Model biases
The model tended to over-estimate Anet and under-estimate gs , although the correlations between the measured and modelled data were quite strong (R> 0.9 for both) suggesting that our modeling correctly captured the behaviour of Anet andgs . These biases have several possible sources: 1) the RACiR measurements were obtained as soon as the leaf chamber stabilized, which means that while the RACiR data may reflect the underlying biochemistry, the stomata will not have adjusted to the irradiance and the vapor pressure deficit in the leaf chamber that drives the Medlyn et al . (2011) stomatal conductance model. Since the modeling assumes a steady state, this could contribute to the biases. 2) RACiR is known to introduce metabolic mismatches that do not affect Vcmax and Jmaxestimates, but may affect Anet values (Stinzianoet al ., 2019ab). These metabolic mismatches might thus contribute to some of the bias, although it should be minimal at the RACiR rates used (i.e. 100 μmol CO2 mol-1min-1). 3) It is possible that respiration in the light is greater than respiration in the dark for some of these species. In this case, modelled Anet would necessarily over-estimate measurements of Anet . 4) We used the general slope from Lin et al . (2015) for the Medlyn et al . (2011) model. It is also possible that there is variation in the stomatal slope parameter (g1) between species (Wolz et al 2017; Miner et al 2017; Franks et al. 2018).