Three-dimensional simulations of atmospheric methyl chloroform: effect of an ocean sink

A global three-dimensional chemical tracer model of the distribution and seasonal cycles of the surface concentration of CH3CC13 is compared with surface observations from the Atmospheric Lifetime Experiment (ALE) for the years 1980-1985. Two-dimensional OH distributions calculated by a photochemical model are empirically adjusted 'from observed trends in the global average and the interhemispheric ratio of methyl chloroform. The effects of the recently discovered ocean sink for methyl chloroform were investigated. The model simulates the 5-year record of observations made at the five ALE sampling sites to generally within +5% of the observed mean. The calculated average global lifetime of methyl chloroform is 5.7 + 0.3 years. The estimated global mean OH concentration is 6.5 + 0.4 x 105 cm -3. However, the inclusion of the ocean sink does not significantly improve the simulation of the observed interhemispheric gradient of methyl chloroform. Atmospheric transport dominates the simulated CH3CC13 seasonal cycle throughout the northern hemisphere but is less important in the southern hemisphere.

atmospheric concentration of CH3CC13 has been measured as part of the Atmospheric Lifetime Experiment (ALE) at five locations ( Figure 1) since 1978 [Prinn et al., 1987]. Because the flux of CH3CC13 to the troposphere can be closely approximated by industrial production figures and its destruction through reaction with OH can be expressed as simple bimolecular reaction rate, the global distribution, seasonal cycle, and trend of CH3CC13 have been used to elucidate the global distribution, seasonal cycle, and trend of atmospheric OH.
Recently, chemical tracer models (CTMs) based upon GCMs have been used to examine the three-dimensional global distribution of CH3CC13. Spivakovsky et al. [1990] and Taylor et al. [1991] simulated CH3CC13 using OH fields generated by their two-dimensional and three-dimensional models. Both groups found that their photochemical models coupled with threedimensional CTMs underpredicted the observed concentration and trend in atmospheric CH3CC13. Spivakovsky et al. concluded that their photochemical model overpredicted OH because of inadequate treatment of OH losses through reactions with nonmethane hydrocarbons. Taylor et al. concluded that the flux of CH3CC13 to the atmosphere is somewhat larger than estimated by Prinn et al. [ 1987].
In this paper, we address this issue with a three-dimensional chemical tracer model based upon the Los Alamos GCM and a two-dimensional photochemical model of tropospheric OH. We simulate the global distribution, trend, and seasonal cycle of CH3CC13 from 1980 to 1985. The results are compared with the ALE record for the same time period. We explore whether the interhemispheric OH concentration ratio can be constrained by the ALE observations in the two hemispheres. We then examine the effects of the recently reported ocean sink for CH3CC13 [Butler et al., 1991].
In section 2 we briefly describe the model configuration.  Table 1 shows the heights and spacings of the vertical levels. Model transport includes horizontal advection, convection, horizontal diffusion, and vertical diffusion. Five precalculated dynamical and thermodynamical variables are required: the horizontal winds (u and v), the vertical wind (w), temperature (T), and the vertical diffusion coefficient (K). These variables are  [Rasmussen et al., 1982]. The inability to estimate the source strength of the former USSR, eastern Europe, and China will result in an unknown but probably small underestimation of the CH3CC13 concentration measured at the ALE sites.
Although the global annual emission rate of CH3CC13 can be approximated from industry-supplied annual production figures, the spatial and seasonal distribution of the emissions are unknown. We globally distribute CH3CC13 emission sources in the same manner as used by Spivakovsky et al. [1990]. Documented electrical power consumption in individual countries is taken to be an index of technological advancement that can be used to apportion CH3CC13 use and, consequently, emissions. This method has been used successfully to simulate CFC-11 and CFC-12 [Prather et al., 1987]. For CH3CC13, the annual release from each grid square was computed as the product of the total annual release of CH3CC13 (as shown in Table 2) and the fractional annual release of CFC-11 from each grid as estimated by Spivakovsky et al. We assumed that the annual emissions were uniformly released throughout the year. The emissions were released into the two lowest vertical levels of the model because the first 1 km of the planetary boundary layer is rapidly mixed vertically. The Chemical Sink of CHsCCls The sole chemical reaction within the troposphere that destroys methyl chloroform is the reaction with OH radicals. Photolysis by UV is an additional loss mechanism that occurs in the stratosphere. In our model these processes are expressed as CH3CC13 + OH --> product with a reaction rate of k 1 CH3CC13 + h¾ --> product with a photolysis rate of J1 The photolysis rate J1 is calculated using the technique employed by Prinn et al. [1975]. Reaction rate k• is 5.0 x 10 -•2 exp(-1800/T) [DeMote et al., 1990]. Following the completion of this work, it was reported that this reaction rate should be revised downward [Talkudar et al., 1992]. The effect of this revision is discussed below.

Model Initialization
The initial two-dimensional (latitude-altitude) distribution of CH3CC13, corresponding to the monthly mean values of July 1978, was constructed from ALE surface measurements and measured vertical profiles [Khalil and Rasmussen, 1984;Knapska et al., 1985;Rasmussen and Khalil, 1983].

Methodology for Estimating the Global OH Field
We began with monthly average two-dimensional OH distributions as calculated from a two-dimensional photochemical model . Figure 2 shows the initial OH distribution in January and July. The model computes a global tropospheric mean OH concentration of 6.5 x 105 molecules/cm -3. This is consistent with the range of reported values (3 to 10 x 105 molecules/cm -3) as reviewed by Taylor et al. [ 1991 ]. The distributions show that OH has maximum at 20øS and 700 mbar in January and at 20øN and 700 mbar in July. This model computes the interhemispheric OH concentration ratio ([OH]N/[OH]s) to be about 0.8. This distribution, and the hemispheric asymmetry of the OH field, are similar to those found by others [Brasseur et al., 1990;Chameides and Tan, 1981;Taylor et al., 1991]. Figure 3 shows the tropospheric seasonal cycle of OH. The amplitude of the seasonal cycle of OH in the northern hemisphere is somewhat smaller than that in the southern hemisphere because of the higher concentrations of CH4, CO, and nonmethane hydrocarbons in the northern hemisphere. to the higher surface albedo and the enhanced levels of NOy and 03 over land compared to those over the ocean. Zonally extending the two-dimensional OH field longitudinally throughout the three-dimensional transport model is justified because the atmospheric lifetime of CH3CC13 is much longer (years) than the time constant for circumzonal transport (weeks). On average, any given molecule of CH3CC13 will circumnavigate the globe several hundred times before being destroyed by reaction with OH. The expected difference in OH concentrations over oceans and continents is not large enough to induce significant meridional variations in the CH3CC13 distribution. For all of the simulations discussed below, we assumed that there was no temporal trend in OH during the 7-year Simulation.

SIMULATIONS OF CH3CC13 IN THE ATMOSPHERE Case l ' Base Case
Our base case simulation used the OH distribution derived from the two-dimensional model described above. Table 3a summarizes the salient features of the simulation. The lifetime of CH3CC13 in the model is 6 years and the interhemispheric transport time constant is 6 months. This lifetime is consistent with the results of other studies [Prinn et al., 1987;Spivakovsky et al., 1990;Taylor et al., 1991 ]. The interhemispheric time constant is somewhat shorter than the 7-to 12-month range found to give acceptable simulations for CFC-11 [Prather et al., 1987]. We have found that this model, with a 6-month interhemispheric transport time constant, accurately simulates the interhemispheric gradient of CFC-11 [Kao et al., 1992]. Figure 4 shows the calculated surface distribution of CH3CC13 averaged for January and July from 1980 to 1985. Longitudinal concentration gradients associated with sources of CH3CC13 in the industrialized regions of the northern hemisphere are apparent. These features are absent in the southern hemisphere because of the absence of major sources there. Figure 5 shows the calculated meridional cross section of CH3CC13 also averaged for January and July from 1980 to 1985. The distribution is consistent with the majority of CH3CC13 being released at the surface in the northern hemisphere and mixing rapidly vertically throughout the northern troposphere. A tongue of the 105-ppt contour extends from the northern to southern hemisphere. The northern upward branch of the Hadley cell delivers CH3CC13-1aden air from the source region of the northern mid-latitudes to the upper troposphere of the southern hemisphere. The slightly positive vertical gradients of CH3CC13 in the southern hemisphere are a manifest/trion of upper tropospheric transport from the northern to the southern hemisphere. This feature has been observed in vertical profile measurements taken in the southern hemisphere [Fraser et al., 1986]. The concentration of CH3CC13 drops sharply in the stratosphere, where it is destroyed by photolysis. Figure 6 shows the monthly averaged CH3CC13 concentrations measured at the ALE sites together with results of the model simulation from July 1980 through June 1985. ':::::? :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: ß ============================================================ ===================================  Table 3b are positive for all sites. Furthermore, the standard error of the mean model bias is small compared to the bias for all sites except Cape Meares, Oregon. This suggests that the OH concentration fields are generally too low. The relatively good agreement found at Cape Meares, Oregon, is believed to be fortuitous. In an earlier study we showed that this model consistently underestimates CFC-11 by about 7 to 12 ppt at Cape Meares [Kao et al., 1992]. We attributed this underestimation to the model's inability to capture regional scale circulation features that appear to be influencing the CFC-11 observations at the Cape Meares sampling location. Therefore we expect that a good simulation of methyl chloroform would underestimate CH3CC13 at Cape Meares by an approximately proportional amount of about 5 to 8 ppt. The ---30-ppt interhemispheric gradient is underpredicted by 5.8 + 1.3 ppt.

Global OH
We concluded from our base case, case 1, that the twodimensional photochemical model globally underestimates the OH concentrations. We elected to constrain the two-dimensional monthly average OH concentration grid by the global average OH calculated from the observed trend of CH3CC13 at the ALE sites. One advantage of this approach is that global OH is constrained by the ALE observations themselves. Secondly, the effects of the recently discovered ocean sink for CH3CC13 [Butler et al., 1991 ] are also implicitly incorporated into this empirical estimate of the global and annual mean OH concentration. A simple equation balances the increasing emission rate and concentration of CH3CC13 with its chemical destruction by reaction with OH:  By this approach we constrain the global average OH by the relationships discussed above and test the accuracy of the monthly averaged relative meridional and vertical distributions of OH predicted by the two-dimensional photochemical model. For case 2, then, we uniformly increased the monthly average OH concentrations within each grid cell as used in case 1 by 25%.
The interhemispheric ratio is unchanged from that in case 1.  Table 3b. We find that the positive model biases found for Case 1 are now negative at all locations except for Adrigole, Ireland. The interhemispheric gradient of •-30 ppt remains underpredicted by •-5.8 ppt. The mean model biases are all significant at the 95% confidence level, which is to say that there is a <5% chance that the null hypothesis, which is that the model bias is zero, is true. We see that case 2 is improved over case 1 inasmuch as, for three (Ireland, Samoa, and Tasmania) of the five ALE locations, the absolute value of the mean bias was reduced from that in case 1.
The model biases for Oregon and Barbados were substantially degraded. Increasing the global mean OH reduces the atmospheric lifetime of methyl chloroform to 5.4 years, which is just within the accepted limits of 6.3 (+1.2, -0.9) years [Prinn et al., 1987]. We conclude that scaling the OH concentration grid to the global trends in methyl chloroform emissions and concentrations improves the accuracy of the simulation. However, agreement, in a statistical sense, with the observations is not satisfactory.  Figure 5)

pheric mean methyl chloroform concentration, A[CH3CC13]N/At,
has an unknown accuracy because of the proximity of the three northern ALE sites to the sources and the effects of regional circulation. This appears to be less of a problem in the southern hemisphere because the two ALE sites at Samoa and Tasmania obtain very similar values for the annual mean concentration of methyl chloroform. As the observational record increases, the interhemispheric difference in methyl chloroform growth rates will become more precisely known.    Figure 8 and Table 3  ocean. Extrapolating to the global ocean, they estimated that the total oceanic sink is ~5 to 11% of the global loss through reaction with OH.

The methodology employed in cases 2 and 3 for adjusting the [OH]o as well as [OH]N/[OH]s implicitly incorporates this sink.
We then tested how sensitive these empirical adjustments were to the existence of the ocean sink term. For this case, we elected to explicitly explore the effect of including an oceanic sink for methyl chloroform in the model. We set the ocean sink rate at 10% of the global loss rate of methyl chloroform by reaction with OH. We chose this value to explore the maximum effect suggested by Butler et al..[1991]. This sink is modeled as seasonally and spatially invariant with no dependence on wind speed. This simulation is identical in all respects to that described as case 1 (c.f. Table 3a) with the addition of the sink for methyl chloroform at the ocean surface. Figure 9 and Table 3  Our results do not resolve all of these differences but, when taken with those of Kao et al. [1992], they further demonstrate that model transport characteristics may depend significantly on model structure. We have shown that it is possible to constrain (albeit somewhat imperfectly) the interhemispheric OH ratio from observations of the interhemispheric methyl chloroform gradient. As modeling of photochemical production of OH and interhemispheric transport improves and as the ALE data record lengthens with time, the method described here will aid refinement of the interhemispheric distribution of OH.

SEASONAL CYCLES OF CH3C13
We examined the annual cycle of CH3CC13 in our model by using the simulation identified above as case 4, which includes the ocean sink for methyl chloroform. For each ALE site the seasonal cycle of methyl chloroform was calculated as described here. The interannual trends in the ALE observational record, the simulation, and model bias (the difference between the observation and the simulation) were removed by a least squares fit to a fourth-order polynomial. The 5-year mean and standard deviation of the monthly residuals for each of these variables at each ALE site are shown in Figure 10.  (Figure 1 Of) shows that, nevertheless, the phase and amplitude of the modeled seasonal cycle for the entire grid are significantly different from the observations at a single locale within the grid. As with Ireland, the model grid box containing Cape Meares includes major sources in the western United States. Likewise, grid-averaged model dynamics and chemistry can differ significantly from the regional scale dynamics and chemistry affecting a single locale within the grid.

Barbados
For Barbados (Figures 10g and 10h), the observations exhibit a minimum in the fall that is not captured by the simulation (R = -0.07). This feature is attributed to large-scale mixing of air with lower methyl chloroform concentrations from more southern latitudes during the hurricane season [Hartley et al., 1991]. Tropical disturbances such as cyclones and hurricanes are not well simulated in this, nor any, GCM. Consequently, the seasonal cycle of model bias shows a pronounced overprediction in the fall (Figure 10i).

Samoa
For Samoa (Figures 10j and 1 Ok), neither the ALE observations nor the simulation exhibit a strong seasonal cycle. However, the large variance in the residuals of the observations for December through March is a manifestation of significant interannual variability during these months. This has been attributed to the E1 Nifio-Southern Oscillation [Hartley et al., 1991]. •Itfis feature of the atmosphere is not captured by the GCM and, consequently, the correlation between the modeled and observed seasonal cycle is poor (R = -0.33). The model seasonal cycle (Figure 10k) exhibits a small minimum in the southern hemisphere summer and small maximum in the winter. The difference between these two curves (Figure 10/) is relatively featureless with no month exhibiting a statistically significant bias.

Tasmania
For Tasmania (Figure 10m), the ALE observations exhibit a weak seasonal cycle with a small maximum in October and minimum in February. The model exhibits a slightly different cycle with the maximum in July (Figure 10n). The correlation of The incorporation of a seasonally invariant and spatially constant ocean sink into the model does not significantly alter the calculated seasonal cycle of methyl chloroform. We compared the modeled seasonal cycles at the ALE locations for cases 1 and 3 (without and with the ocean sink for methyl chloroform, respectively) and found no statistically significant difference between these two cases. The measurements of ocean sink for methyl chloroform showed no dependence on sea surface temperature [Butler et'al., 1991]. However, Butler et al. note that the temperature range sampled may not have been large enough to discern an effect. They also noted several other physical processes that may influence the flux of methyl chloroform from the atmosphere to the ocean. If any of these processes are seasonally dependent, they may affect the seasonal cycle of methyl chloroform over the oceans.

Role of Chemistry Versus Transport in the Simulated Seasonal Cycle of Methyl Chloroform
Comparison of the modeled seasonal cycle of CFC-11 and CH3CC13 separates the effects of model transport versus model OH on the seasonal cycles of CH3CC13. We use the correlation coefficient (R) between the simulated monthly average CH3CC13 and CFC-11 seasonal cycles to examine the relative importance of atmospheric transport versus atmospheric chemistry in influencing the seasonal cycle of CH3CC13. If CH3CC13 were not destroyed by reaction with OH, the value of R would be unity at all locations because the model source distribution and transport are identical to those used for simulating CFC-11. Departures from unity are a measure of the influence of the OH seasonal cycle on the CH3CC13 seasonal cycle. Figure 11 shows the correlation between the seasonal cycles of CH3CC13 and CFC-11 in 4.5 ø latitude bands at the surface. In the northern hemisphere, the average value of R is 0.75. This indicates that the modeled seasonal cycle of CH3CC13 in the northern hemisphere is strongly influenced by atmospheric transport. The value of R declines across the equator to an average value of 0.1 in the southern hemisphere. We interpret this as evidence that the seasonal cycle of methyl chloroform is weakly influenced by model transport and is, consequently, more strongly influenced by the phase and amplitude of the modeled seasonal cycle of OH.

SUMMARY OF FINDINGS
This model of the global distribution of CH3CC13 simulates the 5-year record of observations made at the five ALE sampling sites, generally to within 5% of the observed annual mean concentration. The calculated average global lifetime, 5.7 _+ 0.3 years, is within the range [6.3 (+1.2, -0.9) years] estimated from ALE measurements [Prinn et al., 1987]. The estimated global mean OH concentration is 6.5 _ 0.4 x 105/cm '3. The reaction rate between OH and methyl chloroform has recently been remeasured and revised downward from the value used in this study [Talkudar et al., 1992]. The effect of this revision is to increase The model confirms that CH3CC13 predominantly released at the surface in the northern hemisphere is quickly mixed throughout the northern troposphere. Transport to the southern hemisphere occurs preferentially between 600 and 300 mbar as part of the Hadley cell circulation. The interhemispheric time constant is -6 months with -270 kt/year being transported from the northern to southern hemisphere. But these values are sensitive to the interhemispheric gradient of OH.
The seasonal cycles of methyl chloroform measured at the ALE sites are not fully captured by the model. Some of the reasons for the model's inability to accurately simulate the seasonal cycle include variability in the observations due to phenomena such as regional scale (i.e., subgrid) circulation patterns, tropical storms, and E1 Nifio-southern-oscillation events that are not simulated by the parent GCM. Our two-dimensional model of the distribution of OH is deficient in its treatment of nonmethane hydrocarbons. The assumed spatial and temporal constancy of the ocean sink for methyl chloroform has not been confn'med by observations.
In the northern hemisphere, the seasonal cycles of CH3CC13 in the model are mainly controlled by the effects of atmospheric dynamics in transporting CH3CC13 from the source regions to the sampling locations. How. ever, in the southern hemisphere, the modeled seasonal cycles of CH3CC13 are more strongly influenced by the amplitude of the seasonal cycle of OH.
Three-dimensional simulations of the global biogeodynamics of atmospheric roethane, nonmethane hydrocarbons, and other trace gases depend on an accurate representation of the global OH field. Models of this type, from which the OH distribution can be inferred, will continue to be necessary until experimental methods , are developed that can map the global distribution of OH and monitor the seasonal concentration changes and annual trends.