Due to their complex structure and heterogeneous properties, volcanic rocks can be challenging to characterize in seismic data. A multidisciplinary approach is necessary to investigate volcanic rocks in the subsurface. In this study, I combined seismic interpretation techniques, machine learning, seismic inversion, rock physics, and geostatistics to develop a 3D model of the buried Miocene Kora Stratovolcano, Taranaki Basin, New Zealand, that includes seismic facies, petrophysical properties and structural features. The main goal of this work is to deliver a workflow for volcanic seismic characterization in sedimentary basins. The results contribute to an improved understanding of volcanism in sedimentary basins and the petroleum reservoir potential of volcanic facies.

To assist seismic recognition of volcanic facies, I propose two machine learning approaches based on seismic attributes and well-log data. The first is an unsupervised neural network that uses seismic attributes to classify volcanic facies in seismic reflection data based on their similar seismic characteristics (e.g., amplitude content and seismic reflector geometry). The second approach utilizes well information (e.g., well log data) as training data and seismic attributes (e.g., instantaneous phase, RMS amplitude, relative acoustic impedance and amplitude contrast) as input to deliver a final facies model of the Kora Stratovolcano. The two approaches are combined to support a final seismic interpretation of Kora, contributing to the investigation of volcanism in the Taranaki Basin and its petroleum system.

To estimate petrophysical properties of Kora and assess its reservoir characteristics, I developed a workflow based on hydraulic flow unit evaluation, rock physics, seismic inversion and geostatistical techniques. The hydraulic units are classified from core samples using porosity and permeability values extracted from the work of Bischoff et al. (2021) and well reports from Arco Petroleum (1988a, b and c). The hydraulic flow units are upscaled to wells using well logs and machine learning approaches. A probabilistic relationship was developed between 3D seismic attributes and hydraulic units determined at well locations. Based on the 3D hydraulic units probabilistic cubes generated, I used a Gaussian stochastic simulation to build geostatistical realizations of porosity and permeability for Kora Stratovolcano. The final porosity and permeability model of Kora can help to understand petrophysical characteristics of volcanic facies and the heterogeneity of Kora. Results suggests satisfactory values of porosity and permeability at the south-western area of Kora. In addition, the final 3D model can be used in fluid flow simulation to investigate the potential of volcanic facies as reservoirs and fluid flow paths in the study area.

Since the reservoir characteristic and flow behaviour of volcanic rocks are strongly influenced by the presence of fractures, I carried out a workflow to characterize the fractures system and faults of Kora Stratovolcano. Seismic attributes are used in an unsupervised neural network approach to construct a deterministic fracture network of Kora. Conventional well-log data are used to estimate fracture porosity and permeability at well locations. To extrapolate fracture properties within Kora, I used a neural network approach and geostatistical modelling conditioned by seismic attributes. The fracture system in Kora is assessed within the structural evolution of Taranaki Basin. High values of fracture permeability are present at the south-eastern area of Kora and around Kora-1 well. This indicates that fractures play an important role on the reservoir quality of volcanic facies in Kora Stratovolcano. The final geological model assists structural investigation of Kora and its fluid flow behaviours.

The main principles of these workflows are combined to provide a robust final geological model of Kora Stratovolcano that can be used to understand volcanism in Taranaki Basin and reservoir quality for different Kora facies. By using the methods employed in this thesis, my proposed study can be used to de-risk exploratory prospects in volcanic sedimentary basins and assist in fluid flow simulations to evaluate energy exploration worldwide.