tests of advection schemes

Message

Barbara · #1 Unread post by **Barbara** » Fri Sep 28, 2007 3:34 pm

Hi All,

I have tested some of the advection schemes by running an idealized upwelling case driven by wind forcing.
The North is land, and South is Wall, East-West are periodic BCs,
The wind is along the channel x-direction,
sloping bottom along channel y-direction.

However, the results show variations of variables (elevation, temperature, etc.) along x-direction.

The best results (almost invariant along-channel) were achieved by defining
UV_U3HADVECTION
TS_U3HADVECTION
TS_C4VADVECTION

When defining 4th order centered for both momentum and tracer or 4th order Akima for the tracer, the results are very noisy, with abnormal high temperature points appearing in lower temperature waters.

Figures and output log files of some of the results could be seen at
http://landsea.ust.hk/~zutt/test/
u3 means 3rd-order upstream horiz. Advection
c4 means 4th-order centered
A4 means 4th-order Akima

Should the advection schemes have so much influence on the flow structures? Could you give me some suggestions, and share your experience on choosing advection schemes under different situations.

Thanks a lot!

jcwarner · #2 Unread post by **jcwarner** » Fri Sep 28, 2007 5:27 pm

did you activate any of :
1)
#define UV_VIS2
#define UV_VIS4
#define MIX_GEO_UV
#define MIX_S_UV

2)
#define TS_DIF2
#define TS_DIF4
#define MIX_GEO_TS
#define MIX_S_TS

Barbara · #3 Unread post by **Barbara** » Sat Sep 29, 2007 3:20 am

Yes, I have also activated

#define UV_VIS4
#define MIX_GEO_UV

and

#define TS_DIF4
#define MIX_GEO_TS

And I have done some tests by increasing TNU4 and VISC4, there is little change.

jcwarner · #4 Unread post by **jcwarner** » Sat Sep 29, 2007 1:04 pm

there was a recent post by eveny logvinov and he was having some similar troubles. I think there may be an issue with mix_geo_uv. I told himm to use mix_s_uv and his troubles went away.
So first try your tests without any horiz mix of momentum and without any horiz mix of tracer. then slowly add those options back in.
I will take a closer look at mix_geo_uv.
Thanks for looking at these!

Barbara · #5 Unread post by **Barbara** » Thu Oct 04, 2007 4:42 am

Thanks for your suggestion!

I have tried a few tests by changing the advection schemes and horiz mix related options by using the ROMS default upwelling case.

If I define:
MIX_GEO_UV, MIX_GEO_TS, UV_VIS4, TS_DIF4,
And setting the visc4 = 0 and tnu4 = 0
The model blows up after about 90 steps.

If I define:
MIX_GEO_UV, MIX_GEO_TS, UV_VIS4, TS_DIF4,
And setting the visc4 > 0 and tnu4 > 0
The model blew up after a few steps.

If I only activated TS_DIF4 and MIX_GEO_TS, without horiz. mix of Momentum, the model could run properly.

Once I activated the UV_VIS4 and MIX_GEO_UV, no matter whether I defined the horiz mix of Tracer or not, model blew up.

If I did not activate the UV_VIS4 or TS_DIF4, the model could run properly. But the result has an alternation of high and low temperature stripes when I define (4th order centered advection schemes for both UV and TS): http://landsea.ust.hk/~zutt/test/temp16 ... MIXgeo.jpg

When I activated MIX_S_UV and MIX_S_TS with the same advection scheme used in the former case, the temperature result still has an alternation of high and low stripes:
http://landsea.ust.hk/~zutt/test/temp16 ... c_MIXs.jpg

The alternation of high and low stripes seems not physically reasonable!

If I used 3rd upstream for UV and 3rd upstream for TS horiz., and 4th centered for TS vertical, there would be no stripes. So, there might also be problems in the advection schemes.