Larry W. Lake
The exploration and production industry has had a long history of purely statistical models, though this practice has almost entirely given way to the algorithmic models now referred to as numerical simulators. On the other end of the spectrum the ultimate simplification will be a simple model that does empirically quantifies the flow physics in the form of simple to evaluate statistical correlation(s).
We propose a radically different approach utilizing well rate fluctuations to predict interwell connectivity. They expressed the total fluid production at a producer as a linear combination of the injection rates at different injectors located in the reservoir. The key conclusions are that injector-producer weights are (a) independent of average well rates, (b) appear to reflect geologic features, (c) require some filtering in the presence of dissipation (indeed, with excessive dissipation the method fails entirely), and (d) can be calculated with several statistical procedures. The best thing about the method is that it can be used on virtually any injection process because injection-production rate data exists on every type of mature project.
Currently, research efforts are focused on demonstrating the technique on a truly definitive data set. The Magnus data (based on daily rate data) was obtained some months ago and application of the technique on a dataset so large would require substantial modifications to the approach for calculating inter-well correlation.