A statistical study, including discriminant analysis, was carried out on a suite of 438 formation waters from Alberta, Canada, which had been analyzed for Cl, Br, I, HCO₃, SO₄, Ca, Mg, and Na. The analyses were divided into 2 populations, depending on whether the initial status of the well was producing oil and/or gas (322 samples) or nonproducing, i.e., abandoned (116 samples). The populations were further subdivided into Paleozoic and Mesozoic groups, and these 2 groups, together with the entire suite of analyses, were subjected to statistical study. With 95% assurance, the Alberta formation waters associated with producible hydrocarbons are chemically different, in a multivariate sense, from the formation waters from abandoned wells. Further, iodi e and magnesium are the most important discriminators in the Paleozoic group, whereas sodium and chlorine are the most important in the Mesozoic group. These discriminators reflect the different organic geochemical, geologic, and hydrodynamic history of the Paleozoic and Mesozoic strata. Reclassification of the analyses using the discriminant functions resulted in only a 0.65 success ratio, thereby indicating that the chemical analysis of formation water is not a completely reliable predictor of the occurrence of hydrocarbons, at least at the Paleozoic-Mesozoic level. The 0.65 probability of successful classification may be increased with the acquisition of a larger data base and consequent statistical analysis at the formation level.

End_of_Article - Last_Page 784------------