Acoustic quality of critical habitats for three threatened whale populations

Authors

  • R. Williams,

    Corresponding author
    1. Sea Mammal Research Unit, Scottish Oceans Institute, University of St Andrews, St Andrews, Fife Scotland, UK
    2. Oceans Initiative, Pearse Island, BC, Canada
    • Correspondence

      Rob Williams. Current address: Fisheries Centre, Marine Mammal Research Unit, AERL, Room 247, 2202 Main Mall, Vancouver, BC V6T 1Z4, Canada. Email: rmcw@st-andrews.ac.uk

    Search for more papers by this author
  • C. W. Clark,

    1. Bioacoustics Research Program, Cornell Lab of Ornithology, Cornell University, Ithaca, NY, USA
    Search for more papers by this author
  • D. Ponirakis,

    1. Bioacoustics Research Program, Cornell Lab of Ornithology, Cornell University, Ithaca, NY, USA
    Search for more papers by this author
  • E. Ashe

    1. Sea Mammal Research Unit, Scottish Oceans Institute, University of St Andrews, St Andrews, Fife Scotland, UK
    2. Oceans Initiative, Pearse Island, BC, Canada
    Search for more papers by this author

  • Editor: Nathalie Pettorelli
  • [The copyright line for this article was changed on 17 July 2017 after original online publication]

Abstract

Habitat loss is a leading cause of biodiversity loss in terrestrial ecosystems. For marine species that rely on acoustic cues to navigate, find food or select mates, sound is a key element of their environment. Chronic forms of human-generated ocean noise have the potential to mask communication signals over substantial fractions of their functional areas for substantial fractions of the year, which makes acoustic masking a qualitatively similar stressor to habitat loss. International policy decisions on chronic ocean noise are evolving, which creates an opportunity to advance the scientific foundation of decision-making. We measured ocean noise levels at 12 sites, chosen for current and predicted intensities of anthropogenic activities and importance to three endangered whale species in Canada's Pacific Ocean: fin, humpback and killer whales. Canada includes sound as a key element of resident killer whale critical habitat, but not for other species. In the frequency bands that killer whales use for social communication, noise levels were highest in legally designated killer whale critical habitats. In contrast, noise levels were generally lower in habitats known to be important for baleen whales, but these quieter areas are not yet given special legal protection. These noise levels translate into potentially serious fractions of lost opportunities for acoustic communication. Median noise levels are high enough to reduce the communication spaces for fin, humpback and killer whales under typical (median) conditions by 1, 52 and 62%, respectively, and 30, 94 and 97% under noisy conditions. As countries begin to articulate their policies to protect acoustic attributes of marine habitats under their jurisdiction, we recommend quantifying loss of communication space, but quantitative targets need to be set. We see two ways forward. Managers could specify limits of acceptable change in terms of population-level impacts, which can be modelled through effects from communication masking and/or disturbance on prey intake. Alternatively, managers can specify targets reflecting amount of habitat to protect for each species, adjusting upward to account for habitat effectively lost from chronic ocean noise.

Introduction

The ocean is filled with biotic and abiotic sounds that can influence survival and reproduction of organisms that perceive acoustic energy (Bass & Clark, 2002). Human activities add noise (i.e. unwanted sound) to the ocean, potentially impacting organisms that have evolved to use sound as their primary modality for sensing their environment, navigating, finding food and selecting mates (Payne & Webb, 1971; Tyack & Clark, 2000). Marine bioacoustics research has been dominated by concerns that high-amplitude, anthropogenic sounds (e.g. military sonar or seismic surveys) could injure individual animals, including fish (Popper et al., 2005) and beaked whales (Jepson et al., 2003; Fernandez et al., 2005). Recent attention includes activities with less intense source levels (SLs) that collectively raise background noise levels by at least an order of magnitude over ecologically large areas for long periods of time – so-called ‘chronic ocean noise’ conditions (Ellison et al., 2011). Commercial shipping can create chronic noise in lower frequencies (< 300 Hz), as the number of ships, vessel size and propulsion power has steadily increased in recent decades (Wenz, 1962; Urick, 1983; Andrew et al., 2002).

As bioacoustic metrics are applied to ecologically appropriate spatio-temporal scales, mitigation methods have been proposed for vulnerable species or their habitats using a variety of management tools. Animal-centric tools treat noise as a threat to individual whales (e.g. observers monitor ‘marine mammal safety zones’ around ships that deploy tactical sonar or conduct seismic surveys). Area-based tools are used to manage noise as a chronic, habitat-level stressor and invoke spatial planning approaches (Erbe, MacGillivray & Williams, 2012; Hatch et al., 2012). Put another way, some management policies aim to preserve aspects of an ocean soundscape as it relates to particular species (e.g. sound energy in frequencies used by a threatened species), whereas others aim to reduce acoustic energy in the environment as a whole, regardless of how that sound may affect species. As more countries adopt policies that use ocean noise as an environmental indicator, scientific guidance is needed to distil the complexities of ocean noise into simple statements about the relative quality of acoustic habitats for different species (Moore et al., 2012).

A species' acoustic ‘communication space’ (the predicted space over which animals can communicate) is decreased with the introduction of anthropogenic sound (Clark et al., 2009). The extent of any decrease can be calculated in the form of a ‘masking metric’ from the perspective of any species (Clark et al., 2009), and involves explicit statements about the noisy and quiet conditions one is contrasting (Mohl, 1981). Masking metrics are calculated relative to a reference condition, assumed to represent either a lack of anthropogenic noise [‘ancient ambient’ (Clark et al., 2009)] or some level of non-discrete, anthropogenic noise [‘historical ambient’ (Hatch et al., 2012)]. A masking metric can be presented in terms of range, area or volume. Here we measure communication space as a two-dimensional area (Clark et al., 2009).

Communication masking metrics can guide management decisions for acoustically sensitive species at risk. For example, the European Union's Marine Strategy Framework Directive (2008/56/EC) outlines general standards for achieving ‘Good Environmental Status’ regarding underwater sound (Borja et al., 2010). One indicator suggests that no year-on-year noise level increases should be permitted in some representative, year-round sample within third-octave bands centred on 63 and 125 Hz (re: 1 μPa, rms) (Van der Graaf et al., 2012).

Canada has made progressive and precautionary policy statements on ocean noise and endangered species conservation by including acoustic attributes in its critical habitat designation for resident killer whales Orcinus orca (Fisheries and Oceans Canada, 2011). Acoustic degradation of critical habitat is therefore recognized as a threat to killer whale recovery, and it is illegal to introduce sufficient noise in critical habitats to ‘destroy’ it. Scientific guidance is needed to advise on noise levels that would constitute habitat destruction (Federal Court, 2010). Canada is interpreting its obligation ‘in order that killer whales can maintain communication, and detect and capture prey while in the area’ (Fisheries and Oceans Canada, 2011). No quantitative targets or thresholds have been defined, but masking metrics would allow policy-makers to articulate their conservation objectives in quantitative terms. In contrast, the US, with shared jurisdiction over southern resident killer whales, does not recognize sound as a primary constituent element of critical habitat under the US Endangered Species Act (National Marine Fisheries Service, 2008). The US specifies acoustic thresholds and ‘do not exceed’ criteria under the Marine Mammal Protection Act, but we are unaware of any US policy that sets more precautionary acoustic thresholds for critically endangered species or their critical habitats than for marine mammals in general (Horowitz & Jasny, 2007). Canada's Species at Risk Act is relatively new, and recovery plans and implementation of proposed critical habitat areas are lagging for many other acoustically dependent species such as fin Balaenoptera physalus and humpback Megaptera novaeangliae whales (Fisheries and Oceans Canada, 2010), so it is unclear whether killer whales will set a precedent for including sound in critical habitat designations for other acoustically sensitive marine species.

The purpose of this study was to evaluate and compare the influence of chronic noise on communication space of three cetacean species across a broad region over multiple years. Our study aimed to provide a quantitative description of ambient noise levels, including spatial and temporal patterns, for important whale habitats in British Columbia (BC), Canada. Using noise levels at 12 sites along the BC coastline as a proxy for a communicating whale, we estimated loss of communication space for fin, humpback and killer whales. We compared habitat-level variability in lost communication area for these whale species and discuss the implications of including acoustic noise metrics as elements of critical habitat, both regionally and in other jurisdictions.

Methods

A brief discussion of our field and analytical methods follows below. For a fuller description, please see Supporting Information Appendix S1.

Acoustic data collection

Underwater recordings were collected between 2008 and 2010 at a total of 12 sites along an ∼840 km (450 nautical mile) expanse of the BC coast (Fig. 1). Locations were selected to sample the acoustic environment for frequency ranges in which endemic species are acoustically active and that were expected to sample across a wide range of ship traffic intensity levels (Williams & O'Hara, 2010). Acoustic data were collected using marine autonomous recording units (MARUs) (Clark & Clapham, 2004) anchored 1.5–2.5 m above the seafloor and programmed to record on a duty cycle (Table 1) limited by battery life or disk space. As our study progressed, we adjusted the sampling rate to capture increasingly higher frequencies, which necessitated shorter deployments or sparser duty cycles due to hard drive limitations (Table 1).

Figure 1.

Map of study region and 12 sites along the British Columbia coast at which underwater recordings were collected in 2008, 2009 and 2010. The colour tone of the coloured circles in left hand columns indicates the median ambient deployment noise level (in dB re: 1 μPa, see colour key in upper right) at each site for frequency bands in which fin whales Balaenoptera physalus (17–28 Hz, green), humpback whales Megaptera novaeangliae (71–708 Hz, red) and killer whales Orcinus orca (1.5–3.5 kHz, blue) produce communication signals (see Table 1 for deployment details). The size of a coloured circle indicates the percentage of expected communication space available to the species at that site throughout the recording period under median ambient deployment noise conditions, relative to the communication space available under median quietest deployment noise conditions (see circle size key in middle right).

Table 1. Summary information for MARU deployments
LocationRationaleLat/LonDepth (m)Deploy dateBurn/retrieve dateNumber of daysSample rate (kHz)Gain (dB)Band-pass filters (Hz)Duty cycle (%)
  1. MARU, marine autonomous recording unit.
Triple Island (Dixon Entrance)Near port of Prince Rupert; important habitat for fin whale Balaenoptera physalus, humpback whale Megaptera novaeangliae and killer whale Orcinus orca54.330, −130.89915829 June 200820 October 20081131628.5 (80.5–146.7)1022 (13 min on/47 min off)
Kitimat (Douglas Channel)Near expanding port of Kitimat; site of proposed oil tanker traffic53.884, −128.74217925 August 201026 September 2010323233.5 (75.5–141.7)10 Hz–11000 Hz Band pass Filter50 (30 min on/off)
Kitkiata Inlet (Douglas Channel)Near expanding port of Kitimat; site of proposed oil tanker traffic53.586, −129.21426424 August 201026 September 2010333233.5 (75.5–141.7)10 Hz–11000 Hz Band pass Filter50 (30 min on/off)
Caamano IslandPresumed low-traffic site; candidate critical habitat for humpback and northern resident killer whales52.910, −129.65619728 June 200825 October 20081191628.5 (80.5–146.7)1022 (13 min on/47 min off)
Broughton ArchipelagoProvincial park; important habitat for Pacific white-sided dolphins and transient killer whales; northern resident killer whales displaced by acoustic harassment devices50.766, −126.5021231 September 201021 September 2010206433.5 (75.5–141.7)No Filters33(20 min on/40 min off)
Blackfish SoundDesignated critical habitat for northern resident killer whales; less constricted shipping lane than Johnstone Strait50.648, −126.81912721 August 201022 September 2010323233.5 (75.5–141.7)10 Hz–11000 Hz Band pass Filter50 (30 min on/off)
Robson Bight (Johnstone Strait)Designated critical habitat for northern resident killer whales; very constricted shipping lane50.495, −126.58818524 May 200830 August 20081591628.5 (80.5–146.7)1022 (13 min on/47 min off)
Robson BightDesignated critical habitat for northern resident killer whales; very constricted shipping lane50.495, −126.59217123 July 200929 August 2009373233.5 (75.5–141.7)100 Hz–11000 Hz Band pass Filter36 (11 m on/19 m off)
Robson BightDesignated critical habitat for northern resident killer whales; very constricted shipping lane50.498, −126.60016621 August 201021 September 2010316433.5 (75.5–141.7)No Filters33 (20 min on/40 min off)
Georgia StraitNear Port Metro Vancouver; heavily urbanized; near southern resident killer whale critical habitat49.214, −123.40928926 May 20088 November 20081661628.5 (80.5–146.7)1022 (13 min on/47 min off)
Haro StraitDesignated critical habitat for southern resident killer whales; constricted shipping lane48.488, −123.19518027 July 200930 August 2009343233.5 (75.5–141.7)100 Hz–11000 Hz Band pass Filter36 (11 m on/19 m off)
Haro StraitDesignated critical habitat for southern resident killer whales; constricted shipping lane48.522, −123.2032267 September 201029 September 2010226433.5 (75.5–141.7)No Filters33 (20 min on/40 min off)

Ambient noise measurements and analysis

Acoustic analyses followed analytical procedures described previously (Clark et al., 2009; Hatch et al., 2012). Daily acoustic data from each MARU were processed and converted into daily spectrograms and daily noise statistics. The lack of information on hearing ability in baleen whales relative to killer whales hinders quantitative comparisons among species (Szymanski et al., 1999; Croll et al., 2001; Erbe, 2002), so we consider noise energy in the frequency bands in which these species produce sounds, without making inference about how they perceive sound. We considered five frequency bands as defined by species-specific sound categories and a proxy for the sound SL at which the sound type is produced: fin whale [17–28 Hz, SL = 180 dB re: 1 μPa @ 1 m, within range of reported values (McDonald & Fox, 1999; Charif et al., 2002; Sirovic, Hildebrand & Wiggins, 2007; Weirathmueller, Wilcock & Soule, 2013)], humpback whale song [71–708 Hz; SL = 170 dB re: 1 μPa @ 1 m (Frankel & Clark, 1998)], and killer whale calls, whistles, and echolocation clicks [1.5–3.5, 5–12 and 18–26 kHz, respectively, using SLs of 155 dB re: 1 μPa @ 1 m for calls, 145 dB for whistles (Miller, 2006) and 195 dB for echolocation clicks (Au et al., 2004)]. These frequency bands correspond roughly to 20-Hz fin whale song notes (Watkins et al., 1987), lower frequency humpback whale song notes (Au et al., 2006), and killer whale burst-pulse calls, whistles, and the low end frequency band for echolocation clicks (Ford, 1989; Au et al., 2004; Miller, 2006), respectively. These illustrative bands differ in width, so sound levels can be compared among sites within a band, but should not be compared directly across bands.

The statistical distributions of noise levels at each of the 12 MARU deployment sites (Table 1) were used to quantify loss of communication space, both at daily time scales and for entire MARU deployments, in order to compare the influence of chronic noise across multiple sites and over multiple years. We computed daily distributions of noise band levels for each MARU, producing statistics describing the daily noisiest, daily median and daily quietest band levels. We used these daily noise values to compute band level noise distributions for each MARU's entire deployment, yielding values for the median quietest, median ambient and median noisiest deployment noise levels associated with each deployment. These three median noise statistic values are used as proxies for chronic noise conditions around each MARU in order to calculate estimates of communication space for each species-specific frequency band and the per cent reduction in communication space at each site for the duration of the recording period relative to median quietest and median ambient conditions.

Testing for statistical significance of among-site differences

We tested for statistical significance of differences in noise levels among sites, within each species-specific frequency band, with linear models using generalized least squares in package nlme (Pinheiro et al., 2013) for mixed-effect model fitting and package MuMin (Barton, 2012) for model ranking and selection using Akaike's information criterion (AIC). Median noise level was modelled as a function of frequency and site, allowing a ‘lag 1’ autocorrelation structure in daily median noise levels (Pinheiro et al., 2013) to remove temporal autocorrelation pattern in the residuals. Note that all of the models we considered gave similar results.

Exploratory analyses of anthropogenic versus natural sound sources

We conducted exploratory analyses to see if between-site differences in ambient noise levels were consistent with anthropogenic or wind factors. We generated long-term spectral averages (‘LTSAs’) from two illustrative sites that experience dramatically different levels of shipping patterns (Williams & O'Hara, 2010; Erbe et al., 2012): the Kitimat area experiences relatively little large-ship traffic, whereas the Haro Strait area experiences a high level of ship traffic (Figs 2 and 3). Wind speed data were extracted for the nearest weather buoy from National Oceanic and Atmospheric Administration's ‘Historical Data Download’ site.1

Figure 2.

Four-panel plot showing noise levels as recorded during 2010 on the marine autonomous recording unit located in Haro Strait and wind speed, where the month/day notations (e.g. ‘09/08’) indicate local midnight. (a) Long-term spectrogram showing the hourly, time-varying distribution of acoustic energy as a linear function of frequency. (b) Long-term spectrogram showing the hourly, time-varying distribution of acoustic energy as a function of frequency using a third-octave band frequency scale. (c) Hourly mean noisiest levels (Leq) within the four species-specific frequency bands (see Methods section for details; Orca-low refers to 1.5–3.5 kHz frequency band, Orca-mid refers to the 5–12 kHz band). (d) Daily mean wind speed. Levels in panels (a), (b) and (c) are in dB re: 1 µPa (with 1 Hz band, third octave band and the previously described species-specific frequency bins, respectively).

Figure 3.

Four-panel plot showing hourly noise levels as recorded during 2010 on the marine autonomous recording unit located near Kitimat and wind speed, where the month/day notations (e.g. ‘09/08’) indicate local midnight. (a) Long-term spectrogram showing the hourly, time-varying distribution of acoustic energy as a linear function of frequency. (b) Long-term spectrogram showing the hourly, time-varying distribution of acoustic energy as a function of frequency using a third-octave band frequency scale. (c) Hourly mean noisiest levels (Leq) within the four species-specific frequency bands (see Methods section for details; Orca-low refers to 1.5–3.5 kHz frequency band, Orca-mid refers to the 5–12 kHz band). (d) Daily mean wind speed. Levels in panels (a), (b) and (c) are in dB re: 1 μPa (with 1 Hz band, third octave band and the previously described species-specific frequency bins, respectively).

Results

A brief discussion of our results follows. For a fuller description, please see Supporting Information Appendix S1.

Noise statistics from each MARU are reported as spectral levels in decibels (dB re: 1 μPa) in five species-specific frequency bands (Fig. 1). The selected model indicated a strong effect of both site and frequency band on noise, even after accounting for temporal autocorrelation, measurement of multiple frequencies per sample and unequal variance. There was strong support from the data (ΔAIC = 1153.61; Fig. 1, Supporting Information Table S1) for persistent between-site differences in ambient noise in all frequency bands (expect the highest frequency band, which had only three deployments) over the second-choice model, which assumed that sites varied randomly around some common underlying value. Ambient noise levels are shown in finer resolution in the Supporting Information Table S1.

Loss of communication space as a result of increased background noise for fin, humpback and killer whales

Within the 17–28 Hz frequency band encompassing fin whale 20-Hz song notes, a singing fin whale loses 30% of its communication space between 0 and 32 km as a result of living in the noisiest BC environments (ratio of call propagation range under quiet to noisy conditions; summarized in Table 2, but see Supporting Information Table S2 for metrics from each MARU deployment in each frequency band). Table 2 shows loss of communication space metrics for all three species and all five frequency bands under various scenarios of putative communication range and noise ratios (i.e. median relative to quiet; noisy relative to quiet). On average, singing fin whales lose < 1% of their communication space due to chronic ocean noise (i.e. the ratio of fin whale call propagation in the six noisiest sites relative to the six quietest sites). Singing humpback whales (71–708 Hz) may lose 80 or 94% of communication space (0–32 and 0–16 km, respectively) in the noisiest environments (Table 2; Supporting Information Table S2), or 35 or 52% under moderate conditions. Killer whales living in the noisiest sites in BC lose up to 97% of their acoustic communication space in the mid-frequency bands (1.5–3.5 kHz) used by killer whales for omnidirectional, burst-pulse, social communication calls (Ford, 1989; Miller, 2006), relative to the median quietest deployment noise condition, calculated over an 8-km range (Table 2; Supporting Information Table S2). At closer distances and in the higher frequencies used for whistles or echolocation clicks, less loss of communication space is expected from chronic noise (Table 2).

Table 2. Average loss of communication space
% Lost communication space
 Communication range (km)All median – noisy conditions (%)All median – noisiest conditions (%)
  1. This table lists the average lost communication space for all marine autonomous recording unit (MARU) deployments for each species-specific sound type and for increasing circular areas of increasing radius around the MARU. A value of lost communication space is given as the percentage of space under median ambient deployment noise conditions (i.e. P50, left column) and median noisiest deployment noise conditions (i.e. P05, right column) relative to noise condition at the MARU with the lowest (median quietest, i.e. P95) noise condition. These values assume the following source levels in the species' call-specific frequency band (see additional details in Methods section): fin whale (180 dB), humpback whale (170 dB), killer whale burst-pulse calls (155 dB), killer whale whistles (145 dB), killer whale echolocation clicks (195 dB). For the low-frequency portion of their song (71–708 Hz), singing humpbacks lose about a third of their communication space under median ambient noise conditions compared with median quietest conditions, and killer whales communicating in their low- (1.5–3.5 kHz) and mid-frequency (5–12 kHz) bands lose more than half of their communication space under the median ambient conditions compared with median quietest conditions. Under the noisiest conditions, the communication space for a singing humpback is reduced to less than 20%, while that for a killer whale communicating in their low- and mid-frequency bands is reduced to less than 10% of the space expected under normally quiet conditions (see Supporting Information Table S2 for loss of communication space values for each MARU deployment). Note that ‘NA’ means that even under quietest conditions, the signal would not be detectible at that range.
Fin whale Balaenoptera physalus0–32130
17–28 Hz0–16017
Humpback whale Megaptera novaeangliae0–325294
71–708 Hz0–163580
Killer whale Orcinus orca low0–86297
1.5–3.5 kHz0–43491
0–21968
0–1955
Killer whale mid0–480NA
5–12 kHz0–230NA
0–11891
0–0.51261
Killer whale high0–200
18–30 kHz0–100

Exploratory analyses of anthropogenic versus natural sound sources

The LTSA from Haro Strait (Fig. 2) was dominated by broadband noise that is characteristic of ship engines. In contrast, the LTSA from Kitimat (Fig. 3) showed a strikingly diurnal pattern in which ship noise was evident during daylight hours, but absent from midnight to early morning. A closer inspection of a 3-day period that included times of high and low wind speeds in both Kitimat and Haro Strait (Fig. 4) showed that high winds raised the ambient noise profile off Kitimat (top panel), but similar wind speeds in Haro Strait did not raise background noise levels, which were dominated by the persistent high levels of noise from ship traffic (bottom panel). We extracted audio files from five representative quiet and noisy periods in each of these sites (Figs 2 and 3), and make these illustrative examples available in the Supporting Information Fig. S1. In each of the noisy periods, ship noise is audible; in each of the quiet periods, no ship noise is audible.

Figure 4.

Two-panel hourly spectrogram plot showing time-varying distribution of acoustic energy (third-octave band frequency scale) as recorded on the marine autonomous recording unit located near Kitimat (top panel) and in Haro Strait (bottom panel) during the same 3 days (19–22 September 2010), which included a period with high winds. The occurrences of vessels are noted on both panels, with vessel traffic near Kitimat occurring during daylight hours and traffic at Haro Strait occurring throughout the 24-h day. The period with wind is evident and noted on the Kitimat panel, but not evident on the Haro Strait panel as a result of the nearly continuous high noise levels from vessel traffic.

Discussion

Our study provides a broadscale overview of noise levels in frequency bands relevant to the sounds produced by three at-risk whale populations in Canada's Pacific region. Statistically significant spatial variability was found in ambient noise levels, even after accounting for temporal autocorrelation. This means that anthropogenic noise events occur often enough in some sites to result in persistent differences in ambient conditions along the BC coast. Critical habitats for both northern and southern resident killer whales (Robson Bight and Haro Strait, respectively) were the noisiest in the frequency bands that killer whales use for social communication, although there is a legal obligation to protect acoustic attributes of these two sites. Sites from Kitimat to Caamano Sound, which have been proposed as critical habitat for humpback whales (Nichol et al., 2010) and are known to be important for fin whales (Williams, Ashe & O'Hara, 2011a), have some of the quietest low-frequency levels observed in our study, but there is no legal requirement to keep those habitats quiet. These quiet areas are poised to become much noisier, given a number of major industrial developments proposed, including oil and gas pipeline construction, associated oil and liquefied natural gas (LNG) tanker traffic, and new and expanded LNG, coal, and cargo container port terminals. Similarly, large development applications are pending for the ports of Vancouver and Prince Rupert, which may make noisy areas even noisier.

Our estimates of the loss of communication space indicate one important area for future research, namely to better quantify the spatial scale at which whales actually use these communication signals. Our results show (Table 2) that the greater the range at which whales are attempting to communicate, the greater the cause for concern about increases to ocean noise levels. We only considered fairly modest ranges relative to the distances over which these sounds can propagate theoretically (Payne & Webb, 1971). Active space of some killer whale calls in quiet conditions has been estimated at 25.8 km (Miller, 2006), and field studies have observed killer whales communicating over several kilometres (Deecke, Ford & Slater, 2005; Saulitis, Matkin & Fay, 2005). Given the logistical constraints of estimating the range at which whales respond to acoustic signals of conspecifics, our analyses provide a precautionary indication that humpback and killer whales living in noisy environments off BC are losing substantial fractions of their opportunities to communicate (Table 2). While additional research is needed to refine the precision and accuracy of our preliminary estimates, our study shows that these numbers are potentially large enough to warrant management action.

The quietest sites in our study (primarily mainland inlets) have substantial periods of quiet that are quieter than any recorded off the US east coast (Hatch et al., 2012). As researchers consider experimental approaches to better quantify ecosystem-level effects of noise (Boyd et al., 2011), unusually quiet places like BC's mainland inlets may present opportunities for conducting ocean noise or quieting experiments, or to consider novel conservation mechanisms to preserve acoustic wilderness sites. Broughton Archipelago was the quietest of our 12 sites in all frequency bands (Fig. 1). This site used to provide important feeding habitat for resident killer whales, until acoustic harassment devices (AHDs), deployed to displace seals from open-net salmon fish farms, inadvertently displaced killer whales from this part of their range (Morton & Symonds, 2002). Although AHDs were turned off in 1999 and Broughton Archipelago is now quieter than adjacent Johnstone Strait (Table 2), resident killer whales have still not returned after more than a decade (A. Morton, pers. comm.). This case study serves as an important reminder that high-amplitude noise can have real and lasting consequences for animal conservation.

The primary objective of our study was to compare the acoustic environments among sites, which required us to make many simplifying assumptions that ignored biological and behavioural mechanisms that whales may have to detect signals in noise, for example, increasing signal duration (Miller et al., 2000) or amplitude (Holt et al., 2008). Because our loss of communication space metric is a ratio, uncertainty in these compensatory mechanisms should not affect our relative ranking of sites. We have not considered physiological stress responses to noise (Rolland et al., 2012), but encourage additional research on this topic. Repeated disturbance from low-amplitude noise disrupts feeding activities of killer whales (Williams, Lusseau & Hammond, 2006), which may be prey limited (Fisheries and Oceans Canada, 2011; Williams et al., 2011b). For humpback and fin whales, audiograms are lacking (Erbe, 2002; Clark & Ellison, 2004), but evidence is growing that baleen whales also use acoustic signals in foraging activities (Sharpe & Dill, 1997; Stimpert et al., 2007; Simon et al., 2010). Given no additional information on compensatory mechanisms, the among-site differences we identify (Fig. 1) are large, and suggest that the concept of acoustic masking warrants consideration in area-based management efforts to protect whales and their habitats (Wilson et al., 2004).

Many initiatives are now calling for precautionary reductions in ocean noise. For example, the ‘Okeanos’ target,2 embraced by the International Maritime Organization and the International Whaling Commission, aims to reduce the contributions of shipping to ambient noise energy in the 10–300 Hz band by 3 dB in this decade. Unlike chemical pollutants or legacies of overfishing, noise energy from individual sources can be readily quantified and is quickly lowered when a source is removed or mitigated. It should be noted that the between-site differences we report here (Fig. 1) are large, relative to the Okeanos 3 dB target. From a practical mitigation standpoint, identifying quiet areas and keeping them quiet will be easier than trying to remove sound sources from noisy areas.

The quietest noise measurements in our study match those reported by Urick (1983). This means that there are places and times today where once-normal windows of quiet still exist. For marine mammals and fishes, these are the levels in which natural and sexual selection occurred, and these are the conditions under which the ‘acoustic arms race’ between predator and prey evolved (Tyack & Clark, 2000). Today, throughout the Northern Hemisphere, those once-normal levels are becoming the exception rather than the rule. The low-frequency ocean noise baseline has shifted, and our perceptions of normal acoustic environments may have shifted as well (Pauly, 1995). Our masking metrics tell us that noise now has to be treated as a habitat-level stressor, and while acousticians have an essential role to play, interdisciplinary and modelling studies are needed to explore ecosystem consequences of noise, much as we do for other anthropogenic stressors, such as urbanization (Soule et al., 1988), overfishing (Jackson et al., 2001) and bleaching of coral reefs (Munday, 2004). This will require acoustic monitoring over spatial and temporal scales at which whales and other highly mobile, migratory and long-lived species live. After all, global shipping activities have risen to dominate ocean soundscapes during roughly the lifetime of a fin whale (Lockyer, 2007). Our study provides 12 snapshots in important whale habitats, representing conditions over several months across 3 years. We believe that our measurements of present-day quietest conditions offer a pragmatic proxy to reconstruct what BC coastal waters may have sounded like in a pre-industrial era, while noisier sites offer a plausible prediction of what quiet sites may sound like as shipping traffic expands.

Future science-policy discussions could proceed in at least two ways. One approach is to model population consequences (Thompson et al., 2013) likely to result from acoustic masking (National Research Council, 2003). For species like killer whales, where it is known how populations respond when prey are plentiful or scarce (Ward, Holmes & Balcomb, 2009; Ford et al., 2010), one could model the mortality equivalent of prey reduction via masking of acoustic foraging signals. This provides contrasting modelled trajectories under quiet and noisy conditions, and the difference could be considered an indirect, population-level effect to be compared with a country's stated objectives (Moore, 2012). Another approach is to consider noise as a ‘cost’ when incorporating anthropogenic noise as a persistent feature of the acoustic environment in marine spatial planning. Systematic conservation planning involves setting targets, for example, the percentage of a species' range to set aside as critical habitat or as a protected area. Spatial conservation prioritization optimizes protected area network design to achieve that target (Moilanen & Wilson, 2009). It would be simple, analytically, to include chronic ocean noise or masking metrics as a cost in a spatial prioritization framework, directing algorithms to select the quietest habitat that is most important to whales. Or, masking could be used to modify the targets themselves, such that areas with greater noisy environments than quiet environments need to be set aside to achieve the same level of protection. We are not advocating any particular policy outcome, but note that without a ‘cost’ of noise, either in terms of population- or environmental-level targets, there will never be an incentive to reduce noise levels in critical habitats of acoustically sensitive species.

Acknowledgments

This study represents field and analytical efforts that span several years. R.W. and E.A. wish to thank many mariners and funders for their help with making the noise measurements at sea. For fieldwork, including valuable donations of boat time, logistical support, local knowledge and feedback on early drafts, we thank MaryAnn Ashe, Dave Bell, Ian Boyd, Nic Dedeluk, Volker Deecke, Graeme Ellis, John Ford, Thomas Gotz, Phil Hammond, Sarah Haney, Stan Hutchings and Karen Hansen, Vincent Janik, Mark Johnson, Nicole Koshure, Jean-Marc LeGuerrier, Bill and Tyson MacKay, Ian and Karen McAllister, Alexandra Morton, Larry Olsen, Chris Picard, Kat Piraino, Marven Robinson, Kathryn Ross, Doug Sandilands, Angela Smith, Paul Spong, Doug Stewart, Helena Symonds, Peter Tyack, Val and Scott Veirs, Jane Watson, James and Don Willson, Janie Wray and Hermann Meuter, and Andrew and Helen Wright. E.A. and R.W. would like to thank the many individuals and foundations for supporting Oceans Initiative's acoustic ecology project at Tides Canada over the years, including Meaghan Calcari, Darcy Dobell, British Columbia Marine Planning Fund, Willow Grove Foundation, Dieter Paulmann (Okeanos), National Fish and Wildlife Foundation, Mountain Equipment Co-op, Patagonia, Erich Hoyt (Whale and Dolphin Conservation), Marisla Foundation and Canadian Whale Institute. We thank Ward Krkoska and Fred Channel for equipment preparation, deployment and recovery; and Peter Dugan, Mike Pitzrick, Chris Pelkie, Chris Tessaglia-Hymes, and Adam Strickhart (Cornell BRP) for software and computer system development, data recovery, and analysis. We would like to thank Doug Sandilands and Stacy Rebich Hespanha (NCEAS) for their help with summarizing a great deal of information in Fig. 1, and Len Thomas for help modelling temporal autocorrelation in daily noise levels. We would like to thank two anonymous reviewers for helpful reviews and suggestions, which improved the paper. R.W. was supported during 2009–2010 as a Canada-US Fulbright Visiting Research Chair at University of Washington (Seattle) and in 2010–2012 as a Marie Curie International Incoming Fellowship for project CONCEAL (FP7: PIIF-GA-2009-253407).

Notes

  1. 1

    http://www.ndbc.noaa.gov/download_data.php?filename=46088h2010.txt.gz&dir=data/historical/stdmet/

  2. 2

    http://www.okeanos-foundation.org/assets/Uploads/Hamburg-shipping-report-2.pdf

Ancillary