Report on CCSM Atmosphere Model Working Group Meeting

Seventh Annual CCSM Workshop, The Village at Breckenridge

Co-Chairs: Bill Collins, Leo Donner and Dave Randall

26 June 2002

The CCSM Atmosphere Model Working Group (AMWG) breakout session was held on the second day of the June 2002 CCSM Workshop. The ten oral and twelve poster presentations addressed various aspects of the new Community Atmosphere Model (CAM2) released with the rest of CCSM-2 on May 17, 2002. One of the primary objectives of the AMWG breakout session was to introduce the new atmospheric model to the scientific community. This model incorporates enhancements to the dynamics, physics, and boundary data sets, including:

• new and significantly improved sea-surface temperatures for 1949 to present
• thermodynamic sea ice
• fractional ice and land
• updated prognostic cloud water
• generalized cloud overlap in radiation
• more accurate formulation of water-vapor radiative properties
• addition of a finite-volume dynamical core
  

The improvements in the climate simulation include more realistic tropical precipitable water, clear-sky longwave fluxes in polar regions, and more realistic forcing in the tropical Pacific.

The oral talks addressed analysis of systematic errors in the simulated climate, proposals for ameliorating or eliminating these biases, and new methods for diagnosing the model. The systematic errors include:

• a double ITCZ and extended cold tongue
• underestimation of tropical variability on short & MJO timescales
• overestimation of land surface temperatures in winter
• underestimation of tropical tropopause temperatures
• biases in the boundary layer clouds and surface fields for the Eastern Pacific
• sign errors in the cloud forcing response to changes in SST

The manifestations of these biases were demonstrated in uncoupled and coupled simulations and in whole-atmosphere simulations with interactive stratospheric chemistry.

At the April AMWG meeting, the Scientific Steering Committee (SSC) requested that we give highest priority to research on reducing or eliminating these systematic errors. This request was discussed during our meeting in June and has been accepted by the AMWG. The group agreed to pursue this research in close collaboration with other working groups. The AMWG co-chairs have agreed to coordinate the experiments and experimental strategy.

The group also presented plans for longer-range development once the significant model biases have been addressed. These include:

• incorporation of new aerosol species
• new radiation parameterizations
• alternative formulations of convective processes and transport
• new formulations for boundary layer structure and cloud systems
• incorporation of simplified tropospheric chemistry
• testing via assimilation and analysis of forecast errors
• testing analysis of initial error growth
• exploration of alternative formulations of atmospheric dynamics

Oral presentations:

Model diagnostics

Dave Williamson described a new, ARM-funded initiative to characterize errors in CAM and other GCMs by running short forecasts and comparing the model simulations against observations. The idea is to use state-of-the-art operational analyses to initialize the model state, preferably in regions where the analyses can be evaluated against extensive in situ data. Short-term forecasts from climate models initialized with the analyses will then be compared against observations from, for example, ARM sites. Errors in the short-term forecasts should be dominated by errors in physics parameterizations assuming the errors in the analyzed fields are sufficiently small. Dave presented preliminary results from his efforts to run CAM in forecast mode.

Model characteristics:

Chris Bretherton discussed the simulation of the eastern tropical Pacific. He showed that while the simulation of shortwave cloud forcing had improved, the warm bias in SSTs persists in CCSM2. The net energy flux at the ocean surface actually increases in CAM2 relative to CCM3. Chris attributed this increase to an underestimation of the latent heat flux compared to NCEP analyses. The cause appears to be an excessively moist PBL with mixing ratios under stratocumuli as large as 1 g/kg. His conclusion was that the PBL is not entraining enough air from the free troposphere.

Jeff Kiehl presented an analysis of cloud forcing response to SST anomalies from CAM and observational data sets. The analysis was based on a composite of warm and cold events. The model reproduces the changes in precipitation, total cloud amount, and outgoing longwave radiation. However, the sign of the response in absorbed shortwave radiation is incorrect in CAM relative to ERBE. Since the observed cloud response is strongly correlated with changes in high cloud amount, Jeff concluded that the high-cloud ice water concentrations are too low. He recommended moving to prognosing the phase of condensed water as well as its total amount.

Byron Boville addressed the issue of the cold tropopause bias in CAM. He showed that the cold bias of 6K is independent of dynamical core and is present in January and July. He used a radiative equilibrium model to infer the changes in radiative heating required to eliminate the bias and found that an increase of 0.1 K/day between 80-300mb would be sufficient. He also found that increases in condensed water path by 0.2 g/m^2 in high cloud would warm the tropopause by 3K. Based upon these sensitivity studies, he recommended that future modifications include the latent heat of fusion, better continuity of stratiform cloudiness, and explicit prognostic treatment of cloud phase.

Bruce Briegleb used results from the CCSM2 polar atmosphere simulation to suggest ways to reduce the warm surface temperature bias at high latitudes during winter. The results indicate that low cloud amount over the arctic is overestimated by 50% compared to Warren's cloud atlas. The downwelling longwave is overestimated by 20 W/m^2 and the downwelling shortwave underestimated by 40 W/m^2 compared to ECMWF analysis for January. Bruce recommended that we attempt to reduce the low cloud amount to eliminate these biases.

New modeling initiatives:

Leo Donner presented evidence that his convection scheme may alter the simulation of the MJO and ITCZ structure. Early results from the GFDL AM2 model indicate that precipitation patterns are changed considerably with the introduction of the Donner scheme, although the ITCZ was not as well-developed as originally hoped. The scheme differs from other cumulus parameterizations in its inclusion of mesoscale circulations, PDFs of entrainment coefficients, and a tendency closure. Leo plans to continue experimentation with the Donner convection scheme in CAM2.

Fabrizio Sassi discussed preliminary simulations with the Whole Atmosphere Community Climate Model (WACCM) with interactive chemistry. The simulations showed clearly that the bias in tropopause temperatures are producing an unrealistic simulation of the water vapor tape recorder and leading to dehydration of the stratosphere and mesosphere. The resulting biases in HOX can make it difficult to reproduce the mesosphere ozone temperature. Cold biases in the arctic stratosphere lead to denitrification, creation of NAD, and the formation of an ozone hole. In discussion, Byron Boville noted that the bias in polar stratospheric temperatures may result from an underestimation of gravity wave drag, which was tuned for a different dynamical core.

Phil Duffy presented results from high-resolution, untuned integrations of CAM at T170 and T239 resolution. He analyzed the changes in the simulations relative to T42 using Taylor diagrams. Generally, increased resolution reduced pattern errors in all twenty quantities he studied, particularly when the model was run at T239. There was significant improvement in DJF precipitation and much more realistic regional distribution of precipitation over the continental U.S.

Xiaoqing Wu concluded the breakout session with a report on his work with cumulus momentum transport (CMT) in CAM. The zonal-mean distribution of precipitation in the Atlantic and Pacific seems to be more realistic with the incorporation of CMT, and the seasonal migration of precipitation across the equator is apparently closer to observations. Xiaoqing concluded with his plan to extend this work to other GCM modeling efforts.

Participant List