To address the low efficiency of conventional methods for parameter inversion of rockfill dams, a rockfill dam parameter inversion method based on a Kriging surrogate model optimization algorithm is proposed. The method first extracts a small number of initial sample points from the parameter space using the Latin hypercube sampling method to establish a relatively coarse Kriging surrogate model. New sample points are then selected according to various infill criteria to update the sample set, thereby obtaining a higher-accuracy Kriging surrogate model to perform optimization until the convergence conditions are met. During the process of gradually adding the number of sampling points, this method can ensure that the newly added sample points fall within the parameter space with the greatest potential for optimal solutions, reducing the randomness of sample point selection of traditional surrogate models and thus improving inversion analysis efficiency. The verification results from an engineering case show that the proposed method can reduce the number of finite element model calculations, improve the sampling efficiency of the surrogate model, shorten the parameter inversion time, and enhance the average calculation accuracy at inversion analysis points.