Seismic wave equation modeling is an important task in seismic exploration, especially for those methods based on analysis of the full wavefield. However, the computational time and memory costs are prohibitive for large models and models that include very fine geological features. In this abstract, we propose a multiscale method to solve the elastic wave equation based on multiscale basis functions determined from appropriate local problems with boundary conditions that can effectively represent the influence of complex subgrid variations in material properties. Our method then solves the elastic wave equation on a coarse grid, and the computation of the multiscale basis function is a one-time cost, resulting in a great decrease in the computational cost when simulations are repeated for different source positions, for example. Two numerical examples show that our method can accurately approximate the fine grid solutions.