the towed platform shall be listed as one such

activity. Groups of similar industrial activities may be combined into

industrial “clusters,” to simply discussion in the rest of the planning

document. It is recommended that clusters consist of activities that share

similar duty cycles, bandwidths, and 1-minute averaged received levels of 6

dB or less.

3.1.5: The planning document shall list the minimum detection ranges

to be achieved by the system. Each combination of species and industrial

cluster defined in 3.1.3 and 3.1.4 shall have a separate detection range

listed, and the manner by which this range has been assigned shall be

reported. For example, the detection range may be obtained by (i) a preset

requirement provided by a regulatory agency or contractor; (ii) derived by

the operator, contractor, or regulator, based on propagation simulations of

various industrial sources, combined with estimates of PTS and TTS

thresholds of specific species; or (iii) estimates based simply on past

empirical experience.

3.1.6: The planning document shall list key features of the hardware

system, including the hydrophones, array cable, signal conditioning, and

recording system. See “Hardware” section below and on spreadsheet for

specific reporting requirements and recommendations.

3.1.7: The planning document shall describe the software used by

human operators to monitor the acoustic data, as well as software intended

for use for automated detection and classification. See “Software” section

below and on spreadsheet for specific reporting requirements and

recommendations.

3.1.8: The planning document shall describe by what means the

bearing or range of bioacoustic signals from the towed array will be

estimated. See “Localization” section below and on spreadsheet for details.

3.1.9: The planning document shall list the expected monitoring

schedule of PAM operations (e.g. day only vs. day/night), the number of

human observers to be used during operations, the qualifications of the

observers, shift schedules, shift procedures, and data integrity protocols.

See “Operations” section below and on spreadsheet for specific reporting

requirements and recommendations.

3.2 Planning document simulations

In addition to the topics listed above, the planning document shall present

the results of simulations that illustrate the expected detection performance

of the system using the best prior information available. A simulation shall

be required for each combination of species and industrial cluster defined in

3.1.3 and 3.1.4. The simulation may be as simple as the sonar equation, or

as complex as a numerical acoustic propagation model combined with

empirical noise samples, but every simulation in the planning document shall

describe the following factors:

3.2.1: Each simulation shall list the source levels and directivity

assumed for each species cluster, and time series for typical calls for each

species cluster shall be generated, using digital sampling rates used by the

recording system. Amplitude-scaled recordings of the species in question

are recommended for producing these time series, but calls may also be

simulated by synthetic FM sweeps or impulses, provided that the algorithm

for generating the artificial signals is specified.

3.2.2: Each simulation will describe the acoustic propagation model

used for each species cluster. The model may be theoretical or derived from

empirical measurements. It is not recommended to use a simple spherical

spreading model for situations where the ocean depth is less than the

detection range, where the acoustic wavelength of the lowest frequency

component is 5% or more of the ocean depth, or where a significant

thermocline is expected in the area.

3.2.3: Inputs to any numerical propagation model shall be listed,

including bottom bathymetry, bottom composition, and sound speed profiles.

The document shall cite databases or other resources where these inputs

have been obtained.

3.2.4: The simulation shall list expected received levels of noise

generated by each industrial cluster on individual hydrophones for the towed

array system, for each bandwidth defined by each species cluster. These

received levels may be empirical measurements from previous operations, or

theoretical estimates using estimated source levels and appropriate

propagation modeling. It is strongly recommended that actual empirically-

measured time series from similar industrial clusters be used to estimate

received levels, in order to capture the non-stationary characteristics of each

noise source. If no empirically-based measurements are available, then the

planning document shall use published source levels for a given industrial

cluster, converting into estimated received levels using a propagation loss

model.

3.2.5: The simulation shall generate a time series representing the

noise generated by the various industrial clusters. If empirical data samples

are available, then the time series shall be filtered over an appropriate

species cluster bandwidth, and then the amplitude of the time series shall be

adjusted to convert received levels of the empirical measurements to

estimated received levels on the upcoming operation. The propagation

model used to make these adjustments shall be specified.

If no empirical measurements are available, then synthetically-

generated noise may be used to model the time series generated by the

activity, using filtered white noise, FM sweeps, or gated pulses, depending on

the time-characteristics of the noise source.

The noise time series may incorporate directional characteristics of the

industrial noise, if available.

3.2.6: If no significant source of industrial noise is expected for a given

species cluster and industrial cluster, then the so-called “Wenz” curve may

be used to generate “colored” noise time series with appropriate power

spectral densities listed in the curves. Industrial noise shall not be

considered a significant source of background noise if the power spectral

density of the noise lies 10 dB below the Wenz prediction of the species

cluster bandwidth-limited noise at sea state 1.

3.2.7: Time series of flow noise self-generated by the towed array

shall be estimated, using either empirical data measurements at the planned

tow speed (recommended) or by generating appropriately band limited white

noise, scaled to an appropriate amplitude.

3.3 Simulation outputs

For every simulation discussed in the previous section, a synthetic

spectrogram shall be generated, and a performance curve shall be

generated for each automated detector to be applied to the species cluster

used in the simulation.

Spectrogram: The time series generated for each species cluster

(3.2.1) shall be scaled in amplitude to match the expected received level of

the sounds when generated at the appropriate detection range specified in

3.1.5 at a representative calling depth for each species cluster. These call

time samples shall be combined with the industrial, ambient, and flow noise

time series discussed in 3.2.5, 3.2.6, and 3.2.7 respectively, to generate a

final simulated time series for each combination of species and industrial

noise cluster. If array processing is to be used in the PAM operation, then

array gain and other directionality enhancements may be applied to

generate the final “simulated” time series.

A spectrogram shall be generated from each simulated time series and

presented in the planning document. The spectrogram should cover a time

period over which a human observer would be expected to monitor real-time

data during the actual operation. The purpose of the spectrogram is to

provide a visual indication that a human operator would be able to detect a

representative signal generated by a species cluster at the specified

minimum detection range.

Performance curve:

If an automated detector and classifier (ADC) is to be used in any

capacity in the PAM operation, then a performance curve shall be plotted in

the planning document, for every combination of species cluster and

industrial cluster. The performance curve shall be generated by generating

1000 one-minute simulated time series realizations, that combine the

biological, industrial, ambient, and flow noise time series discussed in 3.2.1,

3.2.5, 3.2.6, and 3.2.7. The ADC will be applied to each realization, and a

Detection Error Tradeoff Graph (false negative rate vs. false positive rate)

shall be generated for the ADC. ADC settings that will be “default” values

during the upcoming operation shall be marked on the DET graph. The

purpose of the graph is to provide insight into what tradeoffs are expected

between the false negative rate (“miss rate”) and the false positive rate

(“false alarm”). This standard shall not specify a particular false negative

rate or false positive rate; rather, these performance metrics shall be

specified in the planning and follow-on validation documents.

3.4 Timing of planning document

The planning document shall be completed and delivered to the

contractor with enough lead time to allow the contractor (and potentially

regulators) to make go/no-go decisions on using PAM.

4.0 Hardware

This section lists required and recommended factors regarding towed array

system hardware, and lists specifications that shall be described in the

planning and/or validation documents. (See also attached spreadsheet for

division of key points into required, required to report, s recommended

columns).

4.1 Hydrophones

At least one hydrophone in each tow cable shall have a sensitivity that

spans the vocalization bandwidth of each species cluster defined in 3.1.2.

[The minimum sensitivity within a given bandwidth will be 200 uPa/V?] At

least one hydrophone in each tow cable shall be calibrated over a bandwidth

sufficient to characterize all noise sources that overlap the combined

bandwidths of all species clusters1. Every towed array cable in the

operation shall have one extra, or backup hydrophone.

The PAM operator shall report the following information about

hydrophones in the planning document, as a function of frequency: noise

floor, calibration curve with respect to frequency, directivity, mechanical

depth limit, and flow noise compensation characteristics. If multiple

hydrophones in a towed array cable share the same specifications, then the

number of hydrophones that share a given set of specifications shall be

reported as well.

4.2 Array geometry

For each towed cable in the water, the planning document shall report

the number of hydrophones in the cable, the spacing between hydrophones,

and the distance between the stern of the vessel and the hydrophone closest

to the vessel. If multiple cables are deployed from a single platform, or if

cables are deployed from multiple platforms, then the relative positions of all

1 While in principle a system with localization capability may be able to verify

detection ranges with uncalibrated hydrophones, a calibrated system is

needed to verify that the various noise source levels estimated in the

planning document are consistent with actual measurements.

hydrophones with respect to each other should be specified, whenever

information from the cables will be combined for localization purposes.

The planning document shall list the expected tow depths of the

hydrophones during expected tow speed conditions. We recommend that

the tow depths be listed as a function of several tow speeds.

4.3 Auxiliary sensors

The planning document shall list the existence and location of auxiliary

sensors in each tow cable, including depth, inclination and/or acceleration

sensors. The resolution and sampling rate of each sensor shall be reported.

We recommend that a tow cable have at least a depth sensor embedded

within the cable that can be periodically sampled every 10 seconds2.

4.4 Cable properties

The tow cable-operating load shall be twice the load expected during

maximum towing speeds.

The planning document shall list the following mechanical properties of

the towing cable: Maximum operating load, cable diameter, linear density,

and maximum radial curvature permitted. The use of weights, anchors, or

other submerging technology shall be described. The model and make of

the deployment and retrieval winch shall be specified, including the

2 As discussed below, the primary motivation for measuring towed array depth is to determine whether the system is above or below a strong thermocline.

maximum load permitted. The planning document shall specify whether the

electrical connection needs to be broken when operating the winch.

The planning document shall list the following electrical properties of

the towing cable: number and gauge of conducting wires, arrangement (e.g.

twisted single-pair), whether a grounding shield exists, and whether data are

transmitted by analog or digital means. If an analog-to-digital convertor is

employed at the hydrophones, the type of converter, the number of bits per

sample, and the sampling rate shall be specified. See Section *** for

requirements on analog/digital conversion. We recommend that cables

transmitting analog data use twisted single pairs that are in turn surrounded

by a grounding sheath.

The planning document shall describe the make and model of any

connectors used to connect tow cables together, or to connect the tow cable

to any deck cable. All connectors (including deck connector) shall be able to

survive full immersion in water without disrupting or shorting the signal. The

length of deck cable needed to connect the tow cable to the recording

electronics shall also be specified.

4.5 Signal conditioning and recording equipment

Acoustic data shall be recorded with a minimum of preprocessing or

filtering before being recorded by a dedicated digital recording system (not

used for multiple purposes like manual monitoring or manual detection). The

acoustic data shall pass through the recording system before being fed

through any computer running automated detection algorithms or subjected

to further filtering and processing for manual review. Signals shall be passed

between equipment using BNC cables (and RCA cables shall never be used).

The minimum bits per sample for any ADC system shall be 16, and the

sampling rate of any recording system shall exceed twice the maximum

frequency expected from all the species clusters. The frequency-dependent

gain and phase characteristics (transfer function) of the total conditioning

and recording system shall be measured and shall be reported in the

planning document, to permit raw numbers of digital time series to be

calibrated in terms of Pascal. **Recording on tape or other analog recorders

shall/shall not be permitted(?)**.

The recording system shall have the ability to record data

continuously. If data are not being recorded continuously, then the

recording system shall be able to buffer 30 seconds of data, so that if the

recording system is activated the previous 30 seconds shall also be stored.

The planning document shall provide a flow chart illustrating any signal

conditioning that takes place prior to data recording. The make and model

of the recorder, its ADC type, number of bits per sample, and sampling rate

shall be reported, along with the make and model of each piece of

equipment that the signal passes through before being recorded.

If any auxiliary sensors are being recorded, including GPS, the

existence of the data streams shall be listed in the planning document, as

well as the data format and sampling rate. Data shall be sampled at least

once a minute, and must be recorded continuously while tow cable is in

water. The recording system need not be the same system recording

acoustic data. Files storing data shall have a time stamp. We recommend

that at a minimum array depth and ship GPS be logged concurrently with any

acoustic data sampling.

5.0 Software

This section addresses software for both real-time and post-cruise

acoustic data processing. It is divided into software for manual review and

software for automated detection and classification.

5.1 Software for real-time human monitoring

Software intended for real-time monitoring by manual (human)

operators shall have the following features:

-The ability to generate scrolling spectrograms in a “ring buffer”: that

is, the past N seconds of acoustic data are always visible on screen.

-The ability to plot two spectrograms covering two different bandwidths

simultaneously.

-The ability to adjust spectrogram time, frequency, and intensity

scaling on the fly, the ability to adjust the Fourier transform and data

overlap, and an ability to plot either a linear or logarithmic frequency axis.

-The ability to draw bounding boxes around signals of interest, and

permanently store the associated calibrated time series from all

hydrophones in the system.

This section will probably define what non-acoustic information

(metadata) needs to be logged by software.

5.1 Automated detection and classification software

Please see spreadsheet for examples of topics for automated signal

processing.

6.0 Localization

See spreadsheet for some possible examples. In general, the required topics

will be presented first, following by the required to report topics, and then

recommended topics.

7.0 Operator Qualifications

See spreadsheet for some possible examples. In general, the required topics

will be presented first, following by the required to report topics, and then

recommended topics.

8.0 Operations

See spreadsheet for some possible examples. In general, the required topics

will be presented first, following by the required to report topics, and then

recommended topics.

9.0 Performance Validation

The PAM operator shall produce a validation document after cruise

completion that serves as a bookend to the planning document. The

following topics shall be contained in this document:

(A) Location of acoustic data streams, associated auxiliary data

streams, and metadata, along with contact information for obtaining this

data.

(B) Plot of self-noise power spectral density of conditioning and

recording electronics, as well as calibration curve.

(C) Plots of cumulative distribution functions of background noise

levels throughout the entire cruise across entire operational bandwidth, in

terms of power spectral densities, as a function of frequency.

(D) For each species cluster, the validation report shall plot

spectrograms of background noise levels, using times representing 10th,

50th, and 90th percentiles of noise levels computed across each cluster

bandwidth.

(E) Spectrogram examples showing species cluster calls at minimum

required range or greater. If no calls from a particular species cluster were

detected during cruise, then simulated species signals (same as used in

planning document) shall be inserted into spectrograms generated from time

series noise data from 10th, 50th, and 90th percentiles. The purpose of this

requirement is to provide a "common sense check" that platform noise levels

are consistent with predicted detection range, and the simulated

spectrograms generated in the planning document.

(F) "Duty cycle," or fraction of time that PAM system was running and

monitored, shall be reported. Times where mitigation/monitoring not

required may be excluded. The duty cycle includes times when system is

running, and operator effort is on.

(G) Plot of CTD/XBT cast and resulting sound speed profile, for

situations where the cast was required.

(H) Metadata for cruise shall be reported using convention:

http://abirger.github.io/animal-telemetry/passive-acoustic/#toc_5

(I) Validation of automated detectors: how?

Noise profile data shall be made available to regulator upon request, in

order to improve databases of expected platform noise levels.