Statistical properties of groundwater noble gas paleoclimate models: Are they robust and unbiased estimators?

Noble gas paleoclimate reconstructions from noble gas concentrations in groundwater have the promise of providing absolute paleotemperature information, but difficulties in modeling measured noble gas concentrations with simple models have led to a number of competing noble gas temperature (NGT) schemes being developed. These different models make different assumptions about the size and nature of the noble gas air‐saturated water and excess air (EA) components within groundwater, and they yield significantly different NGT values (∼5°C). We use four different NGT models (unfractionated air, closed system equilibration (CE), oxygen depletion, and gas diffusion relaxation) to examine seven published noble gas data sets and find that although different NGT models produce records that are offset with respect to temperature and excess air, the amount of variation of fitted temperature and excess air values within a given record is surprisingly consistent between fitting methods. By using the NGT methods on 5000 synthetically produced data for each model, we show that all but the closed system equilibrium or CE model of Aeschbach‐Hertig et al. (2000) are unbiased estimators, with the CE model tending to overestimate both NGT and the EA component. However, despite offsets in NGT and EA, all four models agree remarkably well when estimating temperature differences, confirming the pattern seen with real data, and this suggests that temperature variations derived from NGT data are quite robust. Patterns of misfit of noble gases are shown for both synthetic and real data, which help to provide diagnostic tools for assessing the appropriateness of different NGT methods for different noble gas data sets.