Data for: Testing the accuracy of new paleoatmospheric CO2 proxies based on plant stable carbon isotopic composition and stomatal traits in a range of simulated paleoatmospheric O2:CO2 ratios

A drive to improve long-term estimates of atmospheric CO2 change through Earth history has led to the development of novel paleoproxy CO2 methods applicable to fossil plants. This paper compares two of these paleoproxy CO2 methods (1) the empirical model of Schubert and Jahren (2012, 2015) termed the C3 plant proxy, which was developed using the hyperbolic relationship observed between plant carbon isotope discrimination (Δ13C) and pCO2 and (2) the mechanistic model of Franks et al. (2014), which utilizes stomatal anatomy measurements and plant carbon isotope composition and is based on established equations for leaf gas exchange and photosynthesis. To date both models lack detailed experimental testing of the robustness and accuracy of their pCO2 predictions in relation to phylogenetic differences in Δ13C between C3 plant groups and/or species and atmospheric O2 concentration, which has co-fluctuated with CO2 in the geological past. Here, we investigate if these novel paleoproxy CO2 approaches can produce phylogenetically independent estimates of pCO2 that are not influenced by variations in atmospheric oxygen. To address this, model estimates of pCO2 were compared with measured CO2 values for ten plant species representing four major vascular plant groups (lycophytes, monilophytes, gymnosperms and angiosperms) grown for 6 months in walk-in plant growth chambers under varying O2:CO2 ratios. Results from the mechanistic model reveal that species-specific plant responses to atmospheric CO2 accounted for the large variability in CO2 predictions between species and overestimations of pCO2 by ~+232 ppm to 940 ppm. Adjustments to the model that involved: (1) corrections to the photorespiratory compensation point (to account for fluctuating oxygen) and (2) removal of the phylogenetic effect on Δ 13C, reduced between-species variability (by 50%) and led to better pCO2 estimates within 58 to 229 ppm of measured values. The C3 plant proxy (empirical approach) produced accurate CO2 estimates within +37 to +71 ppm of measured values, however it was affected by species-specific differences in Δ13C and for some species resulted in negative estimates of pCO2. Sub-ambient O2 (16%) resulted in erroneously high CO2 estimates (~100 ppm higher than the control) for a number of species, as plant responses to decreasing O2 mimicked those of increasing CO2. We conclude that both models (with phylogenetic corrections to Franks et al., (2014)) can produce accurate estimates of pa