Publication Date:
2017-11-16
Description:
Carbonate clumped isotopes offer a potentially transformational tool to interpret Earth's history, but the proxy is still limited by poor inter-laboratory reproducibility. Here, we focus on the uncertainties that result from the analysis of only a few replicate measurements to understand the extent to which unconstrained errors affect calibration relationships and paleoclimate reconstructions. We find that highly precise data can be routinely obtained with multiple replicate analyses, but this is not always done in many laboratories. For instance, using published estimates of external reproducibilities we find that typical clumped isotope measurements (3 replicate analyses) have margins of error at the 95% confidence level (CL) that are too large for many applications. These errors, however, can be systematically reduced with more replicate measurements. Second, using a Monte Carlo-type simulation we demonstrate that the degree of disagreement on published calibration slopes is about what we should expect considering the precision of Δ 47 data, the number of samples and replicate analyses, and the temperature range covered in published calibration. Finally, we show that the way errors are typically reported in clumped isotope data can be problematic and lead to the impression that data is more precise than warranted. We recommend that uncertainties in Δ 47 data should no longer be reported as the standard error of a few replicate measurements. Instead, uncertainties should be reported as margins of error at a specified confidence level (e.g., 68% or 95% CL). These error bars are a more realistic indication of the reliability of a measurement.
Electronic ISSN:
1525-2027
Topics:
Chemistry and Pharmacology
,
Geosciences
,
Physics
Permalink