Methods of provenance determination tested with discriminant function analysis.

Currently popular approaches to provenance identification commonly use mineralogic and chemical compositions of siliciclastic sediments or of detrital clastic grains far graphical solutions. We have statistically tested three deterministic models using the raw data that were originally used to construct classifications far determining the provenance of sands and sandstones. These models are based on l) detrital modes of sandstones (Dickinson et al. 1983), 2) undulosity and polycrystallinity of detrital quartz (Basu et al. 1975), and 3) chemical compositions of muddy sediments and rocks (van de Kamp and Leake 1985). Our results show that these models can successfully classify no more than 85% of their own data. We suggest that additional variables (such as heavy minerals), the combination of different methods, and statistical analysis of data are necessary to improve the reliability of provenance interpretation