In this paper, we address the problem of two dimensional (2-D) direction-of-arrival (DOA) estimation for parallel co-prime arrays. Traditional 2-D DOA estimation methods usually suffer from the tremendous computation burden caused by spectral search and angle pairing. To this end, in this paper we propose an efficient reduced dimensional least spectral search based estimation method with automatic pairing. Specifically, we first utilize the cross-covariance matrix to decouple the 2-D DOA estimation problem into a one-dimensional (1-D) one, and then design a least spectral search based 1-D DOA estimation method according to the relations between true and ambiguous angles. Finally, we estimate the remaining 1-D DOAs via least square criterion with automatic pairing. We evaluate the complexity and present the simulation results to show the effectiveness of the proposed method.