4.2.1 | Combination of sap flux and isotopic measurements.
The key advantage of the sap flux/isotopic approach is that it is independent of eddy covariance. Moreover, it leans on two methods, sap flux (Poyatos et al., 2007) and isotopic measurements (Bowling, Pataki, & Randerson, 2008 and references therein) that have been widely used at many sites by ecosystem ecologists. The sap flux/isotopic approach combines them to estimate GPP at the tree scale, which can then be scaled up to the stand. In simple stand structures, that scaling is relatively easy. In sum, the sap flux/isotopic method provides a simple method to estimate tree GPP that, in combination with measurements of ground vegetation GPP, can provide an alternative estimate for comparison with GPP estimated by EC.
An additional strength of the sap flux/isotopic approach is that it provides more detail on the ecophysiological mechanisms underlying GPP variation. For example, we showed here that \(\ g_{C\hat{\alpha}}\) was very similar between plots whereas WUEi was different between the F and the R plot. This led us to conclude that the photosynthetic activity rather than stomatal conductance was the main driver of GPP variation between plots.
One critical advantage of the sap flux/isotopic method for estimating GPP is that its requirements for the terrain and atmospheric conditions are less restrictive than for EC measurements. It thus provides an empirical method that can be applied in hilly topography, complex canopy structure, and non-turbulent atmospheres.
The sap flux/isotopic method also has several important limitations. In particular, sap flux is a key variable in the GPPiso/SFapproach in order to obtain transpiration. Although the technique describes temporal variation well, its use for quantitative estimates requires accounting for several known sources of variation (Oren, Phillips, Katul, Ewers, & Pataki, 1998). Examples include sap flux trends radially in the stem (Cohen, Cohen, Cantuarias Aviles, & Schiller 2008; Ford, McGuire, Mitchell, & Teskey, 2004; Phillips, Oren, & Zimmermann, 1996; Renninger & Schäfer, 2012), azimuthally around the stem (Cohen et al., 2008; Oren, Phillips, Ewers, Pataki, & Megonigal, 1999), with tree size (Schäfer, Oren, & Tenhunen, 2000), and with local competition (Xiong et al., 2015). In addition, corrections are required when probe length exceed the sapwood depth (Clearwater, Meinzer, Andrade, Goldstein, & Holbrook, 1999). Finally, the probes often require specific calibration (Steppe, De Pauw, Doody, & Teskey, 2010; Sun, Aubrey, & Teskey, 2012). Some corrections have been proposed to reduce uncertainties from random variation (Peters et al., 2018; Steppe et al., 2010; Sun et al., 2012), yet some tree-to-tree variation remains (Oren et al., 1998).
The model we used to estimate transpiration was carefully built to account for these errors. Tor-Ngern et al. (2017) began with high quality data based on careful accounting for radial and azimuthal variations and baseline corrections. They used one of the genera that Granier originally calibrated for. They recognised that the sensors were not specifically calibrated for P. sylvestris , but the values agreed well with previously reported results and were robust to the errors induced by the probes (Lundblad, Lagergren, & Lindroth, 2001; Poyatos et al., 2007). Likewise the data were carefully scaled up to the stand using detailed descriptions of the allometric parameters and tree sizes (Ford et al., 2004; Oren et al., 1998). Because the sap flux/isotopic method is so dependent on quantitative sap flux data, other users must also ensure that their sap flux data remove any bias and thus are accurate as well as precise.
4.2.2 | High variability of VPD impacts on\(\mathbf{g}_{\mathbf{C}\hat{\mathbf{\alpha}}}\)
There was considerable variation in our estimate of\(g_{C\hat{\alpha}}\). Because \(g_{C\hat{\alpha}}\) was calculated as the ratio between transpiration and VPD, low VPD caused high variability and improbable \(g_{C\hat{\alpha}}\) results (Eqn. 6, Ewers, Oren, Johnsen, & Landsberg, 2001; Ewers & Oren, 2000). This was especially true in the early and late growing season. The same phenomenon occurred sporadically during the thermal growing season. It forced us to apply a filter and to replace the inconsistent data inside the thermal growing season by predictions from a simple regression between gC and\(\ \hat{\alpha}\). This filtering and replacement is a common procedure, especially at high latitudes where VPD is low (Emberson et al., 2000; Tarvainen et al., 2015). Although we are satisfied with this solution for the moment, a better means of dealing with low VPD should be sought.