Remote sensing of the vertical distribution of atmospheric water vapor from the TOVS observations: Method and validation

This paper presents a method to remotely sense the vertical distribution of atmospheric water vapor using spaceborne measurements from the TOVS instrument aboard the NOAA polar satellite series. It describes a new approach to the water vapor retrieval scheme in the improved initialization inversion (31) method, The technique is based on a neural network scheme, which is analyzed theoretically. Cross-comparisons of its results with a wide variety of independent observations (in situ measurements or other global data sets, e.g., the special sensor microwave/imager (SSIM/I) retrievals, analyses) are then carried out to fully evaluate the method. It is shown that the mean of the differences between total water vapor contents obtained from each data set represents less than 20% of the global mean value of the water vapor content. Different behaviors between TOVS and SSM/I in tropical situations are also highlighted. Concerning the vertical profile, the standard deviation between water vapor content retrieved by 31 and measured by radiosondes varies from 30% in the 1000-850 hPa laver to less than 40% in the 500-300 hPa layer. The vertical increase of the error is linked to the difficulty of measuring weak values by radiosonde instruments, radiometers, or analyses.