Climate Modeling
Research in the Climate Modeling Section (CMS) encompasses a variety of modeling studies of the physical mechanisms governing the global climate system and the numerical techniques required to represent these mechanisms. Scientists carry out research that includes development and testing of new physical parameterizations of a wide range of physical and chemical processes for the Community Climate Model (CCM). Integral to this is the development of new numerical techniques for use in the CCM. CMS scientists compare the simulated climate with observational and analyzed data to evaluate the accuracy and impact of these parameterizations.
James Hack completed an analysis of the CCM3 cloud optical property parameterization, which includes new techniques for diagnosing cloud liquid water content and cloud particle effective radius. This analysis illustrates the improvements in the surface and Top of Atmosphere (TOA) energy budget, improvements in the stationary wave structure, and improvements in the simulated El Niño-Southern Oscillation (ENSO) response associated with the modifications to the cloud scheme. (This figure (53K) shows the El Niño-Southern Oscillation (ENSO) response.) The analysis also demonstrates that the new diagnostic procedure for cloud liquid water does not alter the basic climate sensitivity of the CCM.
Hack, in collaboration with Jeffrey Kiehl and James Hurrell (Climate Analysis Section, CAS), analyzed and documented the simulated hydrologic cycle in CCM3. Other aspects of the simulation (the dynamical circulation and the surface and TOA energy budget) were also analyzed and documented.
Hack completed an analysis of the implied meridional ocean energy transport in the CCM3. This work shows that the marked improvement in the meridional structure of the energy transport, particularly the direction of Southern Hemisphere transport, is primarily related to the introduction of the Zhang and McFarlane deep cumulus scheme in CCM3. As previous studies have suggested, a portion of the response is attributable to changes in the TOA radiation budget. However, a substantial component of the response is associated with sharp decreases in surface latent heat fluxes over regions of deep intertropical convective zone (ITCZ) convection, where these energy budget changes are not reflected at the TOA. This study illustrates the important role of parameterized physics in determining the detailed surface energy budget. (This figure (13K) shows the Annual Mean Implied Meridional Ocean Energy Transport for Community Climate Model (CCM) 2, CCM3, and from Trenberth.)
Byron Boville and Hurrell compared the atmospheric state simulated
by CCM3 with observed monthly sea surface temperatures (SSTs) for
1979-1993 to that simulated in the 300-year Climate System Model
(CSM) control run. They found that the differences
between CCM3 and CSM1 are quite small in most measures of the
atmospheric circulation, consistent with the accurate and
Kiehl, Hack, and Hurrell completed a study of the energy budget of the CCM3. This work compared both the TOA radiation budget of CCM3 with data from the Earth Radiation Budget Experiment (ERBE) and surface fluxes over the oceans with an observationally-based compilation completed by Scott Doney (Oceanography Section, OS), William Large (OS), and Frank Bryan (OS). At the TOA, the shortwave and longwave fluxes are in better agreement with the ERBE data than CCM2. In particular, the shortwave cloud forcing in the extratropical stormtrack regions is in much better agreement with the ERBE data. The seasonal cycle of the TOA cloud radiative effects is also well simulated. The transient response of the outgoing longwave radiation to tropical Pacific SST anomalies associated with the ENSO phenomena agrees well with the observed anomaly response. The anomaly response in the shortwave cloud forcing is too weak in the current model. Finally, the study finds that despite good agreement in the TOA reflected shortwave radiation, the model allows too much shortwave radiation to reach the surface. This bias is indicative of an underestimate of absorbed shortwave radiation in the atmosphere.
Kiehl completed a study of the simulation of the tropical warm pool region with the CSM. The biases in the simulated tropical Pacific SSTs from the CSM are similar to many coupled model results. The warm pool in the western Pacific is too cold and the eastern Pacific cold tongue extends too far west. By comparing the surface energy budget of the uncoupled CCM3, the uncoupled NCAR CSM Ocean Model (NCOM), and the fully coupled CSM, Kiehl concluded that the major cause of the bias in the CSM model was due to an excess of shortwave radiation reaching the ocean surface. Since the dynamical efficiency of ocean heat transport is small in the warm pool region, the coupled system responds by changing the gradient in SSTs. This change in SST gradient accelerates the surface easterlies, which in turn increases the latent heat fluxes in the warm pool region. Thus, the system reduces the overestimated net surface flux (due to the bias in shortwave flux) in the warm pool region in a non-local, non-linear fashion.
Philip Rasch, with J.E. Kristjansson (University of Oslo), developed a new parameterization for the next generation of the CCM and CSM. The parameterization makes a closer connection than earlier formulations between the meteorological processes that determine condensate formation and the condensate amount. The parameterization allows a substantially wider range of variation in condensate amount than in the standard CCM3 and ties the condensate amount to local physical processes. It also allows cloud drops to form prior to the onset of grid-box saturation and can require a significant length of time to convert condensate to a precipitable form or to remove the condensate. The free parameters of the scheme were adjusted to provide reasonable agreement with TOA and surface fluxes of energy. The parameterization was evaluated by a comparison with satellite and in-situ measures of liquid and ice cloud amounts. (This figure (28K) shows the cloud liquid water path for the two model simulations). The large differences between the observational estimates highlight the difficulties and uncertainties associated with the retrieval using the microwave measurements.
Global ice and liquid water burdens are higher in the revised model than in the control simulation, with an accompanying increase in height of the center of mass of cloud water. Zonal averages of cloud water contents are 20 to 50% lower near the surface and much higher above. The range of variation of cloud water contents is much broader in the new parameterization but is still not as large as measurements suggest. Differences in the simulation between the CCM3 with and without prognostic cloud water are generally small. The largest significant changes found in the simulation were seen in polar regions (winter in the Arctic and all seasons in the Antarctic). The new parameterization significantly changes the Northern Hemisphere winter distribution of cloud water and improves the simulation of temperature and cloud amount there.
The new parameterization adds significantly to the flexibility in
the model and the scope of problems that can be attacked. Such a
scheme is needed for a reasonable treatment of scavenging of
atmospheric trace constituents and cloud aqueous or surface
chemistry. The addition of a more realistic condensate
parameterization provides opportunities for a closer connection
between radiative properties of the clouds and their formation and
dissipation. These processes must be treated for many problems of
interest today
David Williamson and Jerry Olson continued to study the source of the relative cold biases seen in the semi-Lagrangian version of CCM3 compared to the Eulerian version. With the standard CCM 18 vertical levels, simulations produced with the semi-Lagrangian approximations develop a colder tropical tropopause and a colder lower polar troposphere than matching simulations with the Eulerian approximations, all other components of the model being the same. They showed that this difference is primarily due to insufficient vertical resolution in the standard 18-level model that has 3 km spacing near the tropopause. They established a minimal vertical grid of 26 levels with which the semi-Lagrangian and Eulerian approximations create the same tropical structure. The additional resolution was added between 200 and 50 mb, giving a grid spacing of about 1.3 km near the tropopause. Both approximations benefit from the increased resolution in that the convective parameterization is better behaved in the upper troposphere in both, and the tropical temperature bias compared to the National Centers for Environmental Prediction (NCEP) reanalysis is reduced in both. They also showed that the Eulerian approximations are prone to stationary gridscale noise if the vertical grid is not carefully designed. The semi-Lagrangian simulation showed no indication of stationary vertical gridscale noise. In addition, the Eulerian simulation exhibited significantly greater transient vertical gridscale noise than the semi-Lagrangian. This noise is believed to cause the excessive cross-tropopause transport and dispersion of tracers found in the lower stratosphere of the middle atmosphere version of CCM2.
The cause of the differences in the polar lower troposphere temperature simulated by semi-Lagrangian and Eulerian approximations was identified to be due to the different way the vertical advection approximations treat vertical structures found at the tops of marginally resolved inversions, when the vertical velocity is reasonably vertically uniform surrounding the top of the inversion. The Eulerian approximations underestimate the cooling that should occur at the top of the inversion. Compared to the NCEP reanalysis, the low vertical resolution Eulerian simulated temperature looks better than the semi-Lagrangian, but those approximations produce that "better" simulated temperature by an incorrect mechanism. For practical applications, the Eulerian approximations require higher vertical resolution below 800 mb than usually used today in climate models, but the coarser resolution is adequate for the semi-Lagrangian approximations.
The semi-Lagrangian codes have been combined with the other improvements made to the CCM3 with the plan of providing the dynamical core for the next released version, CCM4. The resolution of this version will be T63 on the 128x64 point horizontal grid and 26 levels.
Williamson and Olson have begun developing a two-time-level version of the semi-Lagrangian approximations. Two approaches are being tried. The first as developed by Clive Temperton (European Centre for Medium-Range Weather Forecasts, ECMWF) involves incorporation of the Coriolis term directly in the semi-Lagrangian advection. The second makes the Coriolis term semi-implicit on the ends of the trajectory and results in the need to solve a system of equations coupling the Associated Legendre Functions for a given longitudinal wavenumber. The second approach has been completed in the three-time-level model. The two-time-level version should produce the same simulation as the original three-time-level but provide a computational savings of around 25%, which will help offset the higher cost of CCM4 due to the increased vertical resolution.
Boville and Mariana Vertenstein collaborated with Fabrizio Sassi (Atmospheric Chemistry Division, ACD) and Rolando Garcia (ACD) on the development and testing of the next generation of the middle atmosphere version of the CCM and on an associated Lagrangian parcel model for trajectory calculations. The new middle atmosphere model employs a parameterization of a spectrum of gravity waves. An early version of this parameterization is actually contained in CCM3 (although not used), but several refinements to the wave sources and deposition characteristics are being tested to improve the simulation of the upper stratosphere and mesosphere. The model top has been extended from approximately 75 to approximately 85 km with 52 levels. The full three-dimensional semi-Lagrangian dynamics formulation of Williamson and Olson are being used with T63 horizontal truncation and a 128x64 grid. The parcel model is formulated in isentropic coordinates and works on CCM output interpolated to potential temperature surfaces. It can conveniently run on a workstation once the input files have been prepared.
Gordon Bonan continued his work to develop and apply a land surface process model for use with the CCM3 and the CSM. This model accounts for the ecological effects of different vegetation types and the thermal and hydrologic effects of different soil types.
William Collins (visitor, Scripps Institution of Oceanography) has compared satellite observations of visible and near-infrared radiation reflected by the ocean and atmosphere to fluxes calculated with the CCM3. The satellite data and model simulations are global and extend from 1979 to 1987. The satellite data were collected as part of ERBE on the Nimbus-7 satellite. The CCM3 was modified to simulate the satellite radiometers. The long duration and global extent of the observations allow us to test for cloud absorption on climatological space and timescales.
Under clear sky conditions, the observed and modeled spectral albedos are nearly identical. The observations and simulations diverge with increasing cloud albedo and cloud amount, regardless of cloud type or cloud phase. The differences between the model and satellite data occur for all tropical and subtropical ocean regions and exhibit minimal seasonal and interannual variability. These results are consistent with enhanced absorption of shortwave radiation by clouds relative to models of radiative transfer. The results may also be used to test future general circulation model (GCM) parameterizations of radiative transfer that incorporate proposed mechanisms for enhanced absorption. (This figure (110K) shows the ratio of near-infrared reflected radiation to broadband solar reflected radiation measured by Nimbus-7 during 1984, and the ratio of near-infrared reflected radiation to broadband solar reflected radiation calculated from International Satellite Cloud Climatology Program (ISCCP) cloud data for 1984 using the NCAR Column Radiation Model.)
Charles Zender, Kiehl, and Collins, in collaboration with Francisco Valero, Brett Bush, Shelly Pope, Anthony Bucholtz (all of University of California, San Diego), and John Vitko (Sandia National Laboratory), carried out an analysis of the Atmospheric Radiation Measurement (ARM) Enhanced Shortwave Experiment (ARESE) data. These data were collected during a field program carried out in 1995 at the ARM Southern Great Plains site. Surface and aircraft data were analyzed to understand the shortwave radiation budget in clear and cloudy sky conditions. A detailed spectral radiation model designed by Zender was used to compare with the observed fluxes. For cloud conditions, liquid water path measured by a microwave radiometer was used to constrain the cloud optical depth in the model. On the clear days, the measured aerosol optical depth was employed in the model. On clear sky days, there was good agreement between the measured and observed clear sky fluxes. On a day with complete cloud cover, there was a large underestimation of shortwave absorption by the model. This is direct observational evidence that models underestimate shortwave absorption relative to observations. (The two panels in this figure (9K) demonstrate that observed atmospheric absorption in cloudy skies (datapoints) is approximately 50% greater than the absorption predicted by the state-of-the-art radiative transfer theory (dashed and dotted lines).)
Bonan has used the land surface model (LSM), coupled to the CCM3, to study the effects of land-use practices on the climate of the U.S. and the persistence of droughts and floods in the Mississippi River basin. Two major foci for further 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) and (b) inclusion of ecosystem dynamics (in collaboration with Jon Foley, University of Wisconsin).
Boville collaborated with Alex Medvedev and Gary Klaassen (both at York University) on the testing of their gravity wave spectrum parameterization in the mechanistic version of the Middle Atmosphere Community Climate Model (MACCM2). This model, which runs on a workstation, allowed rapid testing of the parameterization and of the sensitivity of the simulation to parameter choices. The parameterization has now been implemented in the Canadian middle atmosphere model.
Williamson and Olson developed a non-interpolating-in-the-vertical
semi-Lagrangian version of CCM3 for use with the standard 18-level
vertical structure. They showed that the computational characteristics
of these approximations are very similar to those of the Eulerian
model, but the approximations provide the computational savings
associated with interpolating semi-Lagrangian schemes. This version
was provided to the Land Surface Group at the Department of
Atmospheric Sciences, University of Arizona, for
Williamson continued to collaborate with Paul Swarztrauber (Scientific Computing Division, SCD) and John Drake (Oak Ridge National Laboratory) developing a Cartesian method for solving the atmospheric equations in spherical geometry. Second order methods were developed in past years. Recently they have examined the application to larger stencils with the goal of developing stable and high order Cartesian methods for solving partial differential equations on irregular spherical grids.
Earlier experiments with CCM2 showed that simulations up to T170 truncation show no indication of convergence. Williamson, Olson, and John Truesdale devised a strategy to look at convergence of the dynamics in the CCM with increased resolution, without the physical forcing feeding back on the additional smaller scales. They developed a version of CCM2 that allows different resolutions for the dynamics and physical parameterizations. Simulations were made with T31 physics coupled with T63 and T106 dynamics and with T42 physics coupled with T63 and T106 dynamics. Preliminary analysis indicates that the simulations do converge in the tropics, where a very strong resolution signal was seen in CCM2, with increasing dynamical resolution. These results imply that the parameterization is incorrect at either the low or the high resolution. The behavior is different in midlatitudes. There the nonlinear dynamical processes continue to introduce power into the higher waves as the dynamical resolution is increased.
Williamson and Erik Kluzek collaborated with scientists at the Program for Climate Model Diagnostics and Intercomparisons (PCMDI) at the Lawrence Livermore National Laboratory (LLNL) and at the Bureau of Meteorology Research Centre (BMRC) to perform a sampling study to establish the importance or necessity of some of the controversial decisions made concerning the standard output of AMIP II. In addition, they performed a sampling study to determine the error associated with different strategies of averaging/interpolating quadratic products. As a result, several fields were eliminated from the requested data as they would not have been comparable with the differing strategies expected to be used by the various modeling groups. A summary of the results will be posted on the AMIP II home page, and the datasets will be available to the AMIP diagnostic sub-projects for further study.
Williamson worked with Karl Taylor (PCMDI/LLNL) and Francis Zwiers (Canadian Centre for Climate Modeling and Analysis) to develop a method of specifying SSTs and sea-ice concentration such that linearly interpolated values between specified mid-month values will yield the correct monthly average. This method is now recommended for the AMIP II simulations. The description is available on the World Wide Web.
Williamson and Olson continued to collaborate with Joe Sela (NCEP) to develop a semi-Lagrangian version of the NCEP global spectral forecast model. Sela is carrying out pre-operational trials at NCEP.
An initial development release of a single-column version of CCM2 and CCM3 has been completed under the direction of Hack, in collaboration with John Pedretti and Jonathan Petch (visitor, University of Reading). A "beta" distribution has been made available via the World Wide Web. This particular framework has been in use for more than one year at NCAR and at selected remote sites and appears to be achieving the goal of facilitating parameterization development for atmospheric GCMs. This work has been sponsored by the Department of Energy's Computer Hardware, Advanced Mathematics and Model Physics (CHAMMP) program.
Kiehl and Timothy Schneider, in collaboration with Susan Solomon
(NOAA Aeronomy Laboratory) and Robert Portmann (NOAA Aeronomy
Laboratory), have calculated the global distribution of radiative
forcing due to changes in both tropospheric and stratospheric
ozone. The present day global distribution of column ozone is obtained
from satellite observations, while the vertical distribution of ozone
is determined from ozone sonde station data. Both the longwave and
shortwave forcings are determined from a three-dimensional diagnostic
radiation model based on the CCM3 radiation scheme. The model also
accounts for cloud effects and variations in surface albedo. The
Solomon-Garcia chemical model is used to determine the pre-industrial
ozone distributions. The global mean ozone forcing from pre-industrial
to present is 0.28 Wm-2. Locally the forcing due to
increases in tropospheric ozone can be larger than
A formulation used to represent processes controlling the atmospheric sulfur cycle has been developed by Rasch, Mary Barth (ACD and Mesoscale and Microscale Meteorology Division, MMM), and Kiehl. The modules have been integrated for the CCM and an associated offline transport model, Model for Atmospheric Transport and Chemistry (MATCH). A number of 5-year runs have been performed with the CCM to diagnose the importance of various processes controlling sulfate aerosols in the atmosphere. (This figure (24K) shows the annual averaged burden of sulfate that has been tagged by the region of origin for year 4 of the CCM3 sulfur simulation.) For example, the aerosol is tagged in various ways to identify the origin of the aerosol. (Is it of anthropogenic or natural origin? Did it occur through processing by clouds or in clear skies? Did the aerosol originate from emissions over Europe or North America?) The model predicts distributions of DMS, SO2, SO4, and H2O2. The model has been compared to observed distributions of SO2 and SO4 near the surface over Europe and the Eastern U.S. The sulfate emissions predicted by this model will be used in the climate of the 20th century simulations with the CSM discussed in the CSM section. (This figure (13K) shows a comparison of surface values of SO2 and SO4 simulated by the model (ordinate) to the observed value (absicca) at a given location.)
Kiehl, Schneider, Rasch, and Barth (ACD) have carried out an analysis of the direct radiative forcing due to the sulfate distributions determined from the version of CCM3 that includes a sulfur chemistry cycle. The sulfate radiative properties of the sulfate aerosols depend on relative humidity to account for hygroscopic growth of aerosol particles. The forcing is calculated within the CCM3 as a diagnostic during the model simulations. The global annual mean forcing, pre-industrial to present, is -0.57 Wm-2. Locally, the direct effect has a magnitude greater than -3 Wm-2 and offsets the greenhouse forcing due to increases in trace gases. The magnitude of the indirect effect is currently being determined. (This figure (19K) shows the global distribution of the direct, annual mean shortwave forcing (W/m2) caused by anthropogenic sulfate aerosols. This forcing was computed diagnostically in a version of the CCM3 incorporating a model of the sulfer chemistry cycle.)