Atmospheric chemistry: response to human influence

Present understanding of global atmospheric chemistry is reviewed. Models are presented and compared with a wide range of atmospheric observations, with emphasis on the stratosphere. In general, excellent agreement is found between the calculated and observed distributions of long lived trace gases. The abundances of many shorter lived species are also satisfactorily reproduced, including NO2, HNO3, O, O3, OH and ClO. Discrepancies between theory and observation are examined and their significance assessed. The influence of human perturbations due to combustion, agriculture and chlorocarbon releases is discussed with emphasis on O3. Uncertainties associated with present models are highlighted. Combustion related releases of CO cause a decrease in the abundance of tropospheric OH with consequent increase in the concentrations of CH4, H2, CH3Cl and other halocarbons. CO emissions may become sufficiently large during the next century to induce substantial increases in tropospheric ozone on a global scale. Recombination of nitrogen fixed by agriculture and combustion may lead to an enhanced source of atmospheric N2O with a related impact on stratospheric NOx. Chlorocarbon industry provides an important source of stratospheric chlorine, and enhanced levels of stratospheric Clx and NOx may cause a significant reduction in the abundance of atmospheric O3, by as much as 10% during the next century. Perturbations due to various anthropogenic activities interact in a nonlinear fashion and the influence on atmospheric chemistry is correspondingly complex.


INTRODUCTION
Atmospheric chemistry has enjoyed a period of remarkable growth over the past decade. Interest has been stimulated mainly by concern that a variety of human influences might lead to detectable change in the abundance of stratospheric ozone.
Advances have occurred in all important areas of the subject. New techniques have been developed for detection of reactive species at exceedingly low concentration. They have been applied in the laboratory to study a wide range of important reactions. Corresponding developments have provided a wealth of new data on the concentration of important atmospheric species including CH 4 (Ehhalt 1974), CO (Seiler 1974;Volz, Ehhalt, Heidt & Pollock 1976), NO (Ridley, Bruin, Schiff & McConnell 1976), N0 2 (Noxon 1975;Noxon, Whipple & Hyde 1977), HN0 3 (Murcray et al. 1975) N 2 0, (Ehhalt, Heidt, Lueb & Pollock 1975b;Schmeltekopf et al. 1977), OH (Anderson 1976;Burnett 1976), Cl and ClO (Anderson, Margitan & Stedman applicable in 1974, 5 x 10 5 t (Cl) per year, was predicted to cause a 7% reduction in 0 3 at steady state (National Academy of Sciences 1976). Nitrous oxide formed by denitrification of agriculturally related nitrogen could lead to reductions in 0 3 as large as 20% by the end of the next century (McElroy 1976;McElroy et al. 1977).
The calculations on which these estimates were based employed a number of simplifying assumptions. With two exceptions, a single calculation described by the National Academy of Sciences (1976) and a recent study by Kurzeja (1977), mathematical models used diurnally averaged values to describe solar radiation. They omitted complexities associated with nonlinear interactions of perturbations.
It is important that the model used to assess perturbations should provide results consistent with present laboratory and atmospheric constraints. This matter is pursued in § § 2 and 3. Section 2 discusses present understanding of relevant chemical reactions, attempting to identify reactions of potential importance in the atmosphere for which data are either lacking or ambiguous. Models are compared with atmospheric measurements in § 3 in an effort to develop a consistent chemical scheme. The rationale for selection of specific emission models is described in § 4, while atmospheric implications are investigated in § 5.

CHEMISTR y
The chemical processes of importance for the atmosphere have been reviewed extensively in the recent literature. An excellent summary of information available up to mid 197 5 is given by Nicolet (1975). We shall concentrate here on subsequent developments, focusing attention on more significant uncertainties. An overview of the important chemistry is presented schematically in figures 1-4.
A new development in stratospheric chemistry concerns the potential importance of ClN0 3 (Rowland, Spencer & Molina 1976a). Chlorine nitrate is formed by CIO + N0 2 + M-+ ClN0 3 + M, and removed both by photolysis and by reaction with OH. Products of these reactions are unknown. Smith, Chou & Rowland (1977) believe that the photolytic path may involve initial production of ClONO, by hv + CIN0 3 -> ClONO + 0, (6) which should be followed rapidly in the stratosphere  either by or by !tv+ClONO-rClO+NO. This scheme involves no net d1auge in the concentration of stratospheric odd oxygen. A similar conclusion holds for photolysis through !tv + ClN0 3 -+ ClO + N0 2 , the path adopted by the National Academy of Sciences (1976). Photolysis of ClNOa through hv+ClN0 3 -> Cl+NOa, (10) followed by ( 11) provides a possible catalytic mechanism for removal ofodd oxygen. The rate for ( 11) is approximately 30% of the net rate for removal of N0 3 , which proceeds mainly by ( 12) We shall adopt the reaction path favoured by the National Academy of Sciences (1976). The scheme favored by Smith el al. (1977) would give essentially identical results. We shall investigate the possible role of ( 10) and ( 11), which could have a significant impact at high concentrations of Clx (see figure 28). Reaction with OH represents a minor path for removal uf CIN0 3 • We assume that the reaction proceeds through OH+ C.:lNOa -+ HOCl + N0 3 , though an alternate path might involve OH+ ClN0 3 -> H0 2 + CIONO. (14) Reaction (la) could provide a detectable source of HOCl, which may be formed in addition by Reaction ( 15) represents a potentially important sink for 0 3 below about 25 km. Cross sections for photolysis of HOC! have been measured by DeMore (1977). The rate constant for (15) is unknown. A value in excess of 10-11 ems s-1 would imply a major role for HOCl in stratospheric chemistry. 0( 1 D) + N 2 -+ 0( 3 P) + N 2 2.0-11 exp ( + 107 /T) Streit et al. (1976) e 0( 1 D) +0 2 -+ 0( 3 P) +02 2.9-11 exp (+67/T) Streit et al. ( 1976) 3a 0(1D)+N 2 0 -+NO+NO 5.5-11 Davidson et al. (1977) 3b -+N2+02 5.5-11 Davidson et al. (1977 4a O(lD)+CH,-+ OH+CH 3 1.3- 10 Davidson et al. (1977) -+ H 2 +H 2 CO 1. 3-11 Hampson & Garvin (1975) 5 OH+CH.-+ CH 3 +HaO 2.36-12 exp (-1710/T) Davis et al. (1974a) 6 OH+CO -+C0 2 +H 2.10-13f(M) exp (-115/T) Greiner (1969) CFCl 3 + hv -+ CFCl 2 +Cl CC1 4 + hv -> CCl 3 +Cl CH 3 Cl + h1 1 -> CH 3 +Cl CH 3 CCl 3 +h1 1 -> CH 3 CCl 2 +CI references and notes see table lb Graham & Johnston (1978) Hall & Blacet (1952) Cox & Derwent (1976)  A> 1730A Selwyn et al. (1977) values at 243 K A < 1730 A Zelikoff et al. ( 1953), Romand & Mayence (1949) Graham & .Johnston (1978 Griggs (1968), Philen et al. (1977) A Ditchburn & Young (196:.) A > 1800A Thompson et al. (1963) A < 1800A  Sun & Weissler (1954) t Units: cm.3 s-1 for bimolecular reactions; cm 6 s-1 for tcrmolccular reactions. The notation .1.0-11 is intended to be read as 1.0 x 10-11 • ! 1A = 0.1 nm = 10-10 m. We adopted a somewhat smaller value, basing our choice on laboratory studies of the isoelectronic process Assessment of the possible roles of CIN0 3 and HOCl requires an accurate treatment for the diurnal cycle of solar radiation. The important species OH, H0 2 , 0, N0 2 , ClO and ClN0 3 exhibit significant and complex diurnal variations, as shown in figure 5. Models based on theassumption of a diurnal mean insolation lead to an underestimate of the net rate for the key re.action ClO+O-+ Cl+0 2 • (.I!)) The discrepancy arises in part due to inadequacies of the simple models, which tend to overestimate the daytime concentration of CIN0 3 . It may be attributed also to difficulties which arise due to covariance of the concentrations for CIO and 0. Our model allows for backscatter of solar radiation from the lower atmosphere, as described by Yung (1976). Scattered radiation is important at longer wavelengths where it leads to a significant increase in the rates for photolysis of ClN0 3 , HOC! and N0 2 and to an enhanced source for 0( 1 D). Another development of note concerns the rate for reaction of NO with H0 2 , Howard & Evenson (1977), using laser magnetic resonance to monitor H0 2 , find a rate for this reaction at room temperature of 8 x 10-12 cm 3 s-1, significantly larger than indirect measurements reported earlier by Davis, Payne & Stief (1973 b), Cox & Derwent (1975), Hack, Hoyerman & Wagner (1975 and . The reaction is of considerable   where xis the solar zenith angle, C is the 0 2 vertical column and Q is any additional opacity source. The photodissociation rate J 02 is given by importance for the stratosphere. It provides a source for OH and, coupled with photolysis of N0 2 , a source also for odd oxygen. The large discrepancy between new and old measurements for (20) raises questions with regard to the validity of rate expressions reported earlier for other reactions involving H0 2 , for example, and reaction (18). We shall return to this matter later, in § 5. Several investigators Niki, Maker, Savage & Breitenbach 1977;Howard 1977) have raised the possibility that reaction of N0 2 with H0 2 might lead to Vol. 290. A. . The diurnal variations of several species at 22, 30 and 40 km are shown for 30° N latitude and 0" solar declination. The notation (-2) indicates that the number density for that species has been multiplied by 10-2 • formation of pernitric acid, H0 2 N0 2 • The subsequent chemistry of this compound is unknown. It will be treated here according to the scheme followed by H0 2 N0 2 + ltv-? OH+ N0 3 , OH+ H0 2 N0 2 -+ H 2 0 + 0 2 + N02, by thermal decomposition (Graham, Winer & Pitts 1977), or by rain out. The rate for (22) has been measured by Howard (1977). Table 1 includes estimates of rates for (23) and (24). Pernitric acid could represent a sink for odd nitrogen in the lower stratosphere and troposphere. It could provide also a source of tropospheric odd oxygen, through (23) followed by (12), and it might be detectable in the atmosphere with current spectroscopic techniques. With our estimated reaction rates, H0 2 N0 2 plays only a minor role in atmospheric chemistry. Recent studies of the reaction (25) indicate that the rate for this process may be a function of ambient pressure (Cox, Derwent & Holt 1976;Sie, Simonaitis & Heicklen 1976;Chan, Uselman, Calvert & Shaw 1977;Overend & Paraskevopoulos 1977;. The effective rate constant at a pressure of one atmosphere appears to exceed the low pressure value by about a factor of 2, with consequent implications for the budgets of tropospheric CO and OH. The mechanism for reaction of CO with OH at high pressures is not well understood. The laboratory data indicate that the effective reaction rate is a sensitive function of the particular choice of background gas and data to define the behaviour in air are incomplete. Further work is clearly required in order to resolve this issue. The pressure dependent rate expression used here is given in table 1.
Mesospheric OH and 0 3 data give indirect information on the concentration of mesospheric H 2 0, together with some check on the validity of rate expressions adopted for reactions (1)-(4). The concentration of stratospheric OH reflects assumptions regarding the concentration of stratospheric H 2 0, and rates for the H0 2 reactions, (2), (3), (20) and (21) . Measurements of stratospheric Cl, ClO and HCl have implications with regard to the total concentration of stratospheric chlorine, and for the validity of the scheme used to treat the chlorine catalytic cycle. Column densities of N0 2 and HN0 3 provide global perspective on the distribution of nitrogen oxides.
Concentrations of the long lived gases N 2 0, NOx, CH 4 , H 2 , Clx and chlorocarbons were calculated using the concept of quasi-horizontal mixing surfaces ( Yung 1976) to define latitudinal distributions, in combination with a one-dimensional model to specify average vertical structure. Pref erred surfaces for horizontal mixing were identified by using observations of Sr 90 and C 14 (List & Telegadas 1969), supplemented in the equatorial region with recent data for N 2 0 (Schmeltekopf et al. 1977;. The mixing surfaces are shown in figure 6. The one-dimensional model is applied at 30° N latitude where data exist for several important trace species including CH 4 • Observations of CH 4 were used to select a set of effective vertical diffusion coefficients as described by Wofsy & McElroy (1973). These coefficients are included in figure 6.  Yung 1975b;Johnston 1976;Hunten 1975;McElroyetal. 1977) and agree with results for N 2 0 reported by Schmeltekopf et al. (1977). species 1.2-10 2.0-10 1.1-10 1.0-9 5.0-11 1.0-9 Upper boundary conditions: zero flux at 80 km t NO., = NO+ N0 2 + N0 3 + 2 x Nz0 6 + CIN0 3 + HN0 3 + HONO + H0 2 N0 2 • + Cl,,= HCl+Cl+CIO+CIN0 3 +HOCl+Cl 2 • The atmosphere contains gases with a wide range of chemical lifetimes. Our selection oflong lived gases includes only those species for which the chemical lifetime is very much longer than a day. The concentration, ni, of these species at 30° N latitude is assumed to satisfy a continuity relation of the form where <f>i denotes the vertical flux (molecules cm-2 s-1 ) of species i and the symbols ~ and I.,, indicate diurnally averaged values for the production and loss rates of i at altitude z. Equation (26) is solved subject to boundary conditions as summarized in table 2. The chemical terms ~ and Li are influenced in general by the complex diurnal variation of other species. The concentrations of short lived species, j, defined to include gases with chemical lifetimes shorter than relevant times for transport, are found by solving local continuity equations Short lived species include 0( 3 P), 0( 1 D), NO, N0 2 , N0 3 , N 2 0 5 , HN0 2 , HN0 3 , H0 2 N0 2 , ClN0 3 , HCl, Cl, Cl 2 , ClO, HOCl, H, OH, H0 2 and H 2 0 2 • Averaged production and loss rates for species i were obtained by numerical integration of the instantaneous rates Pi(t) and

Li(t).
The coupled set of equations defined by (27) was treated numerically with an implicit finite difference representation adapted from Richtmeyer (1957), This equation must be solved with a precision adequate to ensure conservation of mass to high accuracy over a diurnal cycle. The method adopted here conserves mass to better than 1 part in 10 6 • Periodic boundary conditions are imposed by a simple variational technique (see, for example, Wofsy 1978).
Height profiles at various latitudes as calculated for several of the longer lived species are shown in figures 7 (N:.P), 8 (CH 41 CO) and 9 (CFCl 3 , CF 2 Cl 2 , CCl 4 , CH;iCl). Figure 7 includes a. comparison of model results with vertical profiles for N 2 0 as measured at mid-latitudes by Heidt et al. (1976), Schmeltekopf et al. (1977), . A similar compar.ison with data for CH 4 (Ehhalt 1974;Ehha.lt et al. 1975b;Volz et al. i976) and CO (Volz et al. 1976) is given in figure 8. Model results are in satisfactory accord with observational data for all gases with the exception of CF 2 Cl 2 and CO. The discrepancy which appears for CF 2 Cl 2 at higher altitudes may be attributed to its short lifetime relative lo N 2 0 or CH 4 , and is unimportant for present purposes. The agreement lends confidence to the procedures adopted to simulate both vertical and lateral transport, at least for the lower stratosphere. Extension of the data base to higher altitudes .is obviously desirable. R esults in figures 5 and 7-H were obtained by using the standard chemical model as defined in table 1. The discrepancy between calculated and observed CO is more difficult to explain. The CO chemical lifetime is fairly short (2-8 x 10 6 s) above 20 km and its abundance in the model is controlled largely by chemical production from CH 4 , and loss by reaction (25) (see figure 4). Since (29) and (25) both involve reaction with OH, the CO abundance should be approximately given by This scheme yields CO abundances too small by at least a factor of 5, as shown in figure 8. An additional source of CO is evidently required. Photolysis of C0 2 would have to be 100 times faster than current estimates in order to be significant and we consider this to be unlikely. A possible source of CO might be reaction of 0(1D) with C0 2 , 0( 1 D) +C0 2 -+ 0 2 +CO.  (1975) . Altitude profiles at various latitudes for C~ and CO. The CH 4 cu1vcs were calculated a~ discussed in the text. The methane data were taken at 32° N, with the exception of one profile from 32° S ( e ). The points 'Y are from rocket flights. The different symbols indicate individual profiles, and are from Ehhalt (1974), Ehhall et al. (1975, and Volz et al. (1976). The CO profiles in the troposphere were taken from the data of Seiler & Schmidt (1974). In the stratosphere the solid curve result<> from the standard chemical model, while the dashed curve results from inclusion of the process 0( 1 D) + C0 2 (k = l x I 0-11 ems s-1 ) a'> a source of CO. The CO data were taken at 32° N, 0, • (Volz et al. 1976).
Figures 10 and 11 show profiles as calculated for the concentrations of mesospheric OH and 0 3 • The lifetime for odd oxygen in the mesosphere is relatively short and 0 3 must be treated as a short lived species in this region; sec equation (28) . The diurnal variations of several species ofimportance in mesospheric chemistry (H, OH, H0 2 , 0 3 ) are shown in figure 12. Concentrations of mesospheric OH and 0 3 depend in a direct manner on the a bundance of mesospheric H 2 0. Our model assumes H 2 0 abundances as shown in figure 13. The rise in the mixing ratio of H 2 0 above 20 km reflects conversion of CH 4 to H 2 0 and C0 2 • The results shown in figures 10 and 11 arc in excellent agreement with the available observational data set. Figure 11 includes a model in which the room temperature measurements for reactions (2)-( 4) were taken to apply throughout the atmosphere. Both models are in satisfactory accord with Anderson's (1975)  F1cuRE 9. Altitude profiles at various latitudes for (a) CFCl 3 , (b) CF 2 Cl 2 , (c) CCI, and CH 3 CI. The curves were calculated as discussed in the text for 0° solar declination. The experimental data for CFCl 3 and CF 2 Cl 2 are as follows : • , llll, IB (32° N), 0 (51° N), Heidt et al. ( , 1976 mesospheric 0 3 • Day to night differences are adequately reproduced. Watanabe & Tohmatsu (1976) suggest that 0 3 may exhibit significant seasonal variations, with winter concentrations at 65 km larger than summer values by a factor of 2.8. It is difficult to account for variations of this magnitude unless we invoke correspondingly large changes in the concentration of mesospheric H 2 0 . A model with seasonally invariant H 2 0 would predict a modest winter to summer variation in 0 3 , about 10%, in the opposite sense to that reported by Watanabe & Tohmatsu (1976). Water vapour is known to exhibit significant, factor of 3, variations in the lower stratosphere (Mastenbrook i971) . The measurements of OH (Anderson i976), Cl and ClO (Anderson et al. i977) between 25 and 40 km may reflect an upward extension of this variability. Model results are compared with stratospheric OH, 0, Cl, and ClO measurements in figures 14 and 15, illustrating the dependence of results on the choice of profile for H 2 0. It would appear that the high values for the concentrations of Cl and 010 cannot be attributed solely to variations in the water vapour mixing ratio, unless the mixing ratios for H 2 0 are much larger than 10--s. On the other hand the low values for Cl and ClO are consistent with H 2 0 mixing ratios in the range reported by Mastenbrook (1971 ) . Part of the ClO variability might be attributed to temporal changes in the concentration of total chlorine. We may note in this context that several precursors for stratospheric chlorine, in particular CFC1 3 , CF 2 01 2 , CC1 4 and CH 3 CC1 3 , are distributed quite inhomogeneously in the lower atmosphere with abundances in the northern hemisphere typically twice those for the southern hemisphere (  Measurements of HCl provide important information with regard to the role of ClN0 3 in stratospheric chemistry and may be used to place constraints on the total abundance of stratospheric chlorine. Figure 16 shows a comparison of computed concentrations for acidic chlorine, taken as HCl + ClN0 3 + HOCl, with measurements by Lazrus et al. (1976).  (Rowland et al. 1976a). Major uncertainties remain however and clearly point to the need for a careful observational search for CIN0 3 • Results presented here are in accord with a recent observational upper limit to the concentration of CIN0 3 set by Murcray et al. (1977). Figures 19 and 20 show a comparison as a function of latitude of predicted and observed column densities for stratospheric N0 2 and HN0 3 • The height integrated concentrations of these compounds are relatively insensitive to the more uncertain aspects of the model. T he good agreement of theory with observation exhibited by the figures indicates that the total abundance of odd nitrogen is well predicted by the model, and that the model satisfactorily accounts for the apportionment of NOx among its more abundant constituents. It lends confidence to the validity of procedures adopted to treat lateral transport (see also Wofsy 1978) . number density/cm-a FrGuRE 15. Altitude profiles are shown for Cl and CIO at 30° Nin summer and winter for wet and dry conditions (solid lines) at noon. The broken lines were calculated by using the standard water profile but employed high (-·-· )and low (---) values for methane cons. istent with the range of the data shown in figure 8. The data points are from Anderson et al. (1977) and were taken at 32° Nin December (e, x = 50°), July(®,$, X = 16°), and October (6, .A, X = 35°). Noxon et al. (1977) report very low column abundances for stratospheric N0 2 poleward of 50° N, during winter. The calculations exhibit a significantly smaller decrease of N0 2 in this region. This discrepancy could be resolved in large part by a slight temperature dependence in the photolysis rate of HN0 3 , which is the dominant NOx species at these latitudes. Figure 21 shows a comparison of observed and calculated profiles for 0 8 below 50 km. The gas is treated in this height range as a long lived chemical species by using the one-dimensional diffusion model, equation (26). We did not use mixing surfaces to simulate horizontal transport since chemical lifetimes are too short to permit use of this model above 22 km.
Anthropogenic CO leads therefore to some net increase in the abundance of tropospheric 0 3 • Ozone may be formed also as a by-product ofCH 4 oxidation (Levy 1971;Chameides & Walker 1973), by followed by (33) and (34). The rate for (36) is unknown, however, and we are unable therefore to offer any quantitative assessment of its importance. The recent measurement of a fast rate for (20) establishes a central role for nitrogen oxides in the chemistry of tropospheric 0 3 . Indeed the ozone source strength for the lower troposphere is large enough to cause difficulties in our attempt to define a balanced chemical system. The time constant for 0 3 below 5 km is about 10 days. The production due to (35) and (36) could be balanced by heterogeneous processes either in the atmosphere or at the surface. A quantitative discussion of this possibility is difficult however. Photolysis, followed by lzv + 0 3 ~ 0( 1 D) +0 2 0( 1 D) +H 2 0 ~ OH+OH, and reaction with H0 2 , equation (21), represent the most important gas phase sinks for 0 3 in the troposphere. The model is characterized by a peak broader in altitude than that exhibited by the data. The vertical column density of 0 3 in the model has a value 10.2 x 10 18 cm-2 , which may be compared with the observed value of 7.5 x 10 18 cm-2 • The discrepancy might reflect inadequacies of the chemical model. On the other hand, it should be noted that the difficulty arises in a region of the atmosphere where transport should play an important role for 0 3 • As noted earlier, use of one-dimensional vertical diffusion models is somewhat suspect for 0 3 • Most of the global abundance of atmospheric 0 3 is supplied by transport from a source region confined to a relatively narrow altitude region at low latitudes. A major fraction of the vertical column of 0 3 at higher latitudes must lie in the dynamical regime, as may be inferred In light of this discussion, it is perhaps not surprising that one-dimensional models should fail to reproduce the global mean abundance of 0 3 . The disagreement however emphasizes the difficulty associated with use of one-dimensional models for the assessment of human influence on 0 3 . One might hope that one-dimensional models should give a reasonable estimate for the magnitude of any particular perturbation to 0 8 , even though the model might fail to reproduce details of the vertical distribution. As an alternate approach one might follow the procedures described by McElroy et al. (1974) and seek to define a suitable scaling factor with which to model the influence of perturbations in the dynamical zone. Neither procedure is totally satisfactory. A definitive assessment requires a fully integrated chemical-dynamical model for the lower atmosphere. These perturbations fall into two general classes: those which are considered essential to modern society, and those which may be regarded as largely discretionary. Among the essential perturbations, we shall focus mainly on combustion and agriculture. More discretionary human influences include the use of certain halocarbons as working fluids in refrigeration systems, as propellants in aerosol cans, and as solvents and cleaning fluids.
Combustion of fossil fuels provides important direct sources of atmospheric C0 2 , CO, H 2 , NOx and N 2 0. Biological activity plays a major role in regulating the concentrations of these gases in the natural atmospheric environment.
Respiration and decay release C0 2 at rates which can replace the present concentrations of the gas in the atmosphere on a time scale ofless than 20 years (see, for example, McElroy 1976; Bolin 1977). Combustion provides a source strength for C0 2 equal to approximately 10% of that from respiration and decay (Keeling 1973;Delwiche & Likens i977).
Methane, formed as a by-product of microbial fermentation, is the primary natural precursor for atmospheric CO and H 2 • Carbon monoxide is produced (see figure 4) R eaction (39) is the primary natural source for H 2 , giving a global source strength of magnitude 3.5 x 10 7 t per year. The source strength for CO has magnitude 1. 7 x 10 11 molecules cm-2 s-1 , 5.5 x 10 8 t C per year. These inputs may be compared to current estimates for combustion related sources of H 2 and CO of magnitude 2 x 10 7 t per year and 3 x 10 8 t C per year respectively (Jaffe i973 ;Schmidt 1974;Seiler 1974;Seiler & Schmidt 1974;Penner et al. 1977). The methane-related source strengths for CO and H 2 were estimated by using the pressure dependent rate expression for reaction (25). They are lower therefore than earlier estimates based on a pressure independent value for this coefficient. The revised estimate for the source of CO from CH 4 oxidation, and associated implications for tropospheric OH, introduce some difficulty in attempts to balance the budget for northern hemisphere CO. Either the rate constant for (25) must be less than values assumed here, or there must be sources for CO in the northern hemisphere larger than values usually associated with combustion, by about a factor of 3. Oxidation of terpenes or other hydrocarbon emissions from the terrestrial biosphere could provide an important natural source of CO (cf. Wofsy, McConnell & McElroy 1972) . For example, the hemiterpene, isoprene, is a major component of volatile emissions from many species ofland plants (Sanadze 1963;Rasmussen 1970). Isoprene is likely to react rapidly with OH radicals (cf. , and the ultimate yield of CO could be between 40% and 60%, based on olefin oxidation pathways which we might expect to apply in the atmosphere (Demerjian, Kerr & Calvert 1974;. If global emission rates for isoprene were comparable to those found by Sanadze (1963) and Rasmussen & Jones (1973) under laboratory conditions, photooxidation of isoprene could provide a source for CO comparable to or larger than anthropogenic emissions.
Combustion provides a direct source of fixed nitrogen, NOx, of magnitude 2.3 x 10 7 t N per year at the present time. Fixation of nitrogen for fertilizer contributes an additional source of magnitude 4.2 x 10 7 t N per year. These anthropogenic inputs may be compared to estimates for biological fixation which range from about 1-2 x 10 8 t N per year (Burns & Hardy 1975;Soderlund & Svensson 1976;McElroy 1976;Sweeney, Lui & Kaplan 1977;Delwiche & Likens 1977). Both the combustion and fertilizer sources of fixed nitrogen are growing at a steady rate, as illustrated in figure 23, and it is difficult to avoid the conclusion that anthropogenic effects must play a dominant role in global nitrogen fixation in the near future.
Our estimates for the combustion source of fixed nitrogen before 1970 use data for the United States published by the U .S. Environmental Protection Agency (Cavender, Kircher & Hoffman 1973). These data were scaled by a factor of approximately 2. 7 in order to obtain an estimate of global emissions. The scale factor reflects consumption patterns for various fossil fuels as given by Hamilton (1977). We assumed a growth rate for combustion of fossil fuels equal to 3.6% per year between 1970 and 2000, consistent with projections by the Organization for Economic Co-operation and Development (1976) (see also The National Energy Plan (1977)). The growth rate was taken equal to 2.4% per year between 2000 and 2050, and was lowered to 1 % per year for the period 2050 to 2100 (see model C in figure 23). A second model for combustion nitrogen, model D in figure 23, anticipates stringent regulation of NOx and CO em1ss10ns.
We shall explore two models for future use of nitrogenous fertilizers. The models are similar to those used by McElroy et al. (1977). Both models assume a relatively stable world population after the turn of the century. Model A allows for grain production equivalent to 4 x 10 2 kg per person per year for a population assumed to grow to 6.5 x 10 9 by the year 2000. Model B assumes grain production per person close to current values, 2.5 x 10 2 kg per person per year, with a similar demographic projection.
Combustion is known to provide a direct source for N 2 0 as well as NOx .  report measurements of N 2 0 and C0 2 emanating from the power plants in California. They find ratios N 2 0 to C0 2 of2.05 x 10-4 16-2 for coal and fuel oil plants respectively. Their data, combined with Keeling's (1973) discussion of C0 2 emissions, were used to construct the direct release model for N 2 0 given in figure 23 (model E). Figure 24 shows the r elease patterns for CO, C0 2 and H 2 • We used Seiler's (1974) estimate for CO release and CO to H 2 emission ratios given by Schmidt (1974).  Catalytic converters on automobiles Weiss & Craig r976) are known to provide additional direct sources of N 2 0. The automotive source is currently small, much less than 1 x 10 6 t N per year. We chose to omit this source in light of uncertainties in future development and use of catalytic converters.
The anthropogenic sources of fixed nitrogen shown in figure 23 undergo a complex series of transformations resulting ultimately in recombination of fixed N, with release of both N 2 0 and N 2 • On sufficiently long time scales release of nitrogen as N 2 0 and N 2 should balance the net global source of fixed N.
An estimate for the anthropogenically related source of N 2 0 in any given year requires a relatively complete understanding of the global nitrogen cycle. Discussions of this cycle emphasizing Man's impact are given by McElroy (r976) and McElroy et al. (1977). The N 2 0 source strength may be calculated by using the concept of a well defined delay pattern for recombination of fixed N, along with an estimate for the fraction of recombination events resulting in emission of N 2 0 rather than N 2 • The release patterns adopted here are illustrated in figure 25. The fraction ofrecombination events leading to release of N 2 0 in the present system may be estimated if we assume a balance between natural fixation and recombination, and use estimates for the lifetime of atmospheric N 2 0 to evaluate the current source strength of the gas. We shall adopt for this purpose a natural fixation rate of 1.6 x 10 8 t N per year, consistent with the range of values quoted above, and we shall assume the same yield of N 2 0 from recombination of anthropogenic and natural nitrogen. T he uncertainty in the lifetime of atmospheric N 2 0 is more serious than the ambiguity associated with the fixation rate, with values in the literature ranging from 1.8 a (Hahn & J unge 1977) to an upper limit of about 140 a imposed by stratospheric photolysis. Lifetimes less than about 10 a would appear to require sources for N 2 0 in addition to denitrification. A source due to nitrification would not obviate the difficulty since the sum of the sources from nitrification and denitrification must be balanced by fixation. Lightening discharges offer a further possibility (Griffing 1977;Zipf & Dubin 1976), though quantitative estimates (Griffing 1977) for the associated source strength suggest that it should be small, less than 2 x 10 6 t N per year. We shall investigate models with two values for the N 2 0 lifetime, a low value of 20 a, only slightly larger than the range 4-12 a recommended by Hahn & Junge (1977), and a high value of 140 a, which would apply if stratospheric photolysis were the major sink. Corresponding release rates for N 2 0 are given in figure 26.
With our choice of fixation rate, the yield of N 2 0 relative to N 2 in recombination of fixed N is either 6% or 40% for the long and short lifetimes respectively. A yield of 6% agrees with soil data quoted by the Council for Agricultural Science and Technology (1976). Moreover, the long lifetime and low yield would be consistent with the lack of variability for atmospheric N 2 0 as reported by Weiss, Dowd & Craig (1976) from measurements over the Pacific Ocean, and Goldan et al. (1978) from continental measurements. It disagrees however with relatively large variations of the total N 2 0 column density as measured spectroscopically by Goody (1969).
It should be noted that other investigators (Hahn & J unge 1977; Singh, Salas, Shiegeshi & Crawford 1977 b) using chromatographic techniques find greater variability than that seen by Weiss et al. (1976) and Goldan et al. (1978) . The apparent discrepancies might be attributed to a predominant role for continental influences on the budget of atmospheric N 2 0. Data from this laboratory (Kaplan et al. 1978) indicate that fresh water systems can provide both sources and sinks for N 2 0 . It would seem possible that aqueous systems could play an important role and that biological uptake might represent a major sink (see, for example, Brice, Eggleton & Penkett 1977), though heterogeneous chemistry could also contribute (Ausloos, Rebbert & Glasgow 1977). The issue is clearly unresolved. The weight of the evidence, in our view, appears to point towards a relatively long lifetime for N 2 0, in that it is difficult to account for source strengths or sinks much larger than about 5 x 10 7 t N per year ). FmuRE 26. Projections for release rates of N 2 0, assuming the soui:ces of fixed nitrogen given in figure 23, the delay times for denitrification given in figure 25, a natw·al fixation rate 01 1.6 x 10 8 t N/a and N 2 0 lifetimes as shown. The labels are described in figure 23, and the arrows indicate the source of NzO from natural processes. Figure 27 shows historical data for industrial sources of several halocarbons which are ultimately released to the atmosphere. We focus here on chlorinated species. The possible impact of brominated compounds is discussed elsewhere (Wofsy et al. 1975 b). The compounds CFC1 3 and CF 2 Cl 2 are released to the atmosphere on an average of less than a year following manufacture (McCarthy, Bower &Jesson 1977). The data in figure 27 are from McCarthy et al. (1977) for CFC1 3 and CF 2 C1 2 , from Galbali (1976) for CC1 4 , and from Neely & Plonka (1978) for CH 3 CC1 3 • The figure also summarizes models devised to explore possible future release patterns for CFC1 3 , CF 2 Cl 2 , CC1 4 and CH 3 CC1 3 • We consider two models for CFC1 3 , CF 2 Cl 2 and CH 3 CC1 3 • In model A, release rates are assumed to remain constant at values applicable in 1974. Model B envisages regulation of these compounds and the source is abbreviated accordingly in the period subsequent to 1980 (see figure 27). Trichloroethane is used as an intermediate in the manufacture ofvinylidine chloride and as a cleaning solvent, with the latter accounting for most of the release to the atmosphere. Release rates have grown rapidly in recent years as this compound is used to replace the chloroethylenes C 2 Cl 4 and C 2 HC1 3 • The models in figure 27 envisage continued growth in the release of CH 3 CC1 3 until ~980, followed by slow growth thereafter. McCarthy et al. (1977). Model A assumes continued release at 1974 rates, while model B assumes that the use of these compounds as aerosol propellants is discontinued in 1980. ( b) Historical data and projections for release of CH 3 CCI 3 ( x) and CCI, ( O). The data are from Neely & Plonka (1978) and Galbali (1975) respectively. For CH 3 CC1 3 , the growth patterns given for fossil fuel emissions in figure 24 are adopted; a constant release rate after 1974 is assumed for CCl 4 •

RESULTS
We shall describe results from a number of models exploring consequences of combustion, anthropogenic changes to the nitrogen cycle and the chlorocarbon release patterns described above. The anthropogenic perturbations will be treated by using the standard chemical model summarized in table 1, with diurnal variations incorporated as described in § 3. We begin with a brief discussion of results obtained for steady state conditions, assuming a range of values for tropospheric N 2 0 and stratospheric Clx. Figure 28 presents global average column densities of 0 3 as functions of the N 2 0 mixing ratio, for several concentrations of stratospheric Clx, calculated by using the one-dimensional diffusion model. It shows the dependence of results on procedures used to treat insolation. The mixing ratios for N 2 0 and Clx in the present atmosphere are taken as 3 x 10-7 and 2.3 x 10-9 , respectively. The value for N 2 0 reflects the mean surface concentration of the gas. The value for Clx reflects conditions at an altitude of 40 km.
The model. with diurnally averaged insolation (figure 28b) has somewhat less 0 3 than the more complete model (figure 28a). Differences arise largely at altitudes below 30 km. They may be attributed in part to the larger source of odd oxygen in the standard model due to (20), in part to a reduction in the removal rates for odd oxygen due to (42) FIGURE 28. Global average column densities of ozone as a function of the tropospheric N 2 0 mixing ratio. The labels on the curves indicate the CI., mixing ratios (unit: parts/10 9 by vol.) at 40 km. The steady state calculations with a Cl., mixing ratio of ca. 8.6/10 9 at 40 km result from CFCI 3 and CF 2 Cl 2 fluxes of 341 OOOt/a and 410000 t/a respectively, with the boundary conditions for other constituents taken from table 2. A Cl., mixing ratio of ca. 2.3/10 9 corresponds to CFCI 3 and CF 2 Cl 2 fluxes which are smaller by a factor of 10, and approximate present-day conditions. The results shown in (a) are obtained with the full diurnal model described in the text, which includes backscatter of solar radiation from the lower atmosphere. The results in (b), (c) and ( Concentrations of OH, H0 2 , 0 and NO all tend to peak near noon, whereas the concentration of N0 2 attains its highest value soon after sunset. The diurnally averaged model underestimates therefore the sink for odd hydrogen, reaction ( Figure 29 shows calculated values for the integrated column density of 0 3 , with the standard chemical model, as a function of chlorine concentration at 40 km, for several values of the N 2 0 mixing ratio in the troposphere. As before, these results were obtained by using the onedimensional diffusion model. The average insolation model tends to underestimate 0 3 depletion for any given perturbation. The incremental effect of chlorine is small in both models for high concentrations of N 2 0. This behaviour reflects the importance of CIN0 3 and the role of the reaction ClO+NO-+ Cl+N0 2 , both of which serve to suppress the concentration of CIO. Differences between the average and diurnal models may be attributed largely to an overestimate of ClN0 3 by the average model, as discussed above. Figures 30-32 show the changes in the concentration of 0 3 as a function of altitude which result from perturbations of Clx and N 2 0 . Addition of NOx leads to a larger decrease in 0 3 below 24 km in the standard model as compared with the average insolation model (see figure 31). This difference may be attributed to overestimation of ClN0 3 by the averageinsolation model. Figure 32 shows the influence of enhanced concentrations of tropospheric CO and CH 4 • Tropospheric ozone increases significantly in response to large enhancements of CO, as a Vol. 290. A. consequence of process (35). Increases in ozone also occur in the stratosphere, owing to more efficient termination of the Cl catalytic cycle by the reaction This effect is more marked for large chlorine abundances. We should caution that the computed magnitude of the perturbation to 0 3 may be quite sensitive to uncertainties which currently exist in the chemical model. The point is illustrated in figure 33, which may be compared with figures 28b and 29b. The models are identical in all respects except that: (i) The rate constant for (3) was set equal to 2 x 10-11 cm3 s-1 ; (ii) The rate constant for (21) was increased by a factor of 3; and (iii) The rate constant for reaction ( 41) was increased slightly, with the new rate expression taken as 7 .5 x 10-11 exp ( -737 /T) .
Both models provide satisfactory descriptions of the natural atmosphere. The rate adjustments are consistent with experimental uncertainty, yet the character of the response to the perturbations is quite different. We may note for example that results in figure 33 would suggest that the column density of 0 3 might increase with increasing N 2 0, at current levels of stratospheric Clx. I nherent limitations of the one-dimensional diffusion model were discussed earlier. T he model should provide a satisfactory description of 0 3 above about 28 km, but cannot account for the distribution of 0 3 at lower levels, where dynamical effects are predominant and inadequately described by one dimensional diffusion. One-dimensional diffusion models have  been applied to the assessment of 0 3 perturbations, despite these limitations, with the hope that the models should satisfactorily simulate fractional changes to total 0 3 • As an alternate approach one might assume that the transition from chemical to dynamical control of 0 3 should occur over a narrow altitude range. Models might be used therefore to estimate the fractional change in the concentration of 0 3 at the transition boundary and observed concentrations of 0 3 at lower levels might be scaled appropriately to reflect the influence of perturbations. This procedure is useful to the extent that 0 3 may be regarded as a passive tracer in the dynamical region, and to the extent that we can identify a well defined boundary separating zones of chemistry and dynamics (see McElroy et al. 1974). The relative influences of chemistry and dynamics may be evaluated by comparison of relevant time constants. The chemical time constant may be defined as the time required for photolytic production of a given concentration of 0 3 ,

{45)
where J 0 a is the rate (unit: s-1 ) for photodissociation of 0 2 • Chemical time constants for the unperturbed system are shown for several latitudes in figure 35. Consideration of the observational data for 0 3 indicates that dynamical time constants cannot exceed about 3 months, and are most probably smaller than this by at least a factor of 3 (Oort & Rasmu~en 1971). Implications for 0 3 perturbations are summarized in figure 34.
A comparison of figure 34 with figures 28 and 29 suggests that differences between models are most pronounced in their response to low concentrations of NO:z:. The variance between results obtained with the diffusion and scaling approaches provide some subjective estimate of the inherent uncertainty. The diffusion model allows for enhancement of the source for perturbed 0 3 at low altitudes, a contribution omitted in the scaling model. On the other hand, the scaling model should account to some extent for the complex influence of global dynamics, and has the additional benefit that results should be reliable in the limit of small perturbations.
We turn our attention now to specific human influences on the atmosphere, emphasizing possible effects on 0 3 • We shall make use of the one-dimensional diffusion model, while presenting sufficient information to allow the reader to gauge modifications which would arise with a scaling approach. We consider a variety of models as summarized in table 3. Models describing perturbations associated with combustion are labelled A. Models incorporating effects of chlorocarbons and N 11 0 from recombination of fixed N are designated by symbols B and C, respectively. Numerals are used to distinguish specific models employed to describe possible human influences in each of these broad categories.  ( 45)) is shown as a function of altitude, for 20° solar declina· tion, and the latitudes marked on the curves. The arrow indicates a lifetime of one month.  Possible effects of combustion are summarized in figure 36. Models A1 and A2 indicate the influence of direct releases of CO and H 2 • Models A3 and A4 show incremental effects associated with combustion-related releases of N 2 0. The increase of total ozone ari!es principally from enhanced concentrations of the gas in the troposphere, as shown in figures 32 and 37. In the stratosphere, the mixing ratio of Clx increases in response to reduced concentrations of tropospheric OH. This process compensates for the reduced efficiency of the Clx catalytic cycle, due to enhanced levels of CH 4 as discussed above. Anticipated changes in atmospheric N 11 0, CH 4 , CO, CH 3 Cl and stratospheric Clx are given ju figure 36b. Figure 38 summarizes possible effects of chlorocarbons, for both high (upper panel) and low (lower panel) emission rates. Release of CFC1 3 and CF 2 Cl 2 at constant rates of 341000 t a-1 and 410000 t a-1 might be expected to lead to reductions in column 0 3 by as much as 5% by the year 2000, growing to 15% by 2100. The results shown here were obtained with a model which assumed that stratospheric photolysis was the only removal process for CFCl 3 and CF 2 Cl 2 • The perturbation to ozone would be smaller if there were additional important sinks (cf. Ausloos et al. 1977). Inclusion of CO and H 2 emissions diminishes the perturbations to total ozone, as discussed above. The validity of these models is particularly suspect for large perturbations, in that we do not allow for feedback on atmospheric structure and dynamics due to large changes in 0 3 • Mixing ratios for CH 4 , CFCl 3 , CF 2 Cl 2 , CH 3 CC1 3 and Clx are given in figure 39 for models B2 and B4. The possible impact of agricultural and combustion related sources of N 2 0 is summarized in figure 40. We may note that results for model C1 are similar to those for C2. The 0 3 perturbation is relatively insensitive to the choice of lifetime for atmospheric N 2 0, owing to time delays builtintoourmodelofthenitrogencycle. We have assumed that recombination of fixed N is the major natUl'al source of atmospheric N 2 0 and that the yield for production of anthropogenic N 2 0 is similar to that for the natural system. On the other hand, waters polluted by human nitrogen are often biologically active and have been observed to provide significant prompt sources of atmospheric N 2 0 (Kaplan et al. 1978). Assessment of human influence on N20 is difficult given the gaps in our understanding of processes which affect the cycle of this gas. The impact on 0 3 could be larger, or smaller, than suggested in figure 40. An upper limit to the effect might be estimated if we assume a small natural source for N 2 0, with immediate conversion of anthropogenic nitrogen to this gas. A limit based on this consideration is included in figure 40. This model would imply an increase of N 2 0 from 160 parts in 109 in 1940 to 300/109 today. F1oun.E 38. Cl1anges in the global average ozone column associated with cWorocarbon use (Bl-B4) and combustion (Ai ). The upper panel employs the high emission rates for cWorocarbons given in figure 27 while the lower panel assumes some regulation of cWorocarbon emissions. The labels on the curves are described in table 3. The behaviour of 0 3 in response to multiple coupled perturbations is illustrated in figure   41. We adopt, as a baseline, an atmosphere subject to what may be regarded as the least discretionary of human influences, combustion and agriculture (curve A1 + C1). The figure summarizes then the additional effects which might arise due to various chlorocarbon release patterns. It is clear that the influence of a combination of human practices is intrinsically nonlinear as regards the effect on 0 3 • As seen earlier, a constant release of CFC1 3 and CF 2 Cl 2 at rates prevalent in 1974 might be expected to cause a reduction of column 0 3 by about 15% in the year 2100. Indeed the net perturbation in the fully coupled system is less than that which we derive for the chlorine perturbation alone, 8% as compared to 15%. The 'baseline' calculation shows the ozone column abundance increasing with time. The additional ozone is present in the troposphere, rather than the stratosphere (see figures 32 and 37). The impact of such a perturbation to tropospheric ozone clearly requires careful study. FrcuRE 41. Changes in the global average ozone column associated with multiple perturbations (due to combustion (A), chlorocarbon use (B) and fertili.ze1· use (C)) . The labels are d escribed in table 3.

CONCLUDING REMARKS
Human activities associated with combustion, food production, and chlorocarbon industries may be expected to significantly perturb the chemistry of the atmosphere on a global scale. Increases in the mean concentrations of CO, C0 2 , CFC1 3 , CF 2 Cl 2 and CH 3 CC1 3 are already apparent, and detectable changes should be evident for N 2 0, CH 4 , CH 3 Cl and 0 3 within the next fifty years. The average column abundance of 0 3 could decrease by more than 10% by the year 2100 if present trends should continue, and tropospheric ozone may be significantly enhanced.
Assessment of Man's impact on the atmosphere requires a comprehensive model which must simulate, not only essential aspects of atmospheric chemistry and dynamics, but also the manner in which the atmosphere interacts with land, hydrosphere and biosphere. Our understanding of these systems is fragmentary. H uman influences are strongly interactive, with feedbacks which are both positive and negative. Chlorocarbon emissions will enhance the supply of chlorine to the stratosphere. T he associated reduction in Os may be offset however to some extent by an increase in the stratospheric burdens of NOa: and CH 4 induced by other anthropogenic activity.
Our treatment of human effects on the atmosphere is necessarily incomplete. We ignore for example direct and indirect consequences associated with human exploitation of land, fresh waters, estuaries and oceans and it is clear that these activities could have a predominant influence on the coupled biosphere-atmosphere system.
Our discussion here has sought to describe the atmosphere as an isolated compartment, investigating the consequence for 0 3 of various gas exchanges across the lower boundary. Even this limited problem poses major difficulties.
Our model does not account for the present global distribution of 0 3 • The deficiencies may be attributed in part to uncertainties in our understanding of atmospheric chemistry, in part to the inadequacy of our model for atmospheric dynamics. Lifetimes for N 2 0, Os and halocarbons may be affected by heterogeneous processes in the troposphere. The abundance and distribution of tropospheric OH is poorly determined and of central importance, in that reactions with OH regulate the stratospheric concentrations of CH 4 , CH 3 Cl and CH 8 CC1 3 • Factors which influence the concentration of stratospheric H 2 0 are not defined. The observed variability of stratospheric ClO is puzzling, and raises the possibility of a major gap in our description of the chlorine cycle. Our models do not allow for coupling of dynamics and chemistry, which might influence the residence time of gases in the perturbed stratosphere.
Finally we should emphasize that our understanding of N 2 0 and the global nitrogen cycle is seriously incomplete. This problem merits special attention in that the concentration of N 2 0 may be affected by non-discretionary human activity and in that N 2 0 plays a key role in many important areas of atmospheric chemistry.
T his work was supported by the National Science Foundation and the National Aeronautic and Space Administration under contracts NASA-NSG-2031 and NSF-ATM-22723 respectively. O ne of us (J.A. L.) acknowledges support from the Rockefeller Foundation's Fellowship Program in Environmental Affairs. We are indebted to the Ames Research Center of the National Aeronautic and Space Administration for computational facilities. Computer time was also provided by the National Center for Atmospheric Research, which is supported by the National Science Foundation.