4.3 | Mesophyll conductance influenced GPPiso/SF estimates
The calculation of WUEi would not have been valid if gm had been considered infinite (Seibt et al., 2008; Wingate et al., 2007). Yet gm is still frequently ignored by some global photosynthesis models and ecophysiologists (Hu et al., 2010; Rogers et al., 2017; Zhao et al., 2005), or is embedded within a constant empirical adjustment (Cernusak et al., 2013) most likely due to the challenges in its measurements (Flexas et al., 2008; Pons et al., 2009). Likewise, the global modelling community has been reluctant to account for it because of the lack of consensus about how to measure or model it (Rogers, Medlyn, & Dukes 2014).
We compared three different ways of accounting for gm. Simplest would be to assume a constant mean value (Keenan, Sabate, & Gracia, 2010). For example, we estimated GPP with a constant gm = 0.31 mol CO2 m-2s-1 measured at the site (Stangl et al., 2019). The GPPiso/SF from the assumptions of gm = 0.31 mol CO2 m-2 s-1was not different from the GPPiso/SF from the gm∞ assumption. Perhaps this is because the constant gm value was estimated during sunny days in the summertime and therefore represents the maximal gm, under optimal conditions.
We therefore based our comparison with PRELES on a constant ratio: gm/\(g_{C\hat{\alpha}}\) = 2.67. The ratio has the advantage of allowing gm to vary seasonally. Variation responds to environmental factors (Bickford, Hanson, & McDowell, 2010; Cano, López, & Warren, 2014; Han et al., 2016; Xiong, Douthe, & Flexas, 2018); both diurnal (Bickford et al., 2010; Peguero-Pina et al., 2017; Stangl et al., 2019) and seasonal (Montpied, Granier, & Dreyer, 2009) variations have been reported. The use of a constant gm/\(g_{C\hat{\alpha}}\) ratio was certainly artificial (Xiong et al., 2018), but it is a relatively common assumption (Klein et al., 2016; Maseyk, Hemming, Anger, Leavitt, & Yakir, 2011). We suspect that the higher discrepancies between the GPPiso/SF and GPPPRELES in the fall and to a lesser extent in the spring occurred because the constant ratio did not adequately account for seasonal dynamics in gm. The need to refine our description of gm is confirmed by the uncertainty analysis (Table S1 and Figure S3) The Sobol indices, which describe sources of uncertainty, showed that almost 75% of the GPPiso/SF uncertainty came from the gm/\(g_{C\hat{\alpha}}\) estimate.