Climate Change Research
Climate System Model for Paleoclimate
Bette Otto-Bliesner and Esther Brady, in collaboration with numerous
scientists and programmers in the Climate Modeling Section (CMS) and
Oceanography Section (OS), have developed a low-resolution version of
the Climate System Model (CSM) suitable for long, coupled, baseline
simulations and matrices of sensitivity simulations of interest to the
paleoclimate community. The low-resolution version is T31 for the
atmosphere and land models and x3 (1.8º-3.6º in lattitude,
3.6º in longitude) for the oceans and sea ice. A 130-year
simulation has been finished and is being documented. The CSM
Paleoclimate Working Group has proposed a suite of experiments using
this version of the CSM. The atmospheric results are similar to the
T42,x2 CSM 300-year simulation in terms of the large-scale dynamics
and thermodynamics. Global surface temperature averages
287.8º K. The top of the atmosphere is close to
radiative balance, having an imbalance of less than 0.6 W
m-2. Precipitation associated with the intertropical
convergence zone (ITCZ), monsoons and midlatitude stormtracks is well
simulated. The double ITCZ seen in the higher-resolution CSM is less
pronounced in the low-resolution case due to a less prominent
equatorial cold tongue.
This
table shows global, annual mean climatological statistics from
observations and T31 simulations with the Community Climate Model
version 3 (CCM3) (climatological sea surface temperatures and sea-ice)
and the CSM.
This figure (20K)
shows mean precipitation for DJF (top) and JJA (bottom) averaged for
years 100-109 of T31,x3 CSM simulation.
The sea-ice results are significantly improved from the 300-year simulation
with the T42,x2 CSM. Several factors led to this improvement. An imbalance
in the ice-ocean heat flux calculation during the spinup of the high-resolution
simulations was corrected for our simulation. The resulting Arctic and
Antarctic ice volumes and ice areas are in good agreement with observational
estimates and exhibit a reasonable seasonal variation. Arctic ice volumes
also suggest longer period variability. The air-ice drag parameter of the
atmosphere on the sea ice was reduced to a value in better agreement with
measurements for the low-resolution case, resulting in a significantly reduced
meridional ice transport in the Southern Hemisphere in better correspondence
with satellite-derived estimates.
This figure (7K)
shows total ice volumes and areas for the T31,x3 CSM simulation. The
first 60 years are for the uncoupled ocean-sea ice spinup forced by
atmospheric data from CCM3, followed by 130 years of fully coupled
integration. Gray lines denote maxima and minima areas from satellite
data.
This figure (12K)
shows meridional ice volume transports in the Southern Hemisphere from
year 100 of the T42,x2 and T32,x3 simulations and from satellite
data.
Improvements in the Southern Hemisphere
sea-ice component lead to an improved oceanic simulation of the
Antarctic Circumpolar Current (ACC) transport. Values equilibrate at
approximately 160 Sverdrups (Sv), still greater than observational
estimates of 130 Sv but greatly reduced from results of the T42,x2
300-year simulations without the correction. Trends in the ocean are
small. The global volume averaged temperature cools at a rate less
than -0.05ºC per century over the last 30 years. The global
volume average salinity shows a consistent trend toward saltier values
at a rate of 0.010 parts per thousand per century over the entire
length of the coupled integration. While the trends in the global
volume averaged temperature and salinity are acceptable, there are
larger drifts within the water column that will need to be improved
before conducting a multi-century long integration.
This figure
(16K) shows annual-average barotropic
streamfunction.
Oceanic Circulation of the Campanian (80 Ma)
Brady and Robert DeConto continued to study the oceanic climate of the
Campanian using an ocean general circulation model (GCM) forced by a simulation
of the atmospheric climate from the Global Environmental and Ecological
Simulation of Interactive Systems (GENESIS) climate model. This simulation
revealed an oceanic climate very different than today. Warm, salty bottom
water (temperature of approximately 10ºC compared to approximately
3ºC at present) is formed primarily at high southern latitudes as warm,
saline water originating in the south Atlantic Ocean that is transported
poleward and cooled. Additional experiments have been conducted to test
the sensitivity and robustness of this result to smoother bottom topography
and an open ocean salt exchange with an idealized marginal sea. The main
conclusions are robust to these sensitivity experiments. However, the details
of how the warm saline deep water spread globally to fill the deep ocean were
significantly different in these sensitivity experiments. Papers describing
these results have been submitted to Nature and Geophysical Research
Letters.
This figure
(25K) shows Campanian and present annual sea surface temperatures
(C) simulated by the ocean GCM.
In the summer of 1997, a Significant Opportunities in Atmospheric
and Related Sciences (SOARS) student, Sharon Pérez-Suárez, in
collaboration with Brady, DeConto, and Otto-Bliesner, compared these model
results to proxies for ocean temperature calculated from fossil shells found
in ocean sediment cores. Temperature proxy estimates from oxygen isotopic
ratios from planktonic organisms show much cooler tropical SSTs compared to the
model. This has come to be known as the "Cool Tropic Paradox," suggesting
that the Cretaceous tropical SSTs were cooler than today despite the overall
Cretaceous warmth and higher CO2. However, this so-called paradox
is easily explained when the depth of the habitat (upper 100-200 m) for these
organisms is taken into account. At these depths the tropical thermocline is
affected by equatorial upwelling that brings colder water to the surface.
Comparing to model temperatures at 100 to 200 m depths, the temperature proxy
estimates compare reasonably well. These results were presented at the 1997
American Geophysical Union (AGU) Fall Meeting in San Francisco.
This figure
(11K) shows planktonic temperature estimates superimposed on
zonally-averaged temperature estimates simulated in the top seven
layers of the OGCM.
Participation in the Paleoclimates from Arctic Lakes and Estuaries
(PALE) Project
Benjamin Felzer has completed a present-day
simulation of the North Atlantic region using the Arctic Region
Climate System Model (ARCSyM) mesoscale model forced by the European
Centre for Medium-Range Weather Forecasts (ECMWF). This experiment was
designed to assess the ability of the regional climate model to
accurately simulate the climate of the North Atlantic region for
future paleoclimate simulations. This simulation was driven at the
boundaries of the region by ECMWF reanalysis data from September,
1987, to March, 1990. Several sensitivity experiments using various
precipitation parameterizations were also completed. The model
accurately simulates the primary circulation features in the region,
including katabatic surface winds over Greenland and the Icelandic and
Baffin Bay lows. Precipitation patterns also correctly predict
orographically-forced precipitation maxima along Baffin Island,
Ellesmere Island, and Greenland, and are improved with the addition of
the explicit non-convective scheme, which involves prognostic
equations for cloud water and rainwater mixing ratios. This model is
currently being forced with a 0 and 6 ka before present (BP) CSM
simulation to provide the basis for high-resolution data-model
comparison of 6 ka BP.
Felzer also completed a more realistic simulation of 21 ka BP using the
GENESIS model 2.0, along with the Equilibrium Vegetation Ecology model
(EVE), without an ice sheet in East Siberia, to assess the impact of this
ice sheet. A previous simulation of the Last Glacial Maximum (LGM) was
similar to this experiment except for the presence of the ice sheet. The
Peltier ICE-4G reconstruction of the ice sheets, which is used in most
model simulations of the LGM, includes an ice sheet in East Siberia. There
is very little data to support the existence of this ice sheet. The presence
of an East Siberian ice sheet results in warmer temperatures in Siberia
south of the ice sheet, due to subsidence resulting from anticyclonic flow
during winter and reduced northerly flow during summer. This less harsh
climate results in less polar desert and more tundra with the ice sheet
present.
This figure
(10K) shows winter (DJF) surface winds, ECMWF versus ARCSyM.
ARCSyM accurately simulates the katabatic surface winds over Greenland
and the Icelandic and Baffin Bay lows, though northerly flow over
Baffin Bay is weaker than observed.
This figure
(45K) shows biomes at 21 ka BP with and without the East Siberian
ice sheet. The ice sheet induces warmer temperatures south of the ice
sheet in Siberia, resulting in less polar desert and more tundra
throughout Siberia.
Jon Bergengren and Starley Thompson completed their development and application
of the EVE predictive vegetation model. EVE uses monthly mean climate data to
simulate the equilibrium state of natural vegetation and was developed for
coupling to global climate models. Three papers describing the EVE model and
its initial applications have been submitted for publication. Felzer,
Bergengren, David Pollard, and Thompson collaborated on GENESIS version 2.0
simulations of the climate of 6,000, 10,000, and 21,000 years ago using a fully
interactive version of the EVE predictive vegetation model running
synchronously with the climate model. Some large-scale features of the
predicted vegetation match vegetation changes estimated from paleo records,
but some vegetation model shortcomings were also apparent.
A high-resolution dynamic ice-sheet model was coupled to the GENESIS
Global Climate Model by Pollard and Thompson. The ice-sheet model uses
a standard vertically-integrated flow law to predict ice thickness
versus longitude and latitude, with the bedrock topography responding
towards local isostatic equilibrium under the ice weight with a time
lag of 3000 years. The coupled global climate model ice-sheet model
system has been applied to several geologic periods, including the end
of the last interglacial and the last glacial maximum. Results of the
simulations indicated, as expected, that the quality of the global
climate model-produced surface mass balance determined the quality of
the simulated equilibrium ice sheets. Future work will need to
emphasize improvements in the global climate model mass balance
simulation and better paleo ice sheet reconstructions.
The PCM makes use of model component configurations from Department
of Energy (DOE) Climate Change Prediction Program researchers and uses new
massively parallel processor (MPP) computers. The coupled climate model
will conduct multi-century climate change experiments. Special emphasis
will be given to energy emission scenarios that are of interest to the
DOE.
Ocean Model Component
Robert Chervin, Vincent Wayland, and Anthony Craig have developed
an ocean component with collaboration from Los Alamos National
Laboratories (LANL) and the Naval Postgraduate School (NPS) that uses
the Parallel Ocean Program (POP) model with a displaced North
Pole. The grid has an average resolution of 2/3 degree
latitude and longitude with increased latitudinal resolution near the
equator of approximately 1/2 degree. Because of the
displaced pole, there is relatively higher horizontal resolution in
the eastern North Pacific, in the Arctic Straits near northern Canada
and Greenland, and in the Gulf Stream area. Also, the continents and
bottom topography were carefully modified to obtain realistic flow in
many regions throughout the globe. This model has been spun up with
CCM3 daily five-year forcing. The model is presently running
efficiently on the NCAR Cray T3D, the National Energy Research
Supercomputing Center (NERSC) Cray T3E, and the LANL Silicon Graphics,
Inc. (SGI) Origin 2000. The NCAR Hewlett Packard (HP) parallel
computer version has been developed in collaboration with Richard Loft
and Rodney James of the Scientific Computing Division (SCD). Very
recently, Craig, Chervin, and Wayland converted the model
configuration to the new version of POP that has more ocean
parameterization options. The model in its present form yields an
extraordinary simulation of the Arctic Ocean, tropical Pacific, and
boundary currents, such as the Gulf Stream, with ocean eddies resolved
in most basins.
Sea-Ice Model Component
The sea-ice model has been implemented in an eddy-resolving model by Yuxia
Zhang of NPS and optimized for MPP architecture by Craig. It uses the Zhang
and Hibler ice dynamics with line relaxation for solving the viscous-plastic
ice rheology. The thermodynamics are from the Semtner and Parkinson-Washington
models. The grid is transformed such that the resolution is constant, thus
avoiding the problem of convergence near the pole as on a latitude-longitude
grid. This grid will require an additional interpolation of atmosphere
and ocean variables. The spatial resolution of Zhang's model is about 18
km, which provides a realistic Arctic simulation of eddy resolving ocean
and sea-ice motion. Recently, she has also applied this model in the Antarctic
region, again with realistic eddies simulated. For the coupled model, a sea-ice
model grid with 27 km resolution has been implemented that includes
all of the present day ice-covered areas in both hemispheres, minimizing
the grid space required. John Weatherly and Zhang are improving the
thermodynamical aspects of the sea-ice model by adding more realistic treatment
of snow and sea-ice. The new sea-ice method of Elizabeth Hunke and John
Dukowicz (LANL) has been implemented in the coupled system. This model
component has been converted to the SGI Origin 2000 and HP parallel computer
systems.
Atmospheric Model Component
The atmospheric component is the massively parallel version of the
CCM3. This model includes the latest versions of radiation, boundary
physics, and precipitation physics and is a state-of-the-art
atmospheric component. This model has been coded to run on the T3D,
T3E, SGI Origin 2000, and HP parallel computer system.
Flux Coupler
The method of tying the components together and allowing the exchange
of fluxes and variables is the flux coupler. The flux coupler has undergone
testing and implementation by Tom Bettge, Loft (SCD), John Dennis (SCD),
and Steven Hammond (SCD). Since the grid components are different, there
is an interpolation scheme for passing information between the atmospheric
component grid and the ocean/sea-ice grid that has been developed by Philip
Jones (LANL). It has been successfully adapted to the T3D, T3E, SGI Origin
2000, and HP systems.
Concurrent Versions System (CVS)
Because PCM will be running on a variety of MPP computers, we have
incorporated CVS to allow for keeping track of different model
versions and updates to the model. The Climate and Global Dynamics
(CGD) Division, LANL, NPS, and SCD staff are all using this version
control system.
Coupled Models
Bettge, Gary Strand, and Craig have spun up the ocean and sea-ice models
using the same method used by the CSM to minimize the initial drift of the
coupled system. This method has also been useful in demonstrating
and improving the kind of adjustments that initially occur in the ocean
and ice due to coupling the CCM3, without having to run the more expensive
coupled system. The full PCM system with atmosphere is undergoing testing.
In summary, this new massively parallel coupled climate model
system takes advantage of the latest high performance computer
technology. The model is flexible enough to allow changes for new
components. This effort is complementary to the CSM effort in that the
flux coupler concept will be used, the same spin-up technique will be
used, and the CCM3 will be used; however, the ocean and sea-ice will
be at a much higher resolution with more detailed processes. The use
of this model will be primarily for studying various greenhouse gas
and sulfate aerosol emission strategies of interest to the DOE and the
U.S. Global Change Research Program (USGCRP).
The following are the scientists/programmers involved in the coupled
climate model effort in alphabetical order: J. Arblaster (NCAR), T. Bettge
(NCAR), A. Craig (NCAR), J. Dennis (NCAR), J. Dukowicz (LANL), J. Hack
(NCAR), S. Hammond (NCAR), E. Hunke (LANL), R. James (NCAR), P. Jones (LANL),
R. Loft (NCAR), R. Malone (LANL), M. Maltrud (LANL), W. Maslowski (NPS),
G. Meehl (NCAR), A. Semtner (NPS), R. Smith (LANL), G. Strand (NCAR), W.
Washington (NCAR), V. Wayland (NCAR), J. Weatherly (NCAR), D. Williamson
(NCAR), and Y. Zhang (NPS).
Detailed information on the DOE/PCM modeling effort can be found on
the PCM web page.
Meehl, Arblaster, and Strand used the DOE global coupled
model to show that a number of previously identified regional decadal
mechanisms, acting in unison at a global scale, can produce
internally-generated decadal fluctuations of globally-averaged surface air
temperature. Important processes include advection of ocean heat content
anomalies embedded in the gyre circulations of the Pacific, Atlantic, and
Indian Oceans; associated low-frequency "El Niño-like" signals in
the atmosphere and ocean in the Pacific; global scale tropical-midlatitude
interactions that replenish the ocean heat content anomalies and amplify
the decadal signal; and consequent global energy balance variations. These
results imply not only that it may be possible to distinguish
internally-generated global climate variability from fluctuations due to human
activities but also that enhanced upper ocean monitoring could lead to
improved climate prediction on the decadal timescale.
Weatherly, Bruce Briegleb (CMS), and William Large (OS) are developing the
new version of the CSM sea-ice model
(CSIM), which includes a viscous-plastic ice rheology with resistance
to shear stress. This model uses a rotated spherical grid so that
resolution is nearly uniform over the polar regions, which allows a more
accurate simulation of ice dynamics. Sensitivity tests with the original
CSM sea-ice model have investigated the impacts of the CCM3 wind forcing,
the wind-drag parameter, and ice albedo on the CSM simulation. The recent
CSM simulation of the pre-industrial climate showed significant improvement
in ice transport over previous CSM runs.
This result reflects both the improved air-ice drag parameter
and the forcing from a CCM3 run that uses the observed sea-ice concentration
climatology.
This figure
(16K) shows northward sea-ice transport (zonal mean) from the
Antarctic coast, showing the improvement in the new CSM spin-up,
compared to the previous CSM spin-up, previous coupled run, and
satellite-derived estimates.
Chervin and Craig continued to use the 512 PE Cray T3D at the
Pittsburgh Supercomputing Center (PSC) to evaluate, improve, and
optimize a modified version of the 2/3 degree (on average), displaced
pole, global configuration of the POP. The standard 2/3 degree grid
(384x256x32) was modified (i.e., to 384x288x32) to include increased
latitudinal resolution near the equator to resolve the strong tropical
current systems. This version of POP has been integrated for
approximately 50 years, partly at PSC and partly on the recently
upgraded 128 PE T3D in the NCAR Climate Simulation Laboratory
(CSL). The results of this integration are quite promising and
provided considerable guidance for a spin-up procedure using
atmospheric forcing from CCM3. This version of POP has been included
as the ocean component in the PCM, a coupled model developed as part
of a multi-institutional, distributed research effort for use on
current and future generations of MPPs.
Craig and Chervin tested and validated this version of POP on newly
installed Cray T3Es at Cray Research, Inc., PSC, and NERSC under
"friendly user" status.
Chervin and Craig continued to carefully evaluate the interpolation
schemes being developed by Phillip Jones (LANL). These schemes are critical
for communication of component model information through the flux coupler
and also for the analysis of the performance of all versions of POP that
feature generalized curvilinear coordinates and a displaced pole.
Chervin, Craig, and Wayland tested, validated, and modified a completely
rewritten version of POP (new POP) to meet the PCM requirements for an
ocean component. This major POP revision was undertaken by Matthew Maltrud
and Jones (both of LANL) to improve performance of the model on cache-based MPP
systems. The PCM version of new POP was optimized by Wayland for the NERSC
Cray T3E and demonstrated excellent scalability using 8, 16, 32, 64, and
256 PEs.
Chervin, Craig, and Wayland continued to be active participants in the
PCM effort, both technically and scientifically.
Weatherly, Bettge, Craig, and Strand, in collaboration with Yuxia Zhang
(NPS), have coupled the sea-ice model of both polar regions into the PCM.
It includes the viscous-plastic ice dynamics modified
for parallel processors, using a uniform grid with 27 km resolution. This
separate sea-ice model grid requires additional interpolations of atmospheric
and oceanic fluxes by the PCM flux coupler, which are done in a flux-conserving
manner. Spin-up runs of the sea-ice and ocean components of PCM using CCM3
forcing result in Arctic and Antarctic ice simulations with ice motions
at eddy-permitting scales and spatial variations in ice concentrations
and thickness.
The sea-ice model can also use the new elastic-viscous-plastic ice
rheology of Hunke and Dukowicz (LANL), an explicit scheme with greatly
improved parallel performance that compares well with previous sea-ice
models. Images from this model can be found on the PCM ice page.
Washington, Meehl, and Arblaster have been studying the changes in the ocean
caused by global warming in a coupled climate model. The particular climate
model experiments shown here have both the effects of realistic amounts
of increasing greenhouse gases and increasing sulfate aerosol distributions.
Meehl et al. (1996) have discussed the surface and atmospheric response
of the climate system and intercompared them with other modeling groups as part
of the Intergovernmental Panel on Climate Change (IPCC) study. In this work,
we have examined the limited "sections" of the ocean where observations of
temperatures have been taken at 25N in the North Atlantic by Parrilla et al. (1994) in which they found from 1957, 1981, and 1992 surveys a warming of
0.32ºC in the upper part of the ocean and cooling below. This implies a
warming rate of about 1ºC per century. We have found a similar pattern in
our simulations with a coupled model as shown below.
It should be noted that in both the observed and simulated climate
system, there seems to be a great deal of interannual and decadal
variability superimposed on the climate change signal. This points to
the difficulty of establishing climate signals in the ocean because of
the lack of long-term measurement systems. The most important point is
that the climate model shows an ocean warming pattern at 24N that is
consistent to the observed changes.
This figure
(18K) shows a general warming from 100-3000 m and cooling in the
top and bottom parts of the ocean. The simulated ocean temperature
difference at 24N from 75ºW to 25ºW from a coupled climate
model. The difference is computed from a simulation that has
increasing carbon dioxide and the direct and indirect effects of
sulfate aerosols. The years used are close to Parrilla et
al. (1994) in that we used 1985-95 average minus the 1960-70
average.
Meehl, Washington, Bettge, and Strand used the DOE global coupled GCM
without flux adjustment to perform a set of 75-year sensitivity experiments
with various combinations of increasing CO2 and
the direct and indirect
effects of sulfate aerosols, as well as two integrations simulating the
climate from 1900 to 2035. One included only increasing
CO2, and the other had increasing CO2 and sulfate
aerosols. Results show that the cold start
effect (the delay in climate response due to the thermal inertia of the
ocean) plays a role even on the 30-year timescale. The experiments with
forcing from 1900 to present (with 1900 starting at zero anthropogenic
forcing) were already about 20 to 30% warmer than the sensitivity experiments
with forcing from 1960 to present (with 1960 starting at zero anthropogenic
forcing). The increased initial warming in the experiments started in 1900,
then continued into the future climate simulations. Thus, the spin-up of
the earth's climate to gradually increasing anthropogenic forcing is important
in a time evolution sense at least for the duration of this century.
Experiments must be started at least early this century and run into the
future to give a reasonable estimate of the simulated magnitude of climate
change due to the cold start characteristics of the global coupled model.
Meehl analyzed surface fluxes from the component models and the full
coupled climate model to show that features of the coupled model simulation
are a combination of errors in the component models, as well as errors introduced
due to the dynamic interaction, both local and non-local, between atmosphere
and ocean. Thus, in the coupled model, a good simulation of net surface
heat flux does not always produce a correspondingly accurate simulation
of SST. Conversely, a good simulation of SST in the coupled model can be
associated with surface heat flux errors also due to dynamic adjustments
in the atmosphere and ocean in the coupled simulation.
Meehl and Arblaster analyzed output from the CSM and documented
characteristics of the Asian-Australian monsoons and
El Niño-Southern Oscillation (ENSO) in the coupled model. Typical of
other models of its class, the amplitude of ENSO events in the CSM is about 60%
of the observed amplitude. There are significant global patterns of teleconnections
associated with ENSO events in the CSM as seen in the observations. These
include connections to the Asian-Australian monsoons such that when there
are anomalously warm SSTs in the equatorial eastern Pacific in the model,
the monsoons tend to be weaker than normal as observed.
Meehl is the chairman of the Coupled Model Intercomparison Project
(CMIP), an activity organized by the World Climate Research Programme
under the Climate variability and predictability (CLIVAR)
project. CMIP is designed to intercompare control and CO2
climate change experiments from all current functioning global coupled
climate models internationally. The first phase of CMIP, CMIP I,
involves the control climates from the models and is well
underway. The second phase, CMIP II, will intercompare 1% per
year CO2 increase experiments and is in the midst of
the model data collection activity. NCAR is participating in both
phases of CMIP, with results from the DOE-funded global coupled model
and the CSM, which have already been submitted for analysis by CMIP.
