Pattern Recognition as Guide for Establishing Minimum Sampling Requirements in Regional Stratigraphic Studies: ABSTRACT

Abstract

The recognition of spatial order for facies defined within a sedimentary depositional framework has been accomplished traditionally by descriptive methods. This approach has limited the progress made at attempts to establish minimum sampling requirements necessary to delimit the major depositional patterns in regional stratigraphic studies. A more quantitative approach can be taken if one considers the spatial arrangement of facies as a problem in multivariate pattern recognition. This has led to the development of a statistical method for analyzing multiphase mosaics expressed in map form. Nearest-neighbor theory has been combined with cross-association analysis to provide estimates of geometric parameters defined for different classes of depositional environments. On th basis of areally sampled data, a derived pattern can be judged either as being random, in which more detailed sampling is indicated, or as being nonrandom, in which the observed pattern is compared with one of several reference patterns whose geometric parameters are specified. The sample size required to attain any desired level of correspondence for any given reference pattern can be established. The Mississippi deltaic plain was chosen as a test example of the method. Based on areal sampling, a 10% random sample is sufficient to delineate the major depositional framework, and a followup 30% systematic sample is sufficient to delineate the major facies trends. Considering the increasing cost of data storage End_Page 859------------------------------ and retrieval, this approach has potential value for future planning of regional subsurface stratigraphic studies. End_of_Article - Last_Page 860------------