对流层大气是现代复杂电磁环境中的重要传输媒质，研究对流层人工变态对电磁波传输的影响对于夺取制信息权至关重要。本文以增强对流层散射、干扰敌方链路及反隐身等应用背景为导向，围绕对流层环境的重要电磁参数预报、相干声波扰动对流层对电磁波传播链路的影响以及高速飞行器扰动下对流层大气的电磁散射特征三个方面展开。论文的主要工作及研究成果如下：

根据大气湍流理论，结合大气折射率结构常数廓线的测量数据，验证了大气折射率结构常数的变化具有高度-时间相关性。基于大气折射率结构常数的高度-时间相关性，利用具有自适应特征提取和非线性表征能力的深度卷积神经网络和滑动窗口数据处理技术，提出了一种高精度的大气折射率结构常数廓线预报方法。利用TWP3风廓线雷达测得的四川省绵竹市的大气折射率结构常数数据对所提出的大气折射率结构常数廓线预报模型进行训练和测试，验证了所提大气折射率结构常数廓线预报方法的精度。此外，还设计了相关试验对所提深度卷积神经网络模型的单步预报精度和多步预报精度进行了对比，并分析了该预报方法对不同高度上大气折射率结构常数的预报精度。

基于声波改变晴空大气压强、温度和水汽压的物理机理，分别针对无线电波段和光波波段，建立了相干声波引起大气折射率起伏的理论模型。在无线电波段，可听声波长和电波波长易满足Bragg相干条件，因此根据稳定层相干散射理论，计算得到了声波作用前后对流层大气的散射传输损耗，计算结果与已有的实测结果规律相符，验证了满足相干条件的无线电波的散射效应会明显增强，为声波扰动大气实现对流层散射增强提供了理论支撑。在光波波段，主要研究了声波扰动大气对光波传播链路相位的影响，并建立了附加相位的理论计算模型。通过声波干扰迈克尔逊干涉仪和激光外差系统，验证了声波扰动大气对激光传播的影响。此外，还研究了不同声阵列的指向性因数，对于设计和优化声阵列达到不同的现实目的具有指导意义。

以探测飞行器对大气的扰动来实现间接探测隐身飞行器为出发点，将高速飞行器在飞行过程中碰撞、压缩和摩擦周围大气形成的大气介电常数起伏变化的区域视为不均匀的“弱散射体”，并采用计算电磁学中的数值算法对其电磁散射特征进行研究。利用Ansys Fluent求解得到飞行器以不同速度飞行时引起的空间大气压强的剧烈起伏变化，然后根据大气压强与大气介电常数之间的关系得到介电常数起伏的空间分布。根据大气介电常数空间分布的点云数据提取外轮廓并建立相应的非均匀“弱散射体”模型。采用体积分方程矩量法计算高速飞行器所致“弱散射体”的电磁散射特性，分析了不同入射角和频率下“弱散射体”的雷达散射截面（RCS）随散射角的变化规律。将该“弱散射体”的RCS和大气湍流的RCS进行了对比，表明扰动气流形成的“弱散射体”具备雷达可探测性，对于后续反隐身方法的研究具有一定的参考价值。

提出了一种新的用于外推飞行器所致“弱散射体”频域RCS特性的机器学习方法。首先将目标RCS随频率变化的曲线分解为全局趋势和局部波动两个部分，然后分别采用基于对数线性函数的最小二乘方法拟合全局变化趋势，基于谱混合协方差函数的高斯过程回归对局部波动模式进行自动发掘与外推。最后在导体目标和介质目标上对所提方法的外推性能进行了试验验证，试验结果表明该方法不仅能在导体目标的仿真数据和实测数据上进行准确地外推，还能以比较高的精度外推高速飞行器所致“弱散射体”的频域RCS特性。因此，该外推方法不仅可以在一定程度上缓解使用数值仿真方法进行目标频域RCS特性分析时面临的时间成本和内存消耗问题，并且在解决一些实际问题方面具有很大的潜力，例如在有限的测量条件下获得目标在更宽频带内的RCS响应，以及从非合作目标的有限观测RCS数据中外推出其他频率下的RCS。

The tropospheric atmosphere is an important transmission medium in modern complex electromagnetic environments, and studying the impact of the disturbed tropospheric atmosphere on electromagnetic wave transmission has important strategic significance for seizing information superiority. In this dissertation, the prediction of important tropospheric electromagnetic parameters, the influence of coherent acoustic waves on electromagnetic wave propagation, and the electromagnetic scattering characteristics of the tropospheric atmosphere disturbed by a high-speed aircraft have been systematically researched under the practical application background of enhancing tropospheric scattering, jamming enemy links, anti-stealth, and so on. Main works and achievements are as follows:

According to the theory of atmospheric turbulence and the measurement data of the atmospheric refractive index structure constant profile, the height-time correlation of atmospheric refractive index structure constant has been verified. Then, a high-precision forecast model for atmospheric refractive index structure constant profile is proposed using sliding window technology and deep convolutional neural network, which have excellent capabilities of adaptive feature extraction and nonlinear representation. The proposed forecast model for the atmospheric refractive index structure constant profile has been trained and tested using the measured data, which were measured by the TWP3 wind profile radar in Mianzhu City, Sichuan Province, to verify the accuracy. In addition, experiments have been designed to explore the accuracy differences between one-step-ahead and multi-step-ahead forecasts, as well as the forecast accuracy with height.

Based on the physical mechanism of acoustic waves changing atmospheric pressure, temperature, and water vapor pressure, theoretical models of coherent acoustic wave-induced atmospheric refractive index fluctuations have been established in the radio and optical waveband, respectively. In the radio band, the wavelengths of electromagnetic waves and audible sound wave can easily meet the Bragg condition. According to the coherent scattering theory of a stably layered structure, the scattering and transmission loss of the electromagnetic wave in the tropospheric atmosphere before and after the disturbance of coherent acoustic waves are calculated, and show the same patterns as the existing measured results. This fully demonstrate the fact that the scattered radio waves will be significantly enhanced when the coherent condition is met, which provides a theoretical foundation for using acoustic waves to disturb the tropospheric atmosphere to enhance troposcatter. In the optical waveband, the impact of the acoustic wave-induced atmospheric disturbance on the phase of optical waves has been studied, and the theoretical model for the additional phase has been established. The influence of the acoustic-induced disturbance on laser propagation has been verified by jamming the Michelson interferometer and the laser heterodyne system. In addition, the directivity factors of different acoustic arrays have been studied, which may guide the design and optimization of acoustic arrays for different practical purposes.

Aiming at realizing the indirect detection of stealth aircraft by detecting the aircraft-induced atmospheric disturbances, the region with inhomogeneous atmospheres, which is formed by the collision, compression, and friction effects from the high-speed aircraft, is considered as a "weak scatterer", of which the electromagnetic scattering characteristics have been studied by numerical methods in computational electromagnetics. Firstly, the intense fluctuations of the atmospheric pressure caused by the aircraft at different speeds in the troposphere are obtained by using Ansys Fluent, and the spatial distribution of the dielectric constant fluctuations is based on the relationship between atmospheric pressure and atmospheric dielectric constant. Then, the geometric shape of the region with inhomogeneous atmospheres is extracted based on the previously-obtained point cloud data and a corresponding inhomogeneous "weak scatterer" model is established. The method of moments for solving the volume integral equation is adopted to calculate the electromagnetic scattering characteristics of the "weak scatterer" induced by a high-speed aircraft, thus analyzing its radar cross section (RCS) varying with the scattering angle under different incident angles and frequencies. Finally, the "weak scatterer" formed by the aircraft-disturbed atmosphere demonstrates the radar detectability by the comparison between the RCS of the "weak scatterer" and that of the calm atmosphere, which may provide some certain reference for subsequent research on anti-stealth methods.

A novel machine learning method for extrapolating the target’s radar cross section (RCS) versus frequency is presented. The target’s frequency-RCS curve is decomposed into two parts: the global trend and local fluctuations, the non-linear least squares method based on a log-linear function and the Gaussian process regression employing the spectral mixture (SM) covariance function are adopted to capture them, respectively. Experiments based on conducting and dielectric targets are carried out, and the results show that this method can not only accurately extrapolate the conducting target's frequency-domain RCS responses with simulated and measured data, but also those of the "weak scatterer" caused by a high-speed aircraft. Therefore, the proposed method can be used to alleviate the time- and memory-intensive issues of using numerical methods to explore the target’s RCS characteristics, and has great potential for solving some practical problems, such as obtaining the target’s RCS responses in a wider frequency band under the limited measurement conditions, and extrapolating the RCS responses at other frequencies from the limited observation data of non-cooperative targets.