Clm45BgcCrop case dies due to negative leafn

[billsacks]

Bill Sacks < sacks@ucar.edu > - 2017-07-06 13:21:37 -0600
Bugzilla Id: 2489
Bugzilla CC: andre@ucar.edu, dlawren@ucar.edu, dll@ucar.edu, oleson@ucar.edu, rfisher@ucar.edu,

In the compsetchanges branch (soon to be clm4_5_16_r249), this test is dying:

ERP_D.f10_f10_musgs.IHistClm50Bgc.yellowstone_gnu.clm-decStart

with message:

28: ERROR: Carbon or Nitrogen patch negative = 30.5688132018826
28: -6.01119011968369
28: ERROR: limits = -60.0000000000000 -6.00000000000000
28: ENDRUN:
28: ERROR: carbon or nitrogen state critically negative ERROR in CNPrecisionControl
28: Mod.F90 at line 207

This is from this call:

  ! leaf C and N
  call TruncateCandNStates( bounds, filter_soilp, num_soilp, cs%leafc_patch(bounds%begp:bounds%endp), &
                            ns%leafn_patch(bounds%begp:bounds%endp), &
                            pc(bounds%begp:), pn(bounds%begp:), __LINE__, &
                            c13=c13cs%leafc_patch, c14=c14cs%leafc_patch, &
                            pc13=pc13(bounds%begp:), pc14=pc14(bounds%begp:) )

which I'm pretty sure means that the problem is in leafn_patch.

[billsacks]

Bill Sacks < sacks@ucar.edu > - 2017-07-06 13:26:36 -0600

Oops, I copied the wrong test name. The failure is actually in:

ERS_Ly5_P72x1.f10_f10_musgs.IHistClm45BgcCrop.cheyenne_intel.clm-cropMonthOutput

[billsacks]

Bill Sacks < sacks@ucar.edu > - 2017-07-06 13:28:13 -0600

This test was newly added in this branch, in order to maintain a long production test of CLM45 in the test list.

[billsacks]

Erik Kluzek < erik@ucar.edu > - 2017-09-08 15:23:53 -0600

Danica and Keith found they needed to do the following to get simulations to run:

It sets the cnegcrit to -600 and nnegcrit to -60

That's much larger than we are comfortable with. But, this has been how the model has worked for some time, we just didn't notice it as we didn't check for negative values for C and N variables.

[billsacks]

Erik Kluzek < erik@ucar.edu > - 2017-09-19 10:53:26 -0600

Fixed in clm4_5_16_r255

[billsacks]

Bill Sacks < sacks@ucar.edu > - 2017-10-25 07:45:58 -0600

This test is still failing in my branch off of r260, seemingly for the same reason: the death is from this call:

  call TruncateCandNStates( bounds, filter_soilp, num_soilp, cs%leafc_patch(bounds%begp:bounds%endp), &
                            ns%leafn_patch(bounds%begp:bounds%endp), &
                            pc(bounds%begp:), pn(bounds%begp:), __LINE__, &
                            c13=c13cs%leafc_patch, c14=c14cs%leafc_patch, &
                            pc13=pc13(bounds%begp:), pc14=pc14(bounds%begp:) )

[billsacks]

Bill Sacks < sacks@ucar.edu > - 2017-10-25 07:46:59 -0600

It looks like the proposed changes didn't make it into the code:

real(r8), public :: ccrit = 1.e-8_r8 ! critical carbon state value for truncation (gC/m2)
real(r8), public :: cnegcrit = -6.e+1_r8 ! critical negative carbon state value for abort (gC/m2)
real(r8), public :: ncrit = 1.e-8_r8 ! critical nitrogen state value for truncation (gN/m2)
real(r8), public :: nnegcrit = -6.e+0_r8 ! critical negative nitrogen state value for abort (gN/m2)

[ekluzek]

It's actually set in the namelist defaults.

<ncrit    phys="clm5_0" use_cn=".true.">1.d-9</ncrit>
<cnegcrit phys="clm5_0" use_cn=".true.">-6.d+1</cnegcrit>
<nnegcrit phys="clm5_0" use_cn=".true.">-6.d+0</nnegcrit>
<cnegcrit phys="clm4_5" use_cn=".true.">-6.d+2</cnegcrit>
<nnegcrit phys="clm4_5" use_cn=".true.">-6.d+1</nnegcrit>

But it looks like it doesn't actually make it into the actual namelist for some reason, even though as far as I can tell it should.

[ekluzek]

These are the changes needed to get the namelist working correctly:

diff --git a/bld/CLMBuildNamelist.pm b/bld/CLMBuildNamelist.pm
index a831476..626a0bc 100755
--- a/bld/CLMBuildNamelist.pm
+++ b/bld/CLMBuildNamelist.pm
@@ -2105,7 +2105,7 @@ sub setup_logic_cnfire {
 sub setup_logic_cnprec {
   my ($opts, $nl_flags, $definition, $defaults, $nl, $physv) = @_;
 
-  if ( $physv->as_long() >= $physv->as_long("clm5_0") && &value_is_true($nl->get_value('use_cn')) ) {
+  if ( $physv->as_long() >= $physv->as_long("clm4_5") && &value_is_true($nl_flags->{'use_cn'}) ) {
      add_default($opts, $nl_flags->{'inputdata_rootdir'}, $definition, $defaults,
                  $nl, 'ncrit', 'use_cn'=>$nl_flags->{'use_cn'});
      add_default($opts, $nl_flags->{'inputdata_rootdir'}, $definition, $defaults,
diff --git a/bld/namelist_files/namelist_defaults_clm4_5.xml b/bld/namelist_files/namelist_defaults_clm4_5.xml
index f78b54e..325c358 100644
--- a/bld/namelist_files/namelist_defaults_clm4_5.xml
+++ b/bld/namelist_files/namelist_defaults_clm4_5.xml
@@ -434,6 +434,7 @@ attributes from the config_cache.xml file (with keys converted to upper-case).
 <use_nguardrail phys="clm5_0" use_cn=".true." >.true.</use_nguardrail>
 
 <ncrit    phys="clm5_0" use_cn=".true.">1.d-9</ncrit>
+<ncrit    phys="clm4_5" use_cn=".true.">1.d-8</ncrit>
 <cnegcrit phys="clm5_0" use_cn=".true.">-6.d+1</cnegcrit>
 <nnegcrit phys="clm5_0" use_cn=".true.">-6.d+0</nnegcrit>
 <cnegcrit phys="clm4_5" use_cn=".true.">-6.d+2</cnegcrit>

[billsacks]

@ekluzek this is fixed due to #333 and so should be closed, right?

[ekluzek]

Fixed in clm5.0.dev0004