【作者】 王波；

【导师】 赵振维； 赵犁丰；

【作者基本信息】 中国海洋大学 ， 环境科学， 2007， 硕士

【摘要】 大气环境作为无线电系统信息传输的途径,对信息系统的性能具有重要影响,大气参数剖面的遥感对评估无线电系统性能具有重要应用价值,同时,大气剖面的遥感还可用于气象预报,提高气象预报的准确度。本文系统论述了利用地基微波辐射计探测反演大气参数剖面的原理和反演技术。着重研究了双通道微波辐射计反演大气参数剖面、积分水气和云液态水含量的反演技术及反演算法,并对12通道微波辐射计反演大气参数剖面进行了仿真研究。本文的主要工作有:(1)在传统线性回归算法的基础上提出了非线性回归算法,通过增加回归方程的非线性项因素来提高反演大气剖面的精度,并利用青岛和北京的历史探空数据得到了两地区的非线性回归算法模型。在青岛进行了微波辐射计反演和无线电探空剖面的对比实验,实验结果表明,反演剖面与探空剖面有较好的一致性,且非线性回归算法反演水汽剖面较线性回归算法具有更好的反演精度。(2)由于神经网络算法在处理非线性问题方面具有强大的优势,本文给出了一种利用微波辐射测量反演大气参数剖面的BP神经网络算法,利用青岛地区历史探空数据仿真的大气辐射亮温对神经网络进行了训练,并对利用实测亮温反演的大气剖面与探空实测剖面进行了比较分析,分析结果表明利用微波辐射计反演的大气剖面与实测剖面间有很好的一致性,且较非线性回归算法有更好的反演精度。同时利用神经网络算法进行了12通道微波辐射计反演性能的仿真研究,并与双通道微波辐射计反演结果进行了对比分析,分析结果表明利用多通道微波辐射计可进一步提高大气剖面的反演精度。(3)分析了利用双通道微波辐射计反演大气积分水汽和云中液态水含量的反演方法,并利用青岛历史探空数据得到了青岛地区积分水汽和云液态水反演的线性回归模型,实验表明反演结果与气象探空结果有很好的一致性。(4)利用青岛站和北京站的历史探空数据仿真研究了线性回归算法,非线性回归算法和神经网络算法在不同月份的反演性能,研究表明,不同算法的反演精度随着气象条件的变化具有明显差异,针对不同季节,选用适当的反演算法,可有效提高微波辐射计的反演精度。本文的研究结果可用于微波辐射计的大气环境遥感,对提高微波辐射计的大气环境遥感精度具有重要的参考价值。更多还原

【Abstract】 Atmospheric environment, as the information transmitting path for radio system, has great effects on the performance of information systems, therefore remote sensing profiles of atmospheric parameters is very important for evaluating the performance of radio system. Moreover, atmospheric profiles can be applied to weather forecast, and improve its accuracy. The basic principle and retrieving techniques of remote sensing atmospheric profiles by ground-based microwave radiometer are reviewed in this thesis. Research focuses on the retrieving techniques and algorithms of dual-channel radiometer retrieving profiles of atmospheric parameters, integrated precipitable water vapor and integrated liquid water. The main works are:(1) A non-linear regression algorithm based on conventional linear regression algorithm is presented, which improves the accuracy of retrieving profiles by adding non-linear terms. The non-linear regression models are developed based upon the radiosonde data of Qingdao and Beijing. The results of the experiment in Qingdao show that retrieving profiles are identical with radiosonde profiles, and non-linear regression algorithm has better accuracy in retrieving water vapor profile.(2) Artificial neural network algorithm can easily adapt to non-linear problems, therefore a BP network used in radiometer to retrieve atmospheric profiles is introduced, and trained with the historical radiosonde data and its simulating brightness temperatures of Qingdao. The results show that retrieving profiles gained by BP network algorithm have good agreement with radiosonde data, which are also more accuracy than that gained by non-linear algorithm. Meanwhile, 12-channel radiometer retrieval simulation is made with neural network, and the comparison with the retrieving ones by dual-channel radiometer demonstrates that using multi-channel radiometer can improve the accuracy of profiling.(3) Methods to retrieve the integrated precipitable water vapor(V) and integrated liquid water(L) based on dual-channel radiometer are analyzed, and the linear regression models are developed with the simulation results by historical radiosonde data of Qingdao. Experiment demonstrates that these models can be used with good accuracies.(4) The inversion performances of linear regression algorithm, non-linear regression algorithm and neural network algorithm in different months in an average annual year are simulated by utilizing the historical radiosonde data of Qingdao and Beijing, and the simulation results show that the performances of different inversion methods present discrepancies apparently with the variety of weather condition. Therefore the proper selection of the inversion methods in different seasons may improve the retrieval accuracy.The models proposed in this thesis can be applied to remote sensing atmospheric environment by radiometer, which have important reference value for improving the accuracy of remote sensing atmospheric by radiometer.更多还原