[Dart-dev] DART/branches Revision: 12500

[dart at ucar.edu]

nancy at ucar.edu
2018-04-10 16:18:31 -0600 (Tue, 10 Apr 2018)
820
fix bugs:
- height table was off by a factor of 100 for pressure
- scale height vertical conversion was returning error for all surface obs
 
change behavior to match original cam model_mod:
- allow conversion to scale height even if obs pressure is larger than the
   surface pressure (e.g. obs is located below surface).  this choice should be revisited.

enhancements:
- use binary search to find enclosing levels in height and pressure columns
- rename some routines to be clearer
- allow diagnostic files to optionally omit the grid info to be as small as possible
- remove debug print lines
- replace height/pressure table with two tables from a more accurate source.
   one table top height is 86 km for cam; the other is 1000 km for waccm, waccm-x

fix nml file to be legal syntax (no array subscripts allowed)




Modified: DART/branches/recam/models/cam-fv/model_mod.f90
===================================================================
--- DART/branches/recam/models/cam-fv/model_mod.f90	2018-04-10 15:42:04 UTC (rev 12499)
+++ DART/branches/recam/models/cam-fv/model_mod.f90	2018-04-10 22:18:31 UTC (rev 12500)
@@ -8,25 +8,25 @@
 !> this is the interface between the cam-fv atmosphere model and dart.
 !> the required public interfaces and arguments cannot be changed.
 !>
+!>
+!----------------------------------------------------------------
+!>@todo FIXME: consistent directions for interpolation/fractions
+!>  still to change: we should always pass top level first, then bottom level.
+!>  fraction should be computed and interpreted as 0 = top, 1 = bottom.
+!>  in the code below, there are !x! marks on some lines.  the meanings are:
+!>   !$! - where top and bottom need to be swapped in call or subroutine
+!>         interfaces.
+!>   !*! - where fraction is used and may need the sense to be switched 
+!>         (may be using 1 = top and 0 = bottom)
+!>
 !>@todo FIXME: feedback from WACCM-X meeting
-!>  they need to build a special factor for the height conversion
-!>  it's multiplied by G, so the code in build_heights() has to
-!>  change slightly - they aren't passing in a different R array.
 !>
 !>  test if the constants can be used from types_mod.f90 instead of
-!>  having gravity and r redefined here.
+!>  having gravity and r redefined here?  PROBABLY NOT - every model
+!   uses a slightly different value so what would we put in types?
 !>
-!>  not sure about generic height table for rejecting high obs - but
-!>  it's not clear model_interpolate should be rejecting high obs
-!>  at all. suggestions are we cull them from the input obs_seq before
-!>  the run.  in that case we can take it out!
-!>
-!>  going to try to move the oxygen ion density code to a forward
-!>  operator in an obs_def_xxx_mod.f90.
-!>
 !>  test with identity obs to make sure get_close_xx() doesn't need
 !>  special code to handle them.
-!>
 !----------------------------------------------------------------
 
 module model_mod
@@ -36,7 +36,7 @@
 use      time_manager_mod,  only : set_time, time_type, set_date, &
                                    set_calendar_type, get_date
 use          location_mod,  only : location_type, set_vertical, set_location, &
-                                   get_location, write_location, &
+                                   get_location, write_location, is_vertical, &
                                    VERTISUNDEF, VERTISSURFACE, VERTISLEVEL, &
                                    VERTISPRESSURE, VERTISHEIGHT, &
                                    VERTISSCALEHEIGHT, query_location, &
@@ -46,9 +46,10 @@
                                    vertical_localization_on, get_close_type
 use         utilities_mod,  only : find_namelist_in_file, check_namelist_read, &
                                    string_to_logical, string_to_real,& 
-                                   logfileunit, do_nml_file, do_nml_term, &
+                                   nmlfileunit, do_nml_file, do_nml_term, &
                                    register_module, error_handler, &
-                                   file_exist, to_upper, E_ERR, E_MSG, array_dump
+                                   file_exist, to_upper, E_ERR, E_MSG, array_dump, &
+                                   find_enclosing_indices
 use          obs_kind_mod,  only : QTY_SURFACE_ELEVATION, QTY_PRESSURE, &
                                    QTY_GEOMETRIC_HEIGHT, QTY_VERTLEVEL, &
                                    QTY_SURFACE_PRESSURE, &
@@ -226,21 +227,40 @@
 
 ! default to localizing in pressure.  override with namelist
 integer :: vertical_localization_type = VERTISPRESSURE
+
+! commonly used numbers that we'll set in static_init_model
 real(r8) :: global_model_top 
 real(r8) :: global_ref_pressure
 integer  :: global_nlevels
 
+!>@todo FIXME ask kevin if this is small enough to handle
+! all the possible cam configurations?
+! an arbitrary value to test the model top against to see
+! if we're running cam vs waccm or waccm-x.  it changes the
+! standard atmosphere table we use to convert pressure to height, 
+! and changes the formatting of numbers in dart_log output.
+real(r8), parameter :: high_top_threshold = 0.3_r8  ! pascals
+
 ! things related to damping at the model top
 logical  :: are_damping = .false.
 real(r8) :: damp_weight = 1.0_r8
 real(r8) :: ramp_start
 type(location_type) :: ramp_start_loc 
+logical  :: discarding_high_obs = .false.
+real(r8) :: no_assim_above_height = -1.0_r8 
+real(r8) :: no_assim_above_level  = -1.0_r8 
 
-! Precompute pressure -> height map once based on 1010mb surface pressure.
+!> build a pressure/height conversion column based on a
+!> standard atmosphere.  this can only be used when we
+!> don't have a real ensemble to use, or we don't care
+!> about absolute accuracy.
+
+! Precompute pressure <-> height map once based on either a low-top or
+! high-top table depending on what the model top is.
 ! Used only to discard obs on heights above the user-defined top threshold.