Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
2 Principal Scientific Issues in Modeling Studies GLOBAL CLIMATE SENSITIVITYâSIMPLIFIED MODELS AND EMPIRICAL APPROACHES The sensitivity of climate to changes in CO2 concentrations has been examined through two apparently different approaches: (1) by climate models that account for the energy-balance components of the complete surface-atmo- sphere system and (2) from empirical observations keyed to the surface energy balance alone. Simplified climate models, ranging from zero-dimensional empirical re- lationships to highly parameterized multidimensional models, can be useful for inexpensive studies of climate change and climate mechanisms over a wide range of time scales, if used with appropriate caution. The basic strength of these models is that they permit economically feasible analyses over a wide range of parameter space. Such studies are valuable for focusing and interpreting studies performed with more complex and realistic global models. However, it must be realized that the simplified models are limited in the information they can provide on local or regional climate change. Such detailed inferences can be obtained from three-dimensional general-circulation models (3-D GCM'S) of the type discussed in the Charney report and in the section Model Studies and from empirical studies, although both approaches are severely limited in current capabilities. One-Dimensional Models The one-dimensional (i.e., treating a vertical column through the atmosphere) radiative-convective (1-D RC) models provide a good illustration of the use 15
16 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT TABLE 2.1 Equilibrium Surface Temperature Increase Due to Doubled CO, (300 ppm -> 600 ppm) in l-D RC Models ".* Model Description- AT". (Â°C) / F (W irr2) 1 FAR, 6.5LR, FCA 1.2 1 4.0 2 FRH, 6.5LR, FCA 1.9 1.6 3.9 3 Same as 2, except MALR replaces 6. SLR 1.4 0.7 4.0 4 Same as 2, except FCT replaces FCA 2.8 1.4 3.9 5 Same as 2, except SAF included'' 2.5-2.8 1.3-1.4 6 Same as 2, except VAF included' ~ 3.5 ~ 1.8 "Data from Hansen et al., 1981. 'Model 1 has no feedbacks affecting the atmosphere's radiative properties. The feedback factor / specifies the impact of each added process on model sensitivity to doubled CCK. F is the equilibrium thermal flux into the planetary surface if the ocean temperature is held fixed (infinite heat capacity) when CO2 is doubled; this is the flux after the atmosphere has adjusted to the radiative perturbation within the model constraints indicated but before the surface temperature has increased. TRH, fixed relative humidity: FAH, fixed absolute humidity; 6.SLR, 6.5Â°C km"1 limiting lapse rate; MALR, moist adiabatic limiting lapse rate; FCA, fixed cloud altitude; FCT, fixed cloud temperature; SAF, snow-ice albedo feedback; VAF, vegetation albedo feedback. 'Based on Wang and Stone, 1980. 'Based on Cess, 1978. of simplified models. For example, the basic greenhouse mechanism by which atmospheric CO2 warms the Earth can be analyzed with the help of such models. Indeed, the sensitivity of the surface temperature in l-D RC models to changes in CO2 amount is in general agreement with the sensitivity of more realistic 3-D GCM'S, suggesting that l-D RC models are able to simulate certain basic mechanisms and feedbacks in 3-D GCM'S. It is thus of value to use a l-D RC model to illustrate processes that influence climate- model equilibrium sensitivity and also to compare the results of these models to observed climate variations. The main processes known to influence climate-model sensitivity are summarized in Table 2.1 (Hansen et al., 1981). Note that the change in equilibrium global-mean temperature, "equilibrium sensitivity," deduced from l-D RC models is of the order of 1Â°C for doubled CO2, even in the absence of feedback effects. The increase in atmospheric water vapor that
Principal Scientific Issues in Modeling Studies 17 occurs with rising temperature increases the sensitivity to almost 2Â°C for doubled CO2. The atmospheric lapse rate (vertical temperature gradient) may also change in response to increasing atmospheric temperature and water vapor. At low latitudes, where the lapse rate is nearly moist adiabatic, this represents a negative feedback, while more stable lapse rates at high latitudes lead to a positive feedback on surface temperature. A positive cloud feedback occurs if clouds move to higher altitude with increasing temperature, but changes in cloud cover may lead to either a positive or a negative feedback. Improved empirical data on clouds and better modeling capabilities are needed to evaluate cloud effects. Ice-snow albedo feedback is clearly positive. Preliminary estimates of vegetation albedo feedback are also positive but very uncertain. The sensitivity of the global-mean temperature in simplified climate models to increased atmospheric CO2 is consistent with that found with 3-D GCM'S, i.e., 3 Â± 1.5Â°C (see the section Model Studies). Surface Energy Balance Considerations The surface energy balance approach has been adopted in a number of studies of climate sensitivity applied to the CO2 problem (Callendar, 1938; Plass, 1956a, 1956b; Kaplan, 1960; Moller, 1963; Newell and Dopplick, 1979; Idso, 1980a; Kandel, 1981). Manabe and Wetherald (1967) discussed the difficulties in attempting to estimate global surface temperature response to increased CO2 from consideration of surface energy balance. They demon- strated that the complete surface-atmosphere energy balance must be consid- ered to obtain valid results. Two recent studies (Newell and Dopplick, 1979; Idso, 1980a) yielded a CO2-induced surface warming substantially smaller than the warming obtained from the studies of the complete surface-atmosphere system. On closer examination, it is found that there is no basic disagreement between the empirical evidence used in these studies and the climate-model results. Indeed, the empirical response data provide some valuable tests of the models. Below we provide a summary discussion of how increased CO2 modifies the radiation and energy balance of the complete surface-atmosphere system. In the following section we briefly discuss some of the dissenting inferences of surface-temperature sensitivity. More detailed discussion of aspects of the surface energy balance are given by Ramanathan (1981). Models of the Earth's Complete Energy Balance Numerical climate models treat the energy balance of the complete surface- atmosphere system. For the purposes of the present discussion, it is sufficient
18 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT to focus on but two components of the systemâthe troposphere and the surfaceâbecause the stratosphere's role in the energy balance is relatively small. For example, in the case of doubled CO2, inclusion of the stratosphere and thus the downward infrared flux from the stratosphere into the troposphere would increase the calculated total radiative energy flux into the troposphere by only about 30 percent (Schneider, 1975; Ramanathan et al., 1979). This stratospheric effect is included in most climate-model studies and in the estimates of CO2 heating given later in this report. On time scales longer than several weeks, the troposphere is closely coupled to the surface, particularly on a global-mean basis. The principal processes responsible for this coupling are the exchange of latent, sensible, and radiative heat fluxes between the air and the surface through the boundary layer and the horizontal and vertical convective processes within the troposphere. Because of vertical mixing, in particular, the response of the global-mean temperature of the troposphere and surface depends primarily on the heating perturbation to the surface-troposphere system as a whole and is relatively insensitive to the vertical distribution of the perturbation heating within the system. This conclusion has been reinforced in a series of GCM experiments (Manabe and Wetherald, 1975; Wetherald and Manabe, 1975) in which comparable surface- tropospheric heating perturbations due to increases in both CO2 and solar constant produced similar surface-temperature responses, even though the vertical distribution of heating was significantly different. For a doubling of CO2, radiative calculations show that the surface- troposphere system is subjected to a net radiative heating of about 4 W m~2 before any adjustments in temperature or other climatic variables are allowed to occur. Roughly 2.5 W m~2 of this heating is caused by a reduction in the outgoing infrared radiation from the surface-troposphere system, and most of the remaining heating is due to an increase in the downward infrared emission by the stratosphere (Schneider, 1975; Ramanathan et al., 1979; Hansen et al., 1981). As summarized in the Charney report and in the section One-Dimensional Models, most climate models translate this radiative heating into a warming of the surface equilibrium temperature by about 1 K. if the amount of water vapor in the atmosphere is held fixed. This value increases to about 2 K as a result of the increased water vapor abundances expected to accompany increasing atmospheric temperatures, a positive feedback included in most climate models. Dissenting Inferences from Energy-Balance Models and Empirical Studies Inferences of a relatively small sensitivity of climate to changes in CO2 have been made recently (Newell and Dopplick, 1979; Idso, 1980a, 1980b, 1981).
Principal Scientific Issues in Modeling Studies 19 These papers are based on incomplete methods or observations, and their conclusions appear to be of limited utility in assessing the climatic effects of increased CO2. As indicated above, the sensitivity of climate to a perturbation in some radiative process can best be assessed by considering the entire global Earth- atmosphere system. Approaches centered on the surface, limited regions, or time-limited observations necessarily become more complex because many fluxes and reservoirs of energy must be explicitly and quantitatively taken into account. In determining the effects of increased atmospheric CO2 on the surface, a number of effects are produced by a group of interconnected processes: 1. Additional heating of the surface due to additional radiation from the atmosphere; 2. Reduced cooling of the troposphere due to reduction of radiative heat emission from the troposphere to space; 3. Heating of the surface by additional radiation from a warmer troposphere; 4. Increases in sensible and latent heat fluxes from the surface to the atmosphere, thus moderating the surface-temperature increase, heating the atmosphere, and increasing atmospheric moisture; 5. Increased heating of the troposphere by increased absorption of solar radiation by increased water vapor (and CO2); 6. Increased heating of the surface due to enhanced radiation from the additional moisture in the atmosphere; 7. Exchanges of sensible and latent heat with regions not explicitly treated; and (in nonequilibrium cases) 8. Thermal inertia of the land surface, the atmosphere, and especially the oceans. All of these processes must be taken into account in calculating new equilibrium conditions for the surface and atmosphere in which increased net radiative heating of the surface is balanced by increased net fluxes of sensible and latent heat from the surface and to other regions. It is within this context that we consider the cited studies. Idso compares several sets of empirical observations of changes in downward radiative flux at the surface with corresponding changes in surface temperature. From each case, he computes an empirical "response function" relating a change in downward radiative flux to a change in surface temperature. He finds the values of this parameter to be virtually constant among the cases he considers. He then estimates from a radiation model the change in downward flux that would result from doubling atmospheric CO2 concentration and employs the empirical response function to calculate a
20 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT rather small surface-temperature change. This approach is misleading when applied to estimation of the response of global-mean equilibrium climate to increased CO2: 1. It is clear that the response function can be calculated quite arbitrarily at virtually any value in nonequilibrium situations, according to the choice of data. For example, surface temperatures rise between noon and midafter- noon and between June and July (in the northern hemisphere), although in both cases, incident solar radiation decreases; a negative response function might be implied by one of Idso's methods. 2. Some of the "natural experiments" that Idso employs are on time and space scales clearly inappropriate to the CO2 problem and do not involve the components of the climate system that are important for long-term climate changes. The thermal inertia of the Earth's surface, particularly the 70 percent that is ocean, slows the response of surface temperature to changes in incident radiation. Thus, the observations of temperature changes from day to day or season to season employed by Idso do not reflect the full temperature change that would be experienced at equilibrium if the radiation change were to persist for extended periods. For this reason, the response functions calculated by Idso are necessarily too small for estimation of the equilibrium climatic response to increased CO2. 3. Idso's interpretation of empirical radiation measurements confuses primary forcing and the amplifying feedbacks engendered by that forcing. (a) For example, he compares changes in temperature between winter and summer with corresponding changes in solar radiation received at the surface. However, the increase of surface temperature between winter and summer not only results from the primary change in forcing (increased insolation) but also reflects the additional thermal radiation to the surface due to consequent increases in atmospheric temperature and humidity. Thus, in this case, the primary forcing is correctly identified, and some feedbacks are incompletely included in the response. (b) In another case, Idso (1981) compares the temperature of an airless Earth heated by solar radiation alone with that of today's Earth, which receives energy from both the sun and the atmosphere. Here, however, the true initial forcing would be the hypothetical effect of imposing on rather cool, airless Earth a correspondingly cool but radiatively active atmosphere. The radiation from our present warmer and wetter atmosphere is considerably greater than this conceptual initial forcing, reflecting powerful amplifying feedback processes, and thus represents a mixture of cause and effect. Failure to distinguish clearly and consistently between cause and effect permits erroneous and virtually arbitrary conclusions to be drawn. The empirical phenomena described by Idso are in fact perfectly consistent
Principal Scientific Issues in Modeling Studies 21 with the climate models employed for assessment of CO2 increases. For example, the seasonal cycles of radiation and temperature have been used with success by a number of modelers for empirical validation (e.g., Warren and Schneider, 1979; North and Coakley, 1979; Manabe and Stouffer, 1980; Hansen et al., 1981). Newell and Dopplick (1979) deal with the effect of doubled CO2 in the tropics, primarily the effect on the temperature of the tropical ocean. Holding atmospheric parameters fixed, they calculate an increase in energy received at the surface due to doubled CO2. Using formulas for transfer of sensible, latent, and radiative energy from the surface, they estimate a relatively small temperature increase for the tropical ocean. Somewhat larger increases are computed for tropical land areas. They estimate the change in the surface energy budget produced by the increased atmospheric moisture (resulting from the assumed enhanced evaporation) to be about as large as the initial change due to CO2, doubling the effect of the CO2 itself. Nevertheless, they project only a very small increase in low-latitude surface temperatures. As corroboration, they cite the small temperature changes observed after the 1963 Mt. Agung eruption, which produced an increase in stratospheric aerosols and a decrease in solar energy received at the surface. Newell and Dopplick (1981) have also called attention to paleoclimatic data indicating that tropical ocean temperatures have varied but little in the past, although CO, concentrations are known to have changed. As suggested above, the temperature of the Earth's surface is maintained by a balance between fluxes of sensible, latent, and radiant energy between the surface and the overlying atmosphere. The effect of a perturbation in one component of these fluxes can be evaluated correctly only by a complete, internally consistent, and energy-conserving treatment. The Newell-Dopplick arguments are faulty in this respect. To begin with, they treat the tropics in isolation, without considering the exchange of energy with higher latitudes. Moreover, 1. They fail to take into account satisfactorily the effects of atmospheric changes that would accompany CO2 increase. For example: (a) Newell and Dopplick employ formulas expressing fluxes of sensible and latent heat from the surface in terms of differences in temperature and vapor pressure between the surface and the near-surface air. They estimate the sensitivity of surface temperature to increased CO2 by calculating the increase of surface temperature needed to increase heat fluxes from the surface sufficiently to counterbalance the additional energy input from the atmosphere, while holding atmospheric temperature and absolute humidity fixed. In reality, surface temperature and near-surface air conditions are closely coupled by means of these same fluxes, so that a given change in
22 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT surface temperature gives rise to a much smaller change in the magnitude of the air-surface differences. Thus, their assumption overestimates the mag- nitude of changes in air-surface differences and consequently also overesti- mates the magnitude of surface-air fluxes associated with a change in surface temperature. For this reason, they erroneously conclude that only a small increase in surface temperature would be required to enhance upward fluxes from the surface and balance CO2-induced downward fluxes. In reality, the surface warming is accompanied by increases in the temperature and absolute humidity of the overlying air. Therefore, a much larger increase of surface temperature would be required to balance the surface energy budget. Thus, it is not difficult to appreciate why Newell and Dopplick indicate an extremely small sensitivity of surface temperature to an increase in atmospheric CO2 concentration. (b) The radiative effect of increased atmospheric moisture appears to be greatly underestimated, probably because the moisture and temperature of the air column are not allowed to come completely into equilibrium with the surface. 2. The Mt. Agung observations are, in fact, not inconsistent with the results of models used for estimating climatic sensitivity to increased CO2 (Pollacks a/., 1976; Mass and Schneider, 1977; Hansen etal., 1978, 1981). It would be expected that the heat capacity of the ocean would slow the response, causing the temperature change experienced during the brief residence period of the aerosol to be but a fraction of that which would be realized at equilibrium (Hansen et al., 1978; Hoffert et al., 1980). However, the limited observations of stratospheric aerosol optical depth and the lack of data on quantitative heating properties of aerosols suggest that the Mt. Agung observations cannot be taken as definitive indications of climatic sensitivity. Thus, the results of Newell and Dopplick on the tropical surface energy balance do not refute the inferences of a global climate sensitivity obtained from comprehensive models of the complete global climate system. Questions of a different character were raised by Lindzen et al. (1982), who employed a cumulus convective parameterization in a 1-D RC model. They obtained a sensitivity to CO2 increase about 35 percent smaller than that with their model using fixed-lapse-rate convective adjustment. This conclusion is in close agreement with the results of similar model experiments presented in Table 2.1; use of a moist adiabatic lapse rate in Model 2 in order to reflect cumulus convective processes produces a sensitivity that is smaller than that indicated by Model 2 with fixed lapse rate. Similarly, global 3-D GCM'S also show a relatively small sensitivity in the tropics, where cumulus convective processes are strong and moist adiabatic lapse rates
Principal Scientific Issues in Modeling Studies 23 prevail (Manabe and Stouffer, 1980). The response of global-mean temper- ature is, of course, influenced strongly by the relatively large temperature increases projected for higher latitudes, where cumulus convective processes are less important. Thus, a careful comparison of Lindzen et al. results with available 1-D (with moist-lapse-rate adjustment) and 3-D climate models suggests that the overall magnitude of the CO2 warming does not depend greatly on the details of the convective parameterization employed, although that component of the models is one of many that warrant more careful study. The preceding discussion clearly illustrates the complex nature of the surface energy budget and the dangers involved in inferring global climate sensitivity from local surface observations. However, empirical methods of inferring sensitivity are appealing. The real value of empirical studies is that they are necessary to verify the behavior of climate models. A promising empirical method has been proposed by Cess (1976), who obtains sensitivity estimates from the latitudinal gradient in annually and zonally averaged radiation budgets as obtained from satellite radiation measurements and obtains surface-temperature responses consistent with climate-model studies. To summarize, the sensitivity of climate to increased CO2 obtained from most global climate model studies is entirely consistent with the inferences drawn from surface energy balance studies and empirical approaches, provided that the latter methods account fully for the globally connected energy budget and transport processes within the entire surface-atmosphere system on the appropriate time scales. Some of the important implications of the global sensitivity analysis are summarized below: 1. Because of the strong coupling between the surface and the troposphere, the global-mean surface warming is driven by the CO2 radiative heating of the entire surface-troposphere system and not only by the direct CO2 radiative heating at the surface. 2. The magnitude of the surface-troposphere warming is determined by horizontal advective and vertical convective-radiative interactions between the atmosphere and surface (in particular, the oceans). 3. Some recent studies of the climatic effects of increased CO2 based on the surface energy balance approach have not fully accounted for the processes described in points 1 and 2 above. When these limitations are taken into account, their results are seen to be entirely consistent with those of comprehensive models. However, inconsistent interpretation of incomplete analyses can yield virtually arbitrary conclusions (cf. Moller, 1963). 4. Empirical approaches to estimating climate sensitivity from observations should be encouraged, since they may provide a valuable source for calibrating
24 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT or validating model results. However, such approaches should emphasize global-scale measurements and also should infer sensitivity from natural climate change "experiments" in which long-term ocean-atmosphere inter- actions are involved in the climate change. Satellite radiation budget measurements should be a valuable data base for such endeavors. ROLE OF THE OCEANS Role of the Ocean in the Transient Response of Climate The Charney report emphasized a significant but relatively unexplored ques- tion about the role of the oceans in the climate system. The heat capacity of the upper ocean is potentially great enough to delay for several decades the establishment of new equilibrium temperatures associated with increased atmospheric CO2, with consequent impact both on social implications and on verification strategies. The dominant effect of a sudden doubling of atmospheric CO2, in the absence of ocean warming, would be a net downward flux of heat at the ocean surface of about 4 W m ~2 and an almost imperceptible change in atmospheric temperature over most of the globe (Ramanathan, 1981). If this heat flux were maintained indefinitely, a well-mixed layer 50 m deep would warm to a new equilibrium temperature 2Â°C higher in 3 years, whereas one 500 m deep would take 30 years, and the involvement of the whole ocean, 5000 m deep on average, would take 300 years. (In reality, of course, the net flux is partially compensated by increased fluxes of sensible and latent heat from the ocean to the atmosphere, and the approach to equilibrium is still further delayed.) To what depth would mixing take place before a significant rise in sea-surface temperatures takes place? This question is fundamental and quite distinct from the ocean's role in moderating the rise in atmospheric CO2 concentration itself. Few relevant studies have been published. Manabe and Stouffer (1980) used a physically comprehensive GCM of the atmosphere but a simple oceanic mixed layer of about 70-m depth to simulate the annual cycle. However, in the absence of a deeper ocean, this model was insufficient to describe a longer-term approach toward a new equilibrium. Schneider and Thompson (1981) combined a simplified atmospheric model with a two-box model of the ocean and predicted that 25 percent of the increase in surface temperature following an instantaneous doubling of CO2 would be delayed by at least 20 years. They also suggest that the evolution of the latitudinal temperature contrast would be significantly affected. Although the mechanisms for vertical exchange in the oceans are only
Principal Scientific Issues in Modeling Studies 25 qualitatively understood and are difficult to model reliably, certain broad statements can be made with confidence, which indicate the general nature and probable magnitude of the oceans' role in moderating climatic change. Preliminary assumptions and arguments are presented here as a framework for future discussion and research. The most important assumption is controversial. It is that within the ocean itself an increment of heat behaves approximately as a passive tracer, i.e., it does not substantially alter the exchange processes that transfer it downward. In general, this dynamical assertion is certainly false, but it is relatively plausible in the upper ocean, where vertical mixing is driven predominantly by the wind and seasonal overturning and where the horizontal circulation is due mainly to wind stress rather than gradients in surface temperature. The nature of these exchange processes may be inferred from observations of other tracers in the ocean. For example, Figure 2.1 shows the distribution of tritium as determined by the GEOSECS (1973-1974) cruises along a section in the western Pacific from Antarctica to Alaska, approximately 10 years after the tritium was deposited at the surface, predominantly in the northern hemisphere during a 3-year period around 1964 (Fine et al., 1981). Tritium is a passive tracer, both dynamically and chemically. It is apparent that penetration had already occurred to a depth of 700 m in both hemispheres in mid-latitudes and to a depth of 300 m in the tropics, with the upper two thirds of this volume being of approximately uniform concentration within each hemisphere. Isopleths of constant concentration also coincide to a considerable degree with surfaces of constant potential density. This is consistent with the widely, though not universally, accepted view that below a surface-wind-mixed layer and a seasonal thermocline, which is some 50 m deep in the tropics and 200-300 m deep in mid-latitudes, the dominant exchange processes in the ocean are by quasi-horizontal circulation and lateral mixing on surfaces of constant potential density of water with wintertime surface characteristics. This isopycnal exchange (i.e., lateral on potential density surfaces) leads to ventilation of the remaining volume down to 1000 m. The ventilation time increases with depth, but in the upper few hundred meters it is at most a few years. A corresponding tritium section in the Atlantic shows similar features, except that in addition there is entrainment into the abyssal water north of 50Â° N. There are many pathways whereby a tracer such as tritium, or heat, can be carried downward from the ocean surface, and many reservoirs in which it can accumulate. However, as the concentration in a reservoir builds up, some of the tracer-enriched water is returned to the surface, and that reservoir becomes saturated. Pathways to deeper and larger reservoirs remain active. Thus, the volume of water reached by such pathways is an increasing function of time, but after a long time the specifics of the near-surface pathways are
26 ll - 2 o a o 0 = clir BO uj Â§ O O â g E>-, M â o SO> i> o .Â£*-.Â§ | cÂ£Â°- SH313IAI NI Hld3Q C Ifc- Q W5 .2 Â° -s Â° .X â 2 C Â°, .= - -E u oij -*" tn p , Q. "S ^ ^ 2 o 2 o E E< t; 3 C c z > ~ |5lJi o I s | E _c o -S
Principal Scientific Issues in Modeling Studies 27 unimportant, provided they are efficient enough to transmit the fluxes necessary to fill the deeper reservoirs. The accumulation time can be roughly estimated from the volume of specific reservoirs and the supposed exchange rates of water between them. For a time scale of a few weeks, the relevant volume is to a depth of 10-100 m because of wind-induced vertical mixing. On a time scale of 1 year, it is to the base of the seasonal thermocline, around 50 m in the tropics and 200-300 m in mid-latitude. The tritium data show that on a time scale of 10 years, the volume is to a globally averaged equivalent depth of about 500 m, access to most of that probably being through the surface layers poleward of 45Â° N and 45Â° S. On a time scale of 1000 years, the whole world ocean is involved (Munk, 1966). with an average depth of 5000 m. These time scales are for downward mixing. The feedback of a rise of surface temperature on the downward flux of heat at the surface must also be considered. The surface heat flux into the ocean decreases as the surface temperature rises toward a new equilibrium. The rate of rise of the surface temperature depends on the depth to which mixing takes place. As pointed out earlier, typical thermal time constants are 3 years for a 50-m-deep mixed layer, 30 years for a layer 500 m thick, and 300 years for 5000 m. Near the surface, the thermal time constant is much greater than the mixing time, allowing a thermal anomaly to penetrate with little negative feedback on surface heating. On the other hand, the ventilation time of the abyssal waters is substantially longer than the corresponding thermal time constant, implying that the exchanges with deep waters are relatively minor contributions to the heat budget of the layers above. Thus the main thermocline will remain more closely coupled to the atmosphere above than the abyssal water below. Unfortunately, the crossover between these extremes occurs at intermediate depths near the base of the main thermocline, where the ventilation times are relatively poorly known. From tritium and other data, Jenkins (1980) has estimated ventilation times in the Sargasso Sea equal to a few years above 700 m, about 50 years between 700 and 1500 m, and much longer below. It is not known how applicable these estimates may be to other oceans. The atmospheric effects of CO2 would be perceived by the ocean as a sudden change in downward heat flux at the ocean surface, as a change in wind stress, and as a change in net evaporation minus precipitation. Although variations in wind stress can substantially alter the circulation in the ocean, and hence the surface temperature, the changes calculated from atmospheric models appear to be small. Net evaporation minus precipitation affects the salinity of surface waters. In high latitudes, the latter controls whether winter cooling leads to deep convective overturning, with associated renewal of the abyssal waters, which comprise four fifths of the volume of the ocean, or to the formation of sea ice, which happens if the surface water is relatively
28 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT 25 100 TIME SINCE DOUBLING (years) 1500 FIGURE 2.2 Schematic representation of the time-dependent response of ocean-surface temp- eratures in the tropics and subpolar regions to an instantaneous doubling of atmospheric CO2 sustained indefinitely. fresh. However, since the present rate of renewal of abyssal water is relatively slow, believed to be once per 1000 years, salinity is probably not of dominant importance for the heat budget of the surface layers, except perhaps indirectly through its influence on the extent of sea ice. Attention will be concentrated here on the direct effects of an increased downward surface heat flux of about 4 W irr2. A quantitative treatment of these concepts requires at least consideration of a set of reservoirs at different latitudes, coupled in series and parallel according to the picture of oceanic mixing outlined above, to a simplified atmospheric model capable of distinguishing different latitude belts. It is assumed that for the small changes from present climate that are under consideration, perturbations may be superimposed in an approximately linear manner. This implies that the effects of an arbitrary rise curve in atmospheric CO2 can be inferred once the ocean response to any particular time sequence has been fully analyzed. For definiteness, consider an instantaneous doubling, maintained thereafter. The qualitative nature of the conclusions from the conceptual model described above is indicated in Figure 2.2. The curves show the rise in ocean-surface temperature in the tropics and in the subpolar regions (45-60Â° latitude) as a function of time after a hypothetical doubling of atmospheric CO2. After 1500 years, equilibrium has been reached, with a rise of about 2Â°C in the tropics and 4Â°C in subpolar regions. After 5 years,
Principal Scientific Issues in Modeling Studies 29 the temperature in the tropics has risen significantlyâabout 50 percentâ toward the equilibrium value, but thereafter the rise is much slower, being 80 percent complete after 100 years. In subpolar regions, the initial rate of rise is much slower, being 50 percent complete after 25 years. The decrease in poleward temperature gradient predicted by Manabe and Stouffer (1980) is thus delayed by several decades. The precise numbers shown in this diagram are clearly dependent on the detailed model under consideration. The general shape of the curves and the orders of magnitude are not. The relatively higher equilibrium rises in high latitudes are prescribed by a number of processes as described in the calculations of Manabe and Stouffer (1980). Possible additional effects due to sustained poleward transport of heat by ocean currents are not included. That 25-50 years is required to achieve most of the equilibrium rise in temperate latitudes is a consequence of the rapid downward exchange there, which is well documented by the tritium data. This rise time is substantially shorter in the tropics because on time scales of 1-3 years the upper 50-m layer is largely isolated from the remainder of the ocean, enabling partial equilibration to be achieved. The equilibration is not complete because of heat transfers to other latitude bands, primarily through the atmosphere but to some extent by near-surface ocean currents, and the short-term fractional rise depends on the efficiency assumed for each of these processes. Even after 100 years, the surface layers as a whole have not come to equilibrium because of the slow transfer of heat downward to abyssal depths, both by water sinking in polar regions and by cold water rising elsewhere. Because of uncertainties about the effective rate of such exchange, an estimate of the fraction of the temperature rise that is delayed 100 years or longer must be regarded as very tentative. Regarding heat as a passive tracer is particularly suspect in this case. The discussion so far has been entirely in terms of the transient response to an instantaneous but maintained doubling of atmospheric CO2, which is a highly implausible scenario. However, because of the linearity of the climate system to small perturbations, the response for any other rise curve is readily calculated by convolution of the curves shown in Figure 2.2 (or their more refined equivalent) with the instantaneous equilibrium temperature increase (roughly proportional to the logarithm of the atmospheric CO2 divided by the base value). For example, if a particular atmospheric CO2 rise curve scenario implies an equilibrium surface-temperature rise of 3Â°C in the subpolar regions by A.D. 2050, the actual surface-temperature response can be estimated by the history of the rise by using curves like those shown schematically in Figure 2.2. The rise in equilibrium temperature can be approximated by a finite number of step increases. The actual response of surface temperature in 2050 may then be estimated as the following sum: a
30 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT large fraction of increases in equilibrium temperature before A.D. 2000. about half of increases between A.D. 2000 and 2030, and a small fraction of increases after A.D. 2030. In the tropics the same principles would apply. On the basis of the assumption that tracers can serve as an approximate guide, we conclude that the mixing time scale for the main thermocline is less than the thermal time scale associated with increasing CO2. The waters of the upper thermocline will be closely coupled to the atmosphere on time scales of 1-2 decades. Any estimate of the time response of the climate system to increasing CO2 will have to take into consideration the thermal inertia of the upper thermocline. The ocean thermal response may also cause important regional differences in response other than the latitudinal effects already discussed. For example, the geometry of the ocean-land distribution will be important. The climatic response in areas downwind from major oceans will certainly be different from that in the interior of major continents. It is also reasonable to infer that the much smaller ratio of land to sea area in the southern hemisphere will be significant. Depending on how efficiently the atmosphere exchanges heat across the equator, a slower response to increasing CO2 might be expected in the southern hemisphere. Future modeling studies should stress the regional nature of oceanic thermal inertia, and atmospheric energy transfer mechanisms, taking into account that the local response time is proportional to the ocean's thermal inertia and the rate at which energy is exchanged with the atmosphere. Reliable quantitative estimates of the role of the ocean in the climate system will require a better understanding of the processes that give rise to the entire general circulation in the ocean and substantial improvements in our ability to model it. Progress to this end must be based on a broad range of research: continued oceanographic observations of density distributions, tracers, heat fluxes and currents, quantitative elucidation of the mixing process, substantial theoretical effort, and the development of models adequate to reproduce the relative magnitudes of a variety of competing effects. The problems are difficult, and complete success is unlikely to come quickly. Meanwhile, partially substantiated assumptions like those asserted here are likely to remain an integral part of any assessment. In planning the oceanographic field experiments in connection with the World Climate Research Program, we recommend that particular attention be paid to improving estimates of mixing time scales in the main thermocline. Effects of Sea Ice Sea ice has profound effects on climate in several ways. In winter, it allows the ocean to remain at the freezing point while the air temperature falls to
Principal Scientific Issues in Modeling Studies 31 much lower values. In summer, sea-ice melting locks the surface temperature to the melting point. Manabe and Stouffer (1980) have noted the effect of sea ice on the seasonal dependence of the sensitivity of climate to CO2 increases. They have shown how the existence of melting sea ice and the large thermal inertia of the ocean's mixed layer prevent the summer temperatures in a high-CO2 climate from rising much above present levels. Bolin (1981) has recently pointed out that the vertical exchange processes in Antarctic waters "... are large in magnitude and are crucial for a proper understanding of the global CO2 exchange between the atmosphere and the sea. ..." The role of the Arctic Ocean halocline (Aagard et al., 1981) may also be important in the vertical exchange processes of heat, salt, and CO2. At high latitudes, warming must be confined to the winter season, when CO2-induced reductions of sea-ice thickness result in the increase of upward heat conduction through ice. Sea ice also has an important role in the formation of deep water. Ice may be frozen in one location and carried by currents to melt in another location; this process is important for creating cold, salty water that descends to great depths. However, present knowledge of the interaction of ice formation and deep-water formation is still rudimen- tary. In particular, the relative importance of atmospheric radiation and oceanic mixing processes to the different seasonal sea-ice variations in the northern and southern hemispheres need to be explored. Thus, it will be difficult to say even qualitatively what role sea ice will play in high-latitude response and deep-water formation until the climatic factors that control the areal extent of polar pack ice in the northern and southern hemispheres are known. Field experiments are required to gain fundamental observational data concerning these processes. CLOUD EFFECTS Cloudiness-Radiation Feedback Two uncertain aspects of cloudiness-radiation feedback must be considered. The first is the question of whether cloud amounts, heights, optical properties, and structure will significantly change in response to CO2-induced warming. If such changes are not significant, then obviously there will be no cloudiness- radiation feedback. But if cloud amounts, types (e.g., cumiliform versus stratiform), heights, optical properties, and structure are influenced by climatic change, then both the solar and the infrared component of the radiation budget will be altered; it is the relative role of these probably small changes and their regional distributions that constitutes the second uncertain aspect of the problem.
32 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT If, for example, cloud amounts decrease, then, since cloudy regions are generally brighter than clear-sky regions, the decrease would reduce the planetary albedo, resulting in increased solar heating of the surface-atmo- sphere system. But decreased cloudiness would also reduce the infrared opacity of the atmosphere, resulting in increased infrared cooling of the surface-atmosphere system. Thus, the separate solar and infrared modifica- tions act in opposite directions. A corresponding change in effective cloud height would further modify the outgoing infrared radiation, with infrared cooling being enhanced, for example, by a reduction in effective cloud height, since the lower (and hence warmer) clouds would radiate more energy to space. Employing a GCM that predicts both cloud amount and cloud height, Manabe and Wetherald (1980) have, for doubling and quadrupling of atmospheric CO2, suggested that equatorward of 50Â° latitude, net cloud amount and effective cloud height are reduced by CO2-induced warming, with both effects acting to increase the outgoing infrared radiation. However, this is nearly compensated within their model by the corresponding increase in absorbed solar radiation due to reduced cloud amount. Poleward of 50Â° latitude they find an increase in net cloud amount without any substantial change in effective cloud height. The absence of changes in cloud height, which contributed to the near solar-infrared compensation at lower latitudes, is in effect offset by reduced insolation at higher latitudes, such that again the model predicts near compensation for the changes in absorbed solar and outgoing infrared radiation. However, Manabe and Wetherald (1980) emphasize: "In view of the uncertainty in the values of the optical cloud parameters and the crudeness of the cloud prediction scheme incorporated into the model, it is premature to conclude that the change of cloud cover has little effect on the sensitivity of climate." The potential importance of cloud feedback effects has been emphasized by a number of Australian scientists (e.g., Pearman, 1980). There have also been suggestions (e.g., Petukhov et al., 1975; Charlock, 1981; Hunt, 1981; Wang et al., 1981) that changes in cloud optical properties associated with climatic change might be important in modeling cloudiness- radiation feedback. Thus, the suggestion of solar-infrared compensation is at best tenuous; and even if this were to be the case on a global scale, there may be important regional exceptions. Moreover, if small changes in cloud amount are important, they will be difficult to predict from model calculations. Alternative approaches toward estimating the relative solar-infrared com- ponents of cloudiness-radiation feedback involve empirical studies using Earth radiation budget data. In one such approach, Cess (1976) has suggested solar-infrared compensation, whereas Ohring and Clapp (1980) and Hartmann and Short (1980) have suggested that the solar component dominates over
Principal Scientific Issues in Modeling Studies 33 the infrared component by roughly a factor of 2. Cess employed the satellite data compilation of Ellis and Yonder Haar (1976), while the other two studies utilized radiation budget data derived from NOAA scanning radiometer measurements. Recently, Cess et al. (1982) have reviewed these studies and suggest that the conclusions of solar dominance might be attributable to the NOAA data's being derived from narrow spectral measurements. Clearly, the empirical approaches comprise an important means of studying the cloudiness-radiation feedback problem. The approach by Ohring and Clapp (1980) is particularly attractive. They have employed interannual variability in regional monthly-mean radiation data, from which they estimate the relative solar-infrared cloudiness feedback components by attributing this variability to interannual variability in cloudiness. We recommend re- examination of their conclusions employing radiation budget data that do not suffer the possible deficiencies noted above. In summary, while it is conceivable that cloudiness-radiation feedback does not substantially influence climate sensitivity, it is clear that additional studies of this feedback mechanism are necessary. This will be a difficult problem that will require a multifaceted approach. One should not trust model prediction schemes until they produce meaningful simulations of observed seasonal cloud cover and the seasonal radiation components. At present, there are no published models that do this. While the empirical approach offers an attractive means of attacking the problem, it will require clarification of sampling bias within Earth radiation budget data. Stratus-Sea-Ice Interactions In simple climate models, ice-snow albedo feedback contributes substantially to the CO2 warming at high latitudes. However, it seems likely that in regions where sea ice is reduced, evaporation will increase and possibly lead to increased low-level stratus cloud cover, which would reflect solar radiation and at least partially reduce the albedo feedback. This effect is not certain to occur, because altered atmospheric temperature profiles due to added CO2 could also cause a decrease of cloud cover. Examination of 3-D GCM experiments with doubled CO2 (Manabe and Wetherald, 1980; Hansen et al., 1982) show an increase of cloud cover of several percent in the region where the added CO2 melts the sea ice. The change of planetary albedo is 4 times smaller than the change of ground albedo, as a result of both cloud shielding of the ground and increased local cloud cover with increased CO2. This result must be considered tentative, in view of the great oversimplification of the calculation of clouds in climate models, but it serves to emphasize the possible importance of cloud processes.
34 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT TABLE 2.2 Equilibrium Global-Mean Surface-Temperature Effect of Indicated Changes in Abundance of Several Trace Gases as Computed Using an RC Model (Lacis et al., 1981) Gas Hypothetical Abundance Change Temperature Change (Â°C) CO, 300 ppm â Â» 600 ppm 2.9 N,0 0.28 ppm -> 0.56 ppm 0.6 CH4 1.6 ppm â > 3.2 ppm 0.3 CC12F2 + CC1,F 0 â > 2 ppb each 0.6 0, 25% decrease -0.5 TRACE GASES OTHER THAN CO, Most modeling endeavors concerning the CO2-climate problem address only the question of the climatic response to increasing atmospheric CO2, while the amounts of other atmospheric gases remain fixed. But associated changes, either climatologically or anthropogenically induced, of minor atmospheric constituents can also be of significance. For example, Ramanathan (1975) suggested that chlorofluoromethane (CFM) concentrations of only a few parts per billion (ppb) could produce a significant global warming. Subsequently, Wang et al. (1976) analyzed the climatic response to changes in a number of atmospheric trace gases. Trace atmospheric gases that absorb in the infrared may enhance or counteract CO2 warming if their abundance should change (Ramanathan, 1975; Wang et al., 1976). In general, if these gases increase, they will lead to warming, although for gases that are inhomogeneously mixed, such as ozone (O3), the temperature change could be in either direction depending on changes in the vertical distribution of the gas. A comparison of the equilibrium global-mean surface-temperature effect of changes in several trace gases is shown in Table 2.2, based on calculations with a model having a sensitivity of 2.8Â°C for doubled CO2 (Hansen et al., 1981). While this table illustrates the importance of trace gases, there is a considerable difference between this calculation and other model results for CFM'S. The model results of Ramanathan (1975), Reck and Fry (1978), and Karol (1981) show a global surface warming of 0.8-0.9 K for CFM increase from 0 to 2 ppb. There are several ways in which the amount of a climatologically important atmospheric trace gas might be alteredâfor example, â¢ As a consequence of the direct anthropogenic emission of the gas into the atmosphere, as is the case for CFM'S; â¢ As a result of the anthropogenic emission of gas or gases that, through interactive atmospheric chemistry, alter the amounts of climatologically
Principal Scientific Issues in Modeling Studies 35 important trace gases (e.g., increasing anthropogenic emissions of CO are expected to increase the amounts of tropospheric CH4 and O3 (Rowland and Molina, 1975; Logan et a/., 1978; Hameed et al., 1980), both of which are greenhouse gases (see Table 2.2)); â¢ Owing to biospheric changes caused by CO2 warming, e.g., increased CH4 production by warmer wetland areas or the release of CH4 now trapped as a hydrate in permafrost regions; â¢ Owing to altered atmospheric temperature and water-vapor concentration resulting from increased atmospheric CO2, which would produce changes in atmospheric chemistry and therefore in trace-gas abundances. We begin by discussing the last of the above items. While an increase in atmospheric CO2 would warm the surface and troposphere, the stratosphere would be likely to cool as a consequence of enhanced CO2 emissions. This raises the possibility of a change in stratospheric ozone as a consequence of temperature-dependent stratospheric chemistry. However, model studies (Luther et al., 1977; Boughner, 1978) show that for doubled CO2 the column density of O3 would be increased by only 1-3 percent, which would produce an insignificant climate feedback. A second possible interaction pertains to the troposphere. CO2-induced tropospheric warming would produce increased tropospheric water vapor, which, from a radiation standpoint, would result in the well-known positive climate feedback. But as a consequence of tropospheric chemistry, the increased H2O would reduce tropospheric CH4 and O3. Since CH4 and O3 are both greenhouse gases, this aspect of the process comprises a negative climate feedback. Employing coupled climate- chemical models, both of which are vertically averaged 1-D (latitude) models, Hameed et al. (1980) have found that the global warming produced by a 70 percent increase in atmospheric CO2 would result in 15 and 10 percent reductions, respectively, in tropospheric CH4 and O3. The resulting negative feedback would, however, be quite minor, reducing the CO2-induced global warming by about 10 percent. Such decreases in tropospheric CH4 and O3 do not account for increasing anthropogenic emissions of CO, CH4, and NOV resulting from the fossil-fuel burning that produced the increased atmospheric CO2. Logan et al. (1978) and Hameed et al. (1980) have suggested that such anthropogenic emissions could lead to significant increases in tropospheric CH4 and O3 during the next century, with a corresponding enhancement of the CO2-induced global warming. In the latter study, the CH4 and O3 increases were obtained despite the previously discussed negative chemical feedback due to tropospheric water vapor. It should be emphasized that, although the troposphere contains only 10 percent of the total O3, the infrared opacity of tropospheric O3 is,
36 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT TABLE 2.3 Scenario for Global Trace-Gas Alterations (from Ramanathan. 1980) Mixing Ratios by Volume % O, Change CO, CH, N_,O CFM'S (ppb) Tropo- sphere Period (ppm) (ppm) (ppm) Total Pre-1940 300 600 1.4 3.3 0.3 0.6 0 2.2 0 100 0 -6.5 22nd century because of enhanced pressure broadening, nearly the same as that of stratospheric O, (Ramanathan and Dickinson, 1979). Predictions of O, and CH4 changes, based on biogeochemical cycles, are of course very uncertain. Other human activities such as the growing use of nitrogen fertilizers could lead to increases in atmospheric concentrations of N2O and CFM'S. As illustrated in Table 2.2, both are significant greenhouse gases. Moreover, CFM'S, through interactive stratospheric chemistry, may additionally lead to a reduction in stratospheric O, (Logan et al., 1978). Recently, Ramanathan (1980) assumed a steady-state scenario for alterations in trace gases from their assumed values prior to 1940. and their subsequent climatic impacts, that incorporates the consequences of fossil-fuel burning as well as anthro- pogenic emissions of CFM'S and N2O. The hypothetical increase in trace gases is assumed to occur in the same time period for which a doubling of CO2 occurs. This scenario was compiled from a variety of sources, including carbon-cycle models, photochemical models, and future energy-consumption models; it is summarized in Table 2.3. Although tropospheric O, is assumed to double, the total ozone amount decreases because of increasing CFM'S, which reduce stratospheric Ov This also results in a change in the vertical distribution of stratospheric O,. Thus, although the radiative effects of trace gases are almost additive, their influences can be chemically coupled. The climatic effects of the trace-gas alterations shown in Table 2.3 are quite substantial. The twofold CO2 increase produces, with Ramanathan's RC model,* a global warming of 2.0Â°C, but when the other trace-gas changes are included, global warming is increased to 3.6Â°C. Put another way, the model scenario suggests that roughly 40 percent of the global warming would be due to changes in trace gases other than CO2. While the model scenario is of course highly uncertain, it nevertheless serves to illustrate the possible climatic importance of anthropogenic changes of trace gases. *Note that this model has fixed cloud altitude and is thus less sensitive by about 40-60 percent than the model used to generate the 2 x CO, value in Table 2.2 that fixes cloud temperature.
Principal Scientific Issues in Modeling Studies 37 To date, the only known significant change in trace-gas abundance (other than that of CO2) has been in the CFM'S, which have increased from an essentially zero abundance a few decades ago to 0.3 ppb of CC13F2 and 0.2 ppb of CC1,F (Mendonca, 1979), for which the equivalent greenhouse warming is roughly 0.06Â°C. No major trend of O, abundance (either tropospheric or stratospheric) has been observed. N2O has been monitored for several years, and small but significant increases have recently been detected (Weiss, 1981). Although there have been tentative suggestions of a slight increase in CH4 over the past decade (Hudson and Reed, 1979; Heidt and Ehhalt. 1980: Graedel and McRae, 1980: Rasmussen and Khali. 1981), this gas has not been carefully monitored. But atmospheric chemical models (Logan et al., 1978; Hameed et al.. 1980) indicate that the CH4 abundance may be very sensitive to anthropogenic influences, and it is therefore recommended that careful monitoring of CH4 be initiated. Our knowledge of the atmospheric concentrations of these radiatively active trace constituents before the mid-nineteenth century is quite poor. Information is potentially available from tree rings, corals, and glacial ice, and some estimates of CCK concentrations have been made (e.g., Berner et al., 1980). Better knowledge of past concentrations of these gases would greatly facilitate direct assessment of their effects on global climate. In summary, the estimated net impact of measured changes of trace gases during this century has been an estimated equilibrium warming of 0.1-0.2Â°C, which is substantially smaller than the estimated equilibrium warming of roughly 0.5Â°C for the estimated increase of 45 ppm of CO2 during the same period, both numbers being based on a model with 2.8Â°C sensitivity for doubled CO2. However, since model studies indicate that trace-gas abundances might change significantly in the future, it is important to monitor the primary trace gases, because they could significantly enhance future CO2 warming. ATMOSPHERIC AEROSOLS In addition to changes in atmospheric CO2 and other trace gases, atmospheric aerosols provide another potentially significant source of climate variability. But this problem is far more complex than that involving trace-gas changes, since the radiation effects of the aerosols depend on their composition, size, and vertical and global distributions. Stratospheric aerosols, which persist for a few years following major volcanic eruptions, can produce a substantial reduction in global surface temperature and can explain much of the observed natural climatic variability (Hansen et al., 1981: Gilliland. 1982). Such an aerosol is composed primarily of aqueous sulfuric acid droplets (Rosen, 1971), and its main climatic impact is to backscatter solar radiation, thus
38 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT reducing the amount of solar radiation that is absorbed by the Earth-atmosphere system. There is a partially modifying heating due to the infrared greenhouse effect of the aerosols (Pollack et al., 1976), but the net aerosol effect is surface cooling. Although there are indications of an anthropogenic source of stratospheric aerosols (Hofmann and Rosen, 1980), it is not clear if or when this will become significant in comparison with volcanic sources. The climatic effect of tropospheric aerosols is much less certain. Although anthropogenic aerosols are noticeable in regions near their sources, there does not appear to have been a significant long-term increase in the aerosol level in remote regions of the globe except possibly in the Arctic, where substantial concentrations of anthropogenic aerosols build up during the winter and spring (Roosen et al., 1973; Cobb, 1973: Rahn and McCaffrey, 1980; Shaw, 1981). However, no long-term trend in this anthropogenic aerosol has yet been established. The natural aerosol consists of sulfates, marine aerosols, and windblown dust. Both sulfates and windblown dust could increase as the result of man's activities, the former owing to industrial activity (Bolin and Charlson, 1976) and the latter owing to agricultural activities and desertification. Increased tropospheric sulfates would lead to global cooling. With respect to windblown dust, recent measurements indicate a midsolar imaginary refractive index (which governs absorption) of roughly 0.01 (Patterson et al., 1977; Carlson and Caverly, 1977; and Patterson, 1981), and aerosol-climate models that employ this value (Ohring, 1979; Coakley et al., 1982) suggest that increased windblown dust would also lead to cooling on a global scale, but perhaps with important regional exceptions. Because this aerosol absorbs a significant amount of solar radiation, it could lead to an albedo decrease over highly reflecting regions such as deserts and to an albedo increase over darker regions such as oceans (Coakley et al., 1982). An additional anthropogenic component of tropospheric aerosols is indus- trial soot, which, because it is highly absorbing, would lead to global warming (e.g., Hansen et al., 1981). Again, however, it must be emphasized that insufficient observations have been made to determine global trends for this aerosol component. But recent chemical and optical analyses of Arctic haze indicate that during the spring and early summer the haze particles contain a high concentration of graphite carbon (Porch and MacCracken. 1981, based on work by Rosen), and it has been suggested that this Arctic soot may be influencing the Arctic climate (Budiansky, 1980). Recently. Porch and MacCracken (1982) have modeled the possible effects of carbonaceous aerosols on the Arctic climate, and they found that the springtime Arctic soot could lead to an average heating rate of 0.06 K day1 in the lowest 5 km of the atmosphere under cloud-free conditions. Interestingly enough, this is similar to estimated heating rates at northern latitudes that are due to a doubling of atmospheric CO2.
Principal Scientific Issues in Modeling Studies 39 In summary, variations in stratospheric aerosols have continued and will continue to be dependent on volcanic activity, contributing to natural climatic variability. Although there is currently no evidence that a global trend exists in the components of tropospheric aerosols, future anthropogenic changes in tropospheric aerosols are possible. But the climatic impact of such changes, if indeed they occur, cannot currently be determined. Because of the differing optical properties of the individual components of tropospheric aerosols, one cannot even conclude that possible future anthropogenic changes in aerosol loading would produce global heating or global cooling. Furthermore, increased tropospheric aerosols could influence cloud optical properties (Charlock and Sellers, 1980) and thus possibly modify cloudiness-radiation feedback as discussed in the section Cloudiness-Radiation Feedback. VALIDATION OF CLIMATE MODELS Need for Model Validation Mathematical-physical models, whether in a highly simplified form or as an elaborate formulation of the behavior and interaction of the global atmosphere, ocean, cryosphere, and biomass. are generally considered to be the most powerful tools yet devised for the study of climate. This is in part due to the reproducibility (and in this sense the objectivity) of a model's results, in part to the opportunity to trace cause-and-effect relationships within a consistent framework of interacting processes, and in part to the possibility of performing numerical experiments under a wide variety of conditions. Our confidence in climate models comes from a combination of tests of the correctness of the models' parameterizations of individual processes and comparisons of the models' sensitivity to observed seasonal variations. All models, however, require the parameterization of a number of subgrid-scale processes important to climate, such as cloudiness, precipitation, and the radiative and turbulent heat fluxes in the planetary boundary layer. The more realistic climate models also simulate the transient synoptic-scale eddies of middle and higher latitudes and therefore display an inherent variability or noise in their climatic statistics. Each of the above factors influences the extent to which we can (or should) trust the results of climate models, and all are aspects of the general validation problem. While the absolute accuracy of a model may be of greatest interest to the scientist, a knowledge of the uncertainty of a model's results may be of equal or greater importance to those using or acting on the results. This is especially true in the case of the CO2-climate problem, in which decisions of immense economic and social consequences may be made on the basis of information supplied by models. The validation of climate models is therefore becoming an increasingly important matter and should be undertaken on a
40 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT systematic and sustained basis; a comprehensive validation program would include the assembly of a wide variety of observational data specifically for the purpose of model validation and the development of a validation methodology. The validation of numerical climate models comes from a hierarchy of tests at different levels, many of which are part of the stock in trade and are normally not published. Detection of outright errors in a complex computer code is a formidable task that requires extraordinary care. Tests are also made to see if the behavior in isolation of individual subsystems, such as the boundary layer or the treatment of radiation, resembles that established from field or laboratory observations or from more detailed models based on known physical principles. The parameters used in a well-constructed climate model are mostly explicitly derived from such comparisons and are not subject to arbitrary adjustment or tuning. Every investigator establishes in a more or less systematic manner the sensitivity of the conclusions to the boundary conditions applied (such as the distribution of continents and oceans) and to the parameterizations used. Only the more significant sensitivities are singled out for explicit discussion in the literature, but accepted models are based on the judgments of many individuals independently exploring approaches that are similar in kind but different in detail. An important role is played by highly simplified models that approximately reproduce the behavior of the more elaborate models; such models can be used to explore rapidly a wide variety of situations and to identify key experiments for more elaborate investigation. Only with such a background of testing for internal consistency and reasonable overall structure and behavior does the comparison of model output with actual observational data begin. Present State of Model Validation The primary method for validating a climate model is to determine how well the model-simulated climate compares with observations. Usually some verification against observed data is performed during the preliminary testing of a model, especially in those models with a large number of parameterized processes, and here it is important to distinguish between model calibration or tuning and model validation. When, for example, the local heat capacity in an energy-balance climate model is adjusted to give a realistic phase and amplitude for the seasonal variation of surface temperature, then this temperature variation can no longer be used to verify the model's performance. However, the solutions (if any) for other processes and variables predicted by the model could be used for validation, as could variations of the temperature, provided they were not on the same scales of space and time
Principal Scientific Issues in Modeling Studies 41 used to calibrate the model in the first place. In some models the imposition of particularly influential boundary conditions is also a form of model calibration, in that variables closely associated with the prescribed conditions cannot be used for independent model validation. An example is the surface- air temperature simulated over the oceans in a GCM in which the sea-surface temperature has been prescribed. Most climate models have been given at least preliminary validation in terms of a comparison of climatological averages with the simulated time means of selected climatic variables. In the case of GCM'S, for example, such validation has usually included the sea-level pressure, the temperature and geopotential at one or more levels in the free air, and the total precipitation as simulated for a month or season in comparison with the corresponding climatological distributions. In terms of these and other variables, modern climate models provide a reasonably satisfactory simulation of the large- scale global climate and its average seasonal changes. Even here, however, the different sets of observed data that have been used for validation are not of the same quality, and the effects of uneven data coverage are not taken into account. For many of the variables simulated in a climate model it is difficult to find suitable observational data sets available for model validation, even though these variables may be ones in which there is great practical interest (see the section Detection Strategies). Examples of such data are the cloudiness, the surface evaporation and heat flux, and the soil moisture, commonly simulated in GCM'S, and the parameterized fluxes of heat and moisture, commonly simulated in one- and two-dimensional models. As important as the time-averaged distributions of the climatic variables themselves may be, the simulated variance, covariance, and other higher- order statistics are of equal importance in the validation of a climate model. Only a limited amount of such verification has been made for GCM'S because of the need for extended integrations and the lack of the needed observational statistics. Many of the more highly parameterized models, moreover, address at most only an equilibrium seasonal climate and thus do not simulate either the seasonal or the interannual variability associated with the transient synoptic-scale eddies. The validation of the equilibrium statistics of such models is therefore confined to the comparison of equilibrium fields with long-term seasonal climatology. In addition to verification in terms of time-averaged statistics, an important aspect of climate-model validation is the confirmation of the essential correctness of the major physical parameterizations in the model. Chief among these are the model's treatment of clouds and their effects on radiation, the variations of snowcover and sea ice, and the surface exchanges of heat and moisture; in no climate model do these parameterizations rest on completely satisfactory theoretical or empirical bases. If climate models are
42 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT to project future climate changes correctly in response, say, to increased levels of atmospheric CO2, then their parameterizations must be valid under as wide a range of conditions as possible (i.e., they should work for the right reasons). In contrast to the extended global integrations required for the validation of a climate model's overall performance, the validation of the parameterization of specific physical processes can often be accomplished with specialized observations in a local region for a limited period of time. Such "process" validation is the aim of several large-scale observational programs either now under way or in the planning stages. The behavior of climate models is also influenced to some extent by the treatment of land-surface processes such as albedo and evapotranspiration. These depend on vegetation coverage and may interact with climate change in ways that are as yet poorly understood. This interaction is particularly significant for regional processes such as desertification, deforestation, and the distribution of precipitation. It is recognized that a major uncertainty about predictions of CO2-induced climate change is due to the fact that the atmospheric models have not been coupled to realistic ocean models. In this connection, an important technical issue is to devise computationally efficient schemes to couple climatic subsystems with order-of-magnitude different response times (e.g., the atmosphere and oceanic mixed layer). Asynchronous coupling has been used in this connection, but such schemes can distort the transient behavior of coupled atmosphere-ocean models (e.g., Dickinson, 1981; Ramanathan, 1981). In order to maximize the number of numerical experiments performed within a fixed computing budget, it is important to investigate the errors inherent in various asynchronous coupling schemes. There is also skepticism that any model calibrated to today's climate will be useful for predicting a future climate in which both the ocean and atmosphere could be much different. In the case of the ocean, there is nothing equivalent to the historical data set based on daily global observations that is available for the atmosphere. This view overlooks new sources of data that are now available or will soon become available for testing water-mass models of the ocean. In addition to the traditional hydrographic data for the global ocean, which provide the fields of time-averaged temperature and salinity, satellites and ship-of-opportunity data have now provided rough maps of mesoscale eddy intensity. In addition, the GEOSECS program and the follow-up Transient Tracers Program have provided two invaluable 3-D synoptic pictures of the penetration of tritium into the northern hemisphere oceans since its injection during the weapons tests of the 1960's. Data on other tracers such as bomb-produced I4C and radon are not so extensive but still provide valuable constraints for models. While the existing data base represents an inadequate sampling of the large-scale density structure over
Principal Scientific Issues in Modeling Studies 43 many parts of the world ocean, ocean models are beginning to reach a level of development where these data can be used for verification (Haney, 1979; Huang, 1979; Bryan and Lewis, 1979). Recently, the mesoscale statistics of the North Atlantic have been predicted in a high-resolution model (Holland and Rhines, 1980). For ocean models used in the study of the transient response of climate, anthropogenic tracers such as tritium and bomb-produced I4C are particularly valuable. These tracers show how particles introduced at the surface are carried downward. The data set collected by the GEOSECS program and cooperating groups provides a useful verification for ocean transport models. In view of the importance of the ocean in the response of climate to increasing CO2, it is recognized that the development of ocean models to a level comparable with that of atmospheric models is a matter of urgency. It is recommended that the requirements of ocean models be given high priority in the planning of oceanographic field programs planned by the World Climate Research Program. Much has been learned with atmospheric models without an ocean component. In the same way, much can be learned with an ocean model coupled to a very simple atmospheric model or one with specified boundary conditions. There are many complex problems in developing coupled ocean- atmosphere models, not the least of which is that the atmospheric models have been calibrated without active oceans. There is no need, however, to wait until fully tested, universally accepted coupled models are available before moving beyond the oversimplified 1-D models now being used to interpret the tracer data and estimate transient climate response. An inadequate but growing body of data exists for verifying models of ice-pack formation and movement. Satellites provide data on the extent of the ice, and satellite-tracked beacons are providing a new data set on the movement of ice. Less adequate is the information on ice-thickness distribution in the polar oceans. Models of the ice pack designed for climate-response calculations have been verified against the observed distribution of ice (Parkinson and Washington, 1979; Parkinson and Kellogg, 1979; Hibler, 1981), but the level of verification in a fully integrated climate model is less satisfactory (Manabe and Stouffer, 1980). Planetary Studies One useful test of the greenhouse theory can be obtained from empirical examination of other planets that in effect provide an ensemble of experiments over a wide range of conditions. The atmosphere of Mars, when it is dust- free, is relatively transparent in the infrared, and the greenhouse effect is hardly measurable. However, frequent dusty conditions and the small
44 ^ "= . . c t .2 O > â S O DC Q <^l 3C T^.i f-' ri ri r- c 2 o / / 3 i. =5 Â§- 2 c > g s El UJ rj xi O ri DC r i ri ri r- -o u 1- 5 3 ~ -*' ~ 5 Â§ ~ H- j ri o^ 3O ri ri ^ I |+, o ^ ii 2 3 jL; i t- Â«-. fS y 5 = 1 Â£..= ^ r-i r-\ r-i "c Â§ i S I 's i ,if =c - * R _^ - OJ ~ f 1 3 E It . . J= J^ 'C 4> ^- 3 â Â« ^ r- .^ 2 iN - D 4J ' â | E o ^= â ~ c 5 - i S .2 g â f - ; i ( *- s U e *" e "E ,:->f 4J o t -o o -^ â 1^1 C a: Â«ri o >o â r*-, r- K. f' O Sn ^ odd &" â + ~ UJ fl OJ ,1) t. ^^ S bserved c V E ^1 ". s. ft r- Â« J^ O f? .' ^ (I jl Cj g <-> *" *-' 5 .2 ^ ,e .5 j> -T: en P i> - .-'s " u, Â° 3 UJ CQ ^j C Â«'9-| CD f- OJ |11 S UJ > .i Â« c ed CL
Principal Scientific Issues in Modeling Studies 45 magnitude of horizontal energy transport by the atmosphere permit study of the greenhouse mechanism on a local basis (Gierasch and Goody, 1972). On Earth, water vapor and clouds make the troposphere too opaque for pure radiative transfer of heat with a stable lapse rate. The mean lapse rate is F ~ 5-6Â°C km^1, which is less than the dry adiabatic value (~10Â°C km~') because of latent heat release by condensation as moist air rises and cools and because the atmospheric motions that transport heat vertically include large-scale atmospheric dynamics as well as local convection. The mean radiating level occurs in the midtroposphere, at altitude H ~ 6 km. The atmosphere of Venus is opaque to infrared radiation for most altitudes between the surface and the cloudtops, owing to CO2, H2O, and aerosol absorption. The lapse rate is near the dry adiabatic value for the predominantly CO2 atmosphere (~7Â°C km^1), because of the absence of large latent heat effects and the small effect of large-scale dynamics on the vertical temperature gradient. The cloudtops radiating to space are at altitude H ~ 70 km. Observed surface temperatures of Mars, Earth, and Venus (Table 2.4) confirm the existence, nature, and magnitude of the greenhouse effect. Climatological data being collected by spacecraft at Venus and Mars will permit more precise analyses of radiative and dynamical mechanisms that affect the greenhouse warming. Of course, these planetary tests do not validate the current predictions for CO2 warming on Earth, but they do help provide confidence in the predicted magnitude of the equilibrium greenhouse warming due to a given atmospheric composition. The planets also provide the potential for greater tests of 3-D atmospheric models: it should ultimately be possible to use basically the same model for different planets, changing parameters such as solar constant, planet's gravity, rate of rotation, atmo- spheric density, and composition and in this way obtain a better understanding of basic mechanisms. Alternative Modeling Approaches Laboratory experiments on the behavior of differentially heated rotating fluids have provided insight into a number of basic hydrodynamic processes that are relevant to the circulation of the atmosphere and ocean. For example, the structure of such features as jet streams (Fultz et al., 1959; Dolzhanskiy and Golitsyn, 1977; Hide, 1977; Pfeffer et al., 1980) and the dynamics of baroclinic instability have been examined under a wide range of geophysically relevant conditions. Laboratory experiments can also contribute to our understanding of certain other processes such as small-scale turbulence and mixing. However, laboratory models cannot simulate usefully the majority of the climatically important physical processes such as the effects of changes in radiatively active gases, aerosols, albedo, and other surface characteristics
46 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT and the hydrologic cycle and cloudiness. Nevertheless, laboratory modeling studies (and their numerical counterparts) may be useful for study of the dynamic conditions that may occur in a CCK-enriched climate. They should be interpreted insofar as possible in terms of the processes that need to be properly parameterized in large-scale circulation models. Improvement of Model Validation In order to improve our knowledge of the performance of climate models and to increase their ability to project the climate changes likely to result from increased atmospheric CO2 in particular, we recommend that a climate- model validation methodology be developed and that it be vigorously pursued for as many documented models as possible. As key elements of such a validation methodology, we recommend: â¢ The systematic determination of the statistical properties of the perfor- mance of a hierarchy of climate models, including the geographical and seasonal distribution of the simulated means and higher-order statistics of modeled variables and processes. This evaluation should include a statement of the uncertainty or error bars of all climatic statistics, as determined from control integrations of appropriate length. â¢ The systematic assembly of a standardized climatological data base for model validation, with the statistics of all available climatic variables, processes, and boundary conditions determined on time and space scales consistent with those resolved in climate models. Those elements of such a validation data base that now exist should be identified and efforts made to tap effectively all current observational programs of climatic relevance. â¢ The development of appropriate regional or local observational data sets for the purpose of validating specific climate-model parameterizations, such as stratiform and convective cloudiness, the radiative effects of aerosols, and the subgrid scale fluxes of heat and moisture over both the ocean and vegetated land. â¢ The design and adoption of a set of universal sensitivity tests under standardized conditions, in which the performance of all climate models would be systematically compared with each other as well as with the model's performance under normal or control conditions. Such experiments might include, for example, changes in the solar constant (say, Â±2 percent), the surface albedo (say, Â± 10 percent), and of course changes in the atmospheric CO2 concentration (say, a doubling and a quadrupling). After such intercal- ibration, standardized transient response experiments with time-dependent climate models would provide useful further validation. In these, for example.
Principal Scientific Issues in Modeling Studies 47 a step-function increase of atmospheric CO2 could be prescribed and the response calculated over 50 years. â¢ The development of diagnostic and statistical techniques that would provide greater insight than do present methods into the physical and dynamical processes responsible for a modeled change of climate. Of particular value would be techniques to permit the attribution or "tagging" of climatic changes as due to increased CO2 rather than to other factors. We also recommend that climate-model performance be evaluated in terms of phe- nomenological measures and information as needed in climate-impact as- sessment, such as number of rainy days, storm tracks, frosts, drought, degree days, growing season, and extreme temperatures. â¢ The systematic assembly of climate information from paleoclimates and other sources such as planetary atmospheres, which are helpful in the validation of climate models.