Seismic inversion approaches for reservoir characterization: A comprehensive review

Seismic inversion is a pivotal technique in reservoir characterization, enabling the transformation of seismic reflection data into quantitative rock properties to elucidate subsurface characteristics. In this paper, we reviewed various inversion methods such as post-stack seismic inversion approaches (e.g., band-limited, coloured, sparse spike, and model-based inversion) and pre-stack seismic inversion approaches (e.g., amplitude versus offset (AVO), elastic impedance, and simultaneous inversion), as well as full-waveform inversion (FWI) and machine learning-based (e.g., convolutional neural network (CNN)) methods. Furthermore, we reviewed different approaches of inverse rock physics modelling for the purpose of converting layer elastic properties into layer reservoir properties. Our work offers a good opportunity to compare different inversion methods for further application on a given dataset and geology conditions. We observe that despite the current advancements in seismic inversion, still significant challenges remain, including computational demands, integration of multi-source data, and uncertainty quantification. Therefore, we discussed different challenges in more details in addition to a comprehensive review and discussion on the state-of-the-art seismic inversion techniques by emphasizing on their methodologies, advantages and disadvantages. Then, we highlighted the role of uncertainty quantification, with a focus on the Bayesian inversion and the Ensemble Kalman Filter (EnKF) to enhance the reliability and robustness of the seismic inversion results. Furthermore, we explore future directions, particularly the integration of machine learning to improve seismic reservoir characterization.