DART Manhattan Differences from Lanai Release Notes


This document includes an overview of the changes in the DART system since the Lanai release. For further details on any of these items look at the HTML documentation for that specific part of the system.

The two most significant changes in the Manhattan version of DART are it can support running models with a state vector larger than the memory of a single task, removing a limit from the Lanai version of DART. It also reads and writes NetCDF files directly instead of requiring a conversion from one file to another. There are many other smaller changes, detailed below.

Manhattan supported models:

  • 9var

  • bgrid_solo

  • cam-fv

  • cice

  • clm

  • cm1

  • forced_lorenz_96

  • ikeda

  • lorenz_63

  • lorenz_84

  • lorenz_96

  • lorenz_96_2scale

  • lorenz_04

  • mpas_atm (NetCDF overwrite not supported for update_u_from_reconstruct = .true. )

  • null_model

  • POP

  • ROMS

  • simple_advection

  • wrf

If your model of interest is not on the list consider checking out the ‘Classic’ release of DART, which is Lanai plus bug fixes and minor enhancements. All models previously supported by Lanai are still in DART ‘Classic’.

These are the major differences between the Lanai/Classic and Manhattan releases of DART:

  • Read and write NetCDF restarts

  • Calculation of forward operators

  • Vertical conversion of observation locations

  • Diagnostic file changes

  • State Stucture

  • model_mod interface changes

  • Observation Quantity replaces Kind

  • Perturbation of the state

NetCDF restart files

The programs filter and perfect_model_obs now read/write directly from NetCDF files rather than having to run converters (model_to_dart and dart_to_model). To facilitate this there is a new required call add_domain which must be called during static_init_model. It can be called multiple times in static_model_mod, e.g. once for each NetCDF file that contains state variables. There are three ways to add a domain:

  • From File : This is for models which have NetCDF restart files

    • dom_id = add_domain(template_file, num_vars, var_names, ... )

  • From Spec : Creates a skeleton structure for a domain ( currently only used in bgrid_solo )

    • dom_id = add_domain(num_vars, var_names, ... )

    • call add_dimension_to_variable(dom_id, var_id, dim_nam, dim_size)

    • call finished_adding_domain

  • From Blank : This is for small models such as lorenz_96 and no NetCDF restarts

    • dom_id = add_domain(model_size)

For models without NetCDF restarts, use add_domain(model_size). This is the minimum amount of information needed by DART to create a netdcf file. For models with NetCDF restarts use add_domain(info_file, num_vars, var_names) which lets DART read the NetCDF dimensions for a list of variables from a file (info_file). There are several routines that can be used together to create a domain from a description: add_domain, add_dimension_to_variable, finished_adding_domain. This can be used in models such as bgrid_solo where the model is spun up in perfect_model_obs, but the model itself has variable structure (3D variables with names). See Additions/Changes to existing namelists for how to use NetCDF IO.

Note when using NetCDF restarts, inflation files are NetCDF also. The inflation mean and inflation standard deviation are in separate files when you use NetCDF restarts. See Netcdf Inflation Files for details.

Calculation of forward operators

The forward operator code in model_mod now operates on an array of state values. See Forward Operators for more detail about distributed vs. non-distributed forward operators. In distributed mode the forward operators for all ensemble members are calculated in the same model_interpolate call. In non-distributed mode, the forward operators for all ensemble members a task owns (1-ens_size) are calculated at once.

Vertical conversion of observation and state locations

The vertical conversion of observation locations is done before the assimilation by default. This can be changed by namelist options.

In Lanai this calculation is done in the assimilation as part of get_close_obs if a model_mod does vertical conversion. Note that not all models do vertical conversion or even have a concept of vertical location, but every model_mod must have the following routines:

call set_vertical_localization_coord(vert_localization_coord)

call convert_vertical_obs(ens_handle, num, locs, loc_qtys, loc_types, &
                          which_vert, status)

call convert_vertical_state(ens_handle, num, locs, loc_qtys, loc_indx, &
                            which_vert, istatus)

If there are NOT multiple choices for a vertical coordinate (e.g. cartesian, one dimensional), all these routines can be no-ops.

If there are multiple types of vertical coordinates, the convert routines must be able to convert between them. The ‘set_vertical_localization_coord()’ routine should be called from ‘static_init_model()’ to set what localization coordinate type is being requested.

The three routines related to vertical coordinates/localization choices are:

  • set_vert_localization_coord - sets the vertical localization coordiate (not required if there is no vertical conversion)

  • convert_vertical_obs - converts observation location to required vertical type (does nothing if there is no vertical conversion)

  • convert_vertical_state - converts state vector location to required vertical type (does nothing if there is no vertical conversion)

DART diagnostic file changes

For large models DART format diagnostic files (Prior_Diag.nc and Posterior_Diag.nc) have been replaced with separate files for each copy that would have gone into Prior_Diag.nc and Posterior_Diag.nc.

For Prior_Diag.nc:

  • Mean and standard deviation: preassim_mean.nc preassim_sd.nc

  • Inflation mean and standard deviation (if state space inflation is used): preassim_priorinf_mean.nc preassim_priorinf_sd.nc

  • The number of ensemble members specifed in filter_nml (num_output_state_members): preassim_member_####.nc

For Posterior_Diag.nc:

  • Mean and standard deviation: postassim_mean.nc postassim_sd.nc

  • Inflation mean and standard deviation (if state space inflation is used): postassim_priorinf_mean.nc postassim_priorinf_sd.nc

  • The number of ensemble members specifed in filter_nml (num_output_state_members): postassim_member_####.nc

The num_output_state_members are not written separately from the restarts. Note that restarts will have been clamped if any clamping is applied (given as an arguement to add_domain). This is different to Posterior_Diag.nc which contains unclamped values. Note also that there are 2 more “stages” which might be output, in addition to the preassim and postassim discussed here.

For models with multiple domains the filenames above are appended with the domain number, e.g. preassim_mean.nc becomes preassim_mean_d01.nc, preassim_mean_d02.nc, etc.

Changes to nc_write_model_atts

nc_write_model_atts now has 2 arguments:

  • ncid - open netcdf file identifier

  • domain_id - domain number being written

The calling code will write the model state, so this routine should only add attributes and optionally, non-state information like grid arrays.

This routine will only be called if DART is creating an output NetCDF file from scratch. This may include any of the preassim, postassim, or output files.

Changes to nc_write_model_vars

nc_write_model_vars is currently unused (and in fact uncalled). It remains for possible future expansion.

Model_mod.f90 interface changes

The model_mod.f90 file contains all code that is specific to any particular model. The code in this file is highly constrained since these routines are *called by* other code in the DART system. All routine interfaces – the names, number of arguments, and the names of those arguments – must match the prescribed interfaces exactly. Since not all required interfaces are needed for every model there are default routines provided that can be referenced from a ‘use’ statement and then the routine name can be put in the module ‘public’ list without any code for that routine having to be written in the model_mod.f90 file.

The following 18 routines are required:

  • static_init_model

  • get_model_size

  • get_state_meta_data

  • shortest_time_between_assimilations

  • model_interpolate

  • end_model

  • nc_write_model_atts

  • nc_write_model_vars

  • init_time

  • init_conditions

  • adv_1step

  • pert_model_copies

  • get_close_obs

  • get_close_state

  • convert_vertical_obs

  • convert_vertical_state

  • read_model_time

  • write_model_time

Here is an example of code from the top of a model_mod file, including the modules where the default routines live and the required public list.

use     location_mod, only : location_type, get_close_type, &
                             get_close_obs, get_close_state, &
                             convert_vertical_obs, convert_vertical_state, &
                             set_location, set_location_missing, &
use    utilities_mod, only : register_module, error_handler, &
                             E_ERR, E_MSG
                             ! nmlfileunit, do_output, do_nml_file, do_nml_term,  &
                             ! find_namelist_in_file, check_namelist_read
use netcdf_utilities_mod, only : nc_add_global_attribute, nc_synchronize_file, &
                                 nc_add_global_creation_time, &
                                 nc_begin_define_mode, nc_end_define_mode
use state_structure_mod, only : add_domain
use ensemble_manager_mod, only : ensemble_type
use dart_time_io_mod, only  : read_model_time, write_model_time
use default_model_mod, only : pert_model_copies, nc_write_model_vars

implicit none

! required by DART code - will be called from filter and other
! DART executables.  interfaces to these routines are fixed and
! cannot be changed in any way.
public :: static_init_model,      &
          get_model_size,         &
          get_state_meta_data,    &
          shortest_time_between_assimilations, &
          model_interpolate,      &
          end_model,              &
          nc_write_model_atts,    &
          adv_1step,              &
          init_time,              &

! public but in another module
public :: nc_write_model_vars,    &
          pert_model_copies,      &
          get_close_obs,          &
          get_close_state,        &
          convert_vertical_obs,   &
          convert_vertical_state, &
          read_model_time,        &

Observation quantity replaces kinds

Historically there has been confusion about the terms for specific observation types (which often include the name of the instrument collecting the data) and the generic quantity that is being measured (e.g. temperature). The previous terms for these were ‘types’ and ‘kinds’, respectively.

Starting with the Manhattan release we have tried to clarify the terminology and make the interfaces consistent. The following table lists the original names from the Lanai/Classic release and the replacement routines in Manhattan.

All code that is part of the DART code repository has been updated to use the replacment routines, but if you have your own utilities written using this code, you will need to update your code. Contact us ( dart@ucar.edu ) for help if you have any questions.

public subroutines, existing name on left, replacement on right:

assimilate_this_obs_kind()     =>     assimilate_this_type_of_obs(type_index)
evaluate_this_obs_kind()       =>       evaluate_this_type_of_obs(type_index)
use_ext_prior_this_obs_kind()  =>  use_ext_prior_this_type_of_obs(type_index)

get_num_obs_kinds()      =>  get_num_types_of_obs()
get_num_raw_obs_kinds()  =>  get_num_quantities()

get_obs_kind_index()     => get_index_for_type_of_obs(type_name)
get_obs_kind_name()      => get_name_for_type_of_obs(type_index)

get_raw_obs_kind_index()  =>  get_index_for_quantity(quant_name)
get_raw_obs_kind_name()   =>  get_name_for_quantity(quant_index)

get_obs_kind_var_type()  =>  get_quantity_for_type_of_obs(type_index)

get_obs_kind()      =>  get_obs_def_type_of_obs(obs_def)
set_obs_def_kind()  =>  set_obs_def_type_of_obs(obs_def)

get_kind_from_menu()      =>  get_type_of_obs_from_menu()

read_obs_kind()     =>   read_type_of_obs_table(file_unit, file_format)
write_obs_kind()    =>  write_type_of_obs_table(file_unit, file_format)

maps obs_seq nums to specific type nums, only used in read_obs_seq:
map_def_index()  => map_type_of_obs_table()

removed.  apparently unused, and simply calls get_obs_kind_name():

apparently unused anywhere, removed:

public integer parameter constants and subroutine formal argument names,
old on left, new on right:

kind => quantity

type => type_of_obs

integer parameters:
max_obs_generic  =>  max_defined_quantities  (not currently public, leave private)
max_obs_kinds    =>  max_defined_types_of_obs

Additions/changes to existing namelists


These namelist options used to be in filter_nml, now they are in quality_control_nml.

   input_qc_threshold          = 3,
   outlier_threshold           = 4,
   enable_special_outlier_code = .false.

New namelist variables


   single_file_in               = .false.,
   single_file_out              = .false.,

   input_state_file_list        = 'null',
   output_state_file_list       = 'null',
   input_state_files            = 'null',
   output_state_files           = 'null',

   stages_to_write              = 'output'
   write_all_stages_at_end      = .false.
   output_restarts              = .true.
   output_mean                  = .true.
   output_sd                    = .true.

   perturb_from_single_instance = .false.,
   perturbation_amplitude       = 0.2_r8,

   distributed_state            = .true.






True means that all of the restart and inflation information is read from a single NetCDF file. False means that you must specify an input_state_file_list and DART will be expecting input_{priorinf,postinf}_{mean,sd}.nc files for inflation.



True means that all of the restart and inflation information is written to a single NetCDF file. False means that you must specify a output_state_files and DART will be output files specified in the list. Inflation files will be written in the form input_{priorinf,postinf}_{mean,sd}.nc.


character array

This is used for single file input for low order models. For multiple domains you can specify a file for each domain. When specifying a list single_file_in, single_file_out must be set to .true.


character array

This is used for single file input for low order models. For multiple domains you can specify a file for each domain. When specifying a list single_file_in, single_file_out must be set to .true.


character array

A list of files containing input model restarts. For multiple domains you can specify a file for each domain. When specifying a list single_file_in, single_file_out must be set to .false.


character array

A list of files containing output model restarts. For multiple domains you can specify a file for each domain. When specifying a list single_file_in, single_file_out must be set to .false.


character array

Controls which stages to write. Case-insensitive input. Currently there are six options:

  • input – writes input mean and sd only

  • forecast – before assimilation, before prior inflation is applied

  • preassim – before assimilation, before prior inflation is applied

  • postassim – after assimilation, before posterior inflation is applied

  • analysis – after assimilation, after posterior inflation is applied

  • output – final output from filter which includes clamping and inflation



True means output all stages at the end of filter. This is more memory intensive but requires less time. For larger models IO begins to dominate the overall cost of the assimilation, so writting all stages at the end writes more files in parallel, reducing the IO time. Filenames are defined in output_state_files.



True means output a restart file(s). Filenames are defined in output_state_files.



True means output a restart file which contains the ensemble mean for the stages that have been turned on in stages_to_write. The file name will have the stage with _mean appended.



True means output a restart file which contains the ensemble standard deviation for the stages that have been turned on in stages_to_write. The file name will have the stage with _sd appended.



Read a single file and perturb this to create an ensemble



Perturbation amplitude



True keeps the state distributed across all tasks throughout the entire execution of filter.

NetCDF reads and writes:

For input file names:

  • give input_state_file_list a file for each domain, each of which contains a list of restart files. An example of an ‘input_list.txt’ might look something like :

  • if no input_state_file_list is provided then default filenames will be used e.g. input_member_####.nc, input_priorinf_mean.nc, input_priorinf_sd.nc

For output file names:

  • give output_state_file_list a file for each domain, each of which contains a list of restart files. An example of an ‘input_list.txt’ might for WRF might look something like :

    if you would like to simply like to overwrite your previous data input_list.txt = output_list.txt
  • if no output_state_files is provided then default filenames will be used e.g. output_member_####.nc, output_priorinf_mean.nc, output_priorinf_sd.nc

For small models you may want to use single_file_in, single_file_out which contains all copies needed to run filter.


   buffer_state_io          = .false.,
   single_precision_output  = .false.,

When buffer_state_io is .false. the entire state is read into memory at once if .true. variables are read one at a time. If your model can not fit into memory at once this must be set to .true. .

single_precision_output allows you to run filter in double precision but write NetCDF files in single presision


   distribute_mean  = .true.

In previous DART releases, each processor gets a copy of the mean (in ens_mean_for_model). In RMA DART, the mean is distributed across all processors. However, a user can choose to have a copy of the mean on each processor by setting distribute_mean = .false. . Note that the mean state is accessed through get_state whether distribute_mean is .true. or .false.

Removed from existing namelists

   input_qc_threshold          = 3,
   outlier_threshold           = 4,
   enable_special_outlier_code = .false.
   start_from_restart          = .false.
   output_inflation            = .true.
   output_restart              = .true.

NOTE : output_restart has been renamed to output_restarts. ``output_inflation`` is no longer supported and only writes inflation files if inf_flavor > 1

   single_restart_file_out = .true.
   perturbation_amplitude  = 0.2,
   write_binary_restart_files = .true.,
   netCDF_large_file_support  = .false.


The option to perturb one ensemble member to produce an ensemble is in filter_nml:perturb_from_single_instance. The model_mod interface is now pert_model_copies not pert_model_state. Each task perturbs every ensemble member for its own subsection of state. This is more complicated than the Lanai routine pert_model_state, where a whole state vector is available. If a model_mod does not provide a perturb interface, filter will do the perturbing with an amplitude set in filter_nml:perturbation_amplitude. Note the perturb namelist options have been removed from ensemble_manager_nml