Project

General

Profile

nocturnal boundary layer problem simulations

Added by Mikhail Varentsov over 1 year ago

Dear colleges,

In my simulations with CCLM, model significantly overestimate nocturnal temperature over plain terrain around Moscow, especially for calm and clear nights (in comparison with a big set of weather stations). Correct simulation of nocturnal temperature conditions is critical for my research, so I have been trying to find a set of model parameters solving this problem, but couldn’t do it recently. When I was searching in web for solution, I’ve found that I am not alone with such problem, paper and presentations by the links below also describe the problems with nocturnal boundary layer simulations:

1) http://www.cosmo-model.org/content/model/documentation/newsLetters/newsLetter12/3-rossa.pdf

2) ftp://ftp-anon.dwd.de/pub/DWD/Forschung_und_Entwicklung/CUS2015_presentations_PDF/09_Planetary_Boundary_Layer/2_High-resolution_simulations_of_a_nocturnal_boundary_layer/talk_ines2.pdf

So, I would like to ask you, are there any known way to improve nocturnal boundary layer simulations? Thanks in advance for any advice!

P.S. I have a suggestion that the source of the problem is overestimation of turbulent diffusion coefficients...


Replies (5)

RE: nocturnal boundary layer problem simulations - Added by Hans-Juergen Panitz over 1 year ago

Dear Mikhail,

I agree with your assumption.
See also the paper
Buzzi, Matteo; Rotach, Mathias W.; Holtslag, Matthias; Holtslag, Albert A.M. (2011):
Evaluation of the COSMO-SC turbulence scheme in a shear-driven stable boundary layer
Meteorologische Zeitschrift Vol. 20 No. 3 (2011), p. 335 - 350, DOI: 10.1127/0941-2948/2011/0050

Which model version are you using? What are the values of tuning parameters tkhmin and tkmmin?
In older versions the minimum value for both parameters was 1., which was too high for nocturnal conditions (see the paper above).
In more recent versions of the model you can choose smaller values.

For more "expert" opinions, please contact Mathias Raschendorfer and/or Michael Baldauf from DWD.

Best regards

Hans-Juergen

RE: nocturnal boundary layer problem simulations - Added by Andreas Will over 1 year ago

You report a hard problem in COSMO also related to soil physics, dynamics in the boundary layer and convection parameterization. (Are you using deep convection parameterization?)

The relevance of the land surface scheme has been analysed by Jan-Peter Schulz:
1. Missing dependency of the heat conductivity on the soil water content.
The option dependency on soil water content (after Johansen, 1975) introduced in cosmo_1.14 increases the daily temperature range by 2K over northern Africa.
2. Missing vegetation layer with own temperature
TERRA has no vegetation layer above ground. This is a further reason for too high night temperatures.

Hereto a publication is submitted. Please contact Jan-Peter Schulz (DWD) for more details.

The Land Surface Schemes Veg3D (KIT Karlsruhe, Breil) and CLM (ETH Zürich, Davin) can be used alternatively.

Greetings, Andreas Will

RE: nocturnal boundary layer problem simulations - Added by Mikhail Varentsov over 1 year ago

Hans-Juergen Panitz wrote:

Which model version are you using? What are the values of tuning parameters tkhmin and tkmmin?
In older versions the minimum value for both parameters was 1., which was too high for nocturnal conditions (see the paper above).
In more recent versions of the model you can choose smaller values.
For more "expert" opinions, please contact Mathias Raschendorfer and/or Michael Baldauf from DWD.

Dear Hans-Juergen,

I’m using version 5.0_clm6, with itype_turb = 3, and I’ve tried values of tkhmin and tkmmin both set to 0.4 or 0.1,and haven’t seen significant difference between (in nocturnal temperatures modelling).
Now I will try to apply tuning parameters, obtained from the article you’ve advised me - thank you for this link!

And do you (or anybody here) know, is this problem common for other possibilities of itype_turb?

RE: nocturnal boundary layer problem simulations - Added by Mikhail Varentsov over 1 year ago

Andreas Will wrote:

You report a hard problem in COSMO also related to soil physics, dynamics in the boundary layer and convection parameterization. (Are you using deep convection parameterization?)

The relevance of the land surface scheme has been analysed by Jan-Peter Schulz:
1. Missing dependency of the heat conductivity on the soil water content.
The option dependency on soil water content (after Johansen, 1975) introduced in cosmo_1.14 increases the daily temperature range by 2K over northern Africa.

The Land Surface Schemes Veg3D (KIT Karlsruhe, Breil) and CLM (ETH Zürich, Davin) can be used alternatively.

Dear Andreas,

Thank you for your suggestions. I’m using model with turned on dependency of the heat conductivity on the soil water content (itype_heatcond = 2, do you mean this?) and itype_conv = 0.

And in case we are talking about other surface schemes, are there something like starter package or description about how to run CCLM coupled with them? I’ve read about them, but has no idea how to start with them.

RE: nocturnal boundary layer problem simulations - Added by Mikhail Varentsov about 1 year ago

Dear colleges, hello again!

I’ve successfully tested model with some of your previous suggestions during last months. Idea to use lower values of tkhmin, ktmmin and pat_len was really helpful. Now I continue my investigation with high-resolution domain (1 km), and I faced problem that model becomes numerically unstable and fails with nan-values (If I use reduced values of tkhmin, ktmmin and pat_len). Anso I made some tests with newer turbulent schemes (itype_turb = 5/7). In these tests results were better (in terms of the errors for nocturnal temperatures) than with itype_turb=3 and standard tuning parameters, but worse than with itype_turb=3 and tunned parameters (tkhmin, ktmmin and pat_len). And also experiments with itype_turb = 5/7 seems to have higher numerical stability that with itype_turb=3. In addition, this new schemes seems to be completely insensitive to tuning parameters that I used before.

No I have two main questions:
1) what model options I can use to avoid problems with numerical stability in case of itype_turb = 3 and corrected tuning parameters (dx = 1km, pat_len=50, tkhmin=tkmmin=0.05, dt = 5/10 sec).
2) are there any tuning parameters (ore some constants in the code), which can reduce mixing rates under stable conditions in case of usage itype_turb = 5/7?

Thanks in advance for your advices!

    (1-5/5)