MODULE filter_mod

Overview

Main module for driving ensemble filter assimilations. Used by filter.f90, perfect_model_obs.f90, model_mod_check.f90, and a variety of test programs. See the PROGRAM filter for a general description of filter capabilities and controls.

filter_mod is a Fortran 90 module, and provides a large number of options for controlling execution behavior and parameter configuration that are driven from its namelist. See the namelist section below for more details. The number of assimilation steps to be done is controlled by the input observation sequence and by the time-stepping capabilities of the model being used in the assimilation.

See Welcome to the Data Assimilation Research Testbed for more documentation, including a discussion of the capabilities of the assimilation system, a diagram of the entire execution cycle, the options and features.

Namelist

This namelist is read from the file input.nml. Namelists start with an ampersand ‘&’ and terminate with a slash ‘/’. Character strings that contain a ‘/’ must be enclosed in quotes to prevent them from prematurely terminating the namelist.

&filter_nml
   single_file_in               = .false.,
   input_state_files            = '',
   input_state_file_list        = '',
   init_time_days               = 0,
   init_time_seconds            = 0,
   perturb_from_single_instance = .false.,
   perturbation_amplitude       = 0.2,

   stages_to_write              = 'output'

   single_file_out              = .false.,
   output_state_files           = '',
   output_state_file_list       = '',
   output_interval              = 1,
   output_members               = .true.,
   num_output_state_members     = 0,
   output_mean                  = .true.,
   output_sd                    = .true.,
   write_all_stages_at_end      = .false.,
   compute_posterior            = .true.

   ens_size                     = 20,
   num_groups                   = 1,
   distributed_state            = .true.,

   async                        = 0,
   adv_ens_command              = "./advance_model.csh",
   tasks_per_model_advance      = 1,

   obs_sequence_in_name         = "obs_seq.out",
   obs_sequence_out_name        = "obs_seq.final",
   num_output_obs_members       = 0,
   first_obs_days               = -1,
   first_obs_seconds            = -1,
   last_obs_days                = -1,
   last_obs_seconds             = -1,
   obs_window_days              = -1,
   obs_window_seconds           = -1,

   inf_flavor                   = 0,                       0,
   inf_initial_from_restart     = .false.,                 .false.,
   inf_sd_initial_from_restart  = .false.,                 .false.,
   inf_deterministic            = .true.,                  .true.,
   inf_initial                  = 1.0,                     1.0,
   inf_lower_bound              = 1.0,                     1.0,
   inf_upper_bound              = 1000000.0,               1000000.0,
   inf_damping                  = 1.0,                     1.0,
   inf_sd_initial               = 0.0,                     0.0,
   inf_sd_lower_bound           = 0.0,                     0.0,
   inf_sd_max_change            = 1.05,                    1.05,

   trace_execution              = .false.,
   output_timestamps            = .false.,
   output_forward_op_errors     = .false.,
   write_obs_every_cycle        = .false.,
   allow_missing_clm            = .false.,
   silence                      = .false.,
 /

Particular options to be aware of are: ens_size, cutoff (localization radius), inflation flavor, outlier_threshold, input and output state filenames, obs_sequence_in_name, horiz_dist_only, and the binary or ascii controls for observation sequence file formats. Some of these important items are located in other namelists, but all are in the same input.nml file.

The inflation control variables are all dimensioned 2, the first value controls the prior inflation and the second controls the posterior inflation.

Item	Type	Description
single_file_in	logical	.true. means all ensemble members are read from a single netCDF file (which can only be used with subroutine-callable models). .false. means each member is in a separate file.
input_state_files	character(len=256), dimension(MAXFILES)	A list of the NetCDF files to open to read the state vectors. Models using multiple domains must put the domain and ensemble numbers in the file names. The order and format of those is to be determined. NOT SUPPORTED as of March, 2017.
input_state_file_list	character(len=256), dimension(MAXFILES)	A list of files, one per domain. Each file must be a text file containing the names of the NetCDF files to open, one per ensemble member, one per line.
init_time_days	integer	If negative, use the initial days read from the state data restart file. If positive, override the initial days read from state data restart files. Days since 1 Jan 1601.
init_time_seconds	integer	If negative use the initial seconds read from the state data restart file. If positive, override the initial seconds read from state data restart files. Seconds since midnight.
perturb_from_single_instance	logical	.true. means perturb a single state vector from one restart file to create an ensemble. This may be done by model_mod, if model_mod provides subroutine pert_model_copies. .false. means an an ensemble-sized set of restart files is provided.
perturbation_amplitude	real(r8)	Standard deviation for the noise model used when generating ensemble members. This value is available to the model_mod for use in the required interface pert_model_copies. For more, see pert_model_copies below. Ignored if perturb_from_single_instance = .false.
stages_to_write	character(len=10), dimension(6)	Controls diagnostic and restart output. Valid values are: ‘input’, ‘forecast’, ‘preassim’, ‘postassim’, ‘analysis’, ‘output’, and ‘null’. Input is case-insensitive.
single_file_out	logical	.true. means all ensemble members are written to a single netCDF file. .false. means each member is output in a separate file. Only subroutine-callable models may write a single output file.
output_state_files	character(len=256), dimension(MAXFILES)	A list of the netCDF files to open for writing updated state vectors. Not supported when using multiple domains.
output_state_file_list	character(len=256),	A list of files, one per domain. Each file must be a text file containing the names of the netCDF files to open, one per ensemble member, one per line.
output_interval	integer	Output state and observation diagnostics every ‘N’th assimilation time, N is output_interval.
output_members	logical	.true. means output the ensemble members in any stage that is enabled.
num_output_state_members	integer	Number of ensemble members to be included in the state diagnostic output for stages ‘forecast’, ‘preassim’, ‘postassim’ and ‘analysis’. output_members must be .true.
output_mean	logical	.true. means output the ensemble mean in any stage that is enabled.
output_sd	logical	.true. means output the ensemble standard deviation (spread) in any stage that is enabled.
write_all_stages_at_end	logical	For most cases this should be .false.; data will be output as it is generated for the ‘forecast’, ‘preassim’, ‘postassim’, and ‘analysis’ diagnostics, and then restart data will be output at the end. However, if I/O time dominates the runtime, setting this to .true. will store the data and it can all be written in parallel at the end of the execution. This will require slightly more memory at runtime, but can lower the job cost significantly in some cases.
compute_posterior	logical	If .false., skip computing posterior forward operators and do not write posterior values in the obs_seq.final file. Those are rarely worth examining. Saves time and memory. Posterior inflation is not possible. For backwards compatibility the default .true.
ens_size	integer	Size of ensemble.
num_groups	integer	Number of groups for hierarchical filter. It should evenly divide ens_size.
distributed_state	logical	.true. means the ensemble data is distributed across all tasks as it is read in, so a single task never has to have enough memory to store the data for an ensemble member. Large models should always set this to .true., while for small models it may be faster to set this to .false. This is different from &assim_tools_mod :: distributed_mean.
async	integer	Controls method for advancing model: 0 is subroutine call 2 is shell command 4 is mpi-job script Ignored if filter is not controlling the model advance, e.g. in CESM, WRF, etc
adv_ens_command	character(len=256)	Command sent to shell if async is 2.
tasks_per_model_advance	integer	Number of tasks to assign to each ensemble member advance.
obs_sequence_in_name	character(len=256)	File name from which to read an observation sequence.
obs_sequence_out_name	character(len=256)	File name to which to write output observation sequence.
num_output_obs_members	integer	Number of ensemble members to be included in the output observation sequence file.
first_obs_days	integer	If negative, don’t use. If non-negative, ignore all observations before this time.
first_obs_seconds	integer	If negative, don’t use. If non-negative, ignore all observations before this time.
last_obs_days	integer	If negative, don’t use. If non-negative, ignore all observations after this time.
last_obs_seconds	integer	If negative, don’t use. If non-negative, ignore all observations after this time.
obs_window_days	integer	Assimilation window days; defaults to model timestep size.
obs_window_seconds	integer	Assimilation window seconds; defaults to model timestep size.
All variables named inf_* are arrays of length 2. The first element controls the prior, the second element controls the posterior inflation. See PROGRAM filter for a discussion of inflation and effective strategies.
inf_flavor	character(len=32), dimension(2)	Inflation flavor [prior, posterior] see Inflation Options below.
inf_initial_from_restart	logical, dimension(2)	If .true., get initial mean values for inflation from inflation file. If .false. , use the corresponding namelist value inf_initial.
inf_sd_initial_from_restart	logical, dimension(2)	If .true., get initial standard deviation values for inflation from file. If .false. , use the corresponding namelist value inf_sd_initial.
inf_deterministic	logical, dimension(2)	.true. means deterministic inflation, .false. means stochastic.
inf_initial	real(r8), dimension(2)	Initial value of inflation if not read from restart file.
inf_lower_bound	real(r8), dimension(2)	Lower bound for inflation value.
inf_upper_bound	real(r8), dimension(2)	Upper bound for inflation value.
inf_damping	real(r8), dimension(2)	Damping factor for inflation mean values. The difference between the current inflation value and 1.0 is multiplied by this factor and added to 1.0 to provide the next inflation mean. The value should be between 0.0 and 1.0. Setting a value of 0.0 is full damping, which in fact turns off all inflation by fixing the inflation value at 1.0. A value of 1.0 turns inflation damping off leaving the original inflation value unchanged.
inf_sd_initial	real(r8) dimension(2)	Initial value of inflation standard deviation if not read from restart file. If ≤ 0, do not update the inflation values, so they are time-constant. If positive, the inflation values will adapt through time.
inf_sd_lower_bound	real(r8), dimension(2)	Lower bound for inflation standard deviation. If using a negative value for inf_sd_initial this should also be negative to preserve the setting.
inf_sd_max_change	real(r8), dimension(2)	For inf_flavor 5 (enhanced inflation), controls the maximum change of the inflation standard deviation when adapting for the next assimilation cycle. The value should be between 1.0 and 2.0. 1.0 prevents any changes, while 2.0 allows 100% change. For the enhanced inflation option, if the standard deviation initial value is equal to the standard deviation lower bound, the standard deviation will not adapt in time. See PROGRAM filter for a discussion of how the standard deviation adapts based on different types of inflation.
trace_execution	logical	.true. means output very detailed messages about what routines are being called in the main filter loop. Useful if a job hangs or otherwise doesn’t execute as expected.
output_timestamps	logical	.true. means write timing information to the log before and after the model advance and the observation assimilation phases.
output_forward_op_errors	logical	.true. means output errors from forward observation operators. This is the ‘istatus’ error return code from the model_interpolate routine. An ascii text file prior_forward_op_errors and/or post_forward_op_errors will be created in the current directory. For each ensemble member which returns a non-zero return code, a line will be written to this file. Each line will list the following values: the ensemble member number, local observation number, the key for the observation, and then the istatus return code. Be cautious when turning this option on. The number of lines in this file can be up to the number of observations times the number of ensemble members times the number of assimilation cycles performed. This option is generally most useful when run with a small observation sequence file and a small number of ensemble members to diagnose forward operator problems.
write_obs_every_cycle	logical	For debug use; this option can significantly slow the execution of filter. True means to write the entire output observation sequence diagnostic file each time through the main filter loop even though only observations with times up to and including the current model time will have been assimilated. Unassimilated observations have the value -888888.0 (the DART “missing value”). If filter crashes before finishing it may help to see the forward operator values of observations that have been assimilated so far.
allow_missing_clm	logical	Some models are allowed to have MISSING_R8 values in the DART state. If .true. extra caution is taken (at considerable computational cost) to allow missing values in the DART state. So far, only CLM requires this to be .true.
silence	logical	.true. means output almost no runtime messages. Not recommended for general use, but can speed long runs of the lower order models if the execution time becomes dominated by the volume of output.

Inflation Options

The value for the inf_flavor is a character string. For backwards compatiblity (it was an integer code), the specification of the integer is still supported. Inflation values (for flavors other than 0) will be time-varying only if inf_sd_initial > 0.

inflation option	description
0 ‘0’ ‘NO_INFLATION’	no inflation
2 ‘2’ ‘VARYING_SS_INFLATION’	spatially-varying state-space (gaussian)
3 ‘3’ ‘SINGLE_SS_INFLATION’	spatially-fixed state-space (gaussian)
4 ‘4’ ‘RELAXATION_TO_PRIOR_SPREAD’ ‘RTPS	Relaxation To Prior Spread (Posterior inflation only)
5 ‘5’ ‘ENHANCED_SS_INFLATION’	Enhanced spatially-varying state-space (inverse gamma). Refer to inf_sd_initial for how to set the time evolution options.

Create an initial ensemble from a single file

If the default pert_model_copies routine is used, random noise values drawn from a gaussian distribution with the standard deviation specified by perturbation_amplitude will be added to the data in a single initial ensemble member to generate the rest of the members. This option is more frequently used in the low order models and less frequently used in large models. This is in part due to the different scales of real geophysical variable values, and the resulting inconsistencies between related field values. A more successful initial condition generation strategy is to generate climatological distributions from long model runs which have internally consistent structures and values and then use observations with a ‘spin-up’ period of assimilation to shape the initial states into a set of members with enough spread and which match the current set of observations. Each model_mod is required to provide a pert_model_copies routine which can be used to either pass-through to the default routine or can be customized for that specific model.

Modules used

types_mod
obs_sequence_mod
obs_def_mod
obs_def_utilities_mod
time_manager_mod
utilities_mod
assim_model_mod
assim_tools_mod
obs_model_mod
ensemble_manager_mod
adaptive_inflate_mod
mpi_utilities_mod
random_seq_mod
state_vector_io_mod
io_filenames_mod
forward_operator_mod
quality_control_mod

Files

See the filter overview for the list of files.

Error codes and conditions

Routine	Message	Comment
filter_main	ens_size in namelist is ###: Must be > 1	Ensemble size must be at least 2.
filter_main	inf_flavor= ### Must be 0, 2, 3.	Observation Inflation is no longer supported (i.e flavor 1).
filter_main	Posterior observation space inflation (type 1) not supported.	Posterior observation space inflation doesn’t work.
filter_main	Number of processes > model size.	Number of processes can’t exceed model size for now.
filter_generate_copy_meta_data	output metadata in filter needs state ensemble size < 10000, not ###.	Only up to 10000 ensemble members with state output for now.
filter_generate_copy_meta_data	output metadata in filter needs obs ensemble size < 10000, not ###.	Only up to 10000 ensemble members with obs space output for now.
filter_setup_obs_sequence	input obs_seq file has ### qc fields; must be < 2.	Only 0 or 1 qc fields in input obs sequence for now.
get_obs_copy_index	Did not find observation copy with metadata observation.	Only 0 or 1 qc fields in input obs sequence for now.