Title: GISS Model 3 Development
1GISS Model 3 Development
2Active Areas of Development
- Dynamic Ocean
- Sea Ice Dynamics
- Finer Horizontal Resolution
- Finer Vertical Resolution
- Clouds and Convection
- Parallelization
3(No Transcript)
4(No Transcript)
5(No Transcript)
6(No Transcript)
7(No Transcript)
8Ocean Development
- Finer horizontal resolution
- Better regional simulation
- Variable horizontal resolution
- lt1 in tropics and at poles 2-3 at mid-lat
- To capture ENSO events
- Better depiction of polar processes
9Sea Ice Modeling
- Inclusion of modified Zhang et al. shearing
stress terms for sea ice dynamics - Utilizing Russell approach
- Reduces speed of sea ice advection out of the
Arctic
10Impact of Model Resolutionon Tracer Transports
- Increased Horizontal Resolution
- Similar inter- and intra-hemispheric resolution
- Similar vertical transport in troposphere and
trop/strat exchange (except at high N. Lat.) - Increased Vertical Resolution
- Faster interhemispheric transport
- Slower strat/trop exchange older age of air in
stratosphere, less leaky stratospheric tropical
pipe - Both Increased Vertical and Horizontal Resolution
- Added tropical EKE, effects on interhemispheric
transport and trop/strat exchange
11Finer Horizontal Resolution Currently running 1
x 1 model (53, 102 layers)
12Finer Vertical Resolution
- QBO obtained with 102 layers (and added
directions for gravity wave drag) - Also running 150 layer model
- (Both with tops at the mesopause)
- Note Finer horizontal without finer vertical
produces little impact on trace transports
13(No Transcript)
14(No Transcript)
15Clouds and ConvectionA. Del Genio
- Explicit Updraft Calculation for deep convection
- Explicit Downdraft Calculation (later)
- Cumulus Anvil Dynamics (later)
16GCM cumulus updraft speed diagnosis based on
Gregory (2001)
Observed (Zipser and Lutz 1994)
GCM
Combined with Marshall-Palmer DSD and empirical
size-fallspeed relations for liquid/graupel/ice,
allows for interactive estimates of convective
precipitation efficiency and effect on anvil
cloud feedback
17Stronger thunderstorms and tornadoes in a warmer
climate
Most lightning is over land, where surface is
warmest
Severe storms and tornadoes occur when strong
updrafts combine with strong winds aloft (wind
shear)
?
Change in of storms in 5 yrs, current to warmer
climate
TRMM LIS OTD lightning flash rates used to
evaluate model predictions of storm strength
Stronger updrafts near freezing level imply
lightning will occur 10 more often in storms in
a warmer climate
0C warmer climate
?
FEWER STORMS MORE STORMS
? 0C current climate
Strong updrafts will occur more often with strong
wind shear in a warmer climate, i.e., there will
be more of the strongest tornadoes
Del Genio et al. (2007), submitted to
Geophysical Research Letters
18Parallelization
- Goal To allow Model 3 to be run on distributed
memory machines (in addition to shared memory) - Attempts High Performance Fortran
- Speed-up not impressive
- Cluster Open MP
- Negotiating with GSFC to buy it (not optimistic)
- MPI
- Starting with the dynamics
19Comparison of Tracer Transports
- Meteorology
- Interhemispheric Transport
- Intrahemispheric Transport
- Vertical Mixing within the Troposphere
- Transport from the Troposphere to the
Stratosphere - Transport from the Stratosphere to the
Troposphere - Transport within the Stratosphere
20On-line Tracers
- CO2, CH4, N2O, CFC-11, SF6, Rn222, Bomb 14C, and
O3 - Sources as specified in previous papers
- Stratospheric chemistry from LINOZ (CH4, N2O,
CFC-11, O3) - Tropospheric chemistry Prather (CH4), Mickley
(O3) - Simulations run two ways with and without
stratospheric tracer O3 influencing the radiation - 12 year runs (nominally)
21(No Transcript)
22Results Meteorology
- Temperature
- In the coarser vertical resolution models, S.H.
lower stratosphere polar region not cold enough
in winter, too cold in summer - F102 slightly too cold at tropical tropopause
- Zonal wind
- For Dec-Feb, spread in observations of the
strength of the stratospheric jet makes
comparisons difficult - In most models (not M23) stratospheric summer
easterlies are somewhat weak - Specific humidity
- Water vapor minimum at tropical hygropause is too
low in some models (varies depending on whether
interactiveor not) - Water vapor in mesosphere is too large
(parameterized CH4-gtH20)
23(No Transcript)
24(No Transcript)
25Interhemispheric Transport
- Difference in hemispheric concentration divided
by transport across the equator (corrected for
percentage of source in SH (8 for SF6 and CFC11,
1 for CH4) - Range of a factor of two F102 the quickest, E23
the longest (Model E values are longer in
general) - In Model 3, quicker transports with finer
vertical resolution horizontal resolution has
little effect on transport times - Model studies have found IHT produced 2/3 by
eddies M53 a bit closer to this value - Model 3 has both greater tropical eddy energy and
stronger June-August Hadley Cell intensity both
associated with greater tropical precipitation
over land _at_18N in that season (especially the
finer grid models) in general, ratio of tropical
precip over land/ocean is less in Model E - Upper-level convective divergent outflow has been
suggested as the principal mechanism of IHT
Prather et al., 1987 Hartley and Black, 1995,
including convection over land regions (e.g.,
Amazon, equatorial Africa, India) (Lintner,
2003). The proportion of total vertical transport
above 500mb associated with moist convection (as
opposed to large-scale transport) is 60-70 in
M53, F53, and F102 it is 54 in M23, and about
45 in E20 and E23. Another reason therefore for
the longer IHT times in Model E is its reduced
relative effectiveness of convective transports.
26Observed values 0.7 to 1.8, but 3D values in
model intercomparisons (SF6)are 0.81?0.2. NCEP
reanalysis winds with MATCH Model (CFC11) give
0.8 (3D value). Observed interannual variability
5-6.
27- Observed contribution to ITH 2/3 by eddies
- Observed June-Aug Hadley Cell magnitude
190-270(109kgs-1) - June-Aug Land Precip
- Central America N. India
- Observed 6-10mmd-1 8-10mmd-1
- Model 3 6-10mmd-1 6-10mmd-1
- Model E 3-6 mmd-1 6- 8mmd-1(E20 lower value)
28Intrahemispheric Transport
- The mixing within each hemisphere is primarily a
function of eddy transports from the mid-latitude
sources, with subsequent involvement of the mean
circulation. The standard deviation between the
models of the eddy kinetic energy in the N.H.
troposphere is 6 - Mid-latitude/equatorial ratio For CFC-11
tracer, all the models produce a ratio between
the mid-latitude and equatorial region
concentrations of 1.10-1.14. Observed values for
this ratio covering the analogous time period for
the 1980s are 1.10 (Kaye et al., 1994). Observed
values for the SF6 ratio in the marine boundary
layer are similar (1.06) (Denning et al., 1999),
while ratios for Kr85 are again of similar
magnitude (1.16) (Jacob et al., 1987) - NH Mid-latitude/high latitude ratio, all the
models produce similar values at the surface of
1.07?0.01, which is slightly higher than
indicated by the sparse observations shown in
Denning et al. (1999) (1.02) for an Atlantic
transact during two months. In the N.H., the
similarity in EKE leads to similar
intrahemispheric mixing properties in all the
models - In the Southern Hemisphere, most of the models
have similar magnitudes of EKE except for F102,
which has about 15-20 more - SH Mid-latitude/high latitude Some variation
among the models but not consistent from tracer
to tracer. Observations show little extratropical
gradient in the S.H. for SF6 and CO2 (Denning et
al., 1999 NOAA CMDL sampling network
http//cdiac.ornl.gov/trends/co2/cmdl-flask/cmdl-
flask.html) for which all the models show only
small gradients - Overall, 4x5 versions of Model 3 have slightly
better (hence less) variation between the tropics
and extratropics in the two hemispheres, while
Model E has the largest.
29(No Transcript)
30Vertical Mixing Within the Troposphere (I)
- For 222Rn Model E has lower values 500-300mb,
where F102 has most - Model E has higher values 200-100mb, where F102
has least - In all the models, both large-scale vertical
transport and moist convection remove 222Rn from
the region below 800mb. The large-scale transport
dominates in lifting 222Rn up to about 500mb,
then convective transport dominates above,
although both are generally positive throughout.
The large-scale vertical transport is greater by
eddy transports than by the mean circulation
(which is slow given 5 day e-folding time) - The differences that arise in the vertical
profiles are thus due to differences in eddy and
convective processes - The reduced values in Model E in the 500-300mb
region are primarily due to smaller vertical eddy
convergences, while the higher values at levels
above 200mb result from both eddy and convective
effects. These differences are not due to
differences in convection overall, just over
land, where model E convection extends to
somewhat greater heights. - Model E convective fluxes maximize at a lower
altitude (900mb) than in Model 3 (800mb), and
this profile does affect the convective removal,
which follows the distribution of mass flux. - For F102, the higher values from 500-300mb are
due to convective transports, as are the reduced
values at levels above 200mb.
31(No Transcript)
32(No Transcript)
33Observations for Dec-Feb show EKE fairly close to
the general model average upper troposphere
tropical EKE may be too high in F102
34- Mahowald et al. (1997) used the MATCH model with
winds provided by NCEP and ECMWF. In the Northern
Hemisphere, the ratio of concentration at 300mb
to that at the surface at upper mid-latitudes
varied from 20 (NCEP winds) to 12 (ECMWF data).
The finer resolution models used here (M53, F53,
F102) all had ratios close to 13, close to the
ECMWF value, while the values with the coarser
resolution models were lower (M23, 9.7 E23,
9.2 and E20, 7.1). - However, the MATCH model over-predicted the upper
troposphere values in comparison with specific
observations, by about 2.5 if one were to apply
that to the simulations in general, it would
reduce the observed ratio to 5-8 (assuming the
source was not similarly overpredicted), more in
line with the lower vertical resolution models - Results are very sensitive to model convection
schemes (Mahowald et al. ,1997). The GCMs
discussed here all use a generally similar
scheme, hence their results do not differ as much
from one another as have occurred in other model
comparisons (e.g., IPCC, 1994).
35Vertical Mixing within the Troposphere
(II)Boundary layer processes
- Seasonal variation Only F102 has max in summer,
as in the observations for Chester, Pa., with
accurate maximum values - Other seasonal variations at mid-latitudes are
different - Diurnal variations are similar in the different
models - Peak values do not depend on vertical (or
horizontal) resolution nor on these model physics
differences - Boundary layer heights (peak or minimum values)
do not depend primarily on resolution - There is some increase in surface wind velocity
with finer resolution, primarily horizontal but
also vertical to some extent. Even here, however,
the different details of the boundary layer
formulation in Model E are apparent, as this
model has higher velocities over land, but lower
over many ocean grid boxes, again independent of
vertical resolution issues.
36(No Transcript)
37Dashed lines are observations (Prather and Jacob,
1990)
38(No Transcript)
39Rn222
40(No Transcript)
41Transport from the Troposphere into the
Stratosphere (I)
- The ratio between their tropospheric and
stratospheric burden is a measure of the
transport from the lower to the upper region - The model E runs tend to have higher
stratospheric concentrations, indicating faster
transport from the troposphere to stratosphere - The relative effect is the same with all the
species, but in absolute value, it is largest
with CFC-11 and SF6, and smaller with CH4 and
N2O. When the stratospheric sink is at lower
levels in the stratosphere, as in the case of
CFC-11 (whose peak chemical destruction in the
different vertical resolution models occurs
between 28 and 43 mb), the increased flux from
the troposphere is relatively rapidly destroyed.
When the sink is at higher levels, there is more
time for the species to be advected horizontally,
and back down into the troposphere, minimizing
the difference between the model simulations. For
SF6, the higher model E values are due to the
rapidly increasing source, which prevents the
return flux from stabilizing the distribution. - Vertical transport through 100mb finer
resolution models have less transport or even
negative values right near the equator - In the troposphere and extending up to 150mb, the
vertical velocity distribution is dominated by
the surface characteristics, with rising air over
the continents, sinking air over the eastern
parts of the oceans, and rising air over the
western Pacific. This effect is common to all
the models.
42(No Transcript)
43Value estimated from CH4 observations 8.9
44(No Transcript)
45Transport from the Troposphere into the
Stratosphere (II)
- There is a further tendency for sinking air to
occur in the stratosphere (80mb) above the
region of extensive rising motion from the
western Pacific again this occurs in all the
models. - However, in the finer resolution models, this
sinking extends down to the 100mb level in the
western Pacific, especially in F102, while in the
coarser resolution models it does not. It is this
difference in vertical velocity over the western
Pacific that is the primary reason for the
differences in vertical transport seen in the
equatorial region - Observations have shown that the 100mb region is
the level for which temperature variations of
opposite sign occur in the western Pacific region
with warming below, there is cooling above and
vice versa - The annual 100mb temperatures are colder above
the western Pacific in F102 (83C) than in E23
(-81C) and E20 (-79C), and the subsidence may
be in response to the cooling that has produced
the colder air. - With the greater eddy energy in F102, there is
greater cooling due to eddy energy divergence
(-0.6Cd-1) compared to E23 (0.0Cd-1) and E20
(-0.1Cd-1) which had the least energy M53, with
the next most eddy energy, had the next higher
eddy energy divergence (-0.4Cd-1) in F53 it was
0.3C d-1, and intermediate amounts of
subsidence and reduced vertical transport
46(No Transcript)
47Transport from the Troposphere into the
Stratosphere (III)
- Another factor influencing the net vertical
transport is the transport downward back into the
troposphere in the vicinity of the subtropical
jet. - The result in E23 is quite anomalous compared to
the other models, with very strong downward and
upward transports in the vicinity of the jet. It
is associated with an oscillation in the mean
circulation (it does not show up in the eddy
transports), and extends throughout the
troposphere. It appears to be the result of the
gravity wave drag parameterization, which has a
similar variation, its impact oscillating with
latitude, and resulting in alternating
convergences and divergences. - Even averaging out the oscillations, E23 has
considerably larger downward SF6 transport in the
vicinity of the subtropical jets, to balance its
increased vertical transport through the
equatorial region.
48(No Transcript)
49Transport From the Stratosphere into the
Troposphere
- The observed residence time of bomb 14C (linear
fit to the logarithmic fall-off with time) varied
with duration from the bomb release as the time
from the explosion increased, the atmospheric
circulation was able to transport material to
higher levels in the stratosphere - F102 is the most realistic, and model E runs the
least realistic (interactive runs do better) - Finer grid models have smallest net transport of
ozone from strat into trop - Longer residence time with greater reduction of
eddy energy with altitude in the upper trop/lower
strat - greater reduction with finer resolution
(less upward energy flux also better resolution
of the fall-off with height) - Model E also has a significant component of
downward transport due to the mean circulation
(gravity-wave drag-induced). - Differences also apply to flux of O3 into the
stratosphere - finer resolution (and interactive)
models produce reduced fluxes, and better
comparison with radiosonde data in the
troposphere (though all models still too high at
high latitudes). - The model 3 simulations give slightly larger
downward ozone transports in the Northern
Hemisphere as in observations (e.g., Olsen et
al., 2004), while in model E the transports are
somewhat larger in the Southern Hemisphere. This
is due to greater downward eddy transport of
ozone at S.H. mid-latitudes associated with the
large gravity wave dissipative effects in the
N.H. in model E.
50(No Transcript)
51- Observed short--term (36 month) residence time 4
years - Observed longer term residence time (90 months)
5 years - Observed transport of ozone 400-600 (as high as
800)Tg/yr NOTE observations are through the
tropopause, values here are through 117 mb
52(No Transcript)
53(No Transcript)
54(No Transcript)
55Transport Within the Stratosphere
- Age of air calculated by correlation with SF6
increase at the surface or in the upper
troposphere. - E20 has the youngest air in the upper
stratosphere, influenced by the presence of the
model top. - In general, the finest resolution models have the
oldest age of air throughout the middle
atmosphere. - In the middle stratosphere, E23 has the flattest
distribution, indicative of the most leaky
tropospheric pipe - The finer resolution models have less leaky
tropical pipes, while the Model E values are more
leaky. This can be related at least partly to the
gravity wave drag parameterization in model E (or
Rayleigh friction in E20), which in the lower
stratosphere (including the subtropics) is 10x
stronger than that in Model 3 when acting on a
west wind, the drag forces a poleward flow which
helps mix air out of the tropical pipe region - At 45 mb, Model 3 values are generally better
than the models used for the MMII comparison. - Interactive models have older (and more
realistic) age of air values (as much as 50
older in the S.H., 30 in the N.H.
56(No Transcript)
57(No Transcript)
58(No Transcript)
59(No Transcript)
60(No Transcript)
61(No Transcript)
62Interactive Runs
- Differ in O3 because LINOZ does not have O3-hole
parameterization - Hence they have less O3 in lower stratosphere
- Warmer temperatures, and stronger summer east
winds, are improvements relative to normal model
simulations - Increased stability reduces tropospheric EKE by a
few , reduced upward EP flux into the
stratosphere of 10-15, and a reduction in
stratospheric EKE by 20-30 between 100 and 10mb - This leads to a reduction in EP flux convergence
of 20-30 and reduced stratospheric residual
circulation - less upward flux through the
tropics, less downward flux at high latitudes as
well as older age of air - Effect is felt more strongly in the S.H.
63(No Transcript)
64(No Transcript)
65(No Transcript)
66(No Transcript)
67(No Transcript)
68(No Transcript)
69Step-Mountain Technique Applied to an Atmospheric
C-grid Model, or How to Improve Precipitation
Near MountainsGary L. Russell, Revised for
Monthly Weather Review, 2007/01/10
- Variable Number of Vertical Layers fitting
between - bottom topography and model top
- Step-mountain is best implemented on C-grid
- Reduces systematic horizontal advection errors
- Which reduces erroneous vertical mass fluxes
- Which improves precipitation distribution
- Option Modify topography to provide smoother fit
- from grid box to grid box
70Downward Mass Flux (C Grid)
71(No Transcript)
72(No Transcript)
73OTHER MODEL DEVELOPMENT AREAS
- Aerosols (S. Bauer, D. Koch)
- Dust, sulfate, black carbon, nitrate, sea salt
- Indirect Effects f( Cloud droplet number)
- Testing bin microphysics scheme
- Chemistry (D. Shindell)
- Oxidants (H2O2, O3) interactive with aerosol
chemistry - Heterogeneous Chemistry
- Need hydrocarbons (methane, terpines, isoprene)
to be - interactive with biospheric state
- Boundary Layer (V. Canuto, Y. Chen)
- Cloud conserved variables liquid water potential
temperature, total water (Adrian Lock
parameterizations) - Sea ice (G. Schmidt, D. Rind, R. Bleck)
- Sensitivity too low
- Move to Ocean Grid
74Comparison of Models
- Similarities
- numerical calculation of the dynamical equations
- atmospheric radiation
- atmospheric turbulence (calculated at all levels
of the atmosphere using a second order closure
scheme with non-local turbulence) - the land surface (including vegetation)
- most of the input files (solar irradiance,
aerosol and trace gas distribution, sea surface
temperatures and sea ice, vegetation and soils)
75- Differences
- Calculation of surface fluxes Model 3 uses
virtual potential temp - Clouds and convection different choices
regarding - downdraft entrainment (none is used in Model 3)
- the partitioning between convective precipitation
and detrainment into anvil clouds (a fixed
percentage, 50 is used in model 3, a three part
formulation in model E) - partitioning of subgrid convective and
non-convective areas (which affects the choice of
the cloud formation threshold relative humidity) - cloud water evaporation (none is used in Model E)
- Net effect
- Cloud liquid water content some 70 higher in E23
but cloud cover slightly higher in M23 so its
albedo is slightly higher (30.9 compared with
29.4) - Net radiation over land higher in E23 (8Wm-2)
Net radiation over ocean higher in M23 (5Wm-2)
so global value is similar - Both are improvements in M23, as E23 has
excessive solar radiation over land especially in
the Southern Hemisphere, and deficient radiation
over the ocean, especially in the Northern
Hemisphere (Schmidt et al. ,2006) - in both cases
where the biggest differences with M23 are
located. - Gravity wave drag similar formulation in E23 to
Model 3 but values in Model E are much larger
E20 uses Rayleigh friction gravity waves help
initiate high level cloud cover in Model 3 - Filter on the u,v winds only in x direction in
Model E, in x,y directions in Model 3 - Vertical resolution altitudes below/above 100mb
- E2011/9 E23, M23 13/10 M53, F53 29/24
F10258/44 - Model top all at 0.002mb except E20 (0.1mb)
76Impact of Model Physics (e.g., M23 vs E23)
- E23 has slower interhemispheric transport,
associated with reduced precipitation over land. - The convection in E23 extends to somewhat higher
levels over land in the tropics which affects
tracer transport into the highest tropical
troposphere levels - Both differences lead to reduced EKE in upper
tropical troposphere in Model E, and hence
greater transport into the stratosphere (eddy
energy produces heat divergence and subsidence) - Gravity wave drag is much stronger in Model E,
and this with the largest net transport between
the troposphere and stratosphere. The large
gravity wave drag is associated with the
anomalous oscillating vertical transports in the
upper troposphere/lower stratosphere near the
N.H. subtropical jet in E23, and helps produce
the more leaky tropical pipe in both Model E
versions. It also provides for the (spurious)
more downward ozone transport in the Southern
Hemisphere than in the north, due to its large
effects on eddy energy in the Northern Hemisphere
lower stratosphere. - Boundary layer physics in Model E seems to
provide for good boundary layer height variations
and surface concentrations despite reduced
vertical resolution - Top of the model in E20 helps produce increased
upward transports to the upper stratosphere -
result is more realistic, but method is spurious
77Impact of Model Resolution
- Horizontal Resolution M53 and F53 have very
similar transport characteristics similar
interhemispheric transports, and similar
percentage of eddy/MMC contribution to the ITH,
similar EKE and intrahemispheric transport,
generally similar vertical transport within the
troposphere (somewhat more convective mass flux
in M53, although the flux extends up to similar
levels), very similar transport between the
troposphere and stratosphere, with only slightly
longer residence time for bomb 14C within the
stratosphere. Within the stratosphere, the
leakiness of the tropical pipe is similar, while
the age of air is also quite similar except at
higher northern latitudes, where it is older in
M53. From this perspective, simply increasing the
horizontal resolution does not have much impact
on the tracer transports. - Vertical resolution effects can be tested by
comparison of M23 and M53, as well as F53 and
F102. In contrast to the situation for horizontal
resolution, vertical resolution makes a
noticeable difference for these transports. The
finer vertical resolution runs have faster
interhemispheric transport (associated with both
stronger Hadley circulation and increased
tropical eddy energy), slower transport between
the troposphere and stratosphere, older age of
air, and a less leaky stratospheric tropical
pipe. In many cases the differences are not
large, but they occur both with and without the
ozone-radiation interaction. M53 also has
increased convective mass flux and thus puts more
tracer in the upper troposphere than M23, but
this is not true for F53 vs. F102. - The combined increase in vertical and horizontal
resolution in F102 does have effects that are not
apparent when just horizontal resolution is
increased in particular the tropical EKE is much
larger, the moist convective transports dont
reach to as high a level, and there are some
differences in the boundary layer concentrations
of Rn222. The basic idea that to maximize the
ability for finer scale waves to be generated one
needs increases in both horizontal and vertical
resolutions seems to be borne out by these
experiments and the effect on tracer transports.
78(No Transcript)
79(No Transcript)
80(No Transcript)