Warning, /verification/tutorial_tracer_adjsens/DESCRIPTION is written in an unsupported language. File is not indexed.
view on githubraw file Latest commit 847dcb9b on 2006-04-20 18:28:04 UTC
847dcb9bdb Jean*0001 Setup:
0002 -----
0003
0004 The experiment is in verification/carbon/.
0005 The code/ directory has the extra and modified code to
0006 carry a passive tracer (variables: tr1, gTr1, gTr1Nm1).
0007 Additional mod's are in pkg/ctrl/
0008
0009 The tracer fields is initialised at the surface (ini_tr1.F)
0010 and then carried around by the flow field.
0011 The cost funciton is cost_tracer.F
0012 It basically accumulates all tracers reaching the surface,
0013
0014 The controls so far are
0015 initial temp./salinity and surface fluxes,
0016 also initial tracer (tr1) distribution
0017 (see adjoint/makefile).
0018
0019 I added the common block /adtr1/ to addummy_in_stepping
0020 to be able to monitor adtr1 in addition to adtheta, adsalt, ...
0021
0022 The model picks up from a 400 year spinup.
0023 In a test phase it performs a 1 year forward/adjoint run
0024 (timestep is 1 day).
0025 dumpFreq is set to 30 days, yielding adjoint fields of
0026 theta, salt, uvel, vvel, tr1
0027 gtnm1,gsnm1, gunm1, gvnm1, gtr1nm1
0028
0029 If you plot the adjoint fields, e.g. at initial timestep index 144000
0030 you will see that:
0031 ADJtheta.0000144000.001.001.data
0032 ADJsalt
0033 ADJuvel
0034 ADJvvel
0035 have a checkerboard structure, whereas
0036 ADJtr1.0000144000.001.001.data
0037 seems to look ok.
0038
0039 I had the checkerboard structure also for a simpler
0040 adjoint test case (see verification/vero/).
0041 This is basically a box setup,
0042 cost function is final temp. at one grid point.
0043 The checkerboard structure seems to disappear, if you
0044 set in "data"
0045 tAlpha = 0.
0046 and take a linear eqn. of state
0047 eosType='LINEAR'
