Traveltime inversion for the geometry of aquifer lithologies

Crosswell traveltime tomography can provide detailed descriptions of the geometry and seismic slowness of lithologic zones in aquifers and reservoirs. Traditional tomographic inversions that estimate a smooth slowness field to match traveltime data, provide limited information about the dominant scale of subsurface heterogeneity. We demonstrate an alternative method, called the multiple population inversion (MPI), that co-inverts traveltimes between multiple well pairs to identify the spatial distribution of a small number of slowness populations. We also compare the MPI with the split inversion method (SIM) that was recently introduced to address the same problem. The lithologies and hydraulic parameters for these populations can then be determined from core data and hydraulic testing. The MPI iteratively assigns pixels to a small number of slowness populations based on the histogram of slowness residuals. By constraining the number of slowness values, this method is less susceptible to inversion artifacts, such as those related to slight variations in ray coverage, and can resolve finer scale sedimentary structures better than methods that smooth the slowness field. We demonstrate the MPI in two dimensions with a synthetic aquifer and in three dimensions with the Kesterson aquifer in the central valley of California. In both cases, the constrained inversion algorithm converges to an equal or smaller average traveltime residual than obtained with unconstrained-value tomography. The MPI accurately images the dominant lithologies of the synthetic aquifer and provides a geologically reasonable image of the Kesterson aquifer.