Data–knowledge integrated stochastic modeling of complex Cu-Ni sulfide orebodies: A case study from the Eastern Tianshan Cu-Ni Deposit, Xinjiang, China

• Develops a level set–based implicit method for complex Cu–Ni sulfide ore modeling. • Uses geometric distance and geochemical data to constrain orebody boundaries precisely. • Enables multi-source geological data integration for complex orebody modeling. Constructing complex geological models is one of the core tasks in geosciences. This study proposes an implicit modeling approach that integrates data-driven techniques with prior geological knowledge. The approach is designed to address challenges such as data sparsity, morphological complexity, and inadequate uncertainty representation in the three-dimensional modeling of complex Cu-Ni sulfide orebodies. The method extracts point sets from multi-directional two-dimensional sectional contours and utilizes Chamfer distance-based hard constraints to characterize the geometric features of orebody boundaries. A geochemical grade field is constructed using Kriging interpolation, while a mineralization probability field is generated through a logistic regression model to provide soft constraints that guide boundary evolution. The orebody interfaces are dynamically evolved using a level set function, and multiple realizations are generated through Markov Chain Monte Carlo sampling, effectively capturing the complexity and spatial variability inherent in the mineralization process. A multi-source loss function is developed to simultaneously enhance geometric accuracy and incorporate mineralization trend information, thereby improving the model’s structural validity and spatial continuity. Applying the proposed method to the Tulaergen and Huangshanxi Cu-Ni sulfide deposits in the Eastern Tianshan region of Xinjiang demonstrates its applicability and robustness in modeling complex orebodies and representing multiple solution scenarios. This provides robust technical support for deep resource evaluation, uncertainty quantification, and exploration decision-making.