Acoustic ducting by shelf water streamers at the New England shelfbreak

: Greater sound speed variability has been observed at the New England shelfbreak due to a greater inﬂuence from the Gulf Stream with increased meander amplitudes and frequency of Warm Core Ring (WCR) generation. Consequently, underwater sound propagation in the area also becomes more variable. This paper presents ﬁeld observations of an acoustic near-surface ducting condition induced by shelf water streamers that are related to WCRs. The ﬁeld observations also reveal the subsequent disappearance of the streamer duct due to the passage of a WCR ﬁlament. These two water column conditions are investigated with sound propagation measurements and numerical simulations.

Acoustic ducting by shelf water streamers at the New England shelfbreak

Introduction
The New England shelfbreak (NESB) is a dynamic physical oceanographic region with strong temporal and spatial variations of water temperature and salinity, which can produce significant horizontal and vertical gradients in sound speed (Sperry et al., 2003).Specifically, the shelfbreak front is a region where continental shelf and slope water masses abut.The NESB region has reported warming seas in recent years that can be partially attributed to the increase in Gulf Stream (GS) instabilities forming mesoscale eddies or Warm Core Rings (WCRs) (Chen et al., 2020;Gangopadhyay et al., 2019), which significantly interact with local bathymetry and seasonal shelf water masses.These interactions can entrain submesoscale cold water filaments known as shelf water streamers across shelfbreak front boundaries for tens of kilometers offshore and last for months (Zhang et al., 2023).Shelf streamers have a surface expression and are characterized by fresher cold water from the shelf, which can create acoustic surface ducts with warmer saltier slope water underneath.Therefore, mesoscale and submesoscale dynamics significantly affect acoustic propagation conditions at the shelfbreak.
Understanding water column and seabed variability is critical for studying acoustic propagation at shelfbreak environments.A previous experiment conducted in the NESB area in 1996, the PRIMER experiment (Lynch et al., 2003;Sperry et al., 2003), revealed the close connection between acoustic variations and complicated physical oceanographic conditions.More recently, the New England Shelf Break Acoustics (NESBA) Experiments conducted in 2021 explored the acoustic sensitivity of shelfbreak fronts and complex shelf and upper slope bathymetry (DeCourcy et al., 2022;Ozanich et al., 2022), previously studied primarily via theoretical analysis and numerical simulations (Lin and Lynch, 2012;Decourcy et al., 2019).In addition to shelfbreak fronts, water column sound speed fields can also be influenced by other physical oceanographic processes, such as frontal intrusions and nonlinear internal waves.Marine geological features are equally important when quantifying the physical environment for sound propagation.Specifically, complex bathymetric features such as canyons can cause three-dimensional (3D) sound focusing and scattering, and sediment heterogeneities in the seabed that are highly frequency dependent can result in attenuation and scattering.Sound speed gradients and stratigraphy layering present in the seabed can add to the complexity of underwater sound propagation, and emphasizes the importance of understanding the interrelatedness of both physical oceanographic and geological variables for studying acoustics in shelfbreak regions.
This paper is structured as follows.The in situ measurements showing the near-surface acoustic duct induced by a shelf water streamer and the duct departure due to a WCR filament are presented in Sec. 2. Computational model descriptions and results are discussed in Sec. 3. The paper is then summarized in Sec. 4 with future research directions.a) Author to whom correspondence should be addressed.

In situ measurements
The NESBA Experiments were conducted approximately 178 km south of Woods Hole, MA, from April 2021 to June 2021.A total of 16 moorings were deployed by the Woods Hole Oceanographic Institution (WHOI) in an acoustic network extending from the shelf to the upper continental slope with a bottom depth 100 m down to 900 m to observe acoustic propagation and shelfbreak front dynamics [Fig.1(a)].The mooring observations were accompanied with satellite and shipboard measurements as shown in Fig. 1, where sea surface temperature (SST) images were acquired from the Advanced Very High Resolution Radiometer (AVHRR) server (https://tds.maracoos.org/thredds/catalog/REALTIME-SST.html) at 1 km horizontal resolution within the NESBA mooring network from May 15 to May 20, 2021.These SST observations highlight the surface evolution of water masses and more specifically the presence of a shelf water streamer.Directly eastward of the streamer was another WCR.The SST highlights a filament from the WCR extending near the NESBA mooring network.By May 17 the WCR filament started to advance into the network and by May 19 warmer saltier ring water fully occupied the NESBA network.The satellite observation of these water masses during this 6-day period was supplemented with shipboard lowered and towed conductivity, temperature and depth (CTD) measurements, which verified surface observations with subsurface measurements of shelf streamer, slope, and WCR water.
This study focuses on an upslope acoustic propagation path from the west node in the mooring network [labeled "TM3" in Fig. 1  shelf water streamer presence in the top 50-75 m that produced a near-surface sound duct.The second condition was caused by a WCR filament that changed the near-surface sound speed characteristics resulting in the duct absence along the transmission path.The acoustic measurements showing the two corresponding sound propagation conditions are discussed next.
Acoustic transmissions at TM3 consisted of a sound source at 64 m depth that transmitted 30 4-s long down sweep pulses from 925 to 575 Hz with 6-s gaps between each transmission, repeating every 15 min on even hours.The sound source system used a Chip Scale Atomic Clock (CSAC) to maintain a highly precise time base.This study focused on the received signals at the SHARK VLA, which consisted of 16 hydrophones covering the water column from 64.8 to 143.5 m depth, with 3.75 meter spacing between the top eight hydrophones and 7.5-meter spacing between the bottom eight hydrophones.The array data acquisition system also used a CSAC and sampled at 19531.25 Hz.A matched filter was applied to the acoustic signals to increase signal-to-noise ratio (SNR) for data evaluation with a compressed pulse length of approximately 2.9 ms.
Figures 2(c) and 2(d) show one complete transmission cycle of the acoustic data (5-min long) received by the top hydrophone of the SHARK VLA at depth 64.8 m on May 18 and 20.The received source signals on May 18 show near-surface duct arrivals at approximately 0.1 s later than the first arrival when the streamer was present.This was due to a slower sound speed in the duct.The spread of the near-surface duct arrivals was variable, even from consecutive pings, but always maintained the acoustic arrival delay ducting signature.Early arrivals with fewer bottom bounces maintained high coherence over the 5-min transmission period.However, later arrivals of higher incident angles at the seafloor were affected more by bottom reflections, which were coupled with water column variations that can change the incident angles and therefore yielding scattering and a loss of temporal signal coherence.By May 20, the surface duct induced by the shelf water streamer was no longer present due to the arrival of a WCR filament that affected the sound propagation along the transmission path.Coupled with the WCR water were high amplitude bottom bounce early acoustic arrivals.The transition from ducted early arrivals to stable bottom bounce arrivals occurred quickly as the WCR filament arrived and water column conditions changed, while later arrivals were consistently affected more by seafloor and surface scattering resulting in phase perturbation and incoherency.

Computational models
To better interpret the acoustic in situ observations, numerical simulations of broadband (925-525 Hz) sound propagation were conducted using a parabolic equation (PE) model (Collins, 1993) with realistic environmental parameters representing (1) the streamer condition yielding a near-surface duct and (2) the WCR filament condition resulting in the absence of the duct.Due to the physical oceanographic and geological environment jointly affecting underwater sound propagation and scattering, it is important to have reliable environmental model conditions.To characterize the water column conditions in the acoustic model simulations, sound speed profiles based on 1-day time average temperature and salinity measurements during the acoustic observation period were used [start of 24-h periods denoted by vertical lines in Figs.2(a) and 2(c)].The 25 m bathymetry grid in the model was created using shipboard multibeam measurements along the propagation path.A seabed model was generated using an acoustic dataset collected during the Seabed Characterization Experiment (SBCEX) in May 2022 in the NESBA experiment area yielding seabed attenuation values along the 200 m depth contour on the shelfbreak (Knobles et al., 2023).
The acoustic models consisted of range-independent time-averaged environmental conditions along a defined transmission path to study the streamer-induced near-surface sound duct and the duct absence due to a WCR filament.Although incoherent averaging of arrival patterns results in lower temporal resolution, multipath wavefront patterns are preserved and can be used to cross-examine model results with measured data in this study.The near-surface duct condition is apparent in the model with arrival patterns resembling the data, specifically the approximate 0.1 s pulse delay and the pulse energy distribution over depth and time.The measured later arrivals with multipath wavefront patterns dominated by sub bottom layering structure, established from independent SBCEX data, are also reproduced in the model.
The simulation results highlight major differences in propagation between the streamer near-surface duct and the absence of the duct with the arrival of the WCR filament, identified in the data.The model environment neglects the roughness of the sea surface and seafloor, which can lead to fine scale mismatch between the data and model.The water column temporal and spatial averaging used to create environmental parameters in the acoustic model can also limit detailed acoustic data-model agreement.Further investigation is needed to refine the temporal and spatial variabilities captured in the acoustic models for each water column condition presented here.Short temporal oceanographic variability can result in early arrival spread during the near-surface duct period with relatively unstable quickly changing characteristics when compared to the WCR filament period that was dominated by more stable bottom bounces.The ducted arrival variability is also hypothesized to be affected by fine vertical scale sound speed minima resulting in multiple micro-ducts in the water column, which were not considered in the scope of this study.Ducting characteristics and quantification over greater temporal and spatial scales can also be explored, such as the frequency dependence over the entire duration of the streamer's presence on the shelfbreak.Thus, mesoscale and submesoscale dynamics should be considered in shelfbreak environments when examining large and small acoustic propagation variability in both time and space.

Summary
This study presents observations of an acoustic near-surface duct induced by a shelf water streamer and the duct departure related to a WCR on the NESB.The numerical models have effectively been used to interpret acoustic observations in the field for two water column conditions, linking acoustic propagation with oceanographic mesoscale and submesoscale dynamics.This study highlights near-surface duct arrivals that were more dependent on water column processes, while later pulse arrivals were more closely tied to seafloor properties and structure on the NESB.Fine scale water column and seafloor bottom effects can be explored for future study to refine the presented acoustic models, and to unravel joint effects on the propagation during the presented water column conditions.Increased Gulf Stream variability and WCR occurrence will likely result in more intermittent ducting conditions at the NESB.The potential application of this study can be extended to other shelfbreak areas with increased environmental variability that is coupled with rapid changes in the acoustic environment.
(a)] to the center mooring node [labeled "SHARK" in Fig. 1(a)].The TM3 mooring (located at 40 2.1902 0 N, 71 0.9987 0 W) was deployed at bottom depth 260 m and contained an acoustic source at 64 m depth, and the SHARK mooring (located at 40 5.4172 0 N, 70 51.08820 W) was a L-shape hydrophone array consisting of a vertical line array (VLA) and horizontal line array (HLA) at 148 m bottom depth.Both TM3 and SHARK moorings were equipped with temperature, pressure, and salinity sensors sampling every 30 s.The TM3 and SHARK temperature profiles [Figs.2(a) and 2(b)] characterize two different water column conditions in the period of May 15 through May 21, 2021 occurring at both mooring locations.The primary driving mechanism of the first water column condition was the

Fig. 1 .
Fig. 1.(a) The color image shows sea surface temperature from the Advanced Very High Resolution Radiometer (AVHRR) server (https:// tds.maracoos.org/thredds/catalog/REALTIME-SST.html) on May 15, 2021, at and around the NESBA mooring network (red nodes) with the streamer, cool surface shelf water extending offshore (dark blue), and an adjacent WCR (yellow).(b) The lower left panels are time lapsed snapshots zoomed in on the NESBA mooring network during the period of May 15 to May 20, 2021.(c) The temperature and salinity, in practical salinity units (PSU), plot of shelfbreak water masses from May 15 to May 20 from CTD measurements reveal different water masses, which include the shelf streamer (<34 PSU), slope water (35-35.5PSU), the shelfbreak front (SBF) (>34 and <35 PSU), and the warm and salty (>35.5 PSU) WCR water (labeled accordingly).(d) The lower right panels show the locations of the shipboard lowered CTD (white) and towed CTD (pink).

Fig. 2 .
Fig. 2. Panels (a) and (b) show water column temperature time series from sensors on the TM3 and SHARK mooring.Before the WCR arrived at the mooring locations, there was a strong near-surface duct when the surface streamer was present.Red vertical lines indicate the start of the two 24-h periods used to calculate average water column sound speed profiles for the numerical models investigating the streamer acoustic ducting and the WCR ducting absence.Panels (c) and (d) show the TM3 sound source signal arrivals (after matched filtering) on the top hydrophone of the SHARK VLA at 64.8 m depth during the streamer period (c) and the WCR arrival period (d).
Figure 3(a) shows the averaged sound speed profile and modeled transmission loss (TL) with strongly ducted sound along the source depth across the 15 km range, while Fig. 3(b) shows TL with the duct no longer present due to the WCR filament.With the presence or absence of the duct, both TL plots are suggestive of increased bottom loss within the first 5 km range.The modelled pulse arrivals [Figs.3(e) and 3(f)] agree well with the measured acoustic array data [Figs.3(c) and 3(d)] collected in the lower part of the water column.The measured arrivals are 5-min incoherent averages that smooth out temporal variations of the acoustic signal caused by fine scale water column and boundary scattering.

Fig. 3 .
Fig. 3. Panels (a) and (b) show the PE model simulation results for streamer near-surface ducting condition and WCR filament arrival condition.The curve plots on the left depict averaged sound speed profiles during each water column condition, and the black dots indicate the source depth on TM3.The upper right image plots show transmission loss at 725 Hz (the center frequency of the transmitted signals in the experiment).Panels (c) and (d) depict incoherent average pulse compressed signals, and panels (e) and (f) are the broadband model results, where the white horizontal line indicates the top of the SHARK VLA.