Archival PAM recordings are primarily used for long-term monitoring, with capabilities for recording durations ranging from several weeks up to several years (Sousa-lima et al., 2013). Continuous recordings provide an uninterrupted record of species’ acoustic presence, allowing investigators to evaluate species distribution and occurrence, and changes in animal calls, which provide information on behavioral state (e.g., foraging, reproduction, socializing) in a given area or region. In addition they allow for the evaluation of seasonal, inter and intra annual variation in species presence and occurrence over time. Alternatively, recordings can be duty-cycled (defined as the fraction of time that a PAM system is actively recording) to maximize recording duration at sea while limiting equipment interactions (e.g., retrieval to swap out hard drives or batteries). Duty-cycled data is less preferred, as inevitably some information is lost, and biases are introduced by using a reduced recording schedule. If the duty-cycle listening period and recording interval are not appropriately matched to the duration and timing of animal calls, potential detections may be missed and species occurrence underestimated (Miksis-Olds et al., 2010; Sousa-lima et al., 2013; Thomisch et al., 2015; Stanistreet et al., 2016). For example, in Thomisch et al. (2015), duty-cycling at 50 and 2% showed a decrease in accuracy in both acoustic presence and call rate estimates. If it is necessary to duty-cycle, frequent and shorter recording periods may improve accuracy of daily acoustic presence. Duty-cycling effects are most pronounced for species with low and/or temporally clustered calling activity (Thomisch et al., 2015)—i.e., non-song call types such as the North Atlantic right whale—and are less pronounced for species that click over long time intervals, such as beaked whales (Stanistreet et al., 2016). For higher frequency species, a PAM click detector recorder can be an efficient method for data collection (e.g., Bailey et al., 2010a; Temple et al., 2016; Wingfield et al., 2017). The click detector recorder stores continuous higher frequency clicks of delphinids, harbor porpoise, and other high-frequency odontocetes but does not provide a PAM sound record as is the case for most other recorders.

Archival PAM systems are often moored near the seabed with no surface expression and are returned to the surface by divers or by using an acoustic release mechanism. Acoustic data are therefore only recovered and analyzed at the end of the recorder deployment. Consequently, the analyses that are conducted will be retrospective and not real time. However, the need to wait for data records until retrieval can be resolved by using archival PAM systems with surface expression such as those used by Brandt et al. (2018) allow for more frequent data collection. Archival data can be useful to build long-term monitoring records of the presence of sound producing species, both temporally (seasonal and yearly occurrence) as well as spatially (occurrence in and across different regions). These data can also be valuable for evaluating potential effects of construction as changes in species presence and behavior can be correlated with construction activities.

Real-time PAM systems that enable rapid detection and recognition of marine mammal calls are invaluable for monitoring but are essential components for mitigating potential effects from wind energy development. Real-time acoustic alerts can be used to respond quickly to the presence of protected species in a construction area (e.g., during impact pile driving as long as the presence of multiple construction noise sources do not mask their presence) or in the vicinity of transiting vessels, thereby reducing the risk of vessel strike (e.g., Spaulding et al., 2009; Baumgartner et al., 2019, 2020, 2021; Norris et al., 2019; Kowarski et al., 2020; Wood et al., 2020). Real time is defined here as the relay of PAM data (processed or raw) within an operationally usable time span (e.g., data relay frequency may range from every minute, hour, to daily, depending on how quickly the information is needed for decision-making). In effect, any data from the acoustic detection can be used to optimize, or at least provide, timely information to help direct current operations and/or tracking of a species (Klinck et al., 2012; Baumgartner et al., 2013, 2019, 2020; Kowarski et al., 2020). Real-time PAM can also be used to improve and adjust to noise produced by pile driving, by offering real time feedback on the noise produced by the hammer and the capacity to adjust this if needed. Real-time PAM can be conducted from a variety of platforms, including vessels, surface buoys, autonomous vehicles such as gliders, and drifting buoys. The PAM data will travel from a PAM recording sensor to the receiving station on shore/vessel at regular time intervals agreed upon in the PAM Plan. The frequency of data relay is constrained by the type and cost of data upload. Cell phone towers can be used for data relay if sensors are located close to land, as reception tends to be lost beyond 15 miles offshore. Iridium satellite data transmission (currently the most common type) costs are based on the quantity of data and frequency at which you want to upload, and which data service plan you select. Another possible option is cabled arrays laid out on the seafloor, which can also provide a real-time data feed straight to shore where such an installation is feasible [e.g., Lindsey et al. (2019)].

Passive acoustic monitoring fixed, bottom-mounted archival recorders are moored on or near the ocean floor for several weeks to months (many recorders can now record continuously for 4 to 6 months), and up to several years. They should be spaced at distances that encompass the estimated calling/detection radius of the species of interest when they are within, and in the vicinity of, the operating area. The detection radius will vary depending on a number of factors, including those influencing signal propagation (i.e., water depth and temperature, substrate type, noise levels), as well as source level and directivity differences in calls between individuals and species. An estimate of the minimum number of hydrophones needed for detection of the call types of the species of interest should be made and hydrophones placed with these considerations in mind to minimize potential missed portions of the operating or surrounding areas (Figure 4; Table 3). The percentage of area that is desirable or required to be covered by the hydrophone spacing needs to be considered. An example of such a design can be seen in Supplemental Information I, where we aim to have 100% coverage of the acoustic radius of a calling North Atlantic right whale individual within the lease block areas and 50% coverage outside of the areas. In addition, where required, multi-element bottom-mounted hydrophone arrays can be used to track the movements of calling individuals (e.g., Stanistreet et al., 2013).

Passive acoustic monitoring surface buoys are a valuable technology used for both relaying real-time information at regular intervals as well as collecting long-term archival data from a single location. They have been used effectively for monitoring and for mitigation purposes, spaced at appropriate listening distances for the species of interest, in busy shipping lanes, as well as in numerous other areas, including prospective wind leases (Table 3; Supplementary Information II). As there is a connection to the surface, noise produced by the mooring/floatation system can add unwanted noise and must be carefully considered in the design phase. In the cases where these buoys have been used successfully, the surface recorders were anchored using a special mooring system constructed to reduce noise (e.g., Baumgartner et al., 2019). PAM moored surface buoys are essential for the purpose of real-time mitigation, but they are also extensively used for long-term monitoring of species presence. PAM real-time buoy systems should be placed with similar listening distances in mind as for PAM bottom-mounted recorders. However, depending on the use of the platform and requirements, the number of moored surface buoys may differ. If the intent is to minimize ship strikes in a high vessel transit area, the PAM moored surface buoys should cover the vessel transit lanes (see example in Table 3 and Supplementary Information II). However, if the intent is to listen for the presence of endangered species in the vicinity and within the wind lease construction area, then the number and spacing of buoys should reflect the effective listening area that needs to be covered (see Supplemental Information II for an example).

An Autonomous Underwater Vehicle is a robot that travels underwater without requiring input from an operator. Underwater gliders are a subclass of Autonomous Underwater Vehicles and can have a single PAM recorder or a PAM array placed inside or strapped to the outside of the vehicle. Autonomous Underwater Vehicles are boats or ships that operate on the surface of the water without a crew and can similarly be equipped with PAM recording equipment. PAM Autonomous Underwater Vehicles are very versatile and increasingly used to monitor large spatial areas [with the capacity to cover 10 to 1000 s of kilometers (Baumgartner et al., 2020)] over long time periods (generally from 3 to 6 months) and to relay back information either in real time or through archival recordings. They can be programmed to follow tracklines and navigate to new positions throughout a deployment, either diving up and down the water column to collect oceanographic data or following a straight path at a set depth. Gliders can be categorized into battery-operated gliders and wave-propelled gliders. The former have been demonstrated to be highly effective for real-time monitoring and mitigation for North Atlantic right whales and other baleen whale species (Table 3; Baumgartner et al., 2014, 2020; Kowarski et al., 2020). They have also been shown to be valuable in understanding the spatial distribution of toothed whales, such as beaked whales (Klinck et al., 2012). Wave gliders have had some success in archival monitoring of toothed whales (Küsel et al., 2017) but still require further development. In particular, the self-noise produced by the wave glider continues to limit the detection capability of low-frequency baleen whales (Baumgartner et al., 2021). Further technological development may be able to shield this noise in the future. Similarly, Autonomous Underwater Vehicle technologies, such as sail drones and self-navigating vessels, are in development and will likely present new and innovative solutions to add to the current suite of real-time and archival solutions for PAM in the near future (Klinck et al., 2009; Mordy et al., 2017).

Autonomous Underwater Vehicles are ideally suited for monitoring an area to inform BOEM and NOAA mitigation requirements. Their tracklines can be remotely piloted or redirected if needed while the instrument is out at sea. For example, if the glider locates an area of high vocal activity, it can be instructed to stay in place and only move on once the activity has decreased. Currents and wind can influence coverage of tracklines, and this needs to be taken into consideration during the design process (Figure 5). Again, tracklines should be designed with consideration of the listening radius of the focal species and the total detection radius of the mitigation and monitoring area that is required. Gliders are slow moving instruments (∼ 3 knots) that provide point sample data. If the aim is to understand a small area in great detail so as to not miss the presence of the species of interest, then dense tracklines that are surveyed frequently are recommended. If the intention is to cover a large area to detect individuals before an activity requiring mitigation occurs or to inform BOEM and NOAA mitigation decisions over larger areas, then broader tracklines are needed to be able to cover the detection area in question. This will result in less frequent coverage of a trackline, unless several gliders are used. The detection radius and duration it would take a glider to cover the entire desired mitigation area would need to be considered in determining the confidence in monitoring of the detection area for broadly spaced tracklines over a set time period, for example over 24-hour segments. In contrast to fixed recorders, the information that gliders can provide is sparse but spatially broad (Table 3).

Passive acoustic monitoring drifters involve the deployment of acoustic recorders that have a single hydrophone and/or a vertical array suspended in the water column, depending on their design. They typically have a surface expression that allows for satellite tracking capabilities for subsequent recovery. Due to their surface expression, they can provide either real-time information or archival recordings, depending on what is needed. Their location and movement patterns are dependent on currents and wind. Unlike gliders, they cannot be repositioned or redirected remotely. This technology has potential, but has not yet been extensively used or tested in many studies, although it shows promise for certain applications such as monitoring species presence, estimating abundance and measuring ocean noise metrics (e.g., Barlow et al., 2014; Griffiths and Barlow, 2016; Fregosi et al., 2018). When incorporating vertical line arrays in PAM drifters, it can provide information on vertical bearing angles, thus obtaining the estimates on the depth of vocalizing animals (e.g., Griffiths and Barlow, 2016).

Passive acoustic monitoring drifter placement and density can vary greatly depending on the monitoring goal and the local oceanic conditions. Currents and wind direction would need to be well understood because they will affect each drifter independently. It is important to determine how the oceanographic conditions of the area will affect coverage of the monitoring region using these types of platforms. For example, if the site is near areas of high current, such as the Gulf Stream, the recorders may be quickly displaced when they reach that current. However, in a more sheltered, less dynamic environment, PAM drifters could provide good coverage. Drifters can be outfitted with satellite transmitters and so can be retrieved once their data collection period is over. In contrast to fixed recorders, the information that drifters can provide is sparse but spatially broad; however, unlike Autonomous Underwater Vehicles, they are constrained in the areas they cover by ocean currents, tides, and wind (Table 3).

When it is necessary to know the range, bearing, location, or depth of a vocalizing marine mammal (e.g., for real-time tracking or estimating abundance of individuals), an array of time-synchronized hydrophones is required (Thode, 2004; von Benda-Beckmann et al., 2010; Gillespie et al., 2013; DeAngelis et al., 2017). PAM arrays may be towed behind a vessel that is underway, but in this case, flow noise can obstruct low-frequency sounds from being heard. Generally, lower speeds produce lower flow noise; however, speed must also be considered to maintain horizontal orientation of the towed array. This generally prevents towed arrays from being useful to monitor for low-frequency baleen whale species unless the ability to record low-frequency sounds over the noise is clearly demonstrated. The Acoustical Society of America (ASA) is currently working to develop an American National Standards Institute-approved standard for towed hydrophone arrays. The fundamental goal of this ASA standard is to reduce situations where background noise levels prevent effective PAM. To achieve this goal, the standard employs a suite of strategies to standardize how acoustic measurements are logged, reported, and evaluated (Thode and Guan, 2019). Typical PAM arrays for monitoring purposes can be towed behind a vessel (or Autonomous Underwater Vehicle), in which case the tracklines to be covered by the vessel (or Autonomous Underwater Vehicle) with the array need to be designed with the listening radius of the system and species in mind (Table 3). Other types of arrays, such as bottom-mounted cabled arrays, are not discussed here because they are less frequently used, given the cost and infrastructure needed to lay and maintain them.

When PAM is utilized, the design intent may be to:

All real-time PAM designs need a clear and well thought out and consistent process, including PAM placement and technology type, species detection, integration with other visual data observations, communication of information to Protected Species Operators (PSOs)/shoreside operators, and response of the information/detection. Limitations of each real-time PAM system should be well understood and considered in detail in any PAM Plan. Efficacy of the system used and its capacity to detect the signal/species of interest is essential to developing a successful and credible PAM Plan. Lastly, it is important to note that no one PAM system is capable of answering all needs and that frequently a mixture of PAM systems, technologies and designs are likely needed to address all monitoring and mitigation requirements.

For all PAM technologies, the hydrophones and related hardware need to be calibrated (every 3 to 5 years) and their performance systematically measured and optimized within frequency bandwidths of interest for the particular activity, species, and environment. Calibration data, and relevant settings and sensitivities should be noted for all hardware used in recording/monitoring to ensure consistency among measurements for particular hardware and software [more detail can be found in Biber et al. (2018)]. Array synchronization information (where relevant) should also be documented. This information should be permanently associated with the recordings as metadata.

All hardware should be tested and optimized for low self-noise, including the mooring system. In addition to calibration, the system should be fully tested to ensure adequate sensitivity in the area where it will be deployed and with the type of signals it would receive. Additional environmental data will need to be collected to allow for adequate system evaluation. If this cannot be done at the project site, the system should be fully tested in a comparable location (i.e., an area exhibiting similar depth, temperature, substrate, current, acoustic propagation, and ambient noise, with relevant sound sources).

At a minimum, the following specifications should be measured and reported on:

The system needs to be designed, installed, and operated by those having expertise with the specific PAM technology, including placement in the water, attachment of cables to reduce strumming and noise, acoustic release, suitable anchorage for the conditions, software use, etc., Knowledgeable and experienced personnel should operate the units in all situations.

Ideally, the PAM technology used should have been used for the same purpose in other field efforts and have clear and detailed information available about its previous performance and reliability for PAM purposes. If this is not the case, this information needs to be gathered and provided in publicly available documentation as part of the PAM Plan.

Passive acoustic monitoring data analyses for species presence should occur through either a) visual processing of data by an acoustic expert familiar with the call types of the species of interest, or b) by using comprehensively tested PAM software detector(s) for which performance has been documented, and the performance metrics are publicly available for outside evaluation and have been reviewed and deemed acceptable by a panel or group of experts. Visual review will likely be required to some degree when dealing with acoustic detection of rare species such as North Atlantic right whales and/or for ensuring data quality. Standard performance metrics require evaluation and reporting, such as precision, recall, and accuracy, as well as false detection, false positive, false omission and missed detection rates (Figure 6) (e.g., Baumgartner et al., 2019; Kirsebom et al., 2020; Madhusudhana et al., 2020; Gervaise et al., 2021). PAM software detectors comprise a wide range of custom-built computer programs, aimed at automating the process of detecting target species’ calls in a dataset [see review in Bittle and Duncan (2013), Shiu et al. (2020), Gervaise et al. (2021)]. In both PAM archival and real-time data analysis, in addition to any software detector(s) used, some level of visual confirmation by an acoustic expert often still remains essential to improve accuracy and minimize error in call type reporting.

For PAM archival data analysis, where the objectives tend to focus on retrospective understanding of species presence, movements, or behavior, daily or hourly reporting of species detections are the time frames most frequently used (e.g., Davis et al., 2017, 2020; Stanistreet et al., 2017; Halliday et al., 2019). This method can speed up the process of data analysis by only requiring a positive species confirmation at the hourly or daily level. For example, when evaluating the presence of North Atlantic right whales along the United States East Coast, the presence of three upcalls within a 24-hour time period serves as the determination that at least one whale is present during that day (Davis et al., 2017). Three upcalls, rather than a single upcall, are used in order to decrease the likelihood of incorrect species determination given that other baleen whales can produce similar calls to North Atlantic right whale upcalls (Davis et al., 2017). However, each species and call type will require different levels of additional verification as needed (if any) and decisions made as to what level of certainty is acceptable. For example, automated detectors for 20 Hz fin whales have a high detection accuracy (e.g., Morano et al., 2012b), and, in this case, it may be reasonable to simply take the detector output with no further evaluation.

For PAM real-time data analysis, additional visual verification of the detected sound is likely to be needed since the occurrence of a given call type may influence whether operations are able to continue or are required to shut down (in the situations where shut down is possible). The ability to satisfy this confidence metric can be achieved by carrying out PAM training for operators and through evaluation of detections by analysis experts. A PAM expert is defined as a scientist who has 6 months or more of experience working with the call types of the species of interest, who can distinguish between confounding sounds (Kowarski et al., 2020), and who has experience working with the relevant detection software. Additionally, established and publicly available protocols on how a species is determined present for the specific PAM system and software must be documented in the PAM Plan.; for an example, see DeAngelis et al. (2016).

All PAM Plans should provide clear documentation of the efficacy of their detection capabilities and classification software for the specific signals of interest. Examples of comprehensive testing for real-time and archival PAM can be found in Baumgartner et al. (2019, 2020), Kowarski et al. (2020), and for PAM towed arrays can be found in Gillespie et al. (2013).

The PAM Plan should demonstrate for all PAM systems and moorings that (a) the species’ signal of interest can be heard reliably beyond the self-noise, and (b) any detection and classification software that is used has (1) been tested; (2) clearly documented reliability in detecting a given species; and (3) software performance metrics that are openly available.

Localizing calling species during the construction phase of offshore wind projects would be very useful for satisfying mitigation requirements regarding the location and distance of the species in question, relative to the sound source (e.g., pile driving location). Localization can be carried out by the placement of multiple fixed or mobile omnidirectional hydrophones arranged in a configuration that allows for localization of the vocalizing animal using the difference in the time of arrival of a call (or calls) on multiple time-synchronized sound recorders (e.g., Stanistreet et al., 2013; Hastie et al., 2014; Risch et al., 2014; Gillespie et al., 2020; Gervaise et al., 2021). It can also be achieved by using multiple sensors that can calculate bearing [e.g., directional autonomous seafloor acoustic recorders (DASARs); Greene et al., 2004; Blackwell et al., 2007; Mathias et al., 2012; Blackwell et al., 2013]. For stationary systems, a minimum of three hydrophones placed within a range that guarantees overlapping receptivity (i.e., multiple arrivals) is necessary to localize the positions of vocalizing animals using time-of-arrival methods (e.g., Stanistreet et al., 2013; Tremblay et al., 2019). Sensor positional, timing, and speed accuracy all need to be considered, as well as sensor configuration (geometric dilution of precision). However, for mobile platforms, such as a ship with a linear towed hydrophone array, two or more hydrophones can be used for the calculation of bearings; sequential bearing calculated as the platform moves can be used to estimate the location of calling animals (i.e., time-motion analysis, typically with left-right ambiguity) (e.g., von Benda-Beckmann et al., 2010; von Benda-Beckmann et al., 2013). In most cases, decisions on mitigation measures (e.g., pile driving shutdown when feasible or vessel speed reduction) can be made simply based on the range of calling animals from the noise source, without resolving the left-right ambiguity issues or knowing the bearing of the calling animal. However, under certain situations with anisotropic noise propagations, it would be necessary to know the location of the calling animal. Under such situations, localization and methodology should be included in the PAM Plan to approximate locations of the animal or sound source for purposes of taking action. Bottom-mounted recorders, real-time systems, towed and vertical arrays, and drifters can all be used for localization purposes, depending on the accuracy needed and the species of interest. Determining which system and technology to use requires careful consideration and supporting evidence to demonstrate that the design is appropriate. The analytical component of localization can be highly time consuming, which can be costly, and each array design requires careful documentation of the localization errors of the system.

The measurement of background sound levels—i.e., ambient noise metrics—is an additional and important dataset that can be obtained from acoustic recordings made on any platform. Although the primary focus of this effort is to document the vocalizations of marine mammals, additional acoustic analyses of abiotic acoustic sources in the same recordings can reveal temporal patterns in distinct frequency bands that correlate with other factors such as wind speed. Ambient noise metrics could also document the level of potential increases in ambient noise due to wind farms. Metrics for the coincidental recording of ambient noise should be included in PAM Plans and include factors anticipated to be associated with offshore wind development such as vessel traffic and operational noise. These kinds of ambient noise metrics provide a record of acoustic conditions in a given environment and are essential for understanding changes in the sound levels across different regions and time (Dekeling et al., 2014). Currently, available standards and those in development can be found by searching the Ocean Best Practices Repository (see foot note 1). Measurements of ambient noise metrics can be carried out using a number of open source programs such as PAMGUIDE (Merchant et al., 2015⁷) or MANTA⁸. These programs provide a standard series of measurements at the decidecadal level that can be replicated across projects. An ongoing framework inventory on existing standards for observations of sound in the ocean can be found in the Ocean Best Practices Repository (International Quiet Ocean Experiment WG on Standardization, 2018).

Passive acoustic monitoring archiving is essential in order to provide a long-lasting record of the efforts invested in PAM data collection. PAM archival and real-time datasets require the archiving of several items:

Archiving of acoustic sound recordings is encouraged through NOAA’s National Center for Environmental Information (NCEI) archiving service⁹. PAM metadata are required criteria for archiving at NCEI. The process and metadata details can be provided upon request. These should be used as a guide in PAM Plans for documenting relevant information regarding the field recording effort (deployment and retrieval information), as well as resulting analyses (such as species detections or noise metrics).

All confirmed passive acoustic detections of target species/species, whether from archival or real-time data, must be archived in a publicly accessible location. For the United States East Coast, all species detection data and ambient noise metrics should be reported to the Northeast Passive Acoustic Reporting System via nmfs.pacmdata@noaa.gov. Formatted spreadsheets that follow ISO standards with required detection, measurement, and metadata information are available for submission purposes (see Supplemental Information III for details). When real-time PAM is used during construction for mitigation purposes, a subset of the information required on species detections is expected to be provided and uploaded no later than 24 after the detection. Full acoustic detection data, metadata, and GPS data records must be submitted within 48 h via the formatted spreadsheets. When PAM is used for long-term monitoring, all data (detection data, metadata, GPS data, and ambient noise data) should be provided via the formatted spreadsheets and uploaded within 90 days of the retrieval of the recorder or data collection. The spreadsheets can be downloaded from https://www.fisheries.noaa.gov/resource/document/passive-acoustic-reporting-system-templates. For further assistance, contact nmfs.pacmdata@noaa.gov.

All PAM detections and metadata submitted to the Northeast Passive Acoustic Reporting System are visualized on the Passive Acoustic Cetacean Map¹⁰. We encourage PAM detections to be shared across widely used and recognized platforms and regional web portals; for the United States East Coast, some of these standardized efforts are the Passive Acoustic Cetacean Map (see foot note 10), WhaleMap¹¹, and WhaleAlert¹².