Transcript Coupled General Circulation Modeling

Coupled General Circulation
Modeling
Anthony J. Broccoli
Dept. of Environmental Sciences
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Is There a Better Set of Lower
Boundary Conditions?
• Yes! The lower boundary conditions for the
atmosphere could be determined
interactively in response to processes
internal to the model.
• This goal can be achieved by coupling the
atmosphere to an ocean model.
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Today’s Lecture
•
•
•
•
•
Types of coupled models
Coupling methods
Climate drift
Flux correction/adjustment
Design of coupled model experiments
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Types of Coupled Models
•
•
•
•
Atmosphere-swamp ocean
Atmosphere-mixed layer ocean
Atmosphere-ocean GCM
Earth system models of intermediate
complexity (EMICs)
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Atmosphere-Swamp Ocean
• Ocean is represented as a wet surface
with zero heat capacity.
• Surface temperature is interactively
determined.
• Albedo of swamp surface increases when
temperature falls below freezing.
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Atmosphere-Swamp Ocean
Atmospheric GCM
land
swamp
S (1   )  F  F  SH  LH  0
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Atmosphere-Mixed Layer Ocean
• Ocean is represented as a shallow,
motionless slab of water.
• Mixed layer depth is chosen to represent
seasonal heat storage in upper ocean.
• Ocean temperature is interactively
determined.
• Sea ice thermodynamics are included.
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Atmosphere-Mixed Layer Ocean
Atmospheric GCM
land
mixed layer ocean
T
c p h
 S (1   )  F  F  SH  LH
t
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Atmosphere-Ocean GCM
• Ocean component is a full dynamical
ocean model, including advection,
diffusion, heat storage.
• Relatively complete representation of
physical and dynamical feedbacks
between atmosphere and ocean.
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Atmosphere-Ocean GCM
Atmospheric GCM
land
cons. of momentum,
cons. of mass,
cons. of salt
cons. of thermal energy
equation of state
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
EMICs
• EMICs: Earth System Models of
Intermediate Complexity
• Designed to contain many feedbacks of
full AOGCM but consume far less
computer time.
• Used for climate simulations that require
long time scales (i.e., >1000 years).
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
EMIC Example: Ocean GCM with
Energy Balance Atmosphere
• Developed by A. Weaver and collaborators
at Univ. of Victoria.
• OGCM is coupled to simple atmosphere.
• Atmospheric dynamics represented by
diffusion.
• Highly simplified parameterization of
atmospheric radiation.
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Coupling Methods
• Communication between components is
an essential element of coupled models.
• Model component codes are often
developed separately, so grids can be
different, making regridding necessary.
• Frequency of communication must be
managed, particularly given the difference
in response times of atmosphere and
ocean.
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Coupling Methods: Example
Atmospheric model
Coupling
interface
Surface temperature
Albedo
Wind stress
P-E
Heat flux
Sea ice model
Coupling
interface
Temperature
Salinity
Ocean currents
Stress
Fresh water flux
Heat flux
Ocean model
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Asynchronous Coupling
• Atmosphere is run for a relatively short
period with output archived in “library.”
• Ocean is run (with acceleration methods)
for relatively long period using fluxes from
atmospheric library.
• Cycle can be repeated indefinitely.
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Synchronous Coupling
• Conceptually simple; no acceleration
techniques are used.
• Model components may have different
time steps, but communication occurs at a
fixed interval.
• Typical interval: 1x daily (models without
diurnal variation); 8x daily (with diurnal
variation)
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Climate Drift
• Coupled models are typically constructed
from atmosphere and ocean components
that have been independently developed.
• Stand-alone atmosphere and ocean
components are tightly constrained by
observed boundary conditions.
• When atmosphere and ocean components
are coupled, the resulting climate will often
drift away from a realistic state.
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Climate Drift in GFDL CM2
Zonal Mean SST Error from CM2_a10o2 [K]
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Causes of Climate Drift
Flux Difference [W m-2]
AGCM vs. OGCM
February 26, 2003
CM2_a10o2 SST Error [K]
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Causes of Climate Drift
• Imbalances between atmosphere-ocean
heat fluxes simulated by AGCM and
OGCM when both are run with observed
SSTs.
• Climate feedbacks triggered by flux
imbalances. (Ex: CM2_a10o2 cooling
pattern in midlatitude N.H. → southward
shift in westerlies → error in position of
western boundary currents)
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Flux Corrections/Adjustments
• One ad hoc approach to reducing climate
drift is to adjust for differences in
atmospheric and oceanic component
fluxes by adding a compensating flux at
each grid point.
• This method is known as flux correction
(Sausen et al. 1986) or flux adjustment
(Manabe et al. 1991).
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Calculating Flux Adjustments
• The goal is to determine artificial heat and
water fluxes that vary seasonally and
spatially but do not depend on the state of
the model.
• Method 1: GFDL Three-Step
• Method 2: Coupled Restore
• Method 3: Offline Flux Difference
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Method 1: GFDL Three-Step
• Step 1: Run the
AGCM with
climatological SSTs,
archiving the heat and
water fluxes.
• Step 2: Run the
OGCM with the fluxes
from step 1, while
simultaneously
restoring to observed
T and S.
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Method 1: GFDL Three-Step
• Step 1: Run the
AGCM with
climatological SSTs,
archiving the heat and
water fluxes.
• Step 2: Run the
OGCM with the fluxes
from step 1, while
simultaneously
restoring to observed
T and S.
February 26, 2003
Restoring terms
T
 ...   (Tobs  T )
t
S
 ...   ( S obs  S )
t
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Method 1: GFDL Three-Step
• Step 1: Run the
AGCM with
climatological SSTs,
archiving the heat
and water fluxes.
• Step 2: Run the
OGCM with the
fluxes from step 1,
while simultaneously
restoring to
observed T and S.
February 26, 2003
• Step 3: Couple the
AGCM and OGCM
without restoring,
using the archived
restoring terms from
step 2 as flux
adjustments.
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Method 2: Coupled Restore
• Step 1: Couple the AGCM and OGCM,
then run the coupled models while
simultaneously restoring to observed T
and S, archiving the restoring terms as flux
adjustments.
• Step 2: Deactivate the restoring and run
the coupled AOGCM using the flux
adjustments determined in step 1.
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Method 3: Offline Flux Differences
• Step 1: Run the AGCM with climatological
SSTs, archiving the heat and water fluxes.
• Step 2: Run the OGCM, restoring to
observed T and S. Archive the restoring
fluxes.
• Step 3: The differences between the fluxes
from step 1 and step 2 are the flux
adjustments; these are supplied to the
coupled AOGCM.
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Flux Adjustment: Pros and Cons
Cons
• Flux adjustments are nonphysical.
• There is no guarantee that coupled model
biases are invariant over different climate
states.
• Flux adjustments could distort climate
feedbacks.
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Flux Adjustment: Pros and Cons
Pros
• Flux adjustments minimize climate drift
that would distort climate feedbacks if left
unchecked.
• Flux adjustments allow sensitivity
experiments to be performed while better
models (i.e., those with smaller errors) are
under development.
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Design of Coupled Model
Experiments
• Equilibrium: The goal is to determine the climate
that is in equilibrium with a given set of climate
forcings. (Example: What climate state is in
equilibrium with twice the preindustrial level of
atmospheric CO2?)
• Transient: The goal is to investigate the timedependent response of the climate to a given
(often time-dependent) change. (Example: How
will the climate change in response to projected
increases in CO2 and other human-induced
climate forcings?)
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Types of Experiments
• Forcing-Response: Impose a specific
forcing and see how the model responds.
• Unforced Variability: Allow a model to run,
preferably for a lengthy period, and
examine the spatiotemporal variations that
are generated by the internal dynamics of
the model.
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Design of Coupled Model
Experiments: Issues
• Initialization: How to Start?
• Equilibration: How Long to Run?
• Fidelity: How Good is the Model?
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Initialization
• Not typically an important issue for
atmosphere-only or atmosphere-mixed
layer ocean models.
• More important for AOGCMs; these can
exhibit considerable sensitivity to initial
conditions.
• Issue: How to initialize time-dependent
AOGCM simulations of past climates?
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Equilibration
• Time required varies with model type and
depends on e-folding time of slowest
component of climate system.
– AGCMs: < 1 year
– A-MLO models: ~5 years
– AOGCMs: ~500-1,000 years
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Equilibration
• Integration length must be adequate for
sampling climate statistics.
• Acceleration techniques may be useful in
ocean-only simulations, but should be
used with caution.
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Fidelity
• How well does a model simulate the important
processes of interest?
• Careful comparison of model simulations with
the observed climate record are critical for
assessments of model fidelity.
• Successful performance in such comparisons
can increase our confidence in climate models.
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli
Friday’s Seminar
Date: February 28
Time: 2:00 PM
Place: ENRS Building, Room 223
Title: "The New GFDL Global Atmosphere
and Land Model AM2/LM2"
Speaker: Dr. Stephen Klein, NOAA/GFDL
February 26, 2003
16:375:544
Modeling of Climate Change
Anthony J. Broccoli