Welcome, Guest
Username: Password: Remember me
HARMONIE convection study

TOPIC: Kickoff post for convection studies HARMONIE40h1.2

Kickoff post for convection studies HARMONIE40h1.2 1 year 9 months ago #2041

  • Sander Tijm
  • Sander Tijm's Avatar
  • OFFLINE
  • Administrator
  • Posts: 24
  • Thank you received: 2
Dear all,

this is the first post in the HARMONIE convection studies subject on the HIRLAM/HARMONIE forum.

To summarize everything that has happened before:
- several institutes complain about the convection in HARMONIE 40h1.2
- Sensitivity studies have shown that some of the problems may come from the surface data assimilation, some from the microphysics and some from the turbulence/shallow convection

Results from the sensitivity studies can be found through:
hirlam.org/trac/wiki/Harmonie_40h1/ConvectionCases

And on this webpage there is also a document that describes the possible strategy to improve the convection behaviour of HARMONIE 40h1.2:
hirlam.org/trac/raw-attachment/wiki/Harm...ategy_convection.pdf

If you have any ideas, comments or additional results please share them with us through this forum topic and upload your results to the webpage shown above!

Kind regards,

Sander
The administrator has disabled public write access.
The following user(s) said Thank You: Eirik Mikal Samuelsen

Kickoff post for convection studies HARMONIE40h1.2 1 year 9 months ago #2043

Hi all,

What can we learn from cold start experiments? Okay, if the cold start e.g. is more wet than a warm start and this gives better performance it may indicate that the physics (and/or assimilation) in the model leads to too dry conditions. Hmhm, wouldn't it be better to perturb a warm start for the specific variable in question, e.g. WG2, instead of having the "uncertainty" of all variables in a cold start...?

Okay, so is it so that a wet soil (correctly wet) in general gives too much evapotranspiration in ISBA? If yes, is this mostly/only connected to vegetated areas or does it also relate to bare soil areas? For crops in SURFEX (i.e. C3 and C4 crops) the vegetation fraction, VEG, is given by the equation VEG=(1.0-exp(-0.6*LAI)) which gives VEG=0.5 for LAI=1, VEG=0.85 for LAI=3. Do we get “correct” amount of bare soil over the Netherlands during spring time? Hypothetically, if the bare soil fraction would be overestimated it can lead to excess evaporation…

Okay, but let's assume (or maybe it is really confirmed) that it is the VEG part of the grid box that contributes to excess transpiration. If we assume that the soil at some spring situation is “correctly” wet (in terms of soil-water index, SWI). What can then lead to excess transpiration (repeating ideas)?
Too high LAI as provided by ECOCLIMAP (ecoclimap parameter files).
Too low surface resistance. There can be different aspects behind this:
too low minimum stomatal resistance, RSMIN (specified in ini_data_param.F90), now 40 for C3 and 120 for C4.
Soil moisture access as function of SWI (RACMO idea)
wrong clay/sand leading to wrong wilting point and field capacity (has HSWD data base been tested instead of FAO?) )
too thick root zone (i.e. too deep root depth in ecoclimap parameter files)

Regarding soil moisture increments: Yes, tuning the size of the increments is a possibility...
This is similar to what MetCoOp has done for TG2 soil temperature where the increment factor is changed from 1/(2*pi) to ½, thus increasing the impact on TG2.

In my view it is a tricky balance where to invest development time between cy40, “old surface solution” and cy43 “new surface solution”. In general I would like to survive as much as possible with limited investment in current cy40 surface (SURFEX7.3+OI). In parallel I hope work on EKF scheme will be fruitful but as always it is difficult to say when it is healthy enough to become operational and be useful to tackle possible problems with the OI scheme.

Cheers,

Patrick
The administrator has disabled public write access.

Kickoff post for convection studies HARMONIE40h1.2 1 year 9 months ago #2045

  • Sander Tijm
  • Sander Tijm's Avatar
  • OFFLINE
  • Administrator
  • Posts: 24
  • Thank you received: 2
Dear Patrick,

thanks for your elaborate reply. I think what has to be done first is to check what the exact differences are between the runs that give convection in the MetCoOp case (cold start and warm start without surface DA) and the runs that lack the convection (the oper and warm start run with surface DA).

If you know the cause (is it a too dry surface giving a too dry PBL or is it a too moist surface and too much evaporation leading to a too cold PBL), then you can try to make adjustments to prevent the unwanted effect of the surface DA.

Things that can be checked are:
- surface T2m and Td2m in the area where the convection should start (so well before the convection starts in the model) to check whether the cold start and no surface DA runs are more moist or dryer
- surface fluxes in the same areas
- boundary layer profiles before the convection starts
- soil moisture behaviour in subsequent runs to check the impact of the surface DA. If you look at the horizontal distribution of the soil moisture increments you may be able to couple this to the areas where you see a big difference in fluxes and surface T2m and Td2m.

First this has to be done and after that is finished you can think of ways to improve the behaviour of HARMONIE-AROME 40.

I think that an effort to improve HARMONIE-AROME 40 is necessary, as we are already seeing erroneous surface behaviour for as long as we are running HARMONIE-AROME. In the Netherlands it causes a detrimental effect on the quality of HARMONIE-AROME output when we use it as the basis for our KNMI-app.

Also I do not know how long it will take before we can tag an official CY43 release, but seeing the huge changes this version will absorb I think that it may take us another year or more before we are close to release? An improvement on this behaviour is already long overdue, so we should do something about it ASAP.

So please do some checking first and report back on the findings, then we can discuss the way forward.

Kind regards from a fresh De Bilt,

Sander
The administrator has disabled public write access.

Kickoff post for convection studies HARMONIE40h1.2 1 year 9 months ago #2046

Hi,

We at MET Norway have two ideas for testing in precipitation cases:
  1. the impact of lateral boundary condition for moisture transport into the domain
  2. the sensitivity of cloud microphysics parameterizations

For 1) I have run a short test experiment with ARPEGE and ECMWF boundary files for a extreme precipitation case on Svalbard. This is not a convection case but we hope we can learn about the sensitivity to moisture transport in general from this case. The preliminary results suggest that experiments with ARPEGE boundaries are drier near the boundaries but the precipitation amounts further in the domain are not much affected. I will continue this work and hopefully can draw some conclusions. At the moment, it looks like there is not very high sensitivity to the boundary data used.

For 2), the plan is to run experiments with modifications in the microphysics scheme, made by Bjørg Jenny Engdal. That includes modifications in ice nucleation, autoconversion and collection efficiencies.

Regards,
Teresa
The administrator has disabled public write access.
The following user(s) said Thank You: Eirik Mikal Samuelsen

Kickoff post for convection studies HARMONIE40h1.2 1 year 9 months ago #2047

  • Ulf Andrae
  • Ulf Andrae's Avatar
  • OFFLINE
  • Administrator
  • Posts: 283
  • Thank you received: 30
Sander,

I agree that we should keep investigating the impact of soil and atmospheric processes on the convection misses we see. However I think that we cannot argue that a CY43 version is so far away that we have to work in CY40. In my opinion it's a matter of priorities and where you put your efforts. If we continue to focus on CY40 it will for sure take a long time before CY43 becomes ready for operations. Thus, concerning the surface processes and assimilation I'm more interested to investigate it further in CY43 than in CY40 so that's how I think we should prioritize.

Regards,
Ulf
The administrator has disabled public write access.

Kickoff post for convection studies HARMONIE40h1.2 1 year 2 weeks ago #2248

For convection in relationship with CAO, maybe this article from UK Met-office might be of some relevance also for the Arome model?
doi.org/10.1002/qj.2116

At MET Norway we used the UM 4 km model before we introduced Arome 2.5 km, and this model had similar problem with non-resolved convection as we see in the Arome model in Northern Norway. The typical problem is actually related to quite deep convection on a small horizontal scale. Maybe there should be some parameterization of deep convection as well?
The administrator has disabled public write access.
Time to create page: 0.091 seconds