Performance of the Line-By-Line Radiative Transfer Model (LBLRTM) for temperature, water vapor, and trace gas retrievals: recent updates evaluated with IASI case studies

Abstract. Modern data assimilation algorithms depend on accurate infrared spectroscopy in order to make use of the information related to temperature, water vapor (H2O), and other trace gases provided by satellite observations. Reducing the uncertainties in our knowledge of spectroscopic line parameters and continuum absorption is thus important to improve the application of satellite data to weather forecasting. Here we present the results of a rigorous validation of spectroscopic updates to an advanced radiative transfer model, the Line-By-Line Radiative Transfer Model (LBLRTM), against a global dataset of 120 near-nadir, over-ocean, nighttime spectra from the Infrared Atmospheric Sounding Interferometer (IASI). We compare calculations from the latest version of LBLRTM (v12.1) to those from a previous version (v9.4+) to determine the impact of spectroscopic updates to the model on spectral residuals as well as retrieved temperature and H2O profiles. We show that the spectroscopy in the CO2 ν2 and ν3 bands is significantly improved in LBLRTM v12.1 relative to v9.4+, and that these spectroscopic updates lead to mean changes of ~0.5 K in the retrieved vertical temperature profiles between the surface and 10 hPa, with the sign of the change and the variability among cases depending on altitude. We also find that temperature retrievals using each of these two CO2 bands are remarkably consistent in LBLRTM v12.1, potentially allowing these bands to be used to retrieve atmospheric temperature simultaneously. The updated H2O spectroscopy in LBLRTM v12.1 substantially improves the a posteriori residuals in the P-branch of the H2O ν2 band, while the improvements in the R-branch are more modest. The H2O amounts retrieved with LBLRTM v12.1 are on average 14% lower between 100 and 200 hPa, 42% higher near 562 hPa, and 31% higher near the surface compared to the amounts retrieved with v9.4+ due to a combination of the different retrieved temperature profiles and the updated H2O spectroscopy. We also find that the use of a fixed ratio of HDO to H2O in LBLRTM may be responsible for a significant fraction of the remaining bias in the P-branch relative to the R-branch of the H2O ν2 band. There were no changes to O3 spectroscopy between the two model versions, and so both versions give positive a posteriori residuals of ~ 0.3 K in the R-branch of the O3 ν3 band. While the updates to the H2O self-continuum employed by LBLRTM v12.1 have clearly improved the match with observations near the CO2 ν3 band head, we find that these updates have significantly degraded the match with observations in the fundamental band of CO. Finally, significant systematic a posteriori residuals remain in the ν4 band of CH4, but the magnitude of the positive bias in the retrieved mixing ratios is reduced in LBLRTM v12.1, suggesting that the updated spectroscopy could improve retrievals of CH4 from satellite observations.