Changes in Stratospheric Ozone

The ozone layer in the upper atmosphere is a natural feature of the earth's environment. It performs several important functions, including shielding the earth from damaging solar ultraviolet radiation. Far from being static, ozone concentrations rise and fall under the forces of photochemical production, catalytic chemical destruction, and fluid dynamical transport. Human activities are projected to deplete substantially stratospheric ozone through anthropogenic increases in the global concentrations of key atmospheric chemicals. Human-induced perturbations may be occurring already.

Nerlove, Estimation and Iden ofCobb-Douglas Production Functions (Rand McNally, New York, 1965). 15. Z. Griliches, Am. Econ. Rcv. 76, 141 (1986). 16. These numbers differ from Fig. 1 both because they refer to the specific sample used by Griliches (15) and because they are averages of ratios rather than ratios of totals. 17. Here and subsequently, all statements about statistical "significance' should not be taken literally. Besides the usual issue ofdata mining clouding their interpretations, the "samples" analyzed come close to covering completely the relevant population. Tests of significance are used here as a metric for discussing the relative fit of different versions of the model. In each case, the actual magnitude of the estimated coefficients is of more interest than their precise "statistical significance." 18. E. Mansfield

RALPH J. CICERONE
The ozone layer in the upper atmosphere is a natural feature of the earth's environment. It performs several important functions, including shielding the earth from damaging solar ultraviolet radiation. Far from being static, ozone concentrations rise and fall under the forces of photochemical production, catalytic chemical destruction, and fluid dynamical transport. Human activities are projected to deplete substantially stratospheric ozone through anthropogenic increases in the global concentrations of key atmospheric chemicals. Human-induced perturbations may be occurring already.
T tHE OZONE (03) LAYER IS AN IMPORTANT COMPONENT OF the stratosphere, that part of the earth's atmosphere between altitudes of 10 and 50 km where temperature increases with altitude. Ozone serves as a shield against biologically harmful solar ultraviolet (UV) radiation, initiates key stratospheric chemical reactions, and transforms solar radiation into the mechanical energy of atmospheric winds and heat. Also, downward intrusions of stratospheric air supply the troposphere with the 03 necessary to initiate photochemical processes in the lower atmosphere, and the flux of photochemically active UV photons [wavelength (X) < 315 nm] into the troposphere is limited by the amount of stratospheric 03.
This absorption of solar energy is the cause of the stratospheric vertical temperature gradient. Finally, because 03 molecules absorb radiation at UV, visible, and infrared wavelengths, atmospheric 03 affects the earth's energy budget and temperature.
Solar UV radiation of wavelengths less than 240 nm is absorbed by atmospheric 02 and 03, but for wavelengths between 240 and 320 nm only 03 is effective. Wavelengths less than 320 nm span the photoabsorption spectrum of DNA and can produce deleterious 3 JULY I987 biological effects, including skin cancer (1). Reduced amounts of atmospheric 03 permit disproportionately large amounts of UV radiation to penetrate through the atmosphere. For example, with overhead sun and typical 03 amounts, a 10% decrease in 03 results in a 20% increase in UV penetration at 305 nm, a 250% increase at 290 nm, and a 500% increase at 287 nm (2). Because of the apparent susceptibility of biota to UV radiation, the temporal evolution of paleoatmospheric 02, 03, and photosynthesizing plants was probably intimately linked (3).
Photolysis of 03 initiates much of stratospheric chemistry and includes processes, given by reactions la and lb, that control 03 amounts. 03 + hv -O('D) + 02('A) -O + 02 (la) (lb) The high-energy branch (reaction la; k < 315 nm; h, Planck's constant; v, frequency) produces electronically excited oxygen atoms, O('D), that in turn initiate the free-radical chemistry of the stratosphere (4) (2) (3) (4) (5) The absorption ofsolar UV and visible radiation by 03 represents an important source ofheat for the stratosphere. Absorption and reemission of outgoing planetary and atmospheric infrared radiation by O3 cools most regions of the stratosphere but heats the tropical lower stratosphere. The general circulation patterns of the strato- The author is at the National Center for Atmospheric Research, Boulder, CO 80307. sphere can be explained fairly well by diagnostic models that include only the interactions of 03, 02, and NO2 with UV and visible radiation and the interactions of CG2, 03, and H20 with infrared radiation (5).
Although reaction la is slower in the troposphere than in the stratosphere because less UV light is available, it proceeds and is followed by reaction 2, and thus tropospheric OH radicals are produced. The earth's atmosphere is oxidizing toward hydrocarbons, chlorocarbons, and various compounds of sulfur and nitrogen. This oxidation is initiated by gaseous species such as 03, OH, and HO2, not by 2. The movement of stratospheric 03 downward into the troposphere can initiate lower atmospheric photochemistry and even produce more 03 (6).
The measured spatial and temporal distributions of stratospheric 03 display patterns that are consistent with known photochemical production and consumption processes and the general circulation of the atmosphere, at least on the large scale (such as 150 belts of latitude), averaged seasonally. Figure 1 (7) displays vertically integrated 03 amounts as a function of latitude and season. Vertical integrals (or total column amounts) can be measured by groundbased UV absorption techniques and are important because of the relation between 03 and the flux of UV photons into the lower atmosphere.
Generally, 03 column amounts increase with latitude, especially in winter and spring, although 03 production rates are highest over the equator. Seasonal variations show the largest amplitudes at high latitudes. Ozone concentrations vary with altitude above the earth's surface; peak fractions of about 10-5 by volume are found between 25 and 35 km. The vertical column of 03 is distributed roughly as follows: 0 to 10 km, 10%; 10 to 35 km, 80%; and above 35 km, 10%. These patterns and percentages are nominal; in reality the 03 layer exhibits variability under the dynamic forces of chemical production and loss and fluid motions.

Production and Destruction of Ozone
To produce atmospheric 03 it is necessary to break the 0-0 bond in 02 Once released, oxygen atoms rapidly combine with 02 to form 03 by the reaction 0 + 02 + M -03 + M (6) where M is any third-body molecule, such as N2 or 02-The bonddissociation energy for 02 iS 118 kcal moleand corresponds to a threshold wavelength of 242 nm for photodissociation.
exceeds 2 x 1010 kW, or is at least three times the worldwide human energy usage rate (9). Additional O3 is produced by radiation at wavelengths below 175 nm above the stratosphere and by chemical means in the lower stratosphere and the troposphere. Photochemical production of 03 proceeds through reactions such as reaction 2 that are followed by reactions such as  (11) These are then followed by reaction 6, where carbon monoxide (CO) is present because of natural and industrial sources and nitric oxide (NO) is present largely because of reaction 4 in the stratosphere.
How does nature counterbalance the production of 03? The present view is that most 03 is destroyed catalytically because of the gaseous oxides of nitrogen and hydrogen. There is an important direct process, however, identified more than 50 years ago by S. Chapman (10), that involves only oxygen allotropes and can be represented by reaction lb followed by reaction 12. O(1D) + 03--> 202 (15) reaction 12 is an important O3 loss term from altitudes of 20 km through the upper reaches of the stratosphere and into the upper atmosphere. The bulk rates of reactions 12, 14, and 15 are essentially proportional to the square of the 03 concentration. In this way the effectiveness of these reactions in counterbalancing 03 production increases with O3 concentration. These rates are limited by the availability of oxygen atoms produced by reaction 1. The requirement for oxygen atoms also characterizes most of the gasphase processes that destroy 03 catalytically through oxides of other elements.
Catalytic processes that destroy stratospheric 03 are now thought to be very potent. Natural catalysts counterbalance 03 production, and anthropogenic catalysts can be sufficiently active as to cause a net decrease in the amount of stratospheric 03. Their inclusion into the scientific study of 03 about 15 years ago has led to a much more complicated photochemical scheme than was summarized above (10) and to the realization that biological processes at the earth's surface influence the state of the global atmosphere.
Nitrogen oxides are a central case. Once in the stratosphere, NO consumes 03 through the following gas-phase catalytic cycle NO+ 03-> N02 + 02 (16) 03+hv-O+ 02 (lb) Net reaction: 203 + bv --302 (18) Reactions 16, 17, and lb constitute a catalytic cycle because the NO that is consumed in the first reaction is replaced in the final reaction. The chain length ofthis cycle-that is, the number oftimes the cycle is repeated before an intervening process somehow sequesters the NO or N02-is determined by many other chemical reactions, some of which are shown below. For example, HNO3 can be formed and transported downward to the troposphere (Fig. 2). This NO, cycle represents a major sink for O3 that extends vertically throughout the stratosphere [see Johnston in (10) (lb) (25) (26) (18) Because of the rapidity of reaction 6, the loss of oxygen atoms in reactions 24 and 25 is equivalent to the loss of 03 molecules. These three HO, cycles are most important in the upper stratosphere, although reactions 21 and 22 can also be significant in the lower stratosphere and the troposphere (10).
Gaseous reactions that involve chlorine also destroy O3. The importance of these reactions appears to be relatively small in the natural, unperturbed stratosphere but can be very large in a humanperturbed stratosphere. Key reactions are the following.
Most of the elementary chemical reactions listed here have been well characterized in laboratory studies; reaction rate constants and temperature dependencies have been studied, often by several independent methods (11). Thus it is clear that catalytic destruction of 03 by chlorine, NON, and HO. does occur. The effectiveness of this catalysis is a topic of current research.
The efficiency of each of these 03-destruction processes depends on the fraction of time that the active chemicals exist as NO, NO2, H, OH, HO2, Cl, CGO, Br, and BrO compared to the time they exist in the form of inactive, reservoir species such as HNO3, N205, HNO4, H202, CINO3, HC1, and BrNO3. Reservoir species such as ClNO3 and HNO3 are particularly interesting because through their formation reactions (lb) each sequesters 03-active species from two chemical families. The flow of active chlorine and nitrogen species through reservoir (30) species, and through the atmosphere, is partially sketched in Fig. 2. (13) All the 03-destruction mechanisms outlined above derive from homogeneous gas-phase reactions. Recently, heterogeneous processes have been invoked because of the apparent inability of purely (29) gaseous reactions to account for the observed rapid decreases in 03 concentrations over Antarctica each September (13,14). It has been (27) suggested that heterogeneous processes liberate 03-active species (31) from reservoir species (14). These include (23) could be important, especially in a bromine-perturbed stratosphere. The latter catalytic cycle is especially interesting because its rate is not limited by the ambient concentration of oxygen atoms, as it is for the key NO, cycle (reactions 16 and 17), two of the HO, cycles, and the main chlorine cycle (reactions 27 and 28). Thus reactions 29,27, and 31 make up a cycle that could proceed at night (since no photons are required) and even in the lower stratosphere, where concentrations ofoxygen atoms decrease with altitude. The chemical kinetics of bromine reactions, however, are less well studied than for chlorine (for example, reaction 31 may produce other products), H20 + ClONO2 =0 HNO3 + HOC1 HCI + CIONO2 sa HNO3 + C12 (34) (35) and incorporation of gaseous HNO3 and HCI into polar stratospheric cloud particles by condensation or ion-catalyzed condensation (or both). Although these processes are not well defined experimentally or theoretically, they could increase the efficiencies of chlorine and bromine attack on 03 by regenerating more active species from reservoir species. Such possibilities are attractive because there are more cloud particles in the lower stratosphere over Antarctica than anywhere else. It is extremely important to deter-

Sources of Stratospheric Chemicals
The scientific effort to understand stratospheric 03 and humaninduced perturbations is proceeding across disciplines and in several directions. One concern is the flow of chemicals into the stratosphere from natural and anthropogenic sources. Key species are gases that are relatively inert in the troposphere, such as CH4, N20, CH3Cl, synthetic chlorofluorocarbons (CFCs) and chlorocarbons, and certain organobromine compounds (Table 1). Figure 2 displays examples of large-scale processes that produce and transfer source gases, which undergo irreversible photo-oxidation to yield important gaseous radicals, to the stratosphere. Thus N20 from soil and oceanic microbes enters the lower atmosphere and, through large-scale motions (principally in the tropics), is transported upward to the stratosphere. Most N20 is decomposed through N20 + hv-* N2 + O('D) (36) but about 5% produces NO through N20 + O('D) --2NO (37) as indicated in Fig. 2. Similarly, the synthetic chlorofluorocarbons CC12F2 and CC13F are swept upward into the middle stratosphere, where UV photolysis dissociates them to yield chlorine atoms. As with N20, there are no known tropospheric sinks for CC12F2 and CCl3F, so that nearly 100% of the molecules released at the earth's surface reach the stratosphere. Atmospheric tracer data and spatial pattems of N2O, CFCs, and other gases indicate that, on average, it takes about 5 years for a gas emitted at the earth's surface from midlatitudes of the Northern Hemisphere to travel upward to photochemically active altitudes (25 km and higher) in the stratosphere.
Atmospheric residence times on the order of 100 years characterize N20, CCl2F2, and CC13F, because air at altitudes of 25 km and higher must be exchanged vertically many times to deplete the massive tropospheric reservoir of these species.
The activity of another important gas, methane (CH4), is not depicted in Fig. 2. Methane is not as inert in the atmosphere as nitrous oxide (N20) or CFCs. Rather, it is oxidized through the reaction Fig. 4 Fig. 2 show some of the important reactions that control stratospheric 03 concentrations, including some of the reactions 16 through 37. Chain reactions that are carried by radicals destroy O3 and perform other transformations; these chains are terminated by radical-radical reactions that produce relatively stable species such as HNO3 and HC1. Although NO2 and chlorine atoms can be regenerated from HNO3 and HC1, the time constant Tc for each of these reactions is long enough for downward transport into the troposphere to occur. When polar, soluble species such as HNO3 and HCI come into contact with condensed water and surfaces in the lower atmosphere, they can be deposited at the earth's surface, mostly as H+, NO3, and Cl-. Thus at steady state the amount of nitrogen from NO3ions deposited annually would equal the amount of nitrogen released as NO in the stratosphere from N20 if there were no other sources of NO3ions (such as pollution from combustion).
It is important to know the release rates of N20, CH4, and CFCs because of the control over 03 exerted by the decomposition products of these source gases. Also, N20, CH4, and several CFCs are potent climatic "greenhouse" gases (as is 03) that trap outgoing planetary heat in the earth-atmosphere system, and their atmospheric concentrations are increasing (15). For N20 we have only semiquantitative information. Aerobic microbial nitrification and anaerobic microbial denitrification produce N20 both in soils and in water bodies (16). Combustion of fuels, especially nitrogen-rich fuels, also produces N20. Temporal trends and spatial distributions of N20 are well characterized (16, 17), but the relative roles of processes that cause its increase (microbiological sources that act on excess amounts of nitrogen fertilizer compared to combustion processes) are not clear (16,17).
Microbiological processes are also central to the question of atmospheric CH4 sources. Methanogenesis (an anaerobic process) in ruminant animals, rice-paddy soils, swamps and marshes, and perhaps termite guts furnishes the bulk of CH4; and CH4 is also released from natural gas exploration and transmission and coal mining (16). Microbial CH4 oxidation also limits the flux of CH4 from some soils and the oceans into the atmosphere.
Although our knowledge of microbial production rates of N20 and CH4 is inadequate, and although we know very little of the variations in these rates and their underlying mechanisms, we do know the global totals to which the various sources must add. For a gas X with no atmospheric sources and a steady-state global distribution, the total annual source (TAS) is given by where r is the spatial coordinate, L is the local atmospheric photochemical loss rate for gas X, and the integration is over the entire atmosphere. For N20, solar UV radiation (X < 230 nm) and attack by stratospheric O('D) constitute L, and the total annual source (sink) of N20 is calculated to be 11 (± 2) x 1012 g of nitrogen as N20 (16,18). This steady-state argument is improper, because N20 concentrations are increasing (17) and 3 x 1012 g of nitrogen as N20 is needed annually in addition to the steady-state source to account for the increase. Thus the present sources of atmospheric N20 add to about 14 x 1012 g of nitrogen as N20.
Equation 39 yields more uncertain results when applied to CH4 because the atmospheric loss term L for CH4 is dominated by its tropospheric OH radicals (reaction 38). There are virtually no observational data for tropospheric OH, but if OH concentrations from theoretical models are used with Eq. 39 and additional lines of reasoning, the total annual steady-state CH4 source is 450 (±150) x 1012 g (16). In addition, the atmospheric burden of CH4 has increased about 1% annually in recent years, which implies that a 50 X 1012 to 60 x 1012 g CH4 source increase or sink decrease (or both) operates annually. There are indications from '4CH4 data that 80% or more of the total CH4 source is recent biological activity (15). For CFCs, especially CCI2F2 and CC13F, the situation is simpler. From industrial production statistics and tariff data we know the amounts of these chemicals that have entered the atmosphere since their use became widespread in the 1960s. In recent years, the annual industrial production of CCI2F2 and CC13F has been 450 x 103 and 300 x 103 metric tons, respectively. After corrections for the amounts already destroyed in the stratosphere, for CFC storage times, and for amounts dissolved in oceans, the measured atmospheric concentrations agree closely with those expected from production data (16). In contrast, only semiquantitative information is available for sources of C2C13F3, CC14, CH3CI, and organobromine gases.

Perturbations to Stratospheric Ozone
Stratospheric O3 could be decreased by any process that can lead to increased stratospheric amounts of 03-destroying catalysts (for example, oxides ofnitrogen, chlorine, hydrogen, or bromine). Many possible stimuli have been proposed, principally NO, from nuclear explosions (19), hypothetical fleets of supersonic aircraft (20), solar proton events (21), and increased atmospheric N20 (22) and chlorine from continued use of CFCs (23), volcanoes, and space shutfle rocket exhaust (24). Also, increases in atmospheric CH4 can lead to 03 perturbations through interactions with NO, and ClO, cycles and through production of HO,. Further, increases in carbon dioxide can lead to increases in O3, not through direct chemical reactions but through enhanced radiative cooling of the middle and upper stratosphere. Such a cooling would hasten reaction 6 and decrease the concentration ratio of oxygen atoms to O3 molecules. While it can be said that all these stimuli are now being applied to the atmosphere, the most definitive experiment to date concems solar proton events. Observations that followed the large event of August 1972 showed that O3 concentrations were reduced by about as much as theory predicted, at least in the upper stratosphere (25).
Attempts to predict the future effects of continued increases in stratospheric source gases (CFCs, N20, CH4) have given nse to various mathematical models. Early models were one-dimensional (altitude), but in the last several years two-dimensional (altitude and latitude) models have become available. Some of these two-dimensional models include as many chemical reactions as the onedimensional models and also treat atmospheric motions, latitude variations, and seasonal changes much more realistically. One hundred or more gas-phase reactions are included in modem models. Figure 3 shows the results of calculations of future O3 perturbations caused by continued CFC releases and the changes due to simultaneous CFC releases and increases in atmospheric N20 and CH4. These O3 changes were calculated with a two-dimensional diabatic circulation model (26) that simulates known tracer fields very well but does not account for thermal and fluid dynamical 40 changes due tO 03 changes. In these calculations, steady annual releases of 392 x 103 and 265 x 103 tons were simulated for CC12F2 and CC13F, respectively, and the model was run to steady state. Similarly, N20 and CH4 concentrations at the lower boundary (zero altitude) were fixed at 1.4 x 305 parts per billion (ppb) and 2.0 x 1.6 parts per million, respectively. The steady-state results in Fig. 3 represent the atmosphere in approximately the year 2050.
Simulated CFC releases lead to O3 column decreases at all latitudes in Fig. 3; local O3 concentrations decrease greatly above altitudes of 35 km at all latitudes. Increases in local O3 appear below 25 km in the tropics and are largely due to increased penetration of radiation at wavelengths less than 240 nm that produces new O3. Larger 03 column decreases are calculated for high latitudes (>40') than for low latitudes. When the sustained CFC releases are accompanied by increases in N20 and CH4, similar spatial pattems are calculated but with sharper gradients. Calculated O3 decreases are smaller at high altitudes, and O3 increases are greater at low altitudes. The former is due to decreased attack of chlorine on O3 because of the increased rates of reaction 32 and of the reaction CH4 + Cl --HCl + CH3 (40) both of which sequester chlorine atoms. At low altitudes there is increased troposphere-like 03 production due to increased CH4 and NO. in the model, although other mechanisms also operate (26).
The reference state for the perturbations of Fig. 3 was a stratosphere that included 2.4 ppb of inorganic chlorine, CIX; although this CIX value may characterize the present stratosphere, the earlier unperturbed stratosphere probably had less than 1 ppb of CIX.
The transient temporal approach to future O3 perturbations is an important matter both for testing the scientific theory of atmospheric behavior and for more practical reasons, such as limiting future increases in UV radiation or evaluating climate change. Experimental and analytical efforts are under way to detect temporal changes in stratospheric O3 and related species (27) and to compare the results to the time-dependent analogs of Fig. 3. Trend detection is a major goal of these efforts; existing models require additional verification, and early warnings of larger future 03 changes are needed. No clear trend in the total column of stratospheric O3 has been reported to date except for that over Antarctica (27). Some data show a decrease of 03 concentrations in the upper stratosphere since 1970 (27); it is in the upper stratosphere where the largest and earliest 03 decreases are expected.
Large O3 decreases over Antarctica have been reported (13); this perturbation was not predicted, nor is there an accepted explanation at present. Figure 4 shows the measured monthly means of total O3 measured each October over Halley Bay, Antarctica (76°S), from 1957 through 1984, after Farman and colleagues (13) ofthe British Antarctic Survey (BAS). A decrease of more than 35% has occurred since about 1970. Stolarski and co-workers (13) examined NASA satellite data on total 03 and found that the phenomenon is regional in extent, not just local. The decrease occurs during September as the sun rises, and O3 concentration reaches a minimum in mid-October. Eight years (1979)(1980)(1981)(1982)(1983)(1984)(1985)(1986)) of NASA's October minimum monthly means are plotted as crosses in Fig. 4; these are the lowest monthly means observed in any 2°latitude belt south of 60°S. The NASA analysis also shows a 30% decrease in the O3 minimum and a 20% decrease in the surrounding O3 maximum. It seems clear that, for O3 column decreases of tens of percent to be achieved, 03 must be decreased greatly in the lower and middle stratosphere, not just above. Stolarski and co-workers also found that the deep O3 minimum follows the polar vortex, in which temperatures of the lower stratosphere are very low in winter and early spring (often below 190 K) and stratospheric clouds are much more prevalent than over the Arctic (28). Table 1. Data that characterize four key stratospheric trace gases that enter the atmosphere at the earth's surface. Source designation: N, natural; A, anthropogenic; HC, under human control. For CH4 only 10 to 15% of the total source enters the stratosphere, and the concentration decreases with altitude in the stratosphere. Concentrations are in parts per million (ppm), parts per billion (ppb), and parts per trillion (ppt) by volume. TAS, total annual source (see reaction 39 in text).

Concen-Trend
Residence Theories that attempt to explain the Antarctic 03 "hole," its timing, and its secular change are becoming numerous (13,14). It is important to understand the chemical and physical mechanisms of this perturbation to be able to predict whether it will spread to lower latitudes and worsen, or dissipate and shrink. Atmospheric fluid motions, radiative energy exchange, and atmospheric chemistry are likely to be strongly linked in a successful explanation. From a mostly fluid-dynamical point of view, Mahlman and Fels (29) proposed a mechanism that could trigger many features of the Antarctic 03 decreases. If a natural climatic process could substantially reduce the wintertime planetary-scale disturbance activity in the Southern Hemisphere troposphere, then several southem stratospheric responses would be expected (29): a reduction ofwintertime polar 03, a prolonging of the time span of the polar vortex, reduction of the transport of 03 out of the middle stratosphere, and an increased possibility of polar rising motion shortly after the return of the sun to high latitudes. All these effects are in qualitative agreement with the observed 03 changes, although the reasons for the hypothesized triggering events are not clear and the overall mechanism does not preclude independent chemical mechanisms. If natural dynamical mechanisms can cause secular 03 decreases such as those observed over Antarctica, then our ability to detect other human-induced 03 changes would be seriously challenged because more natural variability would have to be incorporated into our trend-detection schemes.
From the viewpoint of chemical perturbation, the Antarctic phenomenon is stimulating proposals of more complex chemical reactions and reactants than had been included in previous models. As noted above, condensed-phase and heterogeneous processes are being invoked to account for the release of more active chlorine, for example from less active chlorine reservoir species (30). If relatively high ratios of CIO to total inorganic chlorine concentrations do characterize the September and October Antarctic stratosphere, then new catalytic cycles could become significant. The following reaction sequence has been suggested (30,31 (33); surface reactions involving ClO and 03 or CGNO3 and 03 are possible (33); even isomerization of CIOO with 03 is possible. Similar ground-based observations (34) have also demonstrated that the Antarctic stratosphere contains unusually small amounts ofNO2, which implies that solar-proton-produced nitrogen oxides act minimally (if at all) in the destruction of 03 during austral spring. As global atmospheric change becomes a matter offact rather than conjecture, it becomes even more important to advance the scientific study of stratospheric 03 perturbations and our ability to predict them. A growing arsenal of experimental techniques is developing, and new instruments are being applied to characterize the atmosphere and its variations and to laboratory experiments in photochemistry and kinetics (27). It is to be hoped that the speed of acquisition of quantitative understanding and predictive skill exceeds the rate at which atmospheric change occurs. 35. B. G. Gardiner and J. D. Shanklin, Gopys. Ret. Lett. 13, 1199Lett. 13, (1986 The functional or ion of the cerebral cortex is modified dramatcally by sensory experience during early postnatal life. The basis for these modifications is a type of synaptic plasticty that may also contribute to some forms of adult lear ng.Te question of how synapses modify accor to experience has been approached by determining theoretically what is required of a modification mechanism to account for the available experimental data in the developing visual cortex. he resulting theory states precisely how certain variables mght influence synaptic modifications. This insight has led to the development of a biologically plausible molecular modd for synapse moiica in the cerebral cortex. N LTHOUGH ARISTOTLE IDENTIFIED HEART AS THE SEAT OF intellect, reserving for brain the finction of cooling the ead, it is now generally believed that it is brain that is the source of thought, the location of memory, the physical basis of mind, consciousness, and self-awareness: all that make us distinct and human. In recent years it has become increasingly fashionable to treat his complex system as a neural network: an assembly of neurons connected to one another by synaptic junctions that serve to transmit information and possibly to store memory.
Since the contents of memory must depend to some extent on experience, the neural network and, in particular, the synapses between neurons cannot be completely determined genetically. This evident reasoning has led to much discussion about possible modification of synapses between neurons as the physiological basis of learing and memory storage. To properly function, neural network models require that vast arrays of synapses have the proper strengths. A basic problem becomes how these synapses adjust their 42 weights so that the resulting neural network shows the desired properties of memory storage and cognitive behavior.
The problem can be divided into two parts. First, what type of modification is required so that in the course ofactual experience the neural network arrives at the desired state? The answer to this question can be illuminated by mathematical analysis of the evolution of neural networks by means of various learning hypotheses. The second part of this problem is to find experimental justification for any proposed modification algorithm. A question of extraordinary interest is: What are the biological mechanisms that underlie the nervous system modification that results in learning, memory storage, and eventually cognitive behavior? One experimental model that appears to be well suited for the purpose of deteriining how neural networks modify is the visual cort of the cat. The modification ofvisual cortical organization by sensory experience is recognized to be an important component of early postnatal development (1). Although much modifiability disappears after the first few months of life, some of the underlying mechanisms are likely to be conserved in adulthood to provide a basis for learning and mcmory. We have approached the problem of ecperience-dependent synaptic modification by deteminiing theoretically what is required of a mechanism in order to account for the cxperimental observations in visual cortex. This process has led to the formulation of hypotheses, many of which are testable with currently available techniques. In this article we illustrate how the interaction between theory and experiment has suggested a possible M