midas_ensPostProcess

link to source code

Dependency Diagrams:

Direct Dependency Diagram

program  midas_enspostprocess

Purpose

Post-processing program for the local ensemble transform Kalman filter (LETKF). Many aspects of this program are controlled throught the namelist block namEnsPostProc defined in epp_postProcess.

—

Algorithm

The ensPostProcess program performs several post-processing tasks of LETK ensemble analyses and the ensemble trials. The following tasks are performed based on the namelist options.

• Mean and std: Ensemble mean and standard deviation are computed upon reading the ensemble members and after the processes which may alter the mean or the standard deviation.

• RTPP: Relaxation to the prior perturbation will relax the analysis perturbation to the trial perturbation with a given amount (\(\alpha\)) in the namelist. It can be formulated as: \(x^a_i \leftarrow \overline{x^a} + (1-\alpha)(\overline{x^a}-x^a_i)+\alpha(\overline{x^b}-x^b_i)\)

  More details for the RTPP can be found in paper: Impacts of Initial Estimate and Observation Availability on Convective-Scale Data Assimilation with an Ensemble Kalman Filter

• RTPS: Relaxation to the prior spread will relax the analysis ensemble spread (i.e. standard deviation) to the trial ensemble spread with a given amount (\(\alpha\)) in the namelist. With the ensemble spread of analysis (\(\sigma^a\)) and trial (\(\sigma^b\)), it can be formulated as: \(x^a_i \leftarrow \overline{x^a} + (\overline{x^a}-x^a_i)(1+\alpha(\sigma^b-\sigma^a)/\sigma^a)\)

  More details for the RTPS can be found in paper: Evaluating Methods to Account for System Errors in Ensemble Data Assimilation

• Humidity limits: Impose the saturation limits and RTTOV limits on the humidity for each member.

• Recenter: The ensemble analyses from the LETKF are recentered around the EnVar analysis based on the input coefficent. It is also called hybrid gain algorithm and more details can be found in paper: Using the hybrid gain algorithm to sample data assimilation uncertainty

• Subsample: Select 20 ensemble members for the medium range forecasts.

• Random perturbation: Generate homogeneous and isotropic random perturbations and add them to the ensemble analyses.

• Analysis increments: After all the processes above, the analysis incrememnts are computed at the central time for all the members including the control member and the subsampled 20 members.

• Mask analysis increments: For LAM grid, the analysis increments in the blending area are masked out.

–

Input and Output Files	Description of file
flnml	In - Main namelist file with parameters user may modify
ensemble_trial/$YYYYMMDDHH_006_$NNNN	In - Background ensemble member files
ensemble_anal/$YYYYMMDDHH_000_$NNNN	In - Analysis ensemble member files
bgcov	In - Bnmc matrix for the random perturbations
analysis_grid	In - Horizontal grid file on which the random perturbations are generated
$YYYYMMDDHH_recentering_analysis	In - Analysis from EnVar for recentering the ensemble analyses
analinc_mask	In - Mask file for masking the analysis increments for LAM grid
$YYYYMMDDHH_000_inc_$NNNN	Out - Ensemble analysis increments for trials
$YYYYMMDDHH_000_$NNNN	Out - Ensemble analyses recentered and perturbed
$YYYYMMDDHH_000_$NNNN_raw	Out - Ensemble analyses from LETKF
$YYYYMMDDHH_000_analmean	Out - Mean ensemble analysis without perturbation
$YYYYMMDDHH_000_analrms	Out - Analysis ensemble spread (i.e. standard deviation) without perturbation
$YYYYMMDDHH_000_analrms_ascii	Out - Global average of $YYYYMMDDHH_000_analrms
$YYYYMMDDHH_000_analpertmean	Out - Mean ensemble analysis with perturbation and recentering
$YYYYMMDDHH_000_analpertrms	Out - Analysis ensemble spread with perturbation and recentering
$YYYYMMDDHH_000_analpertrms_ascii	Out - Global average of $YYYYMMDDHH_000_analpertrms
$YYYYMMDDHH_006_trialmean	Out - Mean ensemble trial for all time levels
$YYYYMMDDHH_006_trialrms	Out - Trial ensemble spread at 6H
$YYYYMMDDHH_006_trialrms_ascii	Out - Global average of $YYYYMMDDHH_006_trialrms
subspace/$YYYYMMDDHH_000_inc_$NNNN	Out - Subsampled ensemble analysis increments for progs
subspace/$YYYYMMDDHH_000_$NNNN	Out - Subsampled ensemble analyses recentered and perturbed
subspace_unpert/$YYYYMMDDHH_000_$NNNN	Out - Unperturbed subsampled ensemble analyses recentered

–

Synopsis

Below is a summary of the ensPostProcess program calling sequence:

• Initial setups:

  • Read the NAMLETKF namelist and check/modify some values.

  • Various modules are setup: gridStateVector_mod, timeCoord_mod (and set up dates and dateStampList variables for both trials and increments/analyses).

  • Setup horizontal and vertical grid objects from first ensemble member file.

  • Allocate varaibles and read ensemble analyses and trials

• Ensemble postprocess:

  • Compute ensemble mean and spread (i.e. standard deviation) for analysis and trial

  • Perform RTPP and RTPS and recompute analysis spread

  • Recenter ensemble mean and recompute analysis mean and spread

  • Apply humidity saturation limit and RTTOV HU limit and recompute analysis mean

  • Subsample ensemble analysis

  • Add random perturbation to the analyses and compute perturbed analysis spread

  • Add random perturbation to the subsampled analyses and recenter subsampled mean to the global mean.

  • Compute analysis increments for all analyses and subsampled analyses including the ensemble mean

  • Mask LAM analysis increments are recompute the ensemble analyses by adding the increments to the trials

  • Writing the outputs.

Options

List of namelist blocks that can affect the ensPostProcess program.

Program/Module	Namelist	Description of what is controlled
midas_ensPostProcess	NAMENSPOSTPROC	Number of ensemble members and read, write control of trials and analyses and horizontal interpolation degree
enspostprocess_mod	NAMENSPOSTPROCMODULE	Control the postprocess algorithms
timeCoord_mod	NAMTIME	Temporal resolution of the background state and the analysis

Needed modules

• version_mod: MODULE version_mod (prefix=’ver’ category=’8. Low-level utilities and constants’)

• midasmpi_mod: MODULE midasMpi_mod (prefix=’mmpi’ category=’8. Low-level utilities and constants’)

• filenames_mod: MODULE fileNames_mod (prefix=’fln’ category=’7. Low-level data objects’)

• ensemblestatevector_mod: MODULE ensembleStateVector_mod (prefix=’ens’ category=’6. High-level data objects’)

• gridstatevector_mod: MODULE gridStateVector_mod (prefix=’gsv’ category=’6. High-level data objects’)

• gridstatevectorfileio_mod: MODULE gridStateVectorFileIO_mod (prefix=’gio’ category=’4. Data Object transformations’)

• verticalcoord_mod: MODULE verticalCoord_mod (prefix=’vco’ category=’7. Low-level data objects’)

• horizontalcoord_mod: MODULE horizontalCoord_mod (prefix=’hco’ category=’7. Low-level data objects’)

• timecoord_mod: MODULE timeCoord_mod (prefix=’tim’ category=’7. Low-level data objects’)

• utilities_mod: MODULE utilities_mod (prefix=’utl’ category=’8. Low-level utilities and constants’)

• ramdisk_mod: MODULE ramDisk_mod (prefix=’ram’ category=’8. Low-level utilities and constants’)

• enspostprocess_mod: MODULE ensPostProcess_mod (prefix=’epp’ category=’1. High-level functionality’)

Routines called

ver_printnameandversion(), mmpi_initialize(), utl_tmg_start(), ram_setup(), utl_abort(), fln_ensfilename(), tim_setup(), tim_getstamplist(), tim_getdatestamp(), gsv_setup(), hco_setupfromfile(), vco_setupfromfile(), vco_getnumlev(), gsv_allocate(), gio_readfromfile(), ens_allocate(), ens_readensemble(), utl_tmg_stop(), epp_postprocess()