Single Point Simulations

[swieneke@meteo_uni-koeln_de]

Hi all,

I hope someone can help me with the following Issue, because right now I´m really frustrated.

My goal is to run the CESM 1.0.4 with my own datasets. After several unsuccessful attempts I decided to run a supported single-point dataset first. Namely the 1x1_brazil dataset. But unfortunately this isn't working either.

My machine configuration:

<machine MACH="tr32_pgi"
         DESC="TR32 Linux Cluster (pgi), 8 pes/node, batch system is PBS"
         EXEROOT="$HOME/cesm1_0_4/$CASE"
         OBJROOT="$EXEROOT"
         INCROOT="$EXEROOT/lib/include"
         DIN_LOC_ROOT_CSMDATA="$HOME/cesm1_0_4/inputdata"
         DIN_LOC_ROOT_CLMQIAN="$HOME/cesm1_0_4/inputdata/atm_forcing.datm7.Qian.T62.c080727"
         DOUT_S_ROOT="$HOME/cesm1_0_4/output/$CASE/"
         DOUT_L_HTAR="FALSE"
         DOUT_L_MSROOT="UNSET"
         CCSM_BASELINE="$HOME/cesm1_0_4/ccsm_baselines"
         CCSM_CPRNC="$HOME/cesm1_0_4/cprnc_64/cprnc"
         OS="Linux.pgi"
         BATCHQUERY="/usr/local/bin/qstat"
         BATCHSUBMIT="/usr/local/bin/qsub"
         GMAKE_J="1"
         MAX_TASKS_PER_NODE="1"
         MPISERIAL_SUPPORT="FALSE"
 />

./create_newcase -case brazil_test -res pt1_pt1 -compset I -mach tr32_pgi
cd brazil_test
./xmlchange -file env_conf.xml -id CLM_PT1_NAME -val 1x1_brazil
./configure -case
./brazil_test.tr32_pgi.build
./brazil_test.tr32_pgi.run


Thu Aug 23 13:23:28 CEST 2012 -- CSM EXECUTION BEGINS HERE
Thu Aug 23 13:23:30 CEST 2012 -- CSM EXECUTION HAS FINISHED

the run quit after a few seconds. The error occur when the cpl tries to initialize the atm component. The error message of the atm log is:

(datm_comp_init) atm_in = datm_atm_in
(datm_comp_init) decomp = 1d
(datm_comp_init) iradsw = 1
(datm_comp_init) factorFn = unused
(datm_comp_init) restfilm = undefined
(datm_comp_init) restfils = undefined
(datm_comp_init) presaero = T
(shr_strdata_readnml) reading input namelist file: datm_atm_in
(shr_stream_init) Reading file clm_qian.T62.stream.Solar.txt
(shr_file_freeUnit) unit 97 was not in use
(shr_stream_init) Reading file clm_qian.T62.stream.Precip.txt
(shr_file_freeUnit) unit 97 was not in use
(shr_stream_init) Reading file clm_qian.T62.stream.TPQW.txt
(shr_file_freeUnit) unit 97 was not in use
(shr_stream_init) Reading file presaero.stream.txt
(shr_file_freeUnit) unit 97 was not in use
(datm_comp_init) atm mode = CLMNCEP
(shr_strdata_init) calling shr_dmodel_mapSet for fill
(shr_strdata_init) calling shr_dmodel_mapSet for remap
(shr_strdata_init) calling shr_dmodel_mapSet for fill
(shr_strdata_init) calling shr_dmodel_mapSet for remap
(shr_strdata_init) calling shr_dmodel_mapSet for fill
(shr_strdata_init) calling shr_dmodel_mapSet for remap
(shr_strdata_init) calling shr_dmodel_mapSet for fill
(shr_strdata_init) calling shr_dmodel_mapSet for remap
('shr_map_getWts') ERROR: yd outside bounds -7.000000000000000
(shr_sys_abort) ERROR: ('shr_map_getWts') ERROR yd outside 90 degree bounds
(shr_sys_abort) WARNING: calling shr_mpi_abort() and stopping


I know that the longitudes are not allowed to be negative, but the latitudes too? And I worked with the supported data, so I don't think that the negative latitude is the problem. But what could it be then?

 

[swieneke@meteo_uni-koeln_de]

Ok, obviously its not a latitude longitude problem, because I run Simulations for Vancouver and had the same problem. I reckon that the cluster on which I run the Case use more than one processor for this task, and that MPI is crashing because of that. I will check this.
But I have another strange problems:

1. When I changed the env_conf.xml after creating the CASE to CLM_USRDAT_NAME -val 1x1_brazil the Model reported the error:

ProgName:: setting clm_usr_name -- but did NOT find any user datasets: 1x1_brazil
ERROR: generate_resolved.csh error for lnd template
configure error: configure generated error in attempting to created resolved scripts

I searched the Forum and found a post where it is said that you have to delete the numbers in the netcdf input files:
e.g.
from: domain.lnd.1x1pt-brazil_navy.090715.nc
to: domain.lnd.1x1pt-brazil_navy.nc

then it is working. But I wondering why I have to change the name of supported single point datasets.

2. In the UsersGuide CLM4 in CESM 1.0.4. Example 5-1.

Example 5-1. Example of running CLM over a single-point test site in Brazil with the default Qian
atmosphere data forcing.
> cd scripts
> ./create_newcase -case testSPDATASET -res pt1_pt1 -compset I 
-mach bluefire
> cd testSPDATASET
# Configure to run for the test site
> set SITE=1x1_brazil
> ./xmlchange -file env_conf.xml -id CLM_CONFIG_OPTS -val "-sitespf_pt $SITE"
> ./xmlchange -file env_conf.xml -id CLM_PT1_NAME -val $SITE

The model isn't finding the option -sitespf_pt in the config_definitions. I added the Option based on the Information giving in the UsersGuide Chapter3 Adding Resolution Names:

<entry id="sitespf_pt"
valid_values="none,1x1_brazil,1x1_tropicAtl,5x5_amazon,
1x1_camdenNJ,1x1_vancouverCAN,1x1_mexicocityMEX,1x1_asphaltjungleNJ,
1x1_urbanc_alpha,1x1_numaIA,1x1_smallvilleIA"
value="none" category="physics">
Flag to turn on site specific special configuration flags for supported single
point resolutions.
Currently the only special settings are for MEXICOCITY and VANCOUVER, which make
changes to urban parameters.
</entry>

But this does not work either.

 

[slevis]

Dear Sebastian,

I hope somebody else can help with (2) because I don't know the answer.

Regarding (1), I understand your frustration, and I apologize on behalf of all of us. Please realize that the CESM is a non-profit academic community project available to anyone who wishes to use it or contribute to it. We have staff at NCAR to offer limited support for it. While such staff has diminished with budgetary constraints, we support a continually growing user community. We always have more work than we can accomplish. Mistakes like this slip through the cracks unfortunately.

Thank you for your understanding.

Sincerely,
Sam Levis

 

[swieneke@meteo_uni-koeln_de]

Hi Sam,

Iam highly appreciate your work, I was only wondering if the renaming of the provided Datasets will not lead to other errors.

Back to the first error. I ran the model on one Node and CPU but its still crashing. Its not a problem to run the model with global datasets, its only crashing with single point data. Anyone has a tip? Or an idea for a workaround?


edit: I searched the Forum and found this post by Koichi, he had the same Problem as I have. He solved the problem by updating his model to cesm 1.0.1, while I use 1.0.4 i dunno realy know what I could do?

http://forum.cgd.ucar.edu/showthrea...oint-simulation&highlight=ERROR+degree+bounds



Thanks

Sebastian

 

[erik]

Sebastian

I just followed the instructions in the CLM User's Guide to run Example 5.1 on bluefire.ucar.edu and it worked fine with no problems. And I didn't have to add 1x1_brazil to config_definition.xml. I just did the following (I removed the mistakes and things I did like "ls")...

104 16:13 svn co https://svn-ccsm-release.cgd.ucar.edu/model_versions/cesm1_0_4
105 16:15 cd cesm1_0_4/scripts/
107 16:15 ./create_newcase -case testSPDATASET -res pt1_pt1 -compset I -mach bluefire
108 16:15 cd testSPDATASET
109 16:16 set SITE=1x1_brazil
110 16:16 ./xmlchange -file env_conf.xml -id CLM_CONFIG_OPTS -val "-sitespf_pt $SITE"
111 16:16 ./xmlchange -file env_conf.xml -id CLM_PT1_NAME -val $SITE
117 16:18 vi env_conf.xml # change USE_MPISERIAL to TRUE
123 16:18 ./configure -case
124 16:18 testSPDATASET.bluefire.build
125 16:22 testSPDATASET.bluefire.run

Here's the tail of the cpl log file...

(seq_mct_drv) : Model initialization complete

tStamp_write: model date = 10102 0 wall clock = 2012-09-19 16:23:02 avg dt = 0.15 dt = 0.15
memory_write: model date = 10102 0 memory = 84.02 MB (highwater) 0.00 MB (usage) (pe= 0 comps= cpl ocn atm lnd ice glc)
tStamp_write: model date = 10103 0 wall clock = 2012-09-19 16:23:02 avg dt = 0.18 dt = 0.22
memory_write: model date = 10103 0 memory = 85.97 MB (highwater) 0.00 MB (usage) (pe= 0 comps= cpl ocn atm lnd ice glc)
tStamp_write: model date = 10104 0 wall clock = 2012-09-19 16:23:03 avg dt = 0.18 dt = 0.18
memory_write: model date = 10104 0 memory = 85.97 MB (highwater) 0.00 MB (usage) (pe= 0 comps= cpl ocn atm lnd ice glc)
tStamp_write: model date = 10105 0 wall clock = 2012-09-19 16:23:03 avg dt = 0.17 dt = 0.15
memory_write: model date = 10105 0 memory = 85.97 MB (highwater) 0.00 MB (usage) (pe= 0 comps= cpl ocn atm lnd ice glc)
Write restart file at 10106 0
(seq_rest_write) write rpointer file rpointer.drv
(seq_io_wopen) create file testSPDATASET.cpl.r.0001-01-06-00000.nc
tStamp_write: model date = 10106 0 wall clock = 2012-09-19 16:23:03 avg dt = 0.27 dt = 0.67
memory_write: model date = 10106 0 memory = 89.94 MB (highwater) 0.00 MB (usage) (pe= 0 comps= cpl ocn atm lnd ice glc)



(seq_mct_drv): =============== SUCCESSFUL TERMINATION OF CPL7-CCSM ===============
(seq_mct_drv): =============== at YMD,TOD = 10106 0 ===============
(seq_mct_drv): =============== # simulated days (this run) = 5.000 ===============
(seq_mct_drv): =============== compute time (hrs) = 0.000 ===============
(seq_mct_drv): =============== # simulated years / cmp-day = 864.251 ===============
(seq_mct_drv): =============== pes min memory highwater (MB) 89.937 ===============
(seq_mct_drv): =============== pes max memory highwater (MB) 89.937 ===============
(seq_mct_drv): =============== pes min memory last usage (MB) -0.001 ===============
(seq_mct_drv): =============== pes max memory last usage (MB) -0.001 ===============

I did try to run the examples on our machines and verify that they worked. I can't promise I did that for everything though. But, as above it does work for me. That doesn't mean you don't have an issue, but if we can't replicate the issue we aren't going to be able to help much.

Since, you don't have 1x1_brazil in your version of CESM1.0.4, I'm wondering if you really are running with CESM1.0.4? Do this

cd cesm1_0_4 (assuming your top level directory is named that way...)
svn diff --new . --old https://svn-ccsm-release.cgd.ucar.edu/model_versions/cesm1_0_4

The other issue is being able to run and build on your own machine. Which is a much tougher problem. I'm assuming you've already found thesection in the CESM UG on this...

http://www.cesm.ucar.edu/models/cesm1.0/cesm/cesm_doc_1_0_4/c2249.html

But, getting it to build and run on your own machine is NOT neccessarily easy and can take even an expert a lot of time. Especially, since you have to get things like MPI and NetCDF working. Following the guidelines in the CESM UG and the TroubleShooting chapter of the CLM UG will hopefully help. We don't have enough resources to help people port to their own machines, other than giving this kind of general advice.

Good luck!

 

[swieneke@meteo_uni-koeln_de]

Hi Erik,

thanks for your detailed answer. I solved my Problem by re-porting CESM on the basis of a generic machine. And now iam able to run the single point datasets on my own machine.

Thanks a lot.

 

[yangk2010@lzu_edu_cn]

Erik,Hello !I'm trying to run a  single-point case with my own atmosphere forcing. And I have read the CLM user's guide carefully. But three are still  some problems which always bothered me. I hear about that you have more expertise with different methods for running single-point. So I hope you can give me some advice.1. When I tried to use mksurfdata.pl script to create my own input datasets for CLM4.5, it warned me that :      Creating mapping file: map_1km-merge-10min_HYDRO1K-merge-nomask_to_1x1_maqu_nomask_aave_da_c141125.nc
      From input grid: /mnt/Disk16T/yangkai_file/data/inputdata/lnd/clm2/mappingdata/grids/UGRID_1km-merge-10min_HYDRO1K-merge-nomask_c130402.nc
      For output grid: /home/yangkai/Single-point/cesm1_2_0/models/lnd/clm/tools/shared/mkmapgrids/SCRIPgrid_1x1_maqu_nomask_c141125.nc       No support for --netcdf4 or UGRID on machines other than yellowstone/caldera/geyser       So I can just run a single-point case with CLM4.0. How can I solve this problem on my cluster ?  Is there the other way to  create of these mapping files or run CLM4.5 for single-point ?2.  I'm puzzled with anohter problem :  When I began to bulid the case, it would check the inputdata first.  I just wanted to run over a year(eg: from 1997-08-18 to 1998-10-31),  but it still warned me that 1972-2004 period atmospheric forcing is needed.  Could you hlep me to sovle this problem ?3.  In CLM user's guide documentation, it told me atmosphere forcing data should be separated per month called YYYY-MM.nc. For example, my begining date is 1997-09-18, ending date is 1998-10-31, How do I set the time variable ?   And I also saw  the atmopheric forcing of example 5-3,  its forcing data is hourly, and the time variable is set like this :     time = 0,   // time(0)
    0.04166667,   // time(1)
    0.08333334,   // time(2)
    0.125,   // time(3)
    0.1666667,   // time(4)
    0.2083333,   // time(5)
    0.25,   // time(6)
    0.2916667,   // time(7)
    0.3333333,   // time(8)
    0.375,   // time(9)
    0.4166666,   // time(10)
    0.4583333,   // time(11)
    0.4999999,   // time(12)
    0.5416666,   // time(13)
    0.5833333,   // time(14)
    0.625,   // time(15)
    0.6666667,   // time(16)
    0.7083334,   // time(17)
    0.7500001,   // time(18)
    0.7916667,   // time(19)
    0.8333334,   // time(20)     So When I run my own single-pointer case, What value dose the time variable get ?      Thanks in advance for your help !       I'm looking forward to your reply !       Best wishes !

 

[jlchen@lzb_ac_cn]

Hi swieneke, I met the similar problem with you, how did you solved it?

 

[1301110222@pku_edu_cn]

Hi   I try to solve this problem by modify some code.In csm_share/shr_map_mod.F90 line 1110 ‘if (shr_map_checkRad(Ysrc)) then’change to'[/b]if (shr_map_checkRad(Ydst)) then'[/b] Yuanhong

 

[yfan1@uni-goettingen_de]

I had this 'shr_map_getWts error too, which was caused by a wrong mapalgo of CO2 stream. I used mapalgo=bilinear for the CO2 stream which should actually be "nn" nearest neighbor, because the domain of the CO2 file is a single point for the whole globe). There is also a note on the user guide saying that this error could be casued by wrong aerosol prescribed to single-point simultion: http://www.cesm.ucar.edu/models/cesm1.0/clm/models/lnd/clm/doc/UsersGuide/x10455.htmlall the best

 

[syang]

I had this 'shr_map_getWts error too, which was caused by a wrong mapalgo of CO2 stream. I used mapalgo=bilinear for the CO2 stream which should actually be "nn" nearest neighbor, because the domain of the CO2 file is a single point for the whole globe). There is also a note on the user guide saying that this error could be casued by wrong aerosol prescribed to single-point simultion: http://www.cesm.ucar.edu/models/cesm1.0/clm/models/lnd/clm/doc/UsersGuide/x10455.htmlall the best

I met the same problem. Coule you let know how did you solve it?