Physics-Dynamics Test Strategies:
Bridging the Gap with Simplified
Moist Test Cases
Christiane Jablonowski and Diana Thatcher
University of Michigan, Ann Arbor, USA
Physics-Dynamics Coupling Workshop (PDC14), Ensenada, Mexico, 12/3/2014
The Talk at its Crossroads
Effective
Resolution:
What should the
scales be that the
dynamical core
passes to the
physics
(grid-point value,
area-averaged,
sub-sampled)?
What are the
believable scales
in the dynamics?
Test Strategies:
Can we understand some
aspects of the
complex
physics-dynamics
coupling
with simplified
moist test cases?
Partly covered by
Peter Lauritzen’s talk
Topic of this talk
Advertisement
Effective Resolution Papers with Foci on Advection
Physics-Dynamics Coupling:
Session Announcements
• Test strategies: It is important to be able to identify good
coupling schemes from inferior ones early on in the
development cycle. Once the theoretical analysis of the
scheme is complete, how can further evidence be
collected to ensure the chosen scheme performs as
anticipated? The full NWP trial stage usually only offers
limited scope for (costly) change. The difficulty is to
design tests with sufficient signal and validity, without
being too complex such that they are useful in the early
development/evaluation phase.
Test Cases:
Hierarchy with Increasing Complexity
Some Desirable Design Criteria
Test cases should
• be designed for hydrostatic and non-hydrostatic dynamical
cores on the sphere,
ideally: for both shallow and deep atmosphere models
• be easy to apply: analytic initial data
suitable for all grids
formulated for different vertical coordinates
• deal with moisture in a simple way
• reveal information about the physics-dynamics coupling
• be as easy as possible, but as complex as necessary
• be cheap and easy to evaluate
• be relevant to atmospheric phenomena
• have a converged reference solution
• find broad acceptance in the modeling community
Overview of the Approaches
• Short-term deterministic assessments (15 days)
– Moist baroclinic waves with large-scale condensation
– Moist baroclinic waves coupled to the `simple-physics’
package by Reed and Jablonowski (James, 2012)
• Long-term ‘climate’ assessments (multiple years)
– Moist version of the Held-Suarez test with elements of
the `simple-physics’ package
This talk’s goal:
• Convince you that idealized physics processes lead to
reasonable atmospheric circulations.
Long-term goal (partly covered in this talk):
• Evaluate whether idealized physics processes mimic the
behavior of complex physics to aid our understanding.
Questions to Ask
• What is our motivation to pursue idealized approaches?
• Is it reasonable: How does a moist Held-Suarez (HS) aquaplanet simulation compare to a full-physics CAM5 aquaplanet simulation?
• Intercomparison: How do the different CAM5 dynamical
cores compare in moist HS and complex aqua-planet
experiments?
• Unit testing: How does the moist HS configuration compare
to aqua-planet simulations that omit some processes (like
the deep convection parameterization)?
• Can we replicate some aspects of the complex physicsdynamics interactions with the moist HS setup?
• What do we learn about the physics-dynamics coupling?
Motivation:
Results from the Aqua-Planet Experiment (APE)
• Aqua-planet model intercomparison revealed a huge spread in the
GCM circulations and precipitation characteristics
• Impossible to tell whether the APE differences are due to physics
parameterizations or the dynamical cores or both?
• Our test approaches level the playing field (identical physics).
Zonal-mean
time-mean total
precipitation rates
(hemispherically
averaged) in 16 GCMs
in aqua-planet mode,
see
Blackburn at al. (2013)
Adding Simple Large-Scale Condensation
to the Dynamical Core
Dynamics
Process
Variable
PBL mixing
Interaction
Physics
Adding Simple Large-Scale Condensation
• Add a specific humidity field q and transport it as a tracer
• Compute condensation C tendencies to force q and the
temperature T whenever the relative humidity (RH) at a
grid point exceeds a threshold (e.g. RH > 100%):
• The large-scale precipitation Pls removes the water
instantaneously without a cloud stage
Reed and Jablonowski (James, 2012)
Baroclinic Wave: Moisture and
Large-Scale Condensation
Dynamical Core Model Intercomparison Project (DCMIP) 42
Large-scale condensation in a moist version of the
Jablonowski-Williamson (2006) baroclinic wave leads to an
intensification of the baroclinic wave here at day 9
• It rains in the right spots (updraft areas associated with frontal zones)
• Provides a first glimpse at the non-linear physics-dynamics
interactions in the presence of moisture
CAM-FV 1°x1° L30, dx = 110 km
Adding a Simple-Physics Package to the
Dynamical Core
Dynamics
Physics
Process
Variable
PBL mixing
Interaction
Reed and Jablonowski (James, 2012)
Simple-Physics Package: Basic Ideas
• Replace the full-physics with a simple-physics package
• The simple-physics tendencies are
Reed and Jablonowski (James, 2012)
• The fluxes are either
– the bulk aerodynamic surface fluxes (latent and sensible
heat, friction) or
– mimic the turbulence in the boundary layer via a first-order
closure (K-theory with surface wind-speed dependent eddy
diffusivities)
• C is large-scale condensation (no re-evaporation)
Moist Interactions: Baroclinic Wave
Idealized moist baroclinic wave tests expose the behavior of
simulations with complex physical parameterizations (here CAM5)
without radiation
Surface pressure, day 9, CAM-FV 1°L30
Dry
Large-scale condensation
Simple-Physics,
no surface friction
Complex CAM5 physics
no surface friction
Simple-Physics,
with surface friction
Complex CAM5 physics
with surface friction
hPa
Tests based on Jablonowski and Williamson (2006), Simple-physics: Reed and Jablonowski (2012)
Moist Version of the Held-Suarez Test on
an Aqua-Planet (prescribed SST)
Color coding:
Held-Suarez (modified):
• Radiation: Newtonian
Temperature relaxation
H
• Rayleigh friction (PBL
momentum mixing and
surface friction)
Dynamics
Variable
Simple-Physics:
• Surface fluxes of latent
sensible heat
• PBL mixing
of moisture
Interaction
and temperature
• Large-scale
condensation
Thatcher and Jablonowski (in prep.)
Reed and Jablonowski (James, 2012)
PBL mixing
Physics
Moist Held-Suarez and Complex Aqua-Planet
CAM-SE 1° L30: Reasonable - Moist Held-Suarez mimics Aqua-Planet
Moist Held-Suarez with simple-physics
Aqua-Planet with complex CAM5 physics
Temperature
Zonal wind
Thatcher and Jablonowski, in preparation
Moist Held-Suarez and Complex Aqua-Planet
CAM-SE 1° L30: Reasonable - Moist Held-Suarez mimics Aqua-Planet
Moist Held-Suarez with simple-physics
Aqua-Planet with complex CAM5 physics
Specific humidity
Less efficient
upward moisture
transport, but distributions
are similar
Relative Humidity
Thatcher and Jablonowski, in preparation
Moist Held-Suarez and Complex Aqua-Planet
CAM-SE 1° L30: Reasonable – Eddy transports are comparable
Moist Held-Suarez with simple-physics
Eddy heat flux
Eddy kinetic
energy
Aqua-Planet with complex CAM5 physics
Moist Held-Suarez and Complex Aqua-Planet
CAM-SE 1° L30: Reasonable – Physics forcing magnitudes comparable
Moist Held-Suarez with simple-physics Aqua-Planet with complex CAM5 physics
Temperature
tendency
Deep convection peaks
higher up
Large-scale
condensation
Focus on
the tropics
Moisture
tendency
Moist Held-Suarez and Complex Aqua-Planet
CAM-SE 1° L30:
Similar tropical waves
are apparent in the total
precipitation rate
(averaged between 5S-5N)
in moist Held-Suarez (top)
and Aqua-Planet (bottom)
runs (here eastward
traveling Kelvin waves)
Same
KelvinCAM5
wave phase
Aqua-Planet with
complex
physicsspeeds
Moist Held-Suarez with simple-physics
Precipitation is less
organized in the moist HS
experiment due to simplicity
of precipitation
Thatcher and Jablonowski, in preparation
mm/day
Moist HS, Complex Aqua-Planet & Unit Testing
• CAM-SE experiments with and without simple Betts-Miller
(BM) and complex Zhang-McFarlane (ZM) deep convection
• Moist HS replicates complex Aqua-Planet (AP) behavior
With deep
No deep
AP
ZM
deep
AP
Moist
HS
BM
deep
Moist
HS
Total precipitation rate
AP
HS
Intercomparisons & Unit Testing
• Easier unit testing: How does CLUBB (new CAM PBL mixing,
shallow convection, macrophysics) interact with the SE and SLD
dycores and diffusion in CAM5 aqua-planet experiments?
SE
SE
SLD
More
diffusion
Double versus
single ITCZ
Double
ITCZ
Intercomparisons: CAM5 dynamical cores
The Community Atmosphere Model (CAM) provides four different
dynamical cores (based on the primitive equations):
1. Semi-Lagrangian (SLD): two-time-level, semi-implicit semiLagrangian spectral transform model, Gaussian grid
2. Eulerian (EUL): three-time-level, semi-implicit Eulerian
spectral transform dycore, Gaussian grid
3. Finite-Volume (FV): default dynamical core in CAM 5 & CAM
5.1, grid-point-based finite-volume discretization, explicit timestepping scheme, latitude-longitude grid
4. Spectral Element (SE): new default dynamical core (CAM 5.3),
based on continuous Galerkin spectral finite element method,
designed for fully unstructured quadrilateral meshes (cubedsphere grid), locally energy- and mass-conserving, explicit
time-stepping scheme
Intercomparisons: CAM5 dynamical cores
• The kinetic energy (KE) spectra of the moist HS
experiments replicate the KE spectra of the complex CAM5
aqua-planet runs (here with 110-150 km grid spacing)
Intercomparisons: CAM5 dynamical cores
• Moist HS experiments can partly replicate the tropical
precipitation rate characteristics of complex CAM5 aquaplanet runs
Moist HS
Increased
precip.
Increased
convergence
CAM5 Aqua-Planet,
no deep convection
CAM5 Aqua-Planet
Intercomparisons:
CAM5 dynamical
cores
• Meridional
Eddy moisture
transport: v’q’
• Indication that
the spectral
dynamical cores
EUL and SLD show
systematic tropical
differences in
comparison to
grid point models FV
and SE in both moist
HS and aqua-planet
Moist HS
SE
FV
EUL
SLD
Aqua-Planet
Conclusions
• The interactions between the dynamical core and moisture
processes can already be simulated with very simple model
configurations, like large-scale condensation, simple-physics,
or the moist HS test
• Some aspects of the complex GCM behaviors can be
replicated with the simplified physics setups
• Tests give access to an easier understanding of the physicsdynamics coupling
• Using identical physics with various dynamical cores levels
the playing field
• Approach allows unit testing of selected parameterizations
or tests of the physics-dynamics coupling technique
• Test cases hold promise to be useful for community use
References
• Reed, K. A., and C. Jablonowski (2012), Idealized tropical cyclone
simulations of intermediate complexity: a test case for AGCMs, J.
Adv. Model. Earth Syst., Vol. 4, M04001,
doi:10.1029/2011MS000099
• Jablonowski, C., and D. L. Williamson (2006), A Baroclinic
Instability Test Case for Atmospheric Model Dynamical Cores,
Quart. J. Roy. Met. Soc., Vol. 132, 2943-2975
• DCMIP shared workspace and DCMIP test case document:
https://www.earthsystemcog.org/projects/dcmip-2012/test_cases
• Thatcher, D. R. and C. Jablonowski, A moist variant of the Held-
Suarez test for atmospheric model dynamical cores: Aquaplanet
comparison and sensitivity analysis, manuscript in preparation