|Title||The generation of RTTOV regression coefficients for IASI and AIRS using a new profile training set and a new line-by-line database|
|Publication Type||Technical memorandum|
|Secondary Title||ECMWF Technical Memorandum|
New RTTOV regression coefficients have been generated for IASI and AIRS from a database of line-by-line (LBL) transmittances computed using version 11.1 of the LBLRTM LBL model and a new set of 83 training profiles. The LBL database was computed utilizing molecular parameters obtained by merging in different regions of the infrared spectrum data available from the HITRAN2000, HITRAN2004 and GEISA2003 databases. To compute the database of LBL transmittances a new set of 83 training profiles has been selected from a dataset of atmospheric profiles of temperature, water vapor and ozone produced using the operational suite of the ECMWF forecasting system during the period July 2006-June 2007. The profiles of temperature, water vapor and ozone have been supplemented by profiles of CO2, CO and CH4 generated by a number of forecast experiments performed at ECMWF within the context of the GEMS project and by profiles of N2O generated by extrapolating to the surface profiles retrieved from radiances measured by the CLAES instrument. Results for the dependent set of profiles used to train the fast model show that RTTOV can reproduce LBL radiances to a degree of accuracy that is below 0.1 K rms for 98 % of the channels of IASI and AIRS. Errors larger than 0.1K are observed for a small fraction of ozone channels, for tropospheric water vapor channels with moderate absorption and for channels where absorption lines due to CO2, CH4, N2O and CO interfere with absorption lines due to H2O. Errors for a set of profiles independent to the regression coefficients are larger but still below 0.15 K rms for the vast majority of the channels. To improve the simulation of line-by-line transmittances in the region between 2000 and 2250 cm-1 we have replaced the standard RTTOV transmittance model with a new model that uses the predictors dedicated to the evaluation of the transmittances for the solar term only for those channels that are significantly contaminated by solar radiation.