Abstract: Traditional gridding methods struggle to balance computational accuracy and efficiency when processing irregularly distributed magnetic anomaly data. To address this issue, this study applied the classic least-squares collocation method from geodesy to the gridding of ground-based magnetic anomaly data. This application was verified through the test and analysis of the simulation data and the actual coalfield data. The results indicate that the computational accuracy of gridding based on least-squares collocation is dictated by the error estimation of discrete observational data and the selection and fitting of the covariance function. More accurate error estimation contributes to higher-accuracy interpolation. A polynomial function is a simple and effective empirical covariance function for processing magnetic anomaly data. The least-squares collocation method demonstrates more effective noise suppression compared to the Kriging, minimum curvature, and radial basis function methods. Overall, applying the least-squares collocation to the gridding of magnetic anomaly data can enhance the accuracy and efficiency of data processing.