Improving forecast skill by assimilation of AIRS temperature soundings

This paper uses AIRS temperature profiles derived by the AIRS Science Team Version-5 retrieval algorithm. The AIRS Science Team Version-5 retrieval algorithm is being used operationally at the Goddard DAAC in the processing (and reprocessing) of all AIRS data. The AIRS Science Team Version-5 retrieval algorithm contains two significant improvements over Version-4: (1) Improved physics allows for use of AIRS observations in the entire 4.3 μm CO₂ absorption band in the retrieval of temperature profile T(p) during both day and night. Tropospheric sounding 15 μm CO₂ observations are now used primarily in the generation of cloud cleared radiances HR_i. This approach allows for the generation of accurate values of HR_i and T(p) under most cloud conditions. (2) Another very significant improvement in Version-5 is the ability to generate accurate case-by-case, level-by-level error estimates for the atmospheric temperature profile, as well as for channel-by-channel error estimates for HR_i. These error estimates are used for quality control of the retrieved products. We have conducted forecast impact experiments assimilating AIRS temperature profiles with different levels of quality control using the NASA GEOS-5 data assimilation system. Assimilation of quality controlled T(p) resulted in significantly improved forecast skill compared to that obtained from analyses obtained when all data used operationally by NCEP, except for AIRS data, is assimilated. We also conducted an experiment assimilating AIRS radiances uncontaminated by clouds, as done operationally by ECMWF and NCEP. Forecasts resulting from assimilating AIRS radiances were of poorer quality than those obtained assimilating AIRS temperatures.