CLM5 run fail - dtlimit issue

[James King]

Hi all,

I have a transient historical run in CLM5 starting in 1901 which fails after the first few timesteps or so. The error in the atm.log is

(shr_dmodel_readstrm) file ub: /work/n02/n02/jking/cesm/CESM2.1.3/cesm_inputdata/atm/datm7/atm_forcing.datm7.GSWP3.0.5d.v1.c170516/Precip/clmforc.GSWP3.c2011.0.5x0.5.Prec.1901-01.nc 2
(shr_strdata_advance) ERROR: for stream 2
(shr_strdata_advance) ERROR: dt limit1 34310.125000000000 0.12500000000000000 1.5000000000000000
(shr_strdata_advance) ERROR: dt limit2 19010101 5400 19010101 16200
ERROR: (shr_strdata_advance) ERROR dt limit for stream

which implies some issue with the timesteps for the DATM input data. I'm running in CLMGSWP3v1 mode and haven't changed any of the datm namelist settings from the defaults - the input data are also the defaults, though I am running on a non-UCAR HPC (which is a supported machine). The run goes fine from 1850-1901, but fails after being restarted at 1901. I've attached the DATM namelist and log, and the output of ncdump -v time for the 1901-01 precip DATM input file. As anticipated, the data have 3-hourly temporal resolution (despite the header saying 6-hourly) so I assumed that the default value of dtlimit = 1.5 was correct. If it's not, what should I change it to?

Thanks,

James

 

[oleson]

For transient runs like this (with mct which is what I assume you are running with) we usually have to remove or zip the datm restart file when restarting at 1901. From your atm log it looks like it is reading the datm restart file:

i.clm5.fire_spinup_test_histC20_001.datm.rs1.1901-01-01-00000.bin

Try removing that or zipping it so that the model doesn't read it.

 

[James King]

Thanks Keith, it looks like that fixed the problem.