|Title||An inter-comparison of line-by-line radiative transfer models|
|Publication Type||Technical memorandum|
|Secondary Title||ECMWF Technical Memoranda|
The fast RT model, RTTOV, used operationally at ECMWF is based on accurate transmittances generated by the GENLN2 line-by-line model. GENLN2 was adopted for use at ECMWF more than ten years ago and since then no major new versions of the code have been released. The long term maintenance of GENLN2 has now become an issue and it is apparent that there is no commitment to further develop the code. Consequently we have considered the possibility of a different choice of line-by-line model for the training of RTTOV. To this end we have compared the GENLN2 model with the LBLRTM and the RFM line-by-line models with the objective to quantify differences between the forward models and to assess the quality of the spectroscopic data used in the forward model computations by using molecular parameters from the HITRAN2000/2004, TES and GEISA2003 databases. Five test cases have been studied by comparing simulated spectra with spectra measured during the first and the third Convection and Moisture Experiment (CAMEX-1 and CAMEX-3) campaigns, the EAQUATE campaign, the MOTH campaign and with spectra measured at the Atmospheric Radiation Measurement (ARM) site. Results show that LBLRTM can perform calculations significantly faster than RFM and GENLN2 at a same or better level of accuracy. It can compute analytical Jacobians, can perform computations in presence of reflected solar radiation and incorporates the P-R branch line mixing for the CO2 ?2 band with a planned extension to the ?3 band. On this basis we adopted LBLRTM as the line-by-line code for the training of RTTOV. As for the choice of the molecular line parameters to be used in conjunction with LBLRTM, results show that none of the databases considered in the study give consistently better results across the whole infrared spectrum: different databases perform better in different regions of the spectrum; consequently, we adopted a molecular database that is the blend of the HITRAN2000/2004, GEISA2003 and TES.