USING LARGE DATASETS IN FLOOD BASALT STUDIES: THE IMPORTANCE OF “DATA AWARENESS” AS ILLUSTRATED BY THE WANAPUM BASALTS, COLUMBIA RIVER BASALT GROUP

Unraveling the stratigraphy and eruptive history of a CFB province with 100s of individual flows requires the integration of fieldwork with internally consistent geochemical data. The Wanapum Basalt, Columbia River Basalt Group (CRBG), consists of 7 members and ~70 recognizable flows erupted during the waning of the main CRBG eruptions. Flow volumes range from ~10 to >3000 km³.

Nearly 2000 samples were analyzed for major and trace elements over 40 years at the University of Massachusetts-Amherst in order to refine the stratigraphy and eruptive history of the Wanapum Basalt. These samples were analyzed under the same operating conditions using a single XRF for major elements and two XRFs for trace elements. Nearly half of the samples analyzed for trace elements on the older XRF were subsequently re-analyzed when a newer XRF was acquired. This data set allows for an examination of how subtle variations in factors such as calibration may affect the data. Flow heterogeneity using the homogeneity index (Rhodes, 1983) and alteration indices (e.g., Chemical Index of Alteration (Nesbitt and Young, 1989); Mafic Index of Alteration (Babechuk et al., 2014)) are used as illustrative examples.

The Heterogeneity Index (HI) reflects the variation within a flow relative to analytical precision as determined from variations in the BHVO-2 standard used as a monitor of XRF performance. Major element HI’s tend to be significantly larger within a single calibration relative to multiple calibrations. For example, Sand Hollow HI’s vary between ~20 and ~45 for single calibrations between 2007 and 2010 vs. ~3.8 for 15 calibrations between 2007 and 2016. Given the similarity of chemical variations within a flow between calibrations, HI variations highlight its sensitivity to analytical precision. In contrast, alteration indices (AI) tend to be consistent within and between calibrations (within 2 to 3% of average) facilitating comparison of AI’s over multiple calibrations. However, when comparing data obtained using the same analytical technique but from different labs, the potential impact of factors such as differences in sample preparation techniques, the machines used for analysis, and how these machines were calibrated must be considered. This has implications for those working with large geochemical databases (e.g., GEOROC) containing data sourced from multiple labs.