A modelling study of tropospheric distributions of the trace gases CFCl3 and CH3CCl3 in the 1980s

Interhemispheric transport is a key process affecting the accuracy of source quantification for species such as methane by inverse modelling, and is a source of difference among global three-dimensional chemistry transport models (CTMs). Here we use long-term observations of the atmospheric concentration of long-lived species such as CH3CCl3 and CFCl3 for testing three-dimensional chemistry transport models (CTMs); notably their ability to model the interhemispheric transport, distribution, trend, and variability of trace gases in the troposphere. The very striking contrast between the inhomogeneous source distribution and the nearly homogeneous trend, observed in the global ALE/GAGE experiments for both CH3CCl3 and CFCl3 illustrates an efficient interhemispheric transport of atmospherically long-lived chemical species. Analysis of the modelling data at two tropical stations, Barbados (13°N, 59°W) and Samoa (14°S, 124°W), show the close relationship between inter-hemispheric transport and cross-equator Hadley circulations. We found that cross-equator Hadley circulations play a key role in producing the globally homogeneous observed trends. Chemically, the most rapid interaction between CH3CCl3 and OH occurs in the northern summer troposphere; while the most rapid photolysis of CH3CCl3 and CFCl3, and the chemical reactions between CFCl3 and O(1D), take place in the southern summer stratosphere. Therefore, the cross-equator Hadley circulation plays a key role which regulates the southward flux of chemical species. The regulation by the Hadley circulations hence determines the amount of air to be processed by OH, O(1D), and ultraviolet photolysis, in both hemispheres. In summary, the dynamic regulation of the Hadley circulations, and the chemical processing (which crucially depends on the concentration of OH, O(1D), and on the intensity of solar insolation) of the air contribute to the seasonal variability and homogeneous growth rate of observed CH3CCl3 and CFCl3.


Introduction
The atmospheric transport of long-lived chemical species by large-scale dynamics is the most important process in the redistribution of trace gases in the atmosphere. The long-range transport of pollutants away from a source area, by the large-scale general circulations provides a key mechanism for non-indigenous pollution events. Hence, it is crucial to study the performance of global three-dimensional (3-D) CTMs with respect to their ability to model transport processes. One way of testing CTM's performance is by studying its ability to model long-lived trace gases, where emission sources are well known and long-term measurements exist, such as chloro¯uorocarbons (CFCs).
Trace gases of anthropogenic origin, which contain chlorine and bromine atoms can greatly perturb stratospheric ozone concentrations (e.g. see Solomon, 1999). Most of these man-made chemical species are longlived. For example, the inferred lifetime for CFCl 3 is about 43±50 years and 4.1±5.4 years for CH 3 CCl 3 (Kaye et al., 1994). The longevity of these halogenated species greatly enhances the probability and magnitude of their stratospheric in¯ux. Once they reach the middle atmosphere, they will be destroyed by ultraviolet photolysis and by chemical reaction with O( 1 D), releasing active chlorine and bromine atoms into the stratosphere. The subsequent threat to ozone can be examined by detailed examination of catalytic chemistry involving chlorine or bromine-containing compounds (e.g. Wayne, 1993).
The ALE/GAGE experiment, which began in 1978, provides a continuous long-term record of surface observations of the concentrations of the atmospheri-cally long-lived chemical species (Prinn et al., 1983(Prinn et al., , 1992Cunnold et al., 1994). The measurements include two gases which have both anthropogenic and biogenic sources (CH 4 and N 2 O) and ®ve gases of anthropogenic origin (CFCl 3 , CF 2 Cl 2 , CF 2 ClCFCl 2 , CH 3 CCl 3 , and CCl 4 ). The use of this long-term surface observational record of CH 3 CCl 3 and CFCl 3 from global ALE/GAGE network stations, combined with their anthropogenic emissions, which are reasonably well known and the well-established chemistry (CFCs, HFCs, and HCFCs) by laboratory studies, provide a sound platform for a modelling study. The purpose of this modelling study can be summarized by the following: to derive the threedimensional distributions of CFCl 3 and CH 3 CCl 3 ; to study the atmospheric transport processes once these species have been released into the air; and to investigate their trend and variability.
If we consider the following two facts: ®rst there is a near globally homogeneous trend in the concentration of CH 3 CCl 3 and CFCl 3 observed during the ALE/GAGE experiments; second, nearly 95% of the reported sources of those two species are in the Northern Hemisphere with maxima centred around mid-latitudes. Then, the very striking contrast between inhomogeneous source distribution and nearly homogeneous observed trend in the global ALE/GAGE stations, strongly suggests an ecient inter-hemispheric transport of anthropogenic chemical species. This indication of ecient interhemispheric transport also emerged from another modelling study of 85 Kr. (MuÈ ller and Brasseur, 1995).
The variability of the observed surface concentration of chemical species by atmospheric transport processes has been pointed out by other studies (Prather et al., 1987;Prinn et al., 1992;Hartley et al., 1993). The role of vertical transport by cloud convection on the observed seasonal cycle of the surface concentration has been noted by the modelling study of CFCl 3 and CF 2 Cl 2 over the continental areas (Prather et al., 1987). Cloud convection also plays an important role with respect to interhemispheric transport of trace gases (Gilliland and Hartley, 1998). The variation of the northern Atlantic storm tracks has a direct in¯uence on the observations at Mace Head in Ireland (Hartley et al., 1993), and the ENSO event could have a strong in¯uence of the in¯ux of Northern Hemispheric air in the boreal winter (Prinn et al., 1992). Hence a close relationship between speci®c atmospheric transport process and the observed variability of surface concentration has long been inferred. The focus of this modelling study is to speci®cally investigate the observed variability of surface concentration by a variety of atmospheric transport processes. We will concentrate our analysis on the process of interhemispheric transport, because it is the key factor for the variability and globally homogeneous trends of concentration of chemical species.

Model description
The 3-D CTM (Stockwell and Chipper®eld, 1999) used in this study includes parametrizations for transport by subgrid scale processes and chemical reactions which lead to the destruction of CH 3 CCl 3 and CFCl 3 . The parametrizations of sub-grid scale transport in the model include cumulus convection, atmospheric boundary layer processes and sub-grid scale diusive transport.
The cumulus parametrization method (Tiedtke, 1989) used in this study is a mass¯ux scheme (Stockwell and Chipper®eld, 1996). The Tiedtke (1989) cumulus scheme considers the cloud model with the following components: in-cloud updraught, downdraught, large-scale subsidence induced by cumulus motion, entrainment and detrainment processes into in-cloud updraught and downdraught air. The scheme therefore considers only the cumulus scale updraughts, as well as downdraughts, but neglects mesoscale circulations. The closure of the scheme is based on the observational evidence that typical cloud ensembles prevail under certain synoptic conditions. For example, the pre-existence of large-scale low level convergence could moisten and destabilize the environment at low levels, so that small-scale thermals can easily reach the level of free convection and produce deep cumulus convection (Holton, 1992). The key point for the whole mechanism to work is that the low-level large-scale convergence produces a convergence of moisture, which by increasing the equivalent potential temperature in the boundary layer makes the environment more favorable for the development of the cumulus convection.
The transport within the atmospheric boundary layer (ABL) has been parametrized by the local-K scheme (Louis, 1979;Heimann, 1994;Holtslag and Boville, 1993) plus a simpli®ed boundary layer height calculation to enhance vertical transport of chemical species by large eddy transport in the atmospheric boundary layer.
The assumption of the calculation of boundary layer height used in this study is based on the observations of the diurnal variation of boundary layer height (Driedonks, 1982;Deardor, 1972). The lowest boundary layer height is assumed to be 1 km, and a sine wave variation with amplitude of 2 km is added according to the time of the day. The maximum boundary layer height is therefore 3 km at noon with a minimum of 1 km at midnight. This amplitude is comparable with that calculated by the nonlocal scheme (Holtslag and Boville, 1993) where the diurnal amplitude ranges from 0.5 km to 4 km according to dierent geophysical locations. The purpose of the boundary layer height is to give an estimated vertical scale of dry convection, which is then used to provide an additional mechanism for vertical transport of species within the ABL. The eect of horizontal diusive mixing induced by the cumulus convection has been applied to the cloudy model column to account for the enhanced sub-grid scale horizontal entraintment and detrainment of air into the cloud (Stockwell and Chipper®eld, 1999), which it is not possible to properly represent in the low-resolution model, and which was shown to aect the interhemispheric tracer transport over the intertropical convergence ozone (Prather et al., 1987).
A simple gas phase chemistry scheme, includes the reaction between CH 3 CCl 3 and OH, and CFCl 3 with O( 1 D) is shown in Table 1. This chemical scheme and photolysis of the two trace species in the stratosphere are the major sinks for both species. Prescribed two dimensional ®elds of photolysis rates, OH, and O( 1 D), which take into account the seasonal variation of photolysis rates, OH, and O( 1 D), have been used in the chemical calculation. These ®elds were obtained from the Cambridge 2D model which includes a detailed description of tropospheric chemistry (Law and Pyle, 1993) and are updated every 10 days.
The emissions data from 1981 to 1990 of Midgley (1992) have been used here to construct the surface sources of CH 3 CCl 3 and CFCl 3 from 1979 to 1988. The reported emissions for CFCl 3 represents around 85% of the actual emission into atmosphere, therefore the emissions used in the model for CFCl 3 have been scaled by a factor 1/0.85 of the value from Midgley (1992). The actual emission used is summarized in the ®rst and second column of Table 2. For comparison, we also list the value estimated from Kaye et al. (1994). The geographical distribution of the emissions is similar to the one used by Hartley and Prinn (1993).

Experiments
An initialization procedure similar to Golombek and Prinn (1986) has been used in this study, the dierence being is that the vertical pro®les from the 2D model have been used in combination with the observed ALE/ GAGE data. The observed value of CH 3 CCl 3 and CFCl 3 in ALE/GAGE stations is ®rst interpolated, using the cubic spline method, to the model Gaussian latitude and the 2D ®elds are then normalized in accordance with the interpolated data in the lowest model layer. The normalization factor is then used to modify the value vertically. The advantage of this initialization method is that it maintains the original 2D vertical pro®le and also assimilates the observed data.
There is no exchange of species¯ux across the lower and upper model boundaries. The emission is treated as being added uniformly into the lowest model layer. The assumption is that the lowest model layer, about 70 m deep, is very turbulent, and with a time step of 3 h, it is expected that there will be quick mixing after the release of species from the surface through the lowest model layer. The dynamic ®elds, which are relative vorticity, divergence, temperature, speci®c humidity and surface pressure from UGCM's AMIP experiment have been used as an input to the CTM (Gates, 1992;Slingo et al., 1994). The 10 years AMIP archive, which is at T42 resolution (approximately 2:5 by 2:5 ), is ®rst processed to T10 resolution (11:25 by 11:25 ) before being incorporated into the CTM. A complete set of 10 years of dynamical ®elds at T10 resolution is then used in the subsequent integration, which saves both CPU time and memory space. The time step for integration is 3 h. The model resolution is 32 longitude grids, 16 Gaussian latitude grids, and 19 layers in the vertical. The input UGCM data is the one generated from AMIP with an annual SST (sea surface temperature) change during the 10 year integration.

Model results
One of the most striking features from the observations is the near homogeneous trend of the concentrations of CH 3 CCl 3 and CFCl 3 shown in global ALE/GAGE network stations. (See Figs. 1, 2 and the second and fourth column of Table 3). The near homogeneous global trend of chemical species indicates ecient interhemispheric transport in the tropics. Further evidence which supports an ecient inter-hemispheric transport is also found in the modelling study of 85 Kr (see Fig. 4 of MuÈ ller and Brasseur, 1995). Approximately 95% of CFCl 3 and CH 3 CCl 3 emissions come from the Northern Hemisphere industrialized countries. The source of 85 Kr comes exclusively from pressurized water nuclear reactors in the Northern Hemisphere. Table 3 summarizes the results from model and observations. At the end of 10 years of integrations, for CFCl 3 , the dierence between the model and observed level is 32.39 pptv in Ireland (Mace Head) and 25.75 pptv in Tasmania. The meridional gradient between Ireland and Tasmania is about 0.36 pptv degree À1 from the model and 0.28 pptv degree À1 from observations. For CH 3 CCl 3 , the dierence between model and observation is 64.95 pptv in Ireland and Where k CH3CCl3OH 1:75 Â 10 À12 expÀ1550=T a cm 3 sec À1 and k CFCl3O 1 D 2:3 Â 10 À10 b cm 3 sec À1 a DeMore (1992) b Law and Pyle (1993)  33.26 pptv in the Tasmania. The meridional gradient between Ireland and Tasmania is 0.71 pptv degree À1 from the model compared with 0.47 pptv degree À1 from observations. The modelled meridional gradient between Ireland and Tasmania for CH 3 CCl 3 (0.71 pptv degree À1 ) is greater than CFCl 3 (0.36 pptv degree À1 ); while this dierence is less pronounced in the observations where the meridional gradient is 0.47 pptv degree À1 for CH 3 CCl 3 and 0.28 pptv degree À1 for CFCl 3 . Figures 1 and 2 show the results from the 10-year integration for both chemical species and the comparison with measurements. The model agrees well with the measurements in the ®rst 3 years. After that, the gap between the model and measurement becomes wider (a topic we will return to later). It is clear that there is a seasonal cycle for CH 3 CCl 3 , which re¯ects the strong seasonality of the OH ®eld. The model captures this seasonality well, however, the model overestimates the magnitude of the seasonal variability observed in Oregon and underestimates it in Tasmania.
Since CFCl 3 is chemically inactive in the troposphere, its time series shown in Fig. 2 re¯ects the major contributions due to atmospheric transport processes and year-to-year surface emission strength. There are no pronounced seasonal cycles observed for this species,   however, a noticeable seasonable cycle was observed in Oregon, although relatively weak compared with CH 3 CCl 3 . Hence, this con®rms that a possible cause for the pronounced seasonal cycle observed for CH 3 CCl 3 is due to the active chemical sinks in the troposphere. The pronounced seasonal cycle is clearly seen at sites such as Oregon, Barbados, and Tasmania for CH 3 CCl 3 . While no clear seasonal cycle is seen for observed CFCl 3 , the model shows a distinctive seasonal cycle for this species at Oregon and Barbodos. This more pronounced CFCl 3 seasonal cycle in the model than the measurements indicates an additional transport eect can also be responsible for the modelled CH 3 CCl 3 seasonal cycle at these sites. Hence, in addition to the modelled CH 3 CCl 3 seasonal variability driven by the OH seasonal variability, the transport process also contributes to the higher magnitude of seasonality than the measurements at these sites. The 10-year average concentration for the lowest model layer is shown in Fig. 3 for both CH 3 CCl 3 and CFCl 3 . The bold numbers within each ®gure show ALE/ GAGE observed concentrations. The maintenance and seasonal movement of inter-tropical high meridional gradient zones of chemical species is clearly seen in the ®gure. The statistics clearly illustrate the correlation between high concentration areas and high emission sources, which are centred around North America, Europe and Japan. Higher surface concentrations in winter than in summer indicates less active vertical transport of surface emissions into the free troposphere during the winter months.
The seasonal change of surface concentration also reveals a pattern of temporal and spatial variability of chemical species arising from a combination of atmospheric processes. The tongue of lower concentration from low latitude into high latitude and the intrusion of high concentration from high latitude into low latitude in both the eastern North Paci®c Ocean and Atlantic Ocean shows the prevailing in¯uence of the subtropical high in that area during the northern summer season. These are two of three major intrusions of high concentrations of these species from high latitude into low latitude. The other major intrusion of high concentrations from high latitude into low latitude occurs in eastern Asia during the winter season, as seen from Fig. 3. The intense continental Mongolian or Siberian High developed during winter usually produces the well-known northeast monsoon season, a prevailing strong wind which produces the southward intrusion pattern.
The zonally averaged concentration at each model layer for CH 3 CCl 3 and CFCl 3 for the summer and winter season is shown in Fig. 4. The distinct contrast between summer and winter clearly shows the seasonal variation of the Hadley cell circulation in the tropical region. The long-lived species are therefore suitable candidates for seeing the underlying dynamics, a useful quality which also belongs to some dynamic tracers such as isentropic potential vorticity under adiabatic isentropic motions (Hoskins et al., 1985). The signi®cant dierence in the vertical gradient of concentration shows the approximate position of the tropopause. The highly strati®ed structure in the vertical direction of concentration in the stratosphere is distinct from the vertically well mixed structure in the troposphere, which has a stronger meridional gradient. These two distinct types of structure of concentration demonstrate the dierent mixing processes within each region.
The wind ®eld of the lowest model layer is shown in Fig. 5 for the winter and summer season respectively. The circulation pattern which shows the strength and position of the subtropical high in the Eastern North Paci®c Ocean and Atlantic Ocean is clearly seen. These are two major persistent circulation systems which have a direct impact on the transport of species from their surrounding major emission sources. The wind pattern which shows the southwest Asian monsoon is also clearly seen in the area centred around 90 E. The wind pattern in this area completely changes from southwest to northeast in the winter season, which is the northeast monsoon season. The annual reversal of the wind direction results in this area being one where active inter-hemispheric air exchange take place.

Budget analysis
In previous discussions we have shown that the nearly homogeneous growth of surface concentrations (both modelled and measured) at a network of global monitoring locations with respect to the very inhomogeneous surface source distribution between Northern and Southern Hemispheres indicates an eective interhemispheric transport in redistributing species. Interhemispheric transport is a key process aecting the accuracy of source quanti®cation for species such as methane by inverse modelling (Houweling et al., 1999), and is a source of dierence among global threedimensional chemistry transport models (CTMs) (Denning et al., 1999). Vay et al. (1999) showed that the CO 2 distribution in the Southern Hemisphere appeared to be largely determined by the interhemispheric transport as air masses with depleted CO 2 levels characteristic of the Northern Hemisphere were frequently observed south of the Intertropical Convergence Zone (ITCZ). Quay et al. (1999) observed that a high d 13 C value in the Southern Hemisphere compared with the Northern Hemisphere was a result of interhemispheric transport. Hence the study of the role of the interhemispheric transport is very important in characterizing both global pollutant redistribution and source inversion (see also Rhoads et al., 1997). In this section, based on ten years of modelled budget, we examine those key interhemispheric transport processes, namely large-scale advection and cloud convection.
If we consider only the eects of advection, cloud convection, and chemistry, the continuity equation for a chemical species i can be written as Here n i is the number density of species i, and the ®ve terms on the right-hand side of Eq. (1) are the eects of chemistry, cloud convection, horizontal advection by zonal wind u in the east-west direction x, meridional advection by meridional wind v in the north-south direction y, and vertical advection by vertical wind w in the vertical direction z. In the following budget calculation, the eect of transport due to large-scale advection (primarily horizontal and meridional advection), chemistry, and cloud convection during each time step Rt were calculated on each box and stored throughout the 10 years of integration. These 3-D budget terms were then weighted by the volume of the air mass in each model box to give a change of mass (kg) for each species i at each corresponding location (altitude, latitude, and longitude) of the model grid box. Figure 6 shows a budget analysis for CH 3 CCl 3 and CFCl 3 due to large-scale transport at the northern tropical site of Barbados. The seasonal variability of transport is clearly seen in this tropical station. The major mean meridional circulation in the tropics is dominated by the Hadley circulation (Hastenrath, 1985;Palmen and Newton, 1969). In the northern summer, the Hadley circulation in the north is weak and the accompanying mass circulation is about Fig. 5a, b. A 10-year average of surface wind ®elds in the lowest model layer for a the summer and b winter seasons 8 Â 10 17 kg yr À1 ; while the Southern Hemispheric Hadley circulation with its extension across the equator from the south is most vigorously developed and the accompanying mass circulation is about 6 Â 10 18 kg yr À1 (Hastenrath, 1985). In the northern winter, the Hadley circulation in the south is weak and the accompanying mass circulation is about 9 Â 10 17 kg yr À1 ; while the Northern Hemispheric Hadley circulation with its extension across the equator from the north is most vigorously developed and the accompanying mass circulation is about 7 Â 10 18 kg yr À1 (Palmen and Newton, 1969).
The domain of in¯uence by the Southern Hemispheric Hadley circulation in the northern summer is between 15 N and 25 S; while the domain for the Northern Hemispheric Hadley circulation in the northern winter is between 28 N and 5 S. The pattern of the total meridional transport hence reveals a seasonal variability of inter-hemispheric transport by crossequator Hadley cell circulations in the tropics around the year. The negative regions in Fig. 6 indicates the mean meridional transport of chemical species from low-concentration regions to high-concentration regions, while the positive regions indicate transport from high-concentration regions to low-concentration regions.
In the northern winter, the inter-hemispheric transport is dominated by the Northern Hemispheric Hadley circulation. This meridional circulation transports species from northern mid-latitude high chemical source regions to northern tropical relatively low concentration regions. The total eect of this meridional transport produces positive tendencies in Barbados. In the northern summer, the inter-hemispheric transport is dominated by the Southern Hemispheric Hadley circulation. This cross-equator meridional circulation transports species from Southern Hemispheric low concentration regions, across the equator, to northern tropical relatively high-concentration regions. The total eect of this cross-equator meridional transport produces negative tendencies in Barbados.
The dierent transport pattern between winter and summer is mainly caused by the seasonal variation of cross-equator Hadley circulations. This cross-equator circulation produces net seasonal southward or northward inter-hemispheric transport of chemical species. In the northern summer, Barbados is under the in¯uence of the southern cross-equator hemispheric Hadley circulation, while in the northern winter, Barbados is under the in¯uence of the northern cross-equator Hadley circulation.
The vertical transport by cloud convection in Fig. 7 shows a completely dierent pattern of vertical transport between Barbados and Samoa. In Barbados, the most vigorous vertical transport appears to be in the northern winter. The vertical dipole pattern with positive regions from 2 km to 10 km and negative regions below 2 km, illustrate an active upward transport of chemical species from the lower troposphere into the free troposphere. In Samoa, the most vigorous vertical transport appears to be in the Southern Hemisphere winter. The vertical dipole pattern with positive regions below 2 km and negative regions from 2 km to 7 km, illustrate an active downward transport of chemical species from the upper troposphere into the lower troposphere.
The very delicate phase relationship between the active vertical dipole transport pattern by cloud convection and positive transport pattern by total meridional transport in Barbados illustrates an eective combined transport process between large-scale advection and cloud convection. This kind of combined transport process provides an active mechanism for chemical species to maintain an ecient upward mass ux in the northern site, Barbados, and downward mass ux in the southern site, Samoa (Dickerson et al., 1987). The overall pattern of transport by large-scale advection is summarized in Fig. 8 for each hemisphere. These hemispheric averaged results provide a clear distinction between Northern and Southern Hemispheres. Chemically, the Northern Hemisphere accounts for about 95% of the total CH 3 CCl 3 and CFCl 3 Fig. 6a, b. Height-time plots of 10-year average of a CH 3 CCl 3 and (b) CFCl 3 budget (in units of 10 6 Â kg year À1 layer À1 ) at Barbados site due to large-scale transports. The positive contours indicate that the large-scale transport contributes to the increase in both species at this site; while negative contours (shaded areas) represent the decrease in species at the same site emissions. Dynamically, the chemical species derived from the northern emissions are transported to the Southern Hemisphere by inter-hemispheric transport which results in an equivalent growth rate of observed chemical concentration compared with the north. The eect of inter-hemispheric transport is clearly seen in the Southern Hemisphere. The positive pattern illustrates the net transport of chemical species from higher concentration areas to lower concentration areas. This kind of down-gradient transport of species provides an essential mechanism for the maintenance of globally near homogeneous trends of observed surface concentration. The positive patterns seen in the Northern Hemispheric winter demonstrate the active season for the southward transport of high-concentration air from northern high latitudes. The negative patterns seen in the Northern Hemispheric summer demonstrate the active season for the northward transport of lowconcentration air from southern low latitudes. This cross-equator northern Hadley cell circulation may re¯ect the contributions of the monsoons of South Asia and Africa (Hastenrath, 1985). Figure 9 summarizes the averaged eects of vertical transport by cloud convection in the Northern and Southern Hemisphere. We found this result is similar to the convective pattern of Fig. 6. The cloud convection produces upward transport of chemical species in the Northern Hemisphere and downward transport of chemical species in the Southern Hemisphere. The upward transport of chemical species from the lower troposphere into the middle and upper troposphere could have a profound impact on the chemistry of CH 3 CCl 3 and CFCl 3 . The wind speed is higher in the upper troposphere and hence the distance of long-range transport is longer than when both species are constrained in the lower troposphere. For CH 3 CCl 3 , the major chemical reaction is with tropospheric OH radicals. The vertical structure of tropospheric OH is in¯uenced by the competing factors of mixing ratio of water vapour (which decreases rapidly with height), the number density of O 3 (which has a maximum between 600 and 800 hPa), and the magnitude of photolysis rate constants J and reaction rate constants k (which are determined by the solar irradiance, temperature, and clouds). The resulting maximum rate of OH production is at around 800 mbar (Spivakovsky et al., 1990). Hence, the budget of CH 3 CCl 3 in the troposphere could be very sensitive to the variation in vertical structure of OH modi®ed by cloud convection (Thompson et al., 1994).
The chemical reaction for CH 3 CCl 3 in the Northern and Southern Hemisphere (upper panel of Fig. 10) illustrate an active sink due to reaction with by OH Fig. 7a±d. Height-time plots of a, b CH 3 CCl 3 and c, d CFCl 3 budget due to cloud convection at both Barbados (a, c) and Samoa (b, d) sites in the troposphere, and sink due to ultraviolet photolysis in the lower stratosphere. The amplitude of photolysis in the lower stratosphere is consistent with the season, which is a direct response to the seasonal variation of the intensity of incoming solar radiation. Such agreement between season and the strength of photolysis can also be seen from the chemical reaction for CFCl 3 (lower panel of Fig. 10). The major sink for CFCl 3 is exclusively stratospheric photolysis and reaction with O 1 D.
The maximum rate of extinction for CH 3 CCl 3 by the OH radical is 41 kT yr À1 in the northern summer, and 27 kT yr À1 in the southern summer (northern winter). The area for this chemical reaction is located at the altitude between 1 km and 6 km, with maximum amplitude occurring at the height of 3.5 km. The stratospheric photolysis sink is 17 kT yr À1 in the northern summer, and 18 kT yr À1 in the southern summer. The Northern Hemisphere has higher rates of extinction by reaction with OH than the Southern Hemisphere; while the Southern Hemisphere has slightly higher stratospheric sinks for CH 3 CCl 3 . This shows that the active northern summer removal of CH 3 CCl 3 by OH helps to reduce the total amount of that species being transported into the lower stratosphere. Figure 11 shows the mean meridional structure of vertically integrated chemical sinks for both species. Both species have maximum sinks between 60 N and 60 S. The extinction rate in the southern summer (24 kT yr À1 ) is higher than in the northern summer (19 kT yr À1 ) for CFCl 3 . This is consistent with our previous argument of stronger solar insolation at the top of the Southern Hemisphere. The tropospheric sink of CH 3 CCl 3 by OH is closely related to factors like temperature, UV irradiance, clouds, H 2 O, CO, O 3 , CH 4 , and NO t (de®ned as NO NO 2 NO 3 2N 2 O 5 HNO 2 HNO 4 ) (Spivakovsky et al., 1990), and stratospheric O 1 D which is closely related to the stratospheric O 3 concentration and solar insolation.

Further considerations
Based on the analysis shown in this study, we are able to make the following statements. First, the model emissions could be too high for both species. It would seem from Figs. 1 and 2 that the emissions used for the ®rst few years of the model integration are reasonable. But that subsequently, the modelled levels have tended to be too high, resulting in the dierence between model and Hence the year-to-year emissions during the 1979±1988 period are very likely to be lower than used here. For CH 3 CCl 3 , its trends also indicates that modelled levels are too high compared with measurements. Hence, its emissions used here may also be too high. Figure 12 shows a comparison of emissions used in this study with estimated emissions, and emissions used in Krol et al. (1998). The estimated emissions are calculated based on: (1) the observed trend of 90.34 pptv for CFCl 3 and 57.51 pptv for CH 3 CCl 3 between 1979 and 1988 within the troposphere (de®ned as the region below 250 hPa); and (2) the estimated lifetime of CFCl 3 (43±50 years) and CH 3 CCl 3 (4.1±5.4 years). For CFCl 3 , it is clear that the model employed emissions which were about 26%±32% higher than estimated emissions (corresponding to 35±41 pptv increase in CFCl 3 ); while for CH 3 CCl 3 , the model employed emissions which were about 4%±19% higher (corresponding to 13±52 pptv increase in CH 3 CCl 3 ). The modelled averaged concentrations at the end of the 10-year integration are about 23 pptv and 35 pptv higher than observed levels for CFCl 3 and CH 3 CCl 3 , respectively. It is clear that the emissions used here could be too high, however, other factors should be considered.
Second, the externally prescribed 2-D OH, photolysis rates, and O 1 D ®elds may be one of the uncertainties for long-term model integrations for species which are strongly in¯uenced by the temporal and spatial variations of OH. It is somewhat unrealistic for OH concentrations to be constant for 10 years (see also Krol et al., 1998). In addition, though the model shows pronounced seasonal variations of CH 3 CCl 3 at Oregon and Barbados as compared with the measurements, no pronounced seasonal CH 3 CCl 3 variations are seen at Tasmania. It may be that the 2-D OH ®eld is poorly prescribed in the Southern Hemisphere, or that there are sources in the Southern Hemisphere which are not included here.
Third, there may be inadequate model representation of air mass exchange between the lower stratosphere and the upper troposphere. Considering the major sinks of CFCl 3 calculations show that the loss of CFCl 3 can only occur ®rst by the photolysis of CFCl 3 ; and second by the gas phase reaction between CFCl 3 and O 1 D. Let us note two facts: ®rst, the trend of observed surface concentration is nearly the same from north to south; second, almost all of the sources are coming from the north. Hence the observed trend of surface concentration in the Southern Hemisphere is clearly coming from the Northern Hemisphere. The processed air reaching the Southern Hemisphere must clearly experience Fig. 9a±d. The same as in Fig. 8 but for cloud convective transport photolysis and gas phase chemical reactions in the lower stratosphere. Therefore the strength of the transport between troposphere and lower stratosphere and the amount of the air taking part in the chemical reactions in the lower stratosphere will directly in¯uence the amount of air to be processed by O 1 D and photolysis in the stratosphere. Since the magnitude of the increase in modelled averaged concentrations at the end of the 10-year integration (23 pptv) is lower than the magnitude of the increase due to higher emissions used in the model (35±41 pptv), the model chemical sinks in the lower stratosphere are likely to be faster than in reality. Finally, there are possible oceanic sinks for both species. There is already evidence of both CH 3 CCl 3 (Tie et al., 1992) and CFCl 3 in the seawater (e.g. Krol et al., 1998). There are no oceanic sinks for either species considered here. Notice that while the amounts of CFCl 3 are traceable in the ocean, the amount dissolved in the ocean is likely too small to be a signi®cant sink for atmospheric CFCl 3 .

Summary
The global ALE/GAGE network monitoring stations provide a continuous long-term record on the atmospheric concentration of anthropogenic chemical spe-cies.The observational results show some very important properties of man-made chemical species once they have been released into the atmosphere. The most interesting ®ndings from the observations are the globally homogeneous trend (growth rate) and variability (spatial and temporal) of chemical species.
The observations reveal a generally growing surface concentration in the northern winter; while in the northern summer the surface concentration generally decays. Given that the rate of emission is the same within each year, only chemical reactions and dynamical transport could produce northern winter higher atmospheric burdens of CH 3 CCl 3 and CFCl 3 than in the northern summer.
The dynamics of the change (trend and variability) of the chemical concentration is closely correlated to the following factors. Chemically, the variation of concentration is driven by the chemical reaction which is a function of variational concentration of OH, O 1 D, and photolysis which is a function of variational intensity of solar insolation at the top of the atmosphere. During the northern winter, the chemical reactions between CH 3 CCl 3 and OH are slower due to the lower tropospheric OH concentrations, and the reaction between CFCl 3 and O 1 D is slower due to the lower stratospheric O 1 D concentrations, and ®nally the lower stratospheric photolysis rate for both species is slower due to the smaller solar insolation at the top of the atmosphere in the winter. Dynamically, the variation of hemispheric concentration is governed by the variational inter-hemispheric transport which is dominated by the cross-equator Hadley cell circulations. They regulate the strength of inter-hemispheric transport in dierent seasons.