Atmospheric measurements of methyl chloroform provide important constraints on the rate of oxidation of hydrocarbons in Earth's atmosphere. Estimates of the loss of methyl chloroform to the oceans play a small but important role in these calculations. Here, we examine the ocean-atmosphere interaction of methyl chloroform in a global ocean model. Contrary to previous assumptions, these simulations suggest that the high-latitude oceans are currently a source of this chemical to the atmosphere. If confirmed, this finding alters estimates of the change in the atmospheric oxidation rate of hydrocarbons. We highlight the potential usefulness of methyl chloroform as a tracer of ocean circulation.
 Manufacture of the solvent 1, 1, 1-trichloroethane (CCl3CH3, methyl chloroform, here after MC) was first controlled and later essentially banned by amendments to the Montreal Protocol on Substances that Deplete the Ozone Layer. Because its atmospheric lifetime is about five years, the concentration of MC in the atmosphere has fallen from its peak value of ∼130 pptv in 1991/1992 to less than 35 pptv today [e.g., Prinn et al., 2001]. The decline in the concentration of MC is responsible for most of the decrease in stratospheric chlorine levels observed to date [Montzka et al., 1999].
 Atmospheric measurements of the global distribution of MC have provided important constraints on the rate of oxidation of hydrocarbons in Earth's troposphere. Because the location, quantity, and timing of MC release to the atmosphere are well known [McCulloch and Midgley, 2001], measurements of the atmospheric concentration allow its lifetime to be reasonably estimated [e.g., Prinn et al., 2001]. As with many hydrocarbons, this lifetime is determined primarily by gas phase oxidation initiated by hydroxyl radicals (OH). With knowledge of the oxidative lifetime of MC, the lifetimes of numerous other compounds with similar chemistry, such as methane [Hein et al., 1997] and hydrochlorofluorocarbons (HCFCs), can be estimated.
 Recently, inversion of the atmospheric MC record by Prinn et al.  suggested that the atmospheric lifetime of MC had increased markedly in the 1990s. Prinn et al. concluded that the increase in the lifetime resulted from a ∼10–15% decrease in the concentration of OH during this time period. Such a finding is alarming because it implies a significant increase in the atmospheric lifetime of numerous greenhouse gases, including the CFC replacements (HCFCs).
 We report here simulations of the uptake and redistribution of MC by the ocean. As hypothesized by Krol and Lelieveld , our simulations suggest MC began to outgas from the high-latitude oceans following its rapid decrease in the atmosphere in the mid and late 1990s. The change in ocean exchange described here reduces the inferred trend in the atmospheric lifetime of MC. Prinn et al.  suggested that a ‘missing source’ of about 117 Gg (∼1 × 109 moles) during the five year period from 1995 to 2000 would be required to yield no trend in the atmospheric lifetime. The change in the ocean-atmosphere interaction described here accounts for ∼30% of this deficit.
2. Interaction of Methyl Chloroform With Ocean Water
Wine and Chameides  first described the interaction of MC and ocean water. Ship-based measurements of its concentration in surface water and in the overlying air confirmed that the oceans were an important sink for MC [Butler et al., 1991]. Although only sparingly soluble in water [Gossett, 1987], a significant amount of MC has entered the ocean during the last thirty years. Significant transfer from the atmosphere to the ocean occurs in the warm tropical oceans, driven mainly by chemical loss (hydrolysis) in the surface waters. Uptake has also occurred at mid- and high- latitudes, due to the higher solubility of MC in colder waters and two distinct high-latitude water-mass transformation processes: high-latitude deep convection, and wind-forced lateral subduction into the main thermocline of the ocean.
 It is well established that MC is chemically removed in warm ocean water. Observations by Butler et al.  in the tropics showed significantly lower dissolved MC than expected from equilibrium with the atmosphere. They demonstrated that the flux of MC from the atmosphere to the ocean was broadly consistent with physical uptake [Gossett, 1987] and subsequent hydrolysis [Jeffers et al., 1989; Gerkens and Franklin, 1989]. The hydrolysis lifetime in ocean water, however, is strongly dependent on temperature. At 25°C, the lifetime against hydrolysis is about one year. This increases to ∼50 years at 3°C, though this value is quite uncertain because it is determined by extrapolation of the laboratory data from warmer temperatures [Jeffers et al., 1989; Gerkens and Franklin, 1989].
 In previous studies of the atmospheric lifetime of MC, loss from the atmosphere to the ocean has been treated as a simple first order process. Consistent with the recommendation of Butler et al. , a lifetime (1/ka-o) of ∼60 years with respect to the atmospheric reservoir is generally prescribed. Such a simple parameterization is adequate for the component of MC lost by hydrolysis in warm ocean water, where the lifetime in ocean water (∼1 year) is shorter than in the atmosphere (∼5 years). It is, however, inadequate for loss via physical uptake in the cold (∼3°C), high latitude ocean. In this process, the flux is sensitive to both the time derivative of the atmospheric concentration and the rate at which mixing of the surface waters into the deep ocean occurs – as both determine the saturation properties of the surface waters. Thus, when the atmospheric concentration of MC was growing exponentially, the flux into the ocean was proportional to [CH3CCl3]. After 1990, when the atmospheric concentration abruptly began to decrease, loss to the high latitude ocean was a complicated function of both atmospheric MC concentration and ocean dynamics.
3. Global Ocean Model
 In this study, we have used a 60-year simulation of ocean transport from the Miami Isopycnal Coordinate Ocean Model (MICOM) to investigate the uptake of MC. MICOM is a primitive equation numerical general circulation model that describes the evolution of momentum, mass, heat and salt in the ocean [Bleck et al., 1992, 1989]. Further details of the model are provided in auxiliary material.
 We have forced MICOM with observed (and into the future with modeled) concentrations of atmospheric MC. We used the emissions inventory of McCulloch and Midgley  along with the tropospheric and stratospheric loss rates suggested by Prinn et al.  to calculate the concentration in a three box model (northern extra tropics, the tropics and southern extratropics). The exchange rate between the hemispheres was picked such that the latitudinal gradient observed between 1970 and 1990 was captured. In this model, atmospheric measurements of MC are reproduced to better than 10%.
 Tracers (MC and CFC-11) were advected offline using 30-day averages of isopycnal and diapycnal mass fluxes. These were obtained by transforming instantaneous horizontal fluxes in the two non-isopycnic model layers representing the surface mixed layer and a “fossil” mixed layer to density coordinates, integrating over time, and inferring diapycnal fluxes from mass conservation. Air-sea gas exchange was parameterized using the protocol described by Dutay et al. .
Figure 1 shows the calculated, annually averaged flux of MC into and out of the ocean in 1990, 2000, and 2010. In 1990, the flux is everywhere into the ocean. The largest fluxes are at low-latitudes. Outside the tropics, the uptake pattern reflects the thermal asymmetry of the anticyclonic subtropical gyres. In 2000, while significant losses to the ocean are calculated in the tropics, these are nearly balanced by a net source to the atmosphere from the high latitude oceans - particularly in the southern hemisphere. By 2010 the net air-sea flux over most of the low- and mid-latitude ocean is close to zero, reflecting the expectation that the atmospheric burden has dropped to nearly zero. The high-latitude northern and southern oceans remain net sources to the atmosphere, although smaller than in 2000.
 The calculated time series of the net MC flux into the ocean, and an inert tracer (no hydrolysis) with the same solubility, is shown in Figure 2. Also shown is the ocean loss parameterization used in previous inversions [e.g., Prinn et al., 2001]. Prior to 1980, when atmospheric MC concentrations were increasing rapidly, the first-order loss calculated with an assumed 1/ka-o of 60 years agrees well with the total loss calculated here. Between 1980 and 1992, the first-order loss parameterization is slightly larger than the best estimate calculated here. Between 1992 and 2000, the difference between the MICOM-calculated loss and the first-order loss estimate gradually increases. Between 1980 and 2000, the integrated difference is 5 × 108 moles with much of the difference (3 × 108 moles) occurring between 1992 and 2000. We have attempted to estimate the uncertainty in this calculation due to poor knowledge of the hydrolysis rate at cold temperatures. As described in the auxiliary material, the error-bars in the upper panel of Figure 2 correspond to our estimate of the lower and upper bounds of the hydrolysis rate. In the lower panel of Figure 2, the fluxes are shown for various latitude bands. Most of the chemical loss occurs equatorward of 45° while physical uptake, and later release, occurs at high latitudes.
5. Model Evaluation
 There are only a few published measurements of MC in the ocean to evaluate the model representation of the coupled chemistry and physics. For evaluation of the calculated chemical loss at warm temperature, we rely on the tropical and subtropical MC surface saturation measurements described by Butler et al.  and Lobert et al. . In the tropics, the sub-saturation calculated in this work is similar to that reported by Lobert et al.  and somewhat smaller than those of Butler et al. . In cold water, depth profile measurements of MC illustrate that MC is quite conservative (very slow biological or abiological hydrolysis loss). Profiles from the north Pacific are described by Lobert et al. ; additional profiles were obtained during the 1993, 1994, 1995, 1996, and 1998 World Ocean Circulation Experiment (WOCE) repeat cruise tracks in the Labrador Sea (AR07W). In both cases, the profiles are similar to the inert CFC-11 and CFC-12, showing (unlike CCl4) no evidence of rapid loss. See auxiliary material for further discussion and comparisons of our calculations with these and other data.
 To evaluate the calculation of the physical uptake of MC and its subduction through the thermocline we have compared calculations of the air-sea exchange of CFC-11 with measurements of CFC-11 made during the WOCE. Because the atmospheric temporal trajectory of MC (until 1992) and CFC-11 are very similar this is a reasonable comparison. The calculated inventory of CFC-11 in 1995 was 5.6 × 108 moles, remarkably close to the 5.5 ± 1 × 108 mole inventory estimated from WOCE [Willey et al., 2004]. This suggests that, in the absence of chemical loss, the model MC inventory would also be quite accurate.
 In summary, the losses of MC to the tropical ocean appear to be reasonably well represented by this calculation. The fluxes out of the high latitude Southern Ocean in the period 1990–2000 are more uncertain. The Southern Ocean is, in general, poorly simulated by most ocean models [Dutay et al., 2002]. On an encouraging note, a sensitivity run in which gas exchange was set to zero at all latitudes south of 60°S (thereby preventing any tracer from entering the ocean at high southern latitudes) yielded very similar results to the control run. This was mainly due to anomalously large air-sea gradients forming in the region just north of 60°S as water moved across this boundary, which in large part compensated the zero-air-sea-flux requirement. However, the most important test of these conclusions would be measurements of MC depth profiles and surface saturation measurements in the Southern Ocean.
 Consistent with the conclusion of Krol and Lelieveld , this study suggests that accurately assessing trends in the atmospheric lifetime of MC during the period of rapidly changing emissions (1990–2000) is challenging and perhaps impossible. Any reservoir for MC with a lifetime longer than ∼5 years, later released to the atmosphere, will produce a significant error in attempts to infer the atmospheric lifetime of this compound. Here we address one such reservoir – the high latitude ocean. There are, undoubtedly others. Krol and Lelieveld note, for example, that a delay in the emission of just 65 Gg MC from the early to the late 1990s (for example by stockpiling in advance of the Montreal Protocol controls [McCulloch and Midgley, 2001]) would be sufficient to remove the entire OH trend inferred by Prinn et al. . Such delayed emissions have been observed in both Europe [Krol et al., 2003] and in North America [Barnes et al., 2003; Millet and Goldstein, 2004]. Following the rapid decline in its emission in 1990, MC alone is not likely to be a useful tracer of changes in the gas-phase oxidation rates in Earth's troposphere. Its future utility may lie, however, in its pulse-like appearance in the ocean.
 Measurements of the pulse of 14C following atmospheric bomb tests in the 1960s provided significant constraints on ocean circulation [Broecker et al., 1985]. This study suggests that the rapid rise and fall of atmospheric MC may provide a new tracer against which to test models of ocean circulation. Unlike bomb tritium, which also had a pulse-like atmospheric source, MC is a gas whose atmospheric dispersion does not rely on incorporation into raindrops, and hence is more easily modeled. Thus, ocean observations of MC combined with measurements of CFC-11 and CFC-12 may provide a significant constraint on diffusive/advective transport in the oceans on the 50–100 year timescale critical for understanding how the ocean uptake of CO2 may change over the next century [Hall et al., 2002]. The interactions of MC with ocean water are in some ways simpler than 14CO2 (apparently no interaction with the biosphere) and in other ways more complicated (hydrolysis). Both MC and 14C are similar, however, in that the ocean is currently changing from an atmospheric sink to an atmospheric source [Caldeira et al., 1998].