Dramatic increases in the speed of desktop computers over the past few years have significantly improved our ability to analyze large data sets and quantify the relative effects of dependent and independent geologic variables using multiple linear regression and other statistical techniques. The result can be highly accurate predictions of porosity and permeability for observations evaluated relative to a calibration data set. Application of these procedures can also provide: (1) independent predictions for porosity and permeability and (2) estimates for the probability of encountering a certain porosity or permeability range in a specific sandstone interval. The predictive capabilities of the empirical approach are illustrated in a recent paper by Bloch and Helmold (1995), using the southern San Joaquin Basin in California as an example.
Until sandstone diagenetic processes are adequately understood, empirical models may have a distinct advantage over process-oriented models in providing reliable reservoir quality predictions. However, despite recent successes, this type of empirical predictive approach has, and will continue to have several limitations. Important diagenetic processes may not be accounted for by parameters that comprise a given calibration data set, resulting in inaccurate predictions based on that data set. Predictions are likely to be basin-specific, and may even be restricted to particular facies or stratigraphic horizons, thus inherently limited in their application. Finally, a large calibration dataset is required, rendering the approach ineffective in sparsely drilled areas. All of these issues point to the need for improved process-oriented models.