Trace gas trends and their potential role in climate change

,

5. Clarify the sources for differences and discrepancies between published estimates for the trace gas radiative effects.
First, some comments are in order on the scientific necessity of this study because several papers have been published on the trace gas effects. None of the previous studies have included all of the trace gases included in this study. The present study uses new laboratory spectroscopic data made available subsequent to the earlier studies; the present study relies heavily on observed concentrations and trends for the trace gases as opposed to the hypothetical concentrations and increases assumed in most of the earlier studies. The Lacis et al. [1981] study did use observed trends but was restricted to only the effects of chlorofluorocarbons (CFC's), CH,•, and N20. Chamberlain et al. [1982] included the direct effects of most of the trace gases considered in this study, but inferred the temperature changes from the radiative flux change at the surface instead of the changes to the surface-troposphere system. It is the radiative flux change to the surface-troposphere system that governs the surface temperature change [Manabe and Wetheraid, 1967;Ramanathan, 1981Ramanathan, , 1982Hansen et al., 1982]. Furthermore, all of the above studies ignore stratospheric ozone changes due to CFC's.
For estimating their present-day radiative effects, we adopt the observed concentrations as of 1980. With respect to projected increases, we extrapolate the present-day trends to 50 years into the future. This procedure enables us to determine the relative importance of the various trace gases. Furthermore, we examine quantitatively the validity of the optically thin approximation, an important issue because this approximation is employed by all studies to treat the radiative effects of trace gases other than CO2, 03, N20, and CH,•.
For the purposes of this study, the word "climate" refers to surface/troposphere/stratospheric temperatures. The temperature changes are computed from a one-dimensional radiative-convective model described by Ramanathan [1981]. The radiative-convective model provides a convenient framework for examining the other trace gas effects, even though it ignores several important feedback processes arising from atmospheric circulation, oceans, and the cryosphere (e.g., icealbedo feedback). Numerous one-, two-, and threedimensional climate model calculations have estimated the surface warming due to doubled CO: (see summaries by the National Research Council [1982,1983], among several others). Furthermore, the radiative effects of the other trace gases (with the exception of stratospheric ozone changes) are very similar to that of CO:. Hence one-dimensional model estimates for the surface warming effects of the other trace gases and that of CO: (provided both are performed with the same model) can be used to scale the effects for the more realistic general circulation models (GCM), since numerous GCM estimates for the CO2 effects are currently available.

OF TRACE GASES
The earth's atmosphere currently contains trace gases with atmospheric lifetimes that vary from much less than an hour to several hundred years. The abundance of trace gases is therefore dictated as much by their removal rates as by the  In what follows, we emphasize radiatively important gases. Table l a briefly describes the lifetimes, dominant sources, and sinks of trace chemicals that have been identified in the global atmosphere. Many of these properties were listed earlier in an extensive review of scientific literature [World Meteorological Organization, 1982b]. For a large number of species where reaction with hydroxyl radical (OH) is the principal removal mechanism, lifetimes are estimated using an average OH concentration of 7 x 105 molecules/cm 3 and an average global atmospheric temperature of 265 K. While uncertain, this OH average is consistent with the budgets of CH3CC13, CO, and •'•CO [Volz et al., 1981]. For species with lifetimes greater than 20 years in Table la, removal is largely due to photolytic decomposition in the stratosphere. For all of the chlorofluorocarbons in Table la, lifetimes are determined based only on stratospheric photolysis from the computations of Wuebbles [1981] and those summarized by the World Meteorological Organization [1982b]. Only oxygenated species (03, aldehydes) and CH3I absorb UV light in the troposphere (2 > 290 nm). The fully fluorinated species are not decomposed by UV light even in the stratosphere. Their destruction by and large would occur in the mesosphere and ionosphere from absorption of Lyman alpha and Lyman beta radiation at altitudes above 70 km [Cicerone, 1979]. The lifetime of fully fluorinated organics (Table l a) can be in the 500-to 1000-year range. Hydrogen is the only species in Table la where microbial action at soil surfaces provides the major removal process. Ozone destruction also occurs on all surfaces (e.g., soil, water, and snow), but the mechanism of this destruction process is not known.
In addition to providing source, sink, and lifetime information, Table la also presents global average concentrations of species for the atmosphere of year 1980. These data are based on actual measurements that have already been summarized in some detail [World Meteorological Organization, 1982b].
It must be remembered that for species with lifetimes of less than 10 years, significant horizontal (latitudinal) gradients can exist. As an example, the ratio of northern to southern hemis-pheric average concentration is about 1.4 for CH3CC13 ('c i = 8 years) and >2 for C2C1,• (z• =0.5 year). For short-lived species (zi < 0.3 year), vertical gradients can also be expected. These gradients have been taken into consideration, based on available information, in developing the global averages shown in Table la.
The year 2030 concentrations are projected to develop a standard set of conditions, but a probable range is also included. The year 2030 "best estimates" in Table la thus constitute the "standard set." The range of likely variability associated with this standard set is also presented in Table l a.
In the following sections we discuss the information that was utilized in developing the year 2030 projections. The year 1980 characterization is based exclusively on atmospheric measurements.

Carbon Dioxide
The CO2 concentrations in the atmosphere have been measured to be increasing at a rate of approximately 1.5 ppm/yr (see Figure 1.2 of the National Research Council [1983]). Fossil fuel combustion is believed to be a major contributor to this increase. Over the past decades, CO2 release rates due to combustion have increased at a rate of about 4.3%/yr. A recent analysis [Rotty and Marland, 1980] supports a 2.4%/yr increase in CO2 emissions over the next 50 years. Only a fraction (• 50%) of this input is expected to remain airborne. Based on this analysis, Wuebbles [1981] described the CO2 concentration in ppm empirically.

Chlorofiuorocarbons (CFC's)
These chemicals came into major use in the 1960's and initially exhibited a rapid growth (10-15%/yr). The most important CFC's to date have been CF2C12 (F12 or CFC12) and CFC13 (Fll or CFCll). The global emissions of the major CFC's (F12 and Fll) actually declined somewhat from the mid-1970's through 1982 [Chemical Manufacturer's Association, 1983]. Part of this decline may be attributed to a ban on some nonessential usages (e.g., spray cans) of CFC's and due to adverse economic conditions that have prevailed in several industrial nations during this time. CFC emissions increased sharply in 1983. Data for CFC13 and CF2C12 have been presented by Logan et al. [1981] and Cunnold et al. [1983a, b].
The use of CFC's in other more essential industries and in previously less industrialized nations (e.g., refrigeration) is expected to grow. We estimate that a 3%/yr growth rate for all of the relatively inert CFC's, i.e., Fll, F12, and C2C13F 3 (Fl13), C2C12F 4 (Fl14), C2C1F 5 (Fl15), and C2F 6 (Fl16), is a reasonable scenario. The range of concentrations shown in Table la is established based on a O-5%/yr growth. For CHC1F 2 (F22), less severe controls are anticipated, since 60% of the emitted amount could be removed in the troposphere. A 5% growth rate with a 3-7%/yr range is used for computations presented in Table l a.

Chlorocarbons
All chemicals (except carbon tetrachloride) in this category have atmospheric residence times of less than l0 years. Be-cause of the toxic nature of many of these chemicals [Surgeon General, 1980], a rapid growth of emissions cannot be expected. Also, because of relatively fast removal rates, a dramatic buildup of these in the global atmosphere is not likely.
Methyl chloride (CH3C1) is the most abundant natural chlorine carrier; it appears to arise mostly from the world's oceans [Lovelock, 1975;Rasmussen et al., 1980;Singh et al., 1983b], although relatively small man-made sources are known to exist [National Academy of Sciences/National Research Council, 1976] and inadvertent release is possible due to biomass burning and/or reactions between organic matter and chlorinated water, as in rivers. Thus it is unlikely that the sources of atmospheric CH3C1 will increase substantially, and we indicate in Table la  Methylene chloride (CH2C12), a relatively short-lived chemical, is a popular solvent which is expected to undergo rapid growth unless found to be toxic in the future. Its virtual noninvolvement in the stratosphere and its lack of toxicity assures it an excellent growth potential. On the average, a 5%/yr (range of 3-7%/yr) growth rate appears a reasonable projection. This growth rate is also consistent with the growth in the last decade [Bauer, 1979].
Over the last decade, the methyl chloroform (CH3CC13) market has grown at a rate of about 15%/yr. Although it may make an increasing contribution to stratospheric ozone depletion, its market is expected to grow rapidly and a growth rate similar to that of CH2C12 is projected. Methylene chloride along with CH3CC13 are the most likely chemicals to be used for substitution as other more to•ic chemicals (e.g., C2HC13, C2C14) are more severely controlled.
Carbon tetrachloride (CC14) is the longest lived atmospheric chlorocarbon, and its historical emission pattern is more complicated [Singh et al., 1976]. Since the early 1960's, when the toxic effects of CC14 became evident, direct emissions virtually ceased. The present atmospheric CC14 concentration growth rate is between 2%/yr [Simmonds et al., 1983] and 5%/yr [Singh et al., 1983a]. Current emission levels of CC14 could grow at a rate of about 2% (0-3%/yr) over the next 50 years, in a manner similar to those of fluorocarbons because a large fraction of CC14 emitted is during its use in fluorocarbon production.

Fully Fluorinated Species
Three chemicals in this cateogry have been measured at enough locations to characterize global concentrations: CF,• (F14), C2F 6 (Fl16), and sulfur hexafluoride (SF6). These three man-made species are relatively stable chemicals with atmospheric residence times over 500 years. These species are not chemically involved in atmospheric processes below about 50 km. The sources of CF4 and C2F 6 are not at all clear. Inadvertent emissions from carbon-electrode processing of minerals [Cicerone, 1979] are likely, specifically from aluminum processing [Penkett et al., 1981]. Assuming a 2-3% steady growth rate of the aluminum industry over the next 50 years, the year 2030 concentrations of CF• and C2F 6 are shown in Table la. Indeed, a temporal increase of about 2%/yr in CF• concentrations has been measured recently (R. J. Cicerone et al., unpublished manuscript, 1985). This rate of increase is less than that deduced by Cicerone [1979]. SF 6 is a dielectric trend, but an upward trend in tropospheric ozone (2-8 km) seems to have occurred [An•Tell and Korshover, 1983].
Other considerations suggest that tropospheric ozone is increasing in the northern hemispheric troposphere. First, Fishman et al. [1979b] presented evidence that there is appreciable in situ photochemical production of ozone. This evidence includes the fact that ozone is more concentrated in the NH even though there should be faster uptake at the earth's surface in the NH. Also, Fishman et al. noted that there are much stronger seasonal variations in the NH production of ozone. Recent models of tropospheric chemistry that embody this theory predict that tropospheric ozone has already increased and will continue to do so, especially in the NH. Due to increases in combustion releases of NO,,, CO, He, and increased CH4, Lo•Tan et al. [1978] calculated that tropospheric 03 can increase greatly, even 100% in the next century, especially in the middle and upper troposphere. Liu et al. [1980] focused on the consequences of NO,, injections from high flying aircraft. In their view, most photochemical production of ozone occurs above the boundary layer, so direct injections by aircraft are especially effective and ground-level NO,, sources may not lead to much photochemical ozone in the free troposphere. Liu et al. calculated that a 14-30% ozone increase should have occurred in the NH middle and upper troposphere between 1970 and 1980. A more recent analysis of tropospheric ozone production by human activities is given by Crutzen and Gidel [1983].
To summarize the two paragraphs above, there is some observational evidence that NH tropospheric ozone has increased by 0.8-1.5%/yr since about 1967; this evidence is compelling but not conclusive. Photochemical theory applied to emission histories and projections of combustion NO,, suggests that a 1%/yr increase in NH tropospheric ozone is possible. In the SH, given the smaller anthropogenic influences, 03 might not change at all (NH anthropogenic NO,,, a key ingredient for photochemical production of 03, should not influence the SH). For our globally averaged radiative calculations, we will adopt an annual growth rate for tropospheric ozone of 0.25%/yr, although values from zero to 1.5%/yr appear possible, at least for the NH. A nominal 40 ppb of tropospheric ozone, for example, becomes 45 ppb in the year 2030 with a 0.25%/yr growth rate; a 1%/yr growth rate would result in 64 ppb in 2030. Note that we do not assume a constant mixing ratio with altitude in our calculations. Table l a provides the adopted altitude variation in the model calculations which are based on the hemispherical, annual mean ozone data described by Ramanathan and Dickinson [1979].
Stratospheric ozone is also thought to be susceptible to perturbing influences, including man-made chloro-and chlorofluorocarbons, increasing CH 4 and N,.O concentrations (see below) and decreases in stratospheric temperature due to increasing CO,_. For our stratospheric ozone profile for the year 2030, we have taken the computed ozone perturbations listed in Table 2. These ozone changes were calculated with the basic chemistry model of  with fixed-flux lower boundary conditions for N,.O, CH3Cl , and CCI,• but a fixed mixing ratio (1.6 ppm) for CH,•. As discussed later, the present computations account for the feedback between temperature and chemistry. A refined diurnal averaging scheme was employed; it led to less nonlinearity in the ozone-chlorine response curve than reported by Cicerone et al. The ozone change shown in Table 2 Wuebbles, private communication, 1984; also see Wuebbles [1983b]). This uniform 3% growth rate is consistent with those adopted in Table la, except that Table la shows a 2% growth rate for CCI,• and a 5% rate for C,.H3Cl 3. In order to place this slight inconsistency in proper perspective, we note that the calculated future stratospheric C1X mixing ratios depend, not only on future emissions, but also on the vertical eddy-mixing coefficient in the (one-dimensional) model. For the 1980 reference atmosphere, we took C1X = 2.5 ppb.
The ozone changes shown in Table 2 account for the feedback between temperature and chemistry. For this purpose, we iterated the temperature changes computed by the climate model (described later) with the ozone change resulting from the chemistry model. The temperature change calculations include not only the ozone changes but also the increase in all other trace gases shown in Table la (see the "Best Estimate" column). This temperature feedback reduced the computed ozone changes by a nonnegligible amount. For example, without the temperature feedback, the computed ozone change at a few of the levels are +4.2% (10 km), +5.1% (20 km), -7.3% (30 km), -45% (40 km), and -36% (44 km); these changes can be compared with those in Table 2.
Our usage of a fixed mixing-ratio lower boundary condition for CH4 actually assumes an increasing flux of CH,• (to maintain the fixed mixing ratio as C1X increases). A fixed-flux boundary condition for CH,• would have led to larger ozone changes than those shown in Table 2 , 1982]. Principal sources of atmospheric methane appear to be enteric fermentation in ruminant animals, release from organic-rich sediments below shallow water bodies and rice paddies, and quite possibly, production by termites and biomass burning. Also, methane re-agent used in electrical equipment. Its concentrations are also estimated based on a 2-3%/yr growth rate.

Nitrogen Compounds
The most important nitrogen-containing chemical from a climatic viewpoint is N20. The 1980 atmospheric concentration is 301 ppb (0.8 ppb less in the southern hemisphere) as measured by Weiss [1981]. Contrary to many previous estimates, it is now accepted that N20 has a very long atmospheric lifetime (> 100 years) with stratospheric photolysis the only known removal process. Microbial production in soils and oceans has been found to be a source as well as a sink of NeO. The net contribution of soils and oceanic microbes to the atmospheric budget of NeO is not yet clear; Weiss [1981] calculated that the total annual source of atmospheric N20 is about 3 x 10 xe g. Over a 4-year period (1976-1980), Weiss measured a rate of increase of 0.2%/yr for N20 concentrations. Further, he constructed a mathematical model which used an exponentially increasing NeO source function to fit his measurement data. More recent data (R. F. Weiss, private communication, 1984) continue to fit Weiss' mathematical model. The recent record shows more uniformity among data from measurement locations than was apparent in the 1976-1980 record. From the Weiss [1981] mathematical model, we estimate the year 2030 concentration to be 375 ppb and a likely range of 350-450 ppb. This range reflects the considerable existing uncertainty as to the identity of the NeO sources most responsible for the observed NeO concentration trend, e.g., coal combustion and microbial production of NeO through nitrification and denitrification of inorganic nitrogen fertilizers applied to soils. Even with the recent expansion of Weiss's data base, the record is still not adequate to distinguish between these two sources (R. F. Weiss, private communication, 1984). In the future, the atmospheric residence time of N20 could decrease if ozone concentrations decrease above 30-km altitude (Table 2); increased UV light levels just below 30 km would increase the rate of NeO photolysis. Also, while we have adopted Weiss's [1981] semiempirical method for projecting future NeO concentrations, we note that there remain many questions about sources of atmospheric N,_O. For example, if we employ a slower rate of increase of fossil fuel combustion than did Weiss, we arrive at a reduced lower limit for NeO in that year, i.e., 350 ppb. Further, while many studies suggest that only 1-2% of all fertilizer N is released as NeO in the year following fertilization, much higher release rates are possible especially from fertilized organic-rich soils [Duxbury et al., 1982].
Two other nitrogen-containing gases, hydrogen cyanide (HCN) and peroxyacetyl nitrate (PAN), have now been observed in the nonurban troposphere. Infrared absorption measurements with the sun as the source have shown that HCN is present in the northern hemisphere (NH), midlatitude troposphere, and the entire NH stratosphere at about 160 ppt, with little, if any, altitude gradient up to the midstratosphere. These measurements, the atmospheric chemistry and possible sources of HCN, have been discussed by Cicerone and Zellner [1983]. In Table la we show no increase in HCN concentration by 2030, but only because there are no data on its temporal trends and because the identities of its sources are uncertain. Similarly for PAN, we can do little more than note that it has been detected recently in the nonurban troposphere, occasionally at concentrations of 400 ppt. While there is reason to believe that its global concentrations are nonnegligible [Singh and Salas, 1983] and that its precursors (NO,,, The calculations account for the feedback between temperature and chemistry within the model stratosphere (above 10 km) and employ the chemistry model of  and the radiative-convective model used in this study. ethane, and propane) might increase in the future, there is too little information to permit an estimate of PAN's future concentrations.
The other nitrogen species (NH 3 and NOx) have extremely short lifetimes (0.5-5 days). The global distribution of these species, even for the 1980 atmosphere, is poorly characterized [Kley et al., 1981]. As expected with species of such short lifetime, a great deal of variability in atmospheric levels is evident. Although anthropogenic sources of NO,• in the troposphere (auto and aircraft exhaust, high-temperature combustion, soil emissions) may double over the next 50 years, it is unclear if this change would increase the atmospheric abundance of NO,• outside the range of present uncertainty.

Ozone
The climatic effects of ozone change depend very strongly on whether tropospheric or stratospheric ozone is being altered [Ramanathan and Dickinson, 1979;Wang, 1982]. Hence we discuss separately the tropospheric and stratospheric 03 trends.
Focusing first on ozone in the free troposphere (above the planetary boundary layer), there are data and theories that suggest that ozone concentrations are increasing with time. A number of investigators [Logan, 1982;Angell and Korshover, 1983; have reviewed and analyzed data from many ozone-measuring stations supported by Umkehr data. Logan finds that at Uccle (Belgium) at 500-and 700-mbar levels, ozone increased by about 1%/yr between 1969 and 1980. Similarly, at Hohenpeissenberg (Germany) at the 500and 700-mbar levels, ozone increased by 15% from 1967 to 1981. However, at Payerne (only 500 km away) at these same altitudes, no such trend was observed. A preliminary analysis of data from nine NH ozonesonde stations has been performed by Liu  . Based on available data, one cannot distinguish a clear lease in mineral, oil, and gas exploration and gas transmission is growing. Clearly, to be able to predict future levels of atmospheric methane, it is necessary to know the relative importance of the various methane sources and their trends. If, for example, rice agriculture is a dominant source, then trends in cultivated area, plant-strain proportions, irrigation, and multiple cropping and fertilization practices must be discerned as they affect methane release.
The dominant sink of atmospheric methane, tropospheric gaseous OH, may not be unchanging. Increased levels of tropospheric CO or of CH,• itself can suppress OH concentrations, as has been noted by several authors. CO exhibits large hemispheric differences; these patterns and our knowledge of CO sources are reviewed by Logan et al. [1981]. Recently, Khalil and Rasmussen have reported evidence from measurement (in Oregon) of dramatic (6%/yr) increases of atmospheric CO between 1979 and 1983. On the other hand, W.

Seiler (private communication, 1984) has measured little or no change (_< 1%/yr) in CO at several stations in both hemispheres. For a clean, background troposphere a CO increase of x% leads to a depletion of tropospheric OH of x/(4 + 1)%, depending on altitude and various model assumptions, according to A.M. Thompson and R. J. Cicerone (unpublished manuscript, 1984). Combining the (1)source analysis by Logan et al. [1981], (2) information on trends of these sources (e.g., fossil fuel usage, oxidation of anthropogenic hydrocarbons), and (3) the CO data mentioned above, it is clearly possible that CO will increase by 1-2%/yr through 2030 A.D. Such an increase could cause CH,• concentrations to increase faster (through OH suppression) than if only CH,• source increases were considered. Because of the spectral locations of the absorption of CO, the CO increase itself is not of interest here. Its effects on the atmospheric levels of OH, CH,•, and 03 could be very important.
Lacking all the detailed information necessary to understand the presently documented rate of increase of atmospheric CH,• concentrations and to predict the future, we estimate that CH,• will increase by 0.75%/yr between now and 2030; this would lead to a globally averaged CH,• concentration of 2.34 ppm in 2030. Rates of increase of 1.5% and 0.25%/yr would lead to 3.30 and 1.85 ppm in 2030, respectively, as listed in Table la. Beyond the year 2030, when the release of continental-slope sediment methane clathrates might occur due to oceanic warming [Revelle, 1983], faster methane increases are possible.

Nomethane Hydrocarbons
In this category we include alkanes, alkenes, alkynes, aldehydes, ketones, and H2. We pay little attention to simple aromatic compounds. By contrast with the situation for methane, there is too little information available on the concentrations, distributions, and sources of these compounds (except for C2H2 and H2) to justify projections of future concentrations. For acetylene (C2H2), fossil fuel burning (e.g., internal combustion engines, oil-fired heaters) is a known source; its atmospheric residence time is about 4 months, and no significant biological sources are known yet [Rudolph and Ehhalt, 1981]. If its sources are wholly anthropogenic, an annual increase of 1-2% would be a reasonable guess. For ethane (C2H6), ethylene (C2H,0, propane (C3H8) , and propene (C3H6) the existing atmospheric and oceanic surface water data suggest that there are natural as well as anthropogenic sources [-see, e.g., Rudolph and Ehhalt, 1981, and references therein]. Fugitive emission• from oil and gas wells and transmission lines are likely, of course. The state of our measurement data base for higher hydrocarbons, aldehydes (present as oxidation products of hydrocarbons), and acetone is discussed by Penkett [1982].

Sulfur Compounds
Carbonyl sulfide (OCS) is the most abundant gaseous sulfur carrier in the atmosphere. It is nearly uniformly distributed with a measured average concentration of 0.52 ppb. Turco et al. [1980] have examined the sources and sinks of OCS. While there are many remaining questions, they propose that up to 50% of the total source is anthropogenic. If so, OCS concentration could increase in the future, but our present understanding of OCS sinks and its atmospheric residence time is not very complete. No measured trend in OCS concentration is available at this time. Considering the lack of such data and the uncertainties about OCS sources and sinks, we cannot project other than a constant OCS abundance in Table la.
Sulfur dioxide (SO•) is a notorious atmospheric constituent because of its role in acid deposition. In continental boundary layers where its principal source is combustion of S-containing fuels, its concentrations are often 10 ppb. Above the boundary layer its concentration is of order 100 ppt [Maroulis et al., 1980]; its presence there is probably due to escape from the boundary layer below, and to oxidation of other species, e.g., OCS, CS2, and CH3SCH 3. Similar concentrations have been measured in the marine boundary layer [Herrmann and Jaeschke, 1984]. Near major anthropogenic SO2 sources its atmospheric residence time is about 1 day (due largely to gas-to-particle conversion); in the higher troposphere in clear air its residence time is up to 1 week. Because of its very short lifetime and the uncertain future of SO2 emission, it is not at all clear that SO2 concentration will increase in the future. Dimethyl sulfide (DMS) is now known to exist in the oceanic boundary layer; it appears to have microbial sources in oceans that provide a significant DMS flux to the marine atmosphere [Andreae and Raemdonck, 1983]. Because this natural source appears to be the major DMS source and because of the short (• 2 days) atmospheric residence time of DMS, we project no growth in its atmospheric concentrations.
Carbon disulfide (CS2) is known to be present in background surface air at concentrations that vary from 0.03 to 0.08 ppb. However, it is virtually undetectable in the free troposphere. Excited-state oxidation [Wine et al., 1981] can be an important removal process. Oceans may be a major source. No growth projection in CS2 concentrations can be proposed reliably at this time.

Brominated and Iodated Species
Only a handful of species in this class have been measured in the nonurban atmosphere. The species of interest are brominated and iodated methane-and ethane-series molecules: methyl bromide (CH3Br), methylene bromide (CH2Br2) , bromoform (CHBr3), bromotrifluoromethane or (F13B1, CBrF3), methyl iodide (CH3I), and dibromoethane (C2H•Br2, or ethylenedibromide, EDB). CH3Br is apparently a natural species [Lovelock, 1975;Sinqh et al., 1983b]. Man-made emissions have the potential to perturb its global background, but only slightly. Methyl iodide is essentially all natural and, given its concentration, is predicted to remain unchanged. CHBr 3 and CH2Br 2 have been measured only recently [Berq et al., 1984], and too little is known about their sources to permit reasonable future projections. CBrF3 (F13B1) and ethylene dibromide are exclusively anthropogenic. A continued shift toward nonleaded gasoline could offset growth that may occur in fu-migation applications of ethylene dibromide. Because of its high carcinogenic potential, a rapid growth is not likely to be permitted in any case. CBrF 3 (used as a fire extinguisher) is the only brominated organic whose sink is primarily in the stratosphere (where bromine atoms can be efficient ozone destroyers). Despite its very low present abundance it can become an important carrier of organic bromine within the next 50 years. Inorganic species such as HBr, HI, BrO, IO, and NOBr are not discussed here because they are as yet undetected in the atmosphere. Their residence times are probably 5 days or less, and future trends are difficult to predict.

PREINDUSTRIAL ERA CONCENTRATIONS OF GREENHOUSE GASES
It is important to ask if CO2 and the other trace gases should already have caused a global warming. It is very difficult, if not impossible, to answer this question for at least two reasons: (1) there are no direct data on the trace gases of interest from, say, the 1850--1940 time period, and (2) there should be a significant time lag between any increased atmospheric radiative forcing and increased global temperatures, due to oceanic heat capacity. To allow at least a rough estimate of the size of the effect due to increased trace gas concentrations from about 1880 until 1980, we will attempt to estimate the 1880 concentrations of methane, nitrous oxide, chlorofluorocarbons, CC14, and tropospheric ozone. The proposed preindustrial concentrations of the trace gases are shown in Table lb. For CO2, the National Research Council [1983] study suggests that the most likely preindustrial value is between 260 and 290 ppm. For this study, the preindustrial concentration of CO2 is assumed to be 275 ppm.
For CH4, the data of Craiq and Chou [1982] show that CH4 has increased monotonically for the past 400 years; these data are CH4 concentrations in air trapped in dated Greenland ice cores. Craig and Chou noted that there is as much as a 90 year uncertainty in the age of this air, depending on whether air moved freely throught the firn phase of the snow above the firn-closure depth. If the air at the 90-year firn level was zero years old, then the CH4 concentration in the year 1880 was about 1.05 ppm. If the air there was not so young, the Craig and Chou data show that the CH4 concentration in 1880 had to be over 1.05 ppm. If, for example, the air in 90-year old ice at this site were 50 years old instead of zero years old, the implied 1880 CH4 concentration would be 1.10-1.15 ppm in 1880. We assume that the 1880 CH• level was 1.15 ppm.
For nitrous oxide, there are no direct data from pre-1900; indeed, N20 was discovered in the atmosphere only in 1938.
Modern data from 1976--1980 and from 1961-1974 have been used by Weiss [1981] to estimate a preindustrial atmosphere N20 concentration of 281-291 ppb. Accordingly, we assign a value of 285 ppb to N20 for the year 1880. The chlorofluorocarbons and fluorocarbons (CC12F 2, CC13F, and the other compounds listed in Table 1  Therefore we estimate that each of them was absent from the 1880 atmosphere. Carbontetrachloride is more interesting. It is known to be produced by marine organisms [see, e.g., Fenical, 1982], yet its mid-1970's (and present) concentration can be explained by anthropogenic emissions [Sinqh et al., 1976]. Because of the apparent unimportance of current natural sources, we will assume that it was essentially absent from the 1880 atmosphere.
Of the important greenhouse gases, tropospheric ozone is most difficult for which to estimate differences between present-day concentrations and those of one century ago. The surplus of NH over SH ozone, the more pronounced seasonal cycle in the NH ozone data, the strong theoretical case for excess ozone production in the industrialized NH and the hints of a positive trend since 1967, all imply that there was less 0 3 in the 1880 NH troposphere. Detailed examination of these and other factors [see Levy et al., 1985] does not allow one to state with confidence that the hemispheric or global background tropospheric ozone is strongly controlled by photochemical reactions (such as those between hydrocarbons and nitrogen oxides to produce ozone). For example, there is evidence that the NH troposphere receives perhaps 3 times as much ozone from the stratosphere as does the SH troposphere. One would predict higher NH concentrations from this meteorological input of ozone, although the higher surface-destruction rates in the NH would offset some of the additional input. Perhaps 50% more ozone is observed in the NH tropics than in the SH tropics (0-to 12-km altitude; Fishman et al. [1979b]) and 25-50% more in the midlatitudes of the NH at 800-mbar pressure levels than at corresponding SH locations. Another feature of ozone in the NH midtroposphere, the east-west gradient over North America [Chatfield and Harrison, 1977], appears to be evidence for photochemical production over continents. We assume that half of the difference between NH and SH is due to anthropogenic emissions (and that 1880 emission of NOx and hydrocarbons was negligible compared with those in 1980). Even with these assumptions, one is left with uncertainty about vertical profiles. In the upper troposphere, there is more influence from the stratosphere, but there is also significant existing potential for human impact by direct injections from aircraft [Liu et al., 1980]. As a very rough estimate, we will guess that there was 25% less ozone in the 1880 NH troposphere than in 1980 NH troposphere and that SH tropospheric ozone did not change during that century.

DESCRIPTION OF THE CLIMATE MODEL
The direct radiative effects of trace gases are included in this study. The effects due to altered chemistry are included explicitly as far as stratospheric 03 perturbation is considered and implicitly with respect to tropospheric 03, i.e., projected 03 increases can be considered to arise from the projected increases in hydrocarbons, CO, and NO. With respect to the feedback effects, this study accounts for troposphere/ stratosphere radiative interactions and the feedback between temperature and chemistry within the stratosphere. The climate-chemistry interactions in the troposphere and the possible effects of temperature changes on stratospheric H20 are ignored. Both of these feedback effects, while they may be relatively smaller than the direct radiative effects would require coupled photochemical climate models [Callis et al., 1983].
A brief description follows of the radiative-convective model and the source for the spectroscopic data used for the computations.

Radiative-Convective Model
The one-dimensional radiative-convective model described by Ramanathan [1981] is adopted. This model, hereafter referred to as model R, has a surface boundary layer which explicitly allows for the surface-atmosphere exchange of latent and sensible heat and also solves for the boundary layer moisture [Ramanathan, 1981, equations 17-20, Figure 7]. The standard radiative-convective models [e.g., Manabe and Wetheraid, 1967; Ramanathan and Coakley, 1978] ignore these processes and do not treat explicitly the exchange of latent heat flux between the surface and the free atmosphere. The boundary layer moisture and hence the relative humidity are explicitly computed in the model, but the tropospheric relative humidity is prescribed as discussed by Ramanathan [1981]. The mass mixing ratio of H20 above 12 km is prescribed to be 2.5 ppm. Because of the explicit treatment of the boundary layer, the model R computes the surface temperature and the surface air temperature. Standat,t radiativeconvective models compute only one temperature for the lower boundary which can be interpreted as an average of the surface and surface air temperature. In model R the surface air temperature change is larger than that of the surface temperature change by about 10--13%. This point should be noted when comparing the present calculations with the published results.

Radiation Model and Spectroscopic Data
The trace gases, their longwave band centers, and the adopted band strengths are shown in Table 4. The treatment of H20, 03, CO2, and CH4 are as described in model R. For CO2, one of us (J.T.K.) incorporated the more detailed band model of Kiehl and Ramanathan [1983] in model R. The surface warming due to doubled CO2 estimated with the detailed CO2 scheme was in excellent agreement (within 5%) with that estimated from the somewhat simpler scheme in model R. For CH4, model R employs the Donner and Ramanathan [1980] band model. Although the band strength adopted in model R is stronger than the current accepted value by about 35%, the band model parameters were fit to give agreement with laboratory absorptances. Hence the CH4 radiative forcing estimated by this band model agrees within 5-10% of that estimated from a 5 cm-• spectral resolution narrow band model which employs recent [Rothman et al., 1983] line data. Numerous modifications were incorporated in model R to treat the effects of the minor trace gases included in this study and these modifications are described below.

N20. The R scheme used the Donner and Ramanathan
[1980] band model scheme. We have retained this scheme, but included the following N20 bands that were ignored by Donner and Ramanathan [1980]' the two-band systems centered at 1168 cm-•, one of which is the 2v4 band of the four isotopes with bandstrength 8.5 cm-• (cm atm)-• STP and the hot bands of the four isotopes with band strength 1.5 cm-• (cm atm -•) STP. Although these bands are considerably weaker than the fundamental v• band system centered at 1285 cm-•, they contribute as much as 20% of the v• band system to the ..... • 'surface warming due to N20. This disproportionately large contribution by the weak bands arises because N20 is almost in the strong line limit, and hence the opacity scales as the square root of the band strength. Hence for gases whose concentration are large and their band strengths are sufficiently strong that they are in the strong line limit, great care must be exercised in including all the bands whose strengths are smaller by as much as 2 orders of magnitude than the strong fundamental band. This is the reason why the present model incorporates many isotopic and hot bands of trace gases such as CO2, N20, and CH 4.

Other trace gases. The band absorption A is expressed as A= Am[1-•••e-•'/•'-] (1)
where Am is the band width in cm- It can be easily shown that (6) is the exact expression for the band absorption in the optically thin limit. Ramanathan [1975] used this expression to treat the CFC13 and CF2C12 bands.
2. The smeared out line structure limit. In this limit, the line spacing between neighboring lines is much smaller when compared with the line half-width. Consequently, the lines are smeared out and the absorption coefficient follows a smooth variation with wavelength. This limit is adequately satisfied for the CFC13 and CF2C12 bands, as can be inferred from the   Tables 3 and 4 Table 4. These authors give a range of measured values, which are reproduced in Table 4  bands and integrated band strengths were kindly supplied to us by H. Niki (personal communication, 1983), whose values are shown in Table 4 as N.

Band strengths' Except for CFC13, CF2C12, CH3CC13, CFC22, and PAN, we have relied heavily on the published summary by Pugh and Rao [1976], denoted as P in
6. The CFC's and other trace gases besides CO2 are assumed to be mixed uniformly from the surface up to 12 km, above which the mixing ratio is assumed to decrease with a scale height of 3 km. However, as shown by Fabian et al. [1984], the mixing ratios of long-lived species such as CF4, C2Fa, and CF3C1 are nearly constant from the surface to the middle stratosphere (•30 km). In the next section we will examine the sensitivity of the greenhouse effect to the mixingratio profile. CH4, N20, and CO2 are assumed to be uniformly mixed throughout the atmosphere. The observed mixing ratio of CH4 and N20 decreases with height above 12 km. However, as shown in the next section, the computed surface warming due to CH,• and N20 increase is insensitive to their mixing-ratio profile in the stratosphere. The 03 profile is taken from the hemispherical, annual data given by Ramanathan and Dickinson [1979]. In summary, the number of trace gases included in this study, together with the treatment of the overlap between the various gases and the details of the various fundamental, hot, and isotopic bands, make the present study the most comprehensive climate model calculations that have been performed so far for the trace gas climate effects.

Accuracy of the Trace Gas Radiative Treatment
The accuracy of the treatment as given by (1) Table 5, the change in the net (down-up) flux at the tropopause, i.e., the surface-troposphere heating, due to increase of CFC11 from 0 to 2 ppb as computed by the reference model is compared with that obtained from the present scheme, i.e., (1)-(5). Also shown in this table, for comparison purposes, is the result obtained from (6), which is the form of the optically thin limit equation employed by Ramanathan [1975]. It is seen that the present scheme is in excellent agreement with the more detailed calculations. The equation for the optically thin assumption employed by Ramanathan [1975] overestimates the heating by about 16%. However as shown later, the sur- •-CH4 and N20 are assumed to be uniformly mixed from the surface to the top of the atmosphere. However, these calculations, when repeated with observed profiles which show a decrease in the mixing ratio above 12 km, yield AT results identical to those shown in this $For this gas and for all the others following it, the mixing ratio decreases above 12 km with a scale height of 3 km. The computed surface warming is larger by about 15% when the mixing ratio above 12 km is held constant at the surface value. face warming due to increase of CFC's computed with the present accurate scheme is in excellent agreement with that estimated by Ramanathan [1975]. The agreement is because of the following compensating effects:(1) the band strengths used in Rarnanathan [1975] are smaller by about 10% than the recent values used in this study; (2) the 16% error shown in Table 5 is reduced somewhat with the inclusion of clouds; (3) the error in the optically thin approximation is smaller for the other bands of CFCll and 12 whose strengths are considerably smaller than the 11.8-#m CFC11 band. Cess [1982] has also performed similar narrow-band calculations for CFC13 and showed that the smeared-out line structure assumption, which is invoked in arriving at (1), is an excellent approximation for CFC bands. In summary, we conclude that the present scheme is very accurate. The only source of remaining error is the neglect of the temperature dependence of the hot bands of CFC's that have been detected, in the vicinity of the strong CFC13 and CF2C12 bands, by Varanasi and Ko [1977] and Nanes et al. [1980]. While the effects of the hot bands are included in this study, the temperature dependence of their band strengths arising from the temperature dependence of the excited vibrational states is ignored. Nanes et al. [1980], however, suggest only a weak temperature dependence. As a note of caution, we add that the temperature dependence mentioned above should not be confused with the temperature correction that is needed to convert band strengths, S, measured at temperature T to STP conditions. Recall that in (1), the path length w is in cm atm, STP, and hence S measured at a temperature T in the units of

cm-• (cm atm)-• should be multiplied by (T/273) to convert to the units of cm-• (cm atm)-•, STP. In some instances in
the literature, this correction factor has been confused for the temperature dependence of S. The procedure of employing the correction factor (T/273) is rigorous for the fundamental CFC bands. For the hot bands, however, we need an additional term to account for the temperature dependence of the excited vibrational state [e.g., see Kiehl and Ramanathan, 1983, equation 12].

Uniform Increase in Certain Trace Gases
Before presenting the results for the trace gas scenario shown in Table l a, we will discuss results for a hypothetical case of 0-1 ppb increase for several of the trace gases. The purpose of this exercise is twofold: to elucidate the processes that determine the magnitude of the trace gas effects and to identify the most important trace gases from a climate viewpoint. For these two objectives, we avoid specific scenarios to assure that conclusions are not scenario dependent.
The computed surface warming due to a 0-1 ppb increase in 15 different trace gases is shown in Figure 1. The tropospheric 03, CH½, and N20 effects are relatively better known and are shown merely for comparison purposes. Several interesting and rather surprising features of the results shown in Figure 1 are noted below. • Ramanathan [1975], this study, and the GISS models (items 4-6) assume a constant CFC mixing ratio from the ground to 12 km, and above 12 km the mixing ratio decreases exponentially with a scale height of 3 km. Reck  5The FCA model results were not mentioned by Ramanathan [ 1975] but were obtained for the purposes of the present comparison. 6Reck and Fry gave AT results for 1-ppbv increase, which was linearly scaled for the 2-ppbv increase. 7CFC mixing ratio as described by Ramanathan [1975]. See footnote 1 above. 8The lower value is surface temperature change, and upper value is surface air temperature change.
9CFC mixing ratio is constant from surface to top of the atmosphere.  Table 4, the strongest bands of these gases are stronger than the stro_n_gest band.q of CFC13 and CF2C12 by more than a factor of 2. The location of the band center, however, plays a crucial role because of the overlap effect. For example, in the 1295-cm-• region, the absorption by N20, CH,,, and H20 bands saturates this region, and hence trace gases in this spectral region have relatively lesser impact on climate. This is the primary reason why CF,,, although possessing a band in the 1285-cm-• region which is stronger than any other bands in the 8-to 20-ttm region, produces a surface warming of only 0.06 K. 2. The sensitivity of surface temperature to tropospheric ozone has been anticipated earlier [Ramanathan and Dickinson, 1979] but has not received much consideration elsewhere in the literature.
3. Although PAN has several moderately strong bands, its strongest band is in the middle of the strong 6.3-ttm H20 bands. Similarly, although CCI,, and CHC13 have moderately strong bands, their strongest bands are located in the 774-cm -• region, where they are overlapped by CO2, H20 rotation bands, and the H20 continuum. 4. To illustrate the importance of the overlap problem, we show in Table 6 the surface temperature increase with and without the overlap effects. Such results, besides illustrating the contribution from various radiative processes, also facilitate model intercomparison study by enabling the identification of the sources for model differences. We also show in this table, the surface and surface air temperature change for one of the two cases. In general, surface air temperature is larger than the surface temperature change by about 10%. It is clear from Table 6 that for several trace gases, e.g., CF,,, CHC13, CH3CC13, C2F6, the overlap effect ameliorates AT s by factors of 1.5-2.
We will now compare the present estimates of A Ts with other published estimates for a few of the trace gases. Consid-er first CFC13 and CF2C12 for which the differences between the various model estimates of the global surface temperature change are disturbingly large as illustrated in Table 7. The differences shown in Table 7 can arise from differences in the radiative treatment and (or) from differences in the model sensitivity. In order to isolate these two sources, the climate sensitivity parameter, 2, as estimated from various models is shown in Table 8. As explained by Dickinson [1982] and Ramanathan [1982], 2 and ATs are approximately related by ;AF where AF is the radiative forcing of the surface-troposphere system due solely to trace gas increase, i.e., CFC increase in the present example. As described below, Tables 7 and 8 provide the answers for all of the differences in the computed A T•. 1. There is a wide spread in the measured band strengths. For example, Figure 2 shows the measured band strengths by various investigators for the strongest CFC bands. The present study uses the recent measurements of Kaqann et al. [1983], whereas all of the other studies employ the earlier measurements. The values used by Ramanathan [1975] underestimate (compare item 12 with item 10) AT s by 8% while Varanasi and Ko's values employed in the Goddard Institute for Space Studies (GISS) models underestimate (compare item 13 with item 10) A Ts by 10%.

With respect to the radiative treatment, Ramanathan [1975] and Chamberlain et al. [1982] employ (6)
, which is one form of the optically thin approximation. By comparing item 12 with 1, both of which use the same band strengths, it is seen that (6) overestimates A Ts by about 5%. However, the present study which uses an accurate procedure is in excellent agreement with Ramanathan [1975] because of the compensating effects of the smaller band strengths used in that study. The other models cited in Table 7, unfortunately, do not give the equation or the details of their radiative treatment. However, all of these models rely primarily on integrated band strengths and hence must employ the optically thin approximation, but not necessarily (6).
3. The unrealistically large value obtained by Chamberlain et al. [1982] results primarily from their approach of esti- Here, fa __ 2(FCT)/2(FCA); f(empirical) = 2(empirical)/2(FCA). See Table 7 (footnotes 2 and 3) for explanation of FCA and FCT. mating the CFC heating from the change in the net flux at the surface rather than at the tropopause (see Ramanathan [1982] for more details on this topic). Karol et al. [1981], and Reck and Fry [1978] assume the CFC mixing ratio to be uniform from the surface to the model top, whereas Ramanathan [1975], the present study, and the GISS models assume an exponentially decaying mixing ratio (see footnote 1 in Table 7) in the stratosphere. Our model calculations, when repeated with a constant mixing ratio, show that the constant mixing ratio overestimates (compare items 10 and 11) ATs by about 15%. Reducing these constant mixing-ratio model estimates by 15% would bring them closer to the present values.  Table 7).

This brings us to the GISS models. First note from
Their model result of 0.46 K when corrected for the above differences increases to 0.56 (0.46 x 1.11 x 1.1), which is in close agreement with the present estimate of 0.55-0.6 K. In summary, we can account for almost all of the differences between the various model estimates. The analyses lead to the conclusion that, in view of the accurate treatment of the CFC radiative effects and in view of the detailed and up-to-date spectroscopic information incorporated in this study, the present estimates of 0.55-0.6 K for CFC increase from 0 to 2 ppb should be considered as the state-of-the art estimates for a radiative-convective model with fixed cloud top altitude.
For CF4, the only available calculation is that of Wang et al. [1980], who compute a surface temperature change, for 0-1 ppbv change in CF4, of 0.07 K to be compared with the present value of 0.06 K. Although the close agreement is reassuring, there are substantial differences in the CF4 treatment between the two models. Wang Hummel and Reck [1981] have computed the effects of CHCIF 2 (CFC22), CH3CCI3, and CH2CI 2. Their surface warming for CHCIF 2, CH3CCL3, and CH2CI2 are, respectively, (for 0-1 ppb increase) 0.04, 0.02, and 0.01, and the respective results of the present study are 0.05, 0.02, and 0.01 K. The two models seem to be in good agreement. Since Hummel and Reck [1981] have given neither the quantitative details of their radiation model nor the band strengths, it is difficult to rule out the possibility of a fortuitous agreement.
Sensitivity to vertical distribution. The sensitivity of the computed surface temperature change to vertical 03 distribution is discussed in the next section. The discussion here is restricted to those gases that are uniformly mixed in the troposphere, i.e., all of the gases, other than 03, shown in a scale height of 3 km. This nonnegligible sensitivity in the computed surface warming for these trace gases is largely because of the substantial difference in the stratospheric abundance between the uniform mixing-ratio profile and the profile with a 3-km scale height.  Table l a. face cooling due to CFM-induced ozone perturbations. The major source of discrepancy is in the adopted stratospheric 03 perturbation profile. The present profile is based on the most recent chemistry and reaction rates, and it shows that large decreases in middle and upper stratospheric 03 profile are accompanied by somewhat smaller percentage increases in the lower stratosphere. Additional calculations were performed to examine the sensitivity of the computed surface warming to vertical distribution of 03 change, which lead to the following inferences. The profile shown in Table 2 leads to a surface warming of 0.08 K. Roughly, 0.06 K i• due to the 03 decrease above 30 km, and the remainder of 0.02 K is due to the 03 increase below 30 km. Thus the 03 decrease above 30 km as well as the 0 3 increase below 30 km contribute to a warming. The perplexing nature of this result can be understood from the detail analyses given by Ramanathan et al. [1976] and Ramanathan and Dickinson [1979], and hence only a brief discussion is given below.

5.2ß Climatic Effects of the Projected Increases
A decrease in stratospheric 0 3, irrespective of the altitude of the decrease, would lead to an increase in the solar radiation reaching the troposphere, and this solar effect would tend to warm the surface. However, 0 3 also alters the IR (longwave) emission from the stratosphere in two ways' first, the decreased solar absorption (due to 03 decrease) cools the stratosphere' the cooler stratosphere emits less downward to the troposphere. Second, a decrease in 03 reduces the absorption (by the 03 9.6-/•m band) of the surface-troposphere emission. This reduction causes an additional cooling of the stratosphere, which in turn, causes an additional reduction in the downward IR emission by the stratosphere. Thus the IR effects of 03 decrease tend to cool the surface. However, the IR opacity of stratospheric CO2, H20, and 03 is sufficiently strong that the impact of the reduction in IR emission (by the stratosphere) on the surface diminishes with an increase in the altitude of 03 perturbation. On the other hand, the surface warming induced by the solar effect is independent of the altitude of 03 perturbation. Consequently, for a decrease in 03 in the upper stratosphere, the solar effect dominates (leading to a surface warming), while for a decrease in the lower The first four chlorofluorocarbon gases here could contribute significantly to future global warming because of the spectral positions of their absorption bands in the 7-13 ttm atmospheric window region if their band-absorption intensities are large enough. For other gases, e.g., HCN and SO2, their absorption bands are not strong enough to be significant at present or near-present atmospheric concentrations. For other gases such as C2H 6 and CH3F, we have too little information to be able to estimate future trends. stratospheric O3, the IR effect dominates (leading to a surface cooling).
The uncertainty in our computed surface warming due to stratospheric 03 change is best illustrated by the following examples. The profile of O3 change in Table 2 for altitude above 30 km when combined with 10% uniform 03 decrease between 12 and 30 km leads to a surface warming of 0.02 K; whereas the same profile (as in Table 2) above 30 km when combined with a 10% uniform 03 increase between 12 and 30 km leads to a surface warming of 0.1 K. In view of the high sensitivity of the computed temperature change to the vertical 03 profile, our computed estimates for stratospheric 03 change should be viewed with caution because such distributions would be influenced by atmospheric dynamics (whose effects are ignored in this analysis) and of course by remaining uncertainties in model chemistry.
The vertical distribution of the computed atmospheric temperature change is shown in Figure 4. It is clear from this figure that other trace gas effects on temperatures are comparable to CO2 effects, not only for surface warming, but also for stratospheric cooling. The stratospheric cooling, to a large extent, results from the stratospheric 03 reduction. The potential climatic effect of gases that are not explicitly discussed in Figure 3 is summarized in Table 9.

Effects of the Inferred Trace Gas Increases From the Preindustrial to the Present Levels
For the sake of discussion, the concentrations for the year 1880 are associated with the preindustrial levels, and these concentrations have been shown in Table lb, while the observed 1980 concentrations are shown in Table la. The computed equilibrium temperature changes are shown in Table 10. The CO2 increase causes a surface warming of 0.52 K, which is enhanced by 50% due to the increase in the other trace gases. The computed stratospheric cooling due to CO2 in-crease is substantial, but that due to other gases is negligible. The computed stratospheric cooling would be larger had we included the effects of stratospheric 03 decrease due to increases in CFC's. From Table 10b, which shows the contribution of the individual gases, it is seen that CH,,, tropospheric 03, and CFC's are the largest contributors, next to CO2, to the surface warming computed for the period 1880-1980.

SUMMARY
The basic conclusion that can be derived from the present study is that the radiative effects of increases in trace gases (other than CO2) are as important as that of CO2 increase in determining the climate change of the future or the past 100 years. Several tens of man-made chemicals have been detected in the troposphere and about 20 of these have strong absorption features in the 7-to 13-/•m regions of the longwave spectrum. The present-day, taken as the year 1980, concentrations are taken from in situ observations. A careful analysis of the measured trends from early 1970's to 1980 form the basis for the concentrations projected 50 years into the future. Published ice-core CH,• observations, surface-based 03 observations, and other studies are used to infer the trace gas concentrations for the preindustrial era. The equilibrium surface and atmospheric temperature changes estimated with the aid of a radiative-convective model reveal the following features.
1. The preindustrial to present-day increase in CO: causes an equilibrium surface warming of 0.5 K in the model, which is enhanced by a factor of 1.5 by the increases in the other trace gases. CH,•, tropospheric 0 3, and CFC's are the largest contributors to this enhancement. The upper stratospheric cooling due to the CO2 increase is as large as about 3 K.
2. The projected CO• increase from 339 ppm in the year 1980 to 450 ppm in 2030 warms the model surface by 0.7 K, which is enhanced by a factor of about 2.1 by the other trace gases. The factor of trace gas enhancement varies from about 1.5 to 3 depending on the assumed scenario. The trace gases that contribute to this significant enhancement are CFC13 (Fll), CF2C12 (F12), CH4, N20, stratospheric and tropospheric 03. Somewhat smaller but nonnegligible contributions arise from CHC1F 2 (F22), CH3CC13, and CFC13.
3. The stratospheric 03 changes resulting from the assumed increase in CFC's and other chlorine compounds by year 2030 are a large decrease in middle stratospheric 03 accompanied by a slight increase in lower stratospheric 03. Hence although CFC's are estimated to cause only a slight reduction in the column ozone, the significant perturbation to the shape of the 03 profile leads to a nonnegligible surface warming of 0.08 K. Thus the 03 change due to CFC's adds to the surface warming of 0.36 K resulting from the CFC direct radiative effects. These two effects when considered together make CFC's the largest contributors (next to CO2) to the overall surface warming computed in this study. However, because of the strong sensitivity of the computed surface warming to the vertical profile of 03 change, the magnitude of the potential surface warming due to stratospheric 03 change is highly uncertain.
4. All of the other trace gases perturb the vertical atmospheric profile in the same manner as CO2 in the following sense' they warm the surface and the troposphere while cooling the stratosphere (above 20 kin) significantly. However, there is one important difference between the radiative effects of CO2 and the other trace gases' as pointed out by Dickinson et al. [1978], CFC's have a strong warming on the tropical tropopause. Also the studies by Rarnanathan and Dickinson [1979] and Fels et al. [1980] reveal the significant sensitivity of tropical tropopause to 03 perturbations. Warming of the tropical tropopause by 2-3 K could lead to large changes in stratospheric water vapor. 5. On a ppb basis, CF3C1 has the strongest greenhouse effect (exceeding very slightly even that of CFC12) followed closely by CBrF3, CF2C12 (F12), CHF3, CFC13 (Fll), and C2F 6 (Fl16), all of which have effects comparable to that of Possible changes in stratospheric O 3 are ignored. *Surface air temperature change is larger than the surface temperature change by about 10-13%. CFC13 or CF2C12. Gases such as CF,•, CCI,•, and PAN have strong absorption features, but due to the overlap with CH,•, N20, CO2, and H20 bands, these gases are not very effective in enhancing the atmospheric greenhouse effect. However, our conclusion concerning the overlap effects should be considered as tentative. Measurements of narrowband spectroscopic parameters for these other trace gases are currently not available, and such measurements are needed for improving the accuracy of the estimates for the overlap effects. For important species such as C2C13F 3 (Fl13), C2C12F• (Fl14), and C2C1F5 (FllS)even band strengths are not available. 6. The accurate radiation model developed here for CFC13 (Fll) and CF2C12 (F12) helped sort out the differences between the various published studies for the estimated surface warming.
The important implication of this study is that the preindustrial to the present increase in CO2 and the other trace gases might, very likely, have caused a significant perturbation to the radiative heating of the climate system. This perturbation radiative heating induces a warming of about 0.8 K in the present model, whereas it might have induced a warming twice as large in recent GCM's [Washington and Meehl, 1984;Hansen et al., 1984]. These GCM's compute a 4 K global warming due to CO2 doubling as opposed to the 2 K yielded by the radiative-convective model, The 0.8-1.6 K global warming, had it indeed occurred from the preindustrial to the present, should have been detectable above the statistical fluctuations of the climate. This is a controversial issue, and the published papers have contradictory results. Hansen et al. [1982] suggest that the CO2 warming is discernible from observed records of global or hemispherical average temperatures. The statistical analysis of the 70-year homogeneous (in time and in longitude) temperatures for 50-70øN by Madden and Rarnanathan [1980] has failed to reveal the CO2 effect.
Before we can attempt to verify the greenhouse theories of preindustrial to present warming by comparing model results with the observations, the following important issues must be dealt with.
1. The one-dimensional and GCM results pertain only to the equilibrium warming of the surface to a step-function increase in the trace gases. The quantity of interest, for the purpose of verification, is the transient climate response to a time-varying distribution of trace gases. Even assuming that time-dependent trace gas distribution is known (for the past 100 years), our understanding of the ocean mixed layer interactions with the atmosphere and the thermociine as well as the lateral ocean heat transport is too imprecise to estimate reliably the transient response of the climate system.
2. Other climate forcing terms, e.g., solar irradiance, volcanic aerosols, surface radiative properties, can also change on the time scales of interest to this study, and we do not have adequate data bases to estimate their contributions to past climates. Major volcanic events, such as E1 Chichon, can cause an equilibrium global cooling of 0.5-1 K, but the aerosol residence time is about 2 years or less, and as yet, we have not come to grips with the tough issue of estimating the transient response to an episodic forcing.
3. The last issue concerns the source of errors in observations of temperatures, humidities, and other trace gases arising from instrumental and sampling biases. There are no reliable global measurements for key components like lower stratospheric H20 and tropospheric 03.
It is hoped this study will provide more scientific justification for making some key measurements on a long-term basis of trace gas trends (at least the top 12 gases identified here), stratospheric aerosols, stratospheric humidity, and tropical tropopause temperatures. Of equal importance, accurate measurements of narrow-band spectroscopic parameters and band strengths for the trace gases are urgently needed.