NSOP is a repurposed ocean buoy (1.6 m diameter) with a central lifting eyelet (Fig. 1). The top of the buoy is 0.5 m above the sea surface. Mounted on top of the buoy are a line of sight, remotely operated winch (Warrior Winch, model C8000) and a gel battery (Haze, model HZY-S112-230). The winch feeds Kevlar rope through a block and tackle with a 3:1 ratio to reduce rope pay-out speed to ∼ 0.05 m s-1. The block and tackle is attached to the end of an outstretched arm 0.25 m from the outer edge of the buoy. The winch line is attached to an open frame (0.35 m diameter, 0.8 m height) with the capacity to house multiple sensors. Desired sampling depth is targeted using knowledge of the winch pay-out speed. Rope pay-out is then timed with a stopwatch. This approach only approximately regulates the sampling depth because (i) winch pay-out varies slightly depending on the amount of rope on the spool and (ii) variable horizontal current strength affects the vertical versus horizontal position of the sampling frame. To minimize horizontal movement of the sampling frame we attached a 10 kg weight to the base of the frame.

The primary sensor on the sampling frame is a small CTD (Valeport miniCTD) set to sample at a high frequency (> 1 Hz). Under calm conditions it is possible to sample as close as 0.1 m from the air–sea interface when the miniCTD and tubing are mounted near the top of the frame. Rougher conditions demand that the frame be kept deeper (∼ 0.5 m) as motion can momentarily bring the sensors and tubing out of the water. An emergency tag line was attached to the sampling frame in case the winch line failed. Seawater for trace gas analysis was pumped back to the ship at 3.5 L min-1 through a 50 m PVC hose (0.5 in inner diameter). A heavy-duty peristaltic pump (Watson Marlow, model 701IB/R) primed with water from the ship's underway supply was used to overcome the large hydraulic head (∼ 4 m). The open end of the tubing was located at the same depth as the miniCTD. Water arriving to the ship's laboratory was divided, with ∼ 3.0 L min-1 for flow-through analysis (e.g. equilibrator for trace gases) and ∼ 0.5 L min-1 for discrete samples (e.g. total alkalinity).

We assessed the depth resolution capability of NSOP at a particular depth by looking at pressure variations under calm conditions with a fixed amount of winch rope paid out. In calm to moderate conditions (< 2.5 m significant wave height) the amount of vertical movement indicated by the standard deviation (SD) in the depth is ±0.18 m (Fig. S1, Supplement). During four deployments in rough conditions (> 2.5 m significant wave height), the depth variability increased as the sampling frame was lowered (at 5 m, SD was ±0.275 m).

On a large research vessel such as the RRS Discovery, the deployment and recovery of NSOP requires close coordination between the bridge and three personnel on deck. The NSOP was always deployed while the ship was on station and not at the same time as other overboard deployments. Ship orientation during deployments was typically with bow into the wind but also accounted for swell and current direction/speed. The NSOP was lifted by the aft crane (Fig. 1). Once the NSOP had been lowered to the surface it was detached from the crane via a quick release. Two slack lines were looped through eyelets on the free-floating NSOP to maintain its position close to the ship. A third slack line was connected to the top of the buoy and passed through a block on a fully extended crane arm of 7 m to maintain this distance between NSOP and the ship. The slack lines successfully inhibited the tendency of NSOP to drift horizontally without disrupting its ability to ride the swell. The instrument frame acted like a sea anchor and minimized rotation of NSOP. A 4 m lifting strop used for recovery was connected to the lifting eyelet and loosely lashed to the aft slack line. During retrieval, the slack lines were hauled in and the crane and jib arms brought towards the ship to bring NSOP alongside. The lifting strop was then parted from the slack line and attached to the crane to lift NSOP back on deck. For additional photographs of a NSOP deployment and videos of NSOP during a deployment and in operation see Fig. S2 and videos.

Turbulence from the ship's propellers has the potential to mix the water column and destroy any near-surface gradients. The ship did not use the aft thrusters whenever conditions were suitable (mild sea state, weak currents and no local hazards). Keeping the NSOP away from the ship limited disruption of near-surface gradients by the thrusters and reduced the risk of line entanglement in the aft propellers. Our winch did not have a groove bar to feed the rope onto the winch drum, leading to an increased likelihood of snagging during spooling. To minimize snagging, the rope was manually fed onto the winch spool before deployments. Visual monitoring of the NSOP frame, slack lines and winch spool is important during deployment.

The NSOP has been successfully deployed in “moderate” sea states up to Beaufort force 5 (∼ 10 m s-1 wind speed and wave heights of ∼ 2.0 m). Deployment length typically varied from 1 to 3 h.

The NSOP can be used in two profiling modes: “continuous” and “discrete”. Continuous profiling maximizes vertical coverage and involves the winch continuously paying rope in and out at ∼ 0.05 m s-1. A complete down/up profile to 10 m can be conducted in approximately 7 min. Depth resolution during continuous profiling is determined by the measurement response time. Instruments with rapid response times such as the miniCTD temperature and conductivity sensors (0.15 and 0.09 s) have theoretical depth resolutions of 0.75 and 0.45 cm respectively. Actual depth resolution will also be affected by the sampling depth variability of the NSOP instrument frame. A measurement setup with a longer response time (such as for seawater CO2) requires a different approach (see Sect. 2.5).

During discrete profiling, the winch pays out a fixed amount of rope (typically 0.5 m) and the sampling frame is left at a fixed depth. After a fixed sampling period, more rope is paid out. The process is repeated down and then up such that a set of discrete depths are sampled in a “stepped” profile. The discrete profiling depth resolution is determined by the depth fluctuations when sampling at a fixed depth (see Sect. 2.1). Discrete profiles are a more appropriate approach for measurement systems with a longer response time. A discrete profile with 0.5 m steps down to 5 m and back to the surface using a 2.5 min sampling period takes about an hour. The sampling period at each depth and frequency/distribution of depths within the profile can be adjusted to suit sampling priorities.

The maximum deployment time is limited by the capacity of the winch battery. When under no load, the battery allows for approximately 3 h of operation in the continuous mode. Discrete profiling requires substantially less winch usage such that battery drainage is even less of a concern.

The CO2 measurement system (Fig. 2) is a modified version of the system described by Hales et al. (2004). Seawater from the NSOP inlet was passed through the equilibrator (see Sect. 2.3.1) at ∼ 3 L min-1 and the flow rate monitored (Cynergy ultrasonic flow meter, model UF25B). A compressed nitrogen gas supply, maintained at a constant flow rate of 100 mL min-1 (Bronkhurst mass flow controller, model F-201-CV-100) flows through the equilibrator in the opposite direction to the seawater flow. The gas has high water vapour content after equilibration and is dried (Permapure nafion dryer, model MD-110-48S-4). The dried sample then enters the analytical cell of a NDIR LI-COR 7000, which is protected with a 0.2 µm filter (Pall, Acro 50).

CO2 measurements at atmospheric pressure as recommended by Dickson et al. (2007) were not possible due to the nature of the experimental setup. The continuous gas flow through the system caused a small 0.4 kPa pressure increase in the LI-COR measurement cell; this was in good agreement with a similar observation by B. Hales (0.5 kpa > ambient pressure; personal communication, 2014). The elevated pressure was taken to be representative of the equilibrator pressure and was used to obtain the partial pressure of CO2 in the equilibrator (pCO2(eq)).

The LI-COR was calibrated using three CO2 standard gases before and after each NSOP deployment. The concentrations of the standard gases (BOC Ltd) were determined by referencing against US National Oceanic and Atmospheric Administration certified standards (244.91, 388.62, 444.40 ppm) in the laboratory. The seawater temperature at the entry and exit ports of the equilibrator was recorded at 1 Hz (Omega ultra-precise 1/10 DIN immersion RTD) using stackable microcontrollers (Tinkerforge master brick 2.1 and PTC bricklet). Equilibrator temperature probes and the miniCTD temperature sensor were calibrated before and after each cruise against an accurate reference sensor (Fluke, model 5616-12, ±0.011 ∘C) in a stable water bath (Fluke 7321).

DMS was measured with atmospheric pressure chemical ionization mass spectrometry (API-CIMS), using a system modified following Saltzman et al. (2009). Measurements were calibrated using an isotopic liquid standard of tri-deuterated DMS (see Bell et al., 2013 for details). Isotopic standard was injected at 120 µL min-1 into the 3 L min-1 seawater flow from NSOP before it entered the Liqui-Cel equilibrator. Compressed nitrogen gas was passed through the equilibrator in the counter direction to the seawater flow at 1 L min-1. The use of an internal standard meant that any incomplete equilibration of the ambient non-isotopic DMS was also true for the isotope. The gas stream exited the equilibrator and was dried (Permapure nafion dryer, model MD-110-48S-4) before entering the mass spectrometer for analysis. DMS was detected at m/z (mass/charge) 63 and the isotopic standard detected at m/z 66. The concentration of DMS was calculated using the ion signals and relevant flow rates (Bell et al., 2015). This approach has been shown to compare well with other analytical techniques for DMS (Royer et al., 2014; Walker et al., 2016).

We used different approaches to assess the delay between instantaneous miniCTD measurements and water arriving to the ship for analysis. The delay between seawater entering the inlet and reaching the equilibrator was calculated as 114 s using the internal volume of NSOP tubing (0.5 in inner diameter, 54 m length) and a seawater flow rate of 4.15 L min-1. Delay correlation analysis between the NSOP miniCTD temperature sensor and a second sensor positioned at the entrance to the equilibrator gives a similar delay of 112 s. Note that the total delay of the system is greater because it also includes the time that equilibrated gas takes to reach the LI-COR. We determined the total delay by transferring the seawater inlet quickly between two buckets with distinctly different CO2 concentrations and timing how long it took for the signal to be detected by the LI-COR (139 s; Fig. 4).

The response time of the NSOP setup was determined by simulating step changes in gas concentrations. A model fit to the exponential change in signal was used to estimate the response time (Fig. 4). We estimate the system response time (e-folding time) for CO2 as 24 s, which is slightly faster than the 34 s reported by Webb et al. (2016). The e-folding time in the DMS signal is estimated as 11 s, which is consistent with the rapid gas flow rate through the analytical system.

Continuous profiling with the CO2 system and a 24 s response time yields a depth resolution of 1.2 m, which is greater than the required resolution to assess near-surface gradients. DMS has a faster response time than CO2, but in continuous profiling mode this only translates to a depth resolution of 0.6 m, slightly less than the 1.2–2 m reported by Royer et al. (2014). A depth resolution of < 0.5 m was desired to capture upper ocean vertical gradients in CO2 and DMS so NSOP was operated in discrete profiling mode.

During discrete profiling, distinct sample depths were identified from the rapid changes in pressure during depth transitions. Data were binned into discrete depth bins using CTD pressure measurements. Trace gas data were assigned to depth bins after adjusting for the calculated transit time through the NSOP tubing (Sect. 2.5). CO2 data from the beginning (2 e-foldings + 15 s buffer = 63 s) and end (15 s buffer) of each depth bin was excluded from analysis to account for the response time of the system and the transition time between sample depths. The same approach was taken for DMS, where the faster response time resulted in a smaller portion of data excluded at the beginning of each depth bin (2 e-foldings + 15 s buffer = 37 s).

The CO2 mixing ratio (xCO2) measured in the LI-COR is converted to equilibrator fugacity (fCO2(eq)) using calibration standards, in situ seawater salinity, and the pressure and temperature in the equilibrator (SOP 5# Underway pCO2; Dickson et al., 2007). Vertical profiles of seawater CO2 fugacity (fCO2(sw)) are calculated using average equilibrator fugacity (fCO2(eq)), equilibrator temperature (T(eq)) and in situ seawater temperature (T(sw)) at each depth (Takahashi et al., 1993).

The NSOP setup enables vertical profiles of discrete seawater samples to be collected from upstream of the equilibrator, with a split in the tubing diverting ∼ 0.5 L min-1 into a sink. For example, discrete seawater samples (250 ml) have been successfully collected and analysed for Total Alkalinity (TA). Samples were collected and poisoned following best practice recommendations (SOP#1; Dickson et al., 2007). Bottle filling plus one overfill took ∼ 60 s. Start and end times were recorded so that collection depth could be retrospectively determined from the CTD pressure data. Analytical methods and an example depth profile are provided in the Supplement.

NSOP was deployed at 14:05 (UTC) on 30 July 2015. During the 6 h preceding deployment, the ship was on station and encountered persistently strong solar radiance (> 600 W m-2), mild winds (< 6 m s-1) and calm sea state (significant wave height < 1.6 m). This combination of low wind speeds and high irradiance (Fig. S5) is favourable for near-surface stratification (Donlon et al., 2002).

Figure 5 presents the time series data collected by NSOP for depth, temperature, salinity and fCO2(sw). Discrete profiling began at 14:05 (UTC) at 0.7 m depth, which was as close to the surface as the frame could be located without the possibility of breaking the surface. Depth bins were identified based on rapid depth transitions (Fig. 5a). Bottles were filled for discrete samples during the downcast. Profiling lasted 75 min and finished back at the surface at 15:20 (UTC). Seawater temperature was 16.61 ± 0.06 ∘C. At 14:20 (UTC), fCO2(atm) was 398 µatm and fCO2(sw) was 389 µatm at 0.67 m, meaning the ocean was undersaturated with respect to the atmosphere. The temperature and seawater CO2 were the expected magnitude for summer in the Celtic Sea (Frankignoulle and Borges, 2001). Salinity was homogeneous throughout the NSOP deployment, only varying by ±0.004.

Depth-binned salinity and temperature data did not show any significant variability (Fig. 6a). A slight temperature gradient was observed, with 0.15 ∘C difference between 5 m and the surface and a fairly constant reduction with depth (0.03 ∘C per metre). The temperature profile was similar for down- and upcasts, although some continued warming of surface waters was evident in the upcast. The temperature measured by NSOP at 5.15 m depth agrees well with the coincident temperature measured by the bow thermistor at 5.5 m (< 0.02 ∘C difference) (Fig. 6c). There is no evidence that the ship's thrusters/propellers disrupted the near-surface gradients.

We compare the NSOP temperature profile with thermistor readings from a series of Sea-Bird Scientific (SBE 56) sensors (0.3, 0.6, 1.5, 3.5 and 7 m depth) mounted on a nearby temperature chain moored ∼ 2.8 km away (49.403∘ N, -8.606∘ E) from the deployment site . The vertical profile implied by the NSOP deployment agrees with the mooring data (Fig. 6c), and corroborates the warming of the upper few metres of the ocean observed during the deployment. The agreement between these independent datasets suggests that it is unlikely that NSOP caused any significant localized warming of surface waters. The mean difference between NSOP temperature from discrete depths and the mooring sensors is 0.02 ∘C. The surface data from the NSOP upcast show less agreement with the mooring, with NSOP temperatures ∼ 0.05 ∘C lower than the 0.3 m and 0.6 m mooring sensors. During the profile the ship drifted ∼ 1 km from the start position of the profile and a further 0.2 km from the mooring. The small offset between the NSOP surface temperatures and the mooring may be driven by horizontal variability between the deployment and mooring locations. It is also possible that turbulence mixed warm surface waters down into cooler sub-surface layers. Turbulence could have been generated around the NSOP sampling frame or by an increase in wave-driven mixing when the significant wave height increased at ∼ 15:00 UTC (Fig. S4a).

Seawater density (Fig. 7a) was calculated using the salinity and temperature profile data (Fig. 6a, b) and the 1983 Unesco equation of state (Millero and Poisson, 1981). As expected, with little variation in the salinity, changes in the density profile are dominated by temperature. The down- and upcasts for CO2 show excellent agreement below 2.5 m. Surface water (< 2 m) CO2 is 2–4 µatm higher than at 5 m (Fig. 7b). Elevated surface CO2 could be explained by a sustained flux from the atmosphere into a near-surface stratified layer with inhibited deepwater exchange. Under this assumption a vertical gradient in seawater CO2 would need to be established shortly after the temperature gradient. A paired t test showed that the fCO2 measured in the surface bins on the downcast and upcast are were significantly different (p= < 0.001). The deepening of the surface stratified layer could explain the more homogeneous CO2 during the upcast. It is worth noting that in addition to physical processes, plankton trapped within the surface layer could also modify the surface CO2. Trace gas concentrations may also be different in the sea surface microlayer but sampling that close to the surface is beyond the capabilities of NSOP. Complementary measurements of the sea surface microlayer could be made using other state-of-the-art purpose-built sampling platforms such as the Sea Surface Scanner (Ribas-Ribas et al., 2017).

To assess measurement accuracy the NSOP Liqui-Cel CO2 system was compared against an independent CO2 system that had a showerhead equilibrator coupled to the ship's seawater supply pumped from 5.5 m below the sea surface (Hardman-Mountford et al., 2008; Kitidis et al., 2012). Technical issues meant that the underway CO2 system installed on the RRS Discovery was not functioning during the deployment detailed above. However, during a deployment on 19 July 2015, the fCO2(sw) measured by NSOP at 5 m agreed well with independent measurements from the underway system, with difference of 1.7 ± 4.18 µatm. The agreement between the two systems is in line with previous intercomparisons (Ribas-Ribas et al., 2014; Körtzinger et al., 2000).

DMS profiles were collected on a small research vessel on 15 April 2014. The NSOP was deployed within the Plymouth Sound at 12:00 UTC and recovered 95 min later (Fig. 8). In the sheltered environment behind the breakwater the standard deviation in depth was ±0.10 m, smaller than observed during open ocean profiles. Seawater temperature and salinity demonstrate clear structure, with lower temperatures and higher salinities associated with sub-surface water. Two river estuaries (Plym and Tamar) converge and flow out to the open ocean through the Plymouth Sound. We likely observed a freshwater surface lens that was protected from wave-driven mixing and had been warmed over the course of the day. We used a different miniCTD during this deployment and were thus also able to collect fluorescence data (Fig. 8d).

Temperature profiles (Fig. 9a) show a sharp discontinuity in the downcast at ∼ 5 m whereas in the upcast the thermocline had shoaled to ∼ 3.5 m. The salinity profiles suggest similar mixing depths to the temperature profiles, with lower salinity water at the surface (Fig. 9b). The increase in fluorescence with depth (Fig. 9c) is either due to reductions in chlorophyll concentration close to the sea surface or because of quenching of the phytoplankton photosynthetic apparatus, which is often observed in surface waters that experience strong irradiance (Sackmann et al., 2008; Biermann et al., 2015). DMS concentrations reduce steadily with depth (Fig. 9d), which is likely explained by changes in DMS production and consumption rates by the biological community (Galí et al., 2013). The DMS profiles from the upcast and the downcast are very similar, with the largest difference at the very surface. A large difference in the surface-most data point can also be seen in the temperature data, and may reflect mixing with sub-surface waters due to the motion of NSOP or short timescale variations in the physical environment.