Geochemical expert system

Expert System for the Characterization of Rock Types (ESCORT) is an expert system based on Bayesian rules providing probabilities for the occurrence of rock types based on geochemical and nongeochemical data. 

Data on samples to be examined are given in a table which is read by the expert system program. The table includes sample identification, analytical data, and the results of simple calculations such as for charge balance and the sum of analyzed dissolved constituents. It also includes mineral saturation indices and data on the carbonate system calculated by the geochemical modeling program. 

The second module, the carbonate module, examines data on the concentration and distribution of dissolved carbonate species including the results of geochemical modeling. A variety of carbonate data are available for many analyses. These provide redundancies useful to evaluate the carbonate system but which lead to a large number of permutations and combinations of data by which the evaluations can be made. The complexity of the carbonate system makes it particularly suitable for analysis using the inferential techniques typical of expert systems. Thus, the carbonate module is discussed in detail in the following section. 

Having produced a better pH value, a coupled expert system could call a geochemical modeling program and rerun it with the new pH. The newly modeled total dissolved carbonate values and carbonate mineral saturation indices could then be substituted for the old values in the water data set and the expert system rerun to test for improved consistency among the new water data. 

Such a coupled expert system would be able not only to evaluate analytical data, but to adjust them to improve both their internal consistency and the accuracy with which they represent the chemistry of water in the formation. This evaluation and adjustment, if done at all, now requires manual calculation and geochemical modeling by a human expert. Manual calculations are likely to be slower and their results less consistent from expert to expert, or from time to time from a single expert, than results guided by an expert system program. 

The present expert system program discovers possible errors and inconsistencies in a ground-water analysis, but it is restricted to water from carbonate aquifers and it cannot directly interact with geochemical modeling programs. The next step in its development will be to couple the carbonate module to a geochemical model to enable the expert system to adjust an analysis for consistency among the parameters of the carbonate system. The expert system will then be applicable in practice to adjust... 

In principle, such an expert system could also consider the extent to which a wide variety of minerals would influence ground-water chemistry at various temperatures and reaction times, as well as the natural variability and experimental uncertainties associated with the thermochemical data used to model geochemical equilibrium. To develop a consensus in the geochemical community about the categories of information required to evaluate mineral reactivity and the properties appropriate for each of many minerals, would be a time consuming task. Thus, in the near future at least, such enhancements will probably be limited to consideration of tightly defined problems of simple mineralogy. 

---