Fire Model

How to run the Fire Model


At its core, the fire model is a nested UM8.5 model with 5 layers, in which a fire model is coupled to Jules for the 2 highest-resolution runs.


To start, the global model is run from 10am the day before the fire until midnight after the fire.

The output of one run is used to create the initial and boundary conditions for the next higher-resolution nest. There is no back-coupling to the lower resolution models, so each nest can be run independently of the higher-resolution nests.

The following is an example for the nesting for the Black Saturday Fires:

Area Resolution Start
Global 10am the day before
Eastern Australia 4km 12 noon the day before
1.5km 2pm the day before
444m midnight before
144m midnight



Coupled in reference to the fire model means whether the fire model feeds back into the atmospheric model (i.e. temperature, moisture) or not. This is set by a flag in the fire.inp file.

Model Run Sequence

Black Saturday: Fire start is hard coded as 12 noon 7/2/2009

Step Map fire code start finish
1 Global no 2009-02-06_1000 2009-02-08_0000
2 4km no 2009-02-06_1200 2009-02-08_0000
3 1.5km no 2009-02-06_1400 2009-02-08_0000
4 444m no 2009-02-07_0000 2009-02-08_0000
5 444m yes 2009-02-07_0800 2009-02-08_0000
6 144m no 2009-02-07_0000 2009-02-08_0800
7 144m yes 2009-02-07_0800 2009-02-08_0000

Note that the 444m nest runs through till the end of the day without fire code to generate boundary conditions for the 144m nest. The 144m nest without fire only needs to run till 8am, since then the fire-144m will take over.

How to run the model

Apparently CRUNs don't work with the fire model executable. Also, the model run time is varying widely (probably depending on the scale of the fire?), so there's no good assessment on how long a model step will take to complete.

The model is run manually in this way: Submit an NRUN for the complete time frame, request maximum walltime. Model will probably fail, either because the LBC files run out, or the walltime expires. Set up a new model run, refer the last restart-dump as the start dump (and adjust the LBC as required), submit it as another NRUN. Rinse and repeat.


BLST Black Saturday 7/2/2009 - Day of devastating bushfires in Victoria
SRTM Shuttle Radar Topography Mission] Source of detailed topographic data


  • Copied varu.s}} into {{vatw.a and converted it into a compile job. (was later deleted)
  • Several handedits and user stashmaster files were in ~cbe563/um_nesting which was not readable to me. Chermelle changed that for me.
  • Extract failed because FCM Extract Directory was set to $LOCALDATA}}, which I haven't set. Fixed that to {{/scratch/users/$USER
  • Extract failed because UM_SVN_URL}} is set to Template:Svn://fcm2/UM svn/UM/trunk. I changed that to {{fcm:um_tr, and I hope that's okay.
  • scp UMATMOS failed, because it was still trying to use username cbe563. The user details were introduced in lots of hand edit files. Made considerable changes to the handedit files.
  • scp UMATMOS failed, as UM_ROUTDIR}} was set to Template:/working/nwp/... -- changed to {{/short/$PROJECT/$USER/working/...
  • Extract of JULES failed: ! revision keyword not defined

I'm giving up on trying to compile the model, I'll start again, this time just executing.

  • Trying again: Copy varu.s}} to {{vatw.b.
  • Copy Chermelle's ~/um_nesting/vn8.5 directory to my own directory.
  • Changed values in ~/um_nesting/vn8.5/user_details}} and {{~/um_nesting/vn8.5/vanc/expt_details to point to relevant directories, userids, et cetera. Note: Apparently it only works when hard coded.
  • Links in ~/um_nesting/vn8.5}}: {{exptid -> vatw}} and {{vatw -> vanc (apparently it doesn't work any other way.)


  • UM Executable : /short/v46/cbe563/bin/vasfv.exec not readable
  • dd: opening `/short/w35/hxw599/fire/BLST/vatw//vatwb/vatwb.out2': No such file or directory

Compile Executable

  • Copied vasf.v}} to {{vatw.a.
  • Copied Chermelle's JULES changes to ~hxw599/jules/src/jules/src