Skip to main content

Full text of "The NASA/Ames Mars General Circulation Model: Model Improvements and Comparison with Observations"

See other formats


RM Haberle NASA/Ames Research Center (, J.L Hollingsworth, NASA/Ames 
Research Center, A. Colaprete, NASA/ Ames Research Center 
( A.F.C. Bridger, San Jose State University ( C.P. 
McKay, NASA/Ames Research Center (, J.R. Murphy, New Mexico State Univer- 
sity (, J. Schaeffer, Raytheon Corporation (, and R. Freed- 
man, NASA/Ames Research Center ( 

Introduction: For many years, the NASA/Ames 
Mars General Circulation Model (GCM) has been 
built around the UCLA B-grid dynamical core. An 
attached tracer transport scheme based on the aerosol 
microphysical model of Toon et al. (1988) provided 
a tool for studying dust storm transport and feed- 
backs (Murphy et al., 1995). While we still use a B- 
grid version of the model, the Ames group is now 
transitioning to the ARIES/GEOS Goddard C-grid 
dynamical core (Suarez and Takacs, 1995). The C- 
grid produces smoother fields when the model top is 
raised above 50 km, and has a built in transport 
scheme for an arbitrary number of tracers. All of our 
transport simulations are now carried out with the C- 

We have also been updating our physics pack- 
age. Several years ago we replaced our bulk boundary 
layer scheme with a level 2 type diffusive scheme, 
and added a multi-level soil model (Haberle et al., 
2000). More recently we replaced our radiation code 
with a more generalized two-stream code that ac- 
counts for aerosol multiple scattering and gaseous 
absorption. This code gives us much more flexibility 
in choosing aerosol optical properties and radiatively 
active gases. Thus, we have several versions of our 
GCM and these are listed in Table 1 . 

also been coupled to a sophisticated cloud micro- 
physics package (i.e., the Community Aerosol and 
Radiation Model for Atmospheres - CARMA, see 
Colaprete and Toon, 2000) to begin exploring water 
and C0 2 ice cloud formation. However, the version 
we are transitioning to is GCM 2.0, which is now 
undergoing final testing, and check out. 

Model Improvements: The C-grid transport 
scheme advects tracers using the same numerical 
algorithm developed for potential temperature. At 
present, we use this scheme to transport water vapor, 
and an arbitrary number of dust and cloud particles. 
Dust can be lifted into the atmosphere through a 
prescribed source, or a model-predicted parameteriza- 
tion. Once into the lowest layer (nominally 10 m 
thick) dust is vertically mixed by a stability depend- 
ent diffusive scheme followed by a convective ad- 
justment. Water vapor is treated similarly, though 
we do not yet have a good evaporation parameteriza- 
tion for surface ice. Dust is removed by gravitational 
settling; water vapor by precipitation. The latter can 
range in sophistication from simple successive satu- 
ration removal, to a full up CARMA cloud micro- 
physical approach. 

The model-predicted dust lifting schemes are 
based on the work of Murphy (1999) (with details 

Table 1. Versions of the NASA/ Ames Mars General Circulation Model 



t>BL / Soil 







Bulk scheme / 
Single Layer 

Dust/C0 2 - 

H 2 clouds 
prescribed -fixed 

Aerosol model 



scheme/ Multi- 

Dust/C0 2 - 

H2O clouds 
prescribed -fixed 




scheme/ Multi- 

Dust/C0 2 - 







scheme/ Multi- 

Generalized 2- 





GCM 1 .0 was our original model but has been 
retired (though it is still available for comparison 
purposes). GCM 1 .5 was used to interpret Pathfinder 
observations and was the first version of the model 
used to assess the effects of MOLA topography on 
the general circulation. GCM 1.7 is used to compare 

given in Haberle et al, 2002) and Newman et al. 
(2002). Murphy's lifting scheme is parameterized in 
terms of surface stress, whereas the Newman et al. 
scheme is based on a saltation flux calculated from 
the friction velocity. Both schemes are threshold 

THE NASA/AMES GCM: R.M. Haberle et al. 

hard-wired to specific dust and water ice optical 
properties, with C0 2 being the only radiatively active 
gas. Furthermore, the range of surface pressures this 
code can accommodate is limited to < 100 hPa. Our 
new radiation code is based on a generalized two- 
stream solution to the radiative transfer equation 
with gaseous opacities calculated using correlated- 
k's. The two-stream solutions can accommodate Ed- 
dington, Quadrature, Hemispheric Mean, and Delta 
function approximations. We are presently using the 
Quadrature approximation for solar radiation and the 
energy conserving Hemispheric Mean approximation 
in the thermal infrared. 

At present we run with 34 spectral intervals 
from 0.3 to 250 microns. The correlated-k's for these 
intervals are generated from a line-by-line code using 
the HITEMP data base from HITRAN for C0 2 , and a 
version of the Schwenke data base (to include lines 
too weak to appear in HITRAN) for H 2 0. In both 
cases line widths are adjusted to represent C0 2 
broadening. A Voigt profile is used at low pres- 
sures, and a Lorentz profile at high pressures. The 
line widths are extended at high pressures so as to 
include all significant absorption. The abundance of 
the deuterated species for H 2 was adjusted for Mars 
conditions. The line-by-line calculations were then 
windowed, and sorted to produce the k coefficients. 
We use a gauss scheme of 8 & 8 points in each spec- 
tral interval with the dividing point at .95 to extract 
the actual coefficients from the sorted probability 
distribution. The k coefficients have been computed 
for a range of pressures, temperatures, and relative 
humidities that allows us to simulate past as well as 
present Martian climates. A example of how this 
new code compares with Dave Crisp's DART code 
for a pure C0 2 atmosphere is shown in Fig. 1 . 



■ i 











- A — Y 



tja. ecu 



• •— ft 
1 ■ 1 




«& Cr% 


-M -« -X -N -10 10 20 X « » 
HEKU* (DK/50U 

Fig. 1. Solar and IR heating rates computed from the 
new radiation code (black) and Crisp's DART code. 

For aerosols, we calculate the wavelength de- 
Dendent scatterine DroDerties (di~. e. and 0„<) off 

line using Mie theory. The dust scattering properties 
we are presently using are taken from the Ockert-Bell 
et al. (1997) work in the visible and Forget (1998) in 
the infrared. The visible (.67 microns) to infrared 
opacities are scaled to produce a value of 2 at 9 mi- 

During the past several years we have coupled 
the CARMA cloud microphysics package into the C- 
grid (GCM version 1 .7) and have begun exploring 
the behavior of H 2 and C0 2 ice clouds in the pre- 
sent climate system (e.g., Colaprete and Haberle, 
2001). The microphysics model accounts for the 
particle-size dependent processes of nucleation, con- 
densation, sedimentation, and evaporation. The ex- 
pressions used for these processes and the rationale 
for them are given in Colaprete (2000) and references 
therein. The model keeps track of three particle types 
in an arbitrary number of size bins: dust, ice, and 
ice-coated dust. The ice can be water ice or C0 2 ice. 
The code is general enough to handle both. 

Comparison with Observations: We have been 
comparing GCM 1.5, 1.7, and 2.0 with Viking and 
MGS observations. With GCM 1.5 we were able to 
pin down the annual global mean surface pressure on 
Mars. We tuned the polar cap properties until the 
model-predicted surface pressures gave a good fit to 
the Viking Lander 1 and 2 data (Fig. 2). The result- 
ing global mean annual surface pressure was 6. 1 hPa, 
coincidentally (?) indistinguishable from the triple 
point pressure of water. 

Fig. 2. GCM fit to daily averaged Viking Lander 


M MD SO 1?0 1»0 TK" 210 W P6 1» M» JM 
MOLXJ>2 1 7- VL-2 Frt 

- -0.062 

. ■ . . " 


MS Ein»r{ 








Hwrtm m «i *.-! 





. ..... 

, j .*-. i . .— i i i i ■ ■ ' 


3D BO »0 170 

1BO Z1D Z*Q r^C JM JW *" 

surface pressures. 

These simulations have also been compared to 
TES temperature data, where we find reasonable 
agreement with zonal mean values, but significant 
differences with the inferred amplitudes of the winter 
stationary waves. TES data show winter wave 1 am- 

2002), whereas our fixed dust and globally uniform 
GCM simulations produce ~20K in the south and ~ 
4K in the north. We can beat down the wave 1 am- 
plitudes in the south by either lowering the global 
opacity to ~ 0.1, or by running with a relatively clear 
polar atmosphere. The latter is more realistic. In the 
north however, the 4K amplitudes are robust to 
changes in the dust distribution. The northern ampli- 
tudes increase somewhat when the center of the time 
averaging window is moved a little earlier or later in 
the season. But the amplitudes never reach the 8K 
seen in the TES data. Figure 3 shows our GCM re- 
sults when we run using the TES observed opacities. 

bution of observed dust storms is more or less longi- 
tudinally uniform in both hemispheres, which is 
consistent with model predictions in the Northern 
Hemisphere, but less so in the Southern Hemisphere. 
Interestingly, very little lifting is predicted in the 
tropics (between ±30°) where only a few dust storms 
were observed. Overall, the model compares re- 
markably well with observations considering the 
assumptions of uniform surface roughness, threshold 
stress, and atmospheric dust loading. 

Deflotion Potentiol (cm), Prwent Obliquity: Ls-109- 
Uox- 3.32 

^gUtfJtvy&BPr? «ia#**s£| 

WAVE 1 [K) lou vorloble (Jli) U: 090 

Loftu* |0«g) 

Fig. 3. Stationary wave 1 temperature amplitudes for 
southern winter (top) and northern winter (bottom). 

To compare the model results with observed 
dust storm activity we introduce the concept of a 
deflation potential, which we define as the depth of 
dust that could be removed from the surface during a 
specified period of time. The deflation potential from 
one of our fixed dust (tau=0.3) experiments based on 
the Murphy (1999) lifting parameterization for the 
period between L s =109°-274° is shown in Fig. 4. 
Also shown are Cantor et al's (2001) observations of 
local dust storms by the MOC wide angle camera 
during the same period. Both model and observa- 
tions show that dust lifting occurs mostly poleward 
of 30° in either hemisphere. There is also a modest 
correlation between the density of dust storms and 

thf> marrnitiirlf. nf tVie Hoflatinn rvntontiul Tho Hictri. 

I I 

-180 -150 -120 -90 

-JO 30 


120 150 180 

Fig. 4. Deflation potential (contours) and dust 
storms (stars). 

An example of our fully coupled GCM and 
cloud microphysical model is shown in Fig. 5. In 
this simulation we employ the full capability of 
CAR.MA and carry 6 dust bins, 6 cloud bins (water 
and C0 2 ), and 1 water vapor bin. The figure depicts 
the zonally-averaged mass-weighted mean water ice 
cloud particle sizes at Ls=103°. The tropical aphe- 
lion cloud belt is readily simulated. Most of the 
water for these clouds comes from the subliming 

Zona* Awoqefl tfl«1i* Cloud Pwlkle Rortm imenrt) 


l,» 130 1JC 110 Z« W 1M UO M° i90 

Fig. 5. Zonally-averaged ice cloud particle sizes from 

cimiilattrm ncirto P A T? \A A 

THE NASA/AMES GCM: R.M. Haberle et al. 

north polar residual ice cap. The water is transported 
off the cap at low levels and is then swept up in the 
ascending branch of the Hadley circulation where it 
is quickly transported into the southern hemisphere. 
Some of this water precipitates out as it moves 
across the equator thereby moistening the lower at- 
mosphere. Once into the southern hemisphere, the 
remaining water is moved back toward the surface in 
the descending branch of the Hadley circulation 
where some of it condenses out onto the seasonal 
C0 2 ice cap which extends to about 60°S in this 

One aspect of the observations, which our model 
does not compare well with, is the thermal tides. 
The amplitudes of our diurnal and semidiurnal sur- 
face pressure tides are significantly lower than ob- 
served at either Viking lander site, or the Pathfinder 
site. Though this is not necessarily a serious flaw, 
the fact that other GCMs do find good agreement has 
motivated us to better understand the reason for our 
weaker tides. Given the sensitivity of the tides to 
dust heating, this is the obvious thing to explore 
first. So we have begun simulations with GCM 2.0 
to determine the tidal response to different assump- 
tions about the dust radiative properties. We hope to 
report these results at the workshop. 

Toon, O.B, C.P. McKay, and T.P. Ackerman, and 
K Santhanam,1989: Rapid calculation of radiative 
heating rates and photodissociation rates in inho- 
mogeneous multiple scattering atmospheres, J. 
Geophys. Res.,94, 16287-16301. 


Banfield, D., B.J. Conrath, J. Pearl, M.D. Smith, 
P.R. Christensen, and R. John Wilson, 2002. 
Forced Waves in the Martian atmosphere from 
MGS TES Nadir data. Icarus, In Press. 

Cantor, B.A, P.B. James, M. Caplinger, and M.J. 
Wolff 2001. Martian dust storms: 1999 Mars Or- 
biter Camera observations. J. Geophys. Res. In 

Colaprete, A. 2000; Clouds on Mars. Ph.D. Thesis, 
University of Colorado. 149 p. 

Colaprete, A., and R. Haberle, 2001 : Initial results 
from the NASA/ Ames GCM Carbon Dioxide 
cloud model. Abstract, XXVI European Geophysi- 
cal Society meeting, Nice, France. 

Haberle, R.M., J.R. Murphy, and J. Schaeffer 2002. 
Orbital change experiments with a Mars General 
Circulation Model. Icarus. In press. 

Murphy, J.R. 1999. The Martian atmospheric dust 
cycle: Insights from numerical model simulations. 
Fifth International Conference on Mars. Abstract 
6087. Lunar and Planetary Institute, Houston TX. 

Newman, C.E., S.R. Lewis, and P.L. Read 2002. 
Modeling the Martian dust cycle. : Representa- 
tions of dust transport processes, J. Geophys. 
Res., In press. 

Ockert-Bell, M.E., J.F. Bell III, J.B. Pollack, C.P. 
McKay, and F. Forget, 1997: Absorption and scat- 
tering properties of the Martian dust in the solar 
wavelengths, J. Geophys. Res., 102, 9039-9050. 

Toon, O.B, R.P. Turco, D. Westphal, R. Malone, 
and M.S. Liu, 1988. A multidimensional model 
for aerosols: Description of computational analogs,