The mission of the Climate Modeling Section (CMS) is to improve understanding of the global
atmosphere and its role in the climate system, through modeling and observational studies,
and to represent that understanding in the form of improved numerical models of the
atmosphere and larger climate system. The principal modeling tool in CMS is the Community
Climate System Model (CCSM) Community
Atmospheric Model (CAM). CAM is the
global atmospheric model employed in the CCSM, where CMS scientists play a primary role in
its ongoing development.
Members of CMS (Byron Boville, William Collins, James Hack, Philip Rasch, and David Williamson), in collaboration with colleagues in the university and national laboratory community, contributed broadly to the development of the next generation of CAM, previously known as the Community Climate Model (CCM). The new model is based on the CAM2, which was publicly released in June of 2002. As was the case for the CAM2, the new model will include a large number of significant improvements and enhancements to the treatment of physical processes, new capabilities such as a Slab Ocean Model (SOM) configuration, and additional improvements to the software engineering implementation of the model. Some of the improvements to the physics include: a substantially revised prognostic cloud water parameterization that includes separate phases for ice and liquid condensate, advection and sedimentation of condensate, and a consistent treatment of condensate in the microphysics and radiative transfer parameterizations. The representation of direct shortwave aerosol forcing is incorporated using an annually repeating aerosol distribution for sulfate, dust, sea salt, and carbonaceous aerosols. Improved representations of shortwave water vapor absorption, longwave absorption, and emission by greenhouse gases are also now included. The latent heat of fusion is included in all aspects of the thermodynamics involving the phase transformation of water substances, where the model now conserves energy exactly in all physical parameterizations. The shallow/frontal convection parameterization now interacts more closely with the prognostic cloud parameterization by detraining condensate directly into the stratiform clouds. The simulation exhibits a number of important improvements with respect to the CAM2 simulation, including a reduction of the warm bias present in the CAM2 arctic simulations, a reduction of the cold bias at the tropical tropopause, and a significantly more realistic cloud response to tropical sea surface temperature (SST) variations. The new CAM represents a concerted effort by CMS to ameliorate or eliminate several of the more important systematic biases identified in the coupled and uncoupled simulations with CAM2. Assuming that this version is accepted by the CCSM Scientific Steering Committee (SSC), it will be incorporated in the version of CCSM used for NCAR's participation in the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report.
As was the case with CAM2, the new model provides three dynamical cores, including the spectral Eulerian dynamical formulation which uses a semi-Lagrangian transport scheme, the NASA Data Assimilation Office (DAO) finite-volume dynamical core that was incorporated via a collaboration with Shian-Jiann Lin (NASA Goddard Space Flight Center, GSFC), and a two-time-level semi-Lagrangian dynamical formulation employing a linear transform grid. The model continues to be capable of running on a variety of computer architectures popular for climate research applications, and is in the process of being extended to once again exploit re-emerging vector architectures.
Hack lead efforts to incorporate a number of useful and important simulation capabilities in the CAM2 including a perpetual configuration allowing for inverse climate sensitivity experiments (so-called Feedback Analysis for GCM Intercomparison and Observations (FANGIO) experimental configuration) and a SOM capability. The SOM extension, developed in collaboration with James Rosinski (CMS), Bruce Briegleb (Oceanography Section), and Cecelia Bitz (University of Washington), exploits a simplified version of the Community Sea Ice Model (CSIM) sea-ice formulation and fractional representation of surface properties that were incorporated into the CAM2. The SOM development exposed weaknesses in the CAM2 cloud scheme that were corrected and will be included in the next generation CAM. Both of these simulation extensions will be essential tools in establishing the climate sensitivity of the uncoupled model, and identifying the underlying feedbacks associated with the physical approximations.
Boville, in collaboration with Christopher Bretherton (University of Washington), developed a more precise definition for energy conservation in the CAM, which significantly improved the conservation properties of the parameterization suite. Boville, in collaboration with Rasch, also made major modifications to the prognostic cloud scheme that have helped to ameliorate the cold tropopause bias in CAM2, as well as the warm surface temperature bias over the Northern Hemisphere polar regions.
Boville, in collaboration with Lin, has also continued his work on incorporating the NASA DAO finite volume (FV) dynamical core in the CAM. He developed a generic energy fixer that is run as a part of the physical parameterization suite. It accounts for kinetic energy dissipation, which should take place within the dynamics, and any other sources/sinks of energy. He has also examined drag and diffusion processes in the FV core in an attempt to address Northern Hemisphere planetary wave biases.
Figure 1. This figure shows that the new CAM is better able to reproduce the tropical fluctuations in absorbed solar radiation with considerable fidelity in timing and amplitude. This is among the simulation improvements exhibited by the new CAM model.
Collins, Julia Lee-Taylor, David Edwards, and Gene Francis (Atmospheric Chemistry Division, ACD) developed a new parameterization for the near-infrared absorption of shortwave radiation by water vapor. Globally water vapor absorbs more visible and near-infrared radiation than any other atmospheric constituent. Under cloud-free conditions, water vapor absorbs approximately 43 W/m2, or 72%, of the global annual-mean incident solar radiation. Collins' work updates the treatment of absorption in CAM so that it is based upon the most up-to-date information on the line and continuum absorption of visible and near-infrared radiation. The earlier parameterized treatment was based on knowledge of water-vapor spectroscopy, dating back to 1982. This new scheme will be used in the version of CCSM used for the upcoming IPCC simulations. In the tropics, the new scheme reduces the surface insolation and increases the atmospheric flux divergence by over 6 W/m2 in the zonal annual mean. Collins et al. also developed a benchmark shortwave line-by-line (LBL) code parallelized for shared-memory supercomputers. Comparison of the new CAM parameterization against calculations with this LBL code shows that the RMS heating rate differences are less than 0.02 Kelvin per day.
Collins, Lee-Taylor, Edwards, and Francis have also updated the longwave parameterization to make use of the latest information regarding water vapor line and continuum absorption of infrared radiation. The main effects are associated with the ongoing evolution of the water vapor continuum. The new scheme is in excellent agreement with the General Line-by-line Atmospheric Transmittance and Radiance Model (GENLN3) LBL code developed in ACD where RMS errors are in the range of 0.02 Kelvin per day. As part of this effort, Collins and Lee-Taylor added OpenMP parallelism to the GENLN3 code, facilitating hyper-spectral treatments of atmospheric radiative transfer by the GENLN user community.
Collins and Andrew Conley (CMS) have added a prescription for many of the major tropospheric aerosol species to CAM. The distributions of the aerosols are generated using the NCAR Model for Atmospheric Transport and Chemistry (MATCH) chemical transport model (CTM) combined with assimilation of satellite aerosol retrievals. The aerosol species include sea salt, soil dust, sulfate, and black and organic carbon. This model data set is input into CAM and used to compute the effects of aerosols on the shortwave radiative fluxes and heating rates. The new aerosol parameterization represents the first time that the effects of soil dust and carbonaceous aerosols have been combined to heat the atmosphere in CAM. Based upon the MATCH results, aerosols represent the third most important source of shortwave diabatic heating, exceeded only by water vapor and ozone. Collins and Conley are investigating the response of the hydrological cycle and atmospheric circulation to the radiative effects of these aerosols using CAM coupled to a slab ocean model.
Hack and Julie Caron (CMS) worked on the development of higher resolution configurations of the CAM, where the principal effort has been the development of a T85 configuration to be used for the upcoming IPCC activity. This work has been focused on understanding the scaling properties of the CAM2 physical parameterization package and led to extensions to the cloud fraction parameterization. Major advantages of the higher resolution configuration are a warmer troposphere, including the tropical tropopause, improved surface stresses and shortwave surface energy budgets in the eastern ocean upwelling regions, improved equatorial surface wind stresses, and a more realistic poleward shift of the southern hemisphere storm track. These features appear to result in significant improvements to coupled model simulations exhibiting reductions in the warm biases off the western coasts of the continents, reductions in southern ocean SST errors, improvements in the near equatorial upper-ocean temperature structures, an improved semi-annual signal of equatorial Pacific SSTs, an improved Pacific equatorial undercurrent due to increased westward wind stress, and improved surface salinity changes in the Arctic.
Hack and Caron have also begun work on developing a more comprehensive understanding of the diurnal cycle of warm season moist convection in CAM at the process level. Presently, the global model does a very poor job of simulating the correct phase and amplitude of convection under conditions of strong diurnal surface forcing. They are currently exploring several modifications to the parameterization of deep convection in the global model that may improve the timing of convective initiation. The global model also fails to simulate the propagation of diurnally-driven convective events, such as the eastward propagation of convection from the Rocky Mountains across the northern great plains during Northern Hemisphere summer. Since horizontal resolution may play an important role in capturing the propagation of these mesoscale structures, Hack and Caron will carefully examine this feature of the simulation at a variety of global resolutions ranging from standard climate resolutions of 300 Km down to resolutions more typically associated with weather forecasting (e.g., 50-75 Km).
Hack and John Truesdale (CMS) have continued to develop new versions of the CCSM CAM Single Column Model (SCAM). This tool has been identified by many members of the CAM atmospheric modeling community as an essential component in our simulation tools hierarchy. The redesign effort updated the single column model graphical user interface, and reconfigured the new CAM code to more easily allow for single-column execution of the physics package. The new implementation more cleanly separates parameterized physics from the global dynamical driver which is designed to minimize the effort required to keep pace with changes to the global model. The latest SCAM design has reduced the number of interface routines from over 150 to about 30, which will be reduced further as work on the redesign of CAM continues. The new version of the model also maintains the ability to produce bit-for-bit agreement with the more complete 3D model, which will allow for a more flexible range of research activities, such as on solution sensitivity to sampling uncertainty with the new CAM physics.
Rasch, in collaboration with Boville and James McCaa (CMS) explored changes to the boundary layer formulation to account for mixing across cloud topped boundary layers associated with radiative cooling. They also explored the prediction of cloud drop radius and its interaction with radiation and sedimentation physics in the stratiform cloud parameterization. The results of the work are promising and are expected to be incorporated in future versions of the CAM.
Rasch, Kerry Emanuel and Dance Zurovac-Jevtic (Massachusetts Institute of Technology) and John Kain (NOAA) have been exploring two alternate parameterizations of convection in recent versions of CAM. While both of these parameterizations produce reasonable simulations for transport of heat and moisture, they produce very different simulations of tracer transport, compared to each other, and to the standard Zhang and McFarlane parameterization of CAM convection.
An example of this behavior is seen in Figure 2.
Figure 2. These panels show the results of a simulation using SCAM. The first set of panels shows a time series of a radon-like tracer (with a source at the surface and rapid radioactive decay) using the model with the three different convection parameterizations. The simulations indicate very different mechanisms for the transport of species. A similar conclusion can be reached when comparing the simulation of an ozone-like tracer (second set of panels) which has a strong source in the stratosphere and sink near the surface.
Rasch, in collaboration with Natalie Mahowald and Samuel Levis (Terrestrial Sciences Section, TSS), worked on modifying the prognostic aerosol package developed by Rasch, Collins, and collaborators to incorporate a new formulation for mobilization of dust that utilizes the physical description of land properties present in the Community Land Model (CLM).
Rasch and Williamson, with Lin, have compared the behavior of the CAM in transporting trace constituents using three numerical formulations for the solution to equations describing the dynamics of the atmosphere. The study demonstrated a very substantial difference between the numerical formulations, with the representation developed by Lin and Ricky Rood (NASA GSFC) showing substantially reduced (and more realistic) cross-tropopause transport of trace constituents.
Williamson and Jerry Olson (CMS) continue to work on developing the capability to apply the CAM in forecast mode. The objective is gain insight into errors in parameterizations by directly comparing parameterized variables such as clouds and precipitation with observations from field campaigns. The examination and comparison are done early in the forecasts while the forecast state is still near the observed atmospheric state. Thus the resolved scales which force the parameterizations are close to the real atmosphere rather than containing the errors present when the model is in its own climate balance. Their approach is to run CAM in forecast mode using historical analyses from National Weather Prediction (NWP) centers for recent years. By using analyses from several centers, information becomes available on the sensitivity of the parameterized variables to the initial conditions. Williamson and Olson developed methods to map the high-resolution NWP analyses of the atmospheric state variables to the coarse climate model grid and to spin-up the land and parameterized atmospheric variables to be consistent with the atmospheric state variables. In collaboration with scientists from the Program for Climate Model Diagnosis and Intercomparison (PCMDI) at Lawrence Livermore National Laboratory (LLNL), Williamson and Olson have carried out a series of forecasts for periods covered by Intensive Observation Periods (IOPs) of the Atmospheric Radiation Measurement (ARM) Program. The analysis to date has concentrated on the Southern Great Plains (SGP) Oklahoma site.
Figure 3. This figure shows examples from two sets of forecasts, one from April 1997 and the other spanning June and July 1997. The top row shows the 24 hour average precipitation for the sequences of one-day forecasts ending on the day indicated on the abscissa. In April the CAM captures the episodic nature of the ARM precipitation very well. The magnitude is about two-thirds of the observed but the CAM data are an average of the four grid boxes surrounding the ARM box and represent a larger area. Although the area of the ARM region is comparable to a single CAM grid box, the center of the ARM box falls very near the vertices of four CAM grid boxes, thus no CAM box coincides with the ARM box.
Examination of the four individual grid boxes shows that a single grid box can capture the magnitude of each event, but no single box captures them all. The southeast box captures the first event, the northwest box captures the second and the northeast captures the third. Although the CAM does well in April, in the summer months of June and July it rains almost all the time with no resemblance to the observed precipitation. The bottom row shows the vertical profiles of the mean of the forecast relative humidity errors as a function of forecast time for five-day forecasts. In June and July the predominant error forms within 24 hours. After 24 hours the CAM is too moist above 500mb, too dry between 900mb and 500mb, and too moist in the surface layers, in both the relative and specific humidity. April shows a slower evolution of the forecast error and a smaller magnitude.
Williamson and Olson, in collaboration with staff members from PCMDI/LLNL will continue to study the processes involved in the error growth during the first day of the forecasts. Comparisons with the ARM observations of other variables are expected to shed light on the workings of the CAM parameterization suite.
Williamson collaborated with Christiane Jablonowski (University of Michigan) to examine a test case she developed for baroclinic dynamical cores involving a growing baroclinic wave. Jablonowski has developed an analytically specified initial condition, which has a mid-latitude jet that is statically stable everywhere and baroclinicly unstable. Since the nonlinear analytic solution for finite perturbations is not known, Jablonowski and Williamson determined the solution numerically with very fine resolution, using several different dynamical cores. The solutions from different dynamical cores converged to the same solution as the resolution increases, providing an estimate of the true solution. However, because the basic state is unstable, small differences, in addition to the imposed initial perturbation, also grow with time. Thus the reference solution can only be determined to within some uncertainty given by the growth of such additional perturbations. They quantified the uncertainty and showed that the cores converge to within that uncertainty with increasing resolution and hit a limit above which higher resolution does not improve the definition of the reference solution. Thus, the test case, as currently posed, can tell us at what resolution a new scheme is as good as standard reference schemes. It also provides information about a numerical approximation's behavior at coarser resolutions.
Williamson is organizing a coordinated Aqua-Planet Experiment (APE) under the auspices of Working Group on Numerical Experimentation (WGNE) with Richard Neale (NOAA Climate Diagnostics Center), Brian Hoskins and Mike Blackburn (University of Reading), and Peter Gleckler (PCMDI/LLNL). The project is intended to provide a benchmark of current model behaviors, and more importantly, to stimulate research to understand the cause of differences arising from different models, different subgrid-scale parameterization suites, different dynamical cores, and different methods of coupling the two. Participants to date include the Meteorological Office, the Commonwealth Scientific and Industrial Research Organization (CSIRO), the Department of Numerical Mathematics at the Russian Academy of Sciences, the Geophysical Fluid Dynamics Laboratory (GFDL), NASA GSFC, the Institute of Atmospheric Physics at the Chinese Academy of Sciences, Frontier Research System for Global Change, the European Centre for Medium-Range Weather Forecasts (ECMWF), NCAR, the Centre for Global Atmospheric Modelling at the University of Reading, the Main Geophysical Observatory in St. Petersburg, and the Institut fur Physik der Atmosphere, Universitat Mainz and Deutscher Wetterdienst. . Details can be found here.
Andrew Gettelman (CMS) collaborated on an inter-comparison and evaluation of radiation models in the upper troposphere and lower stratosphere in conjunction with Qiang Fu (University of Washington), Piers Forster (University of Reading), and Masatomo Fujiwara (Hokkaido University). The study helps to understand the radiative balance of the tropical upper troposphere, as well as to understand differences between various treatments of radiation, including radiation codes from CCM3 and CAM2. In general, the radiation balance of the tropical tropopause region is well understood and modeled. Radiation models for General Circulation Models (GCMs) do not perform significantly differently than detailed models or line-by-line codes; however, the treatment of water vapor is critical for an accurate representation of the radiation balance. This has important implications for climate modeling.
CMS members Boville and Fabrizio Sassi have worked closely with colleagues in ACD and High Altitude Observatory (HAO) to freeze version 1b of the Whole Atmosphere Community Climate Model (WACCM), which is now available for community use from the NCAR Data Portal site. Because WACCM1b does not have interactive chemistry but uses an offline chemistry/transport model to solve for the composition of middle atmosphere, a version is now under development which includes inline solution of chemistry and the calculation of heating rates below 200 nm. Inclusion of those features is critical in order to carry out planned simulations of the effects of the 11-year solar cycle on the whole atmospheric column. This version of the model employs the Finite Volume dynamical core, which is showing very encouraging early results.
A preliminary version of WACCM2 exhibits a very realistic simulation of the composition of the middle atmosphere. Figure 4 compares the simulated "tape recorder" of water vapor (in ppmv) at the Equator to an observed climatology from the The HALogen Occultation Experiment (HALOE). Note that not only the vertical extent of the tape recorder in the two panels is quite similar (a clear improvement compared to earlier simulations), but also the timing of the extreme dry/wet events is well simulated. The ozone column (contour interval is 20 Dobson Unites (DU) from 130 DU to 570 DU) exhibits pronounced ozone depletion at southern latitudes during fall with a minimum ozone column of 130 DU in mid-October. Tropical values follow the observed seasonal cycle and have a realistic minimum of 270 DU. High column values (greater than 500 DU) at northern latitudes during winter illustrate the effects of a dynamically disturbed northern winter in this particular simulation. Figure 5 shows the zonal mean temperature averaged during December, January, and February (in Kelvin; contour interval is 5 K). Compared to earlier simulations in which the tropical tropopause was too cold with excessively dehydrated air rising in the stratosphere, the tropical tropopause temperature is now closer to observations. This important improvement, resulting from changes to the cloud physics parameterizations implemented by Boville and Rasch, allows a more realistic simulation of stratospheric chemistry. Temperature in the lower stratosphere at high latitudes is about 200 K, which is in agreement with observations. This reflects a realistic simulation of the summer to winter circulation in the middle atmosphere. The stratopause is realistically higher in the winter hemisphere than in the summer hemisphere due to the effects of parameterized gravity waves. The summer mesopause is located near 90 km and attains temperatures of 150 K at the South Pole. This is somewhat warmer than in observations, but it can be corrected with a different implementation of the gravity waves parameterization. A version of WACCM2 is expected to be released to the community some time during 2004.
Figure 4. These panels show features of the water vapor distribution from a recent chemistry simulation conducted with WACCM2 (see text).
Figure 5. Zonal mean temperature averaged
during December, January and February (in Kelvin; contour interval is 5 K).
Gettelman, in collaboration with Bill Randel (ACD) and Steve Pawson (NASA GSFC)
continued investigations of water vapor transport and trends in WACCM. This
work is highlighting the importance of different seasons and different regions
in transporting air between the troposphere and stratosphere, and may explain
some of the unattributed observed trends in stratospheric water vapor.
Gettelman, in collaboration with Andrew Dessler and Sun Wong (University of Maryland) is also developing a capability in WACCM to model the stable isotopes of water vapor, in parallel with efforts as part of the larger CCSM isotopic development activity. These efforts should lead to a further understanding of the exchange of air between the troposphere and stratosphere, and the location and role of condensation processes in dehydrating air reaching the stratosphere.
The majority of work conducted in the Climate Modeling Section has focused on climate model development activities with one notable exception, which is a newly formed scientific collaboration with colleagues at GFDL. Members of the Climate Modeling Section, along with Jeffrey Kiehl (Climate Change Research Section, CCR) have been engaged in an ongoing effort to develop a more complete understanding of climate sensitivity as simulated by the CAM and GFDL atmospheric models. Although a principal goal of this collaboration is to understand model sensitivities and the sources of agreement or difference in climate sensitivity, the interactions have involved detailed investigations of the role of fundamental physical processes in the respective models. The two groups have agreed upon standard experimental frameworks for addressing specific scientific questions, and hope to explore many of these questions by examining the regional sensitivity in the two models. This project falls under a new memorandum of understanding between NCAR and NOAA GFDL, and is expected to continue into the indefinite future.
Collins and David Fillmore (CMS) have continued their research into the development of aerosol analyses using CTMs constrained by data assimilation. They have created the first analysis based upon the aerosol retrievals from the NASA Moderate Imaging Spectroradiometer (MODIS) and Multiangle Imaging Spectroradiometer (MISR) satellite instruments. They are currently exploring novel techniques for assimilating the raw data rather than satellite retrievals using analytic adjoints of the radiative transfer equations. Collins has developed a prototype of the first aerosol analysis based upon space-based lidar (light detection and ranging) instruments. The advantage of lidar is to provide direct remote sensing of the vertical profile of aerosols. At present, the vertical profile of aerosols in CTMs is essentially unconstrained by observations, and hence is entirely dependent on resolved and sub-grid vertical transport parameterizations. Accurate determination of the vertical profiles from lidar data will enable better modeling of wet deposition of aerosols (the major sink in the troposphere), the interactions of aerosols and clouds, mixing by frontal systems and storms, and long-range aerosol transport.
Rasch, in collaboration with Peter Hess, Stacy Walters, and Francis Vitt (ACD) initiated an effort to produce an offline CTM embedded within CCSM framework. This model will import meteorological fields from other sources (e.g., National Centers for Evvironmental Prediction, and ECMWF) into the CAM model instead of using internally generated fields. This provides the opportunity to simulate the transport and chemistry of trace constituents (reactive and passive species and aerosols) using meteorological analysis. These distributions can then be compared with measurements on an episodic basis and used in process studies. This model introduces a new capability into the CCSM framework and reduces the software engineering effort required for maintaining multiple chemical transport models within the same institution. It also allows improvements to the consistency of representations of physical processes within the model and allows all components of the climate system to influence the chemical simulations. It is expected that this model will replace MATCH and the Model for OZone and Related chemical Tracers (MOZART) CTMs. Currently progress includes the successful simulation of simple trace constituents within the CCSM framework using analyzed meteorological fields. The tracers include radon, SF6, an ozone-like tracer, and other diagnostic species. An example can be seen in Figure 6, which compares simulations of radon near 200 mb using the original version of CAM and a version of CAM run as an offline model.
Figure 6. This figure compares simulations of radon near 200 mb using the original version of CAM and a version of CAM run as an offline model.
Gettelman in conjunction with Mahowald
is leading a nascent effort to incorporate isotopes in CCSM. The effort is beginning
by adding the capability to model stable isotopes of water vapor. This specific
effort is a collaboration with David Noone (California Institute of Technology)
and Wong. Isotopes of water vapor are valuable paleoclimate indicators (providing
an integrated temperature history of condensation), as well as an important
diagnostic of the hydrologic cycle in the climate system. This effort is expected
to expand to a community effort in the next year.