Abstract
A new evaluation is made of the uncertainty in pressure when a barometer is applied to an assemblage in a rock. Sources of error considered include: accuracy of the experimentally located, barometric end-member reaction, volume measurement errors, analytical imprecision, uncertainty of electron microprobe standard compositions and a-factors, thermometer calibration errors, variation in garnet and plagioclase activity models, and compositional heterogeneity of natural minerals. Whereas calibration and analytical errors can be treated statistically, variability in activity models and natural compositional heterogeneity probably cannot be.
Our estimates (1σ or quarterwidth) of the propagated uncertainties in pressure corresponding to these sources of error based on analysis of five barometers are barometer calibration uncertainties of ±300 to ±400 bars, volume measurement errors of ±2.5 to ±10 bars, analytical imprecision of ±55 to ±185 bars, thermometer calibration inaccuracy of ±250 to ±1000 bars, variability in activity models of ±60 to ±1500 bars, and natural compositional heterogeneity (sample dependent, but reportedly) of ±150 to ±500 bars. Assuming 1% uncertainties in analytical standard compositions and α-factors results in propagated errors of ±40 bars and ±150 bars (1σ) for the GASP barometer when the same standards and when different standards are used respectively for analyzing unknown phases. The accuracy of a barometer as applied to rocks may typically range from ±600 to ±3250 bars (1σ or quarterwidth), with the most significant sources of error being uncertainty in thermometer calibration and poorly constrained activity models. Continued experimental and empirical work aimed at minimizing thermometer uncertainties and better constraining activity-composition relations should substantially reduce the propagated uncertainty.
