Estimation of Impedance Using Seismic Reflection Data Based on Transdimensional Inversion

Deterministic seismic inversion can obtain optimal results when there is a linear relationship between data and model parameters during local optimization (single optimum solution). For nonlinear geophysical problems and in the presence of multiple local minima for a cost function, global optimization techniques are necessary to characterize the global minimum solution. Stochastic, model-based seismic inversion is a widely used global optimization technique and Markov Chain Monte Carlo (MCMC) method is a natural choice to sample model parameters during the random walk. In this dissertation, I apply a sampling technique called reversible jump Markov Chain Monte Carlo (rjMCMC) to traverse the model space. A key property of this approach is that it automatically changes the layer thicknesses and number of layers, thereby predicting the optimum number of model parameters during inversion. The method applies Bayesian inversion, with rjMCMC sampling, so that it also quantifies the uncertainty in model parameters based on an ensemble of models. I apply Bayesian inversion with rjMCMC sampling for two applications. In the first application, I define upscaling velocity logs as an inversion problem to obtain optimal models and quantify uncertainty of upscaled models at the well location. The upscaled velocity at the well locations can be subsequently used to stabilize velocity inversion during Full waveform Inversion (FWI) for seismic imaging purposes. In the second application, I perform post-stack seismic inversion to obtain shallow impedance structure of the TAMU and Ori volcanoes at the Shatsky Rise oceanic plateau. Since impedance is a rock property, it is used to discriminate basalt rock types, which gives insight into the late-stage evolution of both the volcanoes.