Abstract: The two-step method to determine relative weight ratio factors in joint inversion is derived in this paper. In the first step, without considering relative weight ratio factors, the joint inversion model of two or more kinds of observation data is established. Through Helmert method of variance components estimation, unit weight variances of observation data are equal (or the ratio of unit weight variances is approximate to 1), and the reasonable and theoretical weight matrix is determined. At the sametime, different kinds of observation data are normalized. In the second step, on the base of reasonable weight matrix, the joint inversion model considering relative weight ratio factors is established. The relative weight ratio factors are determined when objective function is minimum. Through simulated data and several solutions, the effectiveness and feasibility of two-step method are proved.