[CIG-MC] Problems with tracer and composition output in CitcomS

[Bao, Xiyuan]

Hi Professor Zhong,
Thank you for your help! Actually I did not switch the sequence at first, instead I defined the tracer flavor in 3.0.3 using the loop control:

if(E->monitor.solution_cycles<=1){
                        E->trace.extraq[j][0][kk]=0.0;
                          if(E->T[j][node+1]>0.7&&E->T[j][node+1]<0.9&&rad>radlow&&rad<radhigh){
                              E->trace.extraq[j][0][kk]=1.0;
                    }

and plotted the output at step 1 (the attachment in the last email are for step 0), so that the main code remained the same . Though the model changed a little bit through step 0 , tracer,comp_nd and comp_el should be consistent , however the output are almost the same as before ,i.e. right tracer , same comp_el&comp_nd missing Africa(the mesh is 257x257x129,notice here comp_el is consistent with comp_nd,but for the downsampled model 65x65x33 comp_el is disaster). But I am going to try to read in T without touching the sequence anyway.

On the other hand, I find the fill_composition function is called right after count_tracers_of_flavors, and comp_el is computed with tracer flavor, I guess something wrong might happen at this step.

Best,
Bao,Xiyuan ,
School of Earth and Space Science,
University of Science and Technology of China.
________________________________
From: Shijie Zhong [shijie.zhong at colorado.edu]
Sent: Wednesday, August 09, 2017 12:02 AM
To: cig-mc at geodynamics.org; Bao, Xiyuan
Cc: Liu, Lijun
Subject: Re: [CIG-MC] Problems with tracer and composition output in CitcomS


It looks like that the compositional field defined for tracers did not get projected to elements and nodes properly. I suspect that the trouble may have been caused by your switching the sequence of those two functions or subroutines. Can you not switch the sequence, at least as a test, to see whether tracer’s C, nodal C and elemental C are consistent? I suspect that everything would be ok, if you do not switch the sequence. In this case, you can think about defining T earlier somehow without switching the sequence.

As in any code, switching the sequence of subroutines can be tricky, and may mess up the logic and flow.

Also, about CitcomS at CIG, I would recommend version 3.1.1 if you would like to run simple benchmarks.

Good luck.



Shijie Zhong, Professor
Department of Physics
University of Colorado at Boulder
Boulder, CO 80309
http://mathis.colorado.edu/szhong
Tel: 303-735-5095; Fax: 303-492-7935


From: CIG-MC <cig-mc-bounces at geodynamics.org> on behalf of "Bao, Xiyuan" <xiyuan at illinois.edu>
Reply-To: "cig-mc at geodynamics.org" <cig-mc at geodynamics.org>
Date: Tuesday, August 8, 2017 at 4:22 PM
To: "cig-mc at geodynamics.org" <cig-mc at geodynamics.org>
Cc: "Liu, Lijun" <ljliu at illinois.edu>
Subject: [CIG-MC] Problems with tracer and composition output in CitcomS

Hi there,

I'm currently researching LLSVPs with a global model using CitcomS but I met a weird output.

First , I tried CItcomS 3.0.3, I have attached my Tracer_setup.c. In the lab65x33.cfg file , you can find my initial settings , among them the age part is useless(We changed some related contents in the code, but I shut down them in the cfg file.

1.To pick out LLSVPs, I use temperature(T) and radius(R) as a filter of tracer flavor, in the function count_tracers_of_flavors in lib/Tracer_setup.c:


                  if(E->monitor.solution_cycles==0){
                        E->trace.extraq[j][0][kk]=0.0;
                          if(E->T[j][node+1]>0.7&&E->T[j][node+1]<0.9&&rad>radlow&&rad<radhigh){
                              E->trace.extraq[j][0][kk]=1.0;
                    }
You see, I set 2 flavors, 0 for ambient, 1 for LLSVPs.

2.And of course to activate the filter I need to read in the T before set tracer flavor , so I change the sequence of 2 functions:
For different configure way,

1)on the server Blue Waters, configure file using :
./configure MPICC=cc CC=cc MPICXX=CC CXX=CC LDFLAGS="-dynamic",

move (E->problem_initial_fields)(E) in front of initialize_tracers(E) in initial_conditions in the code /lib/Instructions.c ;


2)on the server Keeling, configure file using :
 CFLAGS="-g -O2 -I$PWD/python/pythia-0.8.1.15-py2.6.egg/" ./configure,

move self.initTemperature() in front of   self.initTracer() in the function launch(self) in the code CitcomS/Components/IC.py.


I use buoy_type = 1 , so comp_el is just a ratio between number of tracer 1 and number of all tracers. The temperature input is in Temp.pdf , all those images below are at CMB.


Then strange thing happened, see output map of tracer flavor in the image Tracer.pdf ,  the African LLSVP at the center is complete, so is Pacific one , but look at comp_nd output in Comp_nd.pdf, African part misses a lot , comp_el in Comp_el.pdf is even a mess , I really cannot figure out why. Comp_nd should be consistent with Tracer flavor, but somehow I got such strange output , I even tried the latest version 3.3.1 but met the same problem, except for one thing, comp_el here is basically the same with comp_nd(see Comp_el_new.pdf , initial settings are in the file newlab). BTW, the way I plot comp_el is to change its name to comp_nd and use autocombine.py. Could you please offer me some clue to solve this? Thank you very much!








Bao,Xiyuan ,
School of Earth and Space Science,
University of Science and Technology of China.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.geodynamics.org/pipermail/cig-mc/attachments/20170809/d3e00980/attachment-0001.html>