IODP Proceedings    Volume contents     Search

doi:10.2204/iodp.proc.320321.203.2012

Example calibration of data: BaSO4

A final step—calibration with chemical analyses of the sediments—is required in order to properly estimate chemical composition with XRF. Figure F5 illustrates the basic process with shipboard Ba data, comparing the XRF BaSO4 estimate from the preliminary linear calibration with the shipboard measurements. The shipboard data set is not ideal for this calibration because the shipboard and XRF data were often measured on different holes but matched by the common meters composite depth used to build the splice. Nonlinear distortions of adjacent sediment columns, either by coring or local sedimentation variability, add noise to the calibration by matching samples at different sediment horizons (Hagelberg et al., 1995). Compositional differences may actually exist between the shipboard ICP-AES and XRF NMS data that we used for the initial test calibrations. Also, we did not have a sufficient number of samples to reserve some for a test data set to determine errors. Nevertheless, the correlation between the shipboard ICP-AES data and XRF Ba data is reasonably good and will be improved by measuring more samples along the splice.

A better example calibration exists for CaCO3 data (Lyle and Backman, submitted). Lyle and Backman use an extensive discrete CaCO3 data set to calibrate an XRF CaCO3 estimate. Only half of the data are used for the calibration. A reserved half of the data set is used to assess the quality of the calibration. Reserving data allowed Lyle and Backman to show that the CaCO3 estimate agreed with the measured CaCO3 by ±5 wt% (1 standard deviation).