Probabilistic Approach for 3D Conceptual Modelling Using Multi Geophysical Data

Abstract

Production sustainability of geothermal resources requires accurate and high definition conceptual model focusing on temperature and permeability. Probabilistic approach facilitates investigation and analysis of complex subsurface structures regarding geothermal reservoirs, cap rocks, recharge areas and faults associated with hydrothermal system. In this paper, we present a case study of 3D conceptual modelling at the Onikobe geothermal field in northeast Japan. There has been multimodal dataset consisted of surface seismic, magnetotelluric (MT), airborne gravity gradiometry (AGG) and gravity surveys in addition to borehole data. These geophysical data provide subsurface structures regarding rock properties to interpret hydrothermal system. Especially, surface seismic data have valuable information on not only velocity structure, but also fault and fracture images represented by geometric attributes that are automatically extracted such as Thinned Fault Likelihood (Hale, 2013), dip azimuth, and curvature attributes (Roberts, 2001). Thus, the geometric attributes are significant to understand spatial distributions of fault-controlled geothermal reservoirs. These seismic attributes and rock properties can be quantitatively integrated by means of multi-attribute analysis (MAA). We have implemented a probabilistic MAA that with the artificial neural network (ANN) algorithm developed by Hampson et al. (2001) for a cost-effective solution of the non-linear problem in geothermal modelling. In this workflow, the probabilistic approach makes it possible to provide a detailed geothermal model while taking advantage of higher spatial resolution of 3D surface seismic data. Another probabilistic approach, Bayesian classification, is a robust tool for permeability assessment in the conceptual modelling. We performed Bayesian classification based on non-parametric probability density functions (PDFs) using resistivity, P-wave impedance and Thinned Fault Likelihood attribute. The comprehensive interpretation from the detailed litho-facies model is consistent with the observed distributions of temperature, permeability and hydrothermal fluids beneath the Onikobe caldera. It reveals that the probabilistic conceptual modelling yields significant benefits for geothermal exploration and exploitation.