United Kingdom Meteorological Office: Model UKMO HadAM3 (2.5x3.75 L19) 1998a

Contact Information

Experimental Implementation

Model Output Description

Model Characteristics

Contact Information

Modeling Group

AMIP Representative(s)

• AMIP representative: Dr. Vicky Pope
• Mail address: Hadley Centre for Climate Prediction and Research, United Kingdom Meteorological Office, London Road, Bracknell, Berkshire RG12 2SY, United Kingdom.
• Phone: +44-1344-854655
• Fax: +44-1344-854898.
• Internet e-mail address: vdpope@meto.gov.uk.

Experimental Implementation

Simulation Period

Earth Orbital Parameters

Calendar

Radiative Boundary Conditions

• As specified for AMIP II, the solar constant is 1365 Wm^-2 (with both seasonal and diurnal cycles present).

• The carbon dioxide concentration is the AMIP-specified 348 ppmv.  AMIP-recommended concentrations of methane (1650 ppbv) and nitrous oxide (306 ppbv) are specified.  Other gas concentrations are 209500 ppm for O₂, 222 pptv for CFC-11, and 382 pptv for CFC-12.

• The monthly ozone concentration is the recommended zonal mean climatology of Wang et al. (1995)[59] interpolated to the model grid.

• The aerosol concentration is specified as a  background value, without seasonal variation, after Cusack et al. (1998)[49]. See also Chemistry.

Ocean Surface Boundary Conditions

• As recommended, these boundary conditions, obtained from PCMDI, are spatially interpolated at the model's horizontal resolution and temporally interpolated to daily values, so as to preserve monthly means.

• Fractional values of sea ice less than 0.3 are set to zero.  If sea ice is present,  the AMIP II SSTs are not used for calculating the radiative and turbulent fluxes from the underlying ocean; instead, a value of -1.8 deg C (the melting temperature of sea ice) is prescribed.  See also Sea Ice.

Orography/Land-Sea Mask

• The model orography is derived from the AMIP-recommended U.S. Navy 10' x 10' data set, and its global-average height is 230.203 m. The raw data are averaged over the model grid boxes; polewards of 60 degrees latitude, a 1-2-1 east-west filter is also applied. Sub-gridscale orographic variances, which impact the parameterization of gravity wave drag, also are computed from the raw data.

• The land-sea mask (the same as that used in the coupled model HadCM3) is derived from the U.S. Navy data set. Each grid box of the mask is defined as either completely land or sea (i.e. no fractional land/sea values are allowed).

Atmospheric Mass

Spinup/Initialization

• The HadAM3 model was run from 00Z 1 February 1978 to 00Z 1 January 1979 using additional spinup SSTs provided by PCMDI.

• The initial dump used for the start of this spinup was an operational analysis for 00Z 1 February 1997 (i.e. from a 19-level, 0.883 x 1.25 degree resolution model analysis).

• The snow depth was taken from climatology for 1 February,  and then was allowed to evolve during the spinup.

• The start conditions for soil moisture/temperature(s) for 1 February 1978 were taken from the model climatology for the last 5 years of a HadAM3 AMIP I run.

Computational Performance
To simulate 1 day, the AMIP II experiment required about 2 minutes per processor, assuming use of all 36 processors. (Actual processing time was highly variable, however, depending on the computer load.)
 

Model Output Description

Calculation of Standard Output Variables

• Method for calculation of percentage time that a pressure surface is below ground: The selected pressure surface is compared against the actual surface pressure at land and ocean grid points to explicitly determine the percentage time below ground. (In cases of deep depressions, for example, the 1000 hPa surface may be below ground a percentage of time.)

• Method for calculation of monthly mean tendencies at 17 WMO standard pressure levels:

• Temperature tendency due to total diabatic heating: Not available.
• Temperature tendencies due to shortwave and longwave radiation: Mean model-level temperatures are computed before and after shortwave and longwave radiation output, and differenced. These model-level tendencies are interpolated to the 17 WMO pressure levels using the mean model surface pressure.
• Temperature tendency due to moist and dry convection: Same method as above. (Note, dry and moist components are not distinguished by the model.)
• Temperature tendency due to large-scale/stratiform precipitation: Same method as above.
• Total moisture tendency due to diabatic processes: Not available.

• Method for calculation of cloud properties:

• Cloud water/ice: prognostically determined, available on model levels.
• Extinction coefficient (cloud optical thickness/layer depth): Not available.
• Cloud emittance:  Not available.

• Method for calculation of surface variables : Following the procedure of Geleyn (1988)[69] (same as in the AMIP I model).

• Method for calculation of mean sea-level pressure: The temperature of the lowest atmospheric level is vertically extrapolated to sea level using a constant lapse rate. The mean sea-level pressure then is obtained from the model surface pressure through use of the altimeter equation.

• Method for calculation of clear-sky radiation and cloud radiative forcing:  Following the recommended procedure of Potter et al. (1992)[70].

• Method for calculation of potential vorticity: Following the method of Anderson and Roulstone (1991)[71].

• Method for calculation of planetary boundary layer height: The top of the turbulent mixing layer is derived from the level at which the profile of the local bulk Richardson number exceeds a critical value of unity (same as in the AMIP I model).

Sampling Procedures

Interpolation Procedures

• Algorithm for interpolation of standard output variables to 17 WMO pressure surfaces: Model-level dynamical variables and their products fields are interpolated to the WMO pressure surfaces following Goddard (1991)[72]. Because of their highly nonlinear character, model-level cloud variables are not interpolated to the WMO pressure surfaces.

• Algorithm for treatment of variables on pressure surfaces below ground:

• For U- and V-wind components,  vertical velocity, specific humidity,  and relative humidity, the value of the field on the below-ground pressure surface is equated to the value at the bottom model level for the same grid point.
• For temperature and geopotential height, the values on below-ground pressure surfaces are obtained by extrapolation, subject to the following assumptions:
(a) a constant lapse rate (gam) of 0.0065 K/m;
(b) a reference temperature taken to be that at the model level just above the model boundary layer;
(c) application of the standard altimeter equation,  i.e.

T(surface) = T(ref level) [p(surface)/p(ref level)]^{(gam R/g)} or

T(i) = T(ref level) [p(i)/p(ref level)]^{(gam R/g)} , where i is the level below the surface.

Since this expression for temperature also can be written as

T(i) = T (surface) + gam [z(surf) - z(i)] ,

combining the above relations gives the height of  the below-ground pressure level as

z(i) = z(surf) + [T(surf)/gam] [ 1 - [p(i)/p(surf)]^{(gam R/g)}]
 
 

Output Data Structure/Format/Compression

• The raw model output data were written in the standard UKMO Unified Model "pp" format, preceded by a 64-word header record. The data were packed (using the UKMO WGDOS algorithm) to an accuracy that depends on the field type, as specified by a model look-up table (maximum accuracy: 32 bits).

• These data were supplied to PCMDI in netCDF format, organized according to the LATS data structure.

Model Characteristics

AMIP II Model Designation

Model Lineage

Model Documentation

Numerical/Computational Properties

Horizontal Representation

Horizontal Resolution

2.5x3.75 degrees latitude-longitude.
 

Vertical Domain

Vertical Representation

Vertical Resolution

Time Integration Scheme(s)

Smoothing/Filling

• To prevent numerical instability, orography is smoothed in high latitudes (see Orography/Land-Sea Mask), and Fourier filtering is applied to mass-weighted velocity and to increments of potential temperature and total moisture.

• Negative moisture values are filled by removing moisture from surrounding grid points, as available. Otherwise,  a global adjustment is implemented: negative moisture values are removed by summing the mass-weighted positive values in each horizontal layer, and rescaling them to ensure global moisture conservation after the negative values are reset to zero.

Dynamical/Physical Properties

Equations of State

Diffusion

• In the model troposphere, linear conservative horizontal diffusion is applied at sixth-order (Del⁶) to atmospheric moisture and winds, and to liquid water potential temperature (cf. Cullen et al. 1991[4]). Horizontal diffusion is switched off near steep topography.

• Stability-dependent, second-order vertical diffusion of momentum and of conserved cloud thermodynamic and water content variables (to include the effects of cloud-water phase changes on turbulent mixing), operates only in the planetary boundary layer. The diffusion coefficients are functions of the vertical wind shear (following mixing-length theory), as well as surface roughness length and a bulk Richardson number that includes buoyancy parameters for the cloud-conserved quantities (cf. Smith 1990b [16]). See also Cloud Formation, Planetary Boundary Layer, and Surface Fluxes.

Gravity Wave Drag

• Orographic gravity wave drag is parameterized as proportional to the vertical divergence of the wave stress, following Gregory et al. (1998)[54]. Near the surface, the stress is proportional to the product of a representative mountain wave number, the square of the wave amplitude (taken to be the subgrid-scale orographic variance--see Orography/Land-Sea Mask), and the density, wind, and Brunt-Vaisalla frequency evaluated in near-surface layers. Low-level gravity wave drag also is impacted by the anistropy of the sub-gridscale orography and by wave breaking due to hydraulic jumps and trapped lee waves.

• At higher levels, the stress is given by this surface value weighted by the projection of the local wind on the surface winds: as this projection approaches zero, the stress also approaches zero. Otherwise, if the minimum Richardson number falls below 0.25, the gravity wave is assumed to break. Above this critical level, the wave is maintained at marginal stability, and a corresponding saturation amplitude is used to compute the stress.

Chemistry

Radiation

• The shortwave fluxes are calculated by a two-stream method, where the spectrum is divided into 6 bands with the following boundaries (units of microns): 0.20-0.32; 0.32-0.69 (2 bands);  0.69-1.19; 1.19-2.38; and 2.38-5.00. Within each band, data are treated using correlated k-distribution methods (cf. Cusack et al. 1999[60]). Gaseous absorption data are derived from HITRAN92 (cf. Rothman et al. 1992[75]), except for ozone where the data are derived from LOWTRAN7 (cf. Kneizys et al. 1988[76]). Rayleigh scattering by gases and scattering/absorption by a climatological background aerosol also are represented. See also Chemistry and Radiative Boundary Conditions.

• The longwave fluxes also are calculated by a two-stream method, where the spectrum is divided into 8 bands with the following boundaries (units of m^-1): 0-40000; 40000-55000; 55000-80000 (excluding 59000-75000); 59000-75000; 80000-120000 (excluding 99000-112000); 99000-112000; 120000-150000; and 150000-300000. Gaseous absorption data are derived from HITRAN92 and continuum absorption by water vapor is treated using version 2.1 of the CKD continuum model after Clough et al. 1989[77]. Cf. Cusack et al. (1999)[60] and Edwards (1996)[62] for further details.

• Cloud-radiative interactions are treated consistently in the shortwave and longwave. The radiative properties of clouds are parameterized separately for water and ice as functions of condensed water content and effective radius. These parameterizations follow the functional forms of Slingo (1989)[20], but with newly derived coefficients for each shortwave/longwave spectral band. The effective radii of water droplets are calculated following Martin et al. (1994)[61] , while the effective radius of ice particles (treated as spheres) is fixed at 30 microns. Stratiform and convective cloud are distinguished, with the latter treated as a vertical tower. Clouds in adjacent layers are assumed to overlap maximally, and otherwise to overlap randomly, with the vertical coherence of stratiform and convective cloud accounted for. Mixed phase clouds are represented by horizontally adjacent regions of ice and water cloud. Cf. Edwards and Slingo (1996)[50] for further details.  See also Convection and Cloud Formation.

Convection

• Moist and dry convection are both simulated by the mass-flux scheme of Gregory (1990) [6] and Gregory and Rowntree (1990)[26] that is based on the bulk cloud model of Yanai et al. (1973) [25].

• Convection is initiated if a parcel in vertical layer k has a minimum excess buoyancy beta that is retained in the next higher level k+1 when entrainment effects and latent heating are included. The convective mass flux at cloud base is taken as proportional to the excess buoyancy; the mass flux increases in the vertical for a buoyant parcel, which entrains environmental air and detrains cloud air as it rises. Both updrafts and downdrafts are represented, the latter by an inverted entraining plume with initial mass flux related to that of the updraft, and with detrainment occurring over the lowest 100 hPa of the model atmosphere.

• When the parcel is no longer buoyant after being lifted from layer m to layer m + 1, it is assumed that a portion of the convective plumes has detrained in layer m so that the parcel in layer m + 1 has minimum buoyancy b. Ascent continues until a layer n is reached at which an undiluted (without entrainment) parcel originating from the lowest convectively active layer k would have zero buoyancy, or until the convective mass flux falls below a minimum value. Convective momentum transport also is parameterized following Gregory et al. (1997)[53].

Cloud Formation

• The convection scheme determines the vertical extent of subgrid-scale convective cloud, which is treated as a single tower in each grid box. The convective cloud base is taken as the lower boundary of the first model layer at which saturation occurs, and the cloud top as the upper boundary of the last buoyant layer. The fractional coverage of each vertical column by convective cloud is a logarithmic function of the mass of liquid water condensed per unit area between cloud bottom and top (cf. Gregory 1990[6]).

• Stratiform cloud is prognostically determined in a similar fashion to that of Smith (1990a) [27]. Cloud amount and water content are calculated from the total moisture (vapor plus cloud water/ice) and the liquid/frozen water temperature, which are conserved during changes of state of cloud water (i.e., cloud condensation is reversible). In each grid box, these cloud-conserved quantities are assumed to vary (because of unresolved atmospheric fluctuations) according to a triangular statistical distribution, with specified standard deviation. The mean local cloud fraction is given by the part of the grid box where the total moisture exceeds the saturation specific humidity (defined over ice if the local temperature is < 273.15 K, and over liquid water otherwise). See also Radiation and Precipitation.

Precipitation

• Large-scale precipitation forms in association with stratiform cloud. For purposes of precipitation formation and the radiation calculations, the condensate is assumed to be liquid above 0 degrees C, and to be ice below -9 degrees C, with a liquid/ice fraction obtained by linear interpolation for intermediate temperatures (cf. Senior and Mitchell 1993[74] and Gregory and Morris 1996[52]). The rate of conversion of cloud water into liquid precipitation is a nonlinear function of the large-scale cloud fraction and the cloud-mean liquid water content, following Sundqvist (1978 [28], 1981 [29]) and Golding (1986) [30]. The precipitation of ice is a nonlinear function of the cloud-mean ice content, as deduced by Heymsfield (1977) [31]. Liquid and frozen precipitation also form in subgrid-scale convection.

• Evaporation/sublimation of falling liquid/frozen precipitation are modeled after Gregory (1995)[51].  For purposes of land hydrology, surface liquid/frozen precipitation is assumed to be exponentially distributed over each land grid box, with fractional coverage of 0.5 for large-scale precipitation and 0.1 for convective precipitation. Cf. Smith and Gregory (1990)[9] and Dolman and Gregory (1992)[34] for further details.

Planetary Boundary Layer

• Conditions within the PBL are typically represented by the first 5 levels above the surface (centered at about 997, 975, 930, 869, and 787 hPa for a surface pressure of 1000 hPa), where turbulent diffusion of momentum and cloud-conserved thermodynamic and moisture variables may occur (see Diffusion and Cloud Formation). The PBL top is defined either by the highest of these layers, or by the layer in which a modified bulk Richardson number (that incorporates buoyancy parameters for the cloud-conserved variables) exceeds a critical value of unity.

• K-theory diffusivities are calculated from the local Richardson number, the vertical shear, and a mixing length that depends on the PBL depth. See also Surface Characteristics, Surface Fluxes, and Land Surface Processes.

Sea Ice

• AMIP II monthly sea ice extents are prescribed. The ice may occupy only a fraction (at least 0.3) of a grid box, and the effects of the remaining ice leads are accounted for in the surface roughness length, shortwave albedo and longwave emission, and turbulent eddy fluxes. Snow falling on sea ice affects the surface albedo (see Surface Characteristics), but not the ice thickness or thermodynamic properties.

• The spatially variable sea ice thickness is prescribed from climatological data. Ice temperature is prognostically determined from a surface energy balance (see Surface Fluxes) that includes a conduction heat flux from the ocean below. Following Semtner (1976) [36], the conduction flux is proportional to the difference between the surface temperature of the ice and the subsurface ocean temperature (assumed to be fixed at the melting temperature of sea ice, or -1.8 degrees C), and the conduction flux is inversely proportional to the prescribed ice thickness.  See also Ocean Surface Boundary Conditions.

Snow Cover

• Surface snowfall is determined from the rate of frozen large-scale and convective precipitation in the lowest atmospheric layer. On land only, the prognostic mass  of a single snow layer (at constant density 250 kg m^-3) is determined from a budget equation that accounts for accumulation, melting, and sublimation. Snow melts when either the mean temperature of the top soil/snow layer or the diagnostic surface skin temperature exceeds 0 deg C. Snowmelt augments soil moisture, and sublimation of snow  contributes to the continental evaporative flux.

• Snow is assumed to cover an entire grid box (i.e. fractional snow coverage is not represented).  Snow cover affects the roughness, surface albedo, and heat conduction at the land surface. See also Precipitation, Surface Characteristics, Surface Fluxes, and Land Surface Processes.

Surface Characteristics

• Surface types include land, ocean, sea ice, and permanent land ice. Sea ice may occupy a fraction of a grid box. On land, 3 soil texture types (fine, medium, and coarse) and 23 land cover (vegetation) classes are specified from the 1x1-degree data of Wilson and Henderson-Sellers (1985) [37]. Mean soil textures for each grid box are calculated as weighted means of sand, silt, and clay fractions. Effects of these soil textures on thermodynamics and hydrology are treated via parameters derived by Buckley and Warrilow (1988) [38].  Geographically varying fields of vegetation parameters (canopy height and interception capacity, infiltration enhancement factor, leaf area index) also are prescribed after Cox et al. (1999)[48]. Aggregated parameters for each grid box are calculated from area-weighted means of sub-gridscale values.

• On each surface, different roughness lengths are specified for momentum. Over oceans only, additional roughness lengths for heat/moisture, and for free convective turbulence (which applies in cases of very light surface winds under unstable conditions) also are specified. Over land, the momentum roughness length is a function of vegetation and small surface irregularities, following Buckley and Warrilow (1988)[38]; it decreases linearly with snow cover to a minimum of 5 x 10^-4 m. In mountainous regions, an effective momentum roughness length represents unresolved orographic form drag (cf. Milton and Wilson 1996[56]). The aggregated roughness length is calculated using a blending height concept (cf. Mason 1988[55]). Over oceans, the roughness length for momentum is a function of surface wind stress (cf. Charnock 1955 [39]), but is constrained to be at least 10^-4 m; the roughness length for heat/moisture  is a constant 10^-4 m,  and that for free convective turbulence is a constant 1.3 x 10^-3 m. Over sea ice, the momentum roughness length is a constant 3 x 10^-3 m, but it is 0.10 m for the fraction of the grid box with ice leads. Cf. Smith (1990b) [16] for further details.

• The surface albedo of open ocean is a function of solar zenith angle. The albedo of sea ice varies between 0.60 and 0.85 as a linear function of the ice temperature above -5 deg C, and it is also modified by snow cover. Where there is partial coverage of a grid box by sea ice, the surface albedo is given by the fractionally weighted albedos of sea ice and open ocean. Albedos of snow-free and deeply snow-covered land are specified according to the different land cover classes from Buckley and Warrilow (1988)[38], where the deep-snow albedo decreases linearly with surface temperature above +2 deg C. For intermediate snow depths, the land albedo approaches the deep-snow value exponentially, according to an e-folding depth of snow equivalent to 5 mm of water (cf. Cox et al. 1999[48]).

• Longwave emissivity is unity (blackbody emission) for all surfaces. Thermal emission from grid boxes with partial coverage by sea ice is calculated from the different surface temperatures of ice and the open-ocean leads, weighted by the fractional coverage of each. See also Surface Fluxes and Land Surface Processes.

Surface Fluxes

• Surface solar absorption is determined from the surface albedos, and longwave emission from the Planck equation with constant emissivity of 1.0 (see Surface Characteristics).

• The surface turbulent fluxes are formulated as bulk formulae in a constant-flux surface layer, following Monin-Obukhov similarity theory. The surface atmospheric variables required for the bulk formula are taken to be at the first level above the surface (at 997 hPa for a 1000 hPa surface pressure). (For diagnostic purposes, temperature and humidity at 1.5 m and the wind at 10 m are also estimated from the constant-flux assumption.) Following Louis (1979) [42], the drag/transfer coefficients in the bulk formulae are functions of stability (expressed as a bulk Richardson number) and roughness length, where the same transfer coefficient is used for heat and moisture. In grid boxes with fractional sea ice, surface fluxes are computed separately for the ice and lead fractions, but using mean drag and transfer coefficients obtained from linearly weighting these fractions. The partitioning between sensible and latent energy fluxes follows an extended Penman-Monteith approach (cf. Monteith 1973[73]) with surface skin temperature determined diagnostically. The residual ground heat flux is proportional to the difference between this skin temperature and the temperature of the top soil layer, where the heat conductivity decreases with snow cover and soil water/ice content (cf. Smith et al. 1994[58] and Cox et al. 1999[48]).

• The continental moisture flux includes sublimation from snow, evaporation from the wet vegetation canopy and from bare soil, and transpiration by the vegetation roots and stomates:

• The atmospheric evaporative demand is met first by sublimation from snow which occurs at the potential rate (i.e. aerodynamic resistance only), following a bulk formulation.

 
• The canopy is represented as a single-story "big leaf" whose height, leaf area index, and water storage capacity vary with vegetation type. Canopy interception is determined assuming the sub-gridscale rainfall intensity follows an exponential distribution (cf. Dolman and Gregory 1992[34]). Evaporation from the wet canopy occurs at the potential rate, where the wet fractional area  varies linearly with the canopy storage. (A non-zero wet area also is prescribed on bare soil to allow for ponding effects.) Evaporation from bare soil is limited by aerodynamic and soil surface resistance (a constant 100 s m^-1) and by a ``beta'' factor that depends on the volumetric moisture content of the top soil layer: beta is 0 if moisture content is less than a wilting point and is 1 if greater than a critical point, while beta varies linearly for intermediate moisture contents. The wilting and critical points (corresponding, respectively, to soil water suctions of 1.5 MPa and 0.033 MPa) vary geographically according to soil texture types.

 
• The rooting depth for transpiration is assumed to be 1 meter (the top 3 soil layers) for all vegetation types except trees, which can draw moisture from the entire 3-meter soil column. Transpiration also is limited by a bulk stomatal resistance, which is  calculated by the coupled canopy conductance-photosynthesis scheme of Cox et al. (1998)[47] that includes a representation of carbon fluxes.  The bulk resistance depends on surface temperature, vapor pressure deficit, photosynthetically active radiation (PAR), carbon dioxide concentration, and soil moisture availability.  See also Surface Characteristics and  Land Surface Processes.

• Above the surface layer, momentum and cloud-conserved temperature and water variables are mixed vertically within the PBL by local stability-dependent diffusion.  Cf. Smith (1990b[16], 1993 [8]) and Pope et al. (2000)[57] for further details.

Land Surface Processes

• The vegetated and bare-soil fractions of a grid box are both at the diagnosed skin temperature. The subsurface temperatures are updated using a discretized form of the heat diffusion equation that is applied in four vertical layers (thicknesses 0.1, 0.25, 0.65, and 2.0 m, with depth), subject to a zero-flux lower boundary condition. Effects of soil water phase changes on heat capacity are incorporated by the approach of Williams and Smith 1989[63], in which the maximum unfrozen water at a given temperature is derived from the soil water suction curve (cf. Black and Tice 1988[64]). The effective soil thermal conductivity is calculated as a function of water and ice content after Farouki (1981)[65].

• Soil hydrology is based on a finite difference approximation to the Richards equation (cf. Richards 1931[66]) with the same vertical discretization as for soil thermodynamics. The moisture within a layer is incremented by the net incoming/outgoing Darcian fluxes, and by transpiration. The dependencies of soil water suction and hydraulic conductivity on soil moisture follow Clapp and Hornberger (1978)[67], with parameters derived from the Wilson and Henderson-Sellers (1985)[37] soil texture dataset using the regression relations of Cosby et al. (1984)[68].

• Surface runoff occurs whenever the grid-box mean rainfall intensity (computed from an exponential sub-gridscale distribution--cf. Dolman and Gregory 1992[34]) exceeds the hydraulic conductivity for prescribed saturated moisture values that depend on the soil texture types. Drainage also occurs from the base of the soil column at a rate equal to the hydraulic conductivity of the deepest soil layer (i.e. a "free drainage" lower boundary condition is applied); however, lateral flow and recharge from below are not represented. See also Surface Characteristics and Surface Fluxes.

AMIP I/AMIP II Model Differences

References

Return to:

Top of Page

AMIP II Model/Experiment Table of Contents

AMIP II Documentation Directory

AMIP Home Page

LLNL Disclaimers

UCRL-MI-135872