Introduction

Human-made noise in the ocean has been increasing since the 1950s, primarily due to an increase in both the number and size of container ships worldwide1. Global ship noise has been predicted to double every 11.5 years2, although with significant regional differences. However, these shipping trends were radically altered during March 2020, with the onset of the COVID-19 global pandemic leading to worldwide lockdowns peaking during April 20203. The pandemic significantly impacted global vessel traffic, resulting in a 4.1% decrease in global maritime trade in early 2020—the largest decline since 20094. In certain economic zones, marine traffic decreased by ~70%3, with models estimating a 6% reduction in shipping noise energy worldwide in the 63 Hz one-third octave band1.

Radiated noise generated by maritime vessels (typically 0.01–10 kHz) has long been known to significantly contribute to underwater soundscapes5. Specifically, container shipping generates low-frequency noise (<1 kHz) near ports and shipping routes, raising ambient sound levels over distances of 10–100 s of kms6. Previous studies do show that the pandemic led to quieter ocean conditions due to reduced marine traffic, with a decrease in low-frequency (≤125 Hz) ocean sound levels in shallow-water areas of marine tourism and recreation7,8, as well as in deep-water areas due to decreased commercial vessel traffic3,9,10.

Thompson and Barclay9 were one of the first to show that a significant drop in commercial ship traffic in the northeast Pacific Ocean during the COVID-19 pandemic (from 2019 to Summer 2020) was linked to reduced ocean noise conditions. They showed that container ship traffic to the Ports of Vancouver and Seattle during this time decreased by ~13% and 18%, respectively, correlated with reduced ocean sound levels of ~1–2 dB as recorded at cabled hydrophones off Vancouver Island. Similarly, Dahl et al.11 used cabled hydrophone data off the Oregon Coast to show ocean noise levels reduced by 1.6 dB during 2020 relative to the mean of the previous 5 years. Several additional studies (Table 2) have shown similar (range of 0.7–8.4 dB) regional reductions in ocean sound levels due to decreased ship traffic, including Glacier Bay Alaska, U.S.7, Monterey Bay, west coast U.S.10,12, Nova Scotia, eastern Canada13, the southeast U.S. Atlantic coast shelf break14, Bahamas (western Atlantic)15, the French West Indies16, the German Baltic17, as well as the North and Adriatic Seas18. A reduction of 1–3 dB in the 10–100 Hz range during 2020 was also reported across the Atlantic, Pacific and Indian Ocean basins using data from hydroacoustic stations of the International Monitoring System19. Interestingly, Miksis-Olds et al.14 did not find a consistent COVID-19-related decrease in sound levels from 2019 to 2020 off the southeast U.S. Atlantic coast. Even though there was a decrease in international cargo and tanker vessels during this time period, domestic transport and fishing vessels in this area remained fairly constant.

Here we add observations of the effect of reduced ship traffic on ocean sound levels during the pandemic, as well as investigate how widespread these sound conditions were across the northern and deeper regions of the Pacific and Arctic Oceans. We review several years of passive acoustic data from 2018 to 2022 sampled at three stations that are part of the National Ocean and Atmospheric Administration-National Park Service Ocean Noise Reference Station Network (ONRSN)20 hydrophones: Ocean Station Papa (OSP) in the Gulf of Alaska, Olympic Coast National Marine Sanctuary (OCNMS), and the Beaufort Sea in the Alaskan Arctic Ocean. We also present temporally overlapping data from a NOAA-Pacific Marine Environmental Laboratory (PMEL) hydrophone moored at Axial Seamount in the northeast Pacific (Fig. 1 and Table 1). Passive acoustic data are compared to vessel movement data and other sound sources to document and understand ocean soundscape conditions pre, during, and post the COVID-19 pandemic.

Fig. 1: NOAA PMEL hydrophone locations in relation to ship traffic density.
Fig. 1: NOAA PMEL hydrophone locations in relation to ship traffic density.
Full size image

a Map showing NOAA PMEL hydrophone locations in the northeast Pacific in relation to the major container ship traffic lanes in the region. The yellow and red colors on the ship tracks represent a high density of traffic, hundreds of ships per year. Blue circles show the locations of the NOAA-NPS hydrophones at Ocean Station Papa, Axial Seamount, the Olympic Coast Sanctuary, as well as the Alaskan Arctic, Beaufort Sea. The hydrophones were deployed near the major shipping lanes, to be well positioned for monitoring ship noise. b Shows the location of the Beaufort Sea hydrophone in relation to the northeast Pacific hydrophone stations in figure (a). c Shows schematic of hydrophone mooring that collected acoustic data presented here48.

Table 1 Hydrophone locations and deployment dates
Full size table

Results

Observed sound levels at Ocean Station Papa

Although there is some overlap in multiyear sound level trends, our analysis of low-frequency sound levels at OSP indicates both the 63 and 125 Hz one-third octave bands (TOB) exhibit a 1–2 dB (dB re 1 μPa2, hereafter dB) (~10–30%) decrease concurrent with the onset of the COVID-19 pandemic during Spring (April) to Summer (August) 2020 as compared to the pre-pandemic years of 2018–2019 (Fig. 2). The 63 and 125 Hz TOBs sound levels later increased to pre-pandemic levels during Fall and Winter of 2020. However, sound levels during 2021were higher than the previous 3 years, increasing ~2–3 dB during April–September of 2021, with peaks during mid-July to September (most readily observed in the 63 Hz band), and an additional increase during late October to December. During 2022, sound levels decreased to levels similar to 2018–2019, and even comparable to the relatively low levels observed in 2020.

Fig. 2: Median one-third octave bands centered at 63 and 125 Hz recorded during 2018–2022 at Ocean Station Papa.
Fig. 2: Median one-third octave bands centered at 63 and 125 Hz recorded during 2018–2022 at Ocean Station Papa.
Full size image

Two-week running average sound levels for each calendar year are visualized as smoothed solid colored lines overlaid on semi-transparent points that represent daily median sound levels. Breaks in the solid colored lines indicate gaps in the recording. Each of these northeast Pacific hydrophone sites shows a 1–3 dB reduction in sound levels during Spring and Summer 2020, which represents a substantial 20–30% decline during early 2020, consistent with reduced ship traffic levels.

Observed sound levels at Axial Seamount

Observed 63 and 125 Hz TOB sound levels recorded at Axial Seamount were significantly variable during 2018–2021. Most notably, in 2019, a nearby research airgun survey was conducted by the R/V Langseth for 3 weeks during July–August21 increasing low-frequency ambient sound levels (Fig. 3a, b). The airgun survey ranged from 1.5 to 30 km from the Axial hydrophone, generating hourly mean received levels of 75–112 dB re µPa2. Additional information on the survey is provided in the section “Short-duration airgun survey temporarily increased ambient sound levels at axial seamount”.

Fig. 3: Median ocean sound levels sampled at Axial Seamount.
Fig. 3: Median ocean sound levels sampled at Axial Seamount.
Full size image

a Median ocean sound levels in the one-third octave bands centered at 63 and 125 Hz during 2018–2021 sampled at Axial Seamount. Two-week running average sound levels for each calendar year are visualized as smoothed solid colored lines overlaid on semi-transparent points that represent daily median sound levels. Breaks in the solid colored lines indicate data gaps. The solid blue arrow points to spring–summer 2020, when low ambient sound levels were observed. Increase in 2021 levels is likely due to a change in hydrophone depth. b Received sound levels from an airgun survey at Axial Seamount during July–August of 2019 in comparison to received levels during years with no surveys. Low received levels observed during the survey were due to a pause in airgun activity21.

Outside of the seismic survey time window, a 1–2 dB decrease in sound levels is reflected in both the 63 and 125 Hz TOBs from January to late August 2020, relative to 2018–2019 pre-pandemic levels (Fig. 3a). The short data gap from August to September 2020, after which sound levels significantly increase, is due to recovery of the hydrophone and the time required to perform the re-deployment of the hydrophone during the servicing cruise. Following redeployment, sound levels immediately appear 6–10 dB higher than at the end of the previous recording and then increase in mid-September through Fall-Winter of 2020. These elevated levels continued through 2021 (in both TOBs). As noted above, the late 2020 to 2021 sound levels are significantly elevated, by ~6–10 dB, compared to the 2018–early 2020 levels. This difference is likely due to the 625 m difference in moored depths of the hydrophones, which could result in a reduction in sound intensity (spherical loss) of several 10 s of dB at the deeper hydrophone from a vessel or other sound source directly overhead. This estimate of relative loss due to depth differences, however, is not necessarily straightforward when dealing with long-term, ambient sound levels, since multiple sound sources (e.g. ships, wind, waves) are involved over time and source locations are not precisely known.

Unfortunately, the hydrophone deployed at Axial in July 2020 stopped logging data on 5 December 2021, and no early 2022 acoustic data is available for comparison to the other sites in this analysis. Nevertheless, the overall sound levels in the ship noise bands were found to be substantially higher at Axial Seamount in the post-pandemic years as compared to pre-pandemic. The Axial Seamount acoustic data also shows a peak in levels during June–August of 2021, similar to the OSP records.

Observed sound levels at Olympic Coast National Marine Sanctuary

At OCNMS, in the 63 Hz TOB, sound levels during 2020 are significantly lower (1–3 dB) through all four seasons as compared to 2018 and 2019 levels (Fig. 4). An exception to this occurs during early August when levels at OCNMS rise to 3–4 dB higher than 2018–2019. The OCNMS 125 Hz band also shows lower levels during 2020 as compared to 2018–2019, although the levels during these years show significant overlap across seasons, with the exception of high levels during November 2019 and early August 2020. Sound levels during 2021–2022 are similar to 2020 levels during Winter–Spring months, then become significantly lower than levels observed in 2020 during mid-May to mid-August in both TOBs.

Fig. 4: One-third octave band sound levels sampled at Olympic Coast National Marine Sanctuary from 2018 to 2022.
Fig. 4: One-third octave band sound levels sampled at Olympic Coast National Marine Sanctuary from 2018 to 2022.
Full size image

Two-week running average sound levels for each calendar year are visualized as smoothed solid colored lines overlaid on semi-transparent points that represent daily median sound levels. Breaks in the solid colored lines indicate data gaps. A several dB decrease is apparent from Spring 2019 to 2020. However, Spring 2021–2022 levels (shown in yellow and green) also remain low, whereas Fall 2018–2022 are relatively high, possibly due to baleen whale vocal presence (see Fig. 8). Annual ship traffic counts gradually increase from 2019 tp 2022.

Observed sound levels at Beaufort Sea—Arctic Ocean

Significant variability was observed in the 63 and 125 Hz TOB sound levels across all 5 years of recordings in the high Arctic. Sound levels varied by over 10 dB in the same month over multiple years, where 2019 levels were as much as ~10 dB lower during mid-March and mid-September compared to the other years. Sound levels also varied by over 20 dB across seasons. Specifically, during the high sea-ice coverage of the winter-spring seasons, the 63 and 125 Hz TOB sound levels recorded in the Beaufort Sea (Arctic Ocean) were ~80–85 dB, However, sound levels increase to ~90–97 dB during late summer-fall (Fig. 5).

Fig. 5: One-third octave band sound levels sampled in the Beaufort Sea from 2018 to 2022.
Fig. 5: One-third octave band sound levels sampled in the Beaufort Sea from 2018 to 2022.
Full size image

Two-week running average sound levels for each calendar year are visualized as smoothed solid colored lines overlaid on semi-transparent points that represent daily median sound levels. Breaks in the solid colored lines indicate data gaps. A decline is apparent in 2020, though it is not as pronounced as other sites. This is likely due to noise from surface sea-ice which shows significant seasonal variability and can dominate sound levels here. Sound levels at the Beaufort Sea site provide an interesting comparison to the northeast Pacific because of the high level of environmental sound sources rather than anthropogenic ones.

Anthropogenic sound sources: estimates of regional vessel presence

Overall, the AIS data show significant variability in total ship counts (presence) between the four sites where the long-term sound level records were sampled (Table 3). For example, comparing Axial Seamount and OCNMS (both of which are near ship traffic lanes), the area around Axial averaged roughly a thousand ships per year whereas the AIS records at OCNMS showed >100,000 ships entering the highly trafficked Strait of Juan de Fuca. In contrast, there is a relative absence of ships near Ocean Station Papa (Gulf of Alaska) and the Beaufort Sea during all four years. This relative absence in the Beaufort Sea area is understandable given its remoteness and seasonal presence of sea-ice which limits ship access to the area, and our observations align with AIS records analyzed for the same location from 2016 to 201722. OSP is also several hundred kilometers from major northeast Pacific ship traffic lanes. Interestingly, ship traffic near OSP and the Beaufort hydrophones was highest in 2019-2020, which dropped to single digits and zero ships in 2022–2023. This full reduction in container shipping at OSP and Beaufort during 2022 and 2023 is unexpected and could be due to the steady increase of sea-ice into the Beaufort Sea observed over the last decade23 as well as possibly reflecting the disruption to overall maritime activity in the North Pacific during the COVID-19 pandemic19.

Histograms of ship counts were also produced for all four sites during 2019–2022. However, since there were so few ships at OSP and the Beaufort Sea sites, we show only the count histograms for ships near OCNMS and Axial Seamount (Figs. 6 and 7). The OCNMS site overall shows an order of magnitude more ships (>100k) than the next highest totals from Axial Seamount, as well as clear annual peaks in ship counts beginning in May and extending until mid-October, even during the height of the pandemic economic downturn in summer 2020. Axial shows significantly less ship activity than OCNMS but still exhibits annual seasonal variability with the same peak in ship traffic during the late spring-early fall time frame.

Fig. 6: Monthly vessel counts within the Olympic Coast National Marine Sanctuary.
Fig. 6: Monthly vessel counts within the Olympic Coast National Marine Sanctuary.
Full size image

Counts of ships derived from the Automated Identification System44 for all ships within a 222 × 222 km square area centered on the hydrophone mooring location at the Olympic Coast National Marine Sanctuary during 2019–2023 (Table 3). Diagrams show clear seasonal patterns with peak ship activity occurring during May–October.

Fig. 7: Monthly vessel counts at Axial Seamount.
Fig. 7: Monthly vessel counts at Axial Seamount.
Full size image

Counts of ships derived from the AIS44 for all ships within a 222 × 222 km square area centered on the hydrophone mooring location at Axial Seamount during 2019–2023 (Table 3). Histograms here highlight the significantly lower ship count at Axial as compared to OCNMS (entrance to the Straits of Juan de Fuca). A seasonal pattern with peak ship activity occurring during May–October similar to OCNMS is also observed.

Short-duration airgun survey temporarily increased ambient sound levels at axial seamount

The Axial Seamount hydrophone recorded a research airgun survey conducted by the R/V Langseth for 3 weeks during July–August 2019 that temporarily dominated the received levels21 (Fig. 3a). Axial Seamount is a 1500 m deep, intermittently active submarine volcano that has been the focus of geophysical studies for decades24. The survey ranged from 1.5 to 30 km from the Axial hydrophone, generating hourly mean received levels of 75–112 dB re µPa2. Figure 3b shows the hourly mean airgun sound levels in the TOBs. The maximum airgun received levels are significantly lower than the predicted levels for this system (160 dB re 1 µPa-rms at distances of 0.5–6.7 km)21. The periods of significantly reduced received levels seen on 20 and 24 July, and 1 and 7 August (Fig. 3b), were due to the survey being shut down both for maintenance and for sighting of marine mammals within 1 km of the airguns21.

Seasonal presence of blue whale vocalizations

Several decades of recording northeast Pacific blue whale (Balaenoptera musculus) vocalizations25,26 have shown that their fundamental calls can dominate the 15–30 Hz band, while overtones typically range in the 45-60 Hz band but can extend up to 100 Hz27. Moreover, their calling exhibits strong seasonality and is likely a significant contributor to ocean sound levels in the North Pacific when these whales are present from February to April and August to November28. Therefore, in addition to the increased ambient noise levels in the late fall, winter, and early spring months due to increased sea surface wind and waves, baleen whale call energy likely contributes significantly to these seasonal fluctuations in sound levels, especially in the 63 Hz TOB where blue whale overtone energy can be present. We therefore examined blue whale calling rates in the datasets, particularly at OCNMS (Fig. 8). Blue whale calling rates at OCNMS begin rising each year in August and September, and end by February of the following year, which is consistent with blue whale call observations at other sites across the NE Pacific26.

Fig. 8: Blue whale detections in the Olympic Coast National Marine Sanctuary.
Fig. 8: Blue whale detections in the Olympic Coast National Marine Sanctuary.
Full size image

a Average number of blue whale calls detected per day at the Olympic Coast National Marine Sanctuary (OCNMS) site from January 2018 to April 2022. Calling started to rise in August and had its highest levels from October to January, with a peak in November/December each year. b Spectra of OCNMS hydrophone data from May (left) and August (right) of 2020. The high levels of broadband energy (yellow and orange colors) during mid-May and August are likely propeller sounds from nearby container ships. Blue and fin whale vocalizations can also be seen to be consistently present in the 1–100 Hz band during time periods with less ship noise.

Influence of wind, wave and current noise on ambient sound levels

It has long been observed that wind and wave action at the sea surface can contribute substantially to ambient ocean sound levels5, and this is observed in the Arctic regions during seasons with minimum sea ice and open water29,30. Overall, the acoustic signals produced by sea surface wind and wave action are typically in the 200–500 Hz band and drop off rapidly in received levels ±200 Hz around this center frequency5,31. This appears to be confirmed in a recent study of the effect of wind speeds on TOB levels in the Bering Sea32. Wind-induced sound levels in the 63 and 125 Hz TOBs were significantly less than wind-induced sound levels in the 250 Hz TOB (0.47 dB per m s−1 versus 0.86 dB per m s−1, respectively) and are not as significant a contributor to the TOBs as ship noise.

Current speeds have also been shown to be a contributing factor to moored-hydrophone recorded ocean sound levels, especially at locations of constricted water flow such as the Bering Sea31. However, the mooring line used for the hydrophone deployments in this study was made of Vectran™, which is composed of multifilament, liquid-crystal polymer with high tensile strength and low friction. The low friction keeps strumming resonance-noise from the mooring lines at low amplitudes under 10 Hz, and strumming was not viewed as a significant contributor to sound levels in the 63 and 125 TOBs.

Discussion

Analysis of the multiyear broadband spectra and the 63 and 125 Hz TOBs from three deep-ocean (~500-1500 m) sites in the northeast Pacific Ocean, and one from the Arctic Ocean, show several interesting long-term trends in sound sources and levels sampled from 2018 to 2022. Overall, sound levels at OSP, OCNMS, and Axial Seamount range from 80 to 107 dB re µPa2, with OCNMS exhibiting the highest levels annually (90–107 dB re µPa33 in both bands). Moreover, OSP, OCNMS and Axial all also exhibited a several-decibel seasonal variation in sound levels (winter highs and summer lows), which is likely driven by seasonal wind and sea-surface wave effects, as well as baleen whale vocalizations26,34. Our analysis of ocean sound levels during the onset of the COVID-19 pandemic indicates that sound levels in the 63 and 125 Hz one-third octave bands at OSP and Axial Seamount (Figs. 2 and 3) decreased 1–2 dB (~10–30%) during Spring (April) to Summer (August) 2020 as compared to the pre-pandemic years of 2018–2019, consistent with previous studies worldwide (Table 2).

Table 2 Global reductions in Ocean Noise during the COVID-19 Pandemic - 2018-2022
Full size table

In addition, the Axial Seamount hydrophone recorded a research airgun survey during July–August 2019 that temporarily, but significantly, elevated received levels at the seafloor (1500 m depth). Also, the Axial data exhibits a significant difference (~6–10 dB) in the annual levels between 2018–2020 and late 2020–2021. This increase is related to the 625 m difference in hydrophone depth between the two deployments.

The OCNMS data, in particular, shows peak sound levels during May–June and August-September, which are likely driven by nearby vessel noise (Fig. 8). The OCNMS 125 Hz TOB band was also lower during 2020 as compared to 2018-2019, however, levels during these years show significant overlap across seasons, and very high levels during early August 2020 and November 2019. In contrast, the 63 Hz sound levels at OCNMS during 2020 are significantly lower (1–3 dB) through all four seasons as compared to 2018 and 2019 levels, with the exception of early August, when levels at OCNMS rise to 3–4 dB higher than 2018–2019. The drivers of the relatively high TOB levels observed at OCNMS in May and August 2019–2020 are not clear, however manual data review revealed that the increased ambient sound is likely related to vessel activity. Although it is also possible for cetacean vocalizations to increase sound levels in the frequencies, the November–December 2019 2020 timing of the peak in the number of blue whale vocalizations is not consistent with the August sound levels maxima (Fig. 8a), and our analysis did not find increased blue whale calls present in May of 2019 or 2020 that would explain the high sound levels observed during this time. Moreover, it is not likely caused by other vocal cetacean species, since the call frequency range for fin whales is typically below 30 Hz, and humpback calls tend to range across the spectrum without projecting significant energy into any one frequency band28,34.

Total ship counts near OCNMS were lower in 2019 and 2020 as compared to 2021–2023 (Figs. 9, 6 and Table 3), due to port closures, container, and cruise ships were forced to queue at access locations across the U.S. West Coast. The queueing of ships peaked during late spring and summer of 202035, which can also be seen in the May and August 2020 OCNMS hydrophone spectra showing very high levels of broadband ship propeller noise (Fig. 8b). The high sound levels in the 125 Hz TOB might also be in part due to increased wind/wave levels during this time period. Indeed, storms along the Olympic Coast in 2020 were notably intense, with stronger-than-usual wind speeds recorded, leading to more frequent storm surges and high winds compared to previous years36.

Fig. 9: Multiyear, long-term spectral averages (LTSA) of ocean sound levels (1–400 Hz) during 2018–2023 from four autonomous underwater hydrophone sites across the northeast Pacific Ocean and Beaufort-Arctic.
Fig. 9: Multiyear, long-term spectral averages (LTSA) of ocean sound levels (1–400 Hz) during 2018–2023 from four autonomous underwater hydrophone sites across the northeast Pacific Ocean and Beaufort-Arctic.
Full size image

The hydrophone data spans multiple years at each site: a OSP (2018–2022), b Axial (2018–2021), c and d OCNMS (2017–2022), and e Beaufort (2018–2022). Each LTSA visualizes the acoustic impact of natural and anthropogenic sources on overall ambient sound levels. Note, peak sound levels at Axial Seamount during July of 2019 caused by a research airgun survey by R/V Langseth. Overall, the broadband sound levels at each site show seasonal variation due to increased wind/sea states during the winter months. At this scale, reduced sound levels due to anthropogenic sources (container shipping) during summer 2020 is not readily apparent.

Table 3 Annual ship counts
Full size table

The Arctic Ocean-Beaufort Sea site (Fig. 5) also shows a similar reduction in 2020 sound levels, at 1–3 dB lower from April to June than in other years, consistent with the pandemic-related reductions observed at the other hydrophone sites. However, the observed reduction is over a shorter time period (ending in June) as compared to the other sites and also reflects seasonal variation in sea-ice coverage37,38. The 2019 winter levels were even lower at times than 2020, which again likely reflects the significant role sea-ice plays in the Beaufort Sea soundscape.

Sea-ice can create significant levels of sound through the collision of large ice masses, and thermo-mechanical induced fracturing during the annual freeze–thaw cycles. For this study, the Beaufort Sea hydrophone was the only site exposed to sea-ice and thus with recorded sound levels potentially influenced by break-up sounds. In addition to seasonal variation in the contribution to ambient sound levels, recent rapid reductions in sea-ice coverage in the Beaufort Sea over the last few decades37,38 have likely led to significant changes in cryogenic sounds, as well as potentially allowing for an increase in anthropogenic sounds from increased ship and construction activities.

Interestingly, there has been significant import of multi-year ice into the Beaufort Sea during both summer and winter over the last decade, but less multi-year sea-ice now survives during the summer38. Historically, the Beaufort Sea serves as a conduit for multi-year sea-ice transport from the central Arctic to the eastern Arctic39. This process is now thought to be slowed, but not interrupted38. Thus sea-ice transport through the Beaufort Sea is a dynamic seasonal process, which is likely reflected in significant seasonal variation in ambient sound levels due to sea-ice thermo-mechanical breakup and flow40,41.

The post-pandemic Beaufort Sea sound levels exhibit seasonal highs and lows that are consistent with levels observed during 2018-2020 (Fig. 5). For example, sound levels during 2021 and 2022 exhibit annual lows in November-December that are very similar to the three previous years, suggesting the sea-ice freeze-up caused by seasonal low temperatures are the main driver of ocean sound levels, rather than any anthropogenic sources.

Following the onset of the pandemic in early 2020, sound levels at OSP returned to pre-pandemic levels during the Fall and Winter of 2020 (Fig. 2). However, levels during 2021 are significantly higher than the previous 3 years, increasing ~2–3 dB during April to September of 2021, with significant peaks during mid-July to September (most readily observed in the 63 Hz band), and an additional increase during late October to December. Sound levels during 2022 decreased to levels similar to 2018–2019, and are also comparable to the relatively low levels observed in 2020. The cause of the overall increased levels during 2021 is somewhat mysterious, as large vessel traffic, the most likely sound source that would impact sound levels in these frequencies in this area and time is relatively non-existent. It is possible that vessels were moving in the area that did not carry AIS transponders or experienced a transponder malfunction during this time. We are confident that the observed increase is not related to data sampling, as all gain and pre-amp recording levels have been thoroughly reviewed and were calibrated for continuity across the 2018–2022 deployments. It seems possible that a combination of the northeast Pacific weather patterns from late 2020, and seasonal baleen whale presence26, contributed to a relative rise in 2021 levels. Furthermore, changes in weather or vessel activity may drive changes in baleen whale behavior, including movement and vocalization patterns.

Axial Seamount sound levels increased during late 2020 by ~5–10 dB, which is very likely due to changes in the water depth of the hydrophone, although levels still continued to rise from this new baseline offset and remained at these higher levels through 2021 (in both TOBs; Fig. 3). Thus, direct comparison of sound levels at Axial Seamount in the post-pandemic years as compared to pre-pandemic is not valid. However, considering only the 2021 levels, the Axial Seamount hydrophone shows a peak during late June-August of 2021, despite a significant decrease in ship counts in 2021. This relative rise is comparable in timing to increased levels observed at OSP and may be due to similar causal factors.

Interestingly, OCNMS sound levels during 2021–2022 are similar to, or lower during certain periods, than 2020 levels. The exception to this is the rapid increase in levels in May and August of 2019 and 2020 (Figs. 4 and 8) which may be due to queueing of ships nearby. Overall, however, ship traffic near OCNMS stays relatively high, and even increased over the ~4-year time period of this study. Thus, our interpretation is that annual variability of ship traffic may have caused sound levels at OCNMS to remain relatively consistent during the Winter and Spring months, but led to significant changes in levels during the Summer and Fall.

The significant decrease in ship activity throughout the North Pacific in 2020 was likely reflected in the sound level observed at Ocean Station Papa, Axial Seamount, Olympic Coast Sanctuary, and the Beaufort Sea. In many cases, vessels remained within the search area for several days to weeks, which increased the total overall ship count (Table 3). However, we view this extended temporal presence of these ships, which added to the total overall ship count in the area, as accurately representing the observed cumulative underwater noise from nearby vessels6. However, it is also possible that ships at distances >222 km could contribute to the sound levels, especially for large and/or fast-moving vessels.

Our observation of the annual counts of container vessels around each station (Table 3) is inconsistent with reduced shipping activity observed in the Strait of Juan de Fuca-Puget Sound region in 2020 during the pandemic economic downturn9. Although vessel traffic near OSP and within the Beaufort Sea did decrease during 2020, shipping near OCNMS and Axial did not. Moreover, shipping near OSP and within the Beaufort Sea apparently did not recover in the two years following the height of the pandemic economic impacts, whereas vessel counts at Axial Seamount showed an increase in 2020 as compared to one year before and after, and vessels near the OCNMS site actually showed steady increases each year from 2019 to 2023 (Figs. 6 and 7). We interpret the disparity in counts as reflecting proximity to major ship traffic lanes, with Axial and especially OCNMS being very near major congestion points. Given that the sound level decreases observed during 2020 at each site are likely due to a reduction in nearby ship traffic, it seems possible that the annual sound levels observed during the 2019–2022 time period at these four sites may be influenced by noise from ships outside of the 222 × 222 km square search area6. Also, in some regions, for example, the southeastern U.S. coast, sound level decreases in 2020 were observed at sites close to shipping lanes but were inconsistent with the timing of shipping changes related to a COVID-19 slowdown14. Moreover, recent maps of global maritime activity42 show that vessel counts based on AIS tracking miss roughly three-quarters of the large vessels (72–76% of the world’s industrial fishing vessels, and 21–30% of container and energy vessel activity) in public tracking systems, which may also account for the unexpected trends in ship counts observed. Passive acoustic datasets and analytical tools are increasingly providing information about vessel movement in parallel with AIS data, as well as in places and for vessels where AIS information is not available for. Although these datasets may be imprecise at times, coupling AIS and PAM data will likely enable more accurate vessel movement data. Future efforts to compare vessel noise and AIS data, as well as analytical methods for comparison, are necessary to increase the efficiency and accuracy of similar projects going forward.

The 1–2 dB reduction in northeast Pacific and Arctic sound levels during the Spring to Summer of 2020, quantified in this analysis, is very similar to the reduction in low-frequency sound levels observed at the Neptune Canada cabled hydrophones off Vancouver Island and in the Puget Sound43, as well as at an NSF cabled hydrophone located ~87 km west of the Oregon coast10, and through regional studies of sound levels worldwide (Table 2). However, from early 2021 to 2022, sound levels varied significantly across the sites, where OSP exhibited lower levels, apparently due to continued reduced levels of nearby ship traffic, while OCNMS shows similar and/or lower sound levels during 2021–2022 possibly due to anthropogenic activity. Nevertheless, each of the sites reviewed here returned to pre-pandemic sound levels by late 2020, and likely normal variability in levels in the subsequent years.

In comparison to previous societal events that curtailed economic activity, the reduced pandemic ocean sound levels observed during 2020 are not as substantial as observed in the Santa Barbara Channel during the 2007–2010 economic recession (~6 dB decrease)43, nor within the Bay of Fundy in the aftermath of September 11, 2001 (5–10 dB)33. While the Santa Barbara Channel hydrophone deployment depth was ~570 m deep, similar to the hydrophone depths used in our North Pacific study, the Bay of Fundy study was in much shallower waters (100–200 m), suggesting sound levels in the Bay of Fundy may be elevated in comparison. Nevertheless, these observed differences in levels likely reflect both the sociocultural impacts and the specific timeframes of each event, which spanned over three decades.

Methods

Instrumentation

Multiyear hydroacoustic data were sampled via passive acoustic, autonomous underwater hydrophone (AUH) systems moored in, or close to, the ocean sound channel at three northeast Pacific Ocean sites and one in the Arctic Ocean. The AUHs were moored at ~500 m water depth, with the exception of the Axial Seamount hydrophone, which was moored at 1525 m during 2018–2020, and 900 m from 2020 to 2021. The geographic locations, years data were collected, hydrophone depth, and sample rate at each of these sites are shown in Fig. 1 and Table 1. All hydrophones (AUH) used were designed and built by NOAA-PMEL.

Each AUH consists of a single, ceramic, omni-directional hydrophone element (model ITC-1032TM) with a nominal sensitivity of –192 dB re 1 V µPa−1 and a flat frequency response (±1 dB) between 10 Hz and 2 kHz. The hydrophone is attached to a filter/amplifier, clock, and processor (Compact Flash-2; Persistor Instruments, Inc.) which are all powered by an internal battery pack. The OSP, OCNMS and Beaufort instruments all continuously record at a sample rate of 5 kHz with a 2 kHz low-pass cutoff frequency, while the Axial Seamount AUHs recorded at 1 kHz with a 400 Hz low-pass. Incoming signals were conditioned by a pre-amplifier and pre-whitening filter to maximize the dynamic range of the 16-bit acoustic data logging system (see Haver et al.9, for additional details). In all analyses, the effect of the pre-whitening filter was removed to restore actual spectral levels.

Clock timing is regulated by a microprocessor-controlled, temperature-correcting crystal oscillator with an average time drift of 0.5–1.5 min year−1 (for multi-year deployments). The system electronics are housed in titanium or aluminum pressure cases appropriate for deep sound-channel deployments (400–900 m) in temperate to polar regions. This acoustic recording package is then deployed using a subsurface oceanographic mooring, consisting of a low friction line to reduce strumming and a syntactic foam float to suspend the recorder either within the sound channel, or several 10 s of meters off the seafloor. Arrays of these moorings can be efficiently deployed from a regional-class oceanographic vessel.

Acoustic data analysis

The hydrophone moorings at each site are on a 2-year recovery cycle, due to instrument battery life and availability of recovery vessels. The recovery schedule is staged across sites, which sometimes results in temporal coverage gaps. The first step in our data analysis was to calculate the multiyear long-term spectral averages (LTSAs) for each dataset at each of the four hydrophone mooring sites. The LTSAs elucidate long-term, broadband trends in sound levels as well as provide insights into the natural and anthropogenic sound sources driving ambient sound levels (Fig. 9). The spectrum levels were calculated in MATLAB® from raw binary files in 1 s spectral averages at a 1 Hz frequency resolution (1 Hz–2 kHz) and then averaged in hourly windows before conversion to decibels for efficient data analysis of the 4-year-long continuous data set.

Hourly averaged power spectral density levels (1 Hz narrow-band) were summed across 56–71 and 112–141 Hz to calculate sound levels in one-third octave bands (TOB), centered at 63 and 125 Hz, and then averaged over 24-h time windows to obtain daily values. The 63 and 125 Hz bands have been found to be accurate proxy frequency bands to estimate underwater radiated noise from commercial ship traffic22. In addition to daily mean decibel levels, a 14-day moving average was calculated for each site to smooth patterns over time.

Estimates of regional ship counts

To estimate the number of ships to compare to the passive acoustic data sampled in our broader observation area, we searched the Automated Identification System44 for all ships within a 222 × 222 km square area centered on the hydrophone mooring locations at Ocean Station Papa, Axial Seamount, Olympic Coast Sanctuary, and the Beaufort Sea (Table 3). The data provided from this website includes information on ship type (e.g., container versus fishing vessels), course over ground, direction and most importantly the daily presence of the vessel. We also searched the database from 2019 through 2023, a year longer than the acoustic data time period, to provide as comprehensive a view as possible of the regional ship traffic.

We used a square search area of 222 km per side to approximate a box of 2° in latitude (where 1°of latitude is ~111 km) and an equal distance in longitude. The goal was to maintain a similar search area size in latitude and longitude, given a degree of longitude varies in size from ~33 to 77 km between hydrophone sites. An area of this size should also allow for robust accounting of all nearby ships likely contributing to ocean sound levels, given that transmission loss at 222 km is ~47 and 23 dB (for spherical and cylindrical, respectively) in the 63–125 Hz band, and typical container ship noise levels43 are 170–190 dB. We consider both spherical and cylindrical spreading here because the hydrophone sites, with the exception of the Beaufort Sea, are located within temperate waters, and the sound channel is several hundred meters below the sea-surface. Thus, there is likely a significant amount of downward propagation of sound energy before entering the channel.

Analysis of Blue Whale vocalizations

Several recordings were also analyzed for blue whale vocalizations. Blue whale calls were detected using spectrogram correlation45 as implemented in the sound analysis package Ishmael 3.046. The Ishmael detection settings used were a spectrogram FFT size of 3.3 s (16,384 samples) with 50% overlap, a Hann window, a spectrogram equalization time of 10 s, a spectrogram correlation kernel ranging from 44.038 to 41.885 Hz over 15.93 s, a kernel bandwidth of 2 Hz, a detection averaging time 5 s, a detection threshold of 3.5 (unitless), and a maximum call duration 20 s.