Error in the output profile while running JULES

1

Hi,

I am running JULES suite u-dr154. I have made changes to the output_profile (added below), to write half hourly from 2006 to 2015.

““

[namelist:jules_output]
dump_period=1
nprofiles=3
output_dir=‘$OUTPUT_FOLDER’
run_id=‘$ID_STEM’

[namelist:jules_output_profile(1)]
file_period=-2
nvars=26
output_end=‘2017-12-31 00:00:00’
output_main_run=.true.
output_period=-1
output_spinup=.true.
output_start=‘1979-01-01 00:00:00’
output_type=26*‘M’
profile_name=‘month’

var=‘esoil_gb’,‘gpp_gb’,‘smc_avail_tot’,‘t_soil’,
=‘ftl_gb’,‘fsmc_gb’,‘tstar_gb’,‘latent_heat’,‘pstar’,
=‘rad_net’,‘precip’,‘et_stom_gb’,‘sw_net’,‘lw_net’,
=‘smc_avail_top’,‘smcl’,‘fqw_gb’,‘tl1’,‘resp_p_gb’,
=‘q1p5m_gb’,‘gpp’,‘resp_p’,‘fsmc’,‘le’,‘ftl’,‘tstar’
var_name=26*‘’

[namelist:jules_output_profile(2)]
file_period=-2
nvars=24
output_end=‘2017-12-31 00:00:00’
output_main_run=.true.
output_period=-1
output_spinup=.false.
output_start=‘1979-01-01 00:00:00’
output_type=24*‘M’
profile_name=‘ilamb’
var=‘gpp_gb’,‘latent_heat’,‘runoff’,‘ftl_gb’,‘albedo_land’,
=‘sw_net’,‘sw_down’,‘lw_up’,‘lw_net’,‘rad_net’,‘t1p5m_gb’,
=‘precip’,‘lw_down’,‘npp_gb’,‘lai’,‘frac’,‘resp_p_gb’,‘zw’,
=‘canopy_gb’,‘snow_grnd_gb’,‘smc_tot’,‘q1p5m_gb’,‘pstar’,
=‘fqw_gb’
var_name=24*‘’

[namelist:jules_output_profile(3)]
file_period=-2
nvars=26
output_end=‘2015-01-01 00:00:00’
output_main_run=.true.
output_period=1800
output_spinup=.false.
output_start=‘2006-01-01 00:00:00’
output_type=26*‘L’
profile_name=‘lt’
profile_name=‘HH’
var=‘esoil_gb’,‘gpp_gb’,‘smc_avail_tot’,‘t_soil’,
=‘ftl_gb’,‘fsmc_gb’,‘tstar_gb’,‘latent_heat’,‘pstar’,
=‘rad_net’,‘precip’,‘et_stom_gb’,‘sw_net’,‘lw_net’,
=‘smc_avail_top’,‘smcl’,‘fqw_gb’,‘tl1’,‘resp_p_gb’,
=‘q1p5m_gb’,‘gpp’,‘resp_p’,‘fsmc’,‘le’,‘ftl’,‘tstar’
var_name=26*‘’
””

I get the error saying " {MPI Task 3} [FATAL ERROR] register_output_profile: Output cannot end before it has started.
Profile: HH"

The dates given for the profile HH as shown above are correct I believe. What is causing this error.

Thanks for the help in advance.

2

Hi Narender

I can’t see anything obviously wrong with the namelists. Can you point me towards your cylc-run directory? I’ll see if I can work out what is going on.

David

3

Hi David,
Here is the directory for the current simulation: “/home/users/naren047/cylc-run/u-dr154/run23”

Thanks
Narender

4

Hi Narender

Looking at the log files under ~naren047/cylc-run/u-dr154/run23/log/job/19790101T0000Z/S2/06/, it appears that the root of the problem is that your main run ends on 1989-01-01, but jules_output_profile(3) doesn’t start until 2006-01-01.

David

5

Hi David,

Okay. Rebecca Oliver has shared this suite with me. I understand that the suite is to have the main runs in 10 year chunks from 1979 to 2017. The output profile 1 writes the monthly data over this period. However, I want to write half hourly data to compare with the fluxnet data. And it will require a lot of space if write all the output from 1979 to 2017. I want to only output for the period: 2007 to 2014, so that I use less space and get what I need for comparing with fluxnet observations.

What changes do I need to make in the namelist and flow.cylc to do this?

Thanks
Narender

6

Hi Narender

If you want to do separate 10-year runs instead of one long run, you will need to disable jules_output_profile(3) on those runs that don’t overlap the time interval between the profile’s output_start and output_end. Although JULES does the right thing with output time intervals that partially overlap the main run time interval, it fails if the output time interval is completely outside the run time interval (as you’ve discovered). This is arguably a bug.

The simplest way to temporarily deactivate jules_output_profile(3) is to set nprofiles=2 under [namelist:jules_output]. I believe that this results in all higher-numbered output profiles being ignored, although I haven’t tested this. If you know more about jinja2 than I do, you might be able to automate this deactivation.

Or you could just do one long run covering all of 1979 to 2017.

Good luck
David