Opened 10 years ago

Closed 9 years ago

#191 closed help (fixed)

Problem with radiation scheme in UM7.0 with UKCA: loss of values from D1

Reported by: luke Owned by: um_support
Priority: high Component: UM Model
Keywords: Cc:
Platform: UM Version: <select version>

Description

In getting UKCA to work within UM7.0 I have noticed that a problem occurs after the 6th timestep. Using totalview to debug the code showed that the values of arrays THETA and PV_ON_THETA_MLEVS become NaN. Since the radiation schemes are called every 6th timestep I have reset this frequency to every 12th timestep (or rather 4 times a day for 48 timesteps a day) and now the model crashes after the 12th timestep. This behaviour is not seen in a model run without UKCA.

Model without UKCA: XDJTC
VN7.0 job copied from Ros' XDLQC. Runs for 10 days with no problems.

Model with UKCA: XDJTD
Same as C job above but with UKCA additions from MOHC which have been documented on the UM Trac page here: http://puma.nerc.ac.uk/trac/UM/ticket/9 .

Location on HECToR:
/work/n02/n02/luke/xdjtd
/home/n02/n02/luke/um/umui_out/xdjtd000.xdjtd.d08323.t162534.leave (for example)
All files should be readible. Please email me on Luke.Abraham@… for more information if required.

Change History (3)

comment:1 Changed 10 years ago by luke

Job XDJTE (a copy of UKCA-ready XDJTD but with UKCA switched off) compiles, reconfigures, and runs, but much slower than XDJTC (the UM7.0 base job, a copy of Ros's XDLQC). This job takes 5 hours to run 10 days on 16 nodes (32 processors).

It seems that the problems only occur when UKCA is switched on.

comment:2 Changed 10 years ago by luke

(It should be noted that the change in timing for the job is due to the inclusion of 26 advected tracers)

comment:3 Changed 9 years ago by lois

  • Resolution set to fixed
  • Status changed from new to closed
  • UM Version set to <select version>
Note: See TracTickets for help on using tickets.