Opened 4 years ago

Last modified 4 years ago

#165 accepted

G2 ancillaries with Fiji and correct Australian orography

Reported by: Martin Dix Owned by: Martin Dix
Priority: major Component: ACCESS model
Keywords: Cc:

Description (last modified by Martin Dix)

G2 is currently using Met Office ancillaries. These use the standard GLOBE30 orography with all its problems over Australia (wiki:access/HighResolutionAustralianOrography). Also Fiji is missing from the land-sea mask (https://code.metoffice.gov.uk/trac/TI/ticket/7).

Attachments (8)

fiji_orography.png (44.3 KB) - added by Martin Dix 4 years ago.
treefrac_fiji_new.png (43.2 KB) - added by Martin Dix 4 years ago.
treefrac_fiji_std.png (42.6 KB) - added by Martin Dix 4 years ago.
treefrac_hawaii_new.png (40.0 KB) - added by Martin Dix 4 years ago.
treefrac_hawaii_std.png (39.9 KB) - added by Martin Dix 4 years ago.
aust_orog_std.png (78.4 KB) - added by Martin Dix 4 years ago.
aust_orog_new.png (77.3 KB) - added by Martin Dix 4 years ago.
aust_orog_diff.png (87.7 KB) - added by Martin Dix 4 years ago.

Download all attachments as: .zip

Change History (13)

comment:1 Changed 4 years ago by Martin Dix

Owner: set to Martin Dix
Status: newaccepted

comment:2 Changed 4 years ago by Martin Dix

Description: modified (diff)

Changed 4 years ago by Martin Dix

Attachment: fiji_orography.png added

Changed 4 years ago by Martin Dix

Attachment: treefrac_fiji_new.png added

Changed 4 years ago by Martin Dix

Attachment: treefrac_fiji_std.png added

Changed 4 years ago by Martin Dix

Attachment: treefrac_hawaii_new.png added

Changed 4 years ago by Martin Dix

Attachment: treefrac_hawaii_std.png added

comment:3 Changed 4 years ago by Martin Dix

The ancillary program can optionally use a list of mask overrides. Standard Met Office N512 set is /projects/access/apps/cap/7.9/control/changes/mask/mask_changes.n512.

Script ~access/apps/pythonlib/umfile_utils/mask_edit.py allows interactive generation of mask overrides by clicking on a map. Using this generates N512 overrides for Fiji as

&DATAC FIELD_NO=1, POINT_NO=465407, DATA_NEW=.TRUE. /
&DATAC FIELD_NO=1, POINT_NO=465408, DATA_NEW=.TRUE. /
&DATAC FIELD_NO=1, POINT_NO=466430, DATA_NEW=.TRUE. /
&DATAC FIELD_NO=1, POINT_NO=466431, DATA_NEW=.TRUE. /
&DATAC FIELD_NO=1, POINT_NO=466432, DATA_NEW=.TRUE. /
&DATAC FIELD_NO=1, POINT_NO=467454, DATA_NEW=.TRUE. /
&DATAC FIELD_NO=1, POINT_NO=470522, DATA_NEW=.TRUE. /
&DATAC FIELD_NO=1, POINT_NO=470523, DATA_NEW=.TRUE. /
&DATAC FIELD_NO=1, POINT_NO=470524, DATA_NEW=.TRUE. /
&DATAC FIELD_NO=1, POINT_NO=470525, DATA_NEW=.TRUE. /
&DATAC FIELD_NO=1, POINT_NO=471546, DATA_NEW=.TRUE. /
&DATAC FIELD_NO=1, POINT_NO=471547, DATA_NEW=.TRUE. /
&DATAC FIELD_NO=1, POINT_NO=471548, DATA_NEW=.TRUE. /
&DATAC FIELD_NO=1, POINT_NO=471549, DATA_NEW=.TRUE. /
&DATAC FIELD_NO=1, POINT_NO=472570, DATA_NEW=.TRUE. /
&DATAC FIELD_NO=1, POINT_NO=472571, DATA_NEW=.TRUE. /
&DATAC FIELD_NO=1, POINT_NO=472572, DATA_NEW=.TRUE. /
&DATAC FIELD_NO=1, POINT_NO=472573, DATA_NEW=.TRUE. /

Appended to the standard Met Office set to create a new override file /projects/access/apps/cap/7.9/control/changes/mask/mask_changes.n512_access_G2.

Mask and orography generated with this look correct


However the vegetation fractions for Fiji looked strange with much more bare soil than expected. This was also true for Hawaii (and similar at several resolutions).



The CAP vegetation interpolation code searches for the nearest defined points. For Hawaii this is somewhere on the Mexican coast. Apart from any question of how reasonable it is to take values from this far away, a code error in calculating the weights means that many points in the island block get very small values that are rounded down to zero, leaving only bare soil.

http://glcf.umd.edu/data/lc has IGBP vegetation classes derived from MODIS data with proper values for Hawaii and Fiji. The vegetation code doesn't have the same sort of point override as the mask, so the simplest fix was to overwrite the missing values in the master 1km data set with the MODIS values. There's not much variation from year to year. For N512 it doesn't seem worth using the highest resolution (500m) data so I used the more convenient 5' data.

The modified file is /projects/access/data/master/vegetation/cover/igbp/v2/qrdata.igbp.island_fix.

This gives more reasonable looking vegetation fractions,


Note that if I'd used this corrected master vegetation from the start the mask overrides wouldn't have been necessary.

Standard orography over Australia


GA orography over Australia


Difference


Other orography differences

Even using the standard GLOBE30 data the orography generated here showed some small differences compared to the Met Office file (e.g. /g/data/dp9/da/access-g/um/ancil/atmos/n512/orography/globe30/ukmo/v1/qrparm.orog. This was mainly over Antarctica and is likely due to some difference in the filtering. At the moment the source of the difference is still unexplained. To avoid these extra differences complicating the comparison, the Australia + Fiji region was copied from the new orography to the standard orography, so the final file only has differences over the expected region.

To merge files

python ~access/apps//pythonlib/umfile_utils/mergefiles_region.py  -x 312,200 -y 191,150 orog_new orog_std orog_merged
Last edited 4 years ago by Martin Dix (previous) (diff)

Changed 4 years ago by Martin Dix

Attachment: aust_orog_std.png added

Changed 4 years ago by Martin Dix

Attachment: aust_orog_new.png added

Changed 4 years ago by Martin Dix

Attachment: aust_orog_diff.png added

comment:4 Changed 4 years ago by Martin Dix

Soil & dust

The HWSD master soil dataset (2.5 minute = 0.04167 degree resolution) includes both Fiji and Hawaii so there's no problem with trying to fill missing data from distant points.

Other fields

Following fields are not used by the cycling suite but possibly used for cold start?

Soil temperature

Master dataset is only an N96 model run. Fiji values seem to be taken from Queensland coast and look reasonable

Soil moisture and snow

Fiji gets plausible soil moisture values, similar to Queensland coast. Snow simply zero.

SST and ice

Fields identical to standard files except for Fiji being now masked out.

New files

In directory /g/data/dp9/da/access-g/um/ancil/atmos/n512 new files are

  • land_sea_mask/igbp/bom/v1/qrparm.mask
  • orography/globe30/bom/v1/qrparm.orog
  • smc_snow/gswp2_hwsd_vg/bom/v4/qrclim.smow
  • soil_dust/hwsd/bom/v2/qrparm.soil.dust
  • soil_parameters/hwsd_vg/bom/v5/qrparm.soil
  • soil_temp/amip/bom/v3/qrclim.slt
  • sst/hadisst_6190/bom/v1/qrclim.sst
  • vegetation/fractions_igbp/bom/v2/qrparm.veg.frac
  • vegetation/func_type_modis/bom/v2/qrparm.veg.func

Version numbers were chosen to match Met Office files.

Last edited 4 years ago by Martin Dix (previous) (diff)

comment:5 Changed 4 years ago by Martin Dix

The rose suite au-aa388 generates the ancillaries. Rev 1302 generates the versions described above. Rev 1305 uses the modified vegetation to generate the mask too. This includes Fiji though it's not quite the same shape as with the override. This approach isn't really recommended but more saved as a record of what's possible.

Note: See TracTickets for help on using tickets.