Search

The demanding precision of triple oxygen isotope (Δ17O) measurements in water has restricted their measurement to dual-inlet mass spectrometry until the recent development of commercially available infrared-laser analyzers. Laser-based measurements of triple oxygen isotope ratios are now increasingly performed by laboratories seeking to better constrain the source and history of meteoric waters. However, in practice, these measurements are subject to large analytical errors that remain poorly documented in scientific literature and by instrument manufacturers, which can effectively restrict the confident application of Δ17O to settings where variations are relatively large (~ 25-60 per meg). We present our operating method of a Picarro L2140-i cavity ringdown spectrometer during the analysis of low-latitude rainwaters where confidently resolving daily variations in Δ17O (differences of ~10-20 per meg) was desired. Our approach was optimized over ~3 years and uses a combination of published best-practices plus additional steps to combat spectral contamination of trace amounts of dissolved organics, which, for Δ17O, emerges as a much more substantial problem than previously documented, even in pure rainwater. We resolve the extreme sensitivity of the Δ17O measurement to organics through their removal via Picarro's micro-combustion module, whose performance is evaluated each sequence using alcohol-spiked standards. While correction for sample-to-sample memory and instrumental drift significantly improves traditional isotope metrics, these corrections have only marginal impact (0-1 per meg error reduction) on Δ17O. Our post-processing scheme uses the analyzer's high-resolution data, which improves δ2H measurement (0.25 ‰ error reduction) and allows for much more rich troubleshooting and data-processing compared to the default user-facing data output. In addition to competitive performance for traditional isotope metrics, we report a longterm, control standard root-mean-square-error for Δ17O of 11 per meg. Overall performance (Δ17O error of 7 per meg, calculated by averaging 3 replicates spread across distinct, independently calibrated sequences) is comparable to mass spectrometry and requires only ~6.3 h per sample. We demonstrate the impact of our approach using a rainfall dataset from Uganda and offer recommendations for other efforts that aim to measure meteoric Δ17O via CRDS. [ABSTRACT FROM AUTHOR]