3D probabilistic geology differentiation using mixed Lp norm joint inversion constrained by petrophysical information

Abstract

Geology differentiation aims to spatially identify subsurface geologic units via integrating prior geologic information and multiple physical property models obtained from joint (or, separate) inversions. However, uncertainty analysis of geology differentiation results is still largely under-explored. To fill the gap, we have developed an empirical method of quantifying the uncertainties of differentiated geological units. We performed mixed Lp norm joint inversion in multiple times with different regularization terms controlled by two tuning parameters. We obtained a sequence of jointly recovered models that all fit the observed geophysical data but exhibit diverse ranges of features (e.g., differing degrees of smoothness). Prior information based on rock sample measurements was used to accept those recovered models whose values are within the expected ranges; the rest are rejected. We performed geology differentiations for all the accepted models and obtained multiple 3D quasi-geology models. Simple statistics such as standard deviations were then calculated to obtain 3D probabilistic quasigeology models. We applied our method to field data collected over the Decorah area located in Northeast Iowa. We were able to quantify the uncertainties of each geology unit and construct 3D probabilistic geology differentiation models.