Opened 11 years ago

Closed 10 years ago

#430 closed help (fixed)

Unable to run NAE starting from ECMWF

Reported by: oma Owned by: willie
Component: UM Model Keywords: GCR fail to converge; reduced timestep
Cc: Platform:
UM Version: 6.1

Description

Hello,

It's been a week now since I'm unsuccessfully trying three NAE runs with 76 vertical levels starting from ECMWF initial data. One run fails just at the first time step, the second run fails at the third time step, and the third run fails at the fifth time step. Apparently the model is able to complete the jobs with no error messages but the domain is completely filled with NaN's everywhere.

I've tried many things such as reconfiguring and not the ancillary files, etc. This weekend I found that the inconsistency must be somehow related to the land-sea mask or orography as is made evident in the attached figure. The figure shows the temperature at the lowest level after 1 time step.

Interestingly, I've also tried a UK Meso run with 38 vertical levels and this has been successful.

I'll be very grateful if you can help me.

Thanks in advance,

Oscar

Attachments (1)

temp_tstep_1_vlev_1.jpg (74.5 KB) - added by oma 11 years ago.
Temperature on the lowermost vertical level after one time step

Download all attachments as: .zip

Change History (8)

Changed 11 years ago by oma

Temperature on the lowermost vertical level after one time step

comment:1 Changed 11 years ago by willie

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

Hi Oscar,

I have started to look at this. I have tried configuring the land-sea mask and this gets it past the first time step. However, the GCR(2) is failing to converge. I will continue to look at this after HECToR is back from maintenance.

Regards,

Willie

comment:2 Changed 11 years ago by willie

  • Keywords GCR fail to converge; reduced timestep added

Hi Oscar,

I reduced the time step to one minute and it has run successfully for 15 timesteps. A timestep of 2.5 minutes still led to GCR filure to converge.

Let me know if this solves your problem.

Regards,

Willie

comment:3 Changed 11 years ago by oma

Hi Willie,

Can you let me know the job ID so that I can replicate it with the same settings?

Thanks,

Oscar

comment:4 Changed 11 years ago by willie

Hi Oscar,

It is xezgk

Regards,

Willie

comment:5 Changed 11 years ago by oma

Hi Willie,

I tried the smaller time step and didn't work. The model still produces NaN's along coastlines after the first time step, at least for levels below 4000 m. The fact that the NaN's are along the coastline makes me think that some of the fields are artificially displaced during the second reconfiguration. What do you reckon?

Regards,

Oscar

comment:6 Changed 11 years ago by willie

Hi Oscar,

You need to configure the land-sea mask too.

Regards,

WIllie

comment:7 Changed 10 years ago by willie

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