Abstract: Significance: Given the limited research on the statistical properties of marine gravity fields, further investigation is crucial. Therefore, conducting covariance analysis of marine gravity fields utilizing shipborne gravity anomaly data is of significant importance. Objectives: The least squares collocation (LSC) method plays a pivotal role in refining local gravity field models, with the development of covariance models for gravity field quantities serving as the cornerstone. Despite extensive research on covariance models for land gravity fields, studies on marine gravity fields remain relatively scarce. Methods: Initially, a method for computing the covariance function along the survey line direction was introduced, leveraging the spatial distribution characteristics of shipborne gravity measurement data. Subsequently, a comparison of covariance function parameters was conducted under conditions where the variance values of both ground and shipborne gravity anomalies were approximately equal, considering specific topographical and geomorphological factors. This comparison revealed directional differences in marine gravity anomalies. Furthermore, using shipborne gravity measurement data simulated by the global gravity field model, a comparison was made between survey line and grid covariance function parameters, leading to the proposal of a covariance modeling approach based on shipborne gravity anomalies. Ultimately, the covariance function model for the experimental area was constructed using shipborne gravity anomaly data. Conclusions: Directly modeling the covariance function along survey lines for shipborne gravity data is feasible, leading to a regional model. Griding the data before modeling is not recommended. At equal spatial distances, gridded marine gravity anomalies have a larger correlation distance than survey line anomalies, and sea surface anomalies have a larger correlation distance than land anomalies. The marine gridded gravity anomaly covariance has a correlation distance of 40.181 km, while the survey line anomaly covariance has a correlation distance of 19.893 km, providing a reference for gridding marine discrete gravity data. The experimental area's shipborne gravity survey line data covariance function model has a correlation distance of 18.878 km.Application Scenarios: The methods proposed in this study can provide valuable insights and serve as a reference for the precise modeling of marine gravity fields using the least squares collocation method, based on covariance functions.