Abstract
The time rate of change of a wide variety of geophysical variables (atmospheric temperature, humidity, clouds, etc.) can be derived from hyperspectral infrared sensors such as AIRS, now in operation for nearly 13 years. Traditional approaches retrieve geophysical variables on a per-footprint basis using some sort of non-linear inversion scheme that inherently is heavily dependent on a-priori information. Trends of interest to the climate community are then derived from time and space averaging of these 1D-var retrievals. An example of this product would be the AIRS Level 3 monthly averaged atmospheric products. For climate applications, a clear understanding of the error characteristics of these products is required, but the complexity (and non-linearity) of these retrieval systems is extremely problematic. We introduce here an approach to measuring climate-level variability with AIRS by instead doing time and space averaging in the radiance domain, and then converting to geophysical trends directly from averaged radiance trends. This approach makes error estimation much more straightforward since the final geophysical measurement is more directly related to the instrument stability (and inter-calibration if using more than one instrument). We will show results using this approach for all-sky retrievals of AIRS 10-year variability, and compare our results to the AIRS Level 3 data, the ECMWF ERA-Interim re-analysis and the NASA MERRA re-analysis.
Optimal Estimation Retrievals of Decadal Variability from AIRS Radiance Time Derivatives and Comparison to Re-Analysis Products
The time rate of change of a wide variety of geophysical variables (atmospheric temperature, humidity, clouds, etc.) can be derived from hyperspectral infrared sensors such as AIRS, now in operation for nearly 13 years. Traditional approaches retrieve geophysical variables on a per-footprint basis using some sort of non-linear inversion scheme that inherently is heavily dependent on a-priori information. Trends of interest to the climate community are then derived from time and space averaging of these 1D-var retrievals. An example of this product would be the AIRS Level 3 monthly averaged atmospheric products. For climate applications, a clear understanding of the error characteristics of these products is required, but the complexity (and non-linearity) of these retrieval systems is extremely problematic. We introduce here an approach to measuring climate-level variability with AIRS by instead doing time and space averaging in the radiance domain, and then converting to geophysical trends directly from averaged radiance trends. This approach makes error estimation much more straightforward since the final geophysical measurement is more directly related to the instrument stability (and inter-calibration if using more than one instrument). We will show results using this approach for all-sky retrievals of AIRS 10-year variability, and compare our results to the AIRS Level 3 data, the ECMWF ERA-Interim re-analysis and the NASA MERRA re-analysis.