Clouds CPT/GCSS WG4 RCE Intercomparison Specification
Case author:
Chris Bretherton (breth@atmos.washington.edu)
11 June 2004
(inspired in part by K-M Xu)
Case coordinator (to whom you should send output):
Brian Mapes (Brian.Mapes@noaa.gov)
Brian will set up a web site linked to this one for case results.
Contribution deadline
We would like CPT CRM/SCM groups to contribute simulations by 31 Jul 2004 so
that we quickly learn about possible problems in the case specs.
Other GCSS WG4 participants are also welcome to participate; we ask
for results by 31 Aug 2004, though there may be some iteration or
deadline-stretching if that seems necessary. Please email to Brian
Mapes by 31 Jul that you intend to participate so we know who will be
involved.
Physical specs
...generally following Tompkins and Craig (1999, J. Climate, 12, 462-476),
except as noted.
1) No ambient rotation
2) No mean wind (this runs the risk of self-aggregation, but our
experience is that this is not an issue in such a small domain,
and a sheared wind profile has its own complications both in 3D
and 2D).
3) Radiation
Cloud-interactive shortwave and longwave radiation
Insolation: Solar constant of 685 W/m2, zenith angle of 51.7 deg
4) Initial sounding
http://www.atmos.washington.edu/~breth/CPT-public/RCE-intercomp-snd.txt
based on a prior RCE run with SST = 302 K. Surface
pressure of 1010 hPa, vertical grid as specified in C1 below.
5) Surface exchange
Use your model's default Monin-Obukhov-like scheme for computing
surface fluxes. Use a over-ocean specification of surface
roughness, z0 = 10^-4 m if your CRM doesn't have its own
Charnock-like scheme. Note Tompkins and Craig specified an
excessive (for the ocean) specified surface roughness of z0 = 1 cm,
which made air-sea exchange rather more efficient than it really
is. To get atmospheric profiles which are comparable to theirs with
the correct z0, it is necessary to use higher SSTs than they did.
6) Three cases: SST of 300, 302, 304 K
These are 2 K warmer than Tompkins and Craig, to compensate for
their excessive z0. Recommended sensitivity study: SST = 302K with
CO2 doubled from your model's default climatological value.
In the output files these should be labeled with runname = 300,
302, 304, or (2XC02) 302.2.
7) Run length
60 d. The first 30 days are for approach to equilibrium. All
average quantities are to be computed over the equilibrium period
30-60 days.
CRMs
C1) Vertical domain size/resolution
Lz approximately 27.5 km, including a sponge layer of 8.75 km thickness
You can use a vertical resolution of your choosing, but we suggest
64 or more vertical levels, with dz = O(100 m) near the surface,
asymptoting to uniform 400 m dz in the upper
troposphere and a 1 km dz in the sponge. A recommended stretching algorithm
is to calculate layer midpoints z(n) as follows (matlab format):
nz = 64
dz0 = 75
dztrop = 400
dzsponge = 1000
ztropbase = 2000
zspongebase = 20000
zspongetrans = 1500
dz(1) = dz0;
z(1) = dz0/2;
for n = 2:64
dz(n) = dz0 + (dztrop-dz0)*tanh(z(n-1)/ztropbase)...
+(dzsponge-dztrop)*0.5*(1+tanh((z(n-1)-zspongebase)/zspongetrans));
z(n) = z(n-1) + dz(n);
end
Here dz0 = 75 m is the grid spacing at the surface, which transitions to
the tropospheric grid spacing dztrop = 400 m as we move through the
level ztropbase = 2 km. As we move through the layer zspongebase =
20 km, the grid spacing again smoothly transitions to dzsponge = 1 km.
Newtonial damping timescale tau in the sponge (roughly upper 1/3 of domain):
tau = 120sec* 60^((ztop-z)/(0.3*ztop)) for 0.7*ztop < z < ztop
where ztop = z(nz).
C2) Horizontal domain size/resolution
Default: Lx = Ly = 64 km, dx = dy = 2 km (3D).
Lx = 512 km, dx = 2 km (2D)
Sensitivity studies: dx = 1 km, Lx = Ly = 128 km...Our experience
with the CSU model is that neither of these changes has much impact on
the cloud statistics in the simulation, but using a larger domain
starts creating issues/biases associated with domain-scale
self-aggregation and inhomogeneity. Clearly this raises
philosophical issues about the meaningfulness of this exercise, but
swallowing hard and proceeding still seems the best approach.
If you are using a 2D CRM, you'll have to fight self-organized
shear and self-aggregation that can considerably affect the
simulated mean state and horizontal variability (e.g. Held et
al. 1993, JAS, 50, 3909-3927) and conceivably
the SST sensitivity. To do this, use a 512 km domain and nudge on
a 1 hour timescale to the following domain-mean velocity profile:
U = min(0.001*z, 5 m/s), V = 0
The weak lower tropospheric shear in z < 5 km should break up the strong
tendency toward formation of a single stationary convective updraft
you will otherwise see, while retaining zero domain-mean surface
wind as in the 3D spec. Otherwise the output specs are identical to before.
C3) Boundary conditions
Lateral periodic
C4) Desired output (3 netcdf files, with time units of days). Note
that if some outputs are too painful to produce, we would be happy to
see what you can provide conveniently. Just put missing values in
the fields you can't provide.
(i) runname-timeseries.nc: (runname = 300, 302, 304, or 302.2)
Domain-mean hourly time series(ideally hourly-mean, but horizontal
mean from instantaneous hourly snapshot is OK too), days 0-60, of:
PCP (Precipitation, mm/day)
E (Evaporation, mm/day)
FSNT (TOA net downward shortwave radiative flux, W/m2)
FLNT (TOA net upward longwave radiative flux, W/m2)
FSNS (surface net downward shortwave radiative flux, W/m2)
FLNS (surface net upward longwave radiative flux, W/m2)
PW (Precipitable water, kg/m2)
SAV (dry static energy Cp*T+g*z, mass-weighted from 100 hPa to
surface, J/kg)
PS (Surface pressure, Pa)
TLL (Temperature at lowest grid level, K)
qLL (specific humidity at lowest grid level, kg/kg)
WSLL (wind speed at lowest grid level, m/s)
CCF (Column cloud fraction (0-1), defined as the
fraction of columns where condensate that is radiatively
active according to your radiation
scheme has a mixing ratio exceeding 0.005 g/kg in
at least one grid layer.)
These will allow simple checks of approach to steady state, mass,
moisture and energy balance, and surface flux diagnosis
(ii) runname-profiles.nc:
30-60 day mean profiles at model levels of
Z (level height, m)
P (mean pressure, Pa)
RHO (mean density, kg/m3)
T (temperature, K)
Q (water vapor mixing ratio, kg/kg)
RH (relative humidity with respect to water saturation, 0-1)
U (x-velocity, m/s)
V (y-velocity, m/s) [just to check U,V stay near zero!]
QRS (net shortwave heating rate, K/s)
QRL (net longwave heating rate, K/s)
QC (cloud water, kg/kg)
QI (cloud ice, kg/kg)
QR (rain mixing ratio, kg/kg)
QS (snow+graupel+hail mixing ratio, kg/kg)
CF (cloud fraction, defined as the fraction of gridpoint columns
where radiatively active condensate exceeds 0.005 g/kg.)
UF (the updraft area fraction (0-1) of 'updraft' gridpoints with
w > 1 m/s)
DF (the downdraft area fraction (0-1) of 'downdraft' gridpoints with
w < -1 m/s)
UMF (upward mass flux, kg/(m2 s), the domain integrated vertical
mass flux in the updrafts, normalized by dividing by the domain area)
DMF (downward mass flux, kg/(m2 s), the domain integrated vertical
mass flux in downdrafts, normalized by dividing by the
domain area)
UH (updraft mean moist static energy = cp*T +g*z + L*qv
averaged over gridpoints with w > 1 m/s)
DH (downdraft mean moist static energy = cp*T +g*z + L*qv
averaged over gridpoints with w > 1 m/s)
HFLXR (resolved-scale domain averaged moist static energy flux
rho*overline(w'h'), W/m2)
HFLXS (parameterized subgrid domain averaged moist static energy
flux, W/m2)
(iii) runname-snapshots.nc:
instantaneous horizontal snapshots, from days 30.0-60.0 (nt = 31 days),
(output dimension nt x nx x ny for each field) of:
PCP (Precipitation, mm/day)
E (Evaporation, mm/day)
FSNT (TOA net downward shortwave radiative flux, W/m2, )
FLNT (TOA net upward longwave radiative flux, W/m2)
FSNTC (TOA net downward clearsky shortwave radiative flux, W/m2, )
FLNTC (TOA net upward clearsky longwave radiative flux, W/m2)
PW (Precipitable water, kg/m2)
COND (column integrated condensate, QC+QI+QS, kg/m2)
TLL (lowest-level temperature, K)
QLL (lowest-level specific humidity, kg/kg)
WSLL (lowest-level wind speed, m/s)
These will be used to calculate probability distributions (in the
form of percentile plots) of the fields. The clear-sky fluxes may
help us intercompare radiation schemes and do some primitive
version of an ISCCP simulator analysis.
------------------------------------------------------------------------------
SCMs
S1) Vertical resolution
Use operational resolution
Optionally do sensitivity studies with more vertical levels.
S2) Desired output (2 netcdf files, with time units of days)
(i) runname-timeseries.nc: (runname = 300, 302, 304, or 302.2)
Hourly time series (ideally hourly-mean, but instantaneous hourly
snapshot is OK too), days 0-60, of:
PCP (Precipitation, mm/day)
E (Evaporation, mm/day)
FSNT (TOA net downward shortwave radiative flux, W/m2)
FLNT (TOA net upward longwave radiative flux, W/m2)
FSNS (surface net downward shortwave radiative flux, W/m2)
FLNS (surface net upward longwave radiative flux, W/m2)
PW (Precipitable water, kg/m2)
SAV (dry static energy Cp*T+g*z, mass-weighted from 100 hPa to
surface, J/kg)
PS (Surface pressure, Pa)
TLL (lowest-level temperature, K)
qLL (lowest-level specific humidity, kg/kg)
CF (column cloud fraction, 0-1, however the SCM defines it)
(ii) runname-profiles.nc
30-60 day mean vertical profiles at model levels of
Z (level height, m)
P (mean pressure, Pa)
RHO (mean density, kg/m3)
T (temperature, K)
Q (water vapor mixing ratio, kg/kg)
RH (relative humidity with respect to water saturation, 0-1)
U (x-velocity, m/s)
V (y-velocity, m/s) [just to check U,V stay near zero!]
QRS (net shortwave heating rate, K/s)
QRL (net longwave heating rate, K/s)
QC (cloud water, kg/kg)
QI (cloud ice, kg/kg)
QR (rain mixing ratio, kg/kg)
QS (snow+graupel+hail mixing ratio, kg/kg)
CF (cloud fraction, defined however the model chooses for
radiation calculations)