An Inverse Stratigraphic Simulation Model – Is Stratigraphic Inversion Possible?

Abstract

Listen

Translate article

Inversion is a systematic method of determining values of process parameters of a forward model that allow a match between observed and modeled data. Historically, geologists have considered the stratigraphic record to be nonunique. That is, geologists have assumed that it is impossible to determine values for and separate stratigraphic process variables such as eustasy, tectonics and sediment supply that operated to form the stratigraphic record. If stratigraphic data are nonunique, then inversion of stratigraphic data is impossible. In an influential paper. Burton et al. (1987) argued that inversion of stratigraphic data using a stratigraphic forward model is not possible. The purpose of this study was to determine if inversion of stratigraphic data using a stratigraphic forward model is theoretically possible. In this study, we designed a stratigraphic inverse simulation model using a forward stratigraphic model capable of simulating realistic temporal and spatial distributions of fades tracts and stratigraphic surfaces. For numerical optimization, we used a gradient descent method that minimizes errors in the least squares sense. We tested this inverse model on synthetic stratigraphic data which act as a proxy for real-world stratigraphic data, to test multiple aspects of the inverse model. In these experiments, we inverted synthetic stratigraphic data for eustasy, sediment supply, tectonic subsidence, lithosphere flexural rigidity, and initial basin topography. Results from these inversion experiments establish that inversion of stratigraphic data is theoretically possible. We determined limits of convergence, degrees of parameter separatability, nonuniqueness of data, and types of data necessary for inversion. Results suggest that using distributions of facies tracts and stratigraphic surfaces within a genetic sequence stratigraphic framework is necessary for inversion. Results from inverse model experiments also suggest that nonuniqueness of these data types with respect to stratigraphic process parameters such as eustasy, tectonics, sediment supply and depositional topography is bounded. Moreover, the bounds of nonuniqueness are quite small. The next phase of our research is to first test an inverse algorithm that is more appropriate for stratigraphic inversion, and then to test an inverse stratigraphic model using a real stratigraphic data set.