Geophysical models integration using principal component analysis: application to unconventional geothermal exploration

Geophysical data integration covers a wide range of approaches, from visual interpretation of model presented side by side to sophistical statistical analyses such as automatic clustering. We present here a geophysical model integration based on principal component analysis (PCA), which allows to gain insight in a multivariable system. PCA for geophysical models integration define a new set of principal component (PC) models, distributed along new orthogonal axes by solving a eigenvalue problem. We show that PC component models patterns reflect different processes sensed by the geophysical observations. We applied this integration method to models obtained from constrained and joint inversion of gravity, ambient noise and MT data in the framework of unconventional geothermal exploration in Massif Central, France. PCA of the log-resistivity, the density contrast and the Vs velocity model has three independent components. The first one (PC1) representing 69 per cent of the total variance of the system is highly influenced by the parameter coupling enforced in the joint inversion process. PC1 allows to point to geophysical structures that may be related to the geothermal anomaly. The second component (PC2) represents 22 per cent of the total variance and is strongly correlated to the resistivity distribution. PC2 correlation to shallow fault structures suggests that it may be a marker of fracturing. The third component, PC3, accounts for 9 per cent of the total variance and is correlated with velocity structures and anticorrelated with density structures, respectively. The contribution of geophysical properties, mainly sensitive to the elastic properties of rock units, and the good agreement with shallow geological features make PC3 useful for the 3-D description of geological units. This statistical approach helps the interpretation of geophysical models into a limited number of geological processes, one possibly geothermal. Lithology may be derived from PC3, fracturing from PC2 and hydrothermal anomaly from PC1. Nevertheless PCA as a geophysical integration methodology is site-dependent and interpretation relies on a priori knowledge of the local geological mechanisms.