With standard namelist settings, both the longwave and shortwave heating rates are evaluated every model hour. Between hourly evaluations, the longwave and shortwave fluxes and flux divergences are held constant.
In CAM 3.0, insolation is computed using the method of Berger [17]. Using this formulation, the insolation can be determined for any time within years of 1950 AD. This facilitates using CAM 3.0 for paleoclimate simulations. The insolation at the top of the model atmosphere is given by
We represent the annual and diurnal cycle of solar insolation with a repeatable solar year of exactly 365 days and with a mean solar day of exactly 24 hours, respectively. The repeatable solar year does not allow for leap years. The expressions defining the annual and diurnal variation of solar insolation are:
(4.175)  
(4.176)  
(4.177)  
(4.178)  
where
 
(4.179)  
The hour angle in the expression for depends on the calendar day as well as model longitude:
The obliquity may be approximated by an empirical series expansion of solutions for the Earth's orbit
(4.181) 
Since the series expansion for the eccentricity is slowly convergent, it is computed using
(4.182) 
(4.183) 
(4.184) 
The general precession is given by another empirical series expansion
(4.185) 
The calculation of requires first determining two mean
longitudes for the orbit. The mean longitude
at the
time of the vernal equinox is :
(4.186)  
(4.187) 
(4.188)  
The orbital state used to calculate the insolation is held fixed over the length of the model integration. This state may be specified in one of two ways. The first method is to specify a year for computing . The value of the year is held constant for the entire length of the integration. The year must fall within the range of . The second method is to specify the eccentricity factor , longitude of perihelion , and obliquity . This set of values is sufficient to specify the complete orbital state. Settings for AMIP II style integrations under 1995 AD conditions are , , and .
The solar spectrum is divided into 19 discrete spectral and pseudospectral intervals (7 for , 1 for the visible, 7 for , 3 for , and 1 for the nearinfrared following Collins [37]). The CAM 3.0 model atmosphere consists of a discrete vertical set of horizontally homogeneous layers within which radiative heating rates are to be specified (see Figure 3.1). Each of these layers is considered to be a homogeneous combination of several radiatively active constituents. Solar irradiance, surface reflectivity for direct and diffuse radiation in each spectral interval, and the cosine of the solar zenith angle are specified. The surface albedo is specified in two wavebands (0.20.7 m, and 0.75.0 m) and distinguishes albedos for direct and diffuse incident radiation. Albedos for ocean surfaces, geographically varying land surfaces, and sea ice surfaces are distinguished.
The method involves evaluating the Eddington solution for the reflectivity and transmissivity for each layer in the vertical under clear and overcast conditions. The layers are then combined together, accounting for multiple scattering between layers, which allows evaluation of upward and downward spectral fluxes at each interface boundary between layers. This procedure is repeated for each spectral or pseudospectral interval and binary cloud configuration (see ``Cloud vertical overlap'' below) to accumulate broad band fluxes, from which the heating rate can be evaluated from flux differences across each layer. The Eddington scheme is implemented so that the solar radiation is evaluated once every model hour (in the standard configuration) over the sunlit portions of the model earth.
The Eddington approximation allows for gaseous absorption by , , , and . Molecular scattering and scattering/absorption by cloud droplets and aerosols are included. With the exception of , a summary of the spectral intervals and the absorption/scattering data used in the formulation are given in Briegleb [27] and Collins [37]. Diagnostic cloud amount is evaluated every model hour just prior to the solar radiation calculation.
The absorption by water vapor of sunlight between 1000 and 18000 cm is treated using seven pseudospectral intervals. A constant specific extinction is specified for each interval. These extinctions have been adjusted to minimize errors in heating rates and flux divergences relative to linebyline (LBL) calculations for reference atmospheres [3] using GENLN3 [57] combined with the radiative transfer solver DISORT2 [163]. The coefficients and weights have the same properties as a kdistribution method [98], but this parameterization is essentially an exponential sum fit (e.g., Wiscombe and Evans [191]). LBL calculations are performed with the HITRAN2k line database [153] and the Clough, Kneizys, and Davies (CKD) model version 2.4.1 [33]. The Rayleigh scattering optical depths in the seven pseudospectral intervals have been changed for consistency with LBL calculations of the variation of watervapor absorption with wavelength. The updated parameterization increases the absorption of solar radiation by water vapor relative to the treatment used in CCM and CAM since its introduction by Briegleb [27].
For some diagnostic purposes, such as estimating cloud radiative forcing [94] a clearsky absorbed solar flux is required. In CAM 3.0, the clearsky fluxes and heating rates are computed using the same vertical grid as the allsky fluxes. This replaces the 2layer diagnostic grid used in CCM3.
The treatment of aerosols in CAM 3.0 replaces the uniform background boundarylayer aerosol used in previous versions of CAM and CCM. The optics for the globally uniform aerosol were identical to the sulfate aerosols described by Kiehl and Briegleb [88]. In the visible, the uniform aerosol was essentially a conservative scatterer. The new treatment introduces five chemical species of aerosol, including sea salt, soil dust, black and organic carbonaceous aerosols, sulfate, and volcanic sulfuric acid. The new aerosols include two species, the soil dust and carbonaceous types, which are strongly absorbing in visible wavelengths and hence increase the shortwave diabatic heating of the atmosphere.
The threedimensional timedependent distributions of the five aerosol species and the optics for each species are loaded into CAM 3.0 during the initialization process. This provides considerable flexibility to:
In its present configuration, CAM includes the direct and semidirect effects of tropospheric aerosols on shortwave fluxes and heating rates. The first indirect effect, or Twomey et al. [175] effect, is not included in the standard version of CAM 3.0.
The annuallycyclic tropospheric aerosol climatology consists of threedimensional, monthlymean distributions of aerosol mass for:
The climatology is produced using an aerosol assimilation system [39,41] integrated for presentday conditions. The system consists of the Model for Atmospheric Chemistry and Transport (MATCH) [142] and an assimilation of satellite retrievals of aerosol optical depth. MATCH version 4 is integrated using the National Centers for Environmental Prediction (NCEP) meteorological reanalysis at T63 triangular truncation [83]. The satellite estimates of aerosol optical depth are from the NOAA Pathfinder II data set [167].
The formulation of the sulfur cycle is described in Barth et al. [12] and Rasch et al. [143]. The emissions inventory for SO is from Smith et al. [162]. The sources for mineral dust are based upon the approach of Zender et al. [196] and Mahowald et al. [120]. The emissions of carbonaceous aerosols include contributions from biomass burning [113], fossil fuel burning [42], and a source of natural organic aerosols resulting from terpene emissions. The vertical profiles of sea salt are computed from the 10m wind speed [21].
The monthlymean mass path for each aerosol species in each layer is computed in units of kg/m. During the initialization of CAM 3.0, the climatology is temporally interpolated from monthlymean to midmonth values. At each CAM 3.0 time step, the midmonth values bounding the current time step are vertically interpolated onto the pressure grid of CAM 3.0 and then time interpolated to the current time step. The interpolation scheme in CAM 3.0 preserves the aerosol masses for each species to 1 part in 10 relative to the climatology, and it is guaranteed to yield positive definite massmixing ratios for all aerosols.
The stratospheric volcanic aerosols are treated using a single species in the standard model. Zonal variations in the stratospheric mass loading are omitted. The volcanic input consists of the monthlymean masses in units of kg/m on an arbitrary meridional and vertical grid. The time series for the recent past is based upon Ammann et al. [2] following Stenchikov et al. [166].
The three intrinsic optical properties stored for each of the eleven aerosol types are specific extinction, single scattering albedo, and asymmetry parameter. These properties are computed on the band structure of CAM 3.0 using Chandrasekhar weighting with spectral solar insolation. The aerosol types affected by hygroscopic growth are sulfate, sea salt, and hydrophilic organic carbon. In previous versions of CCM and CAM 3.0, the relative humidity was held constant in calculations of hygroscopic growth at 80%. In CAM 3.0, the actual profiles of relative humidity computed from the model state each radiation time step are used in the calculation.
The optics for black and organic carbon are identical to the optics for soot and watersoluble aerosols in the Optical Properties of Aerosols and Clouds (OPAC) data set [69]. The optics for dust are derived from Mie calculations for the size distribution represented by each size bin [196]. The Mie calculations for sulfate assume that it is comprised of ammonium sulfate with a lognormal size distribution. The dry size parameters are a median radius of 0.05 m and a geometric standard deviation of 2.0. The optical properties in the seven HO pseudospectral intervals are averaged consistently with LBL calculations of the variation of watervapor absorption with wavelength. This averaging technique preserves the cross correlations among the spectral variation of solar insolation, water vapor absorption, and the aerosol optical properties. The volcanic stratospheric aerosols are assumed to be comprised of 75% sulfuric acid and 25% water. The lognormal size distribution has an effective radius of 0.426 m and a standard deviation of 1.25.
The bulk formulae of Cess [30] are used to combine the optical properties of the individual aerosol species into a single set of bulk aerosol extinctions, singlescattering albedos, and asymmetry parameters for each layer.
CAM 3.0 includes a mechanism to scale the masses of each aerosol species by userselectable factors at runtime. These factors are global, timeindependent constants. This provides the flexibility to consider the climate effects of an arbitrary combination of the aerosol species in the climatology. It also facilitates simulation of climates different from presentday conditions for which the only information available is the ratio of globally averaged aerosol emissions or atmospheric loadings. A mechanism to scale the carbonaceous aerosols with a timedependent unitless factor has been included to facilitate realistic simulations of the recent past.
CAM 3.0 also includes a runtime option for computing a diagnostic set of shortwave fluxes with an arbitrary combination of aerosols multiplied with a separate set of userselectable scale factors. This option can be used to compute, for example, the aerosol radiative forcing relative to an atmosphere containing no aerosols.
The diagnostic fields produced the aerosol calculation include the columnintegrated optical depth and columnaveraged singlescattering albedo, asymmetry parameter, and forward scattering parameter (in the Eddington approximation) for each aerosol species and spectral interval. These fields are only computed for illuminated grid points, and for nonilluminated points the fields are set to zero. The fraction of the time that a given grid point is illuminated is also recorded. Time averages of, for example, the optical depth can be obtained by dividing the timeaveraged optical depths in the history files by the corresponding daylit fractions.
The option of introducing a globally uniform background sulfate aerosol is retained, although by default the optical depth of this aerosol is set to zero. Its optical properties are computed using the same sulfate optics as are used for the aerosol climatology. However, for consistency with the uniform aerosol in previous versions of CAM 3.0 and CCM3, the relative humidity used to compute hygroscopic growth is set to 80%.
An ice particle effective radius, , is also diagnosed by CAM 3.0. Following Kristjánsson and Kristiansen [97], the effective radius for ice clouds is now a function only of temperature, as shown in Figure 4.2.

For cloud scattering and absorption, the radiative parameterization of Slingo [160] for liquid water droplet clouds is employed. In this parameterization, the optical properties of the cloud droplets are represented in terms of the prognosed cloud water path (CWP, in units of kg m) and effective radius , where is the cloud drop size distribution as a function of radius .
Cloud radiative properties explicitly account for the phase of
water. For shortwave radiation we use the following generalization of
the expression used by Slingo [160] for liquid water clouds. The
cloud liquid optical properties (extinction optical depth, single
scattering albedo, asymmetry parameter and forward scattering
parameter) for each spectral interval are defined as
The radiative properties of ice cloud are defined
by
The treatment of cloud vertical overlap follows Collins [38]. The overlap parameterization is designed to reproduce calculations based upon the independent column approximation (ICA). The differences between the results from the new parameterization and ICA are governed by a set of parameters in the shortwave code (Table 4.1 on page and section 4.8.9). The differences can be made arbitrarily small with appropriate settings of these parameters. The current parameter settings represent a compromise between computational cost and accuracy.
The new parameterizations can treat random, maximum, or an arbitrary combination of maximum and random overlap between clouds. The type of overlap is specified with the same two variables for the longwave and shortwave calculations. These variables are the number of randomoverlap interfaces between adjacent groups of maximallyoverlapped layers and a vector of the pressures at each of the interfaces. The specification of the overlap is completely separated from the radiative calculations, and if necessary the type of overlap can change at each grid cell or time step.
The algorithm for cloud overlap first converts the vertical profile of partial cloudiness into an equivalent collection of binary cloud configurations. Let be the fractional amount of cloud in layer in a profile with layers. The index corresponds to the top of the model atmosphere and corresponds to the layer adjacent to the surface. Let be the number of maximallyoverlapped regions in the column separated by randomoverlap boundaries. If the entire column is maximally overlapped, then , and if the entire column is randomly overlapped, then . Each region includes all layers between and . Within each region, identify the unique, nonzero cloud amounts and sort them into a descending list with . Note than in CAM 3.0, cloud amounts are not allowed to be identically equal to 1. It is convenient to define and . By construction for .
The binary cloud configurations are defined in terms of the sorted cloud amounts. The number of unique cloud binary configurations in region is . The binary cloud configuration in region is given by
The cloud overlap for radiative calculations in CAM 3.0 is
maximumrandom (M/R). Clouds in adjacent layers are maximally
overlapped, and groups of clouds separated by one or more clear layers
are randomly overlapped. The two overlap parameters input to the
radiative calculations are the number of randomoverlap interfaces,
which equals , and a vector of pressures at each
randomoverlap interface. These parameters are determined for each
grid cell at each radiation time step. Suppose there are
groups of vertically contiguous clouds in a given grid cell. The
first parameter
. Let represent the pressure
at the bottom interface of each group of contiguous clouds, and let
denote the surface pressure. Both and increase from
the top of the model downward. Then
For diagnostic purposes, the CAM 3.0 calculates three levels of cloud fraction assuming the same maximumrandom overlap as in the radiative calculations. These diagnostics, denoted as low, middle, and high cloud, are bounded by the pressure levels to 700 mb, 700 mb to 400 mb, and 400 mb to the model top.
The solution for the shortwave fluxes is calculated by determining all possible arrangements of binary clouds which are consistent with the vertical profile of partial cloudiness, the overlap assumption, and the parameters for accelerating the solution (Table 4.1 and section 4.8.9). The shortwave radiation within each of these configurations is calculated using the same Eddington solver introduced in CCM3 [27]. The allsky fluxes and heating rates for the original profile of partial cloudiness are calculated as weighted sums of the corresponding quantities from each configuration. The weights are equal to the horizontal fractional area occupied by each configuration. The number of configurations is given by eqn. (4.200), and the area of each configuration is given by eqn. (4.202). There are two steps in the calculations: first, the calculation of the cloudfree and overcast radiative properties for each layer, and second the combination of these properties using the adding method to calculate fluxes. These two processes are described below.
Details of the implementation are as follows. The CAM 3.0 model atmosphere is divided into layers in the vertical; an extra top layer (with index 0, above the layers specified by CAM 3.0) is added. This extra layer prevents excessive heating in the top layer when the top pressure is not very low; also, as the model does not specify absorber properties above its top layer, the optical properties of the top layer must be used for the extra layer. In CAM 3.0, clearsky and allsky solar fluxes are calculated and output for the top of model (TOM) at layer 1 and the top of atmosphere (TOA) corresponding to layer 0. The TOM fluxes are used to compute the model energetic balance, and the TOA fluxes are output for diagnostic comparison against satellite measurements. The provision of both sets of fluxes is new in CAM 3.0. Layers are assumed to be horizontally and vertically homogeneous for each model grid point and are bounded vertically by layer interfaces. For each spectral band, upward and downward fluxes are computed on the layer interfaces (which include the surface and top interface). The spectral fluxes are summed and differenced across layers to evaluate the solar heating rate. The following discussion refers to each of the spectral intervals.
In general, several constituents absorb and/or scatter in each homogeneous layer (e.g. cloud, aerosol, gases...). Every constituent is defined in terms of a layer extinction optical depth , single scattering albedo , asymmetry parameter , and the forward scattering fraction . To define bulk layer properties, the combination formulas of Cess [30] are used:
The Eddington solution for each layer requires scaled properties for , , , given by the expressions:
To combine layers, it is assumed that radiation, once scattered, is
diffuse and isotropic (including from the surface). For an arbitrary
layer 1 (or combination of layers with radiative properties
,
,
,
)
overlaying layer 2 (or combination of layers with radiative properties
,
, and
), the
combination formulas for direct and diffuse radiation incident from
above are:
Note that the transmissions for each layer ( ) and for the combined layers are total transmissions, containing both direct and diffuse transmission. Note also that the two layers (or combination of layers), once combined, are no longer a homogeneous system.
To combine the layers over the entire column, two passes are made through the layers, one starting from the top and proceeding downward, the other starting from the surface and proceeding upward. The result is that for every interface, the following combined reflectivities and transmissivities are available:
direct beam transmission from topofatmosphere to the  
interface ( is the scaled optical depth from topofatmosphere  
to the interface),  
reflectivity to direct solar radiation of entire atmosphere  
below the interface,  
total transmission to direct solar radiation incident from above  
to entire atmosphere above the interface,  
reflectivity of atmosphere below the interface to diffuse  
radiation from above,  
reflectivity of atmosphere above the interface to diffuse  
radiation from below. 
With these quantities, the upward and downward fluxes at every interface can be computed. For example, the upward flux would be the directly transmitted flux ( ) times the reflection of the entire column below the interface to direct radiation ( ), plus the diffusely transmitted radiation from above that reaches the interface ( ) times the reflectivity of the entire atmosphere below the interface to diffuse radiation from above ( ), all times a factor that accounts for multiple reflections at the interface. A similar derivation of the downward flux is straightforward. The resulting expressions for the upward and downward flux are:
If two or more configurations of binary clouds are identical between TOA and a particular interface, then , , and are also identical at that interface. The adding method is applied once and the three radiative quantities are copied to all the identical configurations. This process is applied at each interface by constructing a binary tree of identical cloud configurations starting at TOA down to the surface. A similar method is used for and , which are calculated using the adding method starting the surface and continuing up to a particular interface. The copying of identical radiative properties reduces the number of calculations of , , and by 62% and the number of calculations of and by 21% in CAM 3.0 with M/R overlap.
The computational cost of the shortwave code has two components: a fixed cost for computing the radiative properties of each layer under clear and overcast conditions, and a variable cost for applying the adding method for each column configuration . The variable component can be reduced by omitting configurations which contribute small terms in the shortwave fluxes. Several mechanisms for selecting configurations for omission have been included in the parameterization. The parameters that govern the selection process are described in Table 4.1.
Any combination of the selection conditions may be imposed. If the parameter , cloud layers with are identified as cloudfree layers. The configurations including these clouds are excluded from the flux calculations. If the parameter , the cloud amounts are discretized by