U.S. Dept. of Commerce / NOAA / OAR / ERL / GFDL *Disclaimer  

 

2. ATMOSPHERIC PROCESSES

GOALS

  2.1 RADIATIVE TRANSFER

ACTIVITIES FY99

     2.1.1 Solar Benchmark Computations

          Benchmark computations were carried out as part of an intercomparison study led by Rangasayi Halthore at the Brookhaven National Laboratory. The downward surface direct and diffuse fluxes, and the upward top-of-the-atmosphere (TOA) solar fluxes, were determined for a clear sky atmosphere with and without aerosols, as well as for an overcast atmosphere. A summary of results for the study are currently available for the clear sky cases only. The mean total spectrum values presented are generally within 2% of the GFDL benchmark results.

           A joint project was initiated with Phil Partain of Colorado State University, to carry out Line-by-Line + Monte Carlo (LBL+Monte Carlo) simulations with different 3-D cloud configurations. The goal is to perform new benchmark calculations for more realistic cloudy atmospheres.

     2.1.2 Shortwave Parameterizations

          A comprehensive description of the new shortwave parameterization has been completed (jg). Fig. 2.1 summarizes the errors that arise in overcast atmospheres. Four quantities are considered: a) the absorbed flux in the cloud; b) the absorbed flux in the atmosphere; c) the downward flux at the surface; and d) the upward flux at the TOA. Three cloud heights are considered: low (800-900 mb) (drops), middle (500-600 mb) (drops), and high (180-200 mb) (ice), with optical depths ranging from 1 to 100. Clouds containing drops assume the "CS" and "CL" size distributions used in the ICRCCM (Intercomparison of Radiation Codes in Climate Models) study, corresponding to small and very large sizes, respectively. Clouds containing ice crystals utilize the "Fu" formulation to derive the scattering properties. The error in the absorbed flux is < 10% for the water cloud cases, but can be larger for the ice cloud cases. For the remaining three quantities shown, the error is < 5%. The parameterization meets the objective of accurately representing the reference solar flux disposition in overcast atmospheres for a variety of cases.

          The stratospheric temperature changes associated with the new shortwave parameterization considering all atmospheric gases were presented previously (A98/P99). Current efforts are focusing on the effect of the improved parameterization for each gas on the simulated stratospheric temperature. Four different atmospheric gaseous profiles have been considered: a) O3 only; b) CO2 + O3; c) CO2 + H2O + O3; and d) CO2 + H2O + O3 + O2 (the full clear-sky case). Separate GCM simulations have been performed for each of these situations utilizing the new and old solar parameterizations and the temperature change due to the new formulation(s) has been determined. Present-day gas concentrations have been used. The temperature change associated with the improvements in the parameterized solar heating rate are shown in Fig. 2.2 for January climatological conditions. Note that in each of the four calculations, the longwave radiative transfer algorithm considers all the gases to be present. Of primary interest are the temperature changes occurring in the tropical and subtropical regions (30°N-30°S).

          An increase in the heating rate in the O3-only case is observed with the new parameterization, but the magnitude of the temperature change is generally less than 2 K. When CO2 is included, an additional warming is observed, increasing with height from 1-2 K in the lower and middle stratosphere to several degrees in the upper stratosphere, in response to the significant additional solar heating associated with the new formulation. Improved accounting of stratospheric heating due to H2O vapor and O2 results in additional warming in the lower and middle stratosphere. However, it is the effect of the improved representation of CO2 on stratospheric heating which is most striking.

          SKYHI's implementation of the new shortwave parameterization has been modified to incorporate snow and rain water in order to more completely characterize the single scattering properties of solar radiative transfer in cloudy, precipitating atmospheres.

     2.1.3 Analysis of Solar Flux Measurements

          Ground-based data provided by the Bondine Surface Radiation Network (BSRN) project has been utilized to compare the measured direct downward solar flux at the surface with corresponding clear sky results obtained by the benchmark solar radiative transfer model. The model's clear-sky value is derived assuming a McClatchey midlatitude summer atmosphere.

          The model's clear sky values consistently overestimate the measured flux. The cases with the largest underestimates are due to the presence of thick clouds in the observations. The maximum measured values, which represent the real clear sky conditions, underestimate the modeled clear-sky value by up to 30-40 W/m2 at high sun angles. This indicates a missing attenuation in the model calculations. Some of this could be explained if the actual vapor profile were more moist than the climatological one. However, even assuming a very moist

profile, differences of at least 20 W/m2 remain. It is unlikely that any gaseous species constitute the missing attenuator. Instead, these deviations strongly suggest aerosols as the missing species in the clear-sky model calculations.

     2.1.4 Clear-Sky Shortwave Radiative Flux

          The investigation of the influence of tropospheric aerosols upon the clear-sky shortwave flux reflected at TOA has continued, using the ERBE observations and the new GFDL shortwave radiation parameterization (1593). The difference between a no-aerosol radiation calculation and the observed reflected flux reveals distinct geographical biases (Fig. 2.3(a)) which are principally attributable to tropospheric aerosols. A sequence of calculations and analyses has been performed to compare the degree to which different types of aerosols contribute to the observed reflected flux. For this purpose, we employ aerosol spatial distributions derived from 3-D chemistry-transport models which include natural and anthropogenic sulfate, natural and anthropogenic dust, and black and organic carbon species. Inclusion of non-sea salt aerosols

reduces the bias considerably in the Northern Hemisphere (NH), indicating the important roles played by these aerosols in the radiative energy balance. Note, however, that these aerosols do not play as big a role in the Southern Hemisphere (SH).

          In addition, sea salt aerosol burdens have been considered, obtained by applying a wind-speed-dependent parameterization to the SSM/I satellite-derived surface windspeeds (Fig. 2.3(a)). Because of the considerable uncertainty concerning sea salt aerosol concentrations, two different relationships are used which yield a "low" and a "high" sea-salt burden. The inclusion of the "low" sea salt concentration (Fig. 2.3(c)) reduces the SH biases significantly, illustrating the marked influence of sea salt even when assumed to be present in small concentrations. If a "high" sea salt burden is used (Fig. 2.3(d)), the SH biases are reduced even further, but the NH reflection substantially exceeds observed values. Finally, if it is assumed that the sulfates over the oceans are subsumed into sea salt, as some recent field observations suggest, then the radiative effects of sulfates over the oceans are suppressed. In this case, Fig. 2.3(e) shows that the radiative biases everywhere are reduced substantially, and that there is an even better agreement with observations.

          There are uncertainties in the estimates attributed to aerosols owing to poor knowledge of several factors, e.g., Fresnel reflection, albedo of whitecaps, phytoplankton, satellite retrieval algorithms, etc. However, it appears unlikely that any of these mechanisms can explain the geographical pattern of biases seen in Fig. 2.3(a) to the extent that the current knowledge of aerosol spatial distributions do. It is interesting to note that the global, annual-mean radiative forcing due to the aerosols estimated here (-6.5 W/m2) is about 40% of that estimated for clouds (-17 W/m2). Uncertainty about the modeled global aerosol distributions makes it difficult to select any particular aerosol assumption as being the most realistic. Nonetheless, all of the assumptions reduce the biases with respect to the ERBE observations by at least one-half in magnitude. The analyses strongly indicate that there is a significant contribution to the reflection from both natural and anthropogenic aerosols, and that there is a substantial geographical signature due to aerosols in the observed radiation budget. While the analyses here cannot unambiguously partition the contributions from the natural and anthropogenic aerosols to the observed radiation budget, it places an upper bound on the effects due to anthropogenic particulates.

     2.1.5 Infrared Parameterization

          The infrared radiative transfer code (1597) has been modified to include the effects of absorption due to aerosols. Work has also continued on improvements to the treatment of cloud infrared effects. Currently, GFDL radiation algorithms specify frequency-independent (gray) cloud properties (infrared emissivity and shortwave reflectivity and absorptivity). To enable the infrared radiative algorithm to use cloud properties derived from microphysics models such as those discussed in (2.2), the algorithm has been modified to incorporate non-grey emissivities. Parameterizations deriving the emissivities from the concentrations and sizes of cloud species (e.g., rain, snow, cloud drops, ice) have been incorporated into the infrared radiative code.

     2.1.6 Intercomparison of Radiation Codes

          Under the auspices of the GEWEX (Global Energy and Water Experiment) Water Vapor Project (GVaP), an intercomparison of radiation codes used in retrieving upper tropospheric humidity (UTH) from observations in the 6.3m water vapor absorption band was performed. This intercomparison was one part of a coordinated effort within GVaP to assess the ability to monitor the distribution and variations of upper tropospheric moisture from space-borne sensors. A total of 23 different codes, ranging from detailed line-by-line (LBL) models, to coarser resolution narrow-band (NB) models, to highly-parameterized single-band (SB) models participated in the study. Forward calculations were performed using a carefully selected set of temperature and moisture profiles chosen to be representative of a wide range of atmospheric conditions. Calculations by LBL models exhibited the greatest consistency, typically agreeing to within 0.5 K in terms of the equivalent blackbody brightness temperature (Tb). The majority of NB and SB models agreed to within 1 K of the LBL models, although a few older models exhibited systematic Tb biases in excess of 2 K. The discrepancies between various models, their association with differences in model physics (e.g., continuum absorption), and their implications for UTH retrieval and radiance assimilation has also been summarized (kd). Since the 6.3 m band is heavily used for satellite remote sensing, this intercomparison provided a common reference for international users of these codes to understand how their satellite retrieval or data assimilation product is affected by their choice of radiation scheme.

PLANS FY00

          In collaboration with Phil Partain of Colorado State University, new benchmark calculations will be performed for more realistic cloudy atmospheres. The BSRN data will be analyzed further to examine differences between measured and model values of the surface solar flux, and also to compare with temporal variations produced by GCMs. As part of the development effort of the Flexible Modeling System (FMS), the changes in the shortwave and infrared physical parameterizations discussed above will be incorporated into the radiative algorithm for the FMS. Further computations of the effect of aerosols on TOA and surface fluxes will be performed, and the quantitative roles of natural and anthropogenic aerosols will be analyzed.

  2.2 CONVECTION-CLOUDS-RADIATION-CLIMATE INTERACTIONS

     2.2.1 Cumulus Parameterization

ACTIVITIES FY99

          Despite major field and conceptual studies in recent years, understanding the elusive relationships between deep convection and radiation remains a challenge for climate studies. A new conceptual framework for cumulus parameterization designed to address this issue has been documented (1133). The frequently used concept of mass-flux cumulus parameterizations was extended to include a statistical treatment of the vertical velocities in cumulus convection. These vertical velocities can be used to drive microphysics at physically appropriate scales, and thereby provide representations of the interactions between cumulus-scale clouds and larger-scale, radiatively important cloud systems (1349). The ultimate goal is to construct a parameterization for deep convection, including associated mesoscale upper-tropospheric cloud systems, that will contribute to realistic modeling in coupled atmosphere-ocean models of both transient variability and radiative fluxes from the surface through TOA.

          This new parameterization has been incorporated in the SKYHI GCM and the resulting distribution of heat and moisture sources has been analyzed, with special attention given to mesoscale aspects. The mesoscale components of deep convective systems were found to contribute significantly to the overall intensity of these systems, in agreement with field observations. An illustration is provided by Fig. 2.4 and Fig. 2.5, which show the mass fluxes in

updrafts associated with cumulus cells and mesoscale circulations, respectively. Note that the mesoscale fluxes are a significant fraction of the total fluxes associated with the cumulus cells. The mass fluxes associated with convective systems, including mesoscale components, detrain much more extensively in the middle troposphere than do convective systems without mesoscale components. The temperature and humidity fields in SKYHI's tropical atmosphere are more realistic with this parameterization, falling within 1°C of observations in both the lower and upper troposphere. Cloud-radiative interactions differ appreciably when mesoscale effects are treated.

PLANS FY00

          Radiative aspects of the new cumulus parameterization will be considered in detail. Interaction with the microphysics of larger-scale cloud systems will be studied by linking to a prognostic microphysics parameterization for large-scale cloud systems. Tracer transport associated with deep convection will be evaluated. The mass fluxes associated with this parameterization differ considerably from those associated with parameterizations lacking mesoscale components and may transport tracers more realistically. Cumulus friction (which represents the effects of deep convection on horizontal momentum) will be added to the parameterization.

     2.2.2 Limited-Area Non-Hydrostatic Models

ACTIVITIES FY99

          Basic analysis of three-dimensional integrations of deep convection and associated mesoscale circulations by the LImited Area Nonhydrostatic Model (LAN) has been completed (1613). Among the issues considered were the structure of tropical convective systems, heat and moisture sources associated with tropical convective systems, and transport and transformation of sulfate by tropical convective systems (1537, 1589). The limited-area non-hydrostatic model was also used to study the relationships between sulfate aerosols and basic microphysical processes, providing new understanding of the basic physical processes underlying indirect aerosol effects (ju).

          Microphysics and radiation in the LAN model were studied by modeling deep convection associated with westerly wind bursts in the tropical western Pacific Ocean. These convective events occurred during TOGA-COARE (Tropical Ocean Global Atmosphere-Coupled Ocean Atmosphere Response Experiment) and the observations were well-characterized with respect to both dynamics and radiation balances at TOA and at the ocean surface. These observations were used to develop new treatments for ice microphysics and its interactions with radiation. Experiments with a new formulation of lateral boundary conditions were also completed. The cloud-system model is now able to reproduce radiative energy balances observed at both the surface and TOA. Figs. 2.6 and 2.7 show surface and TOA radiative fluxes corresponding to various treatments of microphysics, radiation, and boundary conditions. In addition, longwave and shortwave radiation codes were updated.

PLANS FY00

          The LAN model, coupled with sulfate aerosol chemistry, will be used to study the INDOEX (Indian Ocean Experiment) region. Preliminary indications of substantial sulfate transport from the Indian sub-continent to the tropics suggest that significant vertical transport and transformation occurs in that region. Indirect effects of aerosols on upper-tropospheric cloud systems will be studied.

          The cloud-system model will continue to be evaluated using observations from major field programs. The ARM (Atmospheric Radiation Measurements) program will be used to model midlatitude continental convection. Using the microphysics and radiation improvements completed previously (A98/P99), the model will also be used to evaluate the new cumulus parameterization (2.2.1).

     2.2.3 Single Column Modeling Tests of Physical Parameterizations

ACTIVITIES FY99

          In recent years, single column versions of GCMs, known as Single Column Models (SCMs), have become very important tools for evaluation of the physical parameterizations used in GCMs. SCMs are models where the column physics is extracted from the parent GCM and the effects of atmospheric dynamics are specified from observations. It is hoped that errors in parameterization of physical processes such clouds, convection, and turbulence can be identified through careful comparison of SCM results to observations and other finer scale models such as cloud-system models.

          A new SCM has been built using the code included in the FMS (3.1). This SCM now includes not only the column physics of the atmosphere, but also the new land-surface model (1.6.3) and the thermodynamic part of the new sea-ice model (4.1.4). The flexibility of this new SCM should prove to be beneficial in a wide variety of studies.

          The GFDL SCM has been compared to other SCMs through continuing participation in the SCM Intercomparison Group of the ARM program of the Department of Energy (ks). The results indicate that all SCMs produce larger errors than does the cloud-system model which also participated in the comparison. This suggests that the cloud-system model, which simulates some quantities that cannot be easily observed, can be used as a tool to diagnose errors in the parameterizations of GCMs.

PLANS FY00

          Participation in the SCM intercomparison program of ARM will continue. The participants in this program will examine a new case from ARM, which includes substantially enhanced observations, including vertical profile of cloud obtained from the millimeter wavelength cloud radar.

     2.2.4 Subgrid-Scale Variability in Cloud Processes

ACTIVITIES FY99

          Because cloud processes occur on much smaller scales than typically resolved by GCMs, the results of nonlinear processes must be parameterized as a function of the resolved grid-scale variables. Processes which are nonlinear functions of local quantities have the property that to accurately know the grid-scale mean of a process, one must know (i.e., parameterize) the distribution of local quantities.

          One such important process is cloud albedo. Because it is a nonlinear convex function of cloud optical depth, the albedo of the mean cloud optical depth is always greater than the mean of the albedos of the optical depth distribution. (This is known as the "plane-parallel albedo bias".) Thus, if a GCM knows the mean cloud optical depth in a grid-cell, the albedo calculated from it will always be greater than the true albedo of the grid-cell. This bias has been quantified in a study of satellite observations (hx). At horizontal scales typical of GCMs, this study indicated that the plane-parallel albedo bias is significantly smaller than prior estimates, at least for marine stratocumulus clouds. By stratifying satellite observations of the albedo bias by surface observed cloud type, it was shown that cumulus cloud fields have higher variability in cloud optical depth relative to their mean than stratiform clouds. This suggests that the albedo bias in a GCM may in part be parameterized by the predominant cloud type simulated by the GCM in a given model cell.

          Another important nonlinear processes is precipitation. The sensitivity of the parameterized large-scale precipitation to vertical variations of cloud fraction has been examined (gx). The results indicate that the effects of "cloud macrophysics" - how much of stratiform precipitation falls into the clear portions of lower levels where it may evaporate versus falling into the cloudy portions of lower levels where more accretion would occur - has a sizeable impact on the calculated rate of conversion of cloud water to precipitation. As a result of this work, a new parameterization of stratiform precipitation in GCMs was developed which resolves the precipitation flux falling through the model into two streams: the precipitation falling within cloud and the precipitation falling outside of cloud. Such a division leads to a marked improvement of the physical basis of cloud microphysical parameterization in GCMs (ks).

PLANS FY00

          Work will begin on the development of a unified model for subgrid-scale variability of clouds which will allow for a consistent prediction of the nonlinear impacts of clouds on both the mean albedo of a grid cell, as well as the mean microphysical rates of conversion of cloud water to precipitation. Such a unified treatment is needed to improve the physical basis of cloud parameterizations in GCMs.

     2.2.5 Development of Physical Parameterizations for GCMs

ACTIVITIES FY99

          A prognostic cloud parameterization has been prepared for the FMS (3.1). This parameterization couples the treatment of cloud microphysics as it is applied in the Commonwealth Scientific and Industrial Research Organization (CSIRO) GCM with the treatment of cloud macrophysics (i.e., cloud fraction and its effects) as it is applied in the ECMWF GCM. Convective schemes have been modified to provide sources of cloud condensate from the detrainment of cumulus updrafts. This parameterization is undergoing extensive testing and refinement in the context of the B-Grid dynamical core.

          In addition, a vertical diffusion parameterization based upon those used in the NCAR and ECMWF models has been prepared and tested in the FMS. This parameterization consists of a K-profile scheme, where the vertical profile of vertical diffusivity is specified as a function of the surface wind stress and buoyancy. Such schemes are now widely used in GCMs because it is recognized that much transport and mixing is not a sole function of local conditions (e.g., the overshoot of large-eddies in a convective boundary layer and its resulting mixing with air above).

PLANS FY00

          The performance of the prognostic cloud scheme will be tuned in the context of the B-Grid dynamical core. The results of the prognostic cloud parameterization will be documented and the sensitivities of the parameterized clouds to plausible variations in arbitrary parameters will be examined.

  2.3 ATMOSPHERIC CHEMISTRY AND TRANSPORT

ACTIVITIES FY99

     2.3.1 Fast Photochemical Solver Development

          A major portion of the computational effort in simulations by 3-D chemistry-transport models is consumed in chemical kinetics calculations, which repeatedly solve coupled ordinary differential equations. To address this burden, a high-speed fully equivalent operational model (FEOM) for chemical kinetics calculations has been developed. The FEOM consists of a hierarchical correlated-function expansion capturing the input-output relationships of chemical kinetics. As an initial test, this work develops FEOMs for the CO-CH4-NOy-H2O chemistry to generate the time-dependent chemical ozone production and destruction rates for a global chemistry-transport model (GCTM) simulation of tropospheric ozone. The FEOMs are constructed for all GCTM model levels, all 12 months of the year, every 10 degrees of latitude, for two types of surface albedo, and for the full range of tropospheric values of H2O, CO, NOx, and O3. The global ozone fields simulated with the FEOMs in the GCTM are at least as accurate and in some cases better than those obtained using traditional four-dimensional look-up tables (kj). While the computational burdens are comparable for this simple case, a standard n-dimensional interpolation becomes prohibitive, both off-line during the table construction and during the actual simulation, for complex non-methane hydrocarbon chemistry.

     2.3.2 Boundary Layer Transport

          A non-local scheme for vertical diffusion in the atmospheric boundary layer has been implemented in the N30 SKYHI model, following that used in the NCAR climate system model (CSM). This non-local scheme transports moisture and tracers from the surface more rapidly than the local scheme under unstable conditions (in the presence of surface heating or cold air mass passing over a warmer surface). Overall, more moisture is transported to the upper troposphere by the non-local scheme than by the local scheme in SKYHI simulations, and it improves model temperature and humidity in the upper troposphere. One conclusion from this is that one must consider possible sensitivity to the parameterization of vertical diffusion.

     2.3.3 Biomass Burning

          Biomass burning on a wide scale causes significant regional pollution, often with deleterious impacts on the health and safety of the local population. On a global scale, biomass burning may have significant impacts on atmospheric chemistry and global climate. Loading of the atmosphere with nitrogen oxides (NOx = NO+NO2), carbon monoxide (CO), black and organic carbon, mineral ash, and volatile organic compounds, in addition to greenhouse gases such as nitrous oxide (N2O), carbon dioxide (CO2), and methane (CH4), contributes to air pollution, global warming, and acid rain. The oxidation of CH4, CO, and other hydrocarbons in a NOx-enriched environment leads to the production of tropospheric ozone (O3), thus modifying the reactivity of the atmosphere. Though lightning is known to be a natural contributor to biomass fires, today most fires are the result of deliberate human fire-management practices, particularly in the tropical and subtropical regions. With a growing population, the demand for land use is increasing, and the challenge of assessing the role of biomass burning in atmospheric chemistry, climate, and terrestrial ecology is becoming increasingly important.

          The GFDL GCTM was used to quantify the impacts of biomass burning on tropospheric concentrations of CO, NOx, and O3. Updated global sources of biomass burning, which were constructed for use with the GCTM, emit 748 Tg CO/yr and 7.8 Tg N/yr in the surface layer of the model. Both sources include six types of biomass: forest, savanna, fuelwood, agricultural residues, domestic crop residues (burned in the home for cooking and/or heating), and dried animal waste. A temporal distribution for the burning of forest, savanna, and agricultural residues is based upon regional cultural use of fire, vegetation type, local climate, and information gathered from satellite observations, while emissions from the burning of fuelwood, domestic crop residues, and dried animal waste are held constant throughout the year. Based on agreement with observations, particularly of CO, we conclude that the collective uncertainty in the biomass burning sources is much less than the factor of two suggested by previous estimates of biomass burned in the tropics annually.

          Based on model simulations, biomass burning is inferred to be a major source of CO and NOx in the northern high latitudes during the summer and fall, and in the tropics throughout most of the year. Fig. 2.8 shows the percentage impact of biomass burning on CO, NOx, and O3 near the surface and at the 500 mb level during June, July, and August. Although biomass burning contributes more than 50% of both the CO and NOx in the boundary layer over major source regions, it has a much larger global impact on the CO distribution in comparison to either NOx or O3, contributing 15 to 30% of the entire tropospheric CO background. The only significant biomass burning contribution to NOx at 500 mb (due to the short lifetime of NOx in the lower troposphere) is a plume occurring in the SH subtropical free troposphere, stretching from South America to the western Pacific. The largest impacts on O3 are limited to those regions where NOx impacts are large as well. Near the surface, biomass burning contributes up to 45% of the total O3 concentrations over major tropical source regions, at least 15% throughout the year in the tropics, and 10% to 20% throughout the SH

during September through November. At 500 mb, the largest contribution to O3 (20-30%) is found in correlation with the NOx plume during July through November.

     2.3.4 Tropospheric Carbon Monoxide

          The GFDL GCTM was used to examine the evolution and distribution of carbon monoxide (CO) (kn). The objective of the work was to gain an improved understanding of the global carbon monoxide budget and to create a realistic CO distribution for future atmospheric chemistry studies. The sum of all CO sources in the model is 2.5 Pg CO/yr, including fossil fuel use (300 Tg CO /year), biomass burning (748 Tg CO/year), oxidation of biogenic hydrocarbons (683 Tg CO/year), and methane oxidation (758 Tg CO/year). The single sink for CO is destruction by the hydroxyl radical (OH). Although the hydroxyl field is an important determinant of the global CO distribution, its sensitivity to this chemical sink varies spatially, with a 30% increase in OH yielding decreases in CO ranging from 4-23%. Using an OH distribution which corresponds to a methyl chloroform lifetime of 4.8 years, the resultant CO concentrations closely agree with the NOAA/Climate Modeling and Diagnostics Laboratory (CMDL) ground-based measurements (93% of seasonal-average data points agree within ±25%) and flight data from measurement campaigns of the NASA Global Tropospheric Experiment (79% of regional-average data points agree within ±25%).

          Figure 2.9 shows selected time series comparing model results (black squares) with available measurements from NOAA/CMDL global cooperative flask sampling network (brown circles), as well as the contribution of CO from each of the four emission sources individually (red for fossil fuel burning, green for biomass burning, blue for biogenic hydrocarbon oxidation, yellow for methane oxidation). All four sources play a significant role in the magnitude of CO in all regions, and CO's spatial and seasonal variability is a complex interaction among seasonal and spatial variations in CO sources, OH destruction, and atmospheric transport. However, seasonality away from surface source regions is generally governed by CO from fossil fuel consumption and biomass burning, the lifetime of CO, which varies from ~2 weeks in the tropics and over summer continental regions to well over a year at the winter poles, accounts for much of the pole to equator gradient in the NH, and the interhemispheric gradient primarily results from the midlatitude NH location of the fossil fuel source.

     2.3.5 Simulation of the Observed Tropospheric Column Ozone Maximum
              in the Tropical South Atlantic Ocean

          In the early 1990s, a new satellite retrieval technique enabled scientists for the first time to examine horizontal gradients of tropospheric column ozone (TCO) over the global tropics and subtropics. Subsequent analysis detected a large TCO maximum over the tropical South Atlantic Ocean (SAO) during the austral spring concurrent with the biomass burning season. This was a surprising discovery, in that previous speculation assumed the largest ozone enhancements would occur over the continents of South America and Africa, associated with ozone production induced by the emission of precursors (nitrogen oxides) from biomass fires. As a result, an intensive field mission, "Transport and Atmospheric Chemistry near the Equator-Atlantic" (TRACE A), was conducted in 1996. While analyses of the observations suggested that biomass burning was the primary driver of tropical "smog", the burning, by itself, was not sufficient to generate the TCO maximum found far from trace gas sources. It was speculated that meteorological transport, lightning generation of NOx, and upper tropospheric photochemical generation of ozone also played a role.

           The GFDL GCTM has successfully simulated the TCO over the tropical SAO. As seen in Fig. 2.10, an ozone column maximum greater than 40 Dobson units (shown in orange, red, and brown) has been produced off the west coast of Africa which agrees quite well in magnitude and location with the latest analysis of satellite data. Also, the large values

extending eastward from Africa to Australia are qualitatively represented. In addition, the model compares favorably with available ozonesonde data and climatological flow fields. Unlike even an intensive field study such as TRACE A, which has data gaps in space and time, a model provides continuous data at all grid points, making possible a quantitative analysis of the interaction of transport and chemistry.

          An analysis of the model's meteorology has shown that the subtropics and tropics of the SAO exhibit a unique circulation regime which is quasi-stationary in space and time. The basic flow features a surface anticyclone anchored off the African coast near 30°S which produces surface southeast trade winds from the west coast of Africa to the Intertropical Convergence Zone (ITCZ) north of the equator. This divergent anticylonic surface outflow is supported by strong subsidence which can transport air from the upper troposphere to the

boundary layer in 4 to 5 days. Also, the region undergoes a Walker type circulation with rising motion over the continents of South America and Africa, and convergent flow in the upper troposphere over the South Atlantic leading to subsequent sinking motion.

          In order to examine the role of transport and chemistry, an extensive analysis was performed over an area encompassing the TCO maximum (depicted as the rectangle in Fig. 2.10). This area is entirely over the ocean, therefore precluding any direct land based NOx source. It was found that two NOx sources accounted for nearly 90% of the tropospheric reactive nitrogen in the volume, with lightning and biomass burning contributing 50% and 39%, respectively. The large contribution from lightning, an upper tropospheric source, implied significant NOx transport to the region since there is no lightning over the arid SAO. The amount and vertical distribution of NOx directly regulates the amount of ozone production or destruction in the volume, with net production occurring above 500 mb and net loss below. Fig. 2.11 shows the affect on the ozone profile by independently removing the lightning and biomass NOx sources, thereby altering the photochemistry. Lightning controls the upper troposphere and biomass the lower with a cross-over near 685 mb, although both have an influence throughout. This implies an effective transport mechanism to mix continental biomass NOx aloft and out over the ocean, as well as prevailing sinking motion to mix lightning NOx downward.

          The complex roles of transport and chemistry in the distribution of ozone are interdependent in that transport operates on tracer gradients which are constantly being changed by photochemistry. To unravel individual contributions, ozone budgets within both the upper troposphere (UT: 150-606 mb) and lower troposphere (LT: 606 mb-sfc) were calculated for the month of September when the total net ozone tendency of the entire volume was essentially zero. In general, the mass in the UT increased during the month, while being balanced by an equal LT decrease. In the UT, net chemical production dominated transport evacuation, while transport accumulation in the LT was offset by the larger sum of net chemical destruction and surface ozone deposition. An examination of the horizontal and

vertical transport fluxes revealed that horizontal flux convergence in the UT was dominated by a vertical flux of ozone into the LT, which in turn was larger than the lower level horizontal flux divergence.

          Overall, the unique meteorology of this region allows continental ozone and NOx generated by biomass burning and lightning to converge in the UT over the SAO, providing an environment for net photochemical production of additional ozone. In turn, strong subsidence transports the ozone to the LT, where it is removed by net photochemical destruction, deposition, and horizontal fluxes.

     2.3.6 Asian Impacts on Regional and Global Air Quality

          In the wake of new, more stringent ground level ozone standards from the EPA (8-hour average of 80 ppb ozone, not to be exceeded three times per year), there is growing interest in understanding the origin of background ozone that arrives from outside the U.S. As the gap between "non-compliance" and background levels of ozone narrows (surface background ozone over the Pacific in the GCTM typically varies from between 20-50 ppb), the significance of background ozone as a potential factor for non-compliance increases.

          The GFDL/GCTM has been used to study the origin of background ozone, with a particular emphasis on upwind emissions from Asia. Asian emissions of the ozone precursor NOx are expected to grow two or three fold in the next several decades, while emissions from Europe and North America are expected to stabilize. The difference between two global simulations, one with full global emissions and the other lacking Asian emissions, gives an approximation for the "Asian contribution" to ozone. Fig. 2.12 shows the time series of the Asian contribution to ambient ozone arriving over central California for 1990 and projected 2020 NOx emission levels. Both the mid-troposphere (500 mb) and boundary layer (940 mb) time series show a May maximum of Asian pollution over California (time series are 6-hour samples from

the GCTM). Surface ozone contributions in 1990 are a modest 5-6 ppb ozone during late spring and summer but jump to 10-20 ppb by the year 2020. This may be enough to push some areas in the U.S. over the threshold of non-compliance. The Asian signal is much stronger aloft, and it is typically subsidence from this reservoir that maintains the Asian pollution signal in the boundary layer.

          The GFDL/GCTM has been shown to be capable of simulating the frequency, magnitude, and dynamic characteristics of episodic CO transport to North America from Asia as observed at Cheeka Peak Observatory (Washington State). It is currently being used to diagnose the episodic nature of trans-Pacific air pollution transport. Analysis of flow and tracer fields reveal that, after venting from the Asian boundary layer, pollution parcels sometimes remain in highly discrete synoptic structures all the way across the Pacific Ocean. Because current ozone standards are based upon a limited number (maximum of three per year) of ozone exceedences (periods in which the ozone standards are exceeded), the policy implications of an Asian signal that arrives as a constant background enhancement versus one that is highly episodic may differ significantly. For example, Fig. 2.12 shows that episodes of 30-40 ppb ozone from Asia may be possible in 2020, suggesting the potential for ozone exceedences over a wide area. Blame for these exceedences would be placed unreasonably on local emission sources, because there is no way to distinguish background and local pollutants in regional measurements. Work to understand and evaluate the policy implications of such episodic transport continues.

     2.3.7 GCM Simulation of Carbonaceous Aerosol Distribution

          A model of the carbonaceous aerosol cycle has been implemented in the GFDL SKYHI model. The fossil fuel emissions of carbonaceous aerosol (both black and organic) have been calculated on the GFDL SKYHI GCM N30 (3° X 3.6°) grid. These emissions have been derived from a 1° X 1° database of black and organic carbon emissions1 using a simple linear interpolation scheme.

          The carbonaceous aerosol model that has been implemented is similar to that of Cooke and Wilson [19962]. Briefly, the scheme uses two tracers for each carbonaceous aerosol component. These represent the hydrophobic and hydrophilic parts of the carbonaceous aerosol. The sinks for these aerosols are dry and wet deposition, respectively. In addition, there is also a transformation from hydrophobic to hydrophilic aerosol with a time constant of one day. This can be thought of as a representation of the aging of the aerosol, which causes it to become more hydrophilic with time.

          The model was run for three years and the results were compared to surface air concentrations and wet deposition measurements. Initial comparisons indicate a reasonable agreement between the modeled and measured concentrations of this aerosol species. CMDL operates a continuous light absorption photometer at various sites from which it is possible to infer daily black carbon (BC) concentrations. The measurements of absorption have been converted to surface concentrations using a factor of 9.265 m2 g-1.

          The first site from which absorption measurements are used is at Sable Island, Nova Scotia (60°W, 43.9°N) where measurements have been collected since October 1995. The monthly geometric mean and interannual monthly geometric deviation of the observed and modeled concentrations are shown in Fig. 2.13. For the first five months of the year, the modeled values are approximately half that which are measured. While the agreement is somewhat better during the summertime, the model still underpredicts the concentrations. The best agreement is during the final three months of the year. The disagreement in the first few months may imply that the emissions upwind of Sable Island are too low at that time or that the modeled scavenging of the aerosol is too efficient during that period. Some insight on whether the emissions are too low may be found by looking at the modeled and measured BC concentrations at Bondville, Illinois (88.4°W, 40.1°N). Bondville is located on the western edge of a large emission source region, so, if the emissions are too low, the model will underestimate the concentration of the aerosol at this site. However, if the modeled concentrations are reasonable, this implies that the removal processes for the aerosol in the transit between the source region and Sable Island are too efficient. The monthly mean concentrations for the Bondville site are also shown in Fig. 2.13. In this case the observed and modeled BC are in quite good agreement for the first three months and are underestimated during the summer period to the same degree as at Sable Island. The peak in the measured concentrations in October

may be due to interference from dust from farming practices (personal communication, John Ogren (CMDL)). The reasonable agreement between the model and measured concentrations and the possible interference of absorbing dust suggests that the emission of BC is not seriously underestimated in this region, and that the underestimation of the model in predicting BC concentrations at Sable Island is more likely due to the scavenging scheme.

          Measurements of light absorption are also performed at the free tropospheric site of Mauna Loa, Hawaii. These show an underestimation in the modeled concentrations in the first six months of the year. A possible explanation for the underestimation of the BC concentrations may be found by looking at back-trajectories arriving at Mauna Loa in April. Back-trajectories calculated from the SKYHI winds and obtained from CMDL both show a northwesterly bias. However, a greater fraction of the CMDL trajectories reach the Asian continent than do in the SKYHI case. The implication of this is that the BC which is emitted in Asia has a greater chance to be removed by precipitation in the SKYHI model, as it is transported eastwards at this time of year. There is a secondary peak in the modeled concentrations in August which is not seen in the measurements. Back-trajectory analysis for this month shows that both sets of trajectories have an easterly bias, with the bias being much more pronounced in the SKYHI case. In addition, the easterlies are much stronger in the SKYHI case implying that transport of BC from central America is contributing to the secondary modeled peak in concentrations.

          Figure 2.14 shows the monthly mean global column burden distribution of BC (mg m-2) in January. The column burden is useful in that it can be used to calculate the effective single scatter albedo of the atmospheric column. Also, radiative forcing calculations require the vertical profile of the aerosol. The major source areas (eastern Europe and China) of black carbon can be seen clearly and most of the NH is significantly affected by the emission of this carbonaceous aerosol. The SH is not impacted as severely, although there are areas with reasonable burdens of black carbon in the atmosphere.Various sensitivity studies have been conducted to investigate the effect of varying the hydrophobic fraction of the emissions, the transformation time for hydrophobic aerosol to become hydrophilic, and the wet deposition rate of the aerosol. The most sensitive parameter by far is the wet deposition removal rate of the aerosol. This implies that the precipitation scheme which is applied in this, or any other, model needs to be rigorously verified when being applied to aerosol studies.

PLANS FY00

          The FEOM technique for chemical kinetics calculations will be expanded into the domain of complex non-methane tropospheric chemistry. Initially, the number of input variables will be increased from four to six by including PAN and HNO3, and a five-tracer simulation with interacting NOx, HNO3, PAN, O3, and CO will be conducted. The overall aim is to include increasingly comprehensive atmospheric chemical mechanisms such as the carbon-bond mechanism (CBM-IV).

          The contributions to tropospheric ozone from stratospheric ozone, ozone produced in the relatively clean troposphere, and ozone produced in the polluted boundary layer will be quantified, and the dependence of each source of ozone on season and location will be analyzed.

          The analysis of the GFDL GCTM simulation of the tropical South Atlantic Ocean tropospheric ozone maximum will continue with an emphasis on its seasonal growth and decay. Also, analysis of the GFDL GCTM simulated tropical distribution of tropospheric ozone over the Indian Ocean will be initiated, coincident with the observational program INDOEX.

          GCTM simulations of NOx, CO, and O3 employing present sources of CO and NOx and estimates of future emissions will be completed, and the present and future regional and global impacts of increasing Asian emissions will be determined. Specific goals include quantifying the export of NOx and ozone from Asia's polluted boundary layer to the free troposphere, investigating the impact of this export on the balance of ozone production and destruction in the Central Pacific, and quantifying the impact of Asian emissions on North America. This study will be conducted in collaboration with the Center for Global and Regional Environmental Research at the University of Iowa (CGRER).

          In collaboration with CGRER, a study to quantify the regional and global air quality impacts resulting from the Indonesian fires in 1997 will be completed. GCTM simulations of NOx, CO, and O3 with increased burning due to the 1997 El Niño induced drought will be compared with available observations in the Indonesian region. Results from the GCTM will also be compared with results from a regional model that includes the 1997 El Niño meteorology.

          Nitrogen oxide chemistry and transport will be included in the RAINS-Asia model (Regional Acidification and Information Simulation - Asia), a widely-used integrated assessment model for science and policy studies of regional air pollution in Asia. This work is undertaken in collaboration with Gregory Carmichael (CGRER). Model results will be compared with observations and used to investigate issues of future environmental policy in East Asia.

          Sulfate and dust emissions will be implemented in the SKYHI model during the coming year. Global-scale calculations of the distribution of these aerosol components will be compared to measured global distributions. In addition, a comparison of the relative abundance of sulfate and carbonaceous aerosols will be conducted. The direct radiative forcing of all these aerosols will also be calculated.

  2.4 ATMOSPHERIC DYNAMICS AND CIRCULATION

     2.4.1 SKYHI Model Development

ACTIVITIES FY99

          A new standard version of the SKYHI model has been produced. It includes a new optional convection parameterization, additional refinements to the long- and short-wave radiation codes, and numerous computational enhancements designed to provide users with additional flexibility and greater ease-of-use. The code has also been reorganized to allow a straightforward transition to a Fortran 90 module structure.

          Development efforts have been shifted to the task of producing Fortran 90 modules which incorporate the existing SKYHI functionality, and which can be executed within the GFDL FMS framework. The long- and short-wave radiation codes have been converted, and work is under way on the remaining code to be included within the FMS. In addition to the mechanical conversion of the code to Fortran 90 modules, effort is being expended to make the code cleaner and more modular, removing the last vestiges of coding practices that were necessary under Fortran 77 on previous computational platforms.

PLANS FY00

          Additional Fortran 90 modules will be created from the existing SKYHI code and made available within the GFDL FMS. Alternative parameterizations and numerical techniques for various model processes will be assessed as part of the development effort. Creation and testing of one or more FMS-based models configured for use by the Atmospheric Processes group will occur.

     2.4.2 SKYHI Control Integrations and Basic Model Climatology

ACTIVITIES FY99

          Control integrations were continued with versions of the SKYHI model at various spatial resolutions. Most notably, the simulation with the 40-level 0.33°x0.4° latitude-longitude version has continued for almost 10 model months. This has allowed a comparison of the large-scale atmospheric simulation at various resolutions (it). The results display a significant improvement in the simulated zonal-mean circulation in the extratropical middle atmosphere with increasing horizontal resolution (A98/P99). At the very fine 0.33°x0.4° resolution the model simulation is largely, if not completely, free of the biases in extratropical zonal-mean circulation evident in lower-resolution versions. Thus, for example, in low-resolution versions the winter pole in each hemisphere is considerably colder than observed, a bias which is much reduced in the 0.33°x0.4° version. This improvement in the simulation of extratropical mean temperature is accompanied by a substantial increase in the strength of the pole-to-pole residual mean meridional circulation in the middle atmosphere. Interestingly, the sensitivity of tropical residual mean vertical velocity to model resolution is very much smaller than that seen in the extratropics. The important issue of how the strength of tropical upwelling is controlled will be addressed with some model experiments including extra sources of zonal drag (2.4.6).

          As part of a World Climate Research Program (WCRP) project, an intercomparison of the basic features of the simulations in several middle atmospheric GCMs including SKYHI has now been completed (kt).

          An extended control integration using the 40-level 3.6°x3.0° version latitude-longitude version of the model with predicted clouds was begun, with the goal of completing a 50-year integration. The model climatology is to be compared with integrations using the FMS, and as a control for experiments with perturbations in radiatively active species. Initial results show a significant increase in stratospheric water vapor, with equilibrium in water vapor concentration requiring ~15 years of model integration time.

PLANS FY00

          Further analysis of the high-resolution integration is planned, including an attempt to compare some of the small-scale and high-frequency variations in the simulated middle atmosphere with available observations. The extended control integration will be further analyzed.

     2.4.3 Spontaneous QBO-like Tropical Wind Oscillations in SKYHI Simulations

ACTIVITIES FY99

          The SKYHI model, when run at 1°x1.2° horizontal resolution and with 80 levels between the ground and 80 km, was found to display a very strong long-period oscillation in the equatorial zonal-mean zonal winds with properties similar to that of the observed quasi-biennial oscillation (QBO), although with a period only about half that of the observed QBO (A98/P99). This earlier result has been verified by a number of additional SKYHI model experiments. First, a series of experiments has shown that the model displays a QBO-like oscillation only when both the horizontal and vertical resolutions are sufficiently fine. It appears that the 3°x3.6° version will not oscillate, no matter how fine the vertical resolution. Similarly, versions with only 40 levels in the vertical will not oscillate no matter what the horizontal resolution (a result that applies even to the 40-level 0.33° x0.4° version discussed in 2.4.2). The 2°x2.4° version with 80 levels does oscillate and this has been used for a number of additional experiments. In one experiment the horizontal diffusion coefficient in the 2°x2.4° model was increased to that normally used in the 3°x3.6° version. This immediately suppressed the simulated equatorial oscillation, suggesting that the key reason that a fine horizontal resolution is required for equatorial mean wind oscillation is the small subgrid-scale horizontal diffusion that can be used. In another experiment the vertical diffusion in the 80-level model was set to that normally used in the 40-level model. This led to only very small changes in the simulated equatorial oscillation. Thus, it appears that the requirement of fine vertical level-spacing for SKYHI to oscillate is not related to the scaling of vertical diffusion parameterization with model resolution, but presumably has to do with adequately resolving the details of the interaction of the mean flow with vertically-propagating waves.

          Two additional experiments have been conducted with the 2°x2.4°, 80-level model. In one of these, the prescribed sea surface temperatures (SSTs) were increased substantially (by 5°C at the equator) over those of the standard control, and in the other, the SSTs were reduced by an equal amount. Both models still displayed QBO-like oscillations of the mean wind, but the period of the warm (cold) SST experiment was shorter (longer) than in the control. This is consistent with expectations based on simple theories of the QBO. When stronger wave fluxes emerge from the troposphere, the resulting stratospheric mean flow accelerations are predicted to be stronger and each cycle of the mean flow oscillation will then be completed more rapidly.

PLANS FY00

          The various simulations will be analyzed in more detail with a focus on understanding the role of vertically-propagating waves in forcing the QBO-like oscillations.

     2.4.4 Low-Frequency Variability of Simulated Stratospheric Circulation

ACTIVITIES FY99

          Long control integrations of the 40-level 3°x3.6° SKYHI model with prescribed climatological SSTs display significant quasi-decadal variability in the NH middle atmospheric circulation. As an example, Fig. 2.15 shows the winter mean mid-stratospheric North Pole temperature for 34 consecutive years in each of two separate control integrations. There are substantial periods when the winter mean temperatures remain either above or below the long-term mean. This sort of behavior requires the atmospheric circulation to have some effective memory of its earlier states extending over interannual timescales. This, of course, is contrary to the usual view of atmospheric predictability, which would suggest that the effective memory of the atmospheric flow is on the order of a month.

          One possibility that might help explain this behavior is the persistence of anomalies in the zonal-mean flow in the tropical stratosphere. Some results recently published have shown that this effect can lead to significant interannual variability in the extratropical stratospheric circulation in a simple mechanistic stratosphere model, even if the wave flux from the troposphere is held constant from year-to-year. In order to investigate this possibility in SKYHI, the control integration has been repeated in a version of the model with an imposed momentum source that constrains the tropical stratospheric mean flow to be close to the climatological value (taken from the control experiment). This leads to a simulation with a long-term mean climatology very similar to that in the control run, but with a significantly different character to the temporal variability. In particular, the variability in the extratropical stratosphere is strongly suppressed in the constrained model for all periods greater than about 6 months, and is virtually the same as in the control model at shorter periods.

PLANS FY00

          The control and constrained integrations will be extended, and detailed analysis of the simulated interannual variability will continue.

     2.4.5 Horizontal Spectra from High-Resolution SKYHI Integrations

ACTIVITIES FY99

          The 0.33°x0.4° version of SKYHI has been shown (A98/P99) to produce a simulation of the horizontal kinetic energy spectrum in the troposphere that can be characterized as a power law with -3 slope at wavelengths larger than about 500 km, and with a fairly abrupt transition to a -5/3 slope at shorter wavelengths. The transition to a shallower mesoscale regime near 500 km wavelength is in good agreement with available observations. Calculations of the basic tropospheric spectra and comparisons with observations have been completed (1594).

          The analysis of the high-resolution SKYHI results has continued. The spectral kinetic energy budget has been computed in a manner that for each wavenumber identifies contributions to the kinetic energy tendency from nonlinear advective processes, from conversion of available potential energy, from mechanical fluxes through the lower and upper boundaries of the region considered, and from subgrid-scale dissipation. In the troposphere, advective contributions are negative at small wavenumbers and positive over the rest of the spectrum. This is consistent with dominant downscale nonlinear cascade of kinetic energy at all scales. Thus, in this model experiment at least, the shallow mesoscale regime is not caused by upscale cascades from horizontal motions forced at small scales, a speculation that has been frequently advanced in the past.

          The kinetic energy spectra and spectral budgets have also been computed for the stratosphere and mesosphere, and the results compared with those from the troposphere. The spectra become increasing shallow with height. This is consistent with expectations that the circulation at high altitudes should be dominated by vertically-propagating waves, and in particular, waves with large vertical group velocity. For gravity waves, the vertical group velocity scales as the horizontal wavenumber and so it is not surprising that the higher wavenumber end of the spectrum becomes more prominent with height. This overall picture of the dynamics is also consistent with the spectral budgets for the stratosphere and mesosphere, which show that the main forcing of eddy kinetic energy at almost all scales comes from mechanical stresses from below.

          SKYHI results have been included with those from several other middle atmospheric GCMs in an intercomparison of horizontal kinetic energy spectra (jc).

PLANS FY00

          A brief portion of the 0.33°x0.4° SKYHI integration is being rerun with enhanced horizontal diffusion. The results will be analyzed to see how this affects the simulated spectra and spectral budgets. Also, some work has begun in analyzing the temperature and tracer variance spectra in a similar manner.

     2.4.6 Parameterized Gravity Wave Drag in the SKYHI Model

ACTIVITIES FY99

          Work has continued towards tuning a version of the Alexander-Dunkerton gravity wave drag parameterization scheme in the 3°x3.6°, 40-level version of SKYHI. The results obtained so far have been quite promising, with a clear demonstration that the scheme can greatly reduce the largest biases in the control model. The aim is to find a set of parameters that produces a simulated zonal-mean temperature structure that departs by, at most, a few degrees from observations throughout the middle atmosphere and through the year.

PLANS FY00

          The work towards efficient implementation and appropriate tuning of the gravity wave scheme will continue, and some specific issues will be addressed with controlled experiments of various sorts. Among these issues is the sensitivity of results to the spatial resolution of the applied gravity wave drag. For example, experiments will determine whether results change substantially if the scheme is applied independently at each grid point or is applied using some mean flow averaged over a geographical region encompassing several grid points. The sensitivity of tropical upwelling to the details of the parameterized drag will also be examined.

     2.4.7 Effect of Advection Schemes

          An investigation into the effect of the choice of advection scheme on stratospheric age-of-air calculations has been completed. The calculations have been performed on-line with SKYHI GCM together with six advection schemes employed for tracer transport (A98/P99). The age-of-air calculations constitute a stringent test of the advection schemes, as they consist of multi-year integrations of a passive tracer initialized in a localized portion of the atmosphere and are thus very sensitive to diffusion and dispersion errors inherent in the advection schemes. Prior to this work, little was known about the relative performance of various advection schemes in long, three-dimensional stratospheric integrations, as most schemes are usually tested in idealized 1-D or 2-D flows and short integrations or in 3-D simulations of atmospheric tracers with chemical and/or physical sources and sinks. The models' parameterizations also can obscure the intrinsic behavior of advection schemes. Moreover, prior to the measurements of the "age" tracers CO2 and SF6, the only observational data against which the simulations of truly conservative tracers could be compared were the rather sparse radioactive carbon-14 and strontium-90 data from the nuclear bomb tests in the 1960s.

          The age-of-air calculations have revealed qualitative differences between sigma-dot and theta-dot trajectories (i.e., trajectories using the model's sigma-coordinate vertical velocities and potential-temperature-coordinate vertical velocities, respectively) and between the NCAR semi-Lagrangian transport (SLT) scheme and three other contemporary advection schemes. As shown in Fig. 2.16, a wide variety of age distributions, ranging from "peaked" to "flat" and from "young" to "old", can be obtained within the same numerical model, depending on the choice of advection scheme. This sensitivity of model ages to the choice of advection scheme calls for caution in interpreting age-of-air distributions solely in terms of a model's large-scale transport behavior. The growing body of measurements from which the age of stratospheric air can be inferred has been used to assess the realism of the various schemes. In particular, these calculations have shown that the least diffusive schemes (i.e., centered difference schemes) produce most realistic ages, including ages in excess of 10 years, while the monotonic, but more diffusive schemes generate ages that are much too young.

          The different age distributions in Fig. 2.16 computed using theta-trajectories and sigma-trajectories ("flat" and "peaked", respectively) result from high-frequency vertical motions undergone by sigma-trajectories. These sigma-dot motions, although essentially adiabatic, are numerically not constrained to the model isentropic surfaces, and in fact are inconsistent with the diabatic heating rates, both in the model and in the real atmosphere (judging by comparison with the ascent rates inferred from the "tape recorder" signal in tropical water vapor data, and with the diabatic heating rates diagnosed from temperature and constituent data). These motions are illustrated in Fig. 2.17, which shows particle positions after 120 days in the trajectory run used in generating the age distribution in Fig. 2.16. The spurious cross-isentropic excursions of sigma-particles give rise to the peaked age distribution

in Fig. 2.16(e). These excursions were noted previously in off-line trajectory calculations and the speculation then was that they might be caused by the aliasing of kinematic velocities when the winds are sampled every few hours. The present results, obtained with an on-line code, demonstrate that this is cannot be the most likely explanation. These spurious motions are not eliminated by various computational enhancements, such as doubling of the vertical resolution in the dynamical module of the GCM or the use of a Runge-Kutta (rather than Euler forward) time-marching scheme in the trajectory code. In fact, increasing the vertical resolution of the GCM to 80 levels seems to worsen the discrepancy between sigma-dot and theta-dot trajectories. Based on these results, the conclusion is that this spurious transport appears to be intrinsic to sigma-dot trajectories, which do not recognize the location of

isentropic surfaces at each trajectory time step. As a consequence, sigma-trajectories are unrealistic in long-term integrations, at least in the vertical direction. This behavior appears to be a random walk effect caused both by real physical processes (e.g., gravity waves) and by computational noise in the sigma velocities. The behavior of sigma-dot trajectories in Fig. 2.17 illustrates a basic inconsistency between the trajectory calculations and the Eulerian dynamics of SKYHI and other GCMs. This type of trajectory inconsistency is believed to be at least partially responsible for the excessive numerical diffusion and anomalously young ages for the SLT scheme in Fig. 2.16(d). It is likely that the performance of the SLT scheme can be improved when the trajectory calculations that form the essence of the scheme are performed using theta-dot rather than sigma-dot velocities. It is also possible that the numerical diffusion of the SLT scheme is compounded in the SKYHI model (compared with the scheme's "parent" model, the NCAR Community Climate Model), due to SKYHI's use of an explicit time differencing and a short time step. Both of these factors result in more short-term dynamic variability (evident, for example, in the mesoscale spectra), which in turn can exacerbate trajectory interpolation errors.

          Motivated by the age-of-air results, a new investigation is under way on the effect of the choice of advection scheme on the simulations of tropospheric CO2 and SF6 (A98/P99, section 4.4.1). This investigation appears likely to have implications for studies of the global carbon budget.

     2.4.8 GCM Chemical Simulation with an Imposed Tropical
              Quasi-Biennial Oscillation

ACTIVITIES FY99

          More than 16 years of simulation have been completed with a version of SKYHI including a detailed prognostic photochemical code and forced with an imposed momentum source at low latitudes designed to produce a realistic simulation of the QBO in mean zonal flow. This QBO-forced SKYHI experiment has been shown to produce a QBO in total column ozone in the tropics and subtropics that compares very well with observations (A98/P99). Even the observed annual cycle modulation of the ozone QBO is well reproduced. In the past year, a detailed analysis of the QBO modulation of constituent transport and its effects on the ozone chemistry have been conducted. Some interesting QBO effects on both zonal-mean and eddy transports have been found (jx). The shading in Fig. 2.18 shows meridional cross sections of the deviations in zonal-mean ozone mixing ratio from its long-term mean in the simulation. Results are shown for two Januaries at near-opposite phases of the QBO. In the first January (top panel) the equatorial mean wind is strongly westerly near 40 mb. In this month, the equatorial lower stratospheric ozone mixing ratio anomalies are positive, and this can be ascribed to the anomalous downwelling associated with the QBO-induced mean circulation. There is a second maximum in the equatorial ozone anomaly near 6 mb. This is caused by the effects of QBO-related reactive nitrogen transport on the ozone chemistry in this region. Off the equator there are negative anomaly regions centered near 10 mb at 20°N and 20°S. These are ascribed to effects of QBO-induced transport of ozone. Note that the off-equatorial anomaly is strongest in the winter hemisphere. This is attributed to a strong seasonal modulation of the QBO-induced transport, which is much stronger on the winter side of the equator. One component of this transport is depicted by the anomaly residual mean meridional stream functions shown in the figure. The intensification of the mean meridional circulation in the winter hemisphere is obvious. The results for the near-opposite QBO phase in the lower panel are almost exactly the negative of those in the top panel.

PLANS FY00

          Further comparisons of the trace constituent simulations with long-term satellite observations will be conducted. The current lower-resolution simulations have significant deficiencies at high latitudes because of the winter cold pole bias of the model. Once an acceptable gravity wave drag parameterization is incorporated, some aspects of the QBO experiment will be repeated, with a goal of examining QBO effects on transport to high latitudes.

     2.4.9 Observational Studies Related to Middle Atmospheric Issues

ACTIVITIES FY99

          Six years of high-resolution twice-daily radiosonde data at Payerne in Switzerland have now been analyzed for gravity wave signals as a contribution to a WCRP project to establish a gravity wave climatology for the lower stratosphere (A98/P99). Attention has now turned to formulating strategies for characterization of the global results from the more than 200 available stations.

          Some preliminary work has started on a project to employ the high-resolution radiosonde data in an examination of the statistics of the occurrence of layers with small Richardson number. Such regions may indicate the presence of small-scale vertical mixing processes. This mixing has to be parameterized in GCMs, and typically such models have fairly coarse vertical grid spacing. An understanding of the statistics of the occurrence of thin mixing regions and their relation to larger-scale flow conditions will be very useful in formulating appropriate subgrid-scale mixing parameterizations.

          A collaboration has now begun with colleagues at the University of Western Ontario to compare gravity wave signals in their Rayleigh lidar data with simulations from high-resolution versions of SKYHI. A complete set (about 200 nights) of the Western Ontario temperature determinations at half-hourly intervals and 100 m vertical resolution has been obtained. These are now being interpolated onto the SKYHI vertical grids to allow a direct comparison with simulated temperature variations.

          In collaboration with colleagues at Wuppertal University, 18 years of daily determinations of airglow emission from the OH* layer (essentially a proxy for temperature in a layer centered near 87 km altitude) at Wuppertal (51°N) have been analyzed, along with somewhat shorter records for some Scandinavian stations (la). The results indicate the presence of significant quasi-decadal variability which may have analogues in the SKYHI simulations (2.4.4).

          The 27-day rotation of the sun leads to a quasi-periodic variation in various aspects of solar output reaching the earth. A number of recent studies have tried to find evidence for solar rotation effects in satellite observations of middle atmospheric temperature and composition. Here, an attempt is being made to detect the effects of solar rotation on atmospheric tides. In particular, the daily cycle of heating in the ozone layer is known to force the solar semidiurnal tidal oscillations that can be easily observed in hourly surface pressure data (particularly at low latitudes). The present study is making use of the unique 79-year record (1866-1944) of high quality barometric observations from Batavia (modern day Jakarta, 6°S) in conjunction with the record of daily sunspot numbers. Preliminary results suggest that near periods of 27 days, there is a negative correlation between sunspot numbers and the strength of the semidiurnal pressure oscillation. This could be accounted for if the dominant effect results from simple sunspot darkening, i.e., a slightly smaller output in the ultraviolet from the darker and cooler sunspot areas.

PLANS FY00

          Analysis of the high-resolution radiosonde data will continue with a focus on the global gravity wave climatology and on the application of these data to problems of subgrid-scale vertical mixing closure. The analysis of worldwide high-resolution radiosonde data will continue in collaboration with the other participants in the WCRP study.

     2.4.10 Dynamics of the Martian Atmosphere

ACTIVITIES FY99

          The profound influence of dust aerosol heating on the temperature structure of the Mars atmosphere is strikingly indicated by the increase in diurnal average temperature and thermal tide amplitudes following two global dust storms observed by the Viking spacecraft in 1977. The recent identification of, and accounting for, a surface radiance bias in the Viking IRTM 15 mm radiances has made possible the recovery of 0.5 mb (~25 km) temperatures (T15), providing a fresh look at the evolution of diurnal mean and diurnally-varying temperatures (iy). The goal is a description of the feedback processes associated with the spectacular growth and subsequent decay of these global dust storms. The diurnal variation of T15 and the observed semidiurnal surface pressure oscillations at the Viking Lander sites provide significant constraints on the evolving aerosol heating associated with two global dust storms. Dust storm simulations with the Mars GCM have been carried out which satisfy these observational constraints. The model includes an interactive, size-resolved description of the radiatively-active aerosol. These calculations provide insight into the evolving dynamical state of the atmosphere.

          Observations of the NH spring and summer seasons (aphelion) have been re-interpreted to reveal that this is a period of relatively little interannual variability relative to the much more variable perihelion dust storm season (iy). Simulations with parameterized dust injection by dust devils have been carried out to examine the maintenance of atmospheric opacity. A potentially important process for controlling the vertical distribution of dust aerosol in this season is a water ice cloud/dust aerosol interaction. A simple microphysical scheme based on a set of moment continuity equations has been incorporated into the GCM to investigate these processes.

PLANS FY00

          The dust storm simulations will be further analyzed and developed to explore the role of boundary layer phenomena in the rapid raising of dust during global dust storm events. The evolution of temperatures during the Viking mission will be compared with temperatures retrieved by the current Mars Global Surveyor mission and the upcoming Mars Climate Observer mission. The microphysical scheme will be further developed and tested against observations currently being made available from the Mars Global Surveyor spacecraft mission. An improved parameterization of sub-surface sources and sinks of water vapor will be developed to enable a better simulation of the latitudinally-dependent seasonal evolution of Mars water column abundances.

  2.5 CLIMATIC EFFECTS DUE TO ATMOSPHERIC SPECIES

     2.5.1 Lower Stratospheric Ozone and Temperature Trends

ACTIVITIES FY99

          Observations indicate that, over the period 1979-1994, there has been a cooling of the NH lower stratosphere (~50-100 mb). Fig. 2.19 illustrates this for the 45°N latitude belt, as obtained from a variety of datasets, including those based on radiosonde, MSU, and SSU satellite and lidar measurements. The annual-mean trend in the lower stratosphere is coherent amongst the various datasets, both with regard to the magnitude and statistical significance. The vertical profile of the temperature trend indicates that the entire stratosphere (from the tropopause to 1 mb) has undergone a statistically significant cooling over this period. The magnitude of the cooling increases with height (~0.8K per decade in the 20-35 km region and ~2.5 K per decade at 50 km). The various datasets are in reasonable agreement with one another throughout the stratosphere. It should be noted that the lidar observations comprise measurements at a single location, while the satellite observations comprise many samples in the 45°N latitude belt and the slight positive trend obtained from lidar at ~43 km is not deemed significant. A more detailed analysis reveals that the lidar-based trends are unusually sensitive to the end period considered for the trend analysis.

          Model simulations of the temperature changes due to the observed lower stratospheric ozone depletion are consistent with the observations in that region of the stratosphere (A98/P99). However, such simulations do not explain the cooling seen in the observations above the lower stratospheric altitudes. Considerations of the increases in well-mixed greenhouse gases does produce a cooling with increasing height in the stratosphere consistent with the vertical trend profile. However, the lower stratospheric ozone loss and the well-mixed greenhouse gas increases do not explain the magnitude of the cooling seen in the upper stratosphere or the slightly decreased or nearly uniform cooling seen over the 20-35 km region. The result presented here is part of a state-of-the-art assessment reported in Chapter 5 of the "Scientific Assessment of Ozone Depletion: 1998" (WMO, 1999). A comprehensive review of the subject has been completed (kp).

          A new trends profile of ozone losses in the stratosphere has been formulated for application in the SKYHI GCM. In contrast to the previous experiments (1394), the new profile extends from the lower to the middle and upper stratosphere globally. It also accounts for polar night ozone losses in a more rigorous manner than before by employing ozonesonde

observations made at very high latitudes. Further, the period covered is 1979 to 1998 in contrast to the 1979 to 1991 losses considered in the earlier study.

     2.5.2 Radiative Forcing Due to Tropospheric Aerosols and Ozone

          As part of the next IPCC Climate Change Assessment, an intercomparison of the peer-reviewed published forcing estimates due to sulfate aerosols (direct effect) and tropospheric ozone have been conducted. Compared to the 1996 IPCC report, there are now a large number of model studies available. Analyses reveal that the mean value and the range of model estimates for the direct sulfate aerosol forcing continue to remain the same as in the 1996 IPCC report. A new parameter, viz. the ratio of aerosol forcing at TOA to the column burden, has been formulated (1529). This parameter allows one to examine the sensitivity of the forcing per unit mass. Comparison of this parameter between the various simulations reveals that the differences between them arise principally due to the differing meteorology and transport in the various models.

          For tropospheric ozone forcing, the mean and range of estimates from several model studies suggest that the differences between the models is not significant. An encouraging development has been the use of observations in one of the recent studies. It turns out that the observations-based estimate of the forcing is in good agreement with that calculated by the 3-D chemistry-transport models. This further enhances the confidence in the estimate of the tropospheric ozone forcing since pre-industrial times. The ratio of forcing to column burden (i.e., forcing per dobson unit of ozone change) illustrates that the remaining differences between models is chiefly due to the treatment of chemistry and transport processes. There are, however, some significant regional differences among models regarding the amount of tropospheric ozone change in biomass burning regions.

PLANS FY00

          The effects on the stratospheric temperatures due to the newly formulated ozone loss profile will be investigated using the SKYHI GCM. The effects due to ozone changes will be contrasted with those brought about by the increases in the well-mixed greenhouse gases. A comparison of the simulated results with observations will be performed. The analysis of forcings since pre-industrial times as part of the IPCC (2001) Assessment will continue.

     2.5.3 Radiative Effects of Aerosol-Cloud Interactions

ACTIVITIES FY99

          The effects of anthropogenic aerosols on cloud composition and radiative properties were investigated using a detailed microphysical modeling approach. Observations from the Monterey Area Ship Track (MAST) Experiment were used to constrain simulations of the influence of ship-emitted aerosols on the microphysics of clean marine stratocumulus and continentally-influenced marine stratocumulus clouds. Factors which are largely ignored in global assessments of indirect aerosol forcing, such as aerosol size and composition, changes in cloud liquid water content (LWC), and changes in cloud absorption and transmission, are the focus of this modeling effort (jz).

          Cloud development has been simulated based on measured below-cloud aerosol size distributions using a size- and composition-resolved externally-mixed aerosol model3. In this model, the detailed chemical composition of internally- and externally-mixed aerosol populations is represented explicitly in a dual-moment sectional particle scheme. Water is tracked in the vapor phase and with a moving sectional scheme in the liquid phase (independently from nonvolatile components) in order to obtain accurate condensation and subsequent evaporation without losing size information. The model uses a simplified parcel scheme to follow measured lapse rates while predicting supersaturation. Since particles are activated to cloud droplets dynamically, accurate predictions of the number of cloud droplets activated, as well as their chemical composition can be obtained. The model provides a convenient way to track the contributions of various aerosol sources to cloud droplet formation.

          Mie scattering parameters for each cloud are differentiated according to the type of aerosols from which the cloud drops were activated and the activated drop sizes, using a combination of mixing rules for the composite refractive indices. A volume-weighted linear mixing rule is used for the non-absorbing fraction of each aerosol particle (consisting of water, ammonium sulfate, ammonium bisulfate, and sodium chloride), while Maxwell-Garnett theory is used for the absorbing fraction (consisting of organic carbon and black carbon). The Mie scattering parameters then act as inputs to a delta-Eddington exponential-sum-fit solar radiation algorithm (jg).

          Analysis of a clean marine stratocumulus cloud influenced by a ship plume reveals an increase of more than a factor of 4 in cloud optical depth at 550 nm wavelength, with droplets in the 1-7m diameter range activated from plume aerosols making the largest contribution. The increase in optical depth translates into a near doubling of the clean maritime cloud's visible albedo (A98/P99). Analysis of a continentally-influenced marine stratocumulus cloud and ship track reveals a 47% increase in cloud optical depth, with droplets larger than 7m diameter activated from both ship plume and continental aerosols making the largest contributions. The increase in optical depth translates into a 13% increase in the continentally-influenced cloud's visible albedo.

          The influence of clouds on atmospheric absorption was also investigated and found to enhance atmospheric absorption, as shown in Figs. 2.20 and 2.21. The spectral signature of the above-cloud absorption, with peaks in the regions of high cloud albedo, indicates that the extra above-cloud absorption is due to the interaction between scattering off of the cloud top and absorption by ozone and water vapor above the cloud. The spectral signature of the in-cloud absorption, with peaks out to 4 microns wavelength, indicates that the extra in-cloud absorption is due to the interaction between scattering by the cloud drops and absorption by water vapor and liquid water. In the clean ambient cloud (Fig. 2.20), absorption by liquid water swamps the effect of absorbing aerosols within the drops, while in the continentally-influenced ambient cloud (Fig. 2.21), there are absorption peaks in the visible spectral region due to the presence of organic and black carbon.

          Sensitivity studies were also performed with respect to changes in the supersaturation rate, SO2 concentration, and LWC. It was found that increases in both the supersaturation and SO2 concentration lead to increases in cloud optical depth and albedo. However, the largest influence on cloud albedo comes from constraints on the LWC. For the clean marine stratocumulus, the increase in LWC from the ambient cloud to the ship track contributes

significantly to the track optical depth, increasing the albedo by 4-23% over what would be predicted by conventional "Twomey" theory, which assumes no change in LWC.

          Comparisons with parameterizations of the "Twomey" effect indicate that microphysical modeling leads to a significantly larger predicted increase in cloud albedo. The three main sources of this disparity are: 1) parameterized models assume no change in LWC; 2) empirical relationships between cloud drop number concentration and aerosol concentration substantially underestimate cloud drop number concentration for high aerosol loadings; and 3) parameterizations assume that the composition of cloud drops consists of pure water, which underestimates the absorption by almost a factor of 3 in the continentally-influenced cloud.

          Simulations have been performed with GFDL's radiative-convective model (RCM), revealing a decrease in the steepness of the tropospheric lapse rate with increasing low cloud optical depth. A simulation has also begun using SKYHI with fixed SSTs, replacing the optical properties of low clouds alternately with the optical properties of the clean marine stratocumulus from the MAST study, and with those of its highly-polluted ship track. Preliminary analysis shows a globally-averaged decrease in total cloud amount, but an increase in TOA net shortwave flux and a decrease in land surface temperature.

PLANS FY00

          Using the aerosol microphysical model, further sensitivity studies will be performed with respect to updraft velocity, particle dilution, and the fraction of updrafts versus downdrafts. In the RCM, a study of the sensitivity of TOA and surface fluxes to cloud co-albedo will be conducted. In the SKYHI GCM, a comparison between variability of cloud amount in the control run (predicted SKYHI clouds) versus the clean and polluted cloud scenarios will be completed. Efforts are also under way to begin a simple global evaluation of the indirect aerosol effect, binning cloud optical properties according to monthly-mean anthropogenic sulfate column burdens estimated from 3-D chemistry-transport models.

     2.5.4 Radiative Forcing Due to Stratospheric Aerosols

ACTIVITIES FY99

          A spectral-, space-, latitude-, and time-dependent set of aerosol optical parameters for two years following the Mount Pinatubo eruption has been used in the SKYHI GCM to determine the changes in the global radiative balance. The GCM uses recently improved shortwave (jg) and longwave (1597) radiation algorithms. Clear- and total-sky forcings have been analyzed, as well as the temporal evolution of perturbations in the stratospheric heating rates. The total (solar+longwave) forcing following the eruption exceeds -8 Wm-2 over the tropical region and thus significantly disturbs the Earth's radiative balance for more than a year. Model clouds have been found to decrease the radiative forcing by ~50% relative to clear-sky. Solar heating in the near infrared has been found to contribute substantially to the total stratospheric heating throughout the two-year period, thus quantitatively addressing an issue that has been in dispute of late.

          Starting from different initial conditions, an ensemble of four 2-year SKYHI integrations were performed, using predicted clouds and climatologically varying SSTs. One set of integrations was for no-aerosol conditions while the other set included the Pinatubo aerosol perturbation. Fig. 2.22 shows the zonally averaged lower stratospheric temperature anomalies (simulated and observed) due to Pinatubo aerosols at 50 mb where the model anomalies are calculated from an ensemble of 4 runs. The observed anomalies over the 2-year period are evaluated using the 30-year National Centers for Environmental Prediction (NCEP) reanalysis data. The color shading denotes statistical significance at the 99% (red), 95% (green), and 90% (blue) confidence levels. Simulated temperatures increase by about 2-3 K during the first half-year following the eruption, due to the aerosol absorption of terrestrial longwave and solar near infrared radiation. The observations (NCEP reanalysis) show similar increases in temperature. During the first northern winter following the Pinatubo eruption, the simulated temperatures are higher than observations by about 1 K in the 30°S-30°N latitude region. During the second winter, however, the observed temperature anomalies are larger, in the range of about 2-3 K, compared with model simulations which are about 1 K. This difference is particularly pronounced in the tropics. These discrepancies are thought to be attributable to the QBO and ozone change effects, which are present in the observed response. The temperature changes at high latitudes fail the significance test due to the large interannual variability in those regions, a feature which is present in model simulations and observations.

          Figure 2.23 shows the temperature changes at 50 mb obtained from simulations and observations at low (0-20°S) and high latitudes (60°N), as well as the global-mean temperature anomaly over the 2-year period. At low latitudes, the stratospheric warming begins within 2-3 months, and the simulation pattern tracks the observed anomalies for about 9-10 months after the eruption. The aerosol optical depth at 0.55m is greater than 0.25 from September 1991 until April 1992 in this latitude region. A comparison with the time evolution of aerosol

optical depth at 0.55m (not shown) reveals that an unambiguous signature of Pinatubo aerosols is seen only when the aerosol optical depth is greater than about 0.25. After approximately 10 months, the optical depth drops below 0.25, and the volcanic signal is not strong enough to rise above the other possible effects likely present (e.g., QBO, ozone depletion, and other forced and unforced variations). At the high latitudes there tends to be some agreement between the simulations and observations, but both results are statistically insignificant.

          Although the value of optical depth (0.25) stated above is not a critical threshold in general, it serves to make the point that a given volcano's space-time impact on temperature can be discernible only if the forcing is strong enough that it can rise above the natural or unforced variability in the model/observations. This is especially so for the warming observed here in the low latitudes. In this case, a clear volcanic signal is seen for only about 10 months after the eruption, after which effects from the QBO and from ozone depletion possibly begin to contribute significantly to the temperature change pattern. In high latitudes, the large interannual variability present in temperatures inhibits the determination of a quantitative contribution by the Pinatubo aerosols to the warming, irrespective of the value of the volcanic aerosol optical depth. For the global average picture, the warming simulated by the model reproduces the temporal peak and decay of the 50 mb temperature quite well compared to the observations. This occurs because the dynamical forcings are nearly averaged out in the global-mean, and only the radiative effects contribute to the global-mean temperature pattern. Presumably, the use of the space-time observed aerosol properties enables a realistic global-mean aerosol forcing estimate to be obtained and thus to "drive" the model response closer to the observed temperature response.

PLANS FY00

          Using the time-dependent Pinatubo aerosol dataset, a series of 2-year experiments will be conducted with El Niño- and La Niña-type SSTs in SKYHI to study the dependence of the aerosol forcing, the lower stratospheric thermal response, and tropospheric climate response on the SSTs. Calculations using the GFDL FMS framework including the effects of a mixed layer ocean are planned.

     2.5.5 Radiative Effects of Methane, Nitrous Oxide, and the
              Water Vapor Continuum

ACTIVITIES FY99

          A study of the climatic changes simulated by the SKYHI GCM due to the introduction of methane (CH4), nitrous oxide (N2O), and due to use of the "CKD" water vapor continuum has been completed (1597; A96/P97), with the response of the GCM being characterized as either primarily "radiative" or "dynamical". In a "radiative" response, the model temperature change is closely correlated with the initial heating rate change due to the introduction of these species. A "dynamical" response implies a close correlation between the initial heating rate change and the heating rate change when the model has reached equilibrium. The results, illustrated in Fig. 2.24, indicate that the model response is "dynamical" in the troposphere and "radiative" in the middle atmosphere (above the troposphere). The middle atmosphere region shows two distinct relations between temperature change and initial heating rate change, which may be characterized as a difference in the radiative-dynamical damping time. Values for the damping time are ~4.1 days in the lower stratosphere region and ~29 days in the remaining region of the middle atmosphere.

PLANS FY00

          The effects of changes in concentrations of CH4, N2O and water vapor between 1979 and 1997 on the climate will be investigated using the SKYHI GCM.


1. Cooke, W. F., C. Liousse, H. Cachier, and J. Feichter, Construction of a 1°X 1° fossil fuel emission dataset for carbonaceous aerosol and implementation in the ECHAM4 model, J. Geophys. Res., accepted, 1999.

2. Cooke, W. F., and J.J.N. Wilson, A global black carbon aerosol model. J. Geophys. Res., 101, 19,395-19,409, 1996.

3. Russell, L.M., and J.H. Seinfeld, Size- and composition-resolved externally-mixed aerosol model, Aerosol Sci. Technol., 28, 403-416, 1998.

*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).