Changes between Version 13 and Version 14 of AvailableOptionsforInitPDAF


Ignore:
Timestamp:
Feb 21, 2023, 4:42:31 PM (3 years ago)
Author:
lnerger
Comment:

--

Legend:

Unmodified
Added
Removed
Modified

  • AvailableOptionsforInitPDAF

    v13 v14  
    77
    88Below we list the options as they are displayed using `subtype=-1`.
    9 
    10 == SEEK (filtertype=0) ==
    11 
    12 {{{
    13 PDAF     Available options for SEEK:
    14 PDAF     --- Sub-types (Parameter subtype) ---
    15 PDAF       0: Evolve unit modes with finite difference approx. of TLM
    16 PDAF       1: like 0 with modes scaled by eigenvalues, unit U
    17 PDAF       2: Fixed basis vectors; variable U matrix
    18 PDAF       3: Fixed covariance matrix (V and U kept constant)
    19 PDAF       5: Offline mode
    20 PDAF     --- Integer parameters (Array param_int) ---
    21 PDAF       param_int(1): Dimension of state vector (>0), required
    22 PDAF       param_int(2): Ensemble size (>0), required
    23 PDAF       param_int(3): Interval for re-diagonalization of P (>0); optional: default 1
    24 PDAF       param_int(4): 1 for incremental updating, 0 else; optional: default 0
    25 PDAF     --- Floating point parameters (Array param_real) ---
    26 PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required
    27 PDAF       param_real(2): epsilon for finite-difference approx. of TLM, required
    28 PDAF     --- Further parameters ---
    29 PDAF       n_modeltasks: Number of parallel model integration tasks
    30 PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
    31 PDAF           =1 required for subtypes 2 and 3
    32 PDAF       screen: Control verbosity of PDAF
    33 PDAF           0: no outputs
    34 PDAF           1: basic output (default)
    35 PDAF           2: 1 plus timing output
    36 PDAF           3: 2 plus debug output
    37 PDAF     +++++++++ End of option overview for the SEEK filter ++++++++++
    38 }}}
    399
    4010
    … …  
    386356PDAF           0: no smoothing (default)
    387357PDAF           >0: apply smoother up to specified lag
    388 PDAF       param_int(4): not used
     358PDAF       param_int(4): Type of ensemble perturbations, optional
     359PDAF           0: no perturbations (default)
     360PDAF           1: constant standard deviation
     361PDAF           2: relative to ensemble standard deviation
    389362PDAF       param_int(5): Type of forgetting factor; optional
    390363PDAF           0: forgetting factor on forecast ensemble (default)
    … …  
    399372PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required
    400373PDAF       param_real(2): Limit for weigts inflation N_eff/N > param_real(2), optional, default=0.0
     374PDAF       param_real(3): Ensemble perturbation level (>0), required, only used if param_int(4)>0
    401375PDAF     --- Further parameters ---
    402376PDAF       n_modeltasks: Number of parallel model integration tasks
    … …  
    424398PDAF           0: no smoothing (default)
    425399PDAF           >0: apply smoother up to specified lag
    426 PDAF       param_int(4): not used
     400PDAF       param_int(4): Type of ensemble perturbations, optional
     401PDAF           0: no perturbations (default)
     402PDAF           1: constant standard deviation
     403PDAF           2: relative to ensemble standard deviation
    427404PDAF       param_int(5): Type of forgetting factor; optional
    428405PDAF           0: forgetting factor on forecast ensemble (default)
    … …  
    439416PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required
    440417PDAF       param_real(2): Limit for weigts inflation N_eff/N > param_real(2), optional, default=0.0
     418PDAF       param_real(3): Ensemble perturbation level (>0), required, only used if param_int(4)>0
    441419PDAF     --- Further parameters ---
    442420PDAF       n_modeltasks: Number of parallel model integration tasks
    … …  
    451429}}}
    452430
     431
     432== LKNETF (filtertype=11) ==
     433
     434Note: The LKNETF (hybrid local Kalman-nonlinear ensemble transform filter) was added in PDAF V2.1.
     435
     436{{{
     437PDAF     Available options for LKNETF:
     438PDAF     --- Sub-types (Parameter subtype) ---
     439PDAF       0: HNK: 2-step LKNETF with NETF before LETKF
     440PDAF       1: HKN: 2-step LKNETF with LETKF before NETF
     441PDAF       4: HSync: LKNETF synchronous
     442PDAF       5: Offline mode - HNK: 2-step LKNETF with NETF before LETKF
     443PDAF     --- Integer parameters (Array param_int) ---
     444PDAF       param_int(1): Dimension of state vector (>0), required
     445PDAF       param_int(2): Ensemble size (>0), required
     446PDAF       param_int(3): not used
     447PDAF       param_int(4): not used
     448PDAF       param_int(5): Type of forgetting factor; optional
     449PDAF           0: inflate forecast ensemble by 1/forget (default)
     450PDAF           1: inflate forecast ensemble by 1/forget only observed domains
     451PDAF           2: inflate analysis ensemble by 1/forget
     452PDAF           3: inflate analysis ensemble by 1/forget only observed domains
     453PDAF       param_int(6): Type of ensemble transformation matrix; optional
     454PDAF           0: random orthonormal matrix orthogonal to (1,...,1)^T (default)
     455PDAF           1: deterministic transformation
     456PDAF       param_int(7): Type of hybrid weight; optional
     457PDAF           0: fixed value
     458PDAF           1: gamma_lin: (1 - N_eff/N_e)*param_real(2) (default)
     459PDAF           2: gamma_alpha: hybrid weight from N_eff/N>=param_real(2)
     460PDAF           3: gamma_ska: 1 - min(s,k)/sqrt(param_real(3)) with N_eff/N>=param_real(2)
     461PDAF           4: gamma_sklin: 1 - min(s,k)/sqrt(param_real(3)) >= 1-N_eff/N>=param_real(2)
     462PDAF     --- Floating point parameters (Array param_real) ---
     463PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required
     464PDAF       param_real(2): prescribed hybrid weight gamma (usually >0 and <=1), optional, default=1.0
     465PDAF       param_real(3): hybrid norm kappa (>0), optional, default=dim_ens
     466PDAF     --- Further parameters ---
     467PDAF       n_modeltasks: Number of parallel model integration tasks
     468PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
     469PDAF           =1 required for subtypes 2 and 3
     470PDAF       screen: Control verbosity of PDAF
     471PDAF           0: no outputs
     472PDAF           1: basic output (default)
     473PDAF           2: 1 plus timing output
     474PDAF           3: 2 plus debug output
     475PDAF     +++++++++ End of option overview for the LKNETF  ++++++++++
     476}}}
    453477
    454478== PF (filtertype=12) ==
    … …  
    540564PDAF       param_int(4): size of parameterized control vector (for parameterized and hybrid 3D-Var), required
    541565PDAF       param_int(5): size of ensemble control vector (required for ensemble and hybrid 3D-Var),
    542 PDAF       param_int(4): Dimension of parameterized control vector
    543 PDAF     --- Floating point parameters (Array param_real) ---
    544 PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required
    545 PDAF       param_real(2): hybrid weight beta, optional
    546 PDAF           >=0.0 and <=1.0 (default = 0.5)
    547 PDAF     --- Further parameters ---
    548 PDAF       n_modeltasks: Number of parallel model integration tasks
    549 PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
    550 PDAF           =1 required for subtypes 2 and 3
    551 PDAF       screen: Control verbosity of PDAF
    552 PDAF           0: no outputs
    553 PDAF           1: basic output (default)
    554 PDAF           2: 1 plus timing output
    555 PDAF           3: 2 plus debug output
    556 PDAF     +++++++++ End of option overview for 3DVAR ++++++++++}}}
     566PDAF       param_int(6): Solver-specific parameter, optional
     567PDAF           LBFGS: parameter m (default=5)
     568PDAF                Number of corrections used in limited memory matrix; 3<=m<=20
     569PDAF           CG+: parameter method (default=2)
     570PDAF                (1) Fletcher-Reeves, (2) Polak-Ribiere, (3) positive Polak-Ribiere
     571PDAF           CG: maximum number of iterations (default=200)
     572PDAF       param_int(7): Solver-specific parameter, optional
     573PDAF           CG+: parameter irest (default=1)
     574PDAF                (0) no restarts; (n>0) restart every n steps
     575PDAF     --- Floating point parameters (Array param_real) ---
     576PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required;
     577PDAF           (only used for ensemble and hybrid 3D-Var)
     578PDAF       param_real(2): hybrid weight beta, optional (only for hybrid 3D-Var)
     579PDAF           range >=0.0 and <=1.0, =1.0 for pure ensemble 3D-var  (default=0.5)
     580PDAF       param_real(3): Solver-specific parameter, optional
     581PDAF           LBFGS: Limit for stopping iterations (pgtol, default=1.0e-5)
     582PDAF           CG+: convergence parameter eps (default=1.0e-5)
     583PDAF           CG: convergence parameter eps (default=1.0e-6)
     584PDAF       param_real(4): Solver-specific parameter, optional
     585PDAF           LBFGS: Tolerance in termination test (factr, default=1.0e+7)
     586PDAF     --- Further parameters ---
     587PDAF       n_modeltasks: Number of parallel model integration tasks
     588PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
     589PDAF           =1 required for subtypes 2 and 3
     590PDAF       screen: Control verbosity of PDAF
     591PDAF           0: no outputs
     592PDAF           1: basic output (default)
     593PDAF           2: 1 plus timing output
     594PDAF           3: 2 plus debug output
     595PDAF     +++++++++ End of option overview for 3DVAR ++++++++++
     596}}}
     597
     598
     599
     600== SEEK (filtertype=0) ==
     601
     602{{{
     603PDAF     Available options for SEEK:
     604PDAF     --- Sub-types (Parameter subtype) ---
     605PDAF       0: Evolve unit modes with finite difference approx. of TLM
     606PDAF       1: like 0 with modes scaled by eigenvalues, unit U
     607PDAF       2: Fixed basis vectors; variable U matrix
     608PDAF       3: Fixed covariance matrix (V and U kept constant)
     609PDAF       5: Offline mode
     610PDAF     --- Integer parameters (Array param_int) ---
     611PDAF       param_int(1): Dimension of state vector (>0), required
     612PDAF       param_int(2): Ensemble size (>0), required
     613PDAF       param_int(3): Interval for re-diagonalization of P (>0); optional: default 1
     614PDAF       param_int(4): 1 for incremental updating, 0 else; optional: default 0
     615PDAF     --- Floating point parameters (Array param_real) ---
     616PDAF       param_real(1): Forgetting factor (usually >0 and <=1), required
     617PDAF       param_real(2): epsilon for finite-difference approx. of TLM, required
     618PDAF     --- Further parameters ---
     619PDAF       n_modeltasks: Number of parallel model integration tasks
     620PDAF           >=1 for subtypes 0 and 1; not larger than total number of processors
     621PDAF           =1 required for subtypes 2 and 3
     622PDAF       screen: Control verbosity of PDAF
     623PDAF           0: no outputs
     624PDAF           1: basic output (default)
     625PDAF           2: 1 plus timing output
     626PDAF           3: 2 plus debug output
     627PDAF     +++++++++ End of option overview for the SEEK filter ++++++++++
     628}}}