Abstract: Despite its significant advantages over conventional methods in terms of efficiency and cost, the simultaneous-source technique faces the challenge of wavefield aliasing that hinders its application. Effective signals and aliased noises in simultaneous-source data exhibit the differences in coherence on non-common shot gathers. These differences can be characterized by their respective sparsity in the shearlet transform domain. Therefore, this paper proposes an adaptive threshold iteration algorithm based on the shearlet transform for data separation, with the objective of ensuring the quality of subsequent seismic data processing and imaging. The algorithm begins by transforming the pseudo-separated simultaneous-source data on common receiver gathers into the shearlet domain for a sparse representation. Leveraging the distinct distribution characteristics of effective signals and aliased noises in the shearlet domain across different scales, aliased noises are separated by introducing an adaptive Bayesian threshold estimation to shrink the noise-related transformation coefficients. An iterative strategy is incorporated into the algorithm to gradually optimize the separation results. Synthetic and field data tests demonstrate the good performance of the proposed algorithm in simultaneous-source data separation.