Create ocean temperature and salt nudging and damping coefficient files. These files can used to nudge either MOM 0.25 degree, MOM 1 degree or NEMO models towards observations.
Build status:
This tool is written in Python and depends a few different Python packages. It also depends on ESMF_RegridWeightGen program to perform regridding between non-rectilinear grids.
Download ocean-nudge:
$ git clone --recursive https://github.com/nicjhan/ocean-nudge.git
$ cd ocean-nudge/regridder/
$ wget http://s3-ap-southeast-2.amazonaws.com/dp-drop/ocean-regrid/grid_defs.tar.gz
$ tar zxvf grid_defs.tar.gz
Use Anaconda as described below or an existing Python setup.
- Download and install Anaconda for your platform.
- Setup the Anaconda environment. This will download all the necessary Python packages.
$ cd ocean-nudge
$ conda env create -f regrid.yml
$ source activate regrid
Install ESMF_RegridWeightGen. ESMF releases can be found here.
There is a bash script regridder/contrib/build_esmf.sh which the testing system uses to build ESMF. This may be useful in addition to the ESMF installation docs.
Follow the example steps below to nudge MOM or NEMO using ORAS4 monthly data. In addition here is an example that nudges NEMO using GODAS pentad.
Download temperature and salinity fields from the GODAS or ORAS reanalysis dataset. Be sure to download GODAS in NetCDF format. Useful URLs:
- GODAS: http://www.esrl.noaa.gov/psd/data/gridded/data.godas.html
- ORAS4: ftp://ftp.icdc.zmaw.de/EASYInit/ORA-S4/
For example for ORAS4:
$ cd test
$ mkdir -p test_data/input/
$ cd test_data/input/
$ wget ftp://ftp.icdc.zmaw.de/EASYInit/ORA-S4/monthly_orca1/thetao_oras4_1m_2003_grid_T.nc.gz
$ wget ftp://ftp.icdc.zmaw.de/EASYInit/ORA-S4/monthly_orca1/thetao_oras4_1m_2004_grid_T.nc.gz
$ wget ftp://ftp.icdc.zmaw.de/EASYInit/ORA-S4/monthly_orca1/thetao_oras4_1m_2005_grid_T.nc.gz
$ gunzip *.gz
Alternatively, download the test data that comes with the package:
$ cd test
$ wget http://s3-ap-southeast-2.amazonaws.com/dp-drop/ocean-nudge/test/test_data.tar.gz
$ tar zxvf test_data.tar.gz
$ cd test_data/input
Regrid the reanalysis files to the model grid. This can be done with regrid_simple.py or regrid.py found in the regridder directory.
The following commands assume a working directory of test/test_data/input.
For example, for MOM 0.25 degree:
$ cd test/test_data/input
$ ../../../regridder/regrid_simple.py ORAS4 thetao_oras4_1m_2004_grid_T.nc thetao \
MOM oras4_temp_on_mom_grid.nc --regrid_weights oras4_mom_regrid_weights.nc
The format for the regrid.py command is similar to the above but the location of all grid definition files must be given:
$ cd test/test_data/input
$ export GRID_DEFS=../../../regridder/grid_defs
$ ../../../regridder/regrid.py --help
$ ../../../regridder/regrid.py ORAS4 $GRID_DEFS/coordinates_grid_T.nc \
$GRID_DEFS/coordinates_grid_T.nc thetao_oras4_1m_2004_grid_T.nc thetao \
MOM $GRID_DEFS/ocean_hgrid.nc $GRID_DEFS/ocean_vgrid.nc oras4_temp_on_mom_grid.nc temp \
--dest_mask $GRID_DEFS/ocean_mask.nc --regrid_weights oras4_mom_regrid_weights.nc
We can use a bash for-loop to regrid multiple files with a single command:
$ for i in 2003 2004 2005; do \
../../../regridder/regrid.py ORAS4 $GRID_DEFS/coordinates_grid_T.nc \
$GRID_DEFS/coordinates_grid_T.nc thetao_oras4_1m_${i}_grid_T.nc thetao \
MOM $GRID_DEFS/ocean_hgrid.nc $GRID_DEFS/ocean_vgrid.nc oras4_temp_${i}_mom_grid.nc temp \
--dest_mask $GRID_DEFS/ocean_mask.nc --regrid_weights oras4_mom_regrid_weights.nc;
done
And for MOM 1 degree:
$ for i in 2003 2004 2005; do \
../../../regridder/regrid.py ORAS4 $GRID_DEFS/coordinates_grid_T.nc \
$GRID_DEFS/coordinates_grid_T.nc thetao_oras4_1m_${i}_grid_T.nc thetao \
MOM1 $GRID_DEFS/grid_spec.nc $GRID_DEFS/grid_spec.nc oras4_temp_${i}_mom_grid.nc temp \
--dest_mask $GRID_DEFS/grid_spec.nc --regrid_weights oras4_mom_regrid_weights.nc;
done
Note that the model name has been changed from MOM to MOM1 and the grid definition files have been changed.
And for NEMO:
$ for i in 2003 2004 2005; do \
../../../regridder/regrid.py ORAS4 $GRID_DEFS/coordinates_grid_T.nc \
$GRID_DEFS/coordinates_grid_T.nc thetao_oras4_1m_${i}_grid_T.nc thetao \
NEMO $GRID_DEFS/coordinates.nc $GRID_DEFS/data_1m_potential_temperature_nomask.nc \
oras4_temp_${i}_nemo_grid.nc temp \
--regrid_weights oras4_nemo_regrid_weights.nc;
done
Note that in this case because the --regrid_weights option is used the computationally expensive part of the regridding only done once and the whole operation should be relatively fast.
Combine the above regridded reanalysis files into a single nudging source file. Note the reanalyses are monthly averages with a nominal time index in the middle of the month. This means that in order for nudging to start at the beginning of the year data from the previous year is needed - the models create data for the beginning of January by interpolating from December of the previous year. So, for example to nudge the model for the whole of 2004:
e.g. for MOM 0.25 degree:
$ ../../../makenudge.py MOM temp --forcing_files oras4_temp_2003_mom_grid.nc
oras4_temp_2004_mom_grid.nc oras4_temp_2005_mom_grid.nc
For MOM 1 degree:
$ ../../../makenudge.py MOM1 temp --forcing_files oras4_temp_2003_mom_grid.nc
oras4_temp_2004_mom_grid.nc oras4_temp_2005_mom_grid.nc
For MOM there will be two kinds of otuput, the actual nudging source file which ends in _sponge.nc and the 3D relaxation coefficient file which ends in _coeff_sponge.nc.
For Nemo:
$ ../../../makenudge.py NEMO temp --forcing_files \
oras4_temp_2003_nemo_grid.nc oras4_temp_2004_nemo_grid.nc \
oras4_temp_2005_nemo_grid.nc
This will output two files: <input_var_name>_nomask.nc and resto.nc. In order to do temperature and salinity nudging this needs to be done twice, once for temperature (and above) and once for salinity. For example:
$ ../../../makenudge.py NEMO salt --forcing_files oras4_salt_2003_nemo_grid.nc \
oras4_salt_2004_nemo_grid.nc oras4_salt_2005_nemo_grid.nc
Configure the model to use the newly created nudging file.
Copy the *_sponge.nc files from above into the MOM INPUT directory. Then add the following to the input.nml:
&ocean_sponges_tracer_nml
use_this_module = .TRUE.
damp_coeff_3d = .TRUE.
/
Take note of the model output as MOM starts up, there should be output similar to the following:
==> Note from ocean_sponges_tracer_mod: Using this module.
==> Using sponge damping times specified from file INPUT/temp_sponge_coeff.nc
==> Using sponge data specified from file INPUT/temp_sponge.nc
A common error looks like:
FATAL from PE 39: time_interp_external 2: time after range of list,file=INPUT/temp_sponge.nc,field=temp
This means that the time range covered by the nudging/sponge file does not match the model runtime. For example the sponge time may go from 0001-01-01 to 0002-01-01 while the model starts in 2004. Check the time range of the nudging source file with the following commands:
$ ncdump -h temp_sponge.nc
$ ncdump -v time temp_sponge.nc
The NEMO_3.6 ORCA2 configuration is already set up to do global nudging, it does this to maintain temperature and salinity tracers in the Mediterranean. To carry out global nudging with the files generated above:
-
replace the input files: 1_data_1m_potential_temperature_nomask.nc and 1_data_1m_salinity_nomask.nc with the corrosponding temperature and salinity _nomask.nc files generated above. Overwrite the input file resto.nc with the file of the same name generated above.
-
Check the values of following configuration namelist parameters:
&namrun ! parameters of the run
!-----------------------------------------------------------------------
ln_rstart = .false. ! start from rest (F) or from a restart file (T)
&namtsd ! data : Temperature & Salinity
!-----------------------------------------------------------------------
! ! file name ! frequency (hours) ! variable ! time interp. ! clim ! 'yearly'/ ! weights ! rotation ! land/sea mask !
! ! ! (if <0 months) ! name ! (logical) ! (T/F) ! 'monthly' ! filename ! pairing ! filename !
sn_tem = 'data_1m_potential_temperature_nomask', -1 ,'votemper' , .true. , .true. , 'yearly' , '' , '' , ''
sn_sal = 'data_1m_salinity_nomask' , -1 ,'vosaline' , .true. , .true. , 'yearly' , '' , '' , ''
ln_tsd_init = .true. ! Initialisation of ocean T & S with T &S input data (T) or not (F)
ln_tsd_tradmp = .true. ! damping of ocean T & S toward T &S input data (T) or not (F)
!-----------------------------------------------------------------------
&namtra_dmp ! tracer: T & S newtonian damping
!-----------------------------------------------------------------------
ln_tradmp = .true. ! add a damping termn (T) or not (F)
nn_zdmp = 0 ! vertical shape =0 damping throughout the water column
! =1 no damping in the mixing layer (kz criteria)
! =2 no damping in the mixed layer (rho crieria)
cn_resto = 'resto.nc'! Name of file containing restoration coefficient field
/
MOM 1 degree nudged with GODAS pentad. Days 1, 5, and 10. Note that the high latitudes are not nudged due to the limited GODAS domain.
Be aware of this issue pyinstaller/pyinstaller#1781. It may be necessary to downgrade setuptools with the following command:
$ conda install setuptools==19.2
First install pyinstaller:
$ pip install pyinstaller
Create release:
$ cd release
$ pyinstaller makenudge.spec
$ mv dist/makenudge ./makenudge-x.x.x
Also include the regridder in the makenudge release. This is needed to regrid analyses datasets prior to creating the nudging source files.
$ cd ../regridder/release
$ pyinstaller regrid.spec
$ cp -r dist/regrid ../../release/makenudge-x.x.x/
$ cd ../../release/
$ tar czvf makenudge-x.x.x.tar.gz makenudge-x.x.x
When running pyinstaller regrid.spec above there may be an error like:
ERROR: can't find libllvmlite.so, please edit libllvm var in regrid.spec.
This means that pyinstaller can't find libllvmlite.so. By default it looks in the directory $HOME/anaconda2/lib/python2.7/site-packages/llvmlite/binding/. Please edit regrid.spec to specify the location of this library.
Upload tarball to s3:
$ s3put -b dp-drop -p /short/v45/nah599/more_home/ ./makenudge-x.x.x.tar.gz
$ s3cmd setacl --acl-public --guess-mime-type s3://dp-drop/ocean-nudge/release/makenudge-x.x.x.tar.gz
$ wget http://s3-ap-southeast-2.amazonaws.com/dp-drop/ocean-nudge/release/makenudge-0.0.1.tar.gz
$ tar zxvf makenudge-0.0.1.tar.gz
$ export PATH=$(pwd)/makenudge-0.0.1/:$(pwd)/makenudge-0.0.1/regrid/:$PATH
$ makenudge --help
$ regrid --help
$ mkdir test
$ cd test/
$ wget http://s3-ap-southeast-2.amazonaws.com/dp-drop/ocean-regrid/grid_defs.tar.gz
$ tar zxvf grid_defs.tar.gz
$ export GRID_DEFS=$(pwd)/grid_defs
$ wget http://s3-ap-southeast-2.amazonaws.com/dp-drop/ocean-nudge/test/test_data.tar.gz
$ tar zxvf test_data.tar.gz
$ cd test_data/input
$ for i in 2003 2004 2005; do \
regrid ORAS4 $GRID_DEFS/coordinates_grid_T.nc \
$GRID_DEFS/coordinates_grid_T.nc thetao_oras4_1m_${i}_grid_T.nc thetao \
MOM $GRID_DEFS/ocean_hgrid.nc $GRID_DEFS/ocean_vgrid.nc oras4_temp_${i}_mom_grid.nc temp \
--dest_mask $GRID_DEFS/ocean_mask.nc --regrid_weights oras4_mom_regrid_weights.nc; done
$ makenudge MOM temp --forcing_files oras4_temp_2003_mom_grid.nc oras4_temp_2004_mom_grid.nc \
oras4_temp_2005_mom_grid.nc
Note that the regrid and makenudge steps above are quite heavy on CPU, memory and disk. The nudging files can become quite large because they contain 4d fields.