Investigation On Calibration Technique And Remote Sensing Of Atmosphere By Ground-based

The remote sensing of atmosphere is important to the weather forecasting, meteorological, astronomy, communications, space observation correction, atmospheric research and other fields. Ground-based microwave radiometer is an effective method of passively and remotely sensing of the atmosphere. In this thesis, the microwave radiation property of atmosphere, the calibration of radiometer and the remote sensing of atmosphere by microwave radiometer have been investigated. The constant in the approximate relationship between liquid absorption and frequency is improved to be a variable related with the temperature of cloud base. The accurate model of liquid absorption related with the temperature of cloud base is established. And then the improved tipping calibration under cloudy sky conditions is deduced which makes standard tipping calibration to be used under all-weather conditions. Further more, the remote sensing of atmosphere by radiometer and other detection devices under cloudy sky conditions are emphatically investigated. Main works of this thesis are as follows:1. The standard tipping calibration flow which applies to all ground-based microwave radiometers is investigated and an improved tipping calibration which applies to cloudy sky conditions is proposed. The improved tipping calibration makes use of the relationship between liquid absorption and frequency to establish the combination form of two frequencies which eliminates the error caused by horizontal nonhomogeneity of liquid water. The standard tipping calibration is developed into a all-weather method.2. The statistical method to retrieve atmospheric parameters by ground-based radiometer is analyzed. Simulation and measurements are done for retrieving water vapor, temperature and refractivity. According to the method of linear regression, an improved linear regression arithmetic to retrieve the atmospheric refractivity is proposed. By using the definition of refractivity, the dry and wet item of the refractivity are set as the input of surface parameter which can clearly reflect the relativity between the atmosphere refractivity and surface parameter better. The results show that the retrieving precision of the improved linear regression arithmetic is generally increased below 5 km especially for the atmosphere near ground.3. The standard neural network method to retrieve atmospheric parameters is analyzed. The atmosphere is divided into several layers and all layers are set as output to be trained by neural network. This model is according to the mutual impact between different atmospheric layers. In fact, the relationship between different layers is not the same. The further of the distance, the less of the relativity. This leads to some errors for the method of standard neural network. Based on the standard back-propagation artificial neural network, a subsection modeling method is proposed to retrieve atmospheric parameters which effectively uses the relationship between different layers. The results of profiling temperature and water vapor indicate that the retrieval accuracy can be improved by subsection modeling.4. Varied liquid water models are compared and the effect of liquid water radiation to remote sensing of atmosphere is analyzed. Based on the traditional method of modeling liquid water, a new method which is not based on modeling cloud is proposed to retrieve the cloudy atmospheric parameters. All historical data(including cloudy days) are regarded as sunny days, thus just considering the absorption coefficients of oxygen and vapor but the absorption coefficient of liquid water. The combination form of dual-frequency or multi-frequency attenuations are established to be used for regression or training by neural network which can be used for retrieving. Due to eliminating the radiation of liquid water, the retrieval error caused by modeling cloud can be decreased.5. The method of remotely sensing cloudy atmosphere by radiometer combined with ceilometer and infrared thermometer are investigated. The standard orthogonal neural network technique expands the atmospheric profile by a set of natural orthogonal functions. The coefficients of the orthogonal function are estimated by neural network. The information of cloudbase height can be well exploited by this method in cloudy atmosphere. Based on the standard orthogonal neural network, a subsection orthogonal neural network technique is presented. The atmospheric profile is divided into several segments with corresponding coefficients of orthogonal functions. The subsection of coefficients matrix are estimated by neural network respectively and the retrieval accuracy of the method is higher than of standard method. The iteration method to detect the atmosphere by microwave radiometer combined with infrared thermometer is introduced. The training dataset is classified according to the cloudbase height. The information of cloudbase height and vapor density at cloudbase height is effectively used which can improve the retrieval accuracy. Finally, the method of Relevance Vector Machine and Neural Network by microwave radiometer combined with GNSS(Global Navigation Satellite Systems) is introduced. The result indicates that the retrieval accuracy of the combination model is higher than that of using a single instrument.