U.S. Dept. of Commerce / NOAA
/ OAR / GFDL
*Disclaimer
In collaboration with Phil Partain of Colorado State University, line-by-line + Monte-Carlo/Equivalence Theorem (LBL+MCET) computations were performed for overcast atmospheric conditions. Due to the computational time needed for the Monte Carlo algorithm, it is employed only once for a given set of drop single-scattering parameters. Thus, the effect of spectral variations in gas absorption are ignored in deriving cloudy layer reflection and transmission. In order to test the viability of this approximation, as well as the MCET technique in general, differences in the fluxes and heating rates derived from LBL+MCET and line-by-line + doubling-adding (LBL+DA) computations were analyzed as a function of the number of cloudy layers (geometrical thickness). Specifically, the effect of increasing the geometrical thickness in a cloud using the "CS" size distribution used in the ICRCCM (Intercomparison of Radiation Codes in Climate Models) study with a fixed total drop optical depth is examined. The fractional differences in both the absorbed flux in the cloud and the individual cloud layer heating rates increase with increasing geometrical thickness, but remain modest (3%, 5%, respectively) for even the thickest case. However, there is a growing overestimate of the absorbed flux in the atmosphere as the geometrical thickness increases. There is likewise an increasing overestimate of heating above the cloud; the maximum fractional difference exceeds 30% for the two thickest cases. There is a corresponding increase in the underestimation of the reflected top-of-the-atmosphere (TOA) flux. These differences point out some limitations of the LBL+MCET technique as applied presently to plane parallel clouds.
3.1.2 Shortwave Parameterizations
A more realistic assessment of the stratospheric temperature changes in SKYHI (the GFDL troposphere-stratosphere-mesosphere general circulation model (GCM)) due to the improved accounting of CO2 shortwave heating has been made possible by the completion of a 10-year SKYHI GCM integration. A control run that uses the new solar radiation algorithm with CO2 shortwave heating included is compared against the new GCM integration that retains the same shortwave algorithm, but neglects the CO2 heating. The results give a reasonable approximation of the effect of the improved accounting of CO2 heating, since the older parameterization produced a very small stratospheric heating. Fig. 3.1 shows the
annually averaged temperature change in the control run due to the inclusion of CO2 heating. An increase in temperature of several degrees is seen to occur throughout most of the stratosphere with the magnitude increasing with height. Superimposed on this is the region where the statistical significance is estimated to be > 99%. Note that this covers the tropical and most of the middle-latitude stratosphere.
3.1.3 Diagnostic Analyses of Surface Solar Flux Measurements
The Baseline Surface Radiation Network (BSRN) data was further analyzed to determine differences between the measured "clear sky" surface flux for Boulder, Colorado and calculated values from a solar radiative transfer model. For the model calculations, climatological profiles of temperature and moisture are derived from daily NCEP data. Clear-sky observations are identified by stipulating that the difference in the direct surface transmission between the measurement and the model clear-sky value be less than a critical amount. This criterion is applied for all observations comprising each 30 minute time period. In Fig. 3.2 (top), the diurnal variation of the monthly-averaged total (direct + diffuse) "clear sky" "surface flux derived from the observations is compared with the model value for Boulder in July 1994. Note that the measurements are consistently lower than the model value, with differences exceeding 100 W/m2. These differences further suggest that additional attenuators, likely aerosols, need to be incorporated into the model computations to produce better agreement.
The entire record for Boulder from the BSRN dataset was archived in order to study temporal variations in the derived monthly-averaged "clear sky" surface flux. Fig. 3.2 (bottom) shows a comparison of the monthly-averaged "clear sky" total surface flux derived for all months and years (1992-1997) in the sample. The relatively small interannual variations (< 20 W/m2) give further credence to the derived fluxes being clear-sky, since such fluxes are expected to exhibit relatively little variability. Also shown is the corresponding SKYHI clear-sky surface flux for the closest grid point to Boulder. Note that SKYHI consistently overestimates the clear-sky flux by 20-30 W/m2 during most of the year. This confirms the fact that additional attenuators are needed to produce better agreement between calculated and observed fluxes. Smaller differences between SKYHI and the measurements are noted during the autumn season.
3.1.4 Development of Radiative Parameterizations for GCMs
A version of
the SKYHI GCM with 1.2° x 1.0° longitude-latitude resolution suitable
for use on the GFDL T3E system has been prepared. This version has been
employed in a 1-year control integration to determine the effects of model
resolution on cloud climatology (no).
3.2 CONVECTION-CLOUDS-RADIATION-CLIMATE INTERACTIONS
3.2.1 Cumulus Parameterization
The impact of deep convective systems on the thermodynamic and hydrological behavior of the atmospheric general circulation has been studied using this conceptual framework, embedded in the SKYHI GCM (mn). The mesoscale cloud systems produced by deep convection in SKYHI increase upper-tropospheric water vapor and intensify the Walker circulation. The size distribution of these simulated mesoscale cloud systems is consistent with satellite observations, an important result for modeling cloud-radiative interactions realistically. The mass fluxes associated with deep convective systems, including mesoscale clouds, differ appreciably from those of deep convective systems parameterized without them, with detrainment more concentrated in the middle troposphere when mesoscale circulations are included.
Microphysical and radiative aspects of the mesoscale cloud systems have also been parameterized. Convective systems with mesoscale clouds are found to produce much larger shortwave and longwave cloud forcing than those parameterized without them.
Preliminary development of an extension to the new cumulus parameterization to include transport of chemical tracers has been completed. Based on mass fluxes produced by the parameterization when mesoscale clouds are included, it is likely that less tracer transport to the upper troposphere will occur, correcting a problem of excessive upper-tropospheric tracer concentrations that has been noted in several studies using mass-flux parameterizations without mesoscale clouds.
3.2.2 Limited-Area Non-Hydrostatic Models
The convection model was augmented to include aerosol chemistry and transport (1537, 1543, 1589), and used to investigate the effect of deep convection on aerosols in the region of the Indian subcontinent, using extensive observations available from the Indian Ocean Experiment (INDOEX). These studies revealed that deep convection is the major mechanism for removing aerosols originating on the Indian subcontinent from the boundary layer as they are advected toward the Equator. The aerosols subsequently modify the microphysical and radiative properties of upper-tropospheric clouds associated with deep convection.
3.2.3 Moist Convective Turbulence
Surface fluxes provide energy to the atmosphere at the ground and radiative fluxes remove energy by cooling the troposphere. As the heating occurs at a warmer temperature than the cooling, entropy is lost and this reduction must be balanced in a statistically steady state by entropy production due to irreversible processes. If the dominant irreversible process is frictional dissipation of kinetic energy, one can estimate the magnitude of the kinetic energy generation, or the mechanical work performed by the atmosphere from the entropy production, as in the simplest heat engine. This is a good approximation for dry convection, but it overestimates the work performed by moist convection by an order of magnitude. In both 2-D and 3-D models of moist convection, it is found that the dominant irreversible sources of entropy are the diffusion of water vapor and the evaporation of condensate into unsaturated air. It can be argued that this is not simply a result within one particular model, but will be a property of any atmosphere in which the dominant mode of vertical energy transport is latent rather than sensible. Coupled with an earlier finding that the work performed by moist convection is primarily used to lift water and not to generate the kinetic energy of the flow (1689), this analysis implies that a theory for CAPE cannot be developed from the entropy and energy budgets alone.
An additional result emerging from this analysis is that the work performed by moist convection is primarily due to the expansion of the water vapor component of the air, despite the fact that the vapor is at most 2% of the atmosphere by weight. When air rises, the dry component expands and performs work, but most of this is cancelled during subsidence, the residual being dependent on the differences in temperature between upward and downward moving parcels. Water vapor also expands and performs work as it moves upward, but much of the vapor condenses, so there is much less cancellation during descent of the work performed during ascent. The implications of this result for our understanding of moist convective turbulence are currently being examined.
3.2.4 Prognostic Cloud Parameterization
Initial climate
simulations with this parameterization revealed significant differences
with observations in the total amount of solar energy absorbed and the
longwave radiation emitted by Earth. Consequently, two parameters of the
cloud parameterization were adjusted until the global mean energy budget
matched observations. The first parameter tuned is the threshold liquid
cloud drop radius for which rain formation begins. The tuned value of the
threshold radius, 7 
m,
is significantly less than the observed values of 10-12 m.
This result is in common with a number of other GCMs using the same auto-conversion
parameterization. The second parameter tuned is the fraction of condensed
water in cumulus updrafts which becomes cloud condensate (the remaining
portion becomes precipitation mass). For updrafts reaching the upper troposphere,
1% of condensed water becomes cloud condensate whereas for updrafts that
only reach the lower troposphere this fraction is 50%. Although these precipitation
efficiencies were tuned in the present application, in future work they
will be predicted from the new cumulus convection parameterization (3.2.1)
when it is coupled with this prognostic cloud parameterization.
With the tuned parameterization, the B-grid dynamical core of the FMS was integrated for 5 years over historically observed sea surface temperatures (SSTs). One field directly simulated by cloud parameterization, the amount of cloud liquid in a column of air per unit area or liquid water path (LWP), can be directly compared with satellite data, but only over oceans. An example of this is shown in Fig. 3.3, which compares the climatological mean liquid water path for the June-July-August season from the model simulation to two satellite derived observations. The cloud parameterization which, in general, has less than observed LWPs, simulates the maximums of midlatitude oceans and tropical convergence zones. However, the magnitude of the eastern tropical Pacific maximum is under-simulated by the model. This is a consequence of a very important model deficiency, namely the lack of marine stratocumulus in the eastern subtropical oceans.
Additional aspects
of the cloud parameterization that can be compared to observations include
the effective radius of liquid clouds, which is diagnosed in the model
from the liquid water mass and an assumed number density of cloud droplets.
Satellite observations suggest values of 9 to 12 m
for this parameter. However, the model's liquid cloud drop effective radii
are diagnosed to be between 6 and 7 m.
This underestimate occurs
because the auto-conversion effectively limits the radii of the clouds
to the threshold radius which is tuned to be 7 m.
This is considered an important deficiency of the cloud parameterization,
as it limits the model's utility for simulating the indirect effect of
aerosols on cloud properties.
Another aspect of the cloud parameterization that can be evaluated is the distribution of optical depths and cloud top pressures. Comparison of the model to satellite data reveals that at all height levels (low, medium, and high), the amount of optically thick cloud is overestimated by the model, whereas the amount of optically thin cloud is underestimated. The overestimate of cloud optical depths explains in part how the climatological radiation budget can be approximately correct with less than observed cloud amount.
If the parameterization is to be used for climate change simulations, it is important that cloud feedbacks be correctly simulated. One observed feedback is that for small increases in temperature, low cloud optical depths increase for cold clouds over land, but decrease for warmer land clouds and all oceanic clouds. Preliminary diagnosis of the temperature feedbacks of the model simulated low clouds indicates that the parameterization qualitatively reproduces this feature, despite the problems of the simulation which include a lack of marine stratocumulus clouds and cloud drop effective radii which are too small.
3.2.4.2 Diagnostic Assessment of the Simulation of Midlatitude Cloudiness
3.3 ATMOSPHERIC CHEMISTRY AND TRANSPORT
In preparation for a future scalable supercomputer system and the incorporation of an on-line chemistry module from NCAR's "MOZART" (Model for Ozone and Related Chemical Tracers) model, the standard GFDL GCTM (Global Chemical Transport Model) was extensively re-coded. All internal input-output structure on the irregular Kurihara grid has been removed and the model now executes in CPU memory. In addition, any number of tracers can now be transported, dependent on memory and CPU speed.
A reduced form of the on-line chemistry module from NCAR's MOZART model is now being tested in a coupled CO, NOx, HNO3, PAN, O3 version of the GCTM. This will serve as the prototype development for the chemistry module that will be incorporated into the GFDL FMS for chemistry-climate studies.
3.3.3 Tropical South Atlantic Ocean Tropospheric Ozone Maximum
Initial speculation assumed that the TCO formation resulted from the emission of ozone precursors by agricultural biomass burning which were then advected from the continents to the ocean. GCTM results have shown that the maximum is produced by transport in the upper troposphere of ozone and reactive nitrogen (NOx) generated over the continents by both lightning and upward convective mixing of biomass burning products, followed by subsidence and chemical destruction of ozone in the boundary layer (1718). The dominance of lightning generated NOx (accounting for 49% of NOx over the SAO versus 36% for biomass burning) indicates a more reduced effect of human influence on tropical ozone pollution than was previously thought. To clarify this, an integration was run with the biomass burning source eliminated from the GCTM ozone photochemical system. Fig. 3.5(b) shows that even with biomass burning removed, there is still a TCO maximum isolated over the SAO. This suggests that a reduced TCO maximum existed in the SAO prior to the advent of agricultural burning, and man's influence results in an amplification of ozone pollution accumulation.
An examination of the TCO versus time (Fig. 3.5(c) from 30°S to 15°N along longitude 1.2°W (dashed line in Fig. 3.5(a)), reveals the seasonality in the SAO region. During the biomass burning dry season (July through October), there is a maximum located from the equator to 15°S, however, a general maximum is present over the SAO throughout the year, varying from less than 35 DU's in MAY to a high near 45 DU's in September. This implies that transport is acting to accumulate ozone in the SAO throughout the year and NOx from continental lightning and smaller sources provides an environment allowing chemical ozone production. An interesting secondary maximum occurs in February during the Southern Hemisphere wet season. This is a result of increased lightning and a smaller increase in biomass burning NOx transported southward from African agricultural burning north of the equator, which is depicted as a pocket of greater than 37.5 DU's at 5° to 10°N.
3.3.4 Asian Impacts on Regional and Global Air Quality
GFDL/GCTM simulations find that, on average, while current Asian emissions already supply more than 20 ppbv of CO to the Northern Hemisphere, significant contributions (> 20 pptv) of NOx are only found over parts of the North Pacific and Indian Oceans. These same emissions also account for a 5-10 ppbv swath of O3 in the boundary layer across the North Pacific, which extends to over half of the Northern Hemisphere in the middle troposphere. A 2020 "business-as-usual" emission scenario predicts that the average impact of Asian emissions on tropospheric O3 will more than double.
Episodic trans-Pacific pollution events greatly exceed the average impact. The strongest Asian CO episodes over North America (NA), occurring most frequently between February and May, are often associated with disturbances that entrain pollution over eastern Asia and amplify over the western Pacific Ocean. With 55 ppb of Asian CO as a criteria for major events, 3-5 Asian pollution events analogous to those observed at Cheeka Peak, WA are expected in the BL all along the U.S. west coast between February and May during a typical year. In contrast to CO, Asia currently has a small impact on the magnitude and variability of background ozone arriving over NA from the west. Direct and indirect Asian contributions to episodic O3 events over the western U.S. are generally in the 3-10 ppbv range. The two largest total O3 events [> 60 ppbv], while having trajectories which pass over Asia, show negligible impact from Asian emissions. However, this may change. A future [~2020] emission scenario in which Asian NOx emissions increase by a factor of 4 from those in 1990 produces late spring ozone episodes at the surface of California with Asian contributions reaching 40 ppb. Such episodic contributions are certain to exacerbate local NA pollution events, especially in elevated areas more frequently exposed to free tropospheric and more heavily Asian-influenced air.
The role of nitric acid deposition in Asia relative to sulfate deposition has been explored with a regional Lagrangian chemical transport model, ATMOS, developed at the University of Iowa. Reactive nitrogen chemistry has been included in the model, with results compared with seasonal and annual deposition measurements. Sensitivity analyses have been conducted to test the model's response to variations in the rate of horizontal dispersion, the simulation of vertical transport, wet and dry deposition rates, chemical conversion rates, and emissions. Simulations from ATMOS are being used to construct a "source-receptor matrix" for the RAINS-Asia model (Regional Air Pollution Information System-Asia), a widely-used integrated assessment model for science and policy studies of regional air pollution in Asia.
3.3.5 GCM Simulation of Carbonaceous Aerosol Distribution
hydrophobic aerosol decreases the column burden of the aerosol. Although the sensitivity of the global mean column burden was less than 25% in the sensitivity tests, the regional effect can be much greater. However, as the parameter range considered for the tests here is somewhat generous, this range is likely an overestimate. In general, the most remote oceanic regions were the most sensitive to variations in the aerosol model parameters. Of the physical factors examined, the intensity and frequency of precipitation events are critical in governing the column burdens. Biases in the frequency of precipitation are likely the single biggest cause of discrepancies between simulation and observations.
The sensitivity of the mean and monthly variability in surface black carbon concentrations to halving of the wet deposition or halving of the transformation time from hydrophobic to hydrophilic state is examined in Fig. 3.7 at four different geographical locations. For reference, the results from the standard case and that from available observations are also illustrated. With respect to the standard case, the monthly mean burdens increase everywhere for a halving of the wet deposition. At Bondville, IL, which is close to sources, the maximum monthly increases are ~9% (occurs in May). At more remote locations, the monthly-mean increases are greater (range of increase at Sable Island is 11-17%, at Mace Head 24-44%, and at Mauna Loa 58-152%). These values may be compared with the global, annual-mean increase of 32%. Thus, the local and monthly sensitivities differ from that in the global, annual-mean. The modeled mean monthly values tend to show varying degrees of agreement with the available observations, with both deficiencies in the
transport and precipitation simulations playing important roles. At
Sable Island and Bondville, the observed variability tends to be much greater
than any of the three simulations (based on 3-year integrations in contrast
to the observations, which at all sites exceed 3 years). Similarly, with
respect to halving of the transformation time, while the global annual-mean
burden is reduced by 25%, the monthly-mean changes in Bondville tend to
be less. The sensitivity increases at the more distant locations, e.g.,
there is a reduction of 20-50% at Sable Island, 40-61% at Mace Head and
3.5-49% at Mauna Loa. While halving the transformation time does tend to
underestimate the surface concentration, it also reduces the overestimate
of the variability in the monthly-means, especially at Mace Head. It is
concluded that the transformation time is likely less than assumed in the
standard simulation, with a value probably between 0.5 and 1 day.
The relative contributions to tropospheric ozone from stratospheric ozone, ozone produced in the relatively clean troposphere, and ozone produced in the polluted boundary layer have been quantified. Global, regional, and local budgets are being constructed; and the mechanisms, both transport and chemical, by which present and future pollution impact tropospheric O3 are being investigated.
GCTM simulations of NOx, CO, and O3, employing present sources of CO and NOx and detailed estimates of future emissions, have been used to examine the present and future regional and global impacts, both average and episodic, of Asian emissions from fossil fuels, biofuels, and biomass burning. Specific goals include: quantifying the export of NOx, CO, and ozone from Asia's polluted BL to the free troposphere; investigating the impact of this export on the balance of ozone production and destruction over the Northern Hemisphere; quantifying the global air quality impacts resulting from the Indonesian fires in 1997; and quantifying the impact of Asian emissions on North America.
In collaboration with Princeton and Rutgers, the FEOM technique for chemical kinetics calculations will be expanded to include comprehensive atmospheric chemistry mechanisms in high-resolution global and regional simulations of tropospheric chemistry.
We will continue development of a reduced form of the on-line chemistry module from NCAR's MOZART model in a coupled CO, NOx, HNO3, PAN, O3 version of the GCTM.
A source-receptor matrix for reactive nitrogen oxides will be developed for the RAINS-Asia (Regional Acidification and Information Simulation - Asia) integrated assessment model, which will then be used to assess issues of future environmental policy in East Asia with a focus on China, Korea, and Japan.
The radiative forcing due to carbonaceous aerosol concentrations generated in the SKYHI GCM will be evaluated.
3.4 ATMOSPHERIC DYNAMICS AND CIRCULATION
Work is underway to convert additional physics modules previously run within the SKYHI model into Fortran 90 modules, so that they may be incorporated into the FMS.
Efforts to create a troposphere-stratosphere-mesosphere model based on the FMS are continuing. Physical and numerical parameterizations and techniques used in the existing SKYHI model, but which are not yet present in the FMS, are being examined to determine those essential features needed to produce an acceptable model climatology.
3.4.2 SKYHI Control Integrations and Basic Model Climatology
Results indicate the cloud climatology in the new integration is substantially more realistic than that of a corresponding simulation using prescribed clouds, although excessive cloudiness is simulated near the surface. The total cloudiness, outgoing longwave irradiances and reflected solar irradiances have been compared to ISCCP (International Satellite Cloud Climatology Project) cloud data and irradiances measured by ERBE. The interannual variability of the outgoing irradiances and the variances of temperature and moisture are greatly increased in the predicted-cloud simulation, despite the continuing constraint of specified climatological SSTs. A comparison between predicted-cloud and prescribed-cloud simulations has been completed (no).
A number of SKYHI control integrations were continued. Particularly noteworthy are a control integration with a 160-level, 1°x1.2° latitude-longitude resolution model that has now continued for 6 months, and another with an 80-level 2°x2.4° version that has run for 20 years. These represent extended model integrations with an unprecedented combination of fine horizontal and vertical resolution.
The results of the high-resolution integrations will be compared with available observations, including the high-resolution radiosonde data (3.4.9).
3.4.3 Spontaneous QBO-like Tropical Wind Oscillations in SKYHI Simulations
3.4.4 Low-Frequency Variability of Simulated Stratospheric Circulation
ratio has some quasi-random year-to-year variability, but there also appears to be an overall trend, with the mixing ratio dropping by ~10% in the first 15 years and largely recovering over the last 10 years. The model has no interannual variation in the specified chemistry and no externally-forced dynamical variations. Thus, the large apparent trends seen in the figure must be caused by the transport effects of spontaneous internal dynamical variability within the model. The variations found in this model simulation are rather similar to those seen in stratospheric methane and water vapor mixing ratios in observations from the Upper Atmosphere Research Satellite (UARS) during the 1990s. The present model results suggest that natural variability may be a plausible explanation for these decadal-scale changes seen in stratospheric composition.
A review paper discussing observational and modelling issues related to interannual variability in the extratropical middle atmospheric circulation was written (lv). Also, a contribution to an extensive review of biennial variability in the middle atmosphere (ls) was completed.
3.4.5 Horizontal Spectra from High-Resolution SKYHI Integrations
Two new experiments were performed with the 0.33°x0.4° model to examine the spectral energy transfers in transient integrations. Each of the integrations started from an initial condition produced by computing a time-average of the control integration over about a day. This resulted in initial conditions that had energy levels in the mesoscale that were strongly suppressed relative to the control result. One integration used the standard version of the model. This produced a simulation in which the mesoscale energy was restored to the control value over a timescale of about 1/2 day, consistent with the earlier energy budget analysis (ma). The other integration used a model with the convective parametrization turned off. In this integration, the mesoscale energy also recovered, but only to within roughly a factor of 2-3 of that seen in the control run. This suggests that the subgrid-scale convection does play an important role in maintaining the mesoscale, but that even in the absence of the parameterized convection, the model would simulate a shallow mesoscale regime.
3.4.6 Parameterized Gravity Wave Drag in the SKYHI Model
3.4.7 GCM Simulations with an Imposed Tropical Quasi-biennial Oscillation
simulation will be compared in detail with the extensive wind observations taken during 1991-1995 by the Doppler radiometer instrument on the UARS.
3.4.8 Observational Study of Gravity Wave Climatology
dominant horizontal direction of propagation in each month of the year as determined at Keflavik and at Macquarie (55°S, 159°E), another high latitude station, but in the Southern Hemisphere. At Macquarie, the dominant wave propagation directions have a strong westward component throughout the year, while at Keflavik the waves propagate eastward in June and July. The very different behaviors seen in the figure at the two stations probably reflect the contrast in the annual cycles of the large-scale mean flow at each location.
3.4.9 Dynamics of the Martian Atmosphere
Thermal tides play a much more prominent role in the martian atmosphere than in the terrestrial atmosphere and an accounting for diurnal variability is an important aspect of interpreting and comparing spacecraft observations with MGCM results. The amplitude and structure of the diurnal temperature variation is strongly dependent on the atmospheric dust opacity and the state of the zonal mean circulation (1709). The tides include westward propagating, sun-synchronous (migrating) waves driven in response to solar heating and additional non-migrating eastward and westward propagating waves resulting from zonal variations in the thermotidal forcing, most notably associated with high amplitude topography. Temperature profiles derived from MGS radio science occultations have revealed large amplitude tropical waves in the lower atmosphere that MGCM simulations suggest are thermal tides that are strongly modulated by topography. An observed large amplitude density variation in the Mars thermosphere (~130 km) has recently been identified as a diurnal period, eastward propagating Kelvin wave excited by the scattering of the migrating tide off the wavenumber 2 component of topography. Such a wave is predicted to extend from the surface to thermosphere heights with little variation in phase. This result is consistent with Viking lander surface pressure data and TES lower atmosphere temperature data.
3.5 CLIMATIC EFFECTS DUE TO ATMOSPHERIC SPECIES
3.5.1 Lower Stratospheric Ozone and Temperature Trends
3.5.2 Radiative Forcing Due to Ozone
3.5.3 Radiative Effects of Aerosol-Cloud Interactions
The aerosol-induced changes in cloud single-scattering albedo (ssa) predicted using the above modeling framework have been incorporated into GFDL's radiative-convective model (using the Fickian diffusion scheme) to investigate the effects of solar cloud absorption on the surface temperature and lapse rate. When low clouds (between 890 and 660 mb) are perturbed, the surface and atmosphere remain coupled. In this case, as solar cloud absorption increases (as the ssa decreases), the surface temperature increases, with the lapse rate becoming less steep. However, no temperature inversion forms because the maximum in solar heating from the low clouds lies at or below the 3 km maximum in atmospheric emission, and the greenhouse effect prevents the formation of an inversion. The same is true if the low cloud optical depth is increased 10 times to a value of ~45. In this case, the increase in optical depth causes an overall reduction in atmospheric and surface temperature (by 20K) and a slight stabilization of the atmosphere, but no temperature inversion forms despite the increase in solar heating. When all clouds are perturbed simultaneously, distributing the perturbation in solar absorption to higher altitudes, the surface and atmosphere remain coupled down to a cloud ssa of 0.6. Only at this low value, which is unrealistic even for polluted clouds (where the minimum expected ssa is ~0.99), the atmosphere and surface decouple and a very low-lying temperature inversion forms. Here, much of the incident solar radiation is absorbed at a higher altitude and the greenhouse effect is too weak to compensate for the surface cooling.
It is found that the changes in surface temperature and lapse rate are diminished when humidity feedback is turned off. The changes in surface temperature and TOA forcing from increases in solar absorption differ considerably from those due to a doubling of CO2 or a 4% increase in the solar constant.
3.5.4 Radiative Effects Due to Pinatubo Stratospheric Aerosols
3.5.4.1 Experiments Using the SKYHI GCM
3.5.4.2 Coupled Climate GCM Simulations of Mt. Pinatubo Effects
The magnitude and evolution of shortwave and longwave radiative anomalies in the GCM exhibit excellent agreement with the ERBE observed anomalies. Both the model and observations show peak anomalies of ~6 W/m2 over the tropics for ~6 months, which then slowly decay and drift towards higher latitudes. The similarity between the observed and model-simulated anomalies suggests that, at least for the case of Mt. Pinatubo, the direct radiative effect of the aerosols dominates over any indirect effects. Some regions and periods do show residual differences of 1-2 W/m2 between the model and observations which may be attributable to indirect effects or natural variability in cloud cover. Discrepancies may also arise due to uncertainties in the prescribed aerosol optical properties (which are estimated to be ~15%). Nevertheless, even with these uncertainties, the comparison of observations and model simulations with Mt. Pinatubo aerosols offers a useful global experiment for bounding the uncertainty of indirect radiative forcing by aerosols in the upper troposphere.
3.5.5 Radiative Forcing Due to Changes in Stratospheric Ozone
Results indicate that ozone decreases and greenhouse gas increases both play major roles in determining the overall pattern of stratospheric temperature change during the last two decades. Fig. 3.11 displays the annual-mean, zonal-mean temperature changes between the two perturbation simulations and the control simulation with climatological ozone concentrations and 1980 greenhouse gas concentrations. Regions with statistical significance at the 99% level are also shown. Substantial (4-6K) temperature decreases, primarily due to the greenhouse gas increases, are predicted for pressures near 1 hPa. Temperature decreases of ~2-3K (which are statistically significant in the Northern Hemisphere autumn and winter) are found in the Antarctic lower stratosphere (20-100 hPa), mostly due to the observed ozone loss. The tropical (30°N-30°S) region between 5-10 hPa shows a small temperature decrease, with substantial cancellation between positive temperature change due to ozone and negative temperature change due to the greenhouse gases. The annual-mean, zonal-mean pattern of temperature change is similar to observed stratospheric temperature trends (1631); the simulation with the greenhouse gas and ozone perturbations appears closer to the observations in tropical latitudes. Both simulations give substantially different results from the previous simulation (A93/P94; 1394) in which ozone change was restricted to the region below ~30 km.
*Portions of this document contain material that has not yet been formally published and may not be quoted or referenced without explicit permission of the author(s).