LINEARIZED INVERSION OF TIME-LAPSE SEISMIC WAVEFORM DATA USING DETERMINISTIC AND BAYESIAN APPROACHES

Autumn 2016

Spring 2018

Time-lapse seismic data is useful for monitoring production-induced changes in a petroleum reservoir. The thesis will present methods for inversion of time-lapse seismic waveform data in the acoustic approximation. The scatter-integral methods of Born approximation and distorted Born approximation will be used for waveform modelling. Inversion techniques will be tested in a synthetic reservoir experiment where time-lapse strategies will be investigated. Both deterministic and Bayesian solutions will be presented. Since the Bayesian solution gives a probability distribution of the model parameters, this approach is more suitable for uncertainty analysis. Thus, the inverted data also has applications within updating reservoir models that are ensemble-based.

Morten Jakobsen