Known issue with consistency check for tr_aero and ntr_aero

[aliceb]

The issue arises when a user sets

tr_aero = .false.

in user_nl_cice, but doesn't change anything else. This conflicts with the
assumptions built into the default specification of ntr_aero (as far as I can
tell). See below for the diagnosis. This showed up when running CICE with
bounds checking enabled. I'm guessing that there should at least be a runtime
test whether the tr_aero and ntr_aero settings are compatible.

More details,

a) The model is built with

-DNTR_AERO=3

and (from cice/bld/configure)

my $cice_cppdefs = " ... -DNTR_AERO=$ntr_aero";

From config_definition.xml:

"CICE_CONFIG_OPTS are normally set as compset variables (e.g., -ntr_aero 3)
"

and in my env_build.xml.



so nothing is being set and 3 is the default for this compset?

b) in ice_domain_size.F90 the parameters n_aero and max_ntrcr are set to

n_aero = NTR_AERO
max_ntrcr = 18

(so n_aero = 3 here).

c) user_nl_cice specifies

tr_aero = .false.

d) in ice_init.F90

if (tr_aero) then
nt_aero = ntrcr + 1
ntrcr = ntrcr + n_aero*4 !MH 2 for snow soot and 2 for ice
!MH for multiple (n_aero) aerosols
else
nt_aero = max_ntrcr
endif

So nt_aero = 18

e) in ice_mechred.F90

if (n_aero >= 1) then
do iaero=1,n_aero
msoot(m,iaero) = msoot(m,iaero) &
+ vsrdgn(ij)*(c1-fsnowrdg) &
*(trcrn(i,j,nt_aero +4*(iaero-1),n) &
+ trcrn(i,j,nt_aero+1+4*(iaero-1),n))
enddo
endif

so (n_aero = 3 >= 1) and nt_aero + 1 + 4*(iaero-1) = 18 + 1 + 0 = 19, or +4
= 23 or +8 = 27.

However, trcrn is declared as

real (kind=dbl_kind), dimension (nx_block,ny_block,max_ntrcr,ncat), &
intent(inout) :: &
trcrn ! ice tracers

so the max 3rd index is max_ntrcr == 18.

Conclusion: It appears that

-DNTR_AERO=3

and

tr_aero = .false.

are incompatible.