In airborne nonside looking MIMO radar system, conventional statisticalbased spacetime adaptive processing (STAP) algorithms always cannot obtain enough independent identically distributed training data, and fail to estimate the clutter covariance accurately, thus degrading the performance of clutter suppression. In order to reduce the requirement of training data, the concept of clutter range azimuth spectrum (RA) is first presented. Then, the clutter RA is obtained by utilizing a sparse recovery algorithm and a preconstructed range filter. At last, the covariance matrix of clutter is calculated by the estimated clutter RA. Due to the application of the priori information and the relation between the spatial frequency and temporal frequency of clutter, the proposed STAP method based on RA, i.e. RASTAP, can estimate the covariance matrix of clutter with a few training cells accurately. Therefore, the method can suppress the clutter and detect lowmoving target effectively. The theoretical analysis and the experimental simulations demonstrate that the method is effective.