Opened 8 years ago

Closed 8 years ago

#837 closed help (fixed)

unable to get vn7.6 LAM to run over Antarctica

Reported by: anmcr Owned by: willie
Component: UM Mesoscale Keywords: LAM, failure to converge
Cc: Platform:
UM Version: 7.6

Description

Hi,

I have a vn7.6 job running at 4km resolution over a section of Antarctica called Durmont D'urville (xfggh). I am running a case study of a 'strong wind event'. The job is a copy from the one which Mark Weeks at the MetOffice? and Andy Elvidge have used to successfully model the Antarctic Peninsula region. This run fails at the first timestep with the 'GCR(2) error' suggesting a failure for the numerical solution to converge.

I have tried every suggestion to get it to work: increasing the width of the blending zone, initailising from a startdump produced from the 12km model after a few hours; reducing the timestep, any number of combinations of switching off diffusion, targeted diffusion, etc, etc. I want to use vn7.6 as it contains the modifications to account for the cold valley problem, which is critical around Antarctica. Perhaps the one thing I have not tried to get the vn7.6 run working is Grenville's suggestion to simply move the LAM location a small fraction.

As a test I ran the same set up but using vn7.1 (xfggz), which ran without any problem.

So I've ran out of ideas. The fall back option would be to scrap the vn7.6 runs and use the vn7.1 runs, but I am reluctant to do this.

Thanks,

Andrew

Change History (3)

comment:1 Changed 8 years ago by willie

  • Owner changed from um_support to willie
  • Status changed from new to accepted

Hi Andrew,

Something nasty has happened to the start dump xfggh.start. The sea ice field has been corrupted. I found this by cumf'ing the start dump with itself and looking in the summary.

Regards,

Willie

comment:2 Changed 8 years ago by anmcr

Hi Willie,

Thanks for your help. I managed to get this run working.

Best wishes,

Andrew

comment:3 Changed 8 years ago by willie

  • Resolution set to fixed
  • Status changed from accepted to closed
Note: See TracTickets for help on using tickets.