3.5.1 Interfaces for user-defined code

The following table describes the available interfaces for user-defined code in the model, the names and possible arguments of the subroutines involved as well as actions which can be accomplished in these subroutines.
 

Call location

Subroutine name and argument

Possible actions

---MODULE userAll user-defined variables which are used outside the respective local scopes of the user-defined subroutines have to be declared here.

End of parin.f90

user_parin

Declaration of user-defined parameters in the NAMELIST group &userpar. These parameters can be read from the local file PARIN and be used for steering the user-defined code (see chapter 4.0). The user-defined parameter region is already pre-defined in &userpar and must not be changed. Additional NAMELIST groups may be declared if desired. 
The CPU time needed for executing user_parin is included in the local file CPU_MEASURES within the category “initialisation”.

End of header.f90

user_header( io )

Output of the values of the user-defined parameters to the local files HEADER and RUN_CONTROL. This output should be made anyway in order to subsequently check the parameter values used for the respective run. The number of the output unit is passed as an argument and must be used in each WRITE statement (e.g.: “WRITE ( io,…”)). By default, this subroutine only creates output if a user-defined NAMELIST is found in file PARIN (in this case the variable user_defined_namelist_found is set  .TRUE. in user_parin). The names of any user-defined subdomains are displayed by default (see statistic_regions).

End of init_3d_model.f90

user_init

Here, user-defined initialization actions which should be accomplished before the start of the 3D-model can be implemented (e.g. inhomogenous change of the surface temperature, the roughness length, or similar). At this point, all default parameter initializations of the model (chapter 4.1) have already been carried out at all. 

One of the most important actions here is the definition of user-defined subdomains for statistic analysis and output (see statistic_regions and chapter 3.5.3) and of additional time series quantities (see comment line example in the file). 
CPU time for user_init is considered in the local file CPU_MEASURES within the category “initialisation”.

See also user_init_3d_model.

DEFAULT case of SELECT CASE( TRIM(canopy_mode) ) command in the middle of init_3d_model.f90user_init_plant_canopyUser-defined initilisation of the plant canopy model.

Here, the user can define the two three-dimensional arrays lad_s and cdc that stand for the leaf area density and the canopy drag coefficient. By this, the user can configure a plant canopy. This requires definition of at least one new canopy_mode in the SELECT CASE( TRIM(canopy_mode) ) command in user_init_plant_canopy. Note that the canopy_mode 'block' is not allowed here since it is the standard case used in the calling routine init_3d_model.
DEFAULT case of SELECT CASE( TRIM(topography) ) command in the middle of init_grid.f90 user_init_grid

Execution of user-defined grid initializing actions.

Here, the user can define the two-dimensional index array nzb_local that is used to set up a non-flat topography. This requires definition of at least one new topography mode in the SELECT CASE( TRIM(topography) ) command in user_init_grid. Note that the topography modes 'flat' and 'single_building' are not allowed here since these are the standard cases used in the calling routine init_grid.

CPU time for user_init_grid is considered in the local file CPU_MEASURES within the category “initialisation”.

In the middle of init_3d_model.f90user_init_3d_
model
Initialization of the 3d-model.

Allows the complete initialization of the 3d model. The user is responsible to set at least all those quantities which are normally set within init_3d_model.

In the middle of init_particles.f90

user_init_
particles

With this routine e.g. size and color of particles, which have been released initially from a source, can be specified. Like the calling routine init_particles, this routine is only called if particle advection is switched on (see package parameter dt_prel).

End of advec_particles.f90

user_particle_
attributes

Here the particles can be assigned new attributes (diameter, color etc.) after each time step. Like the calling routine advec_particles, this routine is only called if particle advection is switched on (see package parameter dt_prel).

Before the plot of particles and cross-sections in
dvrp_plot.f90

user_dvrp_coltab ( mode )

With this routine, color tables can be adjusted which should be used for output with the dvrp software. Individuell color tables for particles and cross-sections are possible. For this purpose the two arrays interval_values_dvrp and interval_h_dvrp (declared in the module dvrp_variables) have to be preset with appropriate values. The number of color table entries must be assigned to the variable of dvrp_colourtable_entries. The value of the subroutine argument mode defines, whether the color table is valid for particles or for cross-sections (i.e. mode can have the values “particles” or “slicer”). 
This routine only becomes active if dvrp-graphics is switched on (see package parameter dt_dvrp).

Beginning and end of the time integration loop of 
time_integration.f90 as well as after each prognostic equation in
prognostic_equations.f90

user_actions( location )

or

user_actions( i, j, location )

In this routine user actions can be implemented which are to be executed either at each time step or at certain times (defined by the user). Since in each case the routine is called at the beginning and at the end of a time step as well as after each prognostic equation it must be defined at which place (and/or at which places) the respective actions are supposed to be executed. For this purpose an appropriate CASE structure is already inserted in this routine. The location of the calling routine is passed to this routine via the argument location. For example, if called at the beginning of a time step, location'before_timestep' and if called at the end, location 'after_timestep'.

Calculation of user-defined output quantities should be carried out at location = 'after_integration'.

Important: if the cache-optimized version of prognostic_equation.f90 is used (this is the default for IBM-Regatta-Systems), the loop variables i and j must be passed as arguments for all calls within prognostic_equations.

CPU time for user_actions appears within the category “user_actions” in the local file CPU_MEASURES. If the cache-optimized version of prognostic_equation.f90 is used, this time measurement only considers the CPU time which was needed outside of the prognostic equations. Time needed for calls within prognostic_equations is considered within the category "all prog.equations".

Important: this subroutine is written as a FORTRAN module, which uses so-called function overloading. It can be called in two ways: with one argument (location) and with three arguments (i, j, location). The three argument version is used in the cache-optimized version of prognostic_equations. Depending on the number of given arguments the subroutine internally calls user_actions or user_actions_ij. So the user must insert the actions (code extensions) desired for the respective prognostic equation either in the CASE structure of user_actions, or/and in the CASE structure of user_actions_ij!

flow_statistics.f90 before the summation of array hom (horizontal averages)

user_statistics
( sr )

Horizontal averages of vertical profiles of user-defined quantities can be computed here. Also, additional time series quantities can be calculated. They have to be defined before in routine user_init. The routine is called once for each defined statistic region (see statistic_region).

The routine contains some simple examples (as comment lines) in order to demonstrate how to use it.

Profile sums have to be stored in the array sums_l. The profile identification number (second index of array sums_l) must be within the range [ pr_palm+1 , pr_palm+max_pr_user ], where pr_palm is an internal parameter and  max_pr_user is the number of user-defined profiles as given by parameter data_output_pr_user in the respective PALM run.

These additional profiles are further processed (e.g. temporally averaged), like the standard profiles, in flow_statistics.

End of check_parameters.f90 where 2d/3d output quantities are checkeduser_check_data_
output( variable, unit )
Here the physical unit(s) of user-defined 2d/3d output quantities (see data_output_user) have to be set. Input parameter variable contains the string identifier of the respective variable. For those variables not recognized by the user, the parameter unit is set to "illegal", which tells the calling routine that the output variable is not defined and leads to a program abort.

See chapter 3.5.4 about creating user-defined output quantities.
End of check_parameters.f90 where vertical profile output quantities are checked
user_check_data_
output_pr(
variable, var_count, unit )
Here the physical unit(s) and vertical grid (u- or w-grid) of user-defined vertical profile output quantities (see data_output_pr_user) have to be set. Input parameter variable contains the string identifier of the respective variable. Parameter var_count contains the internal profile number and must not be changed by the user. For those variables not recognized by the user, the parameter unit is set to "illegal", which tells the calling routine that the output variable is not defined and leads to a program abort.

See chapter 3.5.4 about creating user-defined output quantities.
netcdf.f90 (4 times) where the vertical grid coordinates for cross sections and volume data are defineduser_define_
netcdf_grid( variable, found, grid_x, grid_y, grid_z )
Set the grid on which user-defined output quantities (see data_output_user) are defined. Input parameter variable contains the string identifier of the respective variable. Allowed values are 'x' and 'xu' for grid_x, 'y' and 'yv' for grid_y, and 'zu' and 'zw' for grid_z. The output parameter found has to be set .TRUE. by the user, otherwise an error message will appear in the job protocol for the respective output quantity.

Example:
If an output quantity is defined at the center of the grid boxes, the following settings have to be used:

grid_x = 'x'
grid_y = 'y'
grid_z = 'zu'
found  = .TRUE.

For defining an output quantity at the grid points where the u-velocity-component is defined, please use

grid_x = 'xu'
grid_y = 'y'
grid_z = 'zu'
found  = .TRUE.

See chapter 3.5.4 about creating user-defined output quantities.
Middle of data_output_2d.f90user_data_
output_2d( av, variable, found, grid, local_pf )
Resorts user-defined quantities (to be output as cross-section data; see data_output_user) with indices (k,j,i) to a temporary array local_pf with indices (i,j,k) and sets the grid on which they are defined. Depending on the value of input parameter av resorting is done for instantaneous (av=0) or time averaged data (av=1). Input parameter variable contains the string identifier of the respective variable. Allowed values for grid are 'zu' and 'zw'. The output parameter found has to be set .TRUE. by the user, otherwise an error message will appear in the job protocol for the respective output quantity.

See chapter 3.5.4 about creating user-defined output quantities.
data_output_3d.f90 at the end of the CASE structureuser_data_
output_3d( av, variable, found, local_pf, nz_do )
Resorts user-defined quantities (to be output as volume data; see data_output_user) with indices (k,j,i) to a temporary array local_pf with indices (i,j,k) and sets the grid on which they are defined. Depending on the value of input parameter av resorting is done for instantaneous (av=0) or time averaged data (av=1). Input parameter variable contains the string identifier of the respective variable. The input parameter nz_do defines the upper limit of vertical grid index k of the output array (see nz_do3d). The output parameter found has to be set .TRUE. by the user, otherwise an error message will appear in the job protocol for the respective output quantity.

See chapter 3.5.4 about creating user-defined output quantities.
End of average_3d_data.f90, middle and end of sum_up_3d_data.f90user_3d_data_
averaging( mode, variable )
Sum up and time-average of user-defined output quantities (see data_output_user) as well as allocation of the arrays necessary for storing the respective averages. Input parameter variable contains the string identifier of the respective variable. Depending on the requested action, input parameter mode has the value 'allocate', 'sum', or 'average'.

See chapter 3.5.4 about creating user-defined output quantities.
DEFAULT case of SELECT CASE( output_variable ) command near the end of data_output_dvrp.f90user_data_
output_dvrp(
output_variable, local_pf )
Resorts user-defined quantities (to be output as dvrp objects; see data_output_user and mode_dvrp) with indices (k,j,i) to a temporary array local_pf with indices (i,j,k). The array local_pf is then handed back to the calling subroutine data_output_dvrp.f90.

See chapter 3.5.4 about creating user-defined output quantities.

End of palm.f90

user_last_
action

User-defined actions which are to be executed at the end of a model run. When user_last_actions is called all model actions are already completed, but the files opened by the model are not closed yet and possibly terminating actions for these files (see subroutine of close_files) are also not carried out yet. 
CPU time for user_last_actions appears in the local file CPU_MEASURES within the category “last actions” (the time necessary for close_files is included).



Last change: $Id: chapter_3.5.1.html 138 2007-11-28 10:03:58Z raasch $