Grain-size data analyzed by multivariate computer techniques allows (1) discrimination of environment of deposition and (2) recognition of sedimentary structures. Multiple discriminant analysis where used in conjunction with variables derived from ¼-phi-sieve grain-size analysis was found to be the most useful multivariate technique.

Grain-size samples from 3 environments (Arkansas River, Great Sand Dunes National Monument, Colorado, and Gulf Coast beaches) were used as "known" environments in discriminant analysis. With the intention of classifying 49 samples from Bijou Creek, Colorado, they were listed in the discriminant analysis program as "unknowns." Thirty-six of 39 "unknowns" were correctly classified by the program as river sediments. Q/mode factor analysis correctly classified 32 of the Bijou Creek samples with known river samples.

Variations in grain-size distributions within a given environment were studied in an Arkansas River sand wave. Foreset beds, climbing ripples, and horizontal laminations were designated as "known" sedimentary structures. By use of multiple discriminant analysis 20 of 21 samples from a second sand wave were classified correctly. Using Q-mode factor analysis all but 3 samples were classified correctly.

By analyzing separately each of the 2 or 3 populations appearing as straight lines on cumulative plots on normal probability paper, greater discrimination between environments was obtained than by using standard grain-size parameters calculated by assuming each sample represented a single population.

End_of_Article - Last_Page 873------------