Abstract : In this work the Gaussian Beam Shooting (GBS) algorithm is complemented with new original formulations, and the ability of this "augmented" GBS algorithm to address specific problems encountered in electromagnetic field computations for ground-based Radar applications is tested. GBS is considered as an alternative to methods (Parabolic Equation, ray based methods) currently used for such computations in complex urban environments, especially when lateral obstacles and Non-Line-Of-Sight (NLOS) targets are involved. The "basic" GBS algorithm makes use of analytical expressions obtained through paraxial approximations. It allows to perform fast computations in complex environments, without suffering from any caustics problems. Reasonably accurate results have been obtained with this method in the millimetric range, e.g. for indoor field calculations. At lower frequencies, such as used in ground Radar systems, "basic" GBS cannot model diffraction effects accurately enough, and Gaussian beam width with respect to obstacle dimensions becomes a problem after some propagation distance. Frame theory is used in this PhD to overcome these limitations. Frame theory provides a rigorous framework for the initial decomposition of radiated fields into a set of Gaussian beams, providing flexible rules to adjust the number and directions of the launched beams. In this thesis, frame theory is used to discretize not only the source field distribution but also incident field distributions over planes or parts of planes of interest, according to encountered obstacles and propagation distances. The radiated fields are then obtained by summation of Gaussian beams launched from these frames called "reexpansion frames". Gaussian beam transformations by finite sized obstacles are addressed by this re-expansion scheme : the incident beams partially impinging on limited areas are successively "re-expanded" on two re-expansion frames, the first one composed of "narrow" windows and the second one of "wide" windows, both defined in the plane containing the limited area. Spatially narrow window frames allow to take into account abrupt transitions in space, and spatially wide window frames radiate in the form of collimated Gaussian beams. The re-expansion formulation proposed in this work is designed for efficient numerical implementation. Approximate analytical expressions are established for expansion coefficients on narrow window frames, and for frame change matrix elements. This formulation has been implemented, and the influence of frame parameters on re-expansion accuracy is analyzed. Finally, the GBS algorithm augmented with successive re-expansions is used to compute fields in simplified scenarios similar to situations encountered in ground-based Radar propagation problems.