3.5 | Comparison of GPP estimates
Our first objective was to compare GPPiso/SF to
GPPPRELES for 2012 and 2013. To simplify the figure, we
chose to represent only the assumption that
gm/\(g_{C\hat{\alpha}}\) = 2.67 (Figure 5), which allows
gm to vary during the season. The seasonal GPP patterns
were similar between PRELES and the sap flux/isotopic method (Figure 5).
Recall that GPPPRELES included understorey vegetation.
Correlation coefficients among methods and plots were all high, with
minimum r = 0.91 (Figure S4). However, the fit was nonlinear; in 2012
and 2013 GPPiso/SF approached an asymptote at high
levels of GPPPRELES (Figure S4). The highest
GPPPRELES values did not match with the highest
GPPiso/SF values; the peak of GPPiso/SFoccurred earlier in the season than those of GPPPRELES.
Interestingly, confidence intervals for GPPiso/SF and
GPPPRELES overlapped most of the time, even during the
fall, when the offset was bigger than the rest of the year. However, the
VPD filters removed many values during the fall, which allowed us to
draw a confidence interval only during small periods at that time. As
previously mentioned, the GPP values were gapfilled to draw a complete
seasonal pattern, at least during the thermal growing season. The
resulting annual sums were higher for GPPiso/SF than for
PRELES on the control plot, but not on the fertilised plot (Figure 6).