IAMG 2001--Cancún Technical Program--Session J

Home Page Technical Program Session J Intro Session J Papers

Compositional Interpretation of Well-logs by J.J. Egozcue*, Universitat Politecnica de Catalunya, Spain, R. Tolosana-Delgado, Universitat Politecnica de Catalunya, Spain, and V. Pawlowsky-Glahn, Universitat de Girona, Spain Wireline logging have become one of the most important data sources used by geologists, because they are relatively cheap to obtain, and allow the implementation of quantitative models. These techniques arose in the oil industry, where they are widely used to infere oil content of porous rock formations: the widely-known Archie model relates logs with rock composition, through a set of functional parameters related to mineralogical composition and geological history of the rock formation. The main point is that, for given functional parameters (e.g. cementation coefficient) well-log data can produce results violating the nature of compositions. This approach avoids this problem by assuming these parameters to be random. Then, a statistical treatment is thought to be necessary because: (1) we can assess estimation error and risk, and (2) we will honor both the nature of compositions (through Aitchison's compositional statistics) and the uncertainty inherent to the parameters values (throug the Bayesian approach). Bayes' Theorem is the tool to treat all this available empyrical information and the wireline data. Thus, a compositional bayesian method (BCM) is built and tested against classical interpretation methods. BCM estimations are better, because they are generally statistically reliable and offer a measure of error estimation, although they are affected by the functional parameters' uncertainty in poor oil rock formations. Paper in PDF format Egozcue, Tolosana-Delgado, and Pawlowsky-Glahn, Acrobat PDF, 1 meg.