Tools for Edge Detection of Gravity Data: Comparison and Application to Tectonic Boundary Mapping in the Molucca Sea

Edge detection techniques for potential field data are effective methods for identifying local and regional geological boundaries. Numerous edge detectors (e.g., derivative-, ratio- and statistic-based methods) have been successively proposed and applied to different scenarios. However, these edge detectors show diverse results, which can confuse interpreters in their filter selection and interpretation schemes. To better understand the capabilities of various edge detection methods and avoid over-interpretation of artifacts, it requires a unified evaluation of different edge detectors with the same test models. In this view, we first present a brief review of the previous edge detection methods. Then, using gravity data as an example, we build 2.5D and 3D models to examine the boundary recognition capabilities of 28 edge detectors. Based on the model test results, we classify the existing edge detectors and discuss the similarities and discrepancies of different detectors. These comparisons help us to infer the optimal edge interpretation by integrating multiple results and screening for false appearances. Finally, we apply edge detection techniques to the earthquake-prone Molucca Sea region and present a refined tectonic boundary division, assisted by the focal-mechanism solutions. Besides, we identified four deep boundaries that may be associated with plate subduction. These boundaries correspond well to the source location of earthquakes at different depths; hence, five depth-dependent earthquake zones are partitioned. In addition to subduction, we suggest that the fault system also contributes to the present-day tectonic configuration around the Molucca Sea. The relationship between the earthquake activity near the subduction zones or faults and the boundaries derived from edge detection provides new insights to study multi-plate convergence using multiple observations.