Changes between Version 15 and Version 16 of AvailableOptionsforInitPDAF


Ignore:
Timestamp:
May 30, 2025, 8:55:08 AM (6 months ago)
Author:
lnerger
Comment:

--

Legend:

Unmodified
Added
Removed
Modified

  • AvailableOptionsforInitPDAF

    v15 v16  
    1 = Available options for the different filter algorithms =
     1= Available options for the different DA methods =
    22
    33
    44[[PageOutline(2-3,Contents of this page)]]
    55
    6 There are different operations for each of the filter algorithms that need to be specified in the call to `pdaf_init`. To display the available options in a compiled assimilation program, one can use run with the specification `subtype=-1`. In this case `pdaf_init` wil display the available options for the selected filter algorithm and exit with an error status of -1.
    7 
    8 Below we list the options as they are displayed using `subtype=-1`.
    9 
    10 
    11 == SEIK (filtertype=1) ==
     6|| This page documents the options for PDAF 3. [[BR]]See the [wiki:AvailableOptionsforInitPDAFuntilPDAF231 page on available options in PDAF 2.3.1] for the options valid for the ealier releases. ||
     7
     8There are different options for each of the filter algorithms that need to be specified in the call to `PDAF_init` or that canbe set using `PDAF_set_iparam` and `PDAF_set_rparam`. To display the available options in a compiled assimilation program, one can use run with the specification `subtype=-1`. In this case `PDAF_init` wil display the available options for the selected filter algorithm and exit with an error status of -1.
     9
     10The options are mainly unified over the different DA methods. Here, the nonlinear ensemble methods have somewhaat different options from those for the ensemble Kalman filter variants. Also the 3D-Var methods have a different set of options.
     11
     12== Overview of options ==
     13
     14=== Options for ensemble Kalman filters ===
     15
     16Overview of integer options for ensemble Kalman filters. The parameters are explained in the list of options as they are displayed using `subtype=-1`, which are shown further below.
     17
     18Parameters shown in italic font have changed in PDAF V3 compared to PDAF V2.3.1. The index position in PDAF V2.3.1 is shown in parenthesis.
     19
     20||= iparam =||= SEIK =||= LSEIK =||= EnKF =||= LEnKF =||= ETKF =||= LETKF =||= ESTKF =||= LESTKF =||= //ENSRF// =||
     21||= 1 =|| dim_p || dim_p || dim_p || dim_p || dim_p || dim_p || dim_p ||dim_p || dim_p
     22||= 2 =|| dim_ens || dim_ens || dim_ens || dim_ens || dim_ens || dim_ens || dim_ens || dim_ens || dim_ens ||
     23||= 3 =||       ||       || //dim_lag(5)// ||   || dim_lag || dim_lag || dim_lag || dim_lag || ||
     24||= 4 =||       ||       || //rank_ana(3)// || //rank_ana(3)// || || || || || ||
     25||= 5 =|| type_forget || type_forget ||   ||   || type_forget || type_forget || type_forget || type_forget || ||
     26||= 6 =|| type_trans  || type_trans  ||   ||   || type_trans || type_trans || type_trans || type_trans || ||
     27||= 7 =|| type_sqrt   || type_sqrt   ||   ||   ||   ||   || type_sqrt || type_sqrt       || ||
     28||= 8 =|| obs_ens || obs_ens || obs_ens || obs_ens || obs_ens || obs_ens || obs_ens || obs_ens || obs_ens ||
     29||= 9 =|| type_obs || type_obs || type_obs || type_obs || type_obs || type_obs || type_obs || type_obs || type_obs ||
     30
     31Overview of real-valued options for ensemble Kalman filters. The parameters are explained in the list of options as they are displayed using `subtype=-1`, which are shown further below
     32                                                               
     33||= rparam =||= SEIK =||= LSEIK =||= EnKF =||= LEnKF =||= ETKF =||= LETKF =||= ESTKF =||= LESTKF =||= //ENSRF// =||
     34||= 1 =|| forget ||forget || forget || forget || forget || forget || forget || forget || forget ||
     35
     36
     37
     38=== Options for nonlinear DA methods ===
     39
     40Overview of integer options for nonlinear DA methods. The parameters are explained in the list of options as they are displayed using `subtype=-1`, which are shown further below.
     41
     42Parameters shown in italic font have changed in PDAF V3 compared to PDAF V2.3.1. The index position in PDAF V2.3.1 is shown in parenthesis.
     43
     44||= iparam =||= NETF =||= LNETF =||= LKNETF =||= PF =||
     45||= 1 =|| dim_p || dim_p || dim_p || dim_p ||
     46||= 2 =|| dim_ens || dim_ens || dim_ens || dim_ens ||
     47||= 3 =|| dim_lag || dim_lag || ||
     48||= 4 =|| type_noise || type_noise || //type_hyb(7)// || type_noise
     49||= 5 =|| type_forget || type_forget || type_forget || ||
     50||= 6 =|| type_trans || type_trans || type_trans || //type_resample(3)// ||
     51||= 7 =|| type_winf  || type_winf  ||  || //type_winf(6)// ||
     52||= 8 =|| obs_ens    || obs_ens || obs_ens || obs_ens ||
     53||= 9 =|| type_obs   || type_obs || type_obs || type_obs ||
     54
     55
     56Overview of real-valued options:
     57                                       
     58||= rparam =||= NETF =||= LNETF =||= LKNETF =||= PF =||
     59||= 1 =|| forget ||forget || forget || //forget(2)// ||
     60||= 2 =|| limit_winf || limit_winf || hyb_g || //limit_winf(3)// ||
     61||= 3 =|| noise_amp || noise_amp  || hyb_k || //noise_amp(1)// ||
     62
     63
     64=== Options for 3D-Var methods ===
     65
     66Overview of integer options for 3D-Var methods. The parameters are explained in the list of options as they are displayed using `subtype=-1`, which are shown further below.
     67
     68Parameters shown in italic font have changed in PDAF V3 compared to PDAF V2.3.1. The index position in PDAF V2.3.1 is shown in parenthesis.
     69
     70||= iparam =||= 3dvar =||= en3dvar =||= hyb3dvar =||
     71||= 1 =|| dim_p || dim_p || dim_p ||
     72||= 2 =|| dim_ens || dim_ens || dim_ens ||
     73||= 3 =|| type_opt || type_opt || type_opt ||
     74||= 4 =|| dim_cvec ||   || dim_cvec ||
     75||= 5 =||    || dim_cvec_ens || dim_cvec_ens ||
     76||= 6 =|| solver param1 || solver param1 || solver param1 ||
     77||= 7 =|| solver param2 || solver param2 || solver param2 ||
     78||= 8 =|| obs_ens || obs_ens || obs_ens ||
     79||= 9 =|| type_obs || type_obs || type_obs ||
     80||= 10 =|| || || ||
     81||= 11 =|| || type_forget || type_forget ||
     82||= 12 =|| || type_trans  || type_trans ||
     83||= 13 =|| || type_sqrt   || type_sqrt ||
     84
     85Overview of real-valued options:
     86||= rparam =||= 3dvar =||= en3dvar =||= hyb3dvar =||
     87||= 1 =|| forget || forget || forget ||
     88||= 2 =||  ||  || beta ||
     89||= 3 =|| solver param3 || solver param3 || solver param3 ||
     90||= 4 =|| solver param4 || solver param4 || solver param4 ||
     91
     92In the overview 'solver param1' to 'solver param4' denote different integer and real-valued parameters that control the different solvers used in the 3D-Var methods.
     93
     94
     95== Local Ensemble Kalman Filters ==
     96
     97=== LESTKF (filtertype=PDAF_DA_LESTKF=7) ===
     98
     99{{{
     100PDAF     Available options for LESTKF:
     101PDAF     --- Sub-types (Parameter subtype) ---
     102PDAF       0: Standard implementation with ensemble integration
     103PDAF       10: Fixed error space basis
     104PDAF       11: Fixed state covariance matrix
     105PDAF     --- Integer parameters (Array param_int) ---
     106PDAF       param_int(1): dim_p
     107PDAF            Dimension of state vector (>0), required
     108PDAF       param_int(2): dim_ens
     109PDAF            Ensemble size (>0), required
     110PDAF       param_int(3): dim_lag
     111PDAF           Size of smoothing lag (>=0), optional
     112PDAF            0: no smoothing (default)
     113PDAF            >0: apply smoother up to specified lag
     114PDAF       param_int(4): not used
     115PDAF       param_int(5) type_forget
     116PDAF           Type of forgetting factor; optional
     117PDAF            0: fixed forgetting factor (default)
     118PDAF            1: adaptive forgetting factor for full domain
     119PDAF            2: locally adaptive forgetting factor
     120PDAF       param_int(6) type_trans
     121PDAF           Type of ensemble transformation matrix; optional
     122PDAF            0: deterministic Omega (default)
     123PDAF            1: random orthonormal Omega orthogonal to (1,...,1)^T
     124PDAF            2: use product of 0 with random orthonomal matrix with eigenvector (1,...,1)^T
     125PDAF              (experimental; for random transformations, 1 is recommended)
     126PDAF       param_int(7) type_sqrt
     127PDAF           Type of transformation matrix square root; optional
     128PDAF            0: symmetric square root (default)
     129PDAF            1: Cholesky decomposition
     130PDAF       param_int(8): observe_ens
     131PDAF           Application of observation operator H, optional
     132PDAF            0: Apply H to ensemble mean to compute innovation
     133PDAF            1: Apply H to ensemble states; then compute innovation from their mean (default)
     134PDAF               param_int(8)=1 is the recomended choice for nonlinear H
     135PDAF       param_int(9): type_obs_init
     136PDAF           Initialize observations before or after call to prepoststep_pdaf
     137PDAF            0: Initialize observations before call to prepoststep_pdaf
     138PDAF            1: Initialize observations after call to prepoststep_pdaf (default)
     139PDAF     --- Floating point parameters (Array param_real) ---
     140PDAF       param_real(1): forget
     141PDAF           Forgetting factor (usually >0 and <=1), required
     142PDAF     --- Further parameters ---
     143PDAF       n_modeltasks: Number of parallel model integration tasks
     144PDAF           >=1 for subtype 0; not larger than total number of processors
     145PDAF           =1 required for subtypes 10 and 11
     146PDAF       screen: Control verbosity of PDAF
     147PDAF           0: no outputs
     148PDAF           1: basic output (default)
     149PDAF           2: 1 plus timing output
     150PDAF           3: 2 plus debug output
     151PDAF     +++++++++ End of option overview for the LESTKF ++++++++++
     152}}}
     153
     154
     155=== LETKF (filtertype=PDAF_DA_LETKF=5) ===
     156
     157{{{
     158PDAF     Available options for LETKF:
     159PDAF     --- Sub-types (Parameter subtype) ---
     160PDAF       0: full ensemble integration;  apply T-matrix analogously to SEIK
     161PDAF       1: full ensemble integration; formulation cf. Hunt et al. (2007) without T matrix
     162PDAF       10: Fixed error space basis; analysis with T-matrix
     163PDAF       11: Fixed state covariance matrix; analysis with T-matrix
     164PDAF     --- Integer parameters (Array param_int) ---
     165PDAF       param_int(1): dim_p
     166PDAF            Dimension of state vector (>0), required
     167PDAF       param_int(2): dim_ens
     168PDAF            Ensemble size (>0), required
     169PDAF       param_int(3): dim_lag
     170PDAF           Size of smoothing lag (>=0), optional
     171PDAF            0: no smoothing (default)
     172PDAF            >0: apply smoother up to specified lag
     173PDAF       param_int(4): not used
     174PDAF       param_int(5) type_forget
     175PDAF           Type of forgetting factor; optional
     176PDAF            0: fixed forgetting factor (default)
     177PDAF            1: adaptive forgetting factor for full domain
     178PDAF            2: locally adaptive forgetting factor
     179PDAF       param_int(6) type_trans
     180PDAF           Type of ensemble transformation matrix; optional
     181PDAF            0: deterministic transformation (default)
     182PDAF            2: use product of 0 with random orthonomal matrix with eigenvector (1,...,1)^T
     183PDAF       param_int(8): observe_ens
     184PDAF           Application of observation operator H, optional
     185PDAF            0: Apply H to ensemble mean to compute innovation
     186PDAF            1: Apply H to ensemble states; then compute innovation from their mean (default)
     187PDAF               param_int(8)=1 is the recomended choice for nonlinear H
     188PDAF       param_int(9): type_obs_init
     189PDAF           Initialize observations before or after call to prepoststep_pdaf
     190PDAF            0: Initialize observations before call to prepoststep_pdaf
     191PDAF            1: Initialize observations after call to prepoststep_pdaf (default)
     192PDAF     --- Floating point parameters (Array param_real) ---
     193PDAF       param_real(1): forget
     194PDAF           Forgetting factor (usually >0 and <=1), required
     195PDAF     --- Further parameters ---
     196PDAF       n_modeltasks: Number of parallel model integration tasks
     197PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
     198PDAF           =1 required for subtypes 10 and 11
     199PDAF       screen: Control verbosity of PDAF
     200PDAF           0: no outputs
     201PDAF           1: basic output (default)
     202PDAF           2: 1 plus timing output
     203PDAF           3: 2 plus debug output
     204PDAF     +++++++++ End of option overview for the LETKF ++++++++++
     205}}}
     206
     207
     208=== LEnKF (filtertype=PDAF_DA_LENKF=8) ===
     209
     210{{{
     211PDAF     Available options for LEnKF:
     212PDAF     --- Sub-types (Parameter subtype) ---
     213PDAF       0: Standard EnKF analysis with covariance localization
     214PDAF     --- Integer parameters (Array param_int) ---
     215PDAF       param_int(1): dim_p
     216PDAF            Dimension of state vector (>0), required
     217PDAF       param_int(2): dim_ens
     218PDAF            Ensemble size (>0), required
     219PDAF       param_int(3): not used
     220PDAF       param_int(4): rank_ana_enkf
     221PDAF           maximum rank for inversion of HPH^T, optional, default=0
     222PDAF            for =0, HPH is inverted by solving the representer equation
     223PDAF            allowed range is 0 to ensemble size - 1
     224PDAF       param_int(5): not used
     225PDAF       param_int(6): not used
     226PDAF       param_int(7): not used
     227PDAF       param_int(8): observe_ens
     228PDAF           Application of observation operator H, optional
     229PDAF            0: Apply H to ensemble mean to compute innovation
     230PDAF            1: Apply H to ensemble states; then compute innovation from their mean (default)
     231PDAF               param_int(8)=1 is the recomended choice for nonlinear H
     232PDAF       param_int(9): type_obs_init
     233PDAF           Initialize observations before or after call to prepoststep_pdaf
     234PDAF            0: Initialize observations before call to prepoststep_pdaf
     235PDAF            1: Initialize observations after call to prepoststep_pdaf (default)
     236PDAF     --- Floating point parameters (Array param_real) ---
     237PDAF       param_real(1): forget
     238PDAF           Forgetting factor (usually >0 and <=1), required
     239PDAF     --- Further parameters ---
     240PDAF       n_modeltasks: Number of parallel model integration tasks
     241PDAF           (>=1; not larger than total number of processors)
     242PDAF       screen: Control verbosity of PDAF
     243PDAF           0: no outputs
     244PDAF           1: basic output (default)
     245PDAF           2: 1 plus timing output
     246PDAF           3: 2 plus debug output
     247PDAF     +++++++++ End of option overview for the LEnKF ++++++++++
     248}}}
     249
     250
     251
     252=== LSEIK (filtertype=PDAF_DA_LSEIK=3) ===
     253
     254{{{
     255PDAF     Available options for LSEIK:
     256PDAF     --- Sub-types (Parameter subtype) ---
     257PDAF       0: full ensemble integration; left-sided application of T
     258PDAF       1: full ensemble integration; explicit ensemble transformation
     259PDAF       10: Fixed error space basis
     260PDAF       11: Fixed state covariance matrix
     261PDAF     --- Integer parameters (Array param_int) ---
     262PDAF       param_int(1): dim_p
     263PDAF            Dimension of state vector (>0), required
     264PDAF       param_int(2): dim_ens
     265PDAF            Ensemble size (>0), required
     266PDAF       param_int(3): not used
     267PDAF       param_int(4): not used
     268PDAF       param_int(5) type_forget
     269PDAF           Type of forgetting factor; optional
     270PDAF            0: fixed forgetting factor (default)
     271PDAF            1: adaptive forgetting factor
     272PDAF            2: locally adaptive forgetting factor
     273PDAF       param_int(6) type_trans
     274PDAF           Type of ensemble transformation matrix; optional
     275PDAF            0: deterministic Omega (default)
     276PDAF            1: random orthonormal Omega orthogonal to (1,...,1)^T
     277PDAF            2: use product of 0 with random orthonomal matrix with eigenvector (1,...,1)^T
     278PDAF              (experimental; for random transformations, 1 is recommended)
     279PDAF       param_int(7) type_sqrt
     280PDAF           Type of transformation matrix square root; optional
     281PDAF            (Only relevant for subtype/=11)
     282PDAF            0: symmetric square root (default)
     283PDAF            1: Cholesky decomposition
     284PDAF       param_int(8): observe_ens
     285PDAF           Application of observation operator H, optional
     286PDAF            0: Apply H to ensemble mean to compute innovation
     287PDAF            1: Apply H to ensemble states; then compute innovation from their mean (default)
     288PDAF               param_int(8)=1 is the recomended choice for nonlinear H
     289PDAF       param_int(9): type_obs_init
     290PDAF           Initialize observations before or after call to prepoststep_pdaf
     291PDAF            0: Initialize observations before call to prepoststep_pdaf
     292PDAF            1: Initialize observations after call to prepoststep_pdaf (default)
     293PDAF     --- Floating point parameters (Array param_real) ---
     294PDAF       param_real(1): forget
     295PDAF           Forgetting factor (usually >0 and <=1), required
     296PDAF     --- Further parameters ---
     297PDAF       n_modeltasks: Number of parallel model integration tasks
     298PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
     299PDAF           =1 required for subtypes 10 and 11
     300PDAF       screen: Control verbosity of PDAF
     301PDAF           0: no outputs
     302PDAF           1: basic output (default)
     303PDAF           2: 1 plus timing output
     304PDAF           3: 2 plus debug output
     305PDAF     --- Internal parameter (defined inside PDAF) ---
     306PDAF       Nm1vsN: Normalization of covariance matrix; default: 1
     307PDAF           0: normalization with 1/(Ensemble size)
     308PDAF              (original SEIK, mainly for compatibility with older studies)
     309PDAF           1: normalization with 1/(Ensemble size - 1)
     310PDAF              (sample covariance matrix consistent with other EnKFs)
     311PDAF     +++++++++ End of option overview for the LSEIK filter ++++++++++
     312}}}
     313
     314
     315
     316=== ENSRF/EAKF (filtertype=PDAF_DA_ENSRF=13) ===
     317
     318{{{
     319PDAF     Available options for ENSRF:
     320PDAF     --- Sub-types (Parameter subtype) ---
     321PDAF       0: ENSRF with serial observation processing (cf. Houtekamer/Mitchell, 2002)
     322PDAF       1: EAKF/2-step local least squares filter (cf. Anderson, 2003)
     323PDAF     --- Integer parameters (Array param_int) ---
     324PDAF       param_int(1): dim_p
     325PDAF            Dimension of state vector (>0), required
     326PDAF       param_int(2): dim_ens
     327PDAF            Ensemble size (>0), required
     328PDAF       param_int(3): not used
     329PDAF       param_int(4): not used
     330PDAF       param_int(5): not used
     331PDAF       param_int(6): not used
     332PDAF       param_int(7): not used
     333PDAF       param_int(8): observe_ens
     334PDAF           Application of observation operator H, optional
     335PDAF            0: Apply H to ensemble mean to compute innovation
     336PDAF            1: Apply H to ensemble states; then compute innovation from their mean (default)
     337PDAF               param_int(8)=1 is the recomended choice for nonlinear H
     338PDAF       param_int(9): type_obs_init
     339PDAF           Initialize observations before or after call to prepoststep_pdaf
     340PDAF            0: Initialize observations before call to prepoststep_pdaf
     341PDAF            1: Initialize observations after call to prepoststep_pdaf (default)
     342PDAF     --- Floating point parameters (Array param_real) ---
     343PDAF       param_real(1): forget
     344PDAF           Forgetting factor (usually >0 and <=1), required
     345PDAF     --- Further parameters ---
     346PDAF       n_modeltasks: Number of parallel model integration tasks
     347PDAF           (>=1; not larger than total number of processors)
     348PDAF       screen: Control verbosity of PDAF
     349PDAF           0: no outputs
     350PDAF           1: basic output (default)
     351PDAF           2: 1 plus timing output
     352PDAF           3: 2 plus debug output
     353PDAF     +++++++++ End of option overview for the ENSRF ++++++++++
     354}}}
     355
     356
     357== Global Ensemble Kalman Filters ==
     358
     359=== ESTKF (filtertype=PDAF_DA_ESTKF=6) ===
     360
     361{{{
     362PDAF     Available options for ESTKF:
     363PDAF     --- Sub-types (Parameter subtype) ---
     364PDAF       0: Standard implementation with ensemble integration
     365PDAF       10: Fixed error space basis
     366PDAF       11: Fixed state covariance matrix
     367PDAF     --- Integer parameters (Array param_int) ---
     368PDAF       param_int(1): dim_p
     369PDAF            Dimension of state vector (>0), required
     370PDAF       param_int(2): dim_ens
     371PDAF            Ensemble size (>0), required
     372PDAF       param_int(3): dim_lag
     373PDAF           Size of smoothing lag (>=0), optional
     374PDAF            0: no smoothing (default)
     375PDAF            >0: apply smoother up to specified lag
     376PDAF       param_int(4): not used
     377PDAF       param_int(5) type_forget
     378PDAF           Type of forgetting factor; optional
     379PDAF            0: fixed forgetting factor (default)
     380PDAF            1: adaptive forgetting factor (experimental)
     381PDAF       param_int(6) type_trans
     382PDAF           Type of ensemble transformation matrix; optional
     383PDAF            0: deterministic Omega (default)
     384PDAF            1: random orthonormal Omega orthogonal to (1,...,1)^T
     385PDAF            2: use product of 0 with random orthonomal matrix with eigenvector (1,...,1)^T
     386PDAF              (experimental; for random transformations, 0 or 1 are recommended)
     387PDAF       param_int(7) type_sqrt
     388PDAF           Type of transformation matrix square root; optional
     389PDAF            0: symmetric square root (default)
     390PDAF            1: Cholesky decomposition
     391PDAF       param_int(8): observe_ens
     392PDAF           Application of observation operator H, optional
     393PDAF            0: Apply H to ensemble mean to compute innovation
     394PDAF            1: Apply H to ensemble states; then compute innovation from their mean (default)
     395PDAF               param_int(8)=1 is the recomended choice for nonlinear H
     396PDAF       param_int(9): type_obs_init
     397PDAF           Initialize observations before or after call to prepoststep_pdaf
     398PDAF           0: Initialize observations before call to prepoststep_pdaf
     399PDAF           1: Initialize observations after call to prepoststep_pdaf (default)
     400PDAF     --- Floating point parameters (Array param_real) ---
     401PDAF       param_real(1): forget
     402PDAF           Forgetting factor (usually >0 and <=1), required
     403PDAF     --- Further parameters ---
     404PDAF       n_modeltasks: Number of parallel model integration tasks
     405PDAF           >=1 for subtype 0; not larger than total number of processors
     406PDAF           =1 required for subtypes 10 and 11
     407PDAF       screen: Control verbosity of PDAF
     408PDAF           0: no outputs
     409PDAF           1: basic output (default)
     410PDAF           2: 1 plus timing output
     411PDAF           3: 2 plus debug output
     412PDAF     +++++++++ End of option overview for the ESTKF  ++++++++++
     413}}}
     414
     415
     416
     417=== ETKF (filtertype=PDAF_DA_ETKF=4) ===
     418
     419{{{
     420PDAF     Available options for ETKF:
     421PDAF     --- Sub-types (Parameter subtype) ---
     422PDAF       0: full ensemble integration; apply T-matrix analogously to SEIK
     423PDAF       1: full ensemble integration; formulation cf. Hunt et al. (2007) without T matrix
     424PDAF       10: Fixed error space basis; analysis with T-matrix
     425PDAF       11: Fixed state covariance matrix; analysis with T-matrix
     426PDAF     --- Integer parameters (Array param_int) ---
     427PDAF       param_int(1): dim_p
     428PDAF            Dimension of state vector (>0), required
     429PDAF       param_int(2): dim_ens
     430PDAF            Ensemble size (>0), required
     431PDAF       param_int(3): dim_lag
     432PDAF           Size of smoothing lag (>=0), optional
     433PDAF            0: no smoothing (default)
     434PDAF            >0: apply smoother up to specified lag
     435PDAF       param_int(4): not used
     436PDAF       param_int(5) type_forget
     437PDAF           Type of forgetting factor; optional
     438PDAF            0: fixed forgetting factor (default)
     439PDAF            1: adaptive forgetting factor (experimental)
     440PDAF       param_int(6) type_trans
     441PDAF           Type of ensemble transformation matrix; optional
     442PDAF            0: deterministic transformation (default)
     443PDAF            2: use product of 0 with random orthonomal matrix with eigenvector (1,...,1)^T
     444PDAF       param_int(7): not used
     445PDAF       param_int(8): observe_ens
     446PDAF           Application of observation operator H, optional
     447PDAF            0: Apply H to ensemble mean to compute innovation
     448PDAF            1: Apply H to ensemble states; then compute innovation from their mean (default)
     449PDAF               param_int(8)=1 is the recomended choice for nonlinear H
     450PDAF       param_int(9): type_obs_init
     451PDAF           Initialize observations before or after call to prepoststep_pdaf
     452PDAF           0: Initialize observations before call to prepoststep_pdaf
     453PDAF           1: Initialize observations after call to prepoststep_pdaf (default)
     454PDAF     --- Floating point parameters (Array param_real) ---
     455PDAF       param_real(1): forget
     456PDAF           Forgetting factor (usually >0 and <=1), required
     457PDAF     --- Further parameters ---
     458PDAF       n_modeltasks: Number of parallel model integration tasks
     459PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
     460PDAF           =1 required for subtypes 10 and 11
     461PDAF       screen: Control verbosity of PDAF
     462PDAF           0: no outputs
     463PDAF           1: basic output (default)
     464PDAF           2: 1 plus timing output
     465PDAF           3: 2 plus debug output
     466PDAF     +++++++++ End of option overview for the ETKF ++++++++++
     467}}}
     468
     469
     470
     471=== EnKF (filtertype=PDAF_DA_ENKF=2) ===
     472
     473{{{
     474PDAF     Available options for EnKF:
     475PDAF     --- Sub-types (Parameter subtype) ---
     476PDAF       0: Full ensemble integration; analysis for 2*dim_obs>dim_ens
     477PDAF       1: Full ensemble integration; analysis for 2*dim_obs<=dim_ens
     478PDAF     --- Integer parameters (Array param_int) ---
     479PDAF       param_int(1): dim_p
     480PDAF            Dimension of state vector (>0), required
     481PDAF       param_int(2): dim_ens
     482PDAF            Ensemble size (>0), required
     483PDAF       param_int(3): dim_lag
     484PDAF           Size of smoothing lag (>=0), optional
     485PDAF           0: no smoothing (default)
     486PDAF           >0: apply smoother up to specified lag
     487PDAF       param_int(4): rank_ana_enkf
     488PDAF           maximum rank for inversion of HPH^T, optional, default=0
     489PDAF            for =0, HPH is inverted by solving the representer equation
     490PDAF            allowed range is 0 to ensemble size - 1
     491PDAF       param_int(5): not used
     492PDAF       param_int(6): not used
     493PDAF       param_int(7): not used
     494PDAF       param_int(8): observe_ens
     495PDAF           Application of observation operator H, optional
     496PDAF            0: Apply H to ensemble mean to compute innovation
     497PDAF            1: Apply H to ensemble states; then compute innovation from their mean (default)
     498PDAF               param_int(8)=1 is the recomended choice for nonlinear H
     499PDAF       param_int(9): type_obs_init
     500PDAF           Initialize observations before or after call to prepoststep_pdaf
     501PDAF            0: Initialize observations before call to prepoststep_pdaf
     502PDAF            1: Initialize observations after call to prepoststep_pdaf (default)
     503PDAF     --- Floating point parameters (Array param_real) ---
     504PDAF       param_real(1): forget
     505PDAF           Forgetting factor (usually >0 and <=1), required
     506PDAF     --- Further parameters ---
     507PDAF       n_modeltasks: Number of parallel model integration tasks
     508PDAF           (>=1; not larger than total number of processors)
     509PDAF       screen: Control verbosity of PDAF
     510PDAF           0: no outputs
     511PDAF           1: basic output (default)
     512PDAF           2: 1 plus timing output
     513PDAF           3: 2 plus debug output
     514PDAF     +++++++++ End of option overview for the EnKF ++++++++++
     515}}}
     516
     517
     518=== SEIK (filtertype=PDAF_DA_SEIK=1) ===
    12519
    13520{{{
    … …  
    16523PDAF       0: full ensemble integration; left-sided application of T
    17524PDAF       1: full ensemble integration; right-sided application of T
    18 PDAF       2: Fixed error space basis
    19 PDAF       3: Fixed state covariance matrix
    20 PDAF       4: Implementation with explicit ensemble transformation
    21 PDAF     --- Integer parameters (Array param_int) ---
    22 PDAF       param_int(1): Dimension of state vector (>0), required
    23 PDAF       param_int(2): Ensemble size (>0), required
     525PDAF       2: full ensemble integration; explicit ensemble transformation
     526PDAF       10: Fixed error space basis
     527PDAF       11: Fixed state covariance matrix
     528PDAF     --- Integer parameters (Array param_int) ---
     529PDAF       param_int(1): dim_p
     530PDAF            Dimension of state vector (>0), required
     531PDAF       param_int(2): dim_ens
     532PDAF            Ensemble size (>0), required
    24533PDAF       param_int(3): not used
    25 PDAF       param_int(4): Apply incremental updating; optional
    26 PDAF           0: no incremental updating (default)
    27 PDAF           1: apply incremental updating
    28 PDAF       param_int(5): Type of forgetting factor; optional
    29 PDAF           0: fixed forgetting factor (default)
    30 PDAF           1: adaptive forgetting factor (experimental)
    31 PDAF       param_int(6): Type of ensemble transformation matrix; optional
    32 PDAF           0: deterministic omega (default)
    33 PDAF           1: random orthonormal omega orthogonal to (1,...,1)^T
    34 PDAF           2: use product of 0 with random orthonomal matrix with eigenvector (1,...,1)^T
     534PDAF       param_int(4): not used
     535PDAF       param_int(5) type_forget
     536PDAF           Type of forgetting factor; optional
     537PDAF            0: fixed forgetting factor (default)
     538PDAF            1: adaptive forgetting factor (experimental)
     539PDAF       param_int(6) type_trans
     540PDAF           Type of ensemble transformation matrix; optional
     541PDAF            0: deterministic Omega (default)
     542PDAF            1: random orthonormal Omega orthogonal to (1,...,1)^T
     543PDAF            2: use product of 0 with random orthonomal matrix with eigenvector (1,...,1)^T
    35544PDAF              (experimental; for random transformations, 1 is recommended)
    36 PDAF       param_int(7): Type of transformation matrix square root; optional
    37 PDAF           (Only relevant for subtype/=3)
    38 PDAF           0: symmetric square root (default)
    39 PDAF           1: Cholesky decomposition
    40 PDAF       param_int(8): Application of observation operator H, optional
    41 PDAF           0: Apply H to ensemble mean to compute residual (default)
    42 PDAF           1: Apply H to all ensemble states and then compute residual from mean of these
    43 PDAF              param_int(8)=1 is the recomended choice for nonlinear H
    44 PDAF     --- Floating point parameters (Array param_real) ---
    45 PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required
    46 PDAF     --- Further parameters ---
    47 PDAF       n_modeltasks: Number of parallel model integration tasks
    48 PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
    49 PDAF           =1 required for subtypes 2 and 3
     545PDAF       param_int(7) type_sqrt
     546PDAF           Type of transformation matrix square root; optional
     547PDAF            (Only relevant for subtype/=11)
     548PDAF            0: symmetric square root (default)
     549PDAF            1: Cholesky decomposition
     550PDAF       param_int(8): observe_ens
     551PDAF           Application of observation operator H, optional
     552PDAF            0: Apply H to ensemble mean to compute innovation
     553PDAF            1: Apply H to ensemble states; then compute innovation from their mean (default)
     554PDAF               param_int(8)=1 is the recomended choice for nonlinear H
     555PDAF       param_int(9): type_obs_init
     556PDAF           Initialize observations before or after call to prepoststep_pdaf
     557PDAF            0: Initialize observations before call to prepoststep_pdaf
     558PDAF            1: Initialize observations after call to prepoststep_pdaf (default)
     559PDAF     --- Floating point parameters (Array param_real) ---
     560PDAF       param_real(1): forget
     561PDAF           Forgetting factor (usually >0 and <=1), required
     562PDAF     --- Further parameters ---
     563PDAF       n_modeltasks: Number of parallel model integration tasks
     564PDAF           >=1 for subtypes 0, 1 and 2; not larger than total number of processors
     565PDAF           =1 required for subtypes 10 and 11
    50566PDAF       screen: Control verbosity of PDAF
    51567PDAF           0: no outputs
    … …  
    63579
    64580
    65 == EnKF (filtertype=2) ==
    66 
    67 {{{
    68 PDAF     Available options for EnKF:
    69 PDAF     --- Sub-types (Parameter subtype) ---
    70 PDAF       0: Full ensemble integration; analysis for 2*dim_obs>dim_ens
    71 PDAF       1: Full ensemble integration; analysis for 2*dim_obs<=dim_ens
    72 PDAF     --- Integer parameters (Array param_int) ---
    73 PDAF       param_int(1): Dimension of state vector (>0), required
    74 PDAF       param_int(2): Ensemble size (>0), required
    75 PDAF       param_int(3): maximum rank for inversion of HPH^T, optional, default=0
    76 PDAF           (for =0, HPH is inverted by solving the representer equation)
    77 PDAF           (if set to >=ensemble size, it is reset to ensemble size - 1)
    78 PDAF       param_int(4): not used
    79 PDAF       param_int(5): Size of smoothing lag (>=0), optional
    80 PDAF           0: no smoothing (default)
    81 PDAF           >0: apply smoother up to specified lag
    82 PDAF     --- Floating point parameters (Array param_real) ---
    83 PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required
    84 PDAF     --- Further parameters ---
    85 PDAF       n_modeltasks: Number of parallel model integration tasks
    86 PDAF           (>=1; not larger than total number of processors)
    87 PDAF       screen: Control verbosity of PDAF
    88 PDAF           0: no outputs
    89 PDAF           1: basic output (default)
    90 PDAF           2: 1 plus timing output
    91 PDAF           3: 2 plus debug output
    92 PDAF     +++++++++ End of option overview for the EnKF ++++++++++
    93 }}}
    94 
    95 
    96 == LSEIK (filtertype=3) ==
    97 
    98 {{{
    99 PDAF     Available options for LSEIK:
    100 PDAF     --- Sub-types (Parameter subtype) ---
    101 PDAF       0: full ensemble integration; left-sided application of T
    102 PDAF       2: Fixed error space basis
    103 PDAF       3: Fixed state covariance matrix
    104 PDAF     --- Integer parameters (Array param_int) ---
    105 PDAF       param_int(1): Dimension of state vector (>0), required
    106 PDAF       param_int(2): Ensemble size (>0), required
    107 PDAF       param_int(3): not used
    108 PDAF       param_int(4): Apply incremental updating; optional
    109 PDAF           0: no incremental updating (default)
    110 PDAF           1: apply incremental updating
    111 PDAF       param_int(5): Type of forgetting factor; optional
    112 PDAF           0: fixed forgetting factor (default)
    113 PDAF           1: adaptive forgetting factor for full domain (experimental)
    114 PDAF           2: locally adaptive forgetting factor (experimental)
    115 PDAF       param_int(6): Type of ensemble transformation matrix; optional
    116 PDAF           0: deterministic omega (default)
    117 PDAF           1: random orthonormal omega orthogonal to (1,...,1)^T
    118 PDAF           2: use product of 0 with random orthonomal matrix with eigenvector (1,...,1)^T
    119 PDAF              (experimental; for random transformations, 1 is recommended)
    120 PDAF       param_int(7): Type of transformation matrix square root; optional
    121 PDAF           (Only relevant for subtype/=3)
    122 PDAF           0: symmetric square root (default)
    123 PDAF           1: Cholesky decomposition
    124 PDAF     --- Floating point parameters (Array param_real) ---
    125 PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required
    126 PDAF     --- Further parameters ---
    127 PDAF       n_modeltasks: Number of parallel model integration tasks
    128 PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
    129 PDAF           =1 required for subtypes 2 and 3
    130 PDAF       screen: Control verbosity of PDAF
    131 PDAF           0: no outputs
    132 PDAF           1: basic output (default)
    133 PDAF           2: 1 plus timing output
    134 PDAF           3: 2 plus debug output
    135 PDAF     --- Internal parameter (defined inside PDAF) ---
    136 PDAF       Nm1vsN: Normalization of covariance matrix; default: 1
    137 PDAF           0: normalization with 1/(Ensemble size)
    138 PDAF              (original SEIK, mainly for compatibility with older studies)
    139 PDAF           1: normalization with 1/(Ensemble size - 1)
    140 PDAF              (sample covariance matrix consistent with other EnKFs)
    141 PDAF     +++++++++ End of option overview for the LSEIK filter ++++++++++
    142 }}}
    143 
    144 
    145 == ETKF (filtertype=4) ==
    146 
    147 {{{
    148 PDAF     Available options for ETKF:
    149 PDAF     --- Sub-types (Parameter subtype) ---
    150 PDAF       0: full ensemble integration; apply T-matrix analogously to SEIK
    151 PDAF       1: full ensemble integration; formulation without T matrix
    152 PDAF       2: Fixed error space basis; analysis with T-matrix
    153 PDAF       3: Fixed state covariance matrix; analysis with T-matrix
    154 PDAF     --- Integer parameters (Array param_int) ---
    155 PDAF       param_int(1): Dimension of state vector (>0), required
    156 PDAF       param_int(2): Ensemble size (>0), required
    157 PDAF       param_int(3): Size of smoothing lag (>=0), optional
    158 PDAF           0: no smoothing (default)
    159 PDAF           >0: apply smoother up to specified lag
    160 PDAF       param_int(4): not used
    161 PDAF       param_int(5): Type of forgetting factor; optional
    162 PDAF           0: fixed forgetting factor (default)
    163 PDAF           1: adaptive forgetting factor (experimental)
    164 PDAF       param_int(6): Type of ensemble transformation matrix; optional
    165 PDAF           0: deterministic transformation (default)
    166 PDAF           2: use product of 0 with random orthonomal matrix with eigenvector (1,...,1)^T
    167 PDAF       param_int(7): not used
    168 PDAF       param_int(8): Application of observation operator H
    169 PDAF           0: Apply H to ensemble mean to compute residual (default)
    170 PDAF           1: Apply H to all ensemble states; then compute residual from mean of these
    171 PDAF              param_int(8)=1 is the recomended choice for nonlinear H
    172 PDAF     --- Floating point parameters (Array param_real) ---
    173 PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required
    174 PDAF     --- Further parameters ---
    175 PDAF       n_modeltasks: Number of parallel model integration tasks
    176 PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
    177 PDAF           =1 required for subtypes 2 and 3
    178 PDAF       screen: Control verbosity of PDAF
    179 PDAF           0: no outputs
    180 PDAF           1: basic output (default)
    181 PDAF           2: 1 plus timing output
    182 PDAF           3: 2 plus debug output
    183 PDAF     +++++++++ End of option overview for the ETKF ++++++++++
    184 }}}
    185 
    186 
    187 == LETKF (filtertype=5) ==
    188 
    189 {{{
    190 PDAF     Available options for LETKF:
    191 PDAF     --- Sub-types (Parameter subtype) ---
    192 PDAF       0: full ensemble integration;  apply T-matrix analogously to SEIK
    193 PDAF       2: Fixed error space basis; analysis with T-matrix
    194 PDAF       3: Fixed state covariance matrix; analysis with T-matrix
    195 PDAF     --- Integer parameters (Array param_int) ---
    196 PDAF       param_int(1): Dimension of state vector (>0), required
    197 PDAF       param_int(2): Ensemble size (>0), required
    198 PDAF       param_int(3): Size of smoothing lag (>=0), optional
    199 PDAF           0: no smoothing (default)
    200 PDAF           >0: apply smoother up to specified lag
    201 PDAF       param_int(4): not used
    202 PDAF       param_int(5): Type of forgetting factor; optional
    203 PDAF           0: fixed forgetting factor (default)
    204 PDAF           1: adaptive forgetting factor for full domain (experimental)
    205 PDAF           2: locally adaptive forgetting factor (experimental)
    206 PDAF       param_int(6): Type of ensemble transformation matrix; optional
    207 PDAF           0: deterministic transformation (default)
    208 PDAF           2: use product of 0 with random orthonomal matrix with eigenvector (1,...,1)^T
    209 PDAF     --- Floating point parameters (Array param_real) ---
    210 PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required
    211 PDAF     --- Further parameters ---
    212 PDAF       n_modeltasks: Number of parallel model integration tasks
    213 PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
    214 PDAF           =1 required for subtypes 2 and 3
    215 PDAF       screen: Control verbosity of PDAF
    216 PDAF           0: no outputs
    217 PDAF           1: basic output (default)
    218 PDAF           2: 1 plus timing output
    219 PDAF           3: 2 plus debug output
    220 PDAF     +++++++++ End of option overview for the LETKF ++++++++++
    221 }}}
    222 
    223 
    224 == ESTKF (filtertype=6) ==
    225 
    226 {{{
    227 PDAF     Available options for ESTKF:
     581
     582
     583
     584
     585
     586== Nonlinear DA Methods ==
     587
     588=== NETF (filtertype=PDAF_DA_NETF=9) ===
     589
     590{{{
     591PDAF     Available options for NETF:
    228592PDAF     --- Sub-types (Parameter subtype) ---
    229593PDAF       0: Standard implementation with ensemble integration
    230 PDAF       2: Fixed error space basis
    231 PDAF       3: Fixed state covariance matrix
    232 PDAF     --- Integer parameters (Array param_int) ---
    233 PDAF       param_int(1): Dimension of state vector (>0), required
    234 PDAF       param_int(2): Ensemble size (>0), required
    235 PDAF       param_int(3): Size of smoothing lag (>=0), optional
    236 PDAF           0: no smoothing (default)
    237 PDAF           >0: apply smoother up to specified lag
    238 PDAF       param_int(4): not used
    239 PDAF       param_int(5): Type of forgetting factor; optional
    240 PDAF           0: fixed forgetting factor (default)
    241 PDAF           1: adaptive forgetting factor (experimental)
    242 PDAF       param_int(6): Type of ensemble transformation matrix; optional
    243 PDAF           0: deterministic omega (default)
    244 PDAF           1: random orthonormal omega orthogonal to (1,...,1)^T
    245 PDAF           2: use product of 0 with random orthonomal matrix with eigenvector (1,...,1)^T
    246 PDAF              (experimental; for random transformations, 0 or 1 are recommended)
    247 PDAF       param_int(7): Type of transformation matrix square root; optional
    248 PDAF           0: symmetric square root (default)
    249 PDAF           1: Cholesky decomposition
    250 PDAF       param_int(8): Application of observation operator H
    251 PDAF           0: Apply H to ensemble mean to compute residual (default)
    252 PDAF           1: Apply H to all ensemble states; then compute residual from mean of these
    253 PDAF              param_int(8)=1 is the recomended choice for nonlinear H
    254 PDAF     --- Floating point parameters (Array param_real) ---
    255 PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required
    256 PDAF     --- Further parameters ---
    257 PDAF       n_modeltasks: Number of parallel model integration tasks
    258 PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
    259 PDAF           =1 required for subtypes 2 and 3
    260 PDAF       screen: Control verbosity of PDAF
    261 PDAF           0: no outputs
    262 PDAF           1: basic output (default)
    263 PDAF           2: 1 plus timing output
    264 PDAF           3: 2 plus debug output
    265 PDAF     +++++++++ End of option overview for the ESTKF  ++++++++++
    266 }}}
    267 
    268 
    269 == LESTKF (filtertype=7) ==
    270 
    271 {{{
    272 PDAF     Available options for LESTKF:
     594PDAF     --- Integer parameters (Array param_int) ---
     595PDAF       param_int(1): dim_p
     596PDAF            Dimension of state vector (>0), required
     597PDAF       param_int(2): dim_ens
     598PDAF            Ensemble size (>0), required
     599PDAF       param_int(3): dim_lag
     600PDAF           Size of smoothing lag (>=0), optional
     601PDAF            0: no smoothing (default)
     602PDAF            >0: apply smoother up to specified lag
     603PDAF       param_int(4): type_noise
     604PDAF           Type of ensemble perturbations, optional
     605PDAF            0: no perturbations (default)
     606PDAF            1: constant standard deviation
     607PDAF            2: relative to ensemble standard deviation
     608PDAF       param_int(5) type_forget
     609PDAF           Type of forgetting factor; optional
     610PDAF            0: forgetting factor on forecast ensemble (default)
     611PDAF            2: forgetting factor on analysis ensemble
     612PDAF       param_int(6) type_trans
     613PDAF           Type of ensemble transformation matrix; optional
     614PDAF            0: random orthonormal matrix orthogonal to (1,...,1)^T (default)
     615PDAF            1: deterministic transformation
     616PDAF       param_int(7): type_winf
     617PDAF           Type of weights inflation; optional
     618PDAF            0: no weights inflation (default)
     619PDAF            1: inflate so that N_eff/N > param_real(2)
     620PDAF       param_int(8): observe_ens
     621PDAF           Application of observation operator H, optional
     622PDAF           Note: This parameter has no influence on the NETF assimilation result
     623PDAF            0: Apply H to ensemble mean to compute innovation
     624PDAF            1: Apply H to ensemble states; then compute innovation from their mean (default)
     625PDAF               param_int(8)=1 is the recomended choice for nonlinear H
     626PDAF       param_int(9): type_obs_init
     627PDAF           Initialize observations before or after call to prepoststep_pdaf
     628PDAF           0: Initialize observations before call to prepoststep_pdaf
     629PDAF           1: Initialize observations after call to prepoststep_pdaf (default)
     630PDAF     --- Floating point parameters (Array param_real) ---
     631PDAF       param_real(1): forget
     632PDAF           Forgetting factor (usually >0 and <=1), required
     633PDAF       param_real(2): limit_winf
     634PDAF           Limit for weigts inflation N_eff/N > param_real(2), optional, default=0.0
     635PDAF       param_real(3): noise_amp
     636PDAF           Ensemble perturbation level (>0), required, only used if param_int(4)>0
     637PDAF     --- Further parameters ---
     638PDAF       n_modeltasks: Number of parallel model integration tasks
     639PDAF           >=1; not larger than total number of processors
     640PDAF       screen: Control verbosity of PDAF
     641PDAF           0: no outputs
     642PDAF           1: basic output (default)
     643PDAF           2: 1 plus timing output
     644PDAF           3: 2 plus debug output
     645PDAF     +++++++++ End of option overview for the NETF  ++++++++++
     646}}}
     647
     648
     649
     650
     651=== LNETF (filtertype=PDAF_DA_LNETF=10) ===
     652
     653{{{
     654PDAF     Available options for LNETF:
    273655PDAF     --- Sub-types (Parameter subtype) ---
    274656PDAF       0: Standard implementation with ensemble integration
    275 PDAF       2: Fixed error space basis
    276 PDAF       3: Fixed state covariance matrix
    277 PDAF     --- Integer parameters (Array param_int) ---
    278 PDAF       param_int(1): Dimension of state vector (>0), required
    279 PDAF       param_int(2): Ensemble size (>0), required
    280 PDAF       param_int(3): Size of smoothing lag (>=0), optional
    281 PDAF           0: no smoothing (default)
    282 PDAF           >0: apply smoother up to specified lag
    283 PDAF       param_int(4): not used
    284 PDAF       param_int(5): Type of forgetting factor; optional
    285 PDAF           0: fixed forgetting factor (default)
    286 PDAF           1: adaptive forgetting factor for full domain (experimental)
    287 PDAF           2: locally adaptive forgetting factor (experimental)
    288 PDAF       param_int(6): Type of ensemble transformation matrix; optional
    289 PDAF           0: deterministic omega (default)
    290 PDAF           1: random orthonormal omega orthogonal to (1,...,1)^T
    291 PDAF           2: use product of 0 with random orthonomal matrix with eigenvector (1,...,1)^T
    292 PDAF              (experimental; for random transformations, 1 is recommended)
    293 PDAF       param_int(7): Type of transformation matrix square root; optional
    294 PDAF           0: symmetric square root (default)
    295 PDAF           1: Cholesky decomposition
    296 PDAF     --- Floating point parameters (Array param_real) ---
    297 PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required
    298 PDAF     --- Further parameters ---
    299 PDAF       n_modeltasks: Number of parallel model integration tasks
    300 PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
    301 PDAF           =1 required for subtypes 2 and 3
    302 PDAF       screen: Control verbosity of PDAF
    303 PDAF           0: no outputs
    304 PDAF           1: basic output (default)
    305 PDAF           2: 1 plus timing output
    306 PDAF           3: 2 plus debug output
    307 PDAF     +++++++++ End of option overview for the LESTKF ++++++++++
    308 }}}
    309 
    310 
    311 == LEnKF (filtertype=8) ==
    312 
    313 {{{
    314 PDAF     Available options for LEnKF:
    315 PDAF     --- Sub-types (Parameter subtype) ---
    316 PDAF       0: Full ensemble integration; analysis with covariance localization
    317 PDAF     --- Integer parameters (Array param_int) ---
    318 PDAF       param_int(1): Dimension of state vector (>0), required
    319 PDAF       param_int(2): Ensemble size (>0), required
    320 PDAF       param_int(3): maximum rank for inversion of HPH^T, optional, default=0
    321 PDAF           (for =0, HPH is inverted by solving the representer equation)
    322 PDAF           (if set to >=ensemble size, it is reset to ensemble size - 1)
    323 PDAF     --- Floating point parameters (Array param_real) ---
    324 PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required
    325 PDAF     --- Further parameters ---
    326 PDAF       n_modeltasks: Number of parallel model integration tasks
    327 PDAF           (>=1; not larger than total number of processors)
    328 PDAF       screen: Control verbosity of PDAF
    329 PDAF           0: no outputs
    330 PDAF           1: basic output (default)
    331 PDAF           2: 1 plus timing output
    332 PDAF           3: 2 plus debug output
    333 PDAF     +++++++++ End of option overview for the LEnKF ++++++++++
    334 }}}
    335 
    336 
    337 == NETF (filtertype=9) ==
    338 
    339 {{{
    340 PDAF     Available options for NETF:
    341 PDAF     --- Sub-types (Parameter subtype) ---
    342 PDAF       0: Standard implementation with ensemble integration
    343 PDAF     --- Integer parameters (Array param_int) ---
    344 PDAF       param_int(1): Dimension of state vector (>0), required
    345 PDAF       param_int(2): Ensemble size (>0), required
    346 PDAF       param_int(3): Size of smoothing lag (>=0), optional
    347 PDAF           0: no smoothing (default)
    348 PDAF           >0: apply smoother up to specified lag
    349 PDAF       param_int(4): Type of ensemble perturbations, optional
    350 PDAF           0: no perturbations (default)
    351 PDAF           1: constant standard deviation
    352 PDAF           2: relative to ensemble standard deviation
    353 PDAF       param_int(5): Type of forgetting factor; optional
    354 PDAF           0: forgetting factor on forecast ensemble (default)
    355 PDAF           2: forgetting factor on analysis ensemble
    356 PDAF       param_int(6): Type of ensemble transformation matrix; optional
    357 PDAF           0: random orthonormal matrix orthogonal to (1,...,1)^T (default)
    358 PDAF           1: deterministic transformation
    359 PDAF       param_int(7): Type of weights inflation; optional
    360 PDAF           0: no weights inflation (default)
    361 PDAF           1: inflate so that N_eff/N > param_real(2)
    362 PDAF     --- Floating point parameters (Array param_real) ---
    363 PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required
    364 PDAF       param_real(2): Limit for weigts inflation N_eff/N > param_real(2), optional, default=0.0
    365 PDAF       param_real(3): Ensemble perturbation level (>0), required, only used if param_int(4)>0
    366 PDAF     --- Further parameters ---
    367 PDAF       n_modeltasks: Number of parallel model integration tasks
    368 PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
    369 PDAF           =1 required for subtypes 2 and 3
    370 PDAF       screen: Control verbosity of PDAF
    371 PDAF           0: no outputs
    372 PDAF           1: basic output (default)
    373 PDAF           2: 1 plus timing output
    374 PDAF           3: 2 plus debug output
    375 PDAF     +++++++++ End of option overview for the NETF  ++++++++++
    376 }}}
    377 
    378 == LNETF (filtertype=10) ==
    379 
    380 {{{
    381 PDAF     Available options for LNETF:
    382 PDAF     --- Sub-types (Parameter subtype) ---
    383 PDAF       0: Standard implementation with ensemble integration
    384 PDAF     --- Integer parameters (Array param_int) ---
    385 PDAF       param_int(1): Dimension of state vector (>0), required
    386 PDAF       param_int(2): Ensemble size (>0), required
    387 PDAF       param_int(3): Size of smoothing lag (>=0), optional
    388 PDAF           0: no smoothing (default)
    389 PDAF           >0: apply smoother up to specified lag
    390 PDAF       param_int(4): Type of ensemble perturbations, optional
    391 PDAF           0: no perturbations (default)
    392 PDAF           1: constant standard deviation
    393 PDAF           2: relative to ensemble standard deviation
    394 PDAF       param_int(5): Type of forgetting factor; optional
    395 PDAF           0: forgetting factor on forecast ensemble (default)
    396 PDAF           1: forgetting factor on forecast ensemble only observed domains
    397 PDAF           2: forgetting factor on analysis ensemble
    398 PDAF           3: forgetting factor on analysis ensemble only observed domains
    399 PDAF       param_int(6): Type of ensemble transformation matrix; optional
    400 PDAF           0: random orthonormal matrix orthogonal to (1,...,1)^T (default)
    401 PDAF           1: deterministic transformation
    402 PDAF       param_int(7): Type of weights inflation; optional
    403 PDAF           0: no weights inflation (default)
    404 PDAF           1: inflate so that N_eff/N > param_real(2)
    405 PDAF     --- Floating point parameters (Array param_real) ---
    406 PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required
    407 PDAF       param_real(2): Limit for weigts inflation N_eff/N > param_real(2), optional, default=0.0
    408 PDAF       param_real(3): Ensemble perturbation level (>0), required, only used if param_int(4)>0
    409 PDAF     --- Further parameters ---
    410 PDAF       n_modeltasks: Number of parallel model integration tasks
    411 PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
    412 PDAF           =1 required for subtypes 2 and 3
     657PDAF     --- Integer parameters (Array param_int) ---
     658PDAF       param_int(1): dim_p
     659PDAF            Dimension of state vector (>0), required
     660PDAF       param_int(2): dim_ens
     661PDAF            Ensemble size (>0), required
     662PDAF       param_int(3): dim_lag
     663PDAF           Size of smoothing lag (>=0), optional
     664PDAF            0: no smoothing (default)
     665PDAF            >0: apply smoother up to specified lag
     666PDAF       param_int(4): type_noise
     667PDAF           Type of ensemble perturbations, optional
     668PDAF            0: no perturbations (default)
     669PDAF            1: constant standard deviation
     670PDAF            2: relative to ensemble standard deviation
     671PDAF       param_int(5) type_forget
     672PDAF           Type of forgetting factor; optional
     673PDAF            0: forgetting factor on forecast ensemble (default)
     674PDAF            1: forgetting factor on forecast ensemble only observed domains
     675PDAF            2: forgetting factor on analysis ensemble
     676PDAF            3: forgetting factor on analysis ensemble only observed domains
     677PDAF       param_int(6) type_trans
     678PDAF           Type of ensemble transformation matrix; optional
     679PDAF            0: random orthonormal matrix orthogonal to (1,...,1)^T (default)
     680PDAF            1: deterministic transformation
     681PDAF       param_int(7): type_winf
     682PDAF           Type of weights inflation; optional
     683PDAF            0: no weights inflation (default)
     684PDAF            1: inflate so that N_eff/N > param_real(2)
     685PDAF       param_int(8): observe_ens
     686PDAF           Application of observation operator H, optional
     687PDAF           Note: This parameter has no influence on the LNETF assimilation result
     688PDAF            0: Apply H to ensemble mean to compute innovation
     689PDAF            1: Apply H to ensemble states; then compute innovation from their mean (default)
     690PDAF               param_int(8)=1 is the recomended choice for nonlinear H
     691PDAF       param_int(9): type_obs_init
     692PDAF           Initialize observations before or after call to prepoststep_pdaf
     693PDAF           0: Initialize observations before call to prepoststep_pdaf
     694PDAF           1: Initialize observations after call to prepoststep_pdaf (default)
     695PDAF     --- Floating point parameters (Array param_real) ---
     696PDAF       param_real(1): forget
     697PDAF           Forgetting factor (usually >0 and <=1), required
     698PDAF       param_real(2): limit_winf
     699PDAF           Limit for weigts inflation N_eff/N > param_real(2), optional, default=0.0
     700PDAF       param_real(3): noise_amp
     701PDAF           Ensemble perturbation level (>0), required, only used if param_int(4)>0
     702PDAF     --- Further parameters ---
     703PDAF       n_modeltasks: Number of parallel model integration tasks
     704PDAF           >=1; not larger than total number of processors
    413705PDAF       screen: Control verbosity of PDAF
    414706PDAF           0: no outputs
    … …  
    420712
    421713
    422 == LKNETF (filtertype=11) ==
    423 
    424 Note: The LKNETF (hybrid local Kalman-nonlinear ensemble transform filter) was added in PDAF V2.1.
     714
     715
     716=== LKNETF (filtertype=PDAF_DA_LKNETF=11) ===
    425717
    426718{{{
    … …  
    429721PDAF       0: HNK: 2-step LKNETF with NETF before LETKF
    430722PDAF       1: HKN: 2-step LKNETF with LETKF before NETF
    431 PDAF       4: HSync: LKNETF synchronous
    432 PDAF     --- Integer parameters (Array param_int) ---
    433 PDAF       param_int(1): Dimension of state vector (>0), required
    434 PDAF       param_int(2): Ensemble size (>0), required
     723PDAF       2: HSync: LKNETF synchronous
     724PDAF     --- Integer parameters (Array param_int) ---
     725PDAF       param_int(1): dim_p
     726PDAF            Dimension of state vector (>0), required
     727PDAF       param_int(2): dim_ens
     728PDAF            Ensemble size (>0), required
    435729PDAF       param_int(3): not used
    436 PDAF       param_int(4): not used
    437 PDAF       param_int(5): Type of forgetting factor; optional
    438 PDAF           0: inflate forecast ensemble by 1/forget (default)
    439 PDAF           1: inflate forecast ensemble by 1/forget only observed domains
    440 PDAF           2: inflate analysis ensemble by 1/forget
    441 PDAF           3: inflate analysis ensemble by 1/forget only observed domains
    442 PDAF       param_int(6): Type of ensemble transformation matrix; optional
    443 PDAF           0: random orthonormal matrix orthogonal to (1,...,1)^T (default)
    444 PDAF           1: deterministic transformation
    445 PDAF       param_int(7): Type of hybrid weight; optional
    446 PDAF           0: fixed value
    447 PDAF           1: gamma_lin: (1 - N_eff/N_e)*param_real(2) (default)
    448 PDAF           2: gamma_alpha: hybrid weight from N_eff/N>=param_real(2)
    449 PDAF           3: gamma_ska: 1 - min(s,k)/sqrt(param_real(3)) with N_eff/N>=param_real(2)
    450 PDAF           4: gamma_sklin: 1 - min(s,k)/sqrt(param_real(3)) >= 1-N_eff/N>=param_real(2)
    451 PDAF     --- Floating point parameters (Array param_real) ---
    452 PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required
    453 PDAF       param_real(2): prescribed hybrid weight gamma (usually >0 and <=1), optional, default=1.0
    454 PDAF       param_real(3): hybrid norm kappa (>0), optional, default=dim_ens
    455 PDAF     --- Further parameters ---
    456 PDAF       n_modeltasks: Number of parallel model integration tasks
    457 PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
    458 PDAF           =1 required for subtypes 2 and 3
     730PDAF       param_int(4): type_hyb
     731PDAF           Type of hybrid weight; optional
     732PDAF            0: fixed value
     733PDAF            1: gamma_lin: (1 - N_eff/N_e)*param_real(2) (default)
     734PDAF            2: gamma_alpha: hybrid weight from N_eff/N>=param_real(2)
     735PDAF            3: gamma_ska: 1 - min(s,k)/sqrt(param_real(3)) with N_eff/N>=param_real(2)
     736PDAF            4: gamma_sklin: 1 - min(s,k)/sqrt(param_real(3)) >= 1-N_eff/N>=param_real(2)
     737PDAF       param_int(5): type_forget
     738PDAF           Type of forgetting factor; optional
     739PDAF            0: inflate forecast ensemble by 1/forget (default)
     740PDAF            1: inflate forecast ensemble by 1/forget only observed domains
     741PDAF            2: inflate analysis ensemble by 1/forget
     742PDAF            3: inflate analysis ensemble by 1/forget only observed domains
     743PDAF            5: adaptive forgetting factor for full domain in LETKF part
     744PDAF            6: locally adaptive forgetting factor in LETKF part
     745PDAF       param_int(6): type_trans
     746PDAF           Type of ensemble transformation matrix; optional
     747PDAF            0: random orthonormal matrix orthogonal to (1,...,1)^T (default)
     748PDAF            1: deterministic transformation
     749PDAF       param_int(7): not used
     750PDAF     --- Floating point parameters (Array param_real) ---
     751PDAF       param_real(1): forget
     752PDAF           Forgetting factor (usually >0 and <=1), required
     753PDAF       param_real(2): hyb_g
     754PDAF           prescribed hybrid weight gamma (usually >0 and <=1), optional, default=0.95
     755PDAF       param_real(3): hyb_k
     756PDAF           hybrid norm kappa (>0), optional, default=dim_ens
     757PDAF     --- Further parameters ---
     758PDAF       n_modeltasks: Number of parallel model integration tasks
     759PDAF           >=1; not larger than total number of processors
    459760PDAF       screen: Control verbosity of PDAF
    460761PDAF           0: no outputs
    … …  
    465766}}}
    466767
    467 == PF (filtertype=12) ==
     768
     769
     770
     771=== PF (filtertype=PDAF_DA_PF=12) ===
    468772
    469773{{{
    … …  
    472776PDAF       0: Standard implementation with ensemble integration
    473777PDAF     --- Integer parameters (Array param_int) ---
    474 PDAF       param_int(1): Dimension of state vector (>0), required
    475 PDAF       param_int(2): Ensemble size (>0), required
    476 PDAF       param_int(3): Resampling type, optional
    477 PDAF           1: probabilistic resamping (default)
    478 PDAF           2: stochastic universal resampling
    479 PDAF           3: residual resampling
    480 PDAF       param_int(4): Type of ensemble perturbations, optional
    481 PDAF           0: no perturbations (default)
    482 PDAF           1: constant standard deviation
    483 PDAF           2: relative to ensemble standard deviation
    484 PDAF       param_int(5): Type of forgetting factor; optional
    485 PDAF           0: forgetting factor on forecast ensemble (default)
    486 PDAF           2: forgetting factor on analysis ensemble
    487 PDAF       param_int(6): Type of weights inflation; optional
    488 PDAF           0: no weights inflation (default)
    489 PDAF           1: inflate so that N_eff/N > param_real(2)
    490 PDAF     --- Floating point parameters (Array param_real) ---
    491 PDAF       param_real(1): Ensemble pert. level (>0), required, only used if param_int(4)>0
    492 PDAF       param_real(2): Forgetting factor (usually >0 and <=1), optional, default=1.0
    493 PDAF       param_real(3): Limit for weigts inflation N_eff/N > param_real(2), optional, default=0.0
    494 PDAF     --- Further parameters ---
    495 PDAF       n_modeltasks: Number of parallel model integration tasks
    496 PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
    497 PDAF           =1 required for subtypes 2 and 3
     778PDAF       param_int(1): dim_p
     779PDAF            Dimension of state vector (>0), required
     780PDAF       param_int(2): dim_ens
     781PDAF            Ensemble size (>0), required
     782PDAF       param_int(3): type_resample
     783PDAF           Resampling type, optional
     784PDAF            1: probabilistic resamping (default)
     785PDAF            2: stochastic universal resampling
     786PDAF            3: residual resampling
     787PDAF       param_int(4): type_noise
     788PDAF           Type of ensemble perturbations, optional
     789PDAF            0: no perturbations (default)
     790PDAF            1: constant standard deviation
     791PDAF            2: relative to ensemble standard deviation
     792PDAF       param_int(5) type_forget
     793PDAF           Type of forgetting factor; optional
     794PDAF            0: forgetting factor on forecast ensemble (default)
     795PDAF            2: forgetting factor on analysis ensemble
     796PDAF       param_int(6): not used
     797PDAF       param_int(7): type_winf
     798PDAF           Type of weights inflation; optional
     799PDAF            0: no weights inflation (default)
     800PDAF            1: inflate so that N_eff/N > param_real(2)
     801PDAF       param_int(8): observe_ens
     802PDAF           Application of observation operator H, optional
     803PDAF           Note: This parameter has no influence on the PF assimilation result
     804PDAF            0: Apply H to ensemble mean to compute innovation
     805PDAF            1: Apply H to ensemble states; then compute innovation from their mean (default)
     806PDAF               param_int(8)=1 is the recomended choice for nonlinear H
     807PDAF       param_int(9): type_obs_init
     808PDAF           Initialize observations before or after call to prepoststep_pdaf
     809PDAF           0: Initialize observations before call to prepoststep_pdaf
     810PDAF           1: Initialize observations after call to prepoststep_pdaf (default)
     811PDAF     --- Floating point parameters (Array param_real) ---
     812PDAF       param_real(1): forget
     813PDAF           Forgetting factor (usually >0 and <=1), required
     814PDAF       param_real(2): limit_winf
     815PDAF           Limit for weigts inflation N_eff/N > param_real(2), optional, default=0.0
     816PDAF       param_real(3): noise_amp
     817PDAF           Ensemble perturbation level (>0), required, only used if param_int(4)>0
     818PDAF     --- Further parameters ---
     819PDAF       n_modeltasks: Number of parallel model integration tasks
     820PDAF           >=1; not larger than total number of processors
    498821PDAF       screen: Control verbosity of PDAF
    499822PDAF           0: no outputs
    … …  
    504827}}}
    505828
    506 == GENOBS (filtertype=100) ==
    507 
    508 Note: GENOBS used filtertype=11 before PDAF V2.0.
     829
     830
     831
     832
     833== Observation Generation ==
     834
     835=== GENOBS (filtertype=PDAF_DA_GENOBS=100) ===
    509836
    510837{{{
    … …  
    513840PDAF       0: Standard implementation with ensemble integration
    514841PDAF     --- Integer parameters (Array param_int) ---
    515 PDAF       param_int(1): Dimension of state vector (>0), required
    516 PDAF       param_int(2): Ensemble size (>0), required
     842PDAF       param_int(1): dim_p
     843PDAF            Dimension of state vector (>0), required
     844PDAF       param_int(2): dim_ens
     845PDAF            Ensemble size (>0), required
     846PDAF       param_int(3): seedset
     847PDAF           seed set index for random number generator, optional
     848PDAF           valid are values between 1 and 20; default=1
    517849PDAF     --- Floating point parameters (Array param_real) ---
    518850PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required, but not used
    … …  
    520852PDAF       n_modeltasks: Number of parallel model integration tasks
    521853PDAF           =1 for GENOBS; not larger than total number of processors
    522 PDAF           =1 required for subtypes 2 and 3
     854PDAF           =1 required for subtypes 10 and 11
    523855PDAF       screen: Control verbosity of PDAF
    524856PDAF           0: no outputs
    … …  
    529861}}}
    530862
    531 == 3DVAR (filtertype=200) ==
    532 
    533 Note: 3D-Var methods were added in PDAF V2.0.
     863
     864
     865
     866== 3D-Var Methods ==
     867
     868=== 3D-Var (filtertype=PDAF_DA_3DVAR=200) ===
    534869
    535870{{{
    … …  
    538873PDAF       0: incremental 3D-Var with parameterized covariance matrix
    539874PDAF       1: 3D ensemble Var using LESTKF for ensemble transformation
    540 PDAF       4: 3D ensemble Var using ESTKF for ensemble transformation
    541 PDAF       6: hybrid 3D-Var using LESTKF for ensemble transformation
    542 PDAF       7: hybrid 3D-Var using ESTKF for ensemble transformation
    543 PDAF     --- Integer parameters (Array param_int) ---
    544 PDAF       param_int(1): Dimension of state vector (>0), required
    545 PDAF       param_int(2): Ensemble size (>0), required
    546 PDAF       param_int(3): Select optimization method (solver), required
    547 PDAF           0: LBFGS (default)
    548 PDAF           1: CG+
    549 PDAF           2: direct implementation of CG
    550 PDAF           3: direct implementation of CG with decomposed control vector
    551 PDAF       param_int(4): size of parameterized control vector (for parameterized and hybrid 3D-Var), required
     875PDAF       2: 3D ensemble Var using ESTKF for ensemble transformation
     876PDAF       3: hybrid 3D-Var using LESTKF for ensemble transformation
     877PDAF       4: hybrid 3D-Var using ESTKF for ensemble transformation
     878PDAF     --- Integer parameters (Array param_int) ---
     879PDAF       param_int(1): dim_p
     880PDAF            Dimension of state vector (>0), required
     881PDAF       param_int(2): dim_ens
     882PDAF            Ensemble size (>0), required
     883PDAF       param_int(3): type_opt
     884PDAF           Select optimization method (solver), required
     885PDAF            1: LBFGS (default)
     886PDAF            2: CG+
     887PDAF            3: direct implementation of CG
     888PDAF            12: CG+ with decomposed control vector
     889PDAF            13: direct implementation of CG with decomposed control vector
     890PDAF       param_int(4): size of parameterized control vector (for 3D-Var and hybrid 3D-Var), required
    552891PDAF       param_int(5): size of ensemble control vector (required for ensemble and hybrid 3D-Var),
    553892PDAF       param_int(6): Solver-specific parameter, optional
    … …  
    560899PDAF           CG+: parameter irest (default=1)
    561900PDAF                (0) no restarts; (n>0) restart every n steps
     901PDAF       param_int(8): observe_ens
     902PDAF           Application of observation operator H, optional
     903PDAF            0: Apply H to ensemble mean to compute innovation
     904PDAF            1: Apply H to ensemble states; then compute innovation from their mean (default)
     905PDAF               param_int(8)=1 is the recomended choice for nonlinear H
     906PDAF       param_int(9): type_obs_init
     907PDAF           Initialize observations before or after call to prepoststep_pdaf
     908PDAF           0: Initialize observations before call to prepoststep_pdaf
     909PDAF           1: Initialize observations after call to prepoststep_pdaf (default)
     910PDAF       param_int(10): not used
     911PDAF       ___Options for ESTKF/LESTKF for En3DVar/hyb3DVar___
     912PDAF       param_int(11) type_forget
     913PDAF           Type of forgetting factor; optional
     914PDAF            0: fixed forgetting factor (default)
     915PDAF            1: adaptive forgetting factor (experimental)
     916PDAF            2: locally adaptive forgetting factor (experimental)
     917PDAF       param_int(12) type_trans
     918PDAF           Type of ensemble transformation matrix; optional
     919PDAF            0: deterministic Omega (default)
     920PDAF            1: random orthonormal Omega orthogonal to (1,...,1)^T
     921PDAF            2: use product of 0 with random orthonomal matrix with eigenvector (1,...,1)^T
     922PDAF              (experimental; for random transformations, 0 or 1 are recommended)
     923PDAF       param_int(13) type_sqrt
     924PDAF           Type of transformation matrix square root; optional
     925PDAF            0: symmetric square root (default)
     926PDAF            1: Cholesky decomposition
    562927PDAF     --- Floating point parameters (Array param_real) ---
    563928PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required;
    … …  
    573938PDAF     --- Further parameters ---
    574939PDAF       n_modeltasks: Number of parallel model integration tasks
    575 PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
    576 PDAF           =1 required for subtypes 2 and 3
     940PDAF           >=1 for subtypes >0; not larger than total number of processors
     941PDAF           =1 required for subtype 0
    577942PDAF       screen: Control verbosity of PDAF
    578943PDAF           0: no outputs
    … …  
    582947PDAF     +++++++++ End of option overview for 3DVAR ++++++++++
    583948}}}
    584 
    585 
    586 
    587 == SEEK (filtertype=0) ==
    588 
    589 **Note:** The SEEK filter is deprecated and will be removed in a future release of PDAF.
    590 
    591 {{{
    592 PDAF     Available options for SEEK:
    593 PDAF     --- Sub-types (Parameter subtype) ---
    594 PDAF       0: Evolve unit modes with finite difference approx. of TLM
    595 PDAF       1: like 0 with modes scaled by eigenvalues, unit U
    596 PDAF       2: Fixed basis vectors; variable U matrix
    597 PDAF       3: Fixed covariance matrix (V and U kept constant)
    598 PDAF     --- Integer parameters (Array param_int) ---
    599 PDAF       param_int(1): Dimension of state vector (>0), required
    600 PDAF       param_int(2): Ensemble size (>0), required
    601 PDAF       param_int(3): Interval for re-diagonalization of P (>0); optional: default 1
    602 PDAF       param_int(4): 1 for incremental updating, 0 else; optional: default 0
    603 PDAF     --- Floating point parameters (Array param_real) ---
    604 PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required
    605 PDAF       param_real(2): epsilon for finite-difference approx. of TLM, required
    606 PDAF     --- Further parameters ---
    607 PDAF       n_modeltasks: Number of parallel model integration tasks
    608 PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
    609 PDAF           =1 required for subtypes 2 and 3
    610 PDAF       screen: Control verbosity of PDAF
    611 PDAF           0: no outputs
    612 PDAF           1: basic output (default)
    613 PDAF           2: 1 plus timing output
    614 PDAF           3: 2 plus debug output
    615 PDAF     +++++++++ End of option overview for the SEEK filter ++++++++++
    616 }}}