“…Direct matrix solvers for large geophysical 3D forward problems [33,28,30,26,24,14,19,8,35,2,15] can be computationally expensive, both in terms of memory and CPU time. A way to overcome this limitation is by reducing the dimensionality of the problem, namely from 3D modeling to a 2.5D [23], 2D [4] and/or 1D [20] approximations that are only suitable for particular geometries.…”