Open data to visualize subduction zones and improve large scale prospectivity interpretation: A case study examining the Wadati-Benioff zone of Cocos-Caribbean subduction in Central America

Well-distributed public domain data are a valuable resource for examining large-scale prospectivity in under-explored districts. We describe a workflow created to visualize and interpret the subduction geometry of the Cocos plate below the Caribbean plate in the context of magmatic-hydrothermal mineral deposit prospectivity. Inputs for this case study consist entirely of freely available hypocenter, volcano, topographic, and mineral occurrence data. These data were sub-set to the area of study but are available worldwide. The workflow begins with objective, data-driven methods to enhance visualization of subduction geometry, and transitions to subjective interpretation of the meaning of this geometry in the context of prospectivity. This work considers the potential influence of the subduction geometry of the Cocos plate below the Caribbean plate on the reactivation of structures in the Caribbean plate that may provide conduits for the ascent of fertile magmas. Public domain data is used to visualize the geometry of the Wadati-Benioff zone between these two plates, and we discuss the possible implications for past and present mineral deposit placement. We propose that segmentation of the obliquely subducting Cocos plate exerts differential stresses on the Caribbean plate, and aids in creation or reactivation of arc-oblique faults, which provide conduits for ascent of fertile magmas from the base of the continental crust to the upper crust. To arrive at this hypothesis, we developed a novel workflow using scalable computational methods applied to hypocenter data to constrain aspects of the Wadati-Benioff geometry. Manifold dimensionality reduction in the form of pairwise controlled manifold approximation (PaCMAP) and density-based clustering in the form of hierarchical density-based spatial clustering of applications with noise (HDBSCAN) was used as an interpretive aid and denoising step, followed by multiscale continuous wavelet transform tessellation (MCWTT) to identify breaks in hypocenter density along the trend of the Wadati-Benioff zone, indicating the location of inferred slab segmentation. These three methods are robust and computationally efficient, scalable to much larger datasets for application in analogous settings. We then visually (subjectively) compared these results with the geometry of the modern volcanic arc, broad scale topographic fabric of the Caribbean plate, and placement of mineral occurrences of magmatic-hydrothermal affinity. The workflow is relevant to seismically active back-arc districts wherein modern subduction geometry can provide insight into the processes governing the placement of magmatic-hydrothermal deposits.