ABSTRACT

Atlas mapping is fundamentally based on a number of existing digital data bases, both public and private. The most important of these are the provincial files for index and stratigraphic data, and the Canstrat files for lithology. Integrated computer processing of these data bases is designed to perform automatic distillation of raw geological data into map input information. In geological investigations, this basic task is routinely undertaken manually by a geologist. However, given the size of the Atlas project, the vast volume of data (~ 200,000 wells), the disparity of provincial data bases, and the degree of redundancy and inconsistency in the data, manual evaluation and filtering of the well data would require commitment of time and effort far exceeding the available resources. Consequently, a series of programs has been developed for the purpose of automating as many steps of the preliminary data assessment as possible. These programs comprise an electronic data processing system that transforms the raw data into first-cut map input, practically without human interaction. The geologist is requested only to supply a few parameters at the programs' initiation.

The core of the system is made up of about 50 procedures that integrate, filter and organize the data for further processing. They perform an integrity check, eliminate redundant and unusable data, correct errors, evaluate and merge revised (cloned) wells, identify stratigraphic synonyms, and build data structures.

The filtered raw data may be directed into one of three streams:

1. Selection of an optimized set of control wells (Atlas data base).
2. Selection of an optimized set of control picks (Atlas standard set).
3. Generation of map input from the entire bank of clean data.

The data derived by options (1) and (2) can be used directly for the generation of map input, or can be dispatched to a set of programs that evaluate data conformity and variability through neighbourhood cross-reference and surface modelling. Reports on missing or potentially erroneous data also can be produced. Feedback to the system (geologists' corrections) is assessed in the same way. Analysis and hierarchical classification of sequence types within a specified stratigraphic slice is yet another option offered by the system, for both the raw and derived data.

To perform the above tasks, the system requires some built-in analytical and pattern recognition capacities. When the program attempts to eliminate an inconsistency in the pick sequence, or to isolate stratigraphic synonyms, it must make a decision that requires some understanding of stratigraphic relationships. Yet the diversity and complexity of the regional stratigraphy precludes the usage of reference tables as the system's guide. Thus, the programs extract the stratigraphic knowledge directly from the data, by examining and storing the adjacency relations between picks, and using the stored information, together with a few optimization algorithms and heuristic rules, to arrive at geologically sensible decisions. It should be emphasized that although most of the post-selection processing concentrates on the 10,000 Atlas control wells, the system undertakes continuous evaluation of each well in the context of its neighbourhood (host township). This allows for properly assessed supplemental data to be brought to bear on the mapping, and minimizes the loss of information induced by selection.

End_of_Record - Last_Page 180-------