Upcalling behaviour and patterns in North Atlantic right whales, implications for monitoring protocols during wind energy development

Abstract

Offshore wind energy is rapidly developing in US waters, with construction underway off Southern New England (SNE), an important region for many species, including the critically endangered North Atlantic right whale (NARW). A data-driven understanding of NARW upcalling behaviour is presented here to help establish proper monitoring protocols for mitigating impacts. Analyses of individual upcalls from 2 years of acoustic recordings showed that NARWs were detected at least 1 day every week throughout both years, with highest NARW presence from October to April. Weeks with more days of acoustic presence typically had more hours with calling activity, but the number of upcalls within a day or hour was variable, reflective of the social function of the upcall. Within SNE, on average, 95% of the time NARWs persisted for 10 days, and reoccurred again within 11 days. An evaluation of the time period over which it is most effective to monitor prior to commencing pile driving activities showed that with 1 h of pre-construction monitoring there was only 4% likelihood of hearing a NARW, compared to 74% at 18 h. Therefore, monitoring for at least 24 h prior to activity will increase the likelihood of detecting an up-calling NARW.

Introduction

The increasing demand for alternative energy sources has turned to solar and wind energy solutions, with plans for offshore wind energy development expanding rapidly in the United States (US; Office of the Press Secretary, 2021; Office of the Press Secretary, 2022). There are currently over 27 offshore wind energy lease areas (WEAs) planned in the Atlantic Ocean along the US east coast, from the Gulf of Maine to waters off South Carolina, with additional lease areas opening up regularly (https://www.boem.gov/renewable-energy/renewable-energy-program-overview; https://www.northeastoceandata.org/data-explorer/?energy-infrastructure|planning-areas).

Offshore WEAs have been operational in other countries, particularly Europe, for decades producing a wide range of studies that assess the effects on marine life, including underwater noise, species displacement, and cumulative stressors (e.g. Bailey et al., 2010; Lindeboom et al., 2011; van Deurs et al., 2012; Popper et al., 2022). Although these studies provide a great background for the extensive WEA development facing the US waters, the western North Atlantic is home to a greater number of marine species, making the problem of potential impacts much more complex. In particular, four of the six baleen whale species inhabiting US east coast waters, including the NARW, in addition to sperm whales (Physeter macrocephalus), are considered endangered under the Endangered Species Act (est. 1973) and Marine Mammal Protection Act (est. 1972). One of the first potential WEA threats facing marine mammals is from pile driving during the construction phase: a process that emits intense, impulsive noise that radiates into the surrounding environment as turbines are hammered into the sea floor (Amaral et al., 2020). Various effects on marine life have been observed from pile driving, ranging from avoidance to behavioural changes in harbour porpoises (Brandt et al., 2011) to displacement and physical injuries in fish (Popper et al., 2006). It is not yet known how pile driving will affect each species that has not yet been exposed to it.

The first major offshore WEA development began in the spring of 2023 in the Southern New England (SNE) area, which consists of nine separate WEAs that cover both state and federal waters (https://www.boem.gov/renewable-energy/state-activities). This area spans waters to the west of Nantucket Shoals, an important region for marine species, particularly for endangered large whale species such as the North Atlantic right whale (NARW), Eubalaena glacialis, who utilize the area for feeding, socializing, and as part of the migratory route (Leiter et al., 2017; Stone et al., 2017; O'Brien et al., 2022). NARWs commonly occur in on-shelf waters in the western North Atlantic, ranging from winter calving grounds off the southeast US coast, to northern summer feeding grounds up through the Gulf of St. Lawrence, Canada. While some individuals undergo clear seasonal migrations, NARWs occur along the entire US east coast during winter months, and are found in some regions throughout the year (Davis et al., 2017). Since 2010, shifts in habitat use have been seen, including increased use of SNE, Cape Cod Bay, and Gulf of St. Lawrence, and decreased use of Gulf of Maine and Scotian Shelf waters for summer feeding grounds (Meyer‐Gutbrod et al., 2021, 2022). The remaining population of <340 individual NARWs face lethal threats of entanglement in fishing gear and ship strike: it is unclear what effects offshore WEA may introduce to the species, but limiting cumulative stressors is essential if this species is to have a chance at survival (Pirotta et al., 2022, 2023; Moore, 2023; Pettis et al., 2023). With such a small population, and where breeding females make up less than half the population, any individuals impacted by increased vessel traffic, noise from construction and operation, or displacement of prey species, could have detrimental effects to NARW recovery (Pace et al., 2017; Christiansen et al., 2020; Reed et al., 2022).

State and federal government agencies, non-government organizations (NGOs), and scientific groups have all put forward a range of requirements and recommendations aimed at minimizing the effects of wind energy development (e.g. Bailey et al., 2014; Kraus et al., 2019; Van Parijs et al., 2021). Amongst others, these include: limiting certain construction activities to times when NARWs are not expected in high densities; having observers survey to ensure animals are not present in the project area prior to and during construction; and long-term monitoring to assess potential effects of offshore wind development (e.g. https://rwsc.org/science-plan/; https://www.nrdc.org/bio/francine-kershaw/monitoring-priorities-offshore-wind; https://www.boem.gov/rodeo). Historically, visual monitoring has been the primary tool used to monitor prior to, during, and after construction activities. As a result, clear standards and training programmes have been developed for visual monitoring (Baker et al., 2013).

Over the last two decades, passive acoustic monitoring (PAM) has emerged as a widespread technology effective for monitoring species distribution and occurrence, as well as behaviour, and for mitigating anthropogenic impacts (Rountree et al., 2006; Van Parijs et al., 2009; Gibb et al., 2019; Davis et al., 2020; Van Parijs et al., 2021). Archival bottom-mounted PAM has been used extensively to understand NARW distribution, occupancy, and effects of anthropogenic noise on their communication space (Parks et al., 2007, 2010; Hatch et al., 2012; Cholewiak et al., 2018) over both large spatial scales (Davis et al., 2017; Simard et al., 2019; Durette-Morin et al., 2022; Kowarski et al., 2022) and within smaller regions (e.g. Muirhead et al., 2018; Charif et al., 2019; Estabrook et al., 2022; Murray et al., 2022). These PAM systems are stationary (remain anchored to the sea floor), and require the retrieval of the instrument to access the recordings. In addition, real-time PAM technologies have also come to the forefront as both a monitoring and mitigation tool allowing for vocalizing species to be acoustically detected in near real time, providing close to instant information on species presence in an area when calling (Bröker et al., 2015; Baumgartner et al., 2019, 2020; Aerts et al., 2022). While PAM is highly effective at documenting calling animals, silent animals, or animals outside of the detection range of the recorders, will be missed.

Unlike visual observer protocols, PAM for certain mitigation purposes is still relatively new, although PAM has been used for some time for seismic survey mitigation (Nowacek et al., 2013). The manner in which different species use sounds for foraging, navigation, social behaviour, and reproduction varies and is often species-specific. For example, odontocetes echolocate frequently and therefore have a higher probability of being acoustically heard compared to baleen whales who are variable callers with temporally clustered vocal activity (Thomisch et al., 2015). As a result, not only are clear standards, protocols, and training of observers still being developed for PAM, they also need to be tailored to each of the primary species of concern. Baleen whale species acoustic behaviour varies from species to species, where identifying a target call (the call used to determine species presence), in addition to observing the context of the call, is just as important (Clark and Gagnon, 2022; Franklin et al., 2022). Certain call types, like the NARW upcall, can be identical to vocalizations made by humpback whales (Megaptera novaeangliae), and thus require experience and training to be able to properly distinguish between species to correctly and confidently identify which species is present (Mussoline et al., 2012; Simard et al., 2019). NARWs produce a variety of vocalizations; however, the upcall, spanning on average 100–400 Hz and lasting 1–2 s, functions as a contact call and is produced by all ages, sexes, and throughout their range (Figure 1; Clark et al., 2010; Parks et al., 2011). The upcall is used extensively in PAM studies as an indicator of NARW acoustic presence (Matthews and Parks, 2021). However, information is still needed in order to improve understanding of NARW upcall rates and behaviour, as calling rates are highly variable and may vary by age, sex, behaviour, region, and season (Parks et al., 2011, 2014; McCordic et al., 2016; Durette‐Morin et al., 2019; Franklin et al., 2022; Parks, 2022).

In order to fulfil US government permitting requirements, wind energy companies are required to submit for approval marine mammal monitoring and mitigation plans, which may include the use of PAM and/or visual monitoring methods for protected species before, during, and after certain construction activities (e.g. NMFS, 2021). If a sighting or detection of a protected species occurs within a specified WEA-related zone, mitigation actions may need to be taken. The “PAM Plan” portion of the monitoring and mitigation plans cover all of the details to meet the acoustic monitoring and mitigation requirements set forth in permits (Van Parijs et al., 2021). In the SNE area, construction and pile driving activities are currently not allowed during the time-of-year restriction from January to April (e.g. NMFS, 2022), based on historical data demonstrating high densities of NARWs in and around the area (e.g. Estabrook et al., 2022; O'Brien et al., 2022).

To ensure mitigation measures based on information from PAM provide adequate protections for NARWs in the SNE region, more detailed information—specific to upcall calling (hereafter referred to as “upcalling”) behaviour in this area—in recent years—is necessary. In particular, information is needed on the upcalling rates, upcalling bout duration, and persistence in the area to inform the length of monitoring via PAM prior to construction to ensure monitoring protocols are designed to effectively detect upcalling NARWs and thus provide guidance for mitigation actions. In this study, we address a number of these needs, specific to upcalls, and provide data-driven results for managers to use for further decision making.

Methods

Bottom-mounted acoustic recorders (SoundTrap500s and SoundTrap600s; Ocean Instruments Inc.) were deployed across six sites in the Southern New England (SNE), USA region, located in the western North Atlantic Ocean, spanning Cox’s Ledge (COX) to Nantucket Shoals (NS) ( Figure 2). Recording sites occurred within, or nearby, the nine SNE wind energy lease areas (WEA), with NS sites purposefully covering a known North Atlantic right whale (NARW) hotspot area and COX01 representing the western extent of the WEAs. NS sites ranged 22–26 km apart, with COX01 spaced 65 km from NS01. Sites varied by deployment date and duration as additional sites were added to the recording region (Table 1). Three sites (COX01, NS01, and NS02) were deployed from February 2021 to November 2022, and three sites (NS03, NS04, and NS05) were deployed from February 2022 to November 2022. A total of 617 days of acoustic data were collected across the six sites. Each SoundTrap, moored 1–3 m above the ocean floor, recorded continuously with a sample rate of 48 (for SoundTraps deployed before July 2021) or 64 kHz (for SoundTraps deployed after July 2021) (Table 1).

Data were processed and analysed for NARW upcalls using similar methods as described in Davis et al. (2017). All acoustic data were low-pass filtered and decimated to 2 kHz, then processed using the Low Frequency Detection Classification System (LFDCS; Baumgartner and Mussoline, 2011). The LFDCS programme creates conditioned spectrograms using a short-time Fourier transform, with a data frame of 512 samples and 75% overlap, resulting in a time step of 64 ms and a frequency resolution of 3.9 Hz. It then traces “pitch tracks” of tonal sounds above a relative amplitude threshold, and uses a multivariate discriminant function to classify these pitch tracks according to a user-defined call library (see Davis et al. 2020; Supplementary Table S1 for specifics on the call library used). Each pitch track, or automatic detection, is then assigned a Mahalanobis distance (MD) to quantify the deviation between a given pitch track and its assigned call type, with a lower MD indicating a better match. For most species, setting an MD threshold of 3.0 or less minimizes the probabilities of both false detections and missed detections (Baumgartner et al., 2013).

Following Davis et al. (2017), all automatic NARW upcall detections with a maximum MD of 3.0 were reviewed by trained acousticians (researchers with 6 months or more of NARW-specific acoustic analysis experience) to assess daily acoustic presence. A day (24-h period, standardized to UTC-5) was considered detected for NARWs if it had at least three pitch-tracked and manually confirmed automatic upcall detections at that site. This criterion was used to be consistent with daily presence protocols from previous studies (Davis et al., 2017) and in order to be conservative and confident in stating NARW presence, while accounting for the possibility of a few incorrectly classified calls. Days that could not meet this criteria, but contained one or two confirmed NARW upcall detections, were deemed “possibly detected” for additional analysis beyond the daily presence review. Davis et al. (2017) found the LFDCS had a missed detection rate of 33%; this study did not further evaluate the missed detection rate specific to this dataset; therefore, the resulting analysis covers minimum NARW upcall presence and likely excludes some days that had NARW upcalls missed by the LFDCS.

All days that were marked as detected (contained three or more confirmed automatic upcall detections), or possibly detected (contained one or two confirmed automatic upcall detections) were manually annotated and every NARW upcall was logged for each of those days, regardless of whether it was automatically detected, using a customized Interactive Data Language (IDL) script within the LFDCS programme. Only days with one or more confirmed automatic upcall detections were then manually annotated. The number of hours with detections and the number of upcalls per hour were then averaged across each deployment to assess seasonal and diel trends. Here, we define diel as trends over a 24-h period, regardless of day or night.

For the above analyses all available data were used, including sites that did not have a full year of recordings. To avoid any potential seasonal bias, only sites that had at least one full year of data (COX01, NS01, and NS02; hereafter referred to as “full year sites”), were used for the following analyses, which averaged values across a calendar year.

To understand how far apart upcalls are spaced, and for how long upcalling activity occurs once an upcall is heard, inter-call intervals (ICIs) were measured between every logged upcall, calculated as the difference between the start time of one upcall and the start time of the next. No distinction was made between whether the upcalls measured came from one or more individuals. To define an upcall bout, we took the weighted mean of the 95th percentile of all ICIs for full-year sites (COX01, NS01, and NS02), using R Statistical Programming Language (R Core Team, 2022). Any upcalls with ICIs less than the weighted mean were considered to occur within a single upcalling bout, and any upcalls with ICIs greater than this amount were considered the start of a separate event and new upcalling bout. Bout lengths and number of upcalls within a bout were then measured across each of the three full-year sites (COX01, NS01, and NS02).

Daily presence results were used to calculate the number of days with consecutive acoustic presence (hereby defined as NARW acoustic persistence) or silence, where silence refers to days with no confirmed automatic detections from LFDCS. NARW reoccurrence, or the number of days following a confirmed automatic upcall detection in which NARWs are most likely to be detected again, was calculated from days with one or more automatic NARW upcall detections. Using full-year sites (COX01, NS01, and NS02), the probability of detecting a NARW again for each day following a confirmed automatic upcall detection was estimated over a 4-week period. The weighted mean of the 95th percentile across each of the three sites was used to determine the average number of days over which NARWs are likely to continue to be detected or remain acoustically detected, following an initial detection: hereby defined as NARW acoustic reoccurrence.

Lastly, we evaluated the likelihood of detecting NARW upcalls when they are acoustically present, prior to the start of anthropogenic activity (i.e. pile driving; wind farm construction). The manually logged NARW upcalls were summarized into hourly presence over the full-year sites (COX01, NS01, and NS02). For days with one or more automatic NARW upcall detections, 12 theoretical “start times” (each daylight hour when pile driving is currently permitted, from 07:00 to 18:00 UTC-5) were used to back-calculate and determine if NARW upcalls were detected 1–24 h before the start time of construction. For each possible start time (07:00, 08:00, 09:00…to 18:00 UTC-5), every hour from 1 to up to 24 cumulative hours prior to the start time was examined to determine if manually logged NARW upcalls were acoustically detected in any sum number of those hours. A cumulative proportion of NARW detection over time, given a known presence within that 24 h, was determined by dividing the cumulative hours with presence by the total number of evaluated NARW hours, for each start time and site. Days without a full 24 h prior to any start time were excluded from this portion of the analysis.

Results

Seasonal occurrence

A total of 2337 recording days were analysed for NARW upcall presence across the six sites using the LFDCS. Despite varying recording effort (ranging from 112 recording days at NS03 to 619 recording days at NS02), all sites had automatic NARW upcall detections. A total of 29% (667 days) of the total recording days had at least one confirmed automatic NARW upcall detection; a total of 22% (513 days) of the total recording days were classified as detected for NARW acoustic presence, having three or more true upcall detections in a 24-h period; and a total of 7% (154 days) were classified as possibly detected, having only 1–2 confirmed upcall detections in a day. NS03 had the highest percentage of days with definite NARW detections (45%), despite having the lowest number of days (112 days) for recording effort (Table 2). COX02 and NS05 had varying recording effort (604 recording days and 277 recording days, respectively), but both had the lowest percentage of days (12%) with NARWs detected.

NARWs were acoustically present for at least 1 day every month on all six sites when data were available, except for September at NS01, where no days were considered detected (Figure 3). NARW acoustic occurrence was highest from November to April. While detections were lowest, but not absent, from May to October, NARWs were consistently detected at all sites, when data were available. During May–October, acoustic occurrence peaked in August–October at NS02, and from July to October at NS04. From June to November, NARWs were detected consistently at the easternmost sites on or near Nantucket Shoals (NS02, NS04, and NS05).

Days that had at least one confirmed automatic NARW upcall detection were further reviewed to analyse and count every upcall. NARWs upcalled throughout the day and night at varying degrees based on season and site location (Figure 4). In general, where NARWs were acoustically present for consecutive days, upcalling activity persisted throughout the day at each site. During periods where they occurred less frequently throughout the week, daily activity varied: days ranged from a few to many upcalls detected throughout the day (Figure 4).

Hourly presence

The average number of hours per day with detected upcalls, or vocal hours, showed similar trends to the average number of days per week with NARW acoustic occurrence, with NARWs typically detected for six or more hours on days they were acoustically present (Figures 3 and 5). In general, months and sites with more days per week with NARW acoustic presence had more hours with upcall detections per day. For all sites, October–April had days with a higher percentage of hours with detections, compared to other months. Days with 6 h or less of upcalls (25% or less of a day with detections) typically occurred between April and October; however, some sites had more hours with upcall detections each day (i.e. NS04 from July to October), and some days had fewer hours with upcall detections, despite high daily presence periods (i.e. NS05 in March).

The total number of upcalls per day was averaged in weekly bins for each site (Figure 6). Across all sites, where data were available, January–April had the highest average number of upcalls per day [min = 1, max = 1549, mean = 176, and standard deviation (SD) = 238 upcalls]. There was little variation in the number of upcalls per day from May to December, despite the peaks in NARW daily acoustic presence (Figure 3) and number of hours with upcalls (Figure 5), suggesting that when NARWs are upcalling, there could be the same number of upcalls produced in 1 h, or spread over 24 h. At NS02, there was an additional peak in the number of upcalls per day from November to December (min = 4, max = 1046, mean = 340, and SD = 255 upcalls), compared to the other sites (for COX01: min = 1, max = 39, mean = 9.1, and SD = 10.6; for NS01: min = 3, max = 173, mean = 67.3, and SD = 74.4), reflecting an increase both in the number of upcalls and the number of hours NARWs were detected during those months. Overall, upcalling activity and the number of upcalls per day throughout the dataset varied and were often low, even in periods of high daily and hourly presence.

The average number of upcalls per hour across each full-year site was averaged to examine diel (24-h) trends (Figure 7). In general, there is minimal variation in the number of upcalls detected across all hours of a day, especially at COX01. A peak in upcall detections were seen at 18:00 UTC-5 at NS01, and at 16:00 and 17:00 UTC-5 at NS02.

ICI and bout length

A total of 79032 ICIs were measured between all upcalls for full-year sites (COX01, NS01, and NS02). The weighted average of 95% of all evaluated upcalls across all three sites was 17.5 min between upcalls (Table 3). NS01 had the lowest overall time between upcalls: 95% of all upcalls at this site occurred within 13.9 min of another upcall. COX01 had the longest ICI with 33.6 min occurring between 95% of the upcalls. The average 95% of all ICIs across sites, 17.5 min, was used to define NARW upcall bouts: any upcalls that had an ICI >17.5 min was determined to be a new upcalling bout. There were 808 NARW upcall bouts observed at COX01, 1239 bouts at NS01, and 1800 bouts at NS02. On average, upcall bouts lasted 20.7 min at COX01, 33.3 min at NS01, and 33.5 min at NS02. At each of the three sites, 95% of upcall bouts lasted <87.2, 145, and 149 min, with maximum bout lengths of 581, 1300, and 954 min, respectively. Using a weighted average, 95% of upcall bouts at all three sites were equal to or shorter than 135 min in duration.

Persistence/Reoccurrence

For each full-year site (COX01, NS01, and NS02), days with at least one confirmed NARW upcall detection were evaluated to measure the number of consecutive days of NARW acoustic presence, or “acoustic persistence”, at that site (Table 4, Figure 8). On average, NARW upcalls were detected 3 days in a row across seasons and sites in SNE. A total of 95% of the time, NARW upcalls were detected from 7 (at COX01) to 12 days (at NS02) in a row or less, with a weighted average of 10 days of consecutive upcall persistence across the three sites. Consecutive days with no automatic NARW upcall detections varied more considerably between the three sites (an average of 13 days for COX01, 9 days for NS01, and 5 days for NS02), with 95% of consecutive days with no upcall detections lasting 31 days or fewer.

The proportion of days detecting a NARW upcall again after a first day with confirmed automatic detections (reoccurrence) was analysed for each full-year site. Reoccurrence varied across sites, where 95% of the time a NARW upcall was detected again within 9 (at NS02) to 31 (at COX01) days after the first day of detection (Table 4, Figure 9). Across the three sites, the weighted average for 95% of days with NARW acoustic presence occurred within 11 days of another confirmed acoustic presence day. Therefore, most of the time in the SNE region, when NARWs are detected, there will likely be another upcall detection in the area within 11 days.

Pre-construction-activity monitoring

To assess the likelihood of detecting an acoustically present NARW prior to an initial start time, 144 (at COX01) to 279 (at NS02) days were evaluated at each full-year site (Figure 10). All of the sites and start times showed the same increasing trend with slight variation. Here, 6-h intervals—with a minimum of 1 h and a maximum of 24 h—were used as realistic time requirements to monitor for NARWs prior to construction. Overall, the more time monitored (in this case, up to 24 h) prior to a start time, the more often a NARW was detected, as there were more cumulative hours with an upcall detection (Figure 11). When averaged across the three sites and modelled start times, an upcalling NARW would only be detected in 74.4% of the hours if monitoring 18 h prior to activity. The proportion of hours with detections for an acoustically present NARW dropped dramatically for less monitoring time: 47.3% for 12 h of monitoring, 22.9% for 6 h, and 3.9% for 1 h. For all full-year sites, we looked at time periods during the time-of-year restriction (when pile driving is not permitted in the region, January–April) and allowed construction periods (May–December) separately, but differences were negligible compared to results across the year as a whole.

Discussion

This study provides a detailed overview of NARW upcall acoustic occurrence in the SNE region, using data collected >2 years prior to the start of WEA construction. With NARWs acoustically present every month across all recording sites, the SNE continues to be an important area, as documented in visual and acoustic surveys since 2011 (e.g. Leiter et al., 2017; Quintana-Rizzo et al., 2021; Estabrook et al., 2022; O'Brien et al., 2022). The National Oceanic and Atmospheric Administration (NOAA) and Bureau of Ocean Energy Management (BOEM) regulations currently restrict pile driving from occurring during the months of January–April, when NARWs occur in high numbers in the area; however, NARWs were detected in equally high amounts throughout the area during the months of October–December, months during which further protections should be considered. NS02, NS03, and NS04 were the sites with the highest number of NARW upcalls, showing high densities both on Nantucket Shoals and along the edge of the Shoals (Figure 2). NARW acoustic presence peaked within the time-of-year restriction (during February–April) at NS02, NS03, and NS04, where there was recording effort. At NS04, the recording site located on Nantucket Shoals, acoustic presence also peaked in July and September. This data reinforces previous visual and acoustic data, which showed high variability across years with occasional presence in July and August (Quintana-Rizzo et al., 2021; Estabrook et al., 2022).

For some species, such as in odontocetes, it is possible to estimate the number of animals present, given the number of acoustic cues recorded (e.g. Buckland et al., 2001; Marques et al., 2019). In contrast, NARWs are social communicators and tend to show high variability in their calling rates regardless of group size (e.g. Clark, 1990; Parks et al., 2011; Root-Gutteridge et al., 2018). Variability of calling rates has been seen across age groups and in other regions (Parks et al., 2011, 2019b; Durette‐Morin et al., 2019) as well as changes in vocalizations in mom and calf pairs (Parks et al., 2019a), making it challenging to estimate the number of animals present from acoustic detections alone. Our results reflected this variability as weeks with a high daily presence could have the same number of upcalls as weeks with fewer presence days. Likewise, some days with acoustic presence had the same number of upcalls over 24 h as were detected within 1 h elsewhere. One animal could be present and consistently calling, or many animals could be passing through, calling infrequently, if at all. With this in mind, NARW upcalls are useful for detecting their presence given that one individual alone can trigger a management action; however, it is not as yet possible to extrapolate to abundance for this species using PAM.

Previous studies have evaluated upcall occurrence for seasonal and diel trends in the Gulf of Maine (Bort et al., 2015), Massachusetts Bay (Clark et al., 2010; Morano et al., 2012; Mussoline et al., 2012), New York Bight (NYB; Murray et al., 2022), and the SNE region (Estabrook et al., 2022), revealing an increase in upcalling activity in evening hours (16:00–20:00 UTC-5) in some regions (NYB, SNE), and in both twilight periods (dawn and dusk) further north in Massachusetts Bay and Gulf of Maine. When comparing NARW acoustic activity over time, Charif et al. (2019) found changes in when and how often NARWs are detected in Massachusetts Bay, showing an earlier, and greater peak of acoustic occurrence (more hours of the day with detections) after 2010 compared to before 2010. This study found minimal diel patterns across the dataset, suggesting that NARWs may have changed their behaviour in the region. Sites that showed the highest diel trend for evening hours, NS01 and NS02, were in similar locations as Estabrook et al. (2022), and had year-round recordings. No discernible diel patterns were found on COX01, which could indicate that NARWs exhibit different diel patterns in locations further northwest. Overall, these results demonstrate that monitoring only during certain selective hours of the day does not guarantee that NARWs are detected and longer monitoring time periods are needed to ensure that their presence is captured.

NARW acoustic persistence and reoccurrence are important contexts to evaluate since they can determine both how long NARWs may stay in a given area as well as how long you need to listen for NARWs to ensure that you will detect an acoustically active animal(s) in an area. This study showed that when NARWs were detected in the SNE, they were likely to persist, or reoccur, within the area for an average of 10 and 11 days, respectively. This supports current slow zone rules, where slow zones are in place or “persist” for 15 days after sighting or detecting a NARW (https://www.fisheries.noaa.gov/feature-story/help-endangered-whales-slow-down-slow-zones). An extension of 15 days is issued whenever a NARW is heard or sighted within the last 7 days of the 15-day period. Given the range of persistence (8–12 days) and reoccurrence (9–31 days) found in SNE, applying these methods to data in other regions would provide a better understanding of persistence in other habitats, which will help to inform the proposed changes to the NARW vessel speed rule to further reduce the likelihood of mortalities and serious injuries to endangered NARWs from vessel collision, currently under consideration (https://www.fisheries.noaa.gov/national/endangered-species-conservation/reducing-vessel-strikes-north-atlantic-right-whales).

Within management discussions, there is considerable emphasis on attempting to localize calling animals to determine if they are within or outside of a given WEA. However, the reoccurrence results demonstrate that whales either remain in an area, or move to and fro between a given area so that once detected, NARWs are likely to be detected again within 11 days or fewer at that same site within the SNE region. If whales are heard by a PAM system monitoring in a WEA, even if they are out of the construction zone at the time of detection, it is not unlikely that they will repeatedly move into, near, or through the area. Johnson et al. (2020) showed that from simulated results, in <24 h after an initial detection, there was a <10% probability of a whale remaining within 5 km of reported position, but a >71% chance that the whale remained within 25 km. The SNE is a known feeding and socializing (Quintana-Rizzo et al., 2021) ground for NARWs, and animals will persist or call differently here than in other areas. Using a similar analysis approach in other areas, where different behaviours may occur, is necessary before applying these conclusions across other WEA regions.

When determining the length of time for which monitoring should occur in order to have a higher likelihood of detecting an acoustically active NARW(s) in an area, for the SNE, it was clear that, for example, monitoring for 24 h would give a 100% likelihood of detecting an upcalling NARW; 18 h would give a 74% likelihood of the species being heard, as opposed to monitoring for 1 h would provide only a 4% likelihood of hearing NARWs. Current monitoring requirements prior to pile driving require 1 h of monitoring prior to pile driving operations commencing (e.g. NMFS, 2022). A longer monitoring period is clearly needed to provide “situational awareness” in order to support a more robust mitigation approach. The data for the SNE WEA shows that the longer you listen, the higher the likelihood of detecting an acoustically active NARW. An additional element to consider is the variable and more sparse calling behaviour of NARWs during time periods where their occurrence is lower and when pile driving is permitted (May–December). During this time period, monitoring for at least 24 h prior to activity is critical to increase the likelihood of detecting an NARW if they are in the area and upcalling.

Passive acoustic monitoring is a highly effective tool for mitigation purposes; however, it is important to understand its applications, how and where it should and shouldn’t be used. Ceballos et al. (2022) showed that PAM was 20 times more reliable than a visual survey at detecting a single animal. Here, we present results on NARW upcall presence, and provide guidance on how to use PAM for monitoring and mitigation of WEA impacts on NARWs. It is important to remember that animals can be present but silent, be producing vocalizations different from the target call (in this case, the upcall), or be calling outside of the detection range of the recorders. NARWs produce a variety of vocalizations beyond the upcall, but they can be harder to identify reliably and their use across different regions has not yet been examined, thus are not used for monitoring and mitigation to the same extent as upcalls (Van Parijs et al., 2009; Franklin et al., 2022). Additionally, detection ranges of calls vary with oceanographic conditions, including bathymetry, bottom type, temperature, salinity, and ambient noise conditions. On average, throughout their range, NARWs can be detected from 10 km away (e.g. Laurinolli et al., 2003; Clark et al., 2010; Van Parijs et al., 2021). Estabrook et al. (2022) estimated detection ranges in the SNE region to be 8–22 km on bottom-mounted recorders, and Johnson et al. (2022) estimated detection ranges from 6 to 27 km on real-time platforms. In all cases, detection ranges were affected by ambient noise levels, and will be heavily impacted by the proximity and amount of pile driving, vessel traffic, and other WEA activity (Van Parijs et al., 2023). Thus, combining visual and acoustic methods when monitoring to ensure an area is “clear” during WEA construction is important, especially during times of increased ambient noise (i.e. pile driving), when NARW acoustic detection may be challenging. However, mitigation measures such as bubble curtains are known to reduce pile driving sound levels and may therefore still allow for effective acoustic detection (Dähne et al., 2017).

Therefore, understanding the target species’ calling behaviour is essential when designing effective monitoring and mitigation strategies aimed at reducing the impact of anthropogenic activities. In this study, we present data on NARW upcalling behaviour for the SNE WEA with the understanding that other WEA regions will require similar analyses and evaluations to assess how widely the trends from this region can be applied across areas or not, and if these SNE trends change in future years. Similarly, in regions where other acoustically active species are of primary concern, comparable assessments should be done to evaluate that species calling behaviour as well.

Acknowledgement

We thank Molly Martin for analysis help and Mark Baumgartner for his invaluable input for methods. Thanks to Nick Sisson, Jaclyn Daly, and Carter Esch for their regulatory expertise. Thank you to Kate Choate and Catherine Dodge for all the QA/QC work on these data. Thanks to all the crew and field teams that helped collect this data, especially Tim Rowell, Tyler Aldrich, Jessica McCordic, Amanda Holdman, Annabel Westell, and Rochelle Gordon.

Conflict of interest

The authors declare no conflicts of interest.

Data availability

All these data were collected with federal funding and are publicly available upon request. We are working to integrate the detection data into the publicly available Passive Acoustic Cetacean Map at https://apps-nefsc.fisheries.noaa.gov/pacm.

Author contributions

GED was responsible for the study design, figure creation, conceptual development, and writing of the manuscript; SCT was responsible for data analysis, figure creation, and reviewing the manuscript; and SMVP was responsible for the study design, conceptual development, and writing of the manuscript.

Funding

Funding was provided by BOEM Award #M19PG00015 and the Northeast Fisheries Science Center’s Wind Energy Program.

Animal ethics and welfare

There are no animal ethics and welfare concerns.

References