Joint inversion of multiple geophysical data using guided fuzzy c -means clustering

ABSTRACT Joint inversion of multiple geophysical data has become an active area of research due to its potential to greatly enhance the fidelity of inverted models. Many open questions and challenges still remain. One of them is how to effectively incorporate into joint inversion multimodal petrophysical information that describes the statistical behavior of physical property values in the parameter domain (i.e., in a crossplot). We have regarded the multimodal petrophysical data as different clusters in the parameter domain and developed an approach that handles multimodal petrophysical information through guided fuzzy c-means (FCM) clustering in the parameter domain. We inverted the petrophysical data in the parameter domain in a similar manner to and simultaneously with the geophysical data in the spatial domain through minimizing one common objective function. Numerical examples have determined that the resulting models from this multidomain joint-inversion strategy are able to reproduce both the geophysical and the petrophysical data. In addition to incorporating a priori multimodal petrophysical information into inversion, guided FCM clustering also allows us to integrate geology differentiation and geophysical inversion into one unified framework that makes these two components positively affect each other. Geology differentiation results were obtained as a direct output from joint inversion. We have also developed a strategy for imposing different clustering constraints in different model regions, allowing region-specific a priori petrophysical information to be incorporated into inversion. We have applied our joint-inversion algorithm to the SEG/EAGE salt model in four different scenarios, and we found that the proposed algorithm produced much better geophysical models and geology differentiation results than separate inversions.