The framework of estimating the clutter rank of multi-dimensional sparse array radar is presented in this paper. The whole array is first divided into several sub-arrays according to the Nyquist interval, and then the number of degree of freedom (NDOF) estimate theory is employed to estimate the clutter rank in each sub-array. Finally, the clutter rank of the overall system is obtained by adding them together. In 2D/3D arrays, the clutter correlation becomes 2D/3D band-limited process. The NDOF theory is thus extended to 2D/3D cases. The Nyquist interval in 2D/3D arrays with or without range ambiguity is also discussed. When no range ambiguity exists, the actual Nyquist interval is larger than the nominal one. And under some circumstances, the clutter ranks of the 2D and 3D array are approximately identical.