-
Notifications
You must be signed in to change notification settings - Fork 0
access_AncillaryFileGeneration
For more recent information see CentralAncillaryProgram
Version 6.6 of the ancillary file generation code is installed on cherax. Unless otherwise specified all paths here refer to cherax (an earlier version of this document described the vn6.3 setup on d2).
UMDP 70 Ancillary File Data Sources https://nf.nci.org.au/facilities/software/UM/7.8/umdoc_system/UM_docs/papers/pdf/p070.pdf
UMDP 73 Ancillary File Creation https://nf.nci.org.au/facilities/software/UM/7.8/umdoc_system/UM_docs/papers/pdf/p073.pdf
There are ancillary files at a number of standard resolutions in the directories vayu:/data/projects/access/ancil and cherax:~access/umdir/vn6.3/ancil/atmos.
Some of these also have a version with a _360 suffix to denote use of a 360 day calendar which is used by the climate model. Note that all the ancillary files contain information about the calendar in their header and the model will complain if the wrong one is used. E.g. there are different mask files in n96 and n96_360 (even though they differ by only a single byte in the header the model won't run if the wrong one is used).
Note that we can not reproduce these files exactly because they were generated with the US Navy 10' orographic dataset which has now been deprecated in favour of the GLOBE orography.
There is also an n96_hadgem1 subdirectory. This has a different land mask to the standard N96 case (generated outside the normal ancillary file framework according to Clive).
The high resolution ("master") files are in cherax:~access//umdir/ancil_master. Some of these were sent by Clive together with the original version of the sortware and some have been obtained more recently from Glenn Greed. See UMDP 70 for more information on data sources. Climatological fields have 12 monthly values.
- qrclim.ice Climatological sea-ice on 1 degree grid
- qrclim.ozone.li Shine and Li ozone, 2.5 degree grid, 47 levels
- qrclim.smc_gswp2 Snow and soil moisture from offline MOSES-2 simulation
- qrclim.smow_amip Monthly snow and soil moisture (4 levels) on N48 grid (from UKMO AMIP experiment)
- qrclim.smow_wr Monthly snow and soil moisture (4 levels) on N216 grid (Willmott and Rowe data)
- qrclim.slt Deep soil temperature, 4 soil levels, N48 grid (from UKMO AMIP experiment)
- qrclim.sst Climatological SST on 1 degree grid
- qrparm.ash.GLOBE Orographic roughness on 5 minute grid
- qrparm.igbp.1km (This one is a formatted direct access file)
- qrparm.orog1.GLOBE Surface height at 1 minute resolution
- qrparm.soilcarbon Soil carbon content at 0.5 degree resolution
- qrparm.whs Volume SMC at wilting and root depth on 1 degree grid
All except qrparm.igbp.1km can be inspected with xconv. However xconv mucks up the dimensions of the ozone file, showing 12 levels and 47 times.
There are namelists specifying a number of global and regional grids in ~access/umdir/ancil_parent/src/ancil/control/namelist/grids. With these as examples it should be possible to set up a namelist for any local grid required.
There are also orography namelists in namelist/orog. There are slight differences in some of the options in the high resolution files. **Check whether these are used in normal jobs **
Changes directory. This contains point fix ups for masks and orography. These may no longer be necessary now that we're using a new orography dataset?
Generation of ancillary files is controlled by a script ~access/umdir/ancil_parent/ancil_top. This generates an cherax batch job with the appropriate parameters to generate the desired files.
Central executable.
export ANCIL_PARENT=~access/umdir/ancil_parent
export ANCIL_MASTER=~access/umdir/ancil_master
The grid is specified with the model argument to the ancil_top script. This specifies a grid namelist from the directory $ANCIL_PARENT/src/ancil/control/grids. E.g. for an N96 grid, run with
$ANCIL_PARENT/bin/ancil_top -model n96 -memory 4000
Output goes to $DATADIR/ancil.vn6.6 by default. The ancil_top script uses the current directory for a couple of temporary files, so make sure to run it from a directory for which you have write permission.
This has to be created first because other calculations use it.
export DATASETS="mask"
export MASKOPTS="-m 4" # (for regular grids)
or
export MASKOPTS="-m 2" (# for rotated grids)
The job takes about 500 CPU s on the cherax just to create the land mask at N96 (most time seems to be spent reading the high resolution data so this isn't very resolution dependent). it also requires about 4000 MB of memory. This job produces files
- qrparm.landfrac
- qrparm.mask
Possibly for NWP don't want a land fraction file?
This uses the global 1km orography qrparm.orog1.GLOBE
export DATASETS="orog"
export OROGOPTS="-d GLOBE1"
N96 requires an 4 GB memory limit and takes about 1200 sec. It's supposed to be possible to run a parallel version of this job, but this hasn't been tested here yet.
export DATASETS='vegsoil'
export VEGSOPTS="-w"
This uses qrparm.soilcarbon and qrparm.whs. Produces
- qrparm.veg.dist
- qrparm.veg.frac
- qrparm.veg.func
- qrparm.soil
- qrparm.veg
-a for AMIP climatology, -g for GSWP2
This requires preexisting qrparm.mask and qrparm.soil files. There are 3 different soil moisture datasets, selectable with SMOWOPTS. The default is the WR dataset. Use -a for AMIP climatology, -g for GSWP2. Existing standard resolution files seem to use the AMIP data.
export DATASETS="smow"
export SMOWOPTS="-a"
export DATASETS="vegfrac"
export VEGFOPTS="-g 4"
(To use the same aggregation method as in the land sea mask calculation).
Uses qrparm.soilcarbon, qrparm.whs, qrparm.igbp.1km
Creates
- qrparm.soil.dust_igbp
- qrparm.soil_igbp
- qrparm.veg.frac_igbp
- qrparm.veg.func_igbp
This takes 1176 s to generate annual mean fields. Adding the -t option to generate seasonal values takes ??
This also requires the model levels to be specified as an argument to ancil_top. Global models use smoothed zonal mean ozone, specified as OZOOPTS arguments.
export DATASETS="ozone"
export OZOOPTS="-s -z"
ancil_top -model n48 -memory 2000 -levels 38
For 50 levels using the file vertlevs_strat_50,
ancil_top -model n48 -memory 2000 -levels 50 -vertical strat_50
On solar, generating an N96 mask takes about 240 sec and needs a memory limit of about 1 GB.
Jeff Cole from the University of Reading has written a program xancil that creates Unified Model ancillary files from NetCDF input files, http://cms.ncas.ac.uk/documents/xancil/
xancil and mkancil executables (version 0.50) are in raijin:~access/bin. They're generic i386 binaries so should also work on most linux machines. xancil is the control program and sets up namelists for mkancil and calls it.
This will create ancillary files for
- Ozone
- Soil moisture and snow depth
- Deep soil temperatures
- Soil parameters
- Vegetation parameters
- Vegetation fractions
- Vegetation Functional types
- Disturbed Vegetation Fraction
- Sea Surface Temperature
- Sea Ice
- Orography
- Land/sea mask
- Land fraction
The program just does a direct conversion from netcdf to UM format with no grid conversion. It should be possible to use the program to generate master ancillary files that are interpolated with CAP, as well as files on model grids directly. Generating an orography file from an ETOPO file worked properly.
We have the source code for xancil and Maciej has modified it to produce CO2 surface emission files. Executables are in d2:~gol124/bin/xancil/ and cherax:~access/umdir/bin. See BuildingXancil for more details.
Note that by default xancil creates a file for UM vn4.5 where the grids ran North to South, rather than South to North as in recent versions. To get a file that works with the current model, set the version number to 6 in the Configuration -> General configuration option. TODO - Change the default.