Coefficient file detailed history

RTTOV coefficient file history log

This page documents the history of changes to coefficient files so that you can understand differences between old and new versions. Only changes since May 2014 are documented. Changes that apply to many coefficient files appear at the top of the page followed by file histories for individual files grouped by sensor type. Individual file histories are linked from the main RTTOV coefficients page in the “Date of file creation” column of the coefficient tables.


General coefficient updates

21/02/2017 – RTTOV v12.1 release

  • All IR optical depth coefficients regenerated using an updated set of training profiles. The reference (background) profiles for CO2, N2O and CH4 reflect contemporary values and the range of training profiles for variable CO2, N2O and CH4 coefficients now encompass concentrations appropriate to the period 1970-202x.
  • All IR optical depth coefficients are available with both v7 predictors (variable O3) and v8 predictors (variable O3+CO2).
  • The v9 predictor visible+IR optical depth coefficients now allow variable O3+CO2.
  • New variable SO2 coefficients are available for hi-res IR sounders.
  • New coefficients are available for some previously unsupported sensors:
    • MTG-LI
    • Meteor-M2 IRFS2
    • Meteor SI
    • Nimbus-6 PMR and SCAMS
    • FY-3D MWHS2, MWRI and MWTS2
  • Coefficients have been updated for some sensors:
    • MTG-FCI – new SRF data available
    • MODIS – changes in Planck-weighting of some channels (Planck-weighted channels are now 4, 5, 6, 10 for v7 predictors and 4, 5, 6, 7, 8, 10 for v8 and v9 predictors)
    • SSMI/S Zeeman – modified set of pressure levels, see file headers for LBL/training information
  • The NLTE bias correction model has changed in v12 and so the NLTE coefficients have been updated in hi-res IR sounder coefficient files. Currently NLTE coefficients are available only for IASI in v12.
  • New IR and visible+IR cloud/aerosol coefficients are available for v12.
  • New PC-RTTOV coefficients are available for IASI which are compatible with the NLTE bias correction.

26/05/2016, all non-zeeman MW coefficients

A major review of the line by line code and procedures has been undertaken, and, as a consequence, a number of bug fixes and improvements have been carried out as detailed below. See the links to the statistics plots below for an indication of the impact on simulated brightness temperatures. New Zeeman coefficients to follow.

MW line by line bug: A bug was discovered in the microwave line by line model which affects a number of sensors. The largest errors are seen in ATMS, MWS and SSMIS where the bug leads to biases of approximately 0.5K in brightness temperature in some channels. The affected sensors are:

  • ATMS
  • ICI
  • GMI
  • MTVZAGY
  • MWHS2
  • MWI
  • MWS
  • SSMIS

183GHz halfwidth T-dependence: In the previous (Sep/Nov 2015) set of MW coefficients the halfwidth of the 183GHz water vapour line used in the line by line model was updated to that in Payne et al (2008); the corresponding temperature dependence of the halfwidth, which was not updated alongside this, has now been installed. This results in differences of order 0.1K in brightness temperature for 183GHz channels.

MWS filter file: The filter file for MWS has been updated according to the latest available information from ESA, with significant changes – MW filter files can be found here.

MHS on NOAA coefficients: Due to a bug in the coefficient generation procedure code, the FILTER_FUNCTIONS header section in the previous versions of the coefficient files rtcoef_noaa_18_mhs.dat and rtcoef_noaa_19_mhs.dat listed specifications for the AMSUB instrument, for which channels 2 and 5 are significantly different (the coefficients themselves were unaffected by this). This has now been corrected.

Other minor changes: The spectral resolution used for the channel averaging of transmittances has been increased in many sensor channels, resulting in some minor differences. Filter files have been reviewed and small errors caused by output precision corrected. The filter file for SSMIS has had minor changes made, so that it now coincides with the specifications given for sensor SN01 (DMSP-20) in Table 4.1 of this JCSDA Office Note, with average bandwidths taken for channels 19-24 where different measured values are given for each passband.

The pages linked below show plots of the mean and RMS difference between brightness temperatures calculated using the new RTTOV coefficients (May 2016) and the previous RTTOV coefficients (Sep/Nov 2015). The brightness temperatures are computed for nadir view (zenith angle 0 degrees) with unit surface emissivity and comprise upwelling atmospheric emission plus the surface emission term. The statistics are calculated for the 83 diverse profile set used to train RTTOV and give an indication of the difference between the new and old coefficients expressed in terms of TOA brightness temperature.

25/09/2015, all IR v7 and v8 predictor coefficients including hi-res IR sounders

A bug was discovered in the coefficient generation related to the 2m water vapour variable which affects all IR v7 and v8 predictor coefficients generated since RTTOV v11.1. The impact of this on simulated brightness temperatures is illustrated in this document. In addition coefficients for some channels of HIRS, IRAS and MODIS are now Planck-weighted as this improves the fit of the optical depth regression:

  • NIMBUS 6 HIRS: channels 9-16 are now Planck-weighted
  • NOAA 5-10, 12 HIRS: channels 10 and 12-16 are now Planck-weighted
  • NOAA 11, 14 HIRS: channels 12-17 are now Planck-weighted
  • NOAA 15-19 HIRS: channels 13-17 are now Planck-weighted
  • METOP 1, 2 HIRS: channels 13-17 are now Planck-weighted
  • IRAS: channels 14-18 are now Planck-weighted
  • MODIS: channels 4-6 (v7/v8 predictor) and channels 23-25 (v9 predictor) are now Planck-weighted
  • For “shifted” coefficients the same channels are Planck-weighted as in the non-shifted files

25/09/2015, all MW coefficients

As described in this document, all MW coefficients were regenerated for the release of RTTOV v11.3 to fix some bugs and introduce improved spectroscopy.

  • Optical depth predictors are now calculated from water vapour in units of ppmv over dry air to be consistent with the IR coefficients.
  • The treatment of gas units was not consistent within the coefficient generation process and this has been fixed.
  • Updates to the oxygen line parameters and the 183GHz line half-width.

In addition some changes were made to a few individual instrument coefficient files:

  • MWS: some channel pass bands have been updated.
  • AMSR-E and AMSR2: very small changes in some channel central wavenumbers due to change in precision (negligible impact).
  • SSMIS on DMSP-19: satellite height changed from 850km to 852.8km (negligible impact).

Hi-res IR sounders


rtcoef_metop_2_iasi_so2.H5, v9 predictors, 101L

  • 20/03/2017
    • There was an error in the calculation of the original SO2 coefficients. The updated coefficients give an improved fit to the line-by-line training data.


rtcoef_metopsg_1_iasing_so2.H5, v9 predictors, 101L

  • 20/03/2017
    • There was an error in the calculation of the original SO2 coefficients. The updated coefficients give an improved fit to the line-by-line training data.


rtcoef_nimbus_4_iris.H5, v8 predictors, 101L

  • 24/06/2015
    • LBL spectral range extended to cover all IRIS channels
    • there are some small differences in output radiances in channels within the previous LBL spectral range (up to 20 microns) due to changes in the precision used in the optical depth regression


rtcoef_clarreo_1_mrfirs.H5, v7/v8 predictors, 54L

  • 01/03/2017
    • Truncate spectral range of channel 1 so that it falls within the LBL transmittance database: previously the optical depth regression coefficients for channel 1 were all zero

IR sensors


rtcoef_goes_16_abi.dat, v7 predictors, 54L

  • 31/08/2016
    • updated channel central wavenumbers and SRFs


rtcoef_himawari_8_ahi.dat, v7 predictors, 54L

  • 11/06/2014
    • updated channel central wavenumbers and SRFs
    • channel 7 (index 1 in file) is no longer Planck-weighted
    • channel 9 (index 3 in file) is now Planck-weighted


rtcoef_envisat_1_atsr-shifted.dat, v7 predictors, 54L

  • 01/12/2017
    • shift in 12µm channel is now 40nm instead of 50nm


rtcoef_iss_1_ecostress.dat, v7 predictors, 54L

  • 01/09/2016
    • updated channel central wavenumbers and SRFs


rtcoef_mtg_1_fci.dat, v7 predictors, 54L

  • 11/06/2014
    • channel 10 (index 2 in file) is no longer Planck-weighted
    • there are some small differences in output radiances due to changes in the precision used in the optical depth regression


rtcoef_sentinel3_1_slstr.dat, v7 predictors, 54L


Visible/IR sensors


rtcoef_goes_16_abi.dat, v9 predictors, 54L

  • 01/09/2016
    • updated channel central wavenmbers and SRFs
  • 11/06/2014
    • thermal emission no longer calculated for channel 6 (ss_val_chn=2)
    • there are some small differences in output radiances due to changes in the precision used in the optical depth regression


rtcoef_himawari_8_ahi.dat, v9 predictors, 54L

  • 11/06/2014
    • updated channel central wavenumbers and SRFs
    • channel 7 is no longer Planck-weighted
    • channel 9 is now Planck-weighted


rtcoef_eos_1_aster.dat, v9 predictors, 54L

  • 16/11/2015
    • coefficients for IR channels are trained for zenith angles up to ~65 degrees instead of ~84 degrees as for the VIS/NIR channels (reduces errors and larger angles are not required for IR channels on instruments in LEO)


rtcoef_ers_1_atsr.dat, v9 predictors, 54L

  • 02/09/2014
    • thermal channels trained over 6 secants instead of 14 secants


rtcoef_ers_2_atsr.dat, v9 predictors, 54L

  • 02/09/2014
    • thermal channels trained over 6 secants instead of 14 secants


rtcoef_envisat_1_atsr.dat, v9 predictors, 54L

  • 02/09/2014
    • thermal channels trained over 6 secants instead of 14 secants


rtcoef_envisat_1_atsr-shifted.dat, v9 predictors, 54L

  • 01/12/2017
    • shift in 12µm channel is now 40nm instead of 50nm


rtcoef_noaa_15_avhrr.dat, v9 predictors, 54L
rtcoef_noaa_16_avhrr.dat, v9 predictors, 54L
rtcoef_noaa_17_avhrr.dat, v9 predictors, 54L
rtcoef_noaa_18_avhrr.dat, v9 predictors, 54L

  • 06/09/2015
    • coefficients for IR channels are trained for zenith angles up to ~65 degrees instead of ~84 degrees as for the VIS/NIR channels (reduces errors and larger angles are not required for IR channels on instruments in LEO)


rtcoef_noaa_19_avhrr.dat, v9 predictors, 54L

  • 17/07/2014
    • add missing 1.6 micron channel
    • coefficients for IR channels are trained for zenith angles up to ~65 degrees instead of ~84 degrees as for the VIS/NIR channels (reduces errors and larger angles are not required for IR channels on instruments in LEO)


rtcoef_metop_1_avhrr.dat, v9 predictors, 54L
rtcoef_metop_2_avhrr.dat, v9 predictors, 54L

  • 06/09/2015
    • coefficients for IR channels are trained for zenith angles up to ~65 degrees instead of ~84 degrees as for the VIS/NIR channels (reduces errors and larger angles are not required for IR channels on instruments in LEO)


rtcoef_mtg_1_fci.dat, v9 predictors, 54L

  • 11/06/2014
    • channel 10 is no longer Planck-weighted
    • thermal emission no longer calculated for channel 8 (ss_val_chn=2)
    • there are some small differences in output radiances due to changes in the precision used in the optical depth regression


rtcoef_fy3_1_iras.dat, v9 predictors, 54L

  • 04/09/2015
    • channels 14-18 are now Planck-weighted


rtcoef_eos_1_modis.dat, v9 predictors, 54L

  • 04/09/2015
    • channels 23-25 are now Planck-weighted
  • 11/06/2014
    • thermal emission no longer calculated for channel 7 (ss_val_chn=2)
    • there are some small differences in output radiances due to changes in the precision used in the optical depth regression


rtcoef_eos_1_modis-shifted.dat, v9 predictors, 54L

  • 04/09/2015
    • channels 23-25 are now Planck-weighted
  • 11/06/2014
    • thermal emission no longer calculated for channel 7 (ss_val_chn=2)
    • there are some small differences in output radiances due to changes in the precision used in the optical depth regression


rtcoef_eos_2_modis.dat, v9 predictors, 54L

  • 04/09/2015
    • channels 23-25 are now Planck-weighted
  • 11/06/2014
    • thermal emission no longer calculated for channel 7 (ss_val_chn=2)
    • there are some small differences in output radiances due to changes in the precision used in the optical depth regression


rtcoef_eos_2_modis-shifted.dat, v9 predictors, 54L

  • 04/09/2015
    • channels 23-25 are now Planck-weighted
  • 11/06/2014
    • thermal emission no longer calculated for channel 7 (ss_val_chn=2)
    • there are some small differences in output radiances due to changes in the precision used in the optical depth regression


rtcoef_landsat_8_oli.dat, v9 predictors, 54L

  • 11/06/2014
    • thermal emission no longer calculated for channel 7 (ss_val_chn=2)
    • changes in channel central wavenumbers (in the 3rd significant figure or smaller)
    • there are some small differences in output radiances due to changes in the precision used in the optical depth regression


rtcoef_sentinel3_1_slstr.dat, v9 predictors, 54L

  • 08/10/2015
  • 11/06/2014
    • thermal emission no longer calculated for channel 6 (ss_val_chn=2)
    • changes in channel central wavenumbers (in the 4th significant figure or smaller)
    • there are some small differences in output radiances due to changes in the precision used in the optical depth regression


rtcoef_jpss_0_viirs.dat, v9 predictors, 54L

  • 16/11/2015
    • coefficients for IR channels are trained for zenith angles up to ~65 degrees instead of ~84 degrees as for the VIS/NIR channels (reduces errors and larger angles are not required for IR channels on instruments in LEO)
  • 11/06/2014
    • thermal emission no longer calculated for channel 15 (ss_val_chn=2)
    • small changes in channel central wavenumbers (in the 7th significant figure or smaller)
    • there are some small differences in output radiances due to changes in the precision used in the optical depth regression


rtcoef_fy2_3_vissr.dat, v9 predictors, 54L
rtcoef_fy2_4_vissr.dat, v9 predictors, 54L

  • 06/09/2015
    • coefficients for IR channels are trained for zenith angles up to ~65 degrees instead of ~84 degrees as for the VIS/NIR channels (reduces errors and larger angles are not required for IR channels on instruments in LEO)

MW sensors


rtcoef_noaa_18_mhs.dat, v7 predictors, 54L

  • 07/05/2014
    • central wavenumbers and band correction coefficients corrected (previously were taken from AMSU-B)
    • constants used in Planck function updated to be consistent with other coefficient files


rtcoef_noaa_19_mhs.dat, v7 predictors, 54L

  • 07/05/2014
    • central wavenumbers and band correction coefficients corrected (previously were taken from AMSU-B)
    • constants used in Planck function updated to be consistent with other coefficient files