Prahastudhi, S, William, R, Bailey, E & de Beer, W 2026, 'A study of seismic tomography for cave shape model estimation', in A van As, D Cumming-Potvin & J Wesseloo (eds), Caving 2026: Proceedings of the Sixth International  Conference on Block and Sublevel Caving, Australian Centre for Geomechanics, Perth, pp. 1-11, https://doi.org/10.36487/ACG_repo/2645_61
(https://papers.acg.uwa.edu.au/p/2645_61_Prahastudhi/)
Abstract: In block caving operations, knowing the shape and the evolution of the cave is critical since these factors influence local and regional ground stability, and impact operational strategy. While open hole logging using a borehole camera is one of the most accurate methods for delineating the cave shape, drilling near the cave is challenging and largely impractical due to operational, economic and physical constraints. Operators therefore need indirect methods such as subsurface imaging. In seismology, passive tomography has several applications: it is used in studying the Earth’s structure, in mining, in the geothermal energy industry and in hydrocarbons extraction. Passive tomography uses seismic waves recorded by seismic arrays. A key advantage is that tomography is not dependent on the number of events, but instead relies on the relative distribution of sensors and events and on the density of ray paths. Tomographic inversion requires an initial velocity model. In applications which aim to track stress evolution around large voids, this seed model is usually a 3D (heterogeneous) velocity model derived from the estimated cave shape. This cave shape is informed by limited open hole measurements and/or height of draw (HOD) estimates that are subject to biases introduced by, for example, the flow model assumed. The use of such a model can lead to the inversion result being driven excessively by the 3D velocity model rather than by actual cave growth. In this study, we investigate time lapse passive tomography imaging in two different producing block caving mines and compare the results obtained with and without the seed 3D velocity model. This allows an assessment of the dependence of results on a priori assumptions and allows for comparison with open hole and HOD estimates.