Dr. Nobuo Sato, Dr. Toshiki Iwasaki, and Dr. Tadashi Tsuyuki, Numerical Prediction Division, Japan Meteorological Agency, 1-3-4 Ote-machi, Chiyoda-ku, Tokyo 100 Japan; Phone: + 81-03-3212-8341; Fax: +81-03-3211-8407; e-mail: /PN=N.SATO/O=JMA/ADMD=ATI/C=JP/@sprint.com (Sato) /PN=T.IWASAKI/O=JMA/ADMD=ATI/C=JP/@sprint.com (Iwasaki) /PN=T.TSUYUKI/O=JMA/ADMD=ATI/C=JP/@sprint.com (Tsuyuki)
JMA GSM8911 (T42 L21) 1993
The JMA GSM8911 model first became operational in November 1989. This version is derived from an earlier global spectral model that is described by Kanamitsu (1983)
[1].
Key documentation of the model is provided by the Numerical Prediction
Division's 1993 Outline of Operational Numerical Weather Prediction at Japan Meteorological Agency (hereafter Numerical Prediction Division 1993)
[2] and by Sugi et al. (1989)
[3].
Spectral (spherical harmonic basis function) with transformation to a Gaussian grid for calculation of nonlinear quantities and some physics.
Spectral triangular 42 (T42), roughly equivalent to a 2.8 x 2.8 degree
latitudelongitude grid.
Surface to 10 hPa. For a surface pressure of 1000 hPa, the lowest atmospheric
level is at a pressure of 995 hPa.
Hybrid vertical coordinates which approximate conventional sigma coordinates at low levels and constant-pressure coordinates at upper levels (cf. Simmons and Burridge 1981)
[4].
There are 21 unevenly spaced hybrid levels. For a surface pressure of 1000 hPa, 6 levels are below 800 hPa and 7 levels are above 200 hPa.
The AMIP simulation was run on a HITAC S-810 computer using a single processor
in the HITAC VOS3/HAP/ES operational environment.
For the AMIP experiment, about 2 minutes of HITAC S-810 computation time per
simulated day.
For the AMIP simulation, the initial model atmospheric state is specified from
the ECMWF FGGE III-B analysis for 1 January 1993, with a nonlinear normal-mode
initialization also applied (cf. Kudoh 1984)
[5]. Soil moisture is initialized according to estimates of Willmott et al. (1985)
[6], and
snow cover/depth according to data of Dewey (1987)
[7].
Semi-implicit leapfrog time integration with an Asselin (1972)
[8] time filter (cf. Jarraud et al. 1982)
[9]. The length of the time step is not fixed, but is reset every 6 hours to satisfy the Courant-Friedrichs-Lewy (CFL) condition for the advection terms. Shortwave radiation is recalculated hourly, and longwave radiation every 3 hours.
Orography is truncated at the T42 model resolution (see Orography). When
the atmospheric moisture content of a grid box becomes negative due to spectral
truncation, its value is reset to zero without any other modification of the
local or global moisture budgets.
For the AMIP simulation, the model history is written every 6 hours, but some
diagnostic variables are stored only once per month because of limited storage
resources.
Primitive equation dynamics are expressed in terms of vorticity, divergence,
temperature, specific humidity, and surface pressure, as formulated by Simmons and Burridge (1981)
[4] for hybrid vertical coordinates.
- Fourth order linear (del^4) horizontal diffusion is applied to
vorticity, divergence, temperature, and specific humidity on the hybrid
vertical surfaces, but with a first-order correction of the temperature and
moisture equations to approximate diffusion on constant-pressure surfaces
(thereby reducing spurious mixing along steep mountain slopes). Diffusion
coefficients are chosen so that the enstrophy power spectrum coincides with
that expected from two-dimensional turbulence theory.
- Stability-dependent vertical diffusion of momentum, heat, and moisture in
the planetary boundary layer (PBL) as well as in the free atmosphere follows
the Mellor and Yamada (1974)
[10] level-2
turbulence closure scheme. The eddy diffusion coefficient is diagnostically
determined from a mixing length formulated after the method of Blackadar
(1962)[11]. See also Planetary Boundary Layer and Surface Fluxes.
Orographic gravity-wave drag is parameterized by two schemes that differ mainly
in the vertical partitioning of the momentum deposit, depending on the
wavelength of the gravity waves. Long waves (wavelengths >100 km) are
assumed to exert drag mainly in the stratosphere (type A scheme), and short
waves (wavelengths approximately 10 km) to deposit momentum only in the
troposphere (type B scheme). In both schemes the gravity-wave drag stress is a
function of atmospheric density, wind, the Brunt-Vaisalla frequency, and
subgrid-scale orographic variance (see Orography). (For the type B
scheme, orographic variance is computed as an average difference of maximum and
minimum heights within each 10-minute mesh.) In the type A scheme, the
deposition of vertical momentum is determined from a modified Palmer et al. (1986)
[12] amplitude saturation hypothesis.
Because the momentum stress of short gravity waves decreases with altitude as a
result of nonhydrostatic effects (cf. Wurtele et al. 1987)
[13], the type B scheme assumes the wave stress
to be quadratic in pressure and to vanish around the tropopause. Cf. Iwasaki et al. (1989a [14], b [15]) for further details.
The solar constant is the AMIP-prescribed value of 1365 W/(m^2). Both
seasonal and diurnal cycles in solar forcing are simulated.
The carbon dioxide concentration is the AMIP-prescribed value of 345 ppm.
Monthly averaged zonal ozone distributions are specified from data of McPeters et al. (1984)
[16]. Radiative effects of water
vapor, but not of aerosols, are also included (see Radiation).
- Shortwave radiation is parameterized differently for wavelengths <0.9
micron (visible) and >0.9 micron (near-infrared). In the visible, absorption
by ozone, Rayleigh scattering by air molecules, and Mie scattering by cloud
droplets are treated. In the near-infrared, water vapor absorption is modeled
after Lacis and Hansen (1974)
[17].
Near-infrared scattering and absorption by cloud droplets are calculated by the
delta-Eddington approximation with constant single-scattering albedo.
- Longwave absorption by water vapor, ozone, and carbon dioxide is
determined from transmission functions of Rodgers and Walshaw (1966)
[18], Goldman and Kyle (1968)
[19], and Houghton (1977)
[20], respectively; pressure broadening effects are also included. Continuum absorption by water vapor is treated by the method of Roberts et al. (1976)
[21]. Transmission in
four spectral bands (with boundaries at 4.0 x 10^3, 5.5 x 10^4,
8.0 x 10^4, 1.2 x 10^5, and 2.2 x 10^5 m^-1) includes overlapping effects of different absorbers. Longwave emissivity of
cirrus cloud is set at 0.80, and that of all other clouds at 1.0 (blackbody
emission). For purposes of the radiation calculations, all clouds are assumed
to be randomly overlapped in the vertical. Cf. Sugi et al. (1989)
[3] for further
details. See also Cloud Formation.
- A modified Kuo (1974)
[22] parameterization is used to simulate deep convection. The criteria for the occurrence of convection include conditionally unstable stratification and positive moisture convergence between the cloud base and top. The cloud base is
at the lifting condensation level for surface air, and the top is at a level
where the cloud and environmental temperatures are identical. The cloud
temperature is determined from the moist adiabatic lapse rate modified by
height-dependent entrainment, as proposed by Simpson and Wiggard (1969)
[23]. In a vertical column, the total moisture available from convergence is divided between a fraction b that moistens the environment and the remainder (1 - b) that contributes to the latent heating (rainfall) rate. The moistening paramenter b is a cubic function of the ratio of the mean relative humidity of the cloud layer to a prescribed critical relative
humidity threshold value (70 percent); if cloud relative humidity is less than
this threshold, b is set to unity (no heating of the environment).
- Shallow convection occurs where the vertical stratification is
conditionally unstable but moisture convergence is negative. It is
parameterized by enhancing the vertical diffusion coefficients after the method of Tiedtke (1983)
[24].
No explicit convective cloud fraction is determined (see Convection).
The stratiform cloud fraction is a quadratic function of the difference between
the local relative humidity and a critical value that is empirically obtained
from satellite observations, and that varies for low, middle, and high clouds
(cf. Saito and Baba 1988)
[25]. See also
Radiation for treatment of cloud-radiative interactions.
The convective precipitation rate is determined from the variable moistening
parameter b in the modified Kuo (1974)
[22] convection scheme (see
Convection). Any remaining supersaturation is removed by large-scale
condensation. No subsequent evaporation of precipitation is simulated. See also
Snow Cover.
The Mellor and Yamada (1974)
[10] level-2 turbulence closure scheme (see
Diffusion) represents the effects of the PBL through the determination
of the Richardson number and the vertical wind shear. The PBL top is not
explicitly computed. See also Surface Fluxes.
Orography is obtained from a U.S. Navy dataset (cf. Joseph 1980
[26] with resolution of 10 minutes arc on a latitude-longitude grid. These data are expressed as a series of spherical
harmonics that are truncated at the T42 model resolution. Orographic variances
that are also obtained from this dataset are used in the parameterization of
gravity-wave drag (see Gravity-wave Drag).
AMIP monthly sea surface temperature fields are prescribed, with daily values
determined by linear interpolation.
Monthly AMIP sea ice extents are prescribed. The ice surface temperature is
predicted by the force-restore method of Deardorff (1978)
[27]. The forcing includes the net balance of
surface energy fluxes (see Surface Fluxes) as well as conduction heating
from the ocean below, which is computed assuming the ice to be a uniform 2-m thick and the ocean to be at the temperature for sea ice formation (about -2
degrees C). Snow is not allowed to accumulate on sea ice (see Snow Cover).
Precipitation falls as snow if the temperature at the lowest atmospheric level
(see Vertical Domain) is <0 degrees C. Snow may accumulate on land,
but not on sea ice. The fractional coverage of a grid box is proportional to
the water-equivalent snow depth up to 0.02 m; at greater depths, the
proportionality constant varies with vegetation type. Snow cover alters the
roughness and the albedo of bare and vegetated ground as well as the heat
capacity and conductivity of soil, but sublimation from snow is not included in
the surface evaporative flux. Snow melts (and contributes to soil moisture) if
the ground surface temperature is >0 degrees C. See also Surface Characteristics, Surface Fluxes, and Land Surface Processes.
- Over land, the 12 vegetation/surface types of the Simple Biosphere (SiB)
model of Sellers et al. (1986)
[28] are
specified at monthly intervals.
- The local roughness length over land varies monthly according to
vegetation type (cf. Dorman and Sellers 1989)
[29]; it decreases with increasing snow depth,
the minimum value being 5 percent of that without snow cover. The surface
roughness of sea ice is a uniform 1 x 10^-3 m. Over oceans, the
roughness length for momentum is a function of the surface wind stress after
Charnock (1955)
[30], while the roughness
length for surface heat and moisture fluxes is specified as a constant 1.52 x
10^-4 m (cf. Kondo 1975)
[31].
- Over land, surface albedos vary monthly according to seasonal changes in
vegetation (cf. Dorman and Sellers 1989)
[29]. The albedo is specified separately
for visible (0.0-0.7 micron) and near-infrared (0.7-4.0 microns) spectral
intervals, and is also a function of solar zenith angle. Following Sellers et
al. (1986)
[28], snow cover alters the surface albedo. Over oceans and sea ice,
albedos are functions of solar zenith angle but are independent of spectral
interval.
- Longwave emissivity is prescribed to be unity (blackbody emission) for all
surfaces. See also Surface Fluxes and Land Surface Processes.
- Solar absorption at the surface is determined from the albedo, and
longwave emission from the Planck equation with prescribed emissivity of 1.0
(see Surface Characteristics).
- The representation of turbulent surface fluxes of momentum, heat, and
moisture follows Monin-Obukhov similarity theory as expressed by bulk
formulae. The wind, temperature, and humidity required for these formulae are
taken to be the values at the lowest atmospheric level (at 995 hPa for a
surface pressure of 1000 hPa). The associated drag/transfer coefficients are
functions of the surface roughness (see Surface Characteristics) and
vertical stability, following Louis et al. (1981)
[32].
- Over vegetated surfaces, the temperature and specific humidity of the
vegetation canopy space of the SiB model of Sellers et al. (1986)
[28] are used as
surface atmospheric values. Over land, the surface moisture flux includes
evapotranspiration from dry vegetation (reflecting the presence of stomatal and
canopy resistances) as well as direct evaporation from the wet canopy and from
bare soil (see Land Surface Processes).
- Land surface processes are simulated by the SiB model of Sellers et al.
(1986)
[28], as implemented by Sato et al. (1989a [33], b [34]).
Vegetation in each grid box may consist both of ground cover and an upper-story
canopy, with the spatial pattern of the ground cover varying monthly. Within
the canopy, evaporative fluxes are computed by the Penman-Monteith method (cf. Monteith 1973)
[35]. Evapotranspiration from
dry leaves includes the detailed modeling of stomatal and canopy resistances.
Direct evaporation from the wet canopy and from bare soil is also treated (see
Surface Fluxes). Precipitation interception by the canopy (with
large-scale and convective precipitation distinguished) is simulated, and
infiltration of moisture into the ground is limited to less than the local
hydraulic conductivity of the soil.
- Soil temperature is predicted in four layers by the force-restore method
of Deardorff (1978)
[27]. Soil liquid moisture is predicted from budget equations in
three layers, and snow and soil ice in four layers. This moisture is increased
by infiltrated precipitation and snowmelt, and is depleted by
evapotranspiration and direct evaporation. Both surface runoff and deep runoff
from gravitational drainage are simulated. See also Surface Characteristics and Surface Fluxes.
Go to JMA References
Return to JMA Table of Contents
Last update April 19, 1996. For further information, contact:Tom Phillips (
phillips@tworks.llnl.gov)
LLNL Disclaimers
UCRL-ID-116384