Lagrangean labels or dye with no mixing

[rduran]

Hi,

I would like to do some experiments with lagrangean labels, i.e. 3 dyes defined as follow:

X(x,y,z,t=0)=x
Y(x,y,z,t=0)=y
Z(x,y,z,t=0)=z

where x is longitude, y is latitude, z is depth, t is for time of course.

I would like to see where these labels are advected to, but I would not want them to get mixed. Is there a way to turn the mixing off for the passive tracers?

Thanks,

Rodrigo.

[kate]

You can turn off horizontal diffusion by setting tnu2/tnu4 to zero for those tracers. The vertical mixing is by default set the same as that for salinity. Turning that off is not so simple.

[arango]

The salinity values are used when you actiavate any of the vertical mixing parameterizations available. Otherwise, a contant vertical mixing coefficient of zero or very small for inert tracers can be specified in ROMS standard input file.

[rduran]

Thanks for the replies, I have a couple of doubts,

As for horizontal mixing, Im not sure how I can set TNU2 to zero only for the passive tracers but not salinity or temperature. I saw the tutorial for ocean.in but don't understand [1:NAT+NPT,Ngrids]

Is the first value for NAT and the second for NPT? i.e. I should have:

TNU2 == 20.0d0 0.0d0


Similarly, how can I set the vertical mixing to zero only for the passive tracers and not the active tracers. The only coefficient I see for vertical mixing in the .in file is AKT_BAK. Did I understand correctly that I either turn off vertical mixing for both active and passive tracers or I choose a common value for both but cannot have one on and the other off?

It seems that this is the case and that it would require a "not so simple" fix. Has anybody gone through this before?


Thanks again,

Rodrigo.

[kate]

[1:NAT+NPT,Ngrids]

This means you should have NAT+NPT values, I think five in your case. The first two are for temperature and salinity, the last three for your tracers. You can do the same for AKT_BAK if that is the vertical mixing being used. However, we use LMD_MIXING and many use GLS_MIXING or similar. In that case, there is a full 4-D array for Akt, with only two tracer levels (temp and salt). I would search on Akt and do something similar to what's done for LMD_NONLOCAL in pre_step3d.F.

[rduran]

Thanks for you reply.

I am using MY25_MIXING I changed AKT values as suggested. My .out file now gives me


0.0000E+00 Akt_bak(03)
.
.
.
0.0000E+00 Akt_bak(05)

Which are the values for dye_01 through dye_03. Am I good to go?

If I change into another closure (LMD or GLS) scheme, then I would have to worry about a 4D array for my dyes, is this correct?


Thanks,

Rodrigo.

[rmueller]

Hi Rodrigo,

I am not familiar with the use of passive tracers as a dense mass of Lagrangian floats, but my experience with using passive tracers would advise against it. I could be wrong, but you may want to consider floats instead.

In my application, an instability in the passive tracer caused the model to crash by outputting garbage that violated IO constraints. This garbage was an instability in the passive tracer only, resulting in a checkerboard mosh-pit of disaster that was not apparent in the active tracer realm. This discrepancy between the active and passive tracers was introduced because I had forgotten to include the passive tracer in my sponge layer. I'm still not sure why this would make a difference, given that my boundary file has tracer=0 and the model had tracer=0 along boundary, but...it did!

In any case, the sensitivity of my model run to diffusion values for the passive tracer suggests to me that you may run into trouble when setting vertical or horizontal diffusion to values (e.g. 0) that may violate numerical conditions for stability. If I were you, I might use floats as well just to test that both are behaving the way you want them to.

That's my 2 cents...hopefully more in the realm of quarters than pennies.

Cheers,
Rachael

[kate]

Rachael is right in that a small amount of diffusion on them is probably not going to kill you, nor be your biggest problem. You will be putting the tracers through the ROMS advection equation. In reading the CICE manual, they advect things like age tracers differently from things like ice volume, because the conservation properties should be different. ROMS advects all tracers alike. Initialize a passive tracer with a uniform value to see what I'm talking about.

[rduran]

Rachael,

That what indeed useful information thanks, and it came just hours after I started wondering why my model had crashed too. By day 148 output garbage from dye_03 (depth label) was too much for the writing routine to handle. Similar to what happened to you.

Im curious to learn more about instability for the passive tracers, I know numerical stability gets complicated with PDE's but Im still curious to know what made the difference between the passive and the active tracers, specially since (if I understood correctly) "ROMS advects all tracers alike". If anybody can point to the right direction I would appreciate it.

There is some papers with the same technique I described above, it seems to me that the key is to keep integration times short enough.

From what I see in my output files, short integrations are useful and open boundaries don't seem to be as open as I wish they were. I need to look into my model options, any boundary suggestions are more than welcome.


As for the dye values, after 35 days or so I start seeing unacceptable something.

I need to do Kate's suggested experiment, play around with diffusion coefficients, set some floats and read some more. My advisor had already suggested starting the experiment latter in the year, but Im not sure 35 days of useful information is good enough, perhaps I can repeat a sequence of 35-day runs. Need to look into that too.

Thank you both for your answers,

Rodrigo.