Unifying perspectives in interfacing, diffusion,
precipitations and cloud-coverJ.-F. Geleyn, I. Bašták-Ďurán, R. Brožková, B. Catry,
P. Marquet and F. Váňa
A1WD, Ljubljana, Slovenia, 13-15/06/2012
The issue about interfacing As proven by practice, it is almost impossible to obtain
that scientists developping ‘process parameterisations’ obey even to the simplest laws of thermodynamics:
Yet it may be important to ensure conservation of energy (and enthalpy) at the level of the full model, for adiabatic motions (i.e. before the dynamics, which is in principle conservative) starts acting. The next slides will give examples of this importance.
So the idea is to ‘enforce’ thermodynamic consistency at the time of the phys-dyn interfacing. The difficulty is of course present only for the case of non-linear quantities:
)/()())(ln(
)()(
)1(
)()()1(
2/
/0
//
TRTLTed
TCCLTL
qRqRR
qqCqqCqCqCC
vsvs
ilpvsvsv
vvtd
siirllvpvtpdp
dTCdCTdTCTCd pppp ...)(
About ‘3MT’ It is a part of the ALARO-0 development, but its
multi-scale character does not restrict its use to its prime target, i.e. the so-called ‘grey-zone’ scales (x~5km)
It can work with several levels of scientific sophistication for:– Convective Entrainement and Closure specifications,– Microphysics,– Thermodynamic adjustment.
Basically it is a way to do ‘as if’ deep convection was resolved but without needing to go to scales where this is true, thanks to:– Prognostic and diagnostic ‘memory’ of convection;– A unique micro-physical treatment beyond all sources
of condensation.
When speaking about ‘convergence’
Phys-dyn interfacing rules are a point of ‘friction’.
Because of the ‘as if’, 3MT is an interesting way of testing things like they should be in AROME, without running AROME.
One can in particular test the impact of enthalpy conservation in a system where all the water cycle explicitely undergoes the transition vapour => cloud => precipitating species!
And the result, even if roughly expected, is instructive:
Impact of replacing Q=d(Cp.T)/dt|phys by Q=Cp.dT/dt|phys (at 9km mesh with 3MT)
Red = improvement
Transparent = degradation
Similar signals for T, Hu
Neutral for wind
~0.3 m loss at 24h-48h in the
troposphere => ~half a year of
overall progress for NWP at mid-
latitudes
Small but stable and
non-neglegible
impact
Impact at high resolution
without enthalpy conservation with enthalpy conservation
Precipitation patterns are roughly the same, but the local intensity may be very different, nearly doubled at maximum
Courtesy of R. Brožková
ALARO test (with 3MT in order to make up for the difference between convection ‘permitting’ and convection ‘resolving’) on 2.3 km mesh (90s time step); 6h precipitation on 18/05/2008 (+12h to +18h)
From Bart Catry’s talk in Radostovice (2007)
From Bart Catry’s talk in Radostovice (2007)
From Bart Catry’s talk in Radostovice (2007)
For the thermodynamic equation, the trick is that, if you combine the latent heat’s local impact, the ‘barycentric’ budgets for water
phases contents and the transport of heat by precipitations, things simplify and you find a Green-Ostrogradsky complete
shape for the enthalpy conservation, with ‘linear terms’ in Lv/s0
for the phase changes impact !!
Internal unifying perspectives for diffusion
As shown in three previous talks, this construction is well advanced. We are unifying the handling of:– dry vs. moist;– SOMs only vs. ‘SOMs + TOMs’;– CCH02, QNSE, EFB (A and B systems when
appropriate);– various formulations for L.
Still missing items:– SCC in the Tompkins spirit (see TOUCANS B);– prognostic handling of L (see next slide)– surface upper-air harmonisation (see next-next slide)
The TOUCANS code structure does not yet make the best out of all this.
Prognostic length-scale Concerning EFB, Sergej Zilitinkievich (private
communication) believes that a prognostic handling of the turbulence time scale (or equivalently -and easier to do in TOUCANS- of the length-scale L) has priority before adding the TPE prognostic equation.
Like in the case of dc2 in the foreseen Tompkins-
like extension, it would be rather trivial to use the basic solver of p-TKE, any of our six or seven current ‘static length-scales’ playing the role of the stationary solution.
)],),([~
(11
LlECRLz
LK
zt
LmiE
Reconciliation of the surface and upper-air schemes
Thanks to the RMC01 filiation of p-TKE (or TOUCANS) this exists already at the level of the stability dependency functions. [SURFEX!]
For the overall intensity we have a contradiction between the role of C3 in the TOUCANS formalism and the (Louis-inherited) fact that CNm=CNh at the surface. This is currently treated (rather heuristically) via the distinction between C3 and C3free (see TOUCANS B). Still in EFB, there is the proposal to use a second von-Karman constant h=C3. in order to solve the problem with minimum algorithmic implications. We should try this soon.
Internal unifying perspectives for
convection Through its harmonisation of resolved and
convective condensation processes (and the downdraft’s handling behind this ‘combination’, alike what is observed in nature) 3MT is already a great step forward.
But the ‘intermediate algorithmic’ reflexion is missing (more for the convective part than for the resolved one).
The move towards CSU could be either an occasion, or a progress or a source of yet unknown problems. Quo vadis?
And there is still the LENTCH-related ‘skeletton in cupboard’!
About LENTCH (1/3) The idea to have a ‘historical’ or even
‘prognostic’ entrainment appeared as a nice way to try and delay our onset of convective activities.
Under the LENTCH logical switch one would try to parameterise the so-called ‘cold pool effect’, not directly, but by installing a link between past evaporation and current level of entrainment in convective ascents:– Low past evaporation would mean high entrainment
and hence delay for an organised onset;– Conversly high past evaporation would be
associated with re-triggering by gust fronts via a low entrainment rate facilitating deep ascents.
About LENTCH (2/3) Doina, Luc and myself tried a lot of solutions
around this concept but they all failed to deliver a solution with equivalent ‘static’ quality and with a noticeable delay of the onset.
Recently Radmila made another trial, abandonning the (too complex?) idea of a 3D tuning of the evaporation => entrainment link, replaced by a 2D modulation of the GCVALFA parameter of 3MT.
Indeed the latter has already some ‘LENTCH’ flavour since it modulates the link between the overal level of entrainment and the maximum reachable buoyancy.
About LENTCH (3/3) Results are mitigated:
– The control of the tuning is greatly improved;– The results in terms of precipitation are mostly
reasonable;– But the delay in onset is apparently still small.
Reference (=3.E-05)One of the ‘new LENTCH’ realisations (more contrast,
no oddity)
Internal unifying perspectives for the
thermodynamic adjustment At least here, the ‘3MT-in-ARPEGE’ exercise has
been fruitful: there is now a demand for an intermediate level handling routine coming from elsewhere than from the ALARO side!
Anyhow the distinction between ACNEBCOND and ACCDEV is going to fade out with TOUCANS. The need for two calls of at least the cloudiness part will also increase (see later).
The difficulty will be to move the R-K scheme (quite dependent on the present structure) towards the new situation (once the latter is consolidated, a single move will be hard enough).
Transversal perspective for the cloud-handling (1/2)
Since we leave the ‘local’ side for the ‘global’ one, we cannot anymore speak of unification, but rather of internal harmonisation for ALARO-1, at least in a first step.
Let us assume that we shall soon have an appropriate SCC (on half levels, beware). We would have displaced the ‘shallow convection staggering problem’ from turbulence to radiative forcing, and this would already be a huge progress!
In the latter case, we may expect that reinterpolation will not be too detrimental, but this must be tested carefully.
Transversal perspective for the cloud-handling (2/2)
Given the fact that ARPEGE radiative-cloud strategy (relying on the prognostic ql/i quantities) seems to work well in 3MT-in-ARPEGE, we would then be close to a full harmonisation, with the following steps yet to be accomplished:– Harmonising the ql/i and cloud-cover input to radiation
(via a first call to the thermodynamic adjustment);– Extending the ‘protection of condensates’ (in either
both or at least in a second call to the thermodynamic adjustment) from ‘deep’ convective only to ‘deep + shallow’;
– Verifying (and ensuring further if necessary) the consistency with the ‘reintroduced’ diffusive ql/i transport in TOUCANS.
– Preparing ourselves for a heavy tuning exercise!!!
Conclusion We have already gone a long way on the path
of algorithmic harmonisation by ‘soft methods’ (started in 2005 by an agreement of principle with HIRLAM, in Tartu).
Yet a lot remains to be done:– Convince other people of the intrinsic interest (i) of
an independent interfacing procedure ensuring conservation laws & (ii) of an algorithmic harmonisation target organised at intermediate levels (neither the full packages, too heavy, nor the detailed processes, easily modularised if the intermediate level’s work is done correctly);
– Clean seriously our own ‘ALARO house’;– Do the feasible steps in the ‘Perestroika’ spirit!