Medusa-Aqua system: simultaneous measurement and evaluation of novel potential halogenated transient tracers HCFCs, HFCs and PFCs in the ocean

This study evaluates the potential usefulness of the halogenated compounds HCFC-22, HCFC-141b, HCFC-142b, HFC-134a, HFC-125, HFC-23, PFC-14 and PFC-116 as the time-dependent oceanographic transient tracers in order to better constrain ocean ventilation processes. We collected seawater samples and improved on an established analytical technique, the Medusa-Aqua system, to simultaneous measure them, and estimate their stability in 20 seawater following previous work on the atmospheric history and solubility. HCFC-22, HCFC-141b, HCFC-142b, HFC-134a and HFC-125 have been measured in profiles in the Mediterranean Sea for the first time. We estimated the historic surface saturation anomalies of transient tracers in the Mediterranean Sea by evaluating the historic record. Their stability in seawater was estimated by analysis of their ocean partial lifetimes, seawater surface saturations and concentrations compared to CFC-12 measurements by a well-established technique. Of the 25 investigated compounds, HCFC-141b was found to be the most promising transient tracer in the ocean; it fulfills several essential requirements by virtue of well-documented atmospheric history, established seawater solubility, inertness in seawater and feasible measurements and indication of conservative behavior in seawater by having mean ages in agreement to be expected from both CFC-12 and SF6 observations. However, more information on degradation is needed to further identify its stability in seawater, and HCFC-141b has restrictions on production and 30 consumption imposed by the Montreal Protocol leading to its decreasing atmospheric mole fractions since 2017. The most potential oceanic transient tracers were PFC-14 and PFC-116 due to their stability in seawater, the long and well-documented atmospheric concentrations histories and constructed seawater solubility functions, although the low solubility in seawater creates challenging measurement conditions (i.e. low concentration). Measurements of PFCs can be potentially improved by modifying the Medusa-Aqua analytical system. With the exception of 35 https://doi.org/10.5194/os-2019-101 Preprint. Discussion started: 11 October 2019 c © Author(s) 2019. CC BY 4.0 License.

ocean's oxygen supply. One possible method to quantitatively describe these processes are based on transient tracer measurements.

Why do we look for new transient tracers?
Transient tracers include chronological transient tracers such as dichlorodifluoromethane (CFC-12), trichlorofluoromethane (CFC-11) and sulfur hexafluoride (SF6), and radioactive transient tracers such as  Helium ( 3 H-3 He), Argon-39 ( 39 Ar), and Carbon-14 ( 14 C), although 39 Ar is assumed to be in steady-state and cannot be regarded as a transient racer in the true meaning of the word. They have been used as oceanic transient tracers to study oceanic processes, such as ventilation, mixing, and circulation processes. CFC-12 and CFC-11 have been used since the 1980s, whereas SF6 has only been used since the late 1990s. These compounds are stable in seawater under most circumstances; their seawater solubility functions are well-established (Warner and Weiss, 1985;Bullister et al., 2002) 10 and their historical atmospheric concentrations over time are known (Walker et al., 2000;Bullister, 2015). However, the industrial uses of CFC-12 and CFC-11 were phased out as a result of the implementation of the 1987 Montreal Protocol on Substances that Deplete the Ozone Layer designed to curtail the degradation of the Earth's ozone layer.
Therefore, the atmospheric concentrations of CFC-12 and CFC-11 have decreased since the early 2000s and the early 1990s, respectively (Bullister, 2015), which has reduced their usefulness as oceanographic transient tracers for recently 15 ventilated water masses. Consequently, SF6 has been added to the suite of commonly measured oceanic transient tracers (Tanhua et al., 2004;Bullister et al., 2006) as it is an inert gas whose atmospheric abundance is increasing. Due to its very high global warming potential, some local restrictions on the production and use of SF6 has been imposed since 2006. However, the concentrations of SF6 in the atmosphere is still increasing, partly due to its long atmospheric lifetime. CFC-12, SF6, and CFC-11 can readily be measured onboard research vessels at a reasonable rate from one 20 seawater sample. 1,1,2-trichloro-1,2,2-trifluoroethane (CFC-113) and carbon tetrachloride (CCl4) have previously been used as oceanic transient tracers, but have now been largely discarded since they have been found to be degraded in warm waters  as well as in low oxygen waters (Wallace and Krysell, 1989;Huhn et al., 2001).
The radioactive isotope 39 Ar is in many ways an ideal tracer for ocean circulation for older water masses, but its use has been impeded by difficult analytics. However, recent technological advancements have increased the feasibility of 25 oceanic 39 Ar observations (Lu et al., 2014;Ebser et al., 2018). With the constraints of the weak signal of 3 H and the decreasing atmospheric mole fraction of CFC-12, presently only SF6 is a relatively reliable transient tracer in the seawater timescale less than 100 years (Fig. 1). Since a combination of multiple transient tracers is needed to constrain ocean ventilation (Stöven and Tanhua, 2014;Holzer et al., 2018), it is useful to explore novel transient tracers with monotonically changing input functions for a better understanding of ventilation and mixing processes in the ocean. 30

Potential alternative transient tracers
There are a few general requirements for a transient tracer: 1) well-quantified sources and sinks, 2) no (or well known) natural background, 3) large dynamic range, 4) feasible measurement techniques, and 5) non-reactive and stable in seawater. In the previous work (Li et al., 2019), we focused on points 1, 2 and 3 for these potential alternative oceanographic transient tracers: hydrochlorofluorocarbons (HCFCs) such as 35 hydrofluorocarbons (HFCs) such as HFC-134a, HFC-125 and HFC-23 and perfluorocarbons (PFCs) such as PFC-14 (CF4) and PFC-116. The atmospheric abundances of most HCFCs, HFCs, and PFCs are increasing. Here we call the potential chronological transient tracers , and PFC-116 as the "Medusa tracers", from the analytical instrument used to measure them. Fortunately, the different atmospheric histories of the potential alternative transient tracers (Li et al., 2019) allow us to explore one or several compounds to replace or supplement the established transient tracers (Fig. 1). 5

Stability of alternative tracers in seawater
Chemical reactions (including hydrolysis), adsorption to particles, and biological degradation processes should be considered with regard to the stability of compounds in seawater. There are few studies of the stability of the Medusa tracers, and close to none on the stability in seawater. One indication of the stability is the chemical structure and the atmospheric lifetime. For instance, PFCs have very long atmospheric lifetimes, i.e., > 50 000 and > 10 000 years for 10 respectively, is stable at temperatures of at least 1200°C and the rate of hydrolysis of CF4 is immeasurably small (Ravishankara et al., 1993;Cicerone, 1979). In addition, there are no indications of biological processes that can break C-F bonds, indicating that PFCs are likely to be very stable in the environment.
The contribution of the partial atmospheric lifetime with respect to oceanic uptake of selected HCFCs and HFCs to the 15 total lifetimes is another indirect indication of the stability in seawater. Such partial atmospheric lifetimes depend on the solubility in seawater and other losses relative to their atmospheric concentration, and are always larger than their total lifetimes. Considering the low fraction of these mainly non-polar compounds in the ocean, a small loss in the ocean is insignificant for the overall budget of the compound, but can still be indicative of the potential as a transient tracer. As suggested from previous studies (Yvon-Lewis and Butler, 2002;Carpenter et al., 2014) using this method, 20 HCFCs and HFCs seem to be stable in seawater, although with large uncertainties (Sect. S1 and Table S1). Note that the partial atmospheric lifetimes with respect to oceanic uptake in Table S1 were calculated only considering the chemical degradation process.
Another route is to compare surface saturations of a tracer with unknown stability to those of a compound that is known to be stable in seawater, e.g., CFC-12. The average surface saturation of HCFCs tends to be higher than those of CFC-25 12 (Table S2). This may suggest that HCFCs are stable enough to be suited as tracers in the ocean.
Based on these discussions, PFCs are stable, while HCFCs and HFCs are potentially stable in the ocean when only considering the chemical degradation process and surface saturation in seawater (only for HCFCs). However, the influences of oxygen dependence and biological degradation processes in seawater have not been investigated (Yvon-Lewis and Butler, 2002). For a compilation of published information on biodegradation in freshwater and soil, see 30 Sect. S2 and Table S3. In summary, not enough information is known on the stability of the selected HCFCs and HFCs in the ocean.

Goals of this study
This study extends the work by Li et al. (2019), with a focus on evaluating if rapid, relatively inexpensive, and accurate measurements are possible as well as if these compounds are conservative in the oceanic environment. We also 35 estimate the historical surface saturation to supplement the input function and discuss differences in tracer input functions and their ability to provide additional information on ventilation. A suite of observations of transient tracers with sufficiently different input functions would support the empiric determination of Transit Time Distributions (TTDs), as reported in Stöven and Tanhua (2014). As the first step towards this, these Medusa tracers have been measured, sometimes for the first time, and the results were interpreted to identify their possible use as transient tracers in the ocean. The Mediterranean Sea was chosen for this study because of its rapid ventilation, which causes transient 5 tracers to penetrate most of the water column. In addition, we report from sampling at a shallow station in the Southwestern Baltic Sea.

Measurement of HCFCs, HFCs and PFCs
Measurement of halogenated compounds is often performed by "gas-solvent extraction" techniques, e.g., purge-and-10 trap where an inert gas is bubbled through a seawater sample to move the analytes from the sample into a cold trap for pre-concentration. By desorbing the content of the trap, the sample can then be injected into a gas chromatograph (GC) for separation and detection. This is a well-established technique that has been used successfully for CFCs and SF6 (Bullister and Weiss, 1988;Bullister and Wisegarver, 2008) ideally achieving accuracies in the order of 1% (Bullister and Tanhua, 2010). However, several HCFCs and HFCs (i.e., HCFC-22, HFC-134a, and HFC-125) have low responses 15 and large uncertainties when they are measured by an Electron Capture Detector (ECD) that is normally used for CFC-12 and SF6 (Lobert et al., 1995;Sousa and Bialkowski, 2001;Beyer et al., 2014). One alternative is to use a mass spectrometer (MS) for detection that has the advantage of scanning for unique masses for different compounds, allowing identification and quantification simultaneously. HCFCs and HFC-134a measurements by GC-MS in seawater samples have also been reported in previous studies (Lobert et al., 1996;Ooki and Yokouchi, 2011). 20

The Medusa-Aqua system
The Medusa-GC-MS system (shortened as the Medusa system) for the precise and simultaneous analysis of a wide range of volatile trace gases in the atmosphere has been developed at the Scripps Institution of Oceanography (Miller et al., 2008). This system includes two traps kept at accurately controlled temperatures to trap the volatile gases. The packing material of the traps and the temperature during the trapping stage are designed in a way that allows for the 25 fractionation of the compounds on two traps. In this way, interferences from atmospheric permanent gases can be avoided and hard-to-measure gases like PFC-14 (CF4) can be measured. This analytical system was designed to automatically and continuously measure air samples at Advanced Global Atmospheric Gases Experiment (AGAGE) remote field stations (Prinn et al., 2018) and is unique in that it provides high accuracy measurements of more than 40 compounds including halocarbons, hydrocarbons and sulfur compounds involved in ozone depletion and/or climate 30 forcing from the same sample. The measurement precisions of most halogenated compounds are less than 0.5% in 2 L ambient air. The Medusa-Aqua system, developed based on the Medusa system, is able to measure the majority of the 40 halogenated compounds in seawater samples.
The Medusa-Aqua system consists of a Medusa system (Miller et al., 2008) and a seawater sample pretreatment system (Fig. 2). The Medusa system consists of a cryogenic pre-concentration unit, an Agilent 6890N gas chromatograph 35 (GC), and an Agilent 5975B quadrupole mass spectrometer (MS). The seawater sample pretreatment system purges gaseous tracers from the water samples before injecting them into the Medusa system, replacing the air sampling device of the original Medusa system.
The main difference between the Medusa and Medusa-Aqua system is that the former uses an air-pump module as the gas sample pretreatment system and the sample volume is determined by an integrating mass flow controller (MFC), 5 while the latter uses a purge module as the seawater sample pretreatment system and a gravimetrically calibrated standard loop for standard gases determination. For the injection of water samples to the system, we use the Ampoule-Cracker-System, as designed by Vollmer and Weiss (2002) and then modified by Stöven (2011).

Sampling and Measurement
Here we describe the sampling and measurement methods for samples collected from cruise MSM72 to the 10 Mediterranean Sea in March and cruise AL516 to the Baltic Sea in September 2018. Over the past years, we have collected samples on a few cruises and improved our method (Sect. S3, Figs. S1-S7, and Tables S4-S6).

Sample collection
Seawater samples were collected throughout the water column in three areas of the Mediterranean Sea ( Alkor from September 12 th to 22 nd , 2018 (Booge, 2018). These seawater samples were collected in glass ampoules (~1.3 L) connected to the Niskin bottles via a stainless steel mounting system (Vollmer and Weiss, 2002). 5 minutes is needed for the seawater to fill up a whole glass ampoule and the sampling process lasted for 15 minutes to allow for 20 the seawater to flush the whole ampoule volume three times. The ampoules were flame-sealed immediately after sampling under a flow of high purity N2 (Air Liquide, grade 6.0, Germany) and then sent back to the laboratory in Kiel for measurement. As seen in Fig. 3, no onboard CFC-12 and SF6 measurements (on a PT-GC-ECD) were conducted on the stations we sampled for the Medusa-Aqua system in the Mediterranean Sea, although they were measured at both adjacent stations located 15 nm away along the cruise track. 25

Gas extraction, separation, and detection
The flow scheme for the Medusa-Aqua system is shown in Fig. 2. Before measurement, each ampoule sample was immersed in a warm water bath at 65 °C overnight to support the transfer of the gases into the headspace and to enhance the purging efficiency. The stem of the ampoule is inserted vertically into the cracking chamber and is held by a nut secured by a Teflon ferrule. Then the cracking chamber is flushed with N2 (grade 6.0) for 10 minutes to flush out 30 ambient air completely. A blank test for the cracking chamber is made by simulating an extraction without breaking the glass ampoule. For analysis, the tip of the ampoule's stem is shattered inside the enclosed cracking chamber by rotating the cracking paddle. A straight purge tube is then inserted down into the ampoule until touching the ampoule bottom for finer bubbles. These bubbles will help strip the dissolved gases out of the seawater, and more importantly from the headspace.
The extraction process is started by purging the gases in the ampoule with N2 (grade 6.0) for 20 minutes at a flow rate of 100 mL min -1 . Water vapor is removed from the sample by passing the gases through two Nafion dryers (N1 and N2) of 1.8 m length and one (N3) of 0.6 m length. The counter-flow rate of Nafion dryer gas (N2, grade 5.0) was set 5 to 120 mL min -1 . After the purge gas is injected into Medusa, the following path is the same as described by Miller et al. (2008). The tracer gases are separated on the main column with helium (Air Liquide, grade 6.0, Germany) as the carrier gas and subsequently detected by the MS. The mass of seawater in the ampoules was calculated as the difference between the full weight of the ampoule before measurement and the empty ampoule (including glass splinters) after rinsing with distilled water and drying in an oven. 10

Standard and calibration
The standard gas is a tertiary standard calibrated by the Scripps Institution of Oceanography (SIO) on the AGAGE relative scale "SIO-R1". For details about the propagation of the standard see Miller et al. (2008). Gravimetric calibration scales and calibrated errors of compounds in the tertiary standard are reported in Table 1 Table 1 are linear in the range of our measurements.
The precision of the measurement is dependent on the size of the ampoules and the amount of tracer in the sample; the sample with a higher amount of tracer has better precisions than those with a lower amount. The precision (or reproducibility) for seawater sample measurements were determined by the relative standard deviations (RSD) of the concentrations for duplicate samples from the Baltic Sea (Table 1). They are 0.4%, 3.1%, 6.1%, 1.8%, 9.7%, and 2.0% 30 for CFC-12, HCFC-22, HCFC-141b, HCFC-142b, HFC-134a, and HFC-125, respectively. To quantify the reproducibility of the analytical set-up, we took seawater from the tropical Atlantic Ocean, and let it equilibrate with the atmosphere in the laboratory. The water was then sampled from Niskin bottles in the same way as during a cruise and flame sealed in ampoules, although for this experiment we used 300 mL samples. The reproducibility for CFC-12, HCFC-22, HCFC-141b, and HCFC-142b measurements from four duplicate samples are 3.7%, 2.0%, 3.5%, and 3.4%, 35 respectively (Table 1).
For the target compounds measured by the Medusa-Aqua system from the Mediterranean Sea and Baltic Sea, the concentration of SF6, PFC-14, and PFC-116 in most seawater samples was lower than the detection limit, and HFC-23 had unstable and non-zero blank values in all measurements, preventing us from evaluating those results. The observations of CFC-12, HCFC-22, HCFC-141b, HCFC-142b, HFC-134a, and HFC-125 (in pmol kg -1 ) measured by the Medusa-Aqua system in seawater from both cruises are shown in Table S7 with quality flags marked. 5

Quality control
In order to evaluate the precision and accuracy of seawater measurements by the Medusa-Aqua system, the observations of CFC-12 measured by the Medusa-Aqua system are compared with those in adjacent stations measured onboard by the PT-GC-ECD system in the Mediterranean Sea (Fig. 4). We performed a two-step quality control procedure on the medusa data where, in the first step outliers were flagged as bad (flag 3 in Table S7), and in a second 10 step we flagged data as likely bad (flag 5) where the Medusa-CFC-12 values are inconsistent with the CFC-12 values from PT-GC-ECD, and as good (flag 2) when they are consistent (Fig. 4). The "inconsistent" means that the misfits of CFC-12 concentrations measured by the two instruments are more than three times the standard deviations (3σ), indicating a possible issue during the sampling process. The two-step quality control process led to 9 samples in the Mediterranean that meets all these criteria and have similar concentration as the PT-GC-ECD observations. For such 15 samples, the averaged difference of CFC-12 concentrations measured by the two different instruments is 5.9 ± 4.6% at roughly the same depth by ignoring their distance differences.

Comparison of instruments measuring CFC-12
As mentioned in the previous section, CFC-12 was measured by both the Medusa-Aqua system and a purge and trap GC-ECD instrument (Syringe-PT-GC-ECD) used onboard the cruise MSM72. The latter is a mature system to measure 20 CFC-12, SF6, and SF5CF3 (Stöven, 2011;Stöven and Tanhua, 2014;Stöven et al., 2016;Bullister and Wisegarver, 2008). For comparison, information on the performance of a similar purge and trap system set-up (Cracker-PT-GC-ECD) to measure flame-sealed ampoules for CFC-12 and SF6 is added to a comparison of the three instrument set-ups (Table 2). Compared to other systems, the Medusa-Aqua system has lower purge efficiency due to its larger sampling volume if only considering a single purge (although we used multiple purge cycles to increase the purge efficiency and 25 reduce the uncertainty); has lower precision than that of the Syringe-PT-GC-ECD but higher than that of the Cracker-PT-GC-ECD system; and can measure more compounds.

Time range
The time range where a tracer can be used as a transient tracer is defined by its input function. For chronological 30 transient tracers, the input functions are described by their atmospheric histories and historical surface saturations. For ideal applicability, atmospheric histories of tracers should increase monotonically in the atmosphere. For older waters, and low tracer concentrations, the precision and detection limits will be limiting factors Stöven et al., 2015). Figure 5 shows the atmospheric histories of HCFC-22, HCFC-141b, HCFC-142b, HFC-134a, HFC-125, HFC-23, PFC-14, PFC-116, CFC-12, and SF6 in the Northern Hemisphere (Bullister, 2015;Li et al., 2019).

Tracer age and the Transit Time Distribution (TTD)
Tracer age (or partial pressure age) is defined as the age of a water parcel based on a purely advective flow in the ocean. Figure  If the relative tracer concentrations are higher than 100% in Fig. 6, there has been a decrease in atmospheric concentrations, e.g., for CFC-12. When the atmospheric history of a compound is not monotonically changing, the equilibrium atmospheric mole fraction (and ultimately the age associated with that mole fraction) calculated from its 15 concentration in the ocean is not unique, reducing its potential as a transient tracer (Li et al., 2019). Thus, the tracer age range is a function of the sampling year. For instance, the useful tracer age range of CFC-12 is 30-80 years and 1-60 years for sampling in 2018 and 2000, respectively (Fig. 6). This indicates that the ability of CFC-12 as a transient tracer for recently ventilated water is decreasing with time, but CFC-12 still provides important time information for intermediate and deep water layers with moderate ventilation timescales. It is worth pointing out that PFCs have a 20 longer tracer age range compared to other compounds, even CFC-12, among the chronological transient tracers (Fig.   6).
We used the well-established ocean ventilation model, the Transit Time Distribution (TTD) model that is based on the Green's function (Hall and Plumb, 1994) describing the propagation of tracer boundary conditions into the interior. This model is based on a one-dimensional flow model with time-invariant advective velocity and diffusivity gradient 25 and is thus known as the one-dimensional Inverse Gaussian Transit Time Distribution (IG-TTD). The key variables are the mean age and the width of the distribution ∆ (Waugh et al., 2003). The ∆/ ratio of the TTD with a range of 0.2-1.8 corresponds to the proportion of advective transport and eddy-diffusive characteristics of the mixing processes for a water parcel; the higher the ∆/ ratio, the more dominant the diffusion and vice-versa. For more details on the TTD model and mean age see Sect. S4 and Figs. S9-S10. 30

Historical surface saturation in the Mediterranean Sea
The historical surface saturation of transient tracers is important to constrain the input function (together with the atmospheric history). To determine this for the Mediterranean Sea, we calculated seawater saturation in the winter mixed layer (WML) from historical cruise data. The depths of the WMLs in summer and winter are shown in Fig. 7  35 for two examples of density profiles. It is the saturation during winter that is relevant for deep and intermediate water formation, and thus for the input functions, not the one in the summer mixed layer. Therefore, only the WML was considered in the calculation of historical surface saturation for all cruises. The depth ranges of WMLs (Fig. S8) and the saturation for CFC-12 and SF6 (Fig. 8) were determined by profiles of temperature, potential density, and CFC-12 and SF6 concentration for every historical cruise in the Mediterranean Sea from 1987 to 2018 (Schneider et al., 2014;5 Li and Tanhua, 2020). Since no clear trend over time could be described, we averaged the mean seawater saturation from every single cruise and determined the historical surface saturation to be 94 ± 6% and 94 ± 4% for CFC-12 and SF6, respectively. For CFC-12, this is different from the situation in the North Atlantic Ocean , and could be an indication of the different oceanographic settings where the inflowing Atlantic Water (to the Mediterranean Sea) takes a long time to equilibrate with the atmosphere. The constant undersaturation through time is 10 then mainly a function of rapid cooling during winter and entrainment of water from below, rather than a rapid change of the atmospheric concentration that can drive undersaturation that varies over time. For the following calculations of ages and evaluation of stability, the historical surface saturations are assumed to be a constant 94% for all tracers in this study as no data exists to determine the historical surface saturation of selected HCFCs and HFCs in the Mediterranean Sea. 15

Observations of the Medusa tracers in the Mediterranean Sea
Based on the reasonable correlation between CFC-12 observations from the Medusa-Aqua system and the onboard PT-GC-ECD system (Sect

Surface saturation of Medusa tracers in seawater
The surface saturation of seawater can be an indicator of the stability of a compound in surface seawater or the 25 correctness of the seawater solubility function. However, saturation is influenced by multiple parameters, such as partial pressures in the atmosphere and surface seawater, the air-sea exchange velocity, the solubility and diffusivity of the gas, bubble injection and/or vertical mixing, and the temperature dependence of these parameters (Lobert et al., 1995;Butler et al., 2016). In the Mediterranean Sea, we flagged all surface samples as suspect due to discrepancies between CFC-12 concentrations as measured by the Medusa-Aqua system and PT-GC-ECD. We, therefore, turn to the 30 samples collected in the Baltic Sea where the seawater saturation of CFC-12 and Medusa tracers in one surface sample and two bottom samples are shown in Table 3. Note that the bottom waters (at about 23.5 meters depth) can be considered as recently ventilated as this water is ventilated on an annual basis. The averaged saturation for the Baltic seawater samples are: CFC-12 (122 ± 8%), HCFC-22 (77 ± 8%), HCFC-141b (74 ± 12%), HCFC-142b (114 ± 2%), HFC-134a (125 ± 23%), and HFC-125 (252 ± 35%). 35

Stability based on interior ocean observations
In order to validate the stability of HCFCs and HFCs, the concentrations of CFC-12 from the adjacent PT-GC-ECD measurements for the Mediterranean samples are vertically interpolated by a piecewise cubic hermite interpolating method on potential density surfaces and averaged by the arithmetic mean of the interpolated profiles Schneider et al., 2014). Then the concentrations of HCFC-22, HCFC-141b, HCFC-142b, HFC-134a and HFC-5 125 and SF6 (measured by the PT-GC-ECD) are plotted against the (interpolated) CFC-12 for the Mediterranean and Baltic samples (Fig. 10). The concentrations of the Medusa tracers from the recent (usually shallow) layers are located in the upper right corner with older (usually deep) samples near the lower left in Fig. 10. In the figure, we added the atmospheric history of the Medusa tracers vs. CFC-12, the theoretical mixing line between contemporary and preindustrial concentrations (boundary line to 2018), as well as the theoretical mixing lines assuming an IG-TTD with 10 ∆/ = 0.2-1.8. All samples will have to fall between the first two lines (i.e., the stability area) if the tracer is conservative in seawater by referring to previous methods (Huhn et al., 2001;Roether et al., 2001). Compounds where the samples fall below the "stability area" are not stable (assuming that CFC-12 is stable), and for samples above it, there are issues with too high values (see below). For instance, HCFC-22 is found in the lower part of the stability area (samples would fall on this lowest line if there were no mixing but only advection in the ocean); HCFC-141b, HFC-15 125, and SF6 are well in the allowed ranges; whereas HCFC-142b and HFC-134a are around or above the upper boundary.
The increased ventilation of the (western) Mediterranean Sea during the last decade tends to result in different effects on CFC-12, which is decreasing in the atmosphere, and the Medusa tracers with increasing atmospheric concentrations.
This argument suggests that we could see slightly higher than expected concentrations (similar to SF6) for the Medusa 20 tracers. This is exactly what we see for HCFC-142b (Fig. 10). We found that the ∆/ ratio determined by the CFC-12/HCFC-141b and CFC-12/HCFC-142b tracer pairs are similar to that of the CFC-12/SF6 pair, indicating promising tracer pairs. The higher than expected concentration of HFC-134a may be caused by 1) contamination during sampling or measurement in the laboratory; 2) problems with solubility functions; 3) some other issues within the measurements in the laboratory causing our measurements to be too high. 25 Although the assumption of time-invariant ventilation is not valid for the Mediterranean Sea, the TTD model can produce indicative results to understand the mean ages estimated from Medusa tracers (assuming ∆/ =1.0) and their comparison to those estimated from CFC-12 and SF6, see Sects. S4-S5 and Figs. S9-S11.

Discussions
The results from this study based on interior ocean observations can be used to evaluate the stability and further 30 determine the potential of the Medusa tracers as oceanic transient tracers, and are as such dependent on reasonably accurate measurements. The comparison between the mean ages calculated from the Medusa tracers and CFC-12 is sensitive to the assumed shape of the TTD, and the differences in input history that make them respond differently to  Table   4 by mainly evaluating their atmospheric history, seawater solubility, ease of measurement, and stability in seawater.

CFC-12. This tracer is included as a reference as it is a commonly used transient tracer. The atmospheric history of
CFC-12 is well-documented (Walker et al., 2000;Bullister, 2015), and the seawater solubility function is well-5 established (Warner and Weiss, 1985). In addition, CFC-12 has been observed for several decades by mature analytical techniques, and its stability in warm waters, as well as poorly oxygenated waters, has been proven. However, the decreasing atmospheric history of CFC-12 limits its ability as an oceanic transient tracer -only two black stars in the column "Atmospheric concentration" since it is well-established but still not ideal for an oceanic transient tracer of recently ventilated waters. confidence in the seawater solubility was marked as medium due to lack of direct experimental seawater solubility data to verify the function. HCFC-22 has been measured on several cruises (Lobert et al., 1996;Yvon-Lewis et al., 2008) by GC-ECD and GC-MS instruments, and in this study by the Medusa-Aqua system. The stability was evaluated by comparison to CFC-12 observations. The clustering of HCFC-22 values in the lower range (Fig. 10) could be an 20 indication of slow degradation in seawater, which is also supported by the weak hydrolysis of HCFC-22 in tropical and subtropical waters reported by Lobert et al. (1995). Therefore, HCFC-22 was determined to be unstable in warm waters; additional experiments are needed to evaluate the stability, especially in poorly oxygenated waters. In addition, HCFC-22 can be replaced by SF6 as a transient tracer since they have similar atmospheric histories ( Fig. 5 and Fig. 6).
These indicate that HCFC-22 might not be suitable as a new transient tracer in the warm ocean, for instance, the 25 Mediterranean Sea, but could be used for colder waters.
HCFC-141b. The atmospheric history is reliably reconstructed (Li et al., 2019). However, the seawater solubility function was constructed for the first time (Li et al., 2019), and the freshwater solubility only matched the data presented in Abraham et al. (2001) at the two temperatures. The low surface saturation measured in the Baltic Sea suggests that there is likely an issue with the solubility function. HCFC-141b has been previously measured on cruises 30 (Lobert et al., 1996;Yvon-Lewis et al., 2008) and also in this study, thus we have high confidence in the ability to measure this compound. HCFC-141b is identified to be potentially stable in seawater since its concentrations are in the expected range in the interior ocean (Fig. 10). The input function of HCFC-141b is different enough from the traditional transient tracers to provide additional information, but since the atmospheric history started to decrease in 2017 (Li et al., 2019), the use of HCFC-141b as a transient tracer for "young" waters will be compromised. All these 35 indicate that HCFC-141b has probably limited potential as a transient tracer in the future.
HCFC-142b. The confidence of the atmospheric history is similar to those of HCFC-141b for the same reasons, while the high HCFC-142b saturation of samples from the Baltic Sea supports the medium confidence of the seawater solubility function. HCFC-142b has been measured on some cruises (Lobert et al., 1996;Yvon-Lewis et al., 2008) and also in this study, rendering us to have the confidence to accurately measure this compound. We have medium confidence in our ability to estimate the stability of HCFC-142b because of higher than the expected concentrations in the interior ocean (Fig. 10), but that could be linked to the slightly too high saturation observed. The input function of HCFC-142b is different from those of most other tracers (only similar to that of HCFC-141b but with a longer time 5 range). Consequently, HCFC-142b has a high potential to be used as a transient tracer.

HFC-134a.
We have high and medium confidences in the atmospheric history and seawater solubility function (Li et al., 2019), respectively. Although the estimated seawater solubility function was constructed based on the modeled salting-out coefficients and the experimental freshwater solubility (Li et al., 2019) that matched both the observations (Deeds, 2008) and model results (Abraham et al., 2001), we have only medium confidence in the seawater solubility 10 function due to the lack of experimental seawater solubility data. However, the samples from the Baltic Sea suggest that the solubility function seems reasonable. HFC-134a was measured by Ooki and Yokouchi (2011) by GC-MS and in this study by the Medusa-Aqua system. It could indicate contamination issues due to higher than expected concentrations (see Sect. 5.4). We have medium confidence in the stability of HFC-134a also because of the higher than expected concentrations (Fig. 10). The compound is not identified as unstable and Fig. 10 suggests that the 15 HFC134a is stable (see Sect. 5.4). In addition, HFC-134a can only be used as a tracer for "young" waters due to its short atmospheric history. Based on all these discussions, HFC-134a has a medium potential to be an oceanic transient tracer.
HFC-125. The atmospheric concentrations of HFC-125 in the early 1990s are unclear (Fig. 6), possibly related to uncertainties in the reconstruction, although this only marginally influences its ability as a transient tracer. Three 20 seawater solubility functions of HFC-125 can be constructed (Li et al., 2019), although only two of them were considered; function 1 is supported by freshwater solubility results from Deeds (2008) as well as stability analysis based on comparison to CFC-12 in this study (Fig. 10), whereas the observations and model results from Abraham et al. (2001) supported function 3. This leads to low confidence in the seawater solubility function of HFC-125. We also consider the ability to measure HFC-125 as low since this compound has been measured for the first time in seawater 25 in this study, and we find almost no vertical gradient (Fig. 9), which is unexpected. Furthermore, observed HFC-125 concentrations in freshwater are inconsistent as indicated by three freshwater solubility functions (Li et al., 2019), which suggests unresolved issues with its measurements in water. Due to the poorly defined solubility and difficulties in measurement, it is difficult to assess the stability of HFC-125 in this work. Also, HFC-125 can only be a tracer for "young" water due to its short atmospheric history. Therefore, we consider that HFC-125 has low potential as a 30 transient tracer in the ocean, although this might be remedied by experimentally determining the seawater solubility function and solving possible measurement issues.
HFC-23. HFC-23 could not be reliably measured in our system due to unstable non-zero blanks (see Sect. 3.4). Therefore, we can not reliably assess the stability of HFC-23 in seawater, and we have low confidence in our ability to measure this compound, although the blank problem might be solved by a different instrument configuration. The 35 atmospheric history of HFC-23 has been constructed (Li et al., 2019;Simmonds et al., 2018), but we have only medium confidence as it does not start from zero (Simmonds et al., 2018) due to limited data. We have only medium confidence in the seawater solubility function for the same reason as for HFC-134a. That is, the freshwater solubility function matched results from Deeds (2008) and Abraham et al. (2001) but the seawater solubility function was not constructed by experimental seawater solubility data. In consequence, unknown stability and current issues with measurements lead to the overall assessment that HFC-23 has low potential as a transient tracer in the ocean.

PFC-14 and PFC-116.
The increasing atmospheric histories of PFC-14 and PFC-116 are well established (Li et al., 2019;Trudinger et al., 2016). The seawater solubility functions have been constructed with medium confidence for 5 PFC-14 as it matches both seawater measurements (Scharlin and Battino, 1995) and freshwater solubility (Clever et al., 2005;Abraham et al., 2001). In contrast, low confidence for the solubility function for PFC-116 is attributed to its freshwater solubility only matching that from Deeds (2008) but not the theoretical assessment from Abraham et al. (2001). PFC-14 and PFC-116 are very stable in the environment, but cannot easily be measured in seawater because of the low solubility (Li et al., 2019), i.e., low concentration in seawater. The high stability and long atmospheric 10 histories make PFCs potentially promising transient tracers in the ocean, although measurements are challenging. A possibility would be to use an ECD that has a high sensitivity for C-F bonds, but with the additional complication of co-elution on the chromatogram.

Conclusions and outlook
This study, combined with the study by Li et al. (2019), provides a method to identify and evaluate if a compound is 15 suitable for use as a transient tracer in the ocean. Selected HCFCs, HFCs, and PFCs were evaluated for their potential as transient tracers. The evaluation mainly considered four aspects: input function (including atmospheric history and historical surface saturation), seawater solubility, feasibility of measurement, and stability in seawater. We also considered how Medusa tracers with different atmospheric histories complement each other when constraining ocean ventilation. For these purposes, we modified an existing analytical system for seawater measurements and observed 20 seawater concentrations of HCFC-22, HCFC-141b, HCFC-142b, HFC-134a, and HFC-125. Unfortunately, the poorly soluble PFCs could not be successfully measured with our current analytical system.
By comprehensive evaluation, we conclude that HCFC-22 is unlikely to be a transient tracer in warm waters, whereas HFC-23 cannot be evaluated as a transient tracer because of the lack of information. Fortunately, these two compounds can be substituted by SF6 that has a similar atmospheric history. On the other hand, both HCFC-141b and HCFC-142b 25 show high potential as transient tracers currently. Considering their similar atmospheric histories and the decreasing atmospheric mole fraction of HCFC-141b, HCFC-142b should be further evaluated by obtaining more reliable solubility and stability information in seawater. HFC-134a and HFC-125 show medium and low potentials as transient tracers, respectively; the former because of higher than expected concentrations pointing to issues on the seawater solubility function and/or the measurements but not identified to be unstable; the latter due to the lack of information 30 on solubility, stability, and feasibility of measurement in seawater. Considering their similar atmospheric histories, HFC-134a is a more promising candidate as a transient tracer. Last but not least, PFC-14 and PFC-116 show large potential as transient tracers once their accurate measurement in seawater is resolved. The high stability and long atmospheric history make it worthwhile to explore improved analytical methods, whether it implies using a more sensitive detector, or larger samples. 35 The main result from this study is that although a couple of the investigated tracers might be useful as oceanic tracers, at least under some circumstances such as cold waters, etc., none of the tracers would be fully qualified at this time.
Although we have a good handle of the input function in the atmospheric history, we recognize that targeted work is needed for all compounds to better constrain the solubility. For stability, we have an idea which ones are possibly stable, but more work is needed here too (possibly laboratory experiments under controlled conditions). There are 5 limitations in the current configuration of the analytical techniques as described, and additional efforts are needed to understand and improve the system. In summary, we are not quite yet in the position to recommend targeted observations of these tracers, but this study provides a useful summary of the current knowledge, including the new research reported on in this manuscript, to guide further, more targeted experiments and studies.
Future work will be further evaluating the potential transient tracers identified in this study and constraining ventilation 10 in the global ocean by combining multiple transient tracers using suitable models to describe the Transit Time Distribution. The next steps include experiments in controlled laboratory conditions to construct seawater solubility functions and explore the stability of the compounds, in particular for warm and/or oxygen-depleted seawater.
Additional work on improving the measurement capacity for PFC-14 and PFC-116 by modifying the Medusa system according to Arnold et al. (2012) or possibly using a more sensitive detector (i.e., ECD). The reliability and 15 reproducibility of the measurements for routine use should be improved, possibly by using the vacuum-sparge method by Law et al. (1994) that would speed up the gas extraction and reduce the number of purge cycles needed.

Author contributions
TT conducted the sampling from cruise MSM72. PL developed the instrument and carried out the measurements. PL interpreted the data and analyzed the results based on the discussion with TT. PL wrote the paper with contributions 25 from TT.

Competing interests
The authors declare that they have no conflict of interest.    likely bad (here we used flag 5, see text). In the following plots, we show all data with a quality flag of "2 (dots)" or "5 (crosses)".      PFC-116 *** * * *** ** a The total number of (black and red) stars represent current knowledge: one star means "largely unknown", two stars "reasonably well resolved" and three stars "well documented or resolved"; the number of black stars represents the possibility of a compound as a transient tracer through current assessments. For instance, the atmospheric history for CFC-12 got 3 stars since we know this 5 very well, but only 2 of those are black since its atmospheric concentrations are decreasing after experiencing a continuous increase.