Here we present a Principal Components (PCs) method of retrieval of the HDO/H2O vertical profile using atmospheric
radiances observed from space by sensor like IMG as well as atmospheric transmittance spectra observed by ground
based FTIR. The method is based on the expansion of the retrieved profile on eigenvectors of covariance matrix of
model profiles extracted from the isotopic Atmospheric General Circulation Model (AGCM). A priori information of
covariance matrix compensates partially the lack of information containing in weighting functions for HDO in lower
atmospheric layer (0-1 km) and layers above 10 km. Error estimation of the retrieval scheme was made using closed
model computations with synthetic spectra ofIMG and known sets of T, H20, HDO profiles and its value is within 8% -
70% for vertical profile and not greater than l8% for columnar value of HDO/H20 ratio. The method was applied to
IMG/ADEOS spectra measured over the ocean in clear sky conditions. Latitudinal distributions ofHDO/H20 profile and
columnar HDO/H20 ratio are retrieved over Pacific Ocean for the time interval from winter of 1996 to summer of 1997.
The retrieved HDO/H20 from IMG/ADEOS data and simulated with isotopic AGCM are in a good agreement.
FTIR is Poker Flat high resolution ground based Fourier transform spectrometer for up-looking observation of
atmosphere in the spectral range from 750-4300 cm-1 with resolution 0.0019 cm-1 and high signal to noise ratio. The
spectrometer is located at the Poker Flat Research Range (Altitude 0.61km; latitude 65.11N; longitude 147.42W) of the
Geophysical Institute at the University of Alaska Fairbanks. Poker Flat FTS is operating from 1999, observation modes
are atmospheric emission and solar radiation absorption. The measured atmospheric traiismittances are supported by
sonde observations of T and water vapour profiles. The HDO/H20 PCs retrieval method was also adapted for using the
high spectral resolution atmospheric transmittances observed by FTIR. Linear regression of PCs of the HDO/H20
profiles was obtained in this case. Error estimation of the retrieval scheme was made using closed model computations
with synthetic spectra ofthe FTIR and known sets of T, H20, HDO profiles and its value is within 6% - 67% for vertical
profile but not greater than lO% for columnar value of HDO/H20 ratio. As an example, HDO/H20 vertical profiles were
retrieved using a few samples of FTIR spectra observed at the Poker Flat Research Range from 2000 to 2004 and
compared with isotope AGCM outputs for Alaska's atmosphere.
Multilayer perceptron (MLP) as universal approximator may be used for fast retrieval of atmospheric parameters such as
vertical profiles of temperature, humidity and concentration of absorbing gases from high-resolution infrared spectra
measured by satellite sensors. On the one hand, the number of spectral channels even necessary for retrieval of particular
atmospheric parameter is very high, so practical use of MLP needs for effective compression of spectral data with
tolerable loss of accuracy. On the other hand, algorithm of error back propagation becomes more effective if the input
data vector contains uncorrelated values with zero means, their covariance are approximately equal, and information
content of training set is maximized. The modified method of principal components (or empirical orthogonal functions
expansion) satisfies to all above requirements. The MLP may be constructed using relevant truncated vectors of principal
components as input and output data. Such MLP has fewer dimensions (the number of input, output and hidden neurons)
and requires less time for training than MLP using the high-resolution spectrum as input vector and set of vertical
profiles of atmospheric parameters as output vector.
The developed technique was applied to AIRS observations to retrieve temperature, humidity and methane content. The
empirical orthogonal functions were obtained as eigenvectors of matrix G = Se-1/2SRSe-1/2, where SR is sample
covariance matrix built on real AIRS measurements over given region, and Se is error covariance matrix characterizing
the sensor. The set of measured and model profiles as well as surface temperature and pressure were used for
construction of empirical orthogonal functions to represent output data of MLP as truncated expansion. Error profiles
and examples of temperature and methane maps are presented.