Documentation of the PMIP models (Bonfils et al. 1998)

PMIP Documentation for CCM3

National Center for Atmospheric Research: Model NCAR CCM3 (T42 L18) 1992

PMIP Representative(s)

Dr. John E. Kutzbach, IES-Center for Climatic Research, University of Wisconsin, 1225 W. Dayton St. Ma-dison, Wisconsin 53706-1695;Phone: 608 262 0392; Fax: 608 262 5964 ; e-mail:

World Wide Web URL:

Model Designation

NCAR CCM3 (T42 L18) 1992

Model Identification for PMIP


PMIP run(s)

0fix, 6fix

Number of days in each month: 31 28 31 30 31 30 31 31 30 31 30 31

Model Lineage

The Atmospheric model is the fourth generation NCAR Community Climate Model (version 3: CCM3). The CCM3 is an evolution of the CCM2 as described by Hack and al (1993,1994) and CCM2 is the historical descendant of the CCM1 model (cf. Williamson et al. 1987 and Hack et al. 1989 ), but with most dynamical and physical parameterizations qualitatively changed. The NCAR CCM2 model is also substantially different from the NCAR GENESIS model.

The model is identical to latest AMIP model exept for different initial conditionsand the Earth's orbital parameters.

Model Documentation

main reference:

Bonan, G.B. (1996). A land surface model (LSM version 1.0) for ecological, hydrological, and atmospheric studies: technical description and user's guide, NCAR Tech. Note NCAR/TN-417+STR, Boulder, CO. 150 pp.

secondary reference(s):

Key documents are the NCAR CCM2 model description by the user's guide by Bath et al. (1992) and by:

Hack, J.J., Boville, B.A., Briegleb, B.P., Kiehl, J.T., Rasch, P.J.,and Williamson, D.L. (1993).

Description of the NCAR Community, Climate Model (CCM2), NCAR Tech. Note NCAR/TN-382+STR, Boulder, CO

Various aspects of the simulated climate with prescribed climatological sea surface temperatures are described by:

Kiehl, J.T. (1994). Sensitivity of a GCM climate simulation to differences in continental versus maritime cloud drop size. J. Geophys. Res. 99, 23107-23115.

Hack, J.J., Boville, B.A., Kiehl, J.T., Rasch, P.J., and Williamson, D.L. (1994). Climate statistics from the National Center for Atmospheric Research community climate model CCM2. J. Geophys. Res. 99, 20785-20813. [11,12]

Other papers that provide details on particular model features include Briegleb (1992) , Briegleb et al. (1986) , and Kiehl and Briegleb (1991) on the radiation parameterizations,Hack (1994) on the convection scheme, Holtslag and Boville (1993) on the simulation ofboundary-layer diffusion, and Williamson and Rasch (1994) on the semi-Lagrangian transport scheme. Model datasets available for analysis atNCAR (including those from the AMIP simulation) are summarized by Williamson (1993) .

Numerical/Computational Properties

Horizontal Representation

Spectral (spherical harmonic basis functions) with transformation to a Gaussian grid for calculation of nonlinear quantities and most of the physics. Advection of water vapor is via shape-preserving semi-Lagrangian transport (SLT) on the Gaussian grid (cf. Williamson and Rasch 1994).

Horizontal Resolution

Like CCM2, the CCM3 is a spectral triangular 42 general circulation model (T42), roughly equivalent to about 2.8 x 2.8 degrees latitude-longitude.

dim_longitude*dim_latitude: 128*64

Vertical Domain

Surface to 2.917 hPa. For a surface pressure of 1000 hPa, the lowest atmospheric level is at a pressure of about 992 hPa.

Vertical Representation

Finite differences in hybrid sigma-pressure coordinates after Simmons and Striifing (1981) , but modified to allow an upper boundary at nonzero (2.917 hPa) pressure. The vertical-differencing formulation conserves global total energy in the absence of sources and sinks. See also Vertical Domain and Vertical Resolution.

Vertical Resolution

The model uses a 18 levels hybrid vertical coordinate that is terrain following at the surface and reduces to a pressure coordinate for levels above 100 mb. Eta values for each level are: 0.004809, 0.01307, 0.03256, 0.06395, 0.09904, 0.1387, 0.1892, 0.2512, 0.3248, 0.4090, 0.5013, 0.5982, 0.6952, 0.7865, 0.8664, 0.9293, 0.9704, 0.9925.

For a surface pressure of 1000 hPa, 4 levels are below 800 hPa and 7 levels are above 200 hPa.

Computer/Operating System

For 0fix and 6fix run, the PMIP simulation was run on the NCAR CRAY-YMP computers using multiple processors in the UNICOS environment.

Computational Performance

For the PMIP experiment, about 7 minutes on a single processor of the NCAR CRAY-YMP computer per simulated day.


The experiment was started from a modern September 1 initial data set and run on the NCAR CRAY-YMP. The modern control used prescribed observed SST's and seaice and a CO2 concentration set at 355ppm.

For 6fix, the experiment was started using modern, spun-up initial conditions obtained from the modern control run.

Time Integration Scheme(s)

A centered semi-implicit time integration scheme (cf. Simmons et al. 1978 ) with an Asselin (1972) frequency filter is used for many calculations, but horizontal and vertical diffusion (see Diffusion), the advection of water vapor by the SLT scheme (see Horizontal Representation), and adjustments associated with convection and large-scale condensation (see Convection and Cloud Formation) are computed implicitly by a

time-splitting procedure. The overall time step is 20 minutes for dynamics and physics, with radiation circulations performed every hour. Cf. Hack et al. (1993) for further details.


Orography is smoothed (see Orography). Because advection of moisture is treated by the SLT scheme (see Horizontal Representation) negative specific humidity values are avoided. In cases where negative mixing ratios would result from application of the countergradient term in the parameterization of nonlocal vertical diffusion of moisture in the planetary boundary layer (PBL) (see Diffusion, Planetary Boundary Layer, and Surface Fluxes), the countergradient term is not calculated. In addition, at each 20-minute time step a "fixer" is applied to the surface pressure and water vapor so that the global average mass and moisture are conserved (cf. Williamson and Rasch 1994) .

Sampling Frequency

For the PMIP simulation, the model history is written every 6 hours.

Dynamical/Physical Properties

Atmospheric Dynamics

Primitive-equation dynamics are expressed in terms of vorticity, divergence, temperature, specific humidity, and the logarithm of surface pressure. Virtual temperature is used where applicable, and frictional/diffusive heating is included in the thermodynamic equation.


In the troposphere, linear biharmonic (Ñ4) horizontal diffusion (with coefficient 1 x 1016 m4/s) is applied to divergence and vorticity on hybrid sigma-pressure surfaces, and to temperature on first-order constant pressure surfaces (requiring that biharmonic diffusion of surface pressure also be calculated on the Gaussian grid). In the stratosphere linear second-order (Ñ2) diffusion is applied to the same variables at the top three levels (with diffusivities increasing with height from 2.5 x 105 to 7.5 x 105 m2 /s). In the top model layer, diffusion is enhanced by a factor of 103 on all spectral wave numbers that violate the Courant-Friedrichs-Lewy (CFL) numerical stability criterion, based on the maximum wind speed.

Above the PBL (see Planetary Boundary Layer) a second-order, stability-dependent local formulation of the vertical diffusion of momentum, heat, and moisture is adopted (cf. Smagorinsky et al. 1965 ). The mixing length is taken to be a constant 30 m, and the diffusivity is as given by Williamson et al. (1987) for unstable and neutral conditions and by Holtslag and Beljaars (1989) for stable conditions. Above the surface layer, but within the PBL under unstable conditions, mixing of heat and moisture (but not of momentum) is formulated as nonlocal diffusion, following Holtslag and Boville (1993) --see Surface Fluxes.

Horizontal and vertical diffusion are calculated implicitly via time splitting apart from the solution of the semi-implicit dynamical equations (see Time Integration Scheme(s)).

Gravity-wave Drag

Orographic gravity-wave drag is generalized and parameterized after McFarlane (1987) . The momentum drag is given by the vertical divergence of the wave stress, which is proportional to the product of the local squared amplitude of the gravity wave, the Brunt-Vaisalla frequency, and the component of the local wind that is parallel to the flow at a near-surface reference level. At this reference level, the wave amplitude is bound by the lesser of the subgrid-scale orographic variance (see Orography) or a wave-saturation value defined by the reference Froude number. Above this level, the gravity-wave stress is assumed to be constant with height (zero vertical divergence), except in regions of wave saturation, where the amplitude is obtained from the local Froude number.

Solar Constant/Cycles

The solar constant is the AMIP-prescribed value of 1365 W/(m2). The orbital parameters and seasonal insolation distribution are calculated after PMIP recommendations. Both seasonal and diurnal cycles in solar forcing are simulated over a repeatable solar year of exactly 365 days (i.e., leap years are not included). The basic CCM3 code uses an algorithm that specifies only modern orbital conditions. This code was revised by R. Gallimore (UW-Madison) to allow for determination of 6K orbital parameters.


The carbon dioxide concentration set is at 355 ppm for modern control and 288 ppm for 6fix. Monthly ozone volume mixing ratios derived by Chervin (1986) from analyses of Dütsch (1978) are linearly interpolated to obtain intermediate values every 12 hours. Radiative effects of oxygen and of water vapor are also included (see Radiation). There is an incorporation of trace gases in the longwave radiation (CH4, N2O, CFC11, CFC12) and an incorporation of background aerosol in the CCM3 version.


The principal changes from the CCM2 are related to the cloud, radiation, convection, and boundary layer parameterizations.

There is an incorporation of radiative properties of ice clouds. Shortwave scattering/absorption is parameterized by the delta-Eddington approximation of Joseph et al. (1976) and Coakley et al. (1983) applied in 18 spectral intervals, as described by Briegleb (1992) . (These include 7 intervals between 0.20 and 0.35 micron to capture ozone Hartley-Huggins band absorption and Rayleigh scattering; 1 interval between 0.35 to 0.70 micron to capture Rayleigh scattering and ozone Chappius-band and oxygen B-band absorption; 7 intervals between 0.70 and 5.0 microns to capture oxygen A-band and water vapor/liquid absorption; and 3 intervals between 2.7 and 4.3 microns to capture carbon dioxide absorption.) Following Slingo (1989) , the shortwave optical properties of clouds for the delta-Eddington approximation (optical depth, single-scattering albedo, asymmetry factor) are specified for 4 spectral ranges (with boundaries at 0.25, 0.69, 1.19, 2.38, and 4.0 microns). These properties depend on the specified effective droplet radius (10 microns) and the liquid water path (LWP), which is a prescribed nonlinear function of latitude and height (cf. Kiehl et al. 1994) .

Longwave absorption by ozone and carbon dioxide is treated by a broad-band absorptance technique, following Ramanathan and Dickinson (1979) and Kiehl and Briegleb (1991) . A Voigt line profile (temperature) dependence is added to the pressure broadening of the absorption lines. Absorption by water vapor (and its overlap with that of ozone and carbon dioxide) are modeled as in Ramanathan and Downey (1986) . Longwave broad-band emissivity of clouds is a negative exponential function of LWP, with all clouds assumed to be randomly overlapped in the vertical. Cf. Hack et al. (1993) for further details. See also Cloud Formation.


If the atmosphere is moist adiabatically unstable, temperature/moisture column profiles are adjusted by a mass-flux convective parameterization (cf. Hack 1994) . The scheme utilizes a three-layer model that provides for convergence and entrainment in the lowest subcloud layer, cloud condensation and rainout in the middle layer, and limited detrainment in the top layer. This scheme is applied by working upward from the surface on three contiguous layers, and shifting up successively one layer at a time until the whole column is stabilized.

The parameterization is based on simplified equations for the three-layer moist static energy that include (among other terms) the convective mass flux, a "penetration parameter" beta (ranging between 0 and 1) that regulates the detrainment of liquid water, and temperature and moisture perturbations furnished by the PBL parameterization (see Planetary Boundary Layer, Diffusion, and Surface Fluxes). Other free parameters in the scheme include minimum values for beta, for the vertical gradient of moist static energy, and for the depth of precipitating convection; a characteristic convective adjustment time scale; and a cloud-water to rain-water autoconversion coefficient. The parameter beta is determined by iteration, subject to constraints that it and the vertical gradient of moist static energy be at least their minimum values, that the convective mass flux be positive, and that the detrainment layer not be supersaturated. The profiles of convective mass flux, temperature, and moisture are then obtained, and the total convective precipitation rate is calculated by vertical integration of the convective-scale liquid water sink.

If a layer in the stratosphere (i.e., at the top three vertical levels) is dry adiabatically unstable, the temperature is adjusted so that stability is restored under the constraint that sensible heat be conserved. Whenever two layers undergo this dry adjustment, the moisture is also mixed in a conserving manner. (In the model troposphere, vertical diffusion provides stabilizing mixing, and momentum is mixed as well--see Diffusion). If a layer is supersaturated but stable, nonconvective condensation and precipitation result (see Precipitation).

Cloud Formation

The cloud prediction formulation is as described by Kiehl et al. (1994) . Cloud amount is diagnostically determined from relative humidity, vertical velocity, atmospheric stability, and the convective precipitation rate, following a modified Slingo (1987) approach. Some modifications to the cloud fraction parametrization include a new convective cloud scheme and modified layered cloud scheme.

Convective cloud base and top are determined by the vertical extent of moist instability (see Convection). In each vertical column, the total fractional cloud amount is a logarithmic function of the convective precipitation rate, but is constrained to be between 0.2 and 0.8. The convective cloud fraction in each layer is determined assuming the cloud is distributed randomly in the vertical. For subsequent diagnosis of the fractional amount of nonconvective cloud (see below), the layer relative humidity is reduced proportional to the fraction of convective cloud present.

In regions of upward vertical motion, the fraction of low-level layer cloud is a quadratic function of the difference between the reduced relative humidity (see above) and a constant threshold value (90 percent). The fraction of midlevel and high-level layer cloud is a quadratic function of the difference between the reduced relative humidity and a threshold value that is a linear function of the squared Brunt-Vaisalla frequency (i.e., it is proportional to the vertical stability).

The fraction of marine stratus/stratocumulus is a function of the strength of the associated low-level inversion and the reduced relative humidity. Cf. Hack et al. (1993) for further details. See also Radiation. There are improved diagnosis of cloud optical properties (maritime versus continental effective radius (Kiehl, 1994), liquid water path) and an incorporation of evaporation of stratiform precipitation.


There is a modified moist convection by the incorporation of the Zhang and McFarlane (1995) moist convection scheme (see Convection). Grid-scale precipitation forms as a result of supersaturation under stable conditions. In this case, the moisture is adjusted so that the layer is just saturated, with the excess condensing as precipitation; the layer temperature is adjusted according to the associated latent heat release. (Moisture and temperature are mutually adjusted in two iterations.) Subsequent evaporation of falling precipitation is not simulated. Cf. Hack et al. (1993) for details.

Planetary Boundary Layer

The PBL height is determined by iteration at each 20-minute time step following the formulation of Troen and Mahrt (1986) ; the diagnosis of the boundary layer height is in the non-local atmospheric boundary layer scheme. Within the PBL, there is nonlocal diffusion of heat and moisture after Holtslag and Boville (1993) ; otherwise (and under all conditions for momentum), properties are mixed by the stability-dependent local diffusion that applies in the model's free atmosphere. See also Diffusion and Surface Fluxes.


For 0fix and 6fix, the raw orography is obtained from the U.S. Navy dataset with resolution of 10 minutes arc on a latitude/longitude grid (cf. Joseph 1980 ). These data are area-averaged to a 1 x 1-degree grid, interpolated to a T119 Gaussian grid, spectrally truncated to the model's T42 Gaussian grid, and then spectrally filtered to reduce the amplitude of the smallest scales.

The subgrid-scale orographic variances required for the gravity-wave drag parameterization (see Gravity-wave Drag) are also obtained from the U.S. Navy dataset. For the spectral T42 model resolution, the variances are first evaluated on a 2 x 2-degree grid, assuming they are isotropic. Then the variances are binned to the T42 Gaussian grid (i.e., all values whose latitude and longitude centers fall within each Gaussian grid box are averaged together), and are smoothed twice with a 1-2-1 spatial filter. Values over ocean are set to zero.


AMIP monthly sea surface temperature fields are prescribed the modern control and the 6ka experiment, with intermediate values determined at every 20-minute time step by linear interpolation.

Sea Ice

AMIP monthly sea ice extents are prescribed for control and the 6 ka experiments, with intermediate values determined at every 20-minute time step by linear interpolation. The temperature of the ice is predicted by the same four-layer scheme as used for soil temperature (see Land Surface Processes), but with a fixed temperature (-2 degrees C) of the underlying ocean rather than a zero-flux condition, as the lower boundary condition. The four layer thicknesses are all 0.5 m, and the ice density, heat capacity, and conductivity are specified uniform constants; however, daily snow cover that is prescribed from climatology (see Snow Cover) alters the thermodynamic properties and thickness of the top layer in proportion to the relative mass of snow and ice. Cf. Hack et al. (1993) for further details.

Surface Characteristics, Surface Fluxes, Land Surface Processes, Snow Cover

Land surface processes are simulated by the LSM land surface model, as described by Bonan (1996).

LSM replaces the CCM2 specification of surface wetness, prescribed snow cover and prescribed surface albedos. LSM also replaces CCM2 fluxes over land with a parameterization that includes hydrological and ecological processes ( e.g. soil water, phenology, stomatal physiology, interception of water by plants).

The land surface model (LSM version 1) is a one-dimensional model of energy, momentum, water, and CO2 exchange between the atmosphere and land, accounting for ecological differences among vegetation types, hydraulic and thermal differences among soil types, and allowing for multiple surface types including lakes and wetlands within a grid cell. Vegetation effects are included by allowing for twelve plant types that differ in leaf and stem areas, root profile, height, leaf dimension, optical properties, stomatal physiology, roughness length, displacement height, and biomass. These 12 plant types are combined to form 28 different vegetated surfaces, each comprised of multiple plant types and bare ground so that, for example, a mixed broadleaf deciduous and needleleaf evergreen forest consists of patches of broadleaf deciduous trees, needleleaf evergreen trees, and bare ground. Lakes and wetland, if present, form additional patches. Soil effects are included by allowing thermal properties (heat capacity, thermal conductivity) and hydraulic properties (porosity, saturated hydraulic conductivity, saturated matric potential, slope of retention curve) to vary as functions of percent sand and percent clay. Soils also differ in color, which affects soil albedos. Consequently, each grid cell in the domain of interest is assigned a surface type, a fraction covered by lakes, a fraction covered by wetlands, a soil texture (percent sand, percent silt, percent clay), and a soil color.

Major features of the model are:

* prescribed time-varying leaf and stem areas

* absorption, reflection, and transmittance of solar radiation, accounting for the different optical properties of vegetation, soil, water, snow, and ice

* absorption and emission of longwave radiation allowing for emissivities less than one sensible and latent heat fluxes, partitioning latent heat into canopy evaporation, soil evaporation, and transpiration

* turbulent transfer above and within plant canopies

* vegetation and ground temperatures that balance the surface energy budget (net radiation, sensible heat, latent heat, soil heat)

* stomatal physiology and CO2 fluxes

* interception, throughfall, and stemflow

* snow hydrology

* infiltration and runoff

* temperatures for a six-layer soil column using a heat diffusion equation that accounts for phase change

* soil water for the same six-layer soil column using a one-dimensional conservation equation that accounts for infiltration input, gravitational drainage at the bottom of the column, evapotranspiration losses, and vertical water flow based on head gradients temperatures for six-layer deep and shallow lakes accounting for eddy diffusion and convective mixing

In coupling to the atmospheric model, the land surface model provides to the atmospheric model, at every time step, surface albedos (direct beam and diffuse for visible and near-infrared wavebands), upward longwave radiation, sensible heat flux, latent heat flux, water vapor flux, and surface stresses. The atmospheric model provides to the land model, at every time step, incident solar radiation (direct beam and diffuse for visible and near-infrared wavebands), incident longwave radiation, convective and large-scale precipitation, and lowest model level temperature, wind, specific humidity, pressure, and height

Last update November 9, 1998. For further information, contact: Céline Bonfils ( )