Pilot Study using Elemental Geochemistry as a Means of Discriminating Chert Stratigraphy in the Roberts Mountains Allochthon, Nevada
AuthorBenge, Elizabeth Marie
AdvisorNoble, Paula J
Geological Sciences and Engineering
StatisticsView Usage Statistics
Biogenic chert beds in the Roberts Mountain Allochthon (RMA) are analyzed for major and trace element geochemistry with the goal of developing a geochemical means to differentiate units. This detailed chemical profile will allow for the identification of unknown cherts through the use of chemical data instead of time intensive, and sometimes difficult fossil identification. The data set consists of 82 samples used for handheld XRF analysis and 66 for ICP-MS and ICP-OES analysis. Samples were collected from 6 localities in 5 ranges, all with known ages based on radiolarian biostratigraphy. These samples are assigned to either the Slaven, Cherry Spring, or upper Vinini chert, or equivalent units. Several analytical methods were employed to test their effectiveness in discriminating the cherts. The methods used are a handheld Niton XRF (pXRF) on hand samples, ICP-OES to determine percent oxide concentrations, and ICP-MS for trace element concentrations. Data were analyzed using multivariate linear discriminant analysis (LDA) to develop a classification model on known samples, and the LDA models were tested with several prediction functions. Based on the sample set size, the Leave One Out Cross Validation method (LOOCV) proved to be the most reliable statistical tool for evaluating model performance between the analytical methods. The pXRF, data on in situ samples, while the fastest and cheapest form of analysis, was the least reliable prediction method correctly predicting 61.6% of the time on a subset of data that removed elements where the majority was below detection limits (LOD). Performing pXRF on powdered samples increased the predictability from 61.6% to 75.4% using the LOOCV test on data where elements returning >50% LOD (lower than detection) were removed. Increased run time on powdered pXRF samples did decrease the number of lower than detection (LOD) values, but did not increase the predictability. The ICP-MS data, returning trace elements, including REE’s is a slightly more reliable method of prediction over the powdered pXRF, although more expensive to run. The LOOCV test of the ICP-MS model correctly predicted units 78.9% of the time. ICP-ES data were not a useful predictor of unit, but provided some information regarding provenance. Together ICP-ES and REE data rule out an open ocean setting for the RMA chert units, and instead point to a continental margin environment of deposition, based on the presence of an Eu anomaly and lack of a Ce anomaly. Collectively, these data show that it is possible to discriminate between the chert units in the RMA using geochemistry, even when correlating chert from the Independence to the Shoshone range, and that there are tradeoffs between cost and reliability between pXRF and ICP-MS data. They further illustrate the utility of these geochemical data as a means of evaluating geologic provenance of chert.