Approach To High Spectral Resolution Infrared Remote Sensing Of Atmospheric Temperature And Humidity Profiles

As the rapid development of numerical weather forecast model and the increasingly concerns of globe change, atmospheric remote sounding are demanded for higher acuracy and virtical resolution of atmospheric profiles that can not obtained by means of current operational sounders. Current operational vertical sounding instrument has lower spectral resolution and relatively flat (i.e.broad) weighting functions, so that the sounder recieves emitted energy in each channel from a thicker layer of the atmosphere, thereby making the discrimination of fine-scale vertical details practically impossible. Estimated profiles from satellite data are of wider coverage on globe and superior spatial resolution in horizontal and short time intervals. The inability of current sounders to match the vertical resolution of general circulation models is the major deficiency and having it to be improved upon become even more important at present. United States and European Meteorological Satellite Organise take the lead in developing high spectral resolution sounders and China also has put her projects into schedule. Consequently, new methods and algorithms are urgently needed for hyperspectral data processing and atmospheric parameters retrieving. For such a science object, the dissertation is devoted to both theoretical and computational approachs to retrieval of atmospheric temperature and humidity profiles from high spectral resolution infrared measurements. And the contributions of the dissertation are:(1) Principal component analysis method (PCA) which commonly used in current researches is discussed, and a prospective independent component analysis method (ICA) is intruduced to hyperspectral data pre-processing and estimation algorithms. ICA is capable of extracting higher order statistics (HOS) and dealing with signal of non-Gauss pdf and therefore more favorable for hyperspectral data processing.(2) Relationships between various physical retrieval methods are investigated. Advantages and disadvantages of them are analyzed. Concepts and Methods for solving atmospheric remote sensing problems are unified under the framework of optimization theory. This makes the algorithm studies of inverting hyperspectral radiances for atmospheric profiles a complete mathematic support and theoretical systematization. Inversion methods of Levernberg-Marquardt and trust region are put forward for the atmospheric temperature and humidity retrieval on the nonlinear optimization basis, which overcome the difficult of solving the problem while the Hessen being singular.(3) A new strategy of layer by layer retrieval of atmospheric temperature profile using neural network is proposed, which is more time economy and yields better results than conventional methods.(4) Sensitivity analysis of retrieval departures showed some useful characteristics of hyperspctral sensing of atmospheric profile. For example, the retrieval errors show spectral coverage dependence and various spectral band would impact on the accuracy of retrieved tepmerature at different atmospheric layers. Under the condition of covering the entire infrared spectral band, when channel numbers is greater than 100, the RMSE (Root Mean Square Errors) is less sensitivity to channel numbers; with less than 100 channels, we have inverse result. Additionally, some valuable results have been achieved in the studies of incorporating use of microwave channels, non-satellite fators and effects due to QingHai-Tibet plateau.(5) Effects of bandwidth and channel numbers to the retrieval accuracy have been investigated with given vertical resolution and then bring forward the minimal parameter requirements in developing our country's hyperspectral sounding instruments. The experiment results suggest that developing interferometer infrared atmospheric sounding instrument with 100 spectral band and with bandwidth of 2-4 wavenumber per centimeter would be able to provide significant improved profiles and meet the requirments of current observation objects.(6) Atmospheric environment parameters (Convective Available Potential Energy, Convective Inhibition) and instability indices are computed using different data involving the temperature and humidity fields retrieved from AIRS cloud cleared radiances, AIRS and ATOVS operationally retrieved temperature and humidity products and ECMWF reanalysis fields. Results suggest that the spatial resolutions (both in vertical and in horizontal) how affect the resullts in the computation of instability indices and reveal the importance of cloudy field-of-view in weather analyzing.(7) algorithm repertory for generating hyperspectral retrieved products.