A very large scatter has been observed for the incipient motion threshold of a gravel bed in lit-erature, which can be explained by a lack of bed surface characterization. In the present work, a laser scan-ning technique is used and criteria to describe the bed surface are proposed in order to compare initial bed ar-rangements of laboratory tests on graved bed incipient motion. Statistical tools, such as probability density function of the bed surface elevations and two-dimensional (2D) structure function, are used to characterize the bed surface. Surveys of relatively well-sorted gravel bed surfaces show that even small bed arrangement changes yield a significant impact on the incipient motion of the gravel particles