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
Modern data assimilation algorithms depend on accurate infrared spectroscopy in or- der 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 critical to im- proving the application of satellite data to weather forecasting. Here we present the re- sults 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 Instrument (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 spectro- scopic 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 sig- nificantly improved in LBLRTM v12.1 relative to v9.4+, and that these spectroscopic updates lead to changes of ∼0.5K in the retrieved vertical temperature profiles below 10hPa, with the sign of the change and the variability among cases depending on alti- tude. 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 residuals in the P-branch of the H2Oν2 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 200hPa, 42%higher near 562hPa, 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 of the H2Oν2 band. There were no changes to O3 spectroscopy be- tween the two model versions, and so both versions gives positive residuals of ∼0.3K in the R-branch of the O3 ν3 band. While the updates to the H2O self continuum em- ployed by LBLRTM v12.1 have clearly improved the match with observations near the CO2 ν3 bandhead, we find that these updates have significantly degraded the match with observations in the fundamental band of CO. Finally, significant systematic 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.