This software was developed within the context of the EUMETSAT Satellite Application Facility on Numerical Weather Prediction (NWP SAF), under the Cooperation Agreement dated 25 November 1998, between EUMETSAT and the Met Office, UK, by one or more partners within the NWP SAF. The partners in the NWP SAF are the Met Office, ECMWF, KNMI and Météo-France.

© EUMETSAT 2006-2013, All Rights Reserved.

Contents

1. Introduction

The philosophy behind the development of this code was to produce a flexible, stand-alone 1DVar retrieval system that may be used for a wide variety of satellite instruments. To facilitate this, there are multiple processing options at various parts of the code with a resultant small increase in complexity compared to what it would be if only one instrument were being considered. It is anticipated that users may want to remove routines that are superfluous to their requirements (or alternatively simply take selected routines) before using this code.

The program functions first by reading in user supplied background and observation data (together with appropriate error covariance matrices). The routines that are provided to do this assume data is supplied in a predefined ASCII format, but these routines can be replaced with ones more specific to the format of data that is to be used.

Routines for channel selection and cloud detection are included but bias correction needs to be performed by the user.

Retrievals are made through the optimal estimation methods of Rodgers (1976). Three minimisation routines are provided with the one that is used depending on the size and linearity of the problem.

A radiative transfer program is required that will supply both radiances and Jacobians and this is accessed through a general purpose interface routine. Currently RTTOV (versions 7.1, 8.7, 9.3, 10.2 and 11.1), Gastropod and RTIASI version 4 (with some modifications) are implemented.

The code is written is European Standard Fortran 90 and advanced F90 features are used as much as possible. The code was originally designed to closely follow the structures in the Met Office's Observation Processing System (OPS) and is partly based on the operational ATOVS 1DVar code written at the Met Office by Andrew Smith.

[ Return to Top ]

2. Compiling and Running the Code.

Instructions for compiling and testing the code can be found in the readme.txt file in the top IASI_1DVar directory.

To run the code successfully a radiative transfer coefficients file must be located in the top IASI_1DVar directory. This coefficients file must be compatible with the radiative transfer model chosen and the instrument whose data is being used. Coefficients files can be downloaded for free from the NWPSAF RTTOV website for example RTTOV v11 coefficients can be downloaded from here.

The supplied makefile is set up to compile with RTTOV-11 by default. In order to use a different radiative transfer model a number of steps need to be completed (it has been noted that this process is more complicated than it could be and that this will be addressed in future versions):

• In the makefile edit the RTM_INC, RTM_MOD and RTM_LIB variables to contain the paths to the include files, module files and object archive (library) for your RT model.
• Set the RTMODEL variable to contain the name of the RT model object archive (e.g. "RTMODEL = rttov10.2" for librttov10.2.a).
• Replace all references to the full interface with the dummy interface of the radiative transfer model you don't wish to use. For example if you don't want to use RTTOV v11 replace all references to IASI_RTTOV11_Interface with IASI_RTTOV11_Interface_Dummy. Also, replace the dummy object of your chosen radiative transfer model with the full interface. For example if you wish to use RTTOV v10 replace all references to IASI_RTTOV10_Interface_Dummy with IASI_RTTOV10_Interface.
• If you are compiling with RTIASI, Gastropod or RTTOV v8 or earlier, remove all references to IASIMod_RTTOV9, IASIMod_RTTOV10 and IASIMod_RTTOV11 in the makefile. Similarly if you are compiling with RTTOV v9, remove all references to IASIMod_RTTOV10 and IASIMod_RTTOV11.
• In ControlData.NL change the RTModelToUse variable to your chosen radiative transfer model.
• Finally, make sure the coefficients file in the IASI_1DVar top directory is compatible with your chosen radiative transfer model. Generally RTTOV v7, v8 & v9 coefficient files are compatible with each other but RTTOV v10 and v11 coefficient files will only work with RTTOV v10 and v11 because of changes to the format of the files.

To compile the default code you should compile RTTOV without the HDF macro enabled. If you have compiled RTTOV with the HDF macro then you will need to use one of the expanded sets of linker flags in the makefile which are commented out by default and also link a hdf5 library in the makefile. If you wish to read in emissivities from an atlas then you will also need to make a few changes to the makefile:

• Linking in a NetCDF library.
  
• Replace all references to IASI_RTTOV11_interface_nonetcdf with IASI_RTTOV11_Interface.
• Use one of the expanded sets of linker flags which are commented out by default.

N.B. Due to some preprocessing directives (e.g. #include) present in RTTOV-10 and RTTOV-11 interface files a C preprocessor is now run during the compilation. The compiler options used are based upon which RTTOV version is specified in the makefile.

[ Return to Top ]

3. Input Files

3a. Control Files: ControlData.NL
3b. Control Files: Retrievals.NL
3c. Data Files: ObsFile.dat
3d. Data Files: Background.dat
3e. Auxiliary Files: Rmatrix.dat
3f. Auxiliary Files: Bmatrix.dat
3g. Auxiliary Files: ChannelChoice.dat

There are two main control files, ControlData.NL and Retrieval.NL, and two main data files, ObsFile.dat and Background.dat. Auxiliary data such as error covariance matrices and channel selection choices are stored for each data type in a directory called xxx_COEFFS_DIR where xxx refers to the data type being processed (e.g., IASI_COEFFS_DIR).

All input files are ASCII and all values are in free-format.

3a. Control Files: ControlData.NL

ControlData.NL is a Fortran90 namelist file containing the input parameters required for the running of the code. The table below lists the parameters that may be specified through this file and which ones would normally be required. As one can see most of these parameters have a default value. This file and the IASI_Read_ControlData that reads it may be easily expanded in future if one wants to add further options. Note that there are comments included in this file as supplied, but some Fortran90 compilers may not allow this, in which case they would need to be removed.

An example ControlData.NL file is:

! AIRS Control file for IASI_1DVar

! ******** THESE COMMENTS MAY NEED TO BE REMOVED IF COMPILING ***********
! ******** WITH F90 (rather than F95) ***********************************

 &Control SoundingType_Text='AIRS'
! Relative path of the coefficients directory 
          Coeffs_Dir='AIRS_COEFFS_DIR/'
! This controls the verbosity of the output 0=Minimal, 40=verbose 
          GeneralMode=0 
! For the minimisation option, 1 is Newtonian, 2 is Marquardt-Levenberg
          Minimisation_Method=1   
          MaxIterations=10
! Cloud liquid water profile in the background, 0=no (default), 1=yes
          Read_CLW_Background=0
! Cloud liquid water retrieval method, 0=LWP retrieval, 1=qtotal retrieval
          Lqtotal=0
! These are cloud cost thresholds:
          CostThresh_Land=20. 
          CostThresh_Sea=20. 
          CostThresh_IRWindow_Land=0. 
          CostThresh_IRWindow_Sea=0. 
          CloudAbsThresh_IRWindow=-10. 
          HighCloudAbsThresh_IRWindow=-55. /

[ Return to Top of Section ]

3b. Control Files: Retrievals.NL

Retrieval.NL controls the variables that are to be retrieved and provides the mapping between the minimization vector, the RT model vector and the B-matrix (Figure 1 gives and example of how the three vectors are mapped for a typical IASI retrieval).

The Retrieval.NL file is a Fortran90 namelist file. The quantities to be retrieved are specified through the presence of two- or three-element matrices in this file. Absence of a quantity from this file means that this element is not to be retrieved.

As an example, a typical Retrieval.NL file is shown below. All possible namelist entries are included here but, in practice, only those quantities that are actually being retrieved need to be included.

! Retrieval set up file for AIRS
!
! ******** THESE COMMENTS MAY NEED TO BE REMOVED IF COMPILING ***********
! ******** WITH F90 (rather than F95) ***********************************

&Retrieval Temperature=1,43,1
           Humidity= 18,26,44
           Ozone= 0,0,0
           Surface_Temperature= 1,70
           Surface_Humidity= 1,71
           Surface_Pressure= 0,0
           Skin_Temperature= 1,72
           Cloud_Liquid_Water= 0,0
           Cloud_Top_Pressure= 0,0
           Cloud_Fraction= 0,0
           Surface_UWind= 0,0
           Surface_VWind= 0,0 /

The three-element matrices are for profiles. The first two elements denote the top levels in the atmosphere and the number of levels that are to be retrieved respectively. The pressure profile of the atmosphere being defined in the Background.dat file. The final element is the position in the background error covariance matrix that corresponds to the top level to be retrieved. If any of these elements are zero, the profile is not retrieved.

In the example given, therefore, 43 levels of temperature are to be retrieved starting with the first (top) level. Only 26 humidity levels are to be retrieved starting at level 26 (and finishing with level 43). Ozone is not to be retrieved at all. In the background error covariance matrix, the first 43 elements are for temperature, while elements 44-69 correspond to the humidity on levels 18-43.

The two-element matrices are for scalar (mostly surface) quantities. If the first element is non-zero the quantity is to be retrieved (provided the second entry is valid). The second entry points to the position in the background error covariance matrix for this quantity. For microwave Liquid Water Path (LWP) retrievals, a covariance matrix not correlated to any other variables is assumed and this single value (0.2, inherited from the SSMIS 1D var code) is hardcoded into the routine FIXME. Hence, if LWP has to be retrieved while the first element has to non-zero, the second element has to remain 0.

Cloud_Top_Pressure and Cloud_Fraction do not require a background error covariance entry as the default is to assume large background errors for these quantities if they are to be retrieved. The retrieval of these cloudy parameters uses some additional code to the normal retrieval. The methodology for this is documented in Appendix C.

In the example given, surface temperature, skin temperature and surface humidity are retrieved.

[ Return to Top of Section ]

3c. Data Files: Obsfile.dat

The Obsfile.dat file contains the observation information. All data are specified in free format with mandatory colons delimiting the explanatory labels in the header.

The file starts with 10 lines that are reserved for users comments. Next comes header information which details the number of observations present in the file plus information on which channels have been used to make up a composite instrument (such as ATOVS).

There are are two optional header entries: "Composite Instruments:" and "Channels".

"Composite Instruments:" is used where there is the possibility of more than one "composite instrument" (a composite instrument being a collection of one (e.g, IASI) or more (e.g., ATOVS) instruments that are treated as a single entity for retrieval purposes). The composite instrument entry will then specify the number of instruments followed on separate lines of text with the names of the composite instruments (these should be identical with the required entries in the RMatrix files). E.g. for ATOVS on NOAA satellites 15 and 16:

This is a test ATOVS observations file

This is based on a RTTOV simulated set of radiances from NOAA-15
which is computed from ADC's background profile surf emis=0 RWS
22nd March 2
001

There are 10 header lines in total here!!!



So this is line 10.
Number of Observations in File:    10
No. of Chans per Observation:      40
Total Number of instruments making up observations : 6
*** In the following Series, Platform and Instrument are defined  *** 
*** according to the relevant RT Model definitions (if required): ***
Composite Instruments: 2
NOAA-15 ATOVS
NOAA-16 ATOVS
Sat. Series   Platform   Instrument First_Channel   Last_Channel  Sat ID
    1           15           0           1               20       15 
    1           15           3          21               35       15
    1           15           4          36               40       15
    1           16           0           1               20       16 
    1           16           3          21               35       16
    1           16           4          36               40       16
----------------------------------------------------------------------
Obs ID:              1 Obs Type:          2 Satellite ID:    15
Latitude:       0.000 Longitude:     0.000 Elevation:        0.0
Surface Type:       1 Sat Zen Angle:   0.000 Solar Zen Ang:   38.000
Brightness Temperatures:
      226.717      218.949      219.645      224.452      236.374      247.804
      257.609      274.562      247.897      269.388      248.331      226.588
      261.211      252.591      241.597      236.591      266.911      271.983
      271.537     -9999.00      152.095      152.105      209.206      245.611
      244.334      233.650      226.083      220.954      216.761      215.937
      217.657      221.861      233.096      245.561      200.407      206.016
      235.261      242.110      252.165      261.941
Obs ID:              2 Obs Type:          2 Satellite ID:    16
Latitude:       0.000 Longitude:     0.000 Elevation:        0.0
Surface Type:       1 Sat Zen Angle:   0.000 Solar Zen Ang:   38.000
Brightness Temperatures:
      225.552      219.724      219.573      226.521      237.666      246.276
      255.913      267.355      250.477      259.886      251.149      229.323
      260.399      248.081      240.887      237.802      264.660      266.901
      266.492     -9999.00      152.369      156.488      217.802      243.712
      243.308      234.084      225.322      220.200      217.276      217.978

Number of Observations in File :    980
No. of Chans per Observation:    8461
Number of instruments making up observations : 1
*** In the following Series, Platform and Instrument are defined  *** 
*** according to the relevant RT Model definitions (if required): ***
Sat. Series   Platform   Instrument First_Channel   Last_Channel  Sat ID
    1            1           1           1             8461        300
----------------------------------------------------------------------

The "Channels" option is used for those cases where only a subset of available instrument channels are required. After this the instrument channels that are actually being used are listed. This is purely to allow the correct channels to be set up in the RT code and in all other places (except the R-matrix - see below) the n input channels a re referred to as channels 1 ... n.

After the headers lines, each observation is listed, prefixed by a sub-header detailing auxiliary data for the observation in question such as observation number, latitude, longitude, satellite view angle, etc.

Please note once more that the colons in the header and sub-header lines are used by the program as delimiters and must be included. Please also note that the second (optional) sub-header line is required when using an emissivity atlas with RTTOV-10 or RTTOV-11. In all other situations it may be omitted. Following the sub-header lines, the observed brightness temperatures are listed (free-format).

This file and the subroutine that reads it may be replaced by one in a format defined by the user.

[ Return to Top of Section ]

3d. Data Files: Background.dat

The Background.dat file starts with a 10 header lines that are reserved for user's comments.

The next three lines are general information for the file:
The first of these lines is the number of profiles contained in the file. This may either be unity (in which case the same background is used for all observations) or be the same as the number of observations (i.e., one background per observation).
The second line is the number of levels for each profile. In the current implementation this should be set to 43.
The third and final line defines the choice for the units for water vapour abundance:

The minimisation itself is always performed using the natural logarithm of the mass mixing ratio q (kg/kg), Ln(q).

Each profile is then listed with three comment lines preceeding each. The profiles are presented in four columns in free-format (pressure in Pa, temperature in Kelvin, Water Vapour in the predefined units and Ozone in ppmv with respect to dry air). Finally surface information is provided (each entry being preceeded by a label ending with a colon). The surface parameters provided, in order, are: Surface Temperature (K), Surface Humidity (in the pre-defined units), Skin Temperature(K), Surface Pressure (Pa), 10m U-Wind (m/s), 10m U-Wind (m/s).

If microwave cloud liquid water retrievals are to be performed (either LWP retrieval or qtotal retrieval), a fifth column containing cloud liquid water profiles (in kg/kg) has to be provided.

[ Return to Top of Section ]

Auxiliary Files:

Auxiliary data found in the xxx_COEFFS_DIR directories include the observation error covariance matrix (in Rmatrix.dat), two background error covariance matrices (in Bmatrix_43L.dat & Bmatrix_51L.dat), and the channel selection file (Channelchoice.dat). "xxx" refers to the observation type which is currently IASI, ATOVS, or AIRS.

3e. Auxiliary Files: Rmatrix.dat

The Rmatrix.dat file contains the observational error covariance matrix used by the 1DVar scheme. Currently one error covariance is used for all observations being processed (for a given instrument, of course) with the values being given in brightness temperature. There is no correction for scene temperature in the code, so the R-matrix values should be appropriate for the expected scene temperature of the channel in question. The forward model error is included in the observational error. This matrix must be positive definite or a fatal error will result.

Three options are available for storing the observational errors. These are full matrix, band-diagonal (of which diagonal is a subset), and eigenvalue/eigenvector. In the case of a diagonal R-matrix, the values in the file represent the variances of the observation errors. The band-diagonal option has been included as one way of representing the errors of apodised interferometer observations, while the eigenvalue format is in anticipation of more sophisticated ways of representing the errors of instruments with many channels and has not been fully tested as yet.

The file is formatted as follows (all entries are in free-format):

The first line describes the instrument (or "Composite Instrument") for which the file is valid. If "Composite Instruments" were set up in the observation file, the names in the observation and R-matrix files should agree. If composite instruments were not set up, the first R-matrix is read in.

The second line is made up of four integers which are (in order):

NOAA-15 ATOVS
    2    40     1     0 
       1       2       3       4       5       6       7       8       9
      10      11      12      13      14      15      16      17      18
      19      20       1      2        3       4       5       6       7 
       8       9      10      11      12      13      14      15       1
       2       3       4       5      
      1.22 0.5  0.5  0.5  0.5  0.5  1.22 6.32 1.4  6.32      
      3.0  4.0  1.22 1.22 0.5  0.4  0.4  0.4  0.4  9.9  
      2.0  2.0  2.0  1.26 0.25 0.25 0.25 0.25 0.4  0.4
      0.5  0.95 1.22 1.22 3.0  8.0  5.0  4.0  4.0  4.0

3f. Auxiliary Files: Bmatrix_43L.dat & Bmatrix_51L.dat

The Bmatrix_??L.dat files contains the background error covariance matrices used by the 1DVar scheme. The full matrix is specified. The number (??) refers to the number of levels the matrix is valid on. In this implementation, there are two B-Matrices specified in each file - one for land and one for sea - but there is no other variability in this matrix allowed for (e.g., no latitudinal variation).

The file itself simply consists of two comment lines followed by the dimension of the matrix (integer). The next lines are the B-Matrix in free format with one line per row/column (the matrix is symmetric).

For the second, land, matrix the same format is repeated (i.e., two comment lines, the number of elements and the matrix itself).

[ Return to Top of Section ]

3g. Auxiliary Files: ChannelChoice.dat

The ChannelChoice.dat file contains the choice of channels to be processed. The first line in the file contains Num_Channels, the number of channels for which selection codes (for cloud detection, retrieval, monitoring) are specified. The following Num_Channels lines in the file contain three columns of integers (plus a fourth, optional, column which can include characters). These columns are:

          Bit Number        Sounding Option
             1                  Surface = Sea
             2                  Surface = SeaIce
             3                  Surface = Land
             4                  Surface = Highland
             5                  Surface = Surface Mismatch
             6                  Cloud = Clear
             7                  Cloud = IRCloudy
             8                  Cloud = MWCloudy *
             9                  Cloud = Rain *
            10                  Cloud = High Cloud

In the following example file the selection codes for 11 channels are specified. HIRS 5-12 are used for retrievals in clear cases and AMSU-A 3-5 in all cases. HIRS-8 is also used as the window channel:

     11
           5    33    3  HIRS-5
           6    33    3  HIRS-6 
           7    33    3  HIRS-7
           8    33   -3  HIRS-8
           9    33    3  HIRS-9
          10    33    3  HIRS-10
          11    33    3  HIRS-11
          12    33    3  HIRS-12
          23  1023    3  AMSU-3
          24  1023    3  AMSU-4
          25  1023    3  AMSU-5

[ Return to Top of Section ]

[ Return to Top]

Output Files:

4a. Retrieved_Profiles.dat
4b. Retrieved_BTs.dat
4c. Additional Output: Minimisation.log and Minimisation_BTs.log
4d. Additional Output: A-Matrix and Am-Matrix

There are two main output files, Retrieved_BTs.dat and Retrieved_Profiles.dat. These are ASCII files and their contents should be obvious from inspection.

Additional diagnostic files are produced if the program is run in DebugMode or higher (set GeneralMode to 30 or more in ControlData.NL). These files are A-Matrix.out, Am-Matrix.out, Minimisation.log, and Minimisation_BT.log.

4a. Output: Retrieved_Profiles.dat

The Retrieved_Profiles.dat file contains the profiles retrieved from the 1DVar stage. If the 1DVar retrieval is not performed (e.g., the field of view is cloudy but only clear retrievals are allowed), the profiles are not written to this file; if the retrieval is performed but does not converge the final retrieved profile is written to this file, however (the number of iterations will be one higher than MaxIterations in this case).

The file is written in ASCII and should be self-explanatory. Both background and retrieved fields are supplied. The retrieved fields that are output are: Temperature Profile, Humidity Profile (in the same units as set up in the Background.dat file), Ozone profile, surface temperature, surface humidity (in same units as in the background file), skin temperature and surface pressure. If a Cloudy Retrieval is being done, cloud top pressure and cloud fraction are also output. If a microwave LWP retreival is being done, that will also be output. In addition the observation number, number of iterations, normalised cost function and normalised cost function gradient are reported.

An abridged example of the contents of this file is:

 Observation =  1
                          Retrieval                        Background
  Pressure (Pa)  T (K)    q (ppmv)     Ozone        T (K)   q (ppmv)      Ozone
     101325.   272.132  0.4296E+04  0.2456E-01   272.070  0.4280E+04  0.2456E-01
     100543.   271.952  0.4227E+04  0.2507E-01   271.854  0.4220E+04  0.2507E-01
      98588.   271.509  0.4061E+04  0.2643E-01   271.311  0.4074E+04  0.2643E-01
          .         .         .          .           .          .         .
          .         .         .          .           .          .         .
          .         .         .          .           .          .         .
         69.   265.252  0.5636E+01  0.6042E+01   265.741  0.5636E+01  0.6042E+01
         29.   256.937  0.5410E+01  0.5971E+01   257.553  0.5410E+01  0.5971E+01
         10.   241.249  0.4361E+01  0.5792E+01   241.635  0.4361E+01  0.5792E+01
Surface Temperature (K):         272.935   272.070
Surface Humidity (ppmv):        5202.160  4279.650
Skin Temperature (K):            272.935   272.070
Surface Pressure (Pa):           101325.  101325.

No. of Iterations:               2
Normalised Cost Function:       0.654 Normalised Gradient:       0.001
 --------------------------------------------------------------
 Observation =  2
                           Retrieval                        Background
  Pressure (Pa)  T (K)    q (ppmv)     Ozone        T (K)   q (ppmv)      Ozone
     101325.   271.698  0.4169E+04  0.2456E-01   272.070  0.4280E+04  0.2456E-01
     100543.   271.480  0.4110E+04  0.2507E-01   271.854  0.4220E+04  0.2507E-01
      98588.   271.008  0.3965E+04  0.2643E-01   271.311  0.4074E+04  0.2643E-01
          .         .         .          .           .          .         .
          .         .         .          .           .          .         .
          .         .         .          .           .          .         .
        261.   242.475  0.4945E+01  0.6106E+01   242.500  0.4945E+01  0.6106E+01
        142.   256.285  0.5429E+01  0.6080E+01   256.212  0.5429E+01  0.6080E+01
         69.   265.855  0.5636E+01  0.6042E+01   265.741  0.5636E+01  0.6042E+01
         29.   257.713  0.5410E+01  0.5971E+01   257.553  0.5410E+01  0.5971E+01
         10.   241.273  0.4361E+01  0.5792E+01   241.635  0.4361E+01  0.5792E+01
Surface Temperature (K):         268.230   272.070
Surface Humidity (ppmv):        4559.714  4279.650
Skin Temperature (K):            268.230   272.070
Surface Pressure (Pa):           101325.  101325.

No. of Iterations:               2
Normalised Cost Function:       0.782 Normalised Gradient:       0.000
 --------------------------------------------------------------

4b. Output: Retrieved_BTs.dat

The Retrieved_BTs.dat file contains the brightness retrieved from the 1DVar stage together with the observed brightness temperatures and those calculated from the background profile. If the 1DVar retrieval is not performed (e.g., the field of view is cloudy but only clear retrievals are allowed), the brightness temperatures are not written to this file; if the retrieval is performed but does not converge the final retrieved brightness temperatures are written to this file, however.

Each observation has three headers: the observation number, the number of channels used in the retrieval and the column titles. Four columns are then reported: the channel number, the background brightness temperature, the observed brightness temperature and the retrieved brightness temperature. The channel numbers reported are those specified in the first column of the ChannelChoice.dat file. Only those channels used in the retrieval are reported.

The following is a typical Retrieved_BTs.dat produced from a retrieval using HIRS 1-19 and AMSU-13:

  Observation =  1
 Number of Channels Used =  20
 Channel  Background  Observed   Retrieved
      1    226.579    226.717    226.167
      2    219.569    218.949    219.310
      3    219.321    219.645    219.153
      4    225.255    224.452    225.000
      5    237.027    236.374    236.813
      6    247.197    247.804    247.077
      7    256.419    257.609    256.574
      8    271.211    274.562    271.996
      9    250.658    247.897    251.042
     10    269.220    269.388    269.773
     11    251.843    248.331    250.209
     12    232.814    226.588    230.007
     13    261.277    261.211    261.721
     14    250.869    252.591    250.952
     15    241.685    241.597    241.568
     16    237.392    236.591    237.163
     17    266.360    266.911    266.937
     18    270.537    271.983    271.343
     19    271.088    271.537    271.877
     32    223.373    221.861    222.736

4c. Additional Output: Minimisation.log and Minimisation_BT.log

These files are output when the program is run in DebugMode or VerboseMode. They are used primarily to aid in bebugging by tracking the iteration-to-iteration progress of the minimisation.

Minimisation.log outputs the elements of the retrieval vector (RT_Params%RTGuess(:) in IASI_Minimize) together with the Marquardt-Levenberg gamma factor (zero if not using Marquardt-Levenberg minimisation) and the (un-normalised) cost function for each iteration. The exact makeup and order of the retrieval vector will vary according to the variables that are to be retrieved, but Fig. 1 gives the makeup of this vector in the default set-up supplied.

Minimisation_BT.log outputs the observed-guess brightness temperature differences for the channels used in the retrieval together with the Marquardt-Levenberg gamma factor (zero if not using Marquardt-Levenberg minimisation) for each iteration.

[ Return to Top of Section ]

4d. Additional Output: A-Matrix and Am-Matrix

The analysis error covariance matrix and the propagated measurement noise matrix are output to A-Matrix and Am-Matrix respectivly. These are the expected retrieval error and the expected contribution to the retrieval error from observation noise respectively as calculated through linear retrieval theory. They are output in the same order as the elements in the input B-matrix and are output for each iteration. The A-matrix is also calculated from the background profile for the first observation only.

For a full discussion of the derivation of the propagated measurement error matrix see Rodgers (1990) where it is referred to as the measurement error covariance, S_M.

[ Return to Top of Section ]

[ Return to Top ]

5. Setting Up a New Observation Type

If a fastmodel other than RTTOV, Gastropod or RTIASI is required, the interface with the new fastmodel will need to be coded up and called from IASI_Fastmodel_Interface.f90. Follow the example of one of the the step-by-step guide in Appendix D.

Additional retrieval variables will require changes to the fastmodel interfaces (at least for the fastmodels to be used), IASIMod_RTModel, IASI_Read_Background, IASI_SetUpBackground, IASI_SetUpRetrievals, IASI_TranslateDataOut and probably IASI_CheckIteration.

Contact nwpsaf@metoffice.gov.uk for questions on changing the code.

[ Return to Top ]

6. Notes, bugs and features

There are many user-defined parameters in this code. Some of these parameters may not have been optimised for the users requirements. This particularly applies to channel selection and and cloud detection channels and thresholds. The user is invited to critically review these.

7. Getting Help

8. Reference.

Rodgers, C.D. (1990). Characterization and error analysis of profiles retrieved from remote sounding measurements. J. Geophys. Res., 95, 5587-5595.

[ Return to Top ]

9. Acknowledgements

This work was aided greatly by the advice and input of Andrew Smith, Matthew Szyndel, Roger Saunders (all of the Met Office), Reinhold Hess (DWD), Vanessa Sherlock (NIWA) and Yoshiaki Takeuchi (JMA).

[ Return to Top ]

Appendices

Appendix A. Structure of Code
Appendix B. Minimization
Appendix C. Cloudy Retrievals
Appendix D. Adding a New Radiative Transfer Model