Search

To accurately reconstruct palaeoenvironmental change through time it is important to determine which rock samples were deposited contemporaneously at different sites or transects, as erroneous correlation may lead to incorrectly inferred processes and rates. To correlate samples, current practice interpolates geological age between datable units along each transect, then temporal signatures observed in geochemical logs are matched between transects. Unfortunately spatiotemporally variable and unknown rates of sedimentary deposition create highly nonlinear space-time transforms, significantly altering apparent geochemical signatures. The resulting correlational hypotheses are also untestable against independent transects, because correlations have no spatially-predictive power. Here we use geological process information stored within neural networks to correlate spatially offset logs nonlinearly and geologically. The same method creates tomographic images of geological age and geochemical signature across intervening rock volumes. Posterior tomographic images closely resemble the true depositional age throughout the inter-transect volume, even for scenarios with long hiatuses in preserved geochemical signals. Bayesian probability distributions describe data-consistent variations in the results, showing that centred summary statistics such as mean and variance do not adequately describe correlational uncertainties. Tomographic images demonstrate spatially predictive power away from geochemical transects, creating novel hypotheses attributable to each geochemical correlation which are testable against independent data. [ABSTRACT FROM AUTHOR]

Geological process models typically simulate a range of dynamic processes to evolve a base topography into a final two-dimensional cross section or three-dimensional geological scenario. In principle, process parameters may be updated to better align with observed geophysical or geological data. However, it is hard to find any process model realisations that fit all observations if data sets are complex and sparse in space or time because the simulations typically depend highly non-linearly on base topography and dynamic parameters. As an alternative, geophysical probabilistic tomographic methods may be used to estimate the family of models of a target subsurface structure that are consistent both with information obtained from previous experiments and with new data (the Bayesian posterior probability distribution). However, this family seldom embodies geologically reasonable images. Here we show that the posterior distribution of tomographic images obtained from travel time data can be fully geological by injecting geological prior information into Bayesian inference and that we can do this near-instantaneously by using trained mixture density networks (MDNs). We invoke two geological concepts as prior information about the possible depositional environment of an imaged target structure: a braided river system and a set of marine parasequences. Each concept is parameterised by the latent parameters of a generative adversarial network. Data from a target structure can then be used to infer the family of compatible latent parameter values using either geological concept using MDNs. Our near-instantaneous MDN solutions closely resemble those found using relatively expensive Monte Carlo methods. We show that while the use of incorrect geological conceptual models provides significantly less accurate results, a classifier neural network can infer which geological conceptual model is most consistent with the data. It is thus demonstrated that even apparently barely related geophysical data may contain information about abstract geological concepts, and that geological conceptual models are key to creating reasonable images from geophysical data. [ABSTRACT FROM AUTHOR]