Setup:
-----

The experiment is in verification/carbon/.
The code/ directory has the extra and modified code to
carry a passive tracer (variables: tr1, gTr1, gTr1Nm1).
Additional mod's are in pkg/ctrl/

The tracer fields is initialised at the surface (ini_tr1.F)
and then carried around by the flow field.
The cost funciton is cost_tracer.F
It basically accumulates all tracers reaching the surface,

The controls so far are
initial temp./salinity and surface fluxes,
also initial tracer (tr1) distribution 
(see adjoint/makefile).

I added the common block /adtr1/ to addummy_in_stepping
to be able to monitor adtr1 in addition to adtheta, adsalt, ...

The model picks up from a 400 year spinup.
In a test phase it performs a 1 year forward/adjoint run
(timestep is 1 day).
dumpFreq is set to 30 days, yielding adjoint fields of
theta, salt, uvel, vvel, tr1
gtnm1,gsnm1, gunm1, gvnm1, gtr1nm1

If you plot the adjoint fields, e.g. at initial timestep index 144000
you will see that:
  ADJtheta.0000144000.001.001.data
  ADJsalt
  ADJuvel
  ADJvvel
have a checkerboard structure, whereas
  ADJtr1.0000144000.001.001.data
seems to look ok.

I had the checkerboard structure also for a simpler
adjoint test case (see verification/vero/).
This is basically a box setup,
cost function is final temp. at one grid point.
The checkerboard structure seems to disappear, if you
set in "data"
  tAlpha = 0.
and take a linear eqn. of state
  eosType='LINEAR'
