Climate and Global Dynamics Division
|NCAR | UCAR | NSF | ASR 98|
Climate Modeling Section
| The Climate Modeling Section's (CMS) research goal is to increase the
understanding of the atmosphere and its role in the climate system through modeling and
observational studies, and to represent that understanding in atmospheric models. This
involves studies of the physical mechanisms governing the global climate system and the
numerical modeling techniques required to represent these mechanisms. CMS scientists
compare observed atmospheric data with model output data to validate and to improve the
models. They also are principally responsible for developing and testing new versions of
the Community Climate Model (CCM) and validating and using the CCM in the study of
atmospheric circulation. The CCM Core Group develops and maintains the standard versions
of the CCM.
Development continues toward the next version of the Community Climate Model (CCM4). Within CMS, work is focused on improvements to the deep convection parameterization, alternative dynamical cores, prognostic cloud water and sulfate aerosol schemes, and interactions between clouds and radiation. Improvements to the CCM are being carried out in collaboration with other members of the Climate System Model (CSM) Atmosphere Model Working Group.
Philip Rasch (CMS) and Jon Egill Kristjansson, (visitor, University of Oslo) have continued to improve the prognostic cloud water parameterization. The parameterization has been updated to account for the relatively rapid growth of ice particles to sizes where sedimentation becomes important. Parameterizations for sedimentation velocities that depend upon crystal size and shape suggested by Heymsfield and Donner (1980) and Mitchell (1996) were explored in the single column model version of the CCM. The radiation and cloud microphysics for the CCM were modified to use a uniform set of assumptions about the ice crystal properties. The radiation and cloud microphysics parameterizations were also modified to make a uniform set of assumptions about warm cloud drop size. Simulations to investigate the sensitivity of the CCM climate have been performed and are now being evaluated.
Gordon Bonan (CMS) continued his study of the ecological and hydrological processes by which natural and human-mediated changes in vegetated landscapes affect climate. Previous modeling studies by Bonan showed that temperate deforestation cools the climate of the U.S. These studies were followed by analyses of observed temperatures for the U.S. in relation to land use. The analyses confirmed the results of the model simulations and showed cropland areas are cooler than forested areas. Bonan continued to develop and apply a land surface process model for use with the NCAR CSM. This model accounts for the ecological effects of different vegetation types and the thermal and hydrologic effects of different soil types. The model was expanded to include river routing, dust emissions, and volatile organic carbon emissions.
Three other major focii for further land model development are: (a) Arctic regions, using data collected during field campaigns in boreal forest and tundra ecosystems to identify the key ecological and hydrological processes important for climate modeling and how to best parameterize these processes (in collaboration with F. S. Chapin III (University of Alaska) and Amanda Lynch (University of Colorado)); (b) inclusion of ecosystem dynamics (in collaboration with Jon Foley (University of Wisconsin); and (c) development of a Common Land Model that combines aspects of several existing land surface models for use with the NCAR CSM (in collaboration with the CSM Land Model Working Group).
James Rosinski (CMS), Jerry Olson (CMS), and David Williamson (CMS) implemented a reduced grid into CCM code and completed testing. They defined an n-digit reduced grid for spectral transform models in which the relative error made by omitting certain associated Legendre functions from the Gaussian quadrature poleward of some latitude is less than 10n. The associated Legendre functions omitted from the quadrature are those that have unrepresentable Fourier wavenumbers at a latitude because of the reduced number of longitudinal points at that latitude. They demonstrated that an adiabatic Eulerian spectral transform model run on an n-digit grid is accurate to n digits for short integrations. The error introduced by the reduced grid grows as expected for a turbulent atmosphere and the growth is not accelerated by the reduced grid. The errors from an adiabatic semi-Lagrangian spectral transform model can be significantly larger than those from an Eulerian model because the interpolation required in the semi-Lagrangian method will not maintain more than a few digits of accuracy. They put the adiabatic model errors in perspective by illustrating errors introduced by arbitrary aspects of model specification, such as the longitude of the first grid point. For the Eulerian adiabatic model, there appears to be no justification in using a higher digit grid than the 2-digit one. A higher digit grid may be desirable for an adiabatic semi-Lagrangian model. The errors in a cosine bell advection test were also studied. All reduced grids they tested have the same error for Eulerian spectral transform advection. The semi-Lagrangian advection showed some increase in error with greater reduction in the grid, but the variation was comparable to the difference in error from choosing different interpolants or different time steps. There is little reason to choose a higher digit grid than the 4-digit grid. In fact the 3-, 2,- and 1-digit grids might be considered satisfactory for semi-Lagrangian advection. Finally, multiple year climate simulations with full and reduced grids were performed. Overall, the mean climates produced by the models run on the reduced grids were all very similar. They saw no indication that the reduced grids introduced pathological errors that contaminated the simulations. Their results indicated that even a 1-digit grid is suitable for climate modeling with both Eulerian and semi-Lagrangian spectral transform models.
Williamson and John Truesdale (CMS) carried out a series of experiments that attempt to identify some aspects of the time truncation error in the complete model. These simulations provide the scientific justification for a new design of the CCM, which clearly separated the dynamical core, subgrid scale parameterizations, and input/output aspects. In the new design, the dynamical core and parameterizations can be coupled in a purely time split manner or in a purely process split manner. They have carried out a set of integrations to determine the effect of the details of the time splitting on the simulated climate, and to provide an indication of the magnitude of that aspect of the time truncation error. The Time Split and Process Split versions of CCM3 were each run for 5 years and the original CCM3 was run for 10 years. In all cases, the parameterizations and dynamical cores were identical, and only the coupling between them was changed. The resolution conformed to the standard CCM3 configuration: T42 spectral truncation, 18 vertical levels, and 20-minute time step. Differences in the global average properties of the simulations were physically insignificant. Possibly because the simulations are relatively short, only 5 years, the regional differences in the modeled climates introduced by the different time approximations were smaller than estimates of natural variability except in a few isolated regions. The differences might be statistically significant given longer simulation averages. However, overall the differences were significantly smaller than differences introduced by changes in parameterizations, such as from CCM2 to CCM3, differences introduced by changes within a parameterization, and differences introduced by the "tuning" of parameterizations during model development to achieve, say, reasonable top of atmosphere energy balances. In addition, the differences were neither consistently good nor bad in terms of simulation quality. Thus, at least for the near term, the errors introduced by the different time approximations allowed in the new design will not be a dominant error in the simulations.
Williamson, Truesdale, and Glenn Grant (CMS) are collaborating with S.-J. Lin and Richard Rood (NASA Data Assimilation Office) to combine the Lin-Rood dynamical core with the CCM physical parameterizations. They are creating a version of the CCM with the new core based on the time split design described above. Once parameterizations proposed for CCM4 are settled, controlled comparisons will be made between the original CCM3 Eulerian spectral transform dynamical core, the semi-Lagrangian spectral transform dynamical core, and the Lin-Rood dynamical core, coupled to that parameterization package. The choice of core for CCM4 will be based on these experiments.
Byron Boville (CMS) collaborated with Wendell Welch (visitor, Yale University), Piotr Smolarkiewicz, and Richard Rotunno (both from NCAR's Mesoscale and Microscale Meteorology Division, MMM) on a study of the largescale flow over periodic mesoscale topography. This study used simple topography and large-scale flow to study the form drag produced by the topography and the partioning between low level drag and vertically propagating gravity waves, as a function of the topographic height and flow parameters. Next, the fine scale modeling results will be used to design a General Circulation Model (GCM) parameterization that will be tested in the CCM.
A model of the atmospheric sulfur cycle was developed by Rasch, Mary Barth (Atmospheric Chemistry Division (ACD) and MMM) and Jeffrey Kiehl (CMS). Three papers are in press describing runs in the NCAR CCM. The papers described the model formulation, a verification exercise, the use of the model as a tool for interpreting the atmospheric sulfur cycle, and estimates of the radiative forcing by sulfate aerosols. The parameterization is also being used in the Model for Atmospheric Transport (MATCH). The sulfur cycle and prognostic cloud water models tested in the CSM coupled ocean atmosphere context by Boville and used in the climate of the 20th and 21st century runs are described elsewhere.
Rasch is collaborating with Yiping Zhang (CMS) in the development of a size resolved sulfate aerosol model. This model explicitly represents the formation, transformation, and removal of small (Aitken) and intermediate (Cloud Condensation Nuclei, CCN) sized aerosol. A mass and a number concentration are predicted for each catagory. The CCM has been run with this parameterization for multiple years. Simulations are currently being compared with measurements of these quantities prior to tying the predicted aerosols more closely to the radiation and cloud microphysics. The explicit calculation of mass and number (and indirectly CCN size) provides an alternate to the empirical relationship between aerosol mass and cloud drop number that are used in most indirect radiative forcing calculations in GCMs.
In addition to the enhancement of the sulfate aerosol formulation itself, the parameterization for wet scavenging was recently modified to account for scavenging property differences between small and large particles in polluted and pristine regions.
Parameterizations for two more catagories of aerosols were developed by Rasch and William Collins (CMS). A simple parameterization for diagnosing sea salt aerosol concentrations using wind speed and a prescribed vertical profile was developed for MATCH. A formulation for carbonaceous aerosols was also developed for MATCH that includes emission of black and organic carbon from fossil fuel and biomass burning sources, as well as the production of organic aerosol from volatile organic compounds (terpenes).
Kiehl has initiated work on a parameterization of sea salt aerosols for the CCM. The parameterization will start from the work of Gong et al. (1998).
Collins worked with Kiehl, James Hack (CMS), and Truesdale to test parameterizations for longwave radiative transfer based upon the correlated-k method. This involved extensive intercomparisons with the existing longwave code and with benchmark line-by-line calculations. Collins and Truesdale developed a method for running similar longwave codes in parallel to facilitate the intercomparison. This and other techniques created for the intercomparison will be formalized into a general procedure for testing other alternative longwave codes. Based upon the results of this study, Collins is rewriting the existing longwave code to permit more general assumptions regarding cloud overlap. The results of the study were reported at the Spring 1999 Atmosphere Model Working Group meeting and the 1999 Climate System Model Workshop in Breckenridge, Colorado. Collins, Boville, Hack, and Erik Kluzek (CMS) also developed a general web-based facility for intercomparison of CCM model simulations.
Hack has continued a detailed analysis of convection and atmospheric boundary
layer processes as simulated in the global model, and has been exploring generalizations
to the CCM deep convection closure in collaboration with Drs. Minghua Zhang (State
University of New York, Stony Brook) and Shaocheng Xie (Lawrence Livermore National
Laboratory). Single column model integrations have identified a modified closure that
addresses some of the deficiencies in the transient behavior of parameterized deep
convection, as well as some of the systematic biases in the mean state. Preliminary tests
in the NCAR CCM3 suggest that many of the improvements identified in the single column
modeling tests carry over to the more complex global model in the form of an improved
climate simulation, although improvements to the deficiencies in transient behavior have
not been as promising.
Collins developed a method for introducing enhanced, or "anomalous," shortwave absorption into the NCAR CSM. The method, termed Generic Enhanced Absorption (GEA), is a very simple method for enhancing shortwave absorption in cloudy atmospheres that requires minimal retuning of the model. The integrations of the CSM with GEA are some of the first experiments with a coupled climate model testing the effects of enhanced absorption on climate state. The introduction of GEA greatly improves the annual mean energy budgets of the tropical Pacific Ocean at the top of atmosphere and surface. The differences between simulations with and without GEA are consistent with the hypothesis advanced by Kiehl (1998). Kiehl suggested that the transient response of the CSM to coupling of the ocean and atmosphere is related to the absence of enhanced absorption.
The results from this study were reported at the Global Energy and Water Cycle Experiment (GEWEX) Cloud System Study (GCSS) Working Group on Numerical Experimentation (GCSS-WGNE) Workshop on Cloud Processes at the European Centre for Medium Range Weather Forecasts (ECMWF), at the Cess Symposium hosted by the Scripps Institution of Oceanography in October 1999, and at the NASA Goddard Space Flight Center in May, 1999. A journal manuscript describing the findings is in preparation, and the presentation at the Cess Symposium will also be published by Cambridge University Press.
Collins has joined the science team developing the NASA Triana satellite to be launched into the L-1 Lagrange point in 2001. The satellite includes a new type of radiometer designed to measure the reflected shortwave and near-infrared radiation from the sunlit hemisphere of the Earth. This data will be used to evaluate the conclusions of Collins (1998) that similar data from Nimbus-7 showed clear evidence of enhanced shortwave absorption in clouds. Collins will be heading the model development and analysis required to interpret the measurements from Triana.
Clear-sky fluxes from the NASA Earth Radiation Budget Experiment (ERBE) have been used extensively to evaluate the GCM estimates of clear-sky longwave radiation (e.g., Kiehl and Briegleb, 1992) and aerosol radiative forcing (e.g., Haywood et al., 1999). There are well-known biases in the satellite values of clear-sky outgoing longwave radiation (OLR) and possibly in the satellite shortwave albedos that have hampered evaluation of GCMs. Collins has demonstrated a simple technique for mitigating or eliminating the biases in the longwave, and has developed a method for quantifying the errors in observational estimates of aerosol radiative forcing. These methods were reported at the Fall 1998 meeting of the American Geophysical Union. Collins and Brian Soden (Geophysical Fluid Dynamics Laboratory) have been funded by NOAA and NASA to apply these techniques to produce a new, more accurate satellite data set for validating GCMs.
Hack and Williamson have been conducting exploratory climate simulations of the NCAR CCM3 at high horizontal resolution (a 170 wave configuration, or four times the standard climate resolution) in collaboration with colleagues in the NCAR Scientific Computing Division and colleagues at the Central Research Institute for the Electric Power Industry in Japan. A substantial effort was invested to ensure that the top of atmosphere and surface energy budgets for the high-resolution model were consistent with the low-resolution version of the CCM3 and available observational data sets. They have succeeded in tuning the cloud parameterization to achieve a highly realistic energy balance, where initial analysis of a short control simulation suggests that this and many other mean features of the simulated climate are very realistically reproduced. Analysis of this simulation suggests that there are aspects of the mean and transient behavior that represent significant improvements over the low-resolution results. In particular, the analyses are focused on the climatology of tropical storms and potential improvements in the Madden-Julian Oscillation for the high-resolution model, as well as other transient phenomena.
Williamson studied properties of the pressure vertical velocity in the reanalyses to help shed light on the issue of convergence of the CCM simulations with increasing resolution. The reanalyses considered were National Centers for Environmental Prediction (NCEP)/NCAR (referred to as NCEP I), NCEP/Department of Energy (DOE) Atmospheric Model Intercomparison Project (AMIP II) (referred to as NCEP II), and ECMWF Reanalysis (ERA). In the tropics, comparisons of the reanalyses were made of temporal, zonal averages over the Atlantic and eastern Pacific Oceans. In mid-latitudes, where transient synoptic systems dominate, probability density distributions were compared over storm track regions. In the tropics, given the differences between the reanalyses, and not having any direct measurements of grid-scale vertical motion, it is difficult to conclude that the reanalyses w represent the atmosphere, even on the T42 and larger scales. It is likely that the model-generated first guess is still dominant over observations in the forecast/analysis system and that the subgrid scale parameterizations play an important role. In mid-latitude storm track regions, the probability density distributions of each reanalysis as a function of scale appear to resemble those from model simulations in that the finer scales contribute to more stronger and less weaker upward vertical motions and a modest increase in subsidence. However, the reanalyses disagree on the probability density distributions of the T42 and larger scales. NCEP II is significantly different from NCEP I and ECMWF. Given these differences and the apparent dependence on subgrid scale parameterization, it remains difficult to say with any confidence which reanalyses represent the atmosphere most closely for the scales commonly included in climate model.
The coupling between chemistry and climate is a strong focus within CMS. The chemical composition of the atmosphere determines a major component of the radiative forcing of the climate system. This composition is determined by processes that depend on climate. Thus, the potential exists for important feedbacks between these two systems. The most comprehensive approach to understanding these interactions is to couple climate and chemistry within a single global model. The CMS is working with atmospheric chemists in ACD and at universities to accomplish this goal.
Kiehl, in collaboration with Timothy Schneider (CMS), Rasch, and Barth completed a study on the direct and indirect forcing due to sulfate aerosols in the CCM3. This study employed a version of the CCM that includes an interactive sulfur chemistry model. They found that the indirect effect is very sensitive to whether one assumes sulfate competes with other sources (e.g., sea salt) of cloud condensation nuclei. This leads to a range of uncertainty in indirect forcing of -0.4 to -1.8 Wm-2. These results are discussed in Kiehl et al. (1999a).
Kiehl, in collaboration with Schneider, Susan Solomon, and Robert Portmann (both from NOAA Aeronomy Laboratory) completed a study of the climate forcing due to changes in tropospheric ozone, stratospheric ozone, and total ozone. This study employed an ozone data set for the pre-industrial time period obtained from a chemical transport model. The present-day ozone data were based on a combination of satellite and ozonesonde data. The calculations indicated that changes in total ozone from pre-industrial to present led to a forcing of 0.29 Wm-2. Tropospheric ozone forcing results in a forcing of 0.3 Wm-2, thus stratospheric ozone changes led to a -0.01 Wm-2 forcing. Sensitivity studies were also carried out for different assumptions concerning pre-industrial ozone levels. This study also found that positive forcing from tropospheric ozone can significantly offset local negative forcing from sulfate aerosols. (This figure (55KB) shows the geographic distribution in forcing (Wm-2) due to: (a) direct effect of sulfate aerosols, and (b) sulfate aerosols plus ozone (SEFDH) forcing for June-July-August for 1870 to 1990.) These results are discussed in Kiehl et al. (1999b).
Kiehl, in collaboration with Hack and V. Ramanathan (University of California San Diego, Scripps Institution of Oceanography, SIO), has begun to look at the role of absorbing aerosols in affecting the atmospheric thermodynamic state over the Indian Ocean region. Measurements during the Indian Ocean Experiment indicate the pervasive presence of absorbing aerosols over a large region. Measured single scattering albedos were typically 0.8, much lower than that resulting from pure sulfate aerosols. The study employs the CCM Single Column Model (SCM) in conjunction with NCEP analyses for the time period of early February, 1999. Simulations indicate that the presence of the aerosol warms the lower to mid troposphere by 3 to 4 K.
Kiehl and Ramanathan are carrying out simulations with the CCM3 to study the effects of
absorbing aerosols on the general circulation in the Indian Ocean region. These studies
employ a prescribed distribution of absorption that matches the magnitude observed during
the Indian Ocean Experiment (INDOEX) time period. Initial simulations indicate that the
presence of the aerosol has a large impact on the position of the deep convection
occurring in the Intertropical Convergence Zone (ITCZ) .
Aerosol Assimilation and Forecasting
A system for forecasting aerosols was developed using a chemical transport model together with an assimilation of satellite aerosol retrievals. Collins, Rasch, and Brian Eaton (CMS) implemented the forecast system to help during the INDOEX. The forecasting system consists of MATCH combined with an assimilation package originally developed (by ACD) for atmospheric chemistry applications. The aerosols forecast by MATCH include sulfate, mineral dust (originally developed by C. Zender (CMS and ACD)), and the carbonaceous, and sea-salt aerosols mentioned above. The model includes a detailed treatment of the sources, chemical transformation, and wet and dry deposition of the aerosol species. For more information, please go to the MATCH webpage.
The aerosol forecasts involve a two-stage process. During the assimilation phase, the total column aerosol optical depth (AOD) is estimated from the model aerosol fields. The model state is then adjusted to minimize differences between the simulated AOD and satellite retrievals of AOD. During the subsequent forecast phase, the aerosol fields are evolved using meteorological forecasts.
During INDOEX, the NCAR team produced 24- and 48-hour forecasts of the total AOD over the Indian Ocean region. The model was integrated using global near-real-time meteorology supplied by NCEP. The meteorological forecast and analysis fields were sampled at 3-hour intervals at T126 horizontal resolution, approximately 0.8 degrees at the equator. The satellite aerosol retrievals were derived from Advanced Very High Resolution Radiometer (AVHRR) data acquired at the INDOEX operations site on Male and from low-resolution AVHRR data supplied by NOAA. The satellite and meteorological data were collected and processed at NCAR, and the forecasts were then transmitted back to Male. Collins and Rasch traveled to Male to interpret the forecasts for the ship and aircraft mission scientists.A comparison of the AODs is shown from the assimilation system and from in situ INDOEX data.(This figure (19KB) shows a comparison of the aerosol prediction of total AOD from the assimilation and a measured optical depth (c/o Satheesh and Ramanathan, manuscript. in preparation) from the Maldive islands during INDOEX.) The data is from a sunphotometer at the Kaashidhoo Climate Observatory (KCO, at 4.96N, 73.47E). The comparison shows that the model was able to reproduce the time series of daily mean AODs with small mean and root mean square errors. This accuracy was critical for mission-planning applications during INDOEX, and Collins and Rasch are planning to improve the accuracy further by assimilating INDOEX observations from surface and airborne instruments.
Fabrizio Sassi (formerly ACD, now with CMS) and Boville (CMS) collaborated on the development of a middle atmosphere version of CCM3 (MACCM3). Sassi updated the gravity source functions in the Lindzen style gravity wave parameterization previously developed by Boville. The current MACCM3 uses Williamson's two time level semi-Lagrangian dynamics algorithm at T63 on a 128x64 (linear) grid and extends from the surface to approximately 85 km with 52 layers. The model solves for the same greenhouse gases as in the CSM (methane, nitrous oxide, CFC11, and CFC12). In addition to the water vapor source from methane oxidation, MACCM3 now includes the effects of water vapor destruction due to Lyman-alpha radiation and the source from molecular hydrogen, based on two-dimensional model solutions provided by Rolando Garcia (ACD). A pair of "age of air" tracers are also included in all MACCM3 simulations. MACCM3 has been used by Takeshi Horinouchi (University of Washington), in collaboration with Sassi and Boville, to study synoptic waves and transport between the tropics and extratropics in the lower stratosphere.
Planning for the upward extension of MACCM3 has been completed and the work has been started in collaboration with Garcia and Raymond Roble (High Altitude Observatory, HAO). The CCM will initially be extended to 120 km, near the turbopause, requiring a nonlocal thermodynamic equilibrium parameterization for longwave radiation. Later the CCM will be extended through the thermosphere to approximately 500 km. The CCM framework will then serve as the basis for thermospheric modeling studies at NCAR. All of the work on numerical algorithms and code development being undertaken for climate modeling will then be of immediate benefit for middle and upper atmosphere modeling.
Williamson, Olson, Boville, and Sassi collaborated with Janusz Eluszkiewicz (Atmosphere
Environmental Research, Inc.) and the Geophysical Fluid Dynamnics Laboratory (GFDL) group
headed by Jerry Mahlman to try to unravel the puzzling results obtained with the GFDL
SKIHI model on ``age of air.'' They observed that the semi-Lagrangian transport
codes produced significantly younger air than the Lin-Rood scheme that were in turn
younger than those from the original SKIHI centered differences. Experiments with a
semi-Lagrangian version of MACCM3 produce air more than twice as old as the GFDL
simulations with semi-Lagrangian transport, but the air is still younger than the other
schemes in the GFDL model. Some of the youth seems to be attributable to monotonic
constraints in the semi-Lagrangian and in the Lin-Rood schemes. However, the actual cause
of the differences remains unknown and investigations continue.
CCM Single Column Model (SCM)
A single-column version of the NCAR CCM2 and CCM3 has undergone additional development under the direction of Hack, in collaboration with John Pedretti (CMS). Over the past year this team has continued to refine the design and implementation of the CCM SCM Graphical User Interface (GUI), thus improving the capabilities, efficiency, and portability of the SCM. These efforts have significantly simplified the process of "building" the SCM using sophisticated scripts that automate a number of the more complicated tasks. The SCM has been ported to SGI, DEC, IBM, and most significantly, LINUX platforms. A client-server version of the model has been developed, the web-based user documentation has undergone significant modifications, and several new forcing data sets using various data derived from the Atmosphere Radiation Measurement (ARM) and the Tropical Ocean Global Atmosphere-Coupled Ocean Atmosphere Response Experiment (TOGA-COARE) field programs have been created. These improvements have been packaged in the form of a new version of the CCM single column model (SCCM) and have been made available to the community via the World Wide Web.
Scientific investigations into boundary-layer and convection parameterizations have exposed strong simulation sensitivities to the details in how the single column model is forced with observations. These sensitivities can mask signals associated with modifications to physical parameterizations designed to improve the transient characteristics of the solutions. Hack and Pedretti have completed the first of several planned investigations into the sources of uncertainty in the numerical solutions (Hack and Pedretti, 1999). These results illustrate the nature of solution uncertainties that arise from non-linearities in parameterized physics, and demonstrate the need to make use of an ensemble methodology when conducting single column modeling investigations. Recent work suggests that certain systematic biases may be robustly identified in SCM frameworks. Identifying deficiencies in the transient response are likely to be far more problematical.
In particular, unconstrained ensemble solutions can exhibit two or more solution states during various phases of the simulation where in some cases these states oscillate about the ensemble mean solution. (This figure (358KB) shows examples of solution distribution time series for the GCSS and ARM case studies.) During these phases of the solution, the ensemble mean has little physical significance since it reflects the average of two or more largely independent solution states and does not necessarily represent a realizable solution. This multiple attractor behavior is characteristic of highly nonlinear systems, and illustrates the need for the statistical characterization of single column model solutions. The most intriguing property of these solutions is the collapse of multiple states back to a single state suggesting the presence of a strong restoring force in the system, which is believed to be associated with the SCM equilibrium state.
This work also shows how an unconstrained solution can quickly establish a systematic bias, which can lead to fortuitously good solutions for some of the observed fields. This kind of behavior has been observed for several configurations of the SCCM model physics. The development of these thermodynamic biases could be of some value if they were representative of the biases that are established in the global model. However, experience indicates this is not the case, where the large biases associated with unconstrained SCM solutions appear to be a consequence of the lack of large-scale dynamical feedbacks. Thus, errors accumulate as the model is integrated forward in time, and in the longer term, the physics is responding to a forcing that is no longer consistent with the simulated basic state. If the integration is continued for a long enough period of time, non-physical structures can develop resulting in large solution bifurcations. These results bring into question whether the conventional SCM approach to examining physical parameterizations is a well-posed experimental framework, since the simulations would appear to be of little value once multiple states appear in the solution.
In a related area of SCM research, Hack and Pedretti have also modified the CCM3 code to allow the extraction of detailed large-scale forcing time series suitable for driving the SCM. This work has already proven to be of value to the exploration of the SCM solution sensitivity to sampling, as well as to observational uncertainties. Preliminary work suggests that uncertainties associated with 3-hour sampling of the forcing introduce significant solution errors after only a few days. This computational framework will be used to identify optimal sampling strategies for field observations in support of single-column modeling, as well as optimal experimental configurations for single-column models.