Acoustic water column data were collected throughout the Arctic Ocean during two expeditions with Swedish Icebreaker (IB) Oden: Leg 2 of the Swedish–Russian–US Arctic Ocean Investigation of Climate–Cryosphere–Carbon Interactions 2014 Expedition (SWERUS-C3) and the Arctic Ocean 2016 Expedition (AO2016).

Leg 2 of SWERUS-C3 departed on 20 August 2014 from Barrow, Alaska, and ended on 4 October in Tromsø, Norway. The expedition covered mainly the shallow areas of the East Siberian Sea continental shelf and shelf slope (Fig. 1). The median water depth of the 78 CTD stations investigated from SWERUS-C3 is 340 m. The hydrography of this area can be characterized as dynamic and seasonally variable as it is influenced by large river runoff, coastally trapped waves, ice formation and melting, and brine rejection in coastal polynyas.

The AO2016 expedition took place between 8 August and 19 September 2016, departing from and returning to Svalbard (Fig. 1). One specific research goal during AO2016 was dedicated to investigating the possibility to detect and map thermohaline stratification using a mid-water wideband echo sounder. The cruise track covered mainly the central Arctic Ocean and the median water depth of 24 CTD stations investigated is 4000 m (Fig. 1). Together, the SWERUS-C3 and AO2016 expeditions spanned much of the breadth and depth of the Arctic Ocean and provided wideband acoustic data in a variety of oceanographic settings.

The wideband water column backscatter data presented here were collected with a Simrad EK80 split-beam scientific echo sounder (SBES) installed on IB Oden. The system was operated continuously during both the SWERUS-C3 and AO2016 expeditions.

The SBES consisted of a Simrad EK80 wideband transceiver transmitting through a standard Simrad ES18-11 transducer installed in the “ice knife” near the bow of the vessel and protected by an ice window. This transducer model is widely installed in fishery research vessels, typically operating at 18 kHz with a -3 dB beamwidth of 11∘. In 2014, the transducer model was tested with a Simrad EK80 wideband transceiver and determined to have a useable two-way frequency response over 15–25 kHz. Thus, a frequency range of 15–25 kHz was used throughout the EK80 data collection period on IB Oden.

Transmit power was maintained at the maximum setting of 2000 W to compensate for losses through the ice protection window and improve signal-to-noise (SNR) characteristics, especially during noisy hull–ice interactions. Transmission pulse lengths were adjusted over a range of 1–8 ms in an effort to minimize the extent of autocorrelation sidelobes (sidelobes are typically minimized with shorter pulses) while maximizing the SNR (better with longer pulses). All EK80 operation was controlled and monitored around the clock using the Simrad user interface to adjust pulse length and range-recording duration. Data were logged in the Simrad .raw format.

Position and attitude information were provided to the echo sounder as an integrated solution by a Seapath Seatex 330 GPS/GLONASS navigation and motion reference system. Vessel motion was minimal (typically less than 1∘ pitch and roll in the data presented here) and thus does not appreciably affect the observations of horizontally oriented backscattering layers occupying broad portions of the beam.

During the AO2016 expedition, a small delay was applied to the EK80 transmit–receive cycle trigger in order to avoid transmission interference from the two other echo sounding systems (Kongsberg EM122 12 kHz multibeam and SBP120 2–7 kHz sub-bottom profiler) in the earliest portion of the EK80 receive cycle corresponding to the upper water column region of interest.

The dataset collected with the EK80 was match filtered with an ideal replica signal using a MATLAB software package provided by the system manufacturer, Kongsberg Maritime (Lars Anderson, personal communication, 2014). After match filtering, ship-related noise was found within the signal band. A band-pass filter with 16 and 22 kHz cutoff frequencies was applied to the data to exclude the noise.

Sound speed profiles were calculated from CTD-derived temperature, salinity, and pressure data using the International Thermodynamic Equation of Seawater (IOC, SCOR and IAPSO, 2010). Ranges from the transducer were then calculated using the cumulative travel times through sound speed profile layers based on the nearest (in time) CTD profile. These ranges were then converted to depths by compensating for the transducer location relative to the static waterline on IB Oden and the heave of the vessel.

The EK80 was calibrated onboard the Oden on 1 September 2015, following a standard method described by Demer et al. (2015). A 64 mm copper sphere of known acoustic properties was suspended on a monofilament line and moved through the SBES field of view. The calibration data were collected in relatively calm seas and atmospheric conditions while the Oden drifted. All propulsion systems were secured during the calibration procedure in order to reduce noise in the water column. A CTD profile was collected immediately before calibration operations.

Utilizing a calibration sphere target strength model based on the work by Faran (1951) and MacLennan (1981) (MATLAB software; Dezhang Chu, personal communication, 2015), a calibration offset (C=8.5 dB, averaged over the transducer beam width) was calculated using a temperature of 0 ∘C and a salinity of 34.5 at the sphere depth of approximately 80 m. This calibration offset represents the difference between the nominal target strength (TS) observed by the EK80, as predicted after match filtering, and the modeled TS of the calibration sphere. The offset is then applied to subsequent measurements of TS, yielding calibrated TS results for the EK80 datasets.

The TS of an ideally smooth layer is a function of both the reflection coefficient (R) and the ensonified area (A). Here, we assume that A is limited by the width of the EK80 beam (rather than the length of the pulse) such that A can be estimated as A(z)=π(tan⁡φz)2, where φ is half the beam width and z is the depth in the sonar reference frame. Following Lurton and Leviandier (2010) the TS for a layer at depth z, with reflection coefficient R, can then be estimated as TS(z)=20log⁡10R+10log⁡10(A(z)). For our estimates of observed R, we simply invert the above equation to solve for R: R=A-1/210TS/20, where TS is the calibrated acoustic backscatter observation from the EK80.

CTD data were collected with a Sea-Bird 911 equipped with dual Sea-Bird temperature (SBE 3) and conductivity (SBE 04C) sensors. The CTD data files were post-processed with SBE Data Processing software, version 7.26.6 (available at http://www.seabird.com/software). The alignment parameter was tuned following the suggested method described in the SBE Data Processing manual (available at http://www.seabird.com/software). All CTD data presented are averaged in 10 cm vertical bins.

The reflection coefficient from CTD data (RCTD) was calculated through RCTD(i)=η(i)-η(i-1)η(i)+η(i-1), where each element i has a corresponding depth z(i), the depth of RCTD(i) is the average of z(i-1) and z(i), and η is the acoustic impedance given by η(z)=V(z)ρ(z), where V is the sound speed and ρ the seawater density. The accuracies of the pressure, conductivity, and temperature sensors are 0.0015 %, 0.0003 S m-1, and 0.001 ∘C, respectively. All conversions (salinity, density, and sound speed) were made according to the International Thermodynamic Equation of Seawater (IOC, SCOR and IAPSO, 2010).

To determine the MLD, we apply the method presented in de Boyer Montégut et al. (2004) in which successively deeper data points in each of the CTD potential temperature profiles are examined until one is found with a potential temperature value differing from the value at the 10 m reference depth by more than the threshold value (ΔT) of ±0.2 ∘C. Using this approach, the MLD is then assumed to be at least 10 m deep, and any shallower well-mixed sections in the water column are not taken into consideration (de Boyer Montegut et al., 2004). The reference depth applied by de Boyer Montegut et al. (2004) was chosen to avoid the diurnal variability of the mixing layer (typically found at depths < 10 m) while keeping the longer-term variability of the mixed layer.

The typical summer MLD of the Arctic Ocean is ∼ 20 m (Steele et al., 2008). By applying a density threshold method for determining the MLD, Toole et al. (2010) reported for the central Canada Basin an average summer MLD of 16 m and an average winter MLD of 24 m. The shallower mean MLD in the SWERUS-C3 data is consistent with the large river runoff into the Siberian shelf seas, which should lead to a generally shallower mixed layer compared to the AO2016 data from the central parts of the Arctic Ocean (Large et al., 1994). Given the dynamic nature of the more coastal Leg 2 SWERUS-C3 cruise track compared to the open-ocean-dominated AO2016 cruise track, we were expecting larger MLD variability in the SWERUS-C3 data. We cannot see such a tendency in our data (Table 2), but again the basis of the statistics is rather poor.

In general, MLD variations between the different regions of the Arctic Ocean covered in this study match well with mean Arctic Ocean MLD based on other field observations (Ilıcak et al., 2016; Peralta-Ferriz and Woodgate, 2015), with shallow MLDs along the East Siberian Sea, slightly deeper MLDs in the Canada Basin, and the deepest MLDs in the central Arctic Ocean. As the emphasis of this paper is mainly on the acoustic method rather than the actual MLD observations, we are hesitant to draw any conclusions based on the MLD statistics presented in Table 2, especially when considering the small number of observations on which the statistics are based.

With the acoustic method we can observe the MLD at a horizontal resolution far exceeding alternative in situ methods, such as CTD profiles. The acoustic method enables the study of internal waves propagating on the layer interface at the base of the mixed layer (left inset, Fig. 2a). Internal waves are a ubiquitous phenomenon in the ocean and drive vertical mixing that is important for the global ocean circulation and primary production (Garret and Munk, 1979; Denman and Gargett, 1983; Munk and Wunsch, 1998). Stranne et al. (2017) observed internal waves that caused vertical undulations of the stratification down to depths of about 300 m. While these deeper internal waves were clearly not excited by the vessel, we cannot exclude the possibility that some of the vertical undulations of the MLD seen here are due to near-surface internal waves generated by the Oden (Nansen, 1905).

The recording duration of the EK80 was set to observe the full water column, resulting in a ping rate of around 0.1 ping s-1 when synchronized with the multibeam echo sounder in deep water (i.e., ping rate is limited by recording range on the outer swath, which can be more than twice the water depth). The ping rate can be set much higher (up to several pings per second) in shallow water or if only data from the shallow part of the water column are to be collected. In our data the MLD is clearly visible while drifting and steaming, but the quality of the data underway would benefit from a higher ping rate; specifically, the highest-frequency temporal and/or spatial variations in MLD are likely undersampled at this lower ping rate while the vessel is moving (right inset Fig. 2a).

The cruise track of the SWERUS-C3 expedition during Leg 2 covers mainly the shallow areas of the East Siberian Sea shelf and shelf slope, an area that is heavily influenced by river runoff (e.g., from the Lena River). The freshwater input (or negative buoyancy flux) to the coastal waters leads to a generally shallower MLD (Large et al., 1994). This is clearly manifested in our data in which the average MLD from the shelf-dominated SWERUS-C3 cruise is approximately two-thirds that of the sea ice-covered, deep-basin-dominated AO2016 cruise (Table 2).

The deep depth limit for detecting ocean stratification with this particular EK80 setup appears to be around 300 m (Stranne et al., 2017), while the shallow depth limit depends on the draft of the hull-mounted transducer and the pulse length. On the Oden, the EK80 transducer is mounted at a draft of 7 m and, depending on pulse length, we generally observe useful data starting at 7.5–12 m of depth from the surface (0.5–5 m from the transducer; Fig. 4d). The amount of data lost at the upper boundary is reduced with shorter pulse length (Fig. 4d); these data also show the better range resolution obtained with a shorter pulse length (Fig. 5), but there is a serious trade-off in terms of reduced SNR (Fig. 4a). More data are needed in order to determine the optimal pulse length for EK80 MLD detection as it also depends on region and platform.

Due to ship draft and the data loss at very close range from the transducer, the shallow MLDs seen in some of the SWERUS-C3 CTD profiles are sometimes difficult to detect acoustically with the EK80 (Fig. S4). This is the most common factor explaining more than 50 % of the failures to acoustically detect the MLD during SWERUS-C3.

In the example shown in Fig. 2, the reflections are likely stemming from the impedance contrast from the ocean stratification alone; this is supported by the close match between the theoretical reflection coefficient calculated from the CTD data and the reflection coefficient derived from the calibrated acoustic backscatter data. This agreement among reflection coefficients is consistent with observations of deeper thermohaline staircase stratification from the central Arctic Ocean presented in Stranne et al. (2017). In the SWERUS-C3 data, biological scatterers are generally identified at CTD stations closer to the coast. Biological scattering can potentially obscure the reflections from the MLD boundary (Fig. 6); at other times, the distribution of biological scatterers may coincide with the ocean stratification and enhance the layer reflections.

At the time of the SWERUS-C3 expedition, we did not yet realize that the EK80 was capable of MLD detection and accordingly nothing was done to optimize the performance of the EK80 to detect ocean stratification in 2014. At four of the SWERUS-C3 CTD stations, the MLD is obscured by noise from an unknown source (Fig. S5) but the source was not identified and no actions were taken to reduce it. This type of noise did not occur in the acoustic data from the later AO2016 cruise.