The DIC and ALK initial-condition files in this directory were extracted from the gx1v6 CESM ecosystem initial-condition file. The reason for replacing the ALK and DIC files is that they contained a sprinkling of unphysical values that caused computations in the dic_carbon_chem.F file to go off the rails. This problem was first encountered in the gx3v6-resolution test cases, but it also has been confirmed to be a problem at the gx1v6 resolution as well. ======================================================================= be1105en/glade/proj3/cseg/inputdata/ocn/pop/gx1v6/ic% ncks -v ALK ecosys_jan_IC_gx1v6_20090416.nc /glade/proj3/cseg/inputdata/ocn/moby/gx1v6/ic/ALK.ecosys_jan_IC_gx1v6_20090416.nc be1105en/glade/proj3/cseg/inputdata/ocn/pop/gx1v6/ic% ncks -v DIC ecosys_jan_IC_gx1v6_20090416.nc /glade/proj3/cseg/inputdata/ocn/moby/gx1v6/ic/DIC.ecosys_jan_IC_gx1v6_20090416.nc Thu Oct 20 10:49:09 MDT 2011 ======================================================================= Details of the problem are below: Date: Thu, 13 Oct 2011 14:31:31 -0600 (MDT) From: Nancy J. Norton To: Stephanie Dutkiewicz Cc: Oliver Jahn , Keith Lindsay , Gokhan Danabasoglu Subject: UPDATE: problem in dic_carbon_chem.F I wanted to update everyone with the status of the problem I was having in my CESM/MOBY prototype ( log10(hnew), hnew < 0 in dic_carbon_chem.F). It turns out that the initial conditions that I had been using had some very unphysical ALK and DIC values in various locations around the globe. At Keith's suggestion, I added the missing_value attribute to the original netcdf files from Steph, then reapplied his regridding tool. The results were noticeably better, but not perfect, and subroutine CALC_PCO2_APPROX still had problems when initialized with them. I didn't track the exact location, but it's quite possible that the problem occured in the marginal seas (Black, Caspian, etc), were values were zero. Instead, following a second suggestion from Keith, I extracted DIC and ALK initial conditions from the data file used by the CESM ecosystem model. These IC's contain data in the Arctic and in the marginal seas that have been constructed based upon observations. I added a FORTRAN trap that aborts the model upon the first instance of hnew <= 0 in CALC_PCO2_APPROX, then set up the prototype so that it was initialized with the ecosystem-model DIC and ALK fields. The problem has now disappeared -- really. I'll add a review of all IC datasets to my task list, but this is probably a subject for us to discuss as a group, in person next month. In the meantime, I'll change the default DIC and ALK dataset names to point to the ecosystem-model-based files.