Changes between Initial Version and Version 1 of OMI_debugging


Ignore:
Timestamp:
Nov 23, 2020, 7:08:21 PM (5 years ago)
Author:
lnerger
Comment:

--

Legend:

Unmodified
Added
Removed
Modified

  • OMI_debugging

    v1 v1  
     1= PDAF-OMI Debugging Information =
     2
     3[[PageOutline(2-3,Contents of this page)]]
     4
     5== Overview ==
     6
     7When implementing an observation with PDAF, or when performing the very first implementation of PDAF with a new model, it is useful to check whether the inputs to the PDAF-routines are correctly used. For this porpose, PDAF-OMI provides a debugging functionality. It allows you to activate debugging output e.g. for a single local analysis domain on a single process of a complex application of a local filter like LEKSTF.
     8
     9
     10== Activating Debugging Output ==
     11
     12Debugging output is activated by the routine `PDAFomi_set_debug_flag`. In particular this call can be inserted in any routines contained in `callback_obs_pdafomi.F90`.
     13
     14For example to activate debugging in `init_dim_obs_l_pdafomi` for the local analysis domain `domain_p=10` and filter process 0, one uses
     15{{{
     16SUBROUTINE init_dim_obs_l_pdafomi(domain_p, step, dim_obs, dim_obs_l)
     17
     18  USE PDAFomi, ONLY: PDAFomi_set_debug_flag
     19  USE mod_parallel_pdaf, ONLY: mype_filter
     20
     21  ...
     22
     23  IF (domain_p==10 .AND. mype_filter==0) THEN
     24    CALL PDAFomi_set_debug_flag(domain_p)
     25  ELSE
     26    CALL PDAFomi_set_debug_flag(0)
     27  ENDIF
     28
     29  ...
     30
     31}}}
     32
     33== Understanding the Debugging Output ==
     34
     35The debugging output mainly writes information about the different variables contained in the full data type `obs_f` allocated as `thisobs` and the local type `obs_l` allocate as `thisobs_l`. For reference we list the full declaration of these types. When reading the debugging output one can check for the meaning of the variables.
     36
     37{{{
     38  TYPE obs_f
     39     ! ---- Mandatory variables to be set in INIT_DIM_OBS ----
     40     INTEGER :: doassim=0                 !< Whether to assimilate this observation type
     41     INTEGER :: disttype                  !< Type of distance computation to use for localization
     42     INTEGER :: ncoord                    !< Number of coordinates use for distance computation
     43     INTEGER, ALLOCATABLE :: id_obs_p(:,:) !< Indices of process-local observed field in state vector
     44
     45     ! ---- Optional variables - they can be set in INIT_DIM_OBS ----
     46     REAL, ALLOCATABLE :: icoeff_p(:,:)   !< Interpolation coefficients for obs. operator (optional)
     47     REAL, ALLOCATABLE :: domainsize(:)   !< Size of domain for periodicity (<=0 for no periodicity) (optional)
     48
     49     ! ---- Variables with predefined values - they can be changed in INIT_DIM_OBS  ----
     50     INTEGER :: obs_err_type=0            !< Type of observation error: (0) Gauss, (1) Laplace
     51     INTEGER :: use_global_obs=1          !< Whether to use (1) global full obs.
     52                                          !< or (0) obs. restricted to those relevant for a process domain
     53
     54     ! ----  The following variables are set in the routine PDAFomi_gather_obs ---
     55     INTEGER :: dim_obs_p                 !< number of PE-local observations
     56     INTEGER :: dim_obs_f                 !< number of full observations
     57     INTEGER :: dim_obs_g                 !< global number of observations
     58     INTEGER :: off_obs_f                 !< Offset of this observation in overall full obs. vector
     59     INTEGER :: off_obs_g                 !< Offset of this observation in overall global obs. vector
     60     INTEGER :: obsid                     !< Index of observation over all assimilated observations
     61     REAL, ALLOCATABLE :: obs_f(:)        !< Full observed field
     62     REAL, ALLOCATABLE :: ocoord_f(:,:)   !< Coordinates of full observation vector
     63     REAL, ALLOCATABLE :: ivar_obs_f(:)   !< Inverse variance of full observations
     64     INTEGER, ALLOCATABLE :: id_obs_f_lim(:) !< Indices of domain-relevant full obs. in global vector of obs.
     65  END TYPE obs_f
     66}}}
     67
     68{{{
     69  TYPE obs_l
     70     INTEGER :: dim_obs_l                 !< number of local observations
     71     INTEGER :: off_obs_l                 !< Offset of this observation in overall local obs. vector
     72     INTEGER, ALLOCATABLE :: id_obs_l(:)  !< Indices of local observations in full obs. vector
     73     REAL, ALLOCATABLE :: distance_l(:)   !< Distances of local observations
     74     REAL, ALLOCATABLE :: ivar_obs_l(:)   !< Inverse variance of local observations
     75     INTEGER :: locweight                 !< Specify localization function
     76     REAL :: lradius                      !< localization radius
     77     REAL :: sradius                      !< support radius for localization function
     78  END TYPE obs_l
     79}}}
     80
     81=== Example output ===
     82
     83For illustration we insert the above call to PDAFomi_set_debug flag into `init_dim_obs_l_pdafomi` in `tutorial/online_2D_serialmodel_omi/`. then we compile and execute the tutorial program with:
     84{{{
     85  mpirun -np 4 ./model_pdaf -dim_ens 4 -filtertype 7 -local_range 5 -locweight 2 -assim_B .true. -assim_A .false.
     86}}}
     87With these settings only observation type B is activated, which are just 3 values in the model domain
     88
     89The first lines of the debugging output looks like this:
     90{{{
     91 ++ OMI-debug set_debug_flag: mype_filter           0 activate          10
     92 ++ OMI-debug:           10 PDAFomi_init_dim_obs_l -- START
     93 ++ OMI-debug:           10    PDAFomi_init_dim_obs_l -- count local observations
     94 ++ OMI-debug init_dim_obs_l:          10   Re-init dim_obs_l=0
     95 ++ OMI-debug init_dim_obs_l:          10   coords_l   1.0000000000000000        10.000000000000000     
     96 ++ OMI-debug cnt_dim_obs_l:           10   thisobs%ncoord           2
     97 ++ OMI-debug cnt_dim_obs_l:           10   thisobs_l%lradius   5.0000000000000000     
     98 ++ OMI-debug cnt_dim_obs_l:           10   Check for observations within radius
     99 ++ OMI-debug comp_dist2:              10   compute Cartesian distance
     100 ++ OMI-debug cnt_dim_obs_l:           10   valid observation with coordinates   5.0000000000000000        8.0000000000000000     
     101 ++ OMI-debug:           10    PDAFomi_init_dim_obs_l -- initialize local observation arrays
     102 ++ OMI-debug comp_dist2:              10   compute Cartesian distance
     103 ++ OMI-debug init_dim_obs_l:          10   thisobs_l%dim_obs_l           1
     104 ++ OMI-debug init_dim_obs_l:          10   thisobs_l%id_obs_l           1
     105 ++ OMI-debug init_dim_obs_l:          10   thisobs_l%distance_l   4.4721359549995796     
     106 ++ OMI-debug:           10 PDAFomi_init_dim_obs_l -- END
     107}}}
     108
     109The first line
     110{{{
     111 ++ OMI-debug set_debug_flag: mype_filter           0 activate          10
     112}}}
     113is from PDAFomi_set_debug_flag showing that debugging is activates with value 10 (which is values fo `domain_p` specified in the call)
     114
     115The next lines are
     116{{{
     117 ++ OMI-debug:           10 PDAFomi_init_dim_obs_l -- START           
     118 ++ OMI-debug:           10    PDAFomi_init_dim_obs_l -- count local observations
     119 ++ OMI-debug init_dim_obs_l:          10   Re-init dim_obs_l=0
     120}}}
     121show that debugging output for PDAFomi_init_dim_obs_l is shown. Most routines show such a 'START' line. The second line shows that a segment of the orutine started, the counting of local observations. The third line states that dim_obs_l=0 is set. This line, as many others shows the name of the subroutine in short form without `PDAFomi` at the beginning of the line.
     122
     123The following lines show variable values
     124{{{
     125 ++ OMI-debug init_dim_obs_l:          10   coords_l   1.0000000000000000        10.000000000000000     
     126 ++ OMI-debug cnt_dim_obs_l:           10   thisobs%ncoord           2
     127 ++ OMI-debug cnt_dim_obs_l:           10   thisobs_l%lradius   5.0000000000000000     
     128}}}
     129First we see that the coordinates `coords_l` of the grid point corresponding to domain_p=10 are (1.0, 10.0). Further we have two dimensions (thisobs%ncoord=2) and the localization radius is set to 5.0.
     130In the following lines
     131{{{
     132 ++ OMI-debug cnt_dim_obs_l:           10   Check for observations within radius
     133 ++ OMI-debug comp_dist2:              10   compute Cartesian distance
     134 ++ OMI-debug cnt_dim_obs_l:           10   valid observation with coordinates   5.0000000000000000        8.0000000000000000     
     135}}}
     136the it is checked which observations lie within the distance thisobs_l%lradius=5.0 from coords_l=(1.0, 5.0). One observation with coordinates (5.0, 8.0) is found. The followign lines
     137{{{
     138 ++ OMI-debug:           10    PDAFomi_init_dim_obs_l -- initialize local observation arrays
     139 ++ OMI-debug comp_dist2:              10   compute Cartesian distance
     140 ++ OMI-debug init_dim_obs_l:          10   thisobs_l%dim_obs_l           1
     141 ++ OMI-debug init_dim_obs_l:          10   thisobs_l%id_obs_l           1
     142 ++ OMI-debug init_dim_obs_l:          10   thisobs_l%distance_l   4.4721359549995796     
     143}}}
     144show that observation arrays are initialized. A Cartesian distance is computerd. `thisobs_l%dim_obs_l` confirms that one local observation was found and it's the first element of the full observation vector (thisobs_l%id_obs_l=1) and the distance is thisobs_l%distance_l=4.4721359549995796.
     145
     146One could now compare this with the input information. Are the values of coords_l correct? Are the coordinates of the first observation (5.0, 8.0)?
     147
     148Later in the debugging output we find e.g.
     149{{{
     150 ++ OMI-debug:           10    PDAFomi_init_obs_l -- Get local vector of observations
     151 ++ OMI-debug g2l_obs:                 10   thisobs%id_obs_l           1
     152 ++ OMI-debug g2l_obs:                 10   obs_l -0.98253999999999997     
     153 ++ OMI-debug:           10    PDAFomi_init_obs_l -- Get local vector of inverse obs. variances
     154 ++ OMI-debug g2l_obs:                 10   thisobs%id_obs_l           1
     155 ++ OMI-debug g2l_obs:                 10   obs_l   4.0000000000000000     
     156}}}
     157These lines are from the internal routine `PDAFomi_init_obs_l`, which initializes the local vector of observations. (This is functionality also existing in the 'traditional' form of implementing the observation. The documention page about [wiki:init_obs_l_pdaf init_obs_l_pdaf] explains what is done here.) Important is that not only the local observation vector (a single value) is initialized, but also the inverse observation variance (4.0 in this example).
     158
     159Similarly
     160{{{
     161 ++ OMI-debug:           10 PDAFomi_g2l_obs -- START Get local observed ensemble member           1
     162 ++ OMI-debug g2l_obs:                 10   thisobs%id_obs_l           1
     163 ++ OMI-debug g2l_obs:                 10   obs_l  0.12725110911115362     
     164 ++ OMI-debug:           10 PDAFomi_g2l_obs -- END
     165}}}
     166shows the localization of ensemble member 1. This is performed in the interval routine `PDAFomi_g2l_obs` (see the page about [wiki:g2l_obs_pdaf g2l_obs_pdaf]). There is analogous output for all of the 4 ensemble states.
     167
     168