A two-step approach consisting of meso- and macro-scale simulations is presented to predict the radar signature of a composite aircraft. At the meso-scale, we develop an efficient method to extract the effective anisotropic surface impedances (ASI), which mimic the electromagnetic properties of composite materials for all incident angles and all polarizations of a radar plane wave at the frequency of interest. The approach is based on an Interface-enriched Generalized Finite Element Method (IGFEM), which uses a virtual model composed of finite elements that do not need to conform with material interfaces, yet provides the same level of accuracy as conforming meshes. At the macro-scale, we employ the extracted ASI to construct a well-conditioned surface integral equation (SIE) to predict the radar signature of the whole aircraft. We implement a parallel Multilevel Fast Multipole Algorithm (MLFMA) to reduce the computational complexity and speed up the process of solving the derived SIE.