Inferring a continuous lithology and mineralogy from multivariate statistical analysis of well-logging data: Examples from some sedimentary structures associated with tectonic plates convergence zones (Ocean Drilling Program Leg 134, 156 and 160)

Well-logging techniques provide a preferential way of investigating deep geological formations, not directly observable. They measure on a continuous basis and in situ numerous petrophysical parameters, as well as oxide contents in 7 major (Si, Ca, Fe, Al, S, K, Ti) and 3 (U, Th, Gd) trace chemical elements. A thorough knowledge of the physics underlying tool operation and the calibration steps is needed to estimate the uncertainty on the measurement. This uncertainty is mainly due to tool operating conditions (borehole diameter and irregularities,, operating speed). When good quality data are available, multivariate clustering techniques, such as k-means, are used as a first approach to visualize the variables provided by well-logging tools. One can also choose linear and non linear inversion methods to infer an accurate enough mineralogical log so it may be used in geological and petrophysical modeling. Wireline logging techniques are typically carried out during oceanic and continental scientific drilling programs, but are almost useless in such unstable geological environments as sedimentary structures associated with tectonic plates convergence zones. The improved state-of-the-art method of Logging-While-Drilling gives a best quality measurement before the borehole conditions worsen. Geological, petrophysical and textural interpretation of well-logging data gives valuable information on sedimentation type, as well as hints on possible mineralogical transformations. The in situ criterion pertaining to the well-logging measurement allows fluids dynamics modeling as well as variation and orientation of tectonic stress, often characterizing these geological environments.