meet trouble with GLS_MIXING

[FengZhou]

Hi, ROMSers,

Normally I used k-e (or k-omega) as my default GLS_MIXING scheme while applied ROMS(3.0) to the Bohai Sea (a small north part of the East China Sea). It performs better than others like k-kl, gen, and KPP. So I did the same choice while applied the model(3.1) to the whole East China Sea including the Changjiang river (Yangtze River). But it blows up after 2-3 days simulations, by contrast, there is no problem with KPP and MY25 for more than 1 years at least.

Does the k-e scheme have more strict limitations of the stability?

platform:
Linux node65 2.6.16.60-0.27-smp #1 SMP Fri Nov 28 21:47:49 CST 2008 x86_64 x86_64 x86_64 GNU/Linux
vendor_id : AuthenticAMD
cpu name : Quad-Core AMD Opteron(tm) Processor 8350

model version:
svn: $LastChangedDate: 2008-12-24 11:42:42 +0800 (Wed, 24 Dec 2008) $
svn: $Id: Version 291 2008-12-24 03:42:42Z arango $

compiler:
PGI

Thank you in advance!

zhou

[jcwarner]

there really should not be substantial differences in GLS between roms 3.0 and 3.1. One thing you could do is look at the gls_corstep.f90 files that are generated during the Build, and compare them between the two versions.

[jacopo]

I experienced too that k-e is less stable than others schemes. In several realistic applications I did, the model runs OK if using MY25, k-w, gen but it blows up with k-e using the very same time step. With a reduced time-step, k-e works fine. This problem occurs mostly during spin up periods.

[FengZhou]

Hi, John,

I don't know MPI well, so I just check others.

I don't think there are changes between these two versions by comparing Build/gls_corstep.f90.


zhou

[ggerbi]

I found the same thing as jacopo. Reducing the time step made it run fine.
Have you tested it with all three stability functions (Kantha-Clayson, Canuto A and Canuto B?)

[FengZhou]

Hi,

I found the same thing as jacopo. Reducing the time step made it run fine.
Have you tested it with all three stability functions (Kantha-Clayson, Canuto A and Canuto B?)

As you said, Canuto A works.

Thank you!