Matt Long (Jun 16 2020 at 13:17): Matt Long (Jun 16 2020 at 13:17):@Michael Levy
I have some forcing files in hand that I think I am happy with.
feventflux_input%filename = '/glade/work/mclong/cesm_inputdata/feventflux_5gmol_POP_tx0.1v3.c200616.nc'fesedflux_input%filename = '/glade/work/mclong/cesm_inputdata/fesedflux_total_reduce_oxic_POP_tx0.1v3.c200616.nc'riv_flux_shr_stream_file = '/glade/work/mclong/cesm_inputdata/riv_nut.gnews_gnm.JRA025m_to_tx0.1v3_nnsm_e333r100_190226.20200616.nc' Michael Levy (Jun 16 2020 at 13:25):@Matt Long Awesome, I'll get them into inputdata and fire off a run. Thanks!
Michael Levy (Jun 16 2020 at 14:19):I'm getting NaNs in iron, possibly from the sediment flux file? My test ran in /glade/scratch/mlevy/SMS_Ld1.TL319_t13.G1850ECOIAF_JRA_HR.cheyenne_intel.pop-highres_JRA_cice.022/run and caseroot is /glade/work/mlevy/codes/CESM/cesm2_2_beta04+GECO_JRA_HR/tests/SMS_Ld1.TL319_t13.G1850ECOIAF_JRA_HR.cheyenne_intel.pop-highres_JRA_cice.022:
4099: open_ocean_volume_2_km 1.361132666438881E+24 4099: rf_volume_2_km 2.357816520390610E+24 4099: ecosys_driver:ecosys_driver_set_interior: NaN in dtracer_module, (i,j,k)=( 4099: 2726 , 1889 , 1 ) 4099: (lon,lat)=( 163.117107990667 , 62.2284853109204 ) ... 4099: Fe 9.921742583549502E-004 NaN ... 4099: Iron Sediment Flux 4099: NaN 4099: delta_z = 1000.00000000000 4099:------------------------------------------------------------------------ 4099: 4099:POP aborting... 4099: Stopping in ecosys_driver:ecosys_driver_set_interior 4099: 4099:------------------------------------------------------------------------
Michael Levy (Jun 16 2020 at 14:19):It's happening at several points, and as far as I can tell they are all at the surface
Matt Long (Jun 16 2020 at 15:10):@Michael Levy, I'll fix this. Thanks for checking.
Michael Levy (Jun 16 2020 at 15:10):Actually, I wonder if the problem is the land mask in feventflux_5gmol_POP_tx0.1v3.c200616.nc; it looks like it's 38% land at the surface, which seems high:
>>> import xarray as xr
>>> import numpy as np
>>> ds = xr.open_dataset('feventflux_5gmol_POP_tx0.1v3.c200616.nc')
>>> np.sum(np.isnan(ds.isel(z_t=0)['FESEDFLUXIN']).data)/(3600*2400)
0.3825826388888889
Matt Long (Jun 16 2020 at 15:14):I think that's just the land mask:
Fe vent flux
Matt Long (Jun 16 2020 at 15:15):I am creating a new file right now
Matt Long (Jun 16 2020 at 15:23):Something is wrong...I am applying fillna and everything is reasonable until I write out the dataset and read it back in. Somehow I am getting -Inf's.... I'll sort it out...hopefully quickly.
Matt Long (Jun 16 2020 at 15:41):Ok, @Michael Levy, this file
/glade/work/mclong/cesm_inputdata/feventflux_5gmol_POP_tx0.1v3.c200616.nc
should now be correct.
Michael Levy (Jun 16 2020 at 15:41):great, trying it now
Michael Levy (Jun 16 2020 at 16:15):Tests with the new file run successfully, thanks for updating it!
Matt Long (Jun 16 2020 at 16:15):thanks, we may still need to change these files...
Matt Long (Jun 16 2020 at 17:44):@Michael Levy, thinking more about this, I would like to have an opportunity to revisit the fesedflux field to ensure comparability with what we use for the 1° spin-up.
Michael Levy (Jun 16 2020 at 17:49):@Matt Long that makes sense. I want to do a longer run to make sure we're happy with all of the inputdata files (ndep, initial conditions, fesedflux, feventflux, riv_nut), but I'm going to wait until I have updated tunings. For starters, I want to do that check with what we are considering compset defaults (no cocco, OMIP initial conditions)... but then I'll repeat for the experiment setup (with cocco, initial conditions interpolated from the 1 degree spin-up)
Michael Levy (Jun 16 2020 at 17:50):I should add that I'm open to suggestions on how long the "longer run" should be. 1 year? 2? 5? I'm fairly confident that the 1-day tests I'm currently doing are insufficient :)
Last updated: May 16 2025 at 17:14 UTC