Under the assumption that the sea bottom is an almost-flat and randomly rough thin layer, a spatial correlation model for bottom reverberation is constructed. At high signal noise ratio, the phase of the spatial correlation function is the product of sound wave number and the vertical vector between two hydrophones. In a nominally horizontal plane, roll and pitch bring on the vertical vector between the hydrophones. Then an equation including roll, pitch and the phase of the spatial correlation function is found. If a parallelogram can be constructed by 3 hydrophones and the rat ios of its acreage to diagonals are not smaller than half of the wavelength, roll and pitch can be obtained analytically or by optimal method. However, the ranges of roll and pitch are restricted because of the phase ambiguity. Using Fisher information matrix, the Cramer-Rao lower bound is obtained. Results from computer simulation and sea test data prove the feasibility of the method. Under the assumption that the sea bottom is an almost-flat and randomly rough thin layer, a spatial correlation model for bottom reverberation is constructed. vertical one between two hydrophones. In a nominally horizontal plane, roll and pitch bring on the vertical vector between the hydrophones. Then an equation including roll, pitch and the phase of the spatial correlation function is found. If a parallelogram can be constructed by 3 hydrophones and the rat ios of its acreage to diagonals are not smaller than half of the wavelength, roll and pitch can be obtained analytically or by optimal method. However, the ranges of roll and pitch are restricted because of the phase ambiguity. matrix, the Cramer-Rao lower bound is obtained. Results from computer simulation and sea test data prove the feasibility of the method.