Chemistry Reference
In-Depth Information
Table 3 presents the correlation matrix between ecophysiological and environmental
variables in the three scales of study. The best associations were present in the relations E
and Gs with Qleaf, reinforcing the behavior displayed in Figure 2. On the plot scale, all
correlations presented were significant and as in the pot scale, the relation between Gs and
VPD was found to be negative, while at the watershed scale it was not consistent.
Variables
Pot
Plot
Watershed
E x Qleaf
0,86**
0,83**
0,87*
E x VPD
0,30
ns
0,63**
0,56*
Gs x Qleaf
0,80**
0,85**
0,66*
Gs x VPD
-0,76**
-0,33**
0,07
ns
** e * = significant correlation at 1% and 5% respectively, ns = non-significant
Table 3.
Simple correlation matrix between ecophysiological and environmental variables on pot, plot
and watershed scales.
3.3. Scaling up: pot, plot and watershed
Analyses of variance among the parameters were significant at 1% probability. The
mathematical equations as well as the comparison between the hourly mean values
observed and simulated by the models are in Tables 4, 5, 6 and 7. Among the relations, E
was associated more evenly with Qleaf, with the highest coefficients of determination (R ²)
when compared to VPD, regardless of the scale of observation. For this reason, Qleaf can be
used more safely than other variables for being more consistent.
In model
E = f (Qleaf)
the mean test was significant for
E
V
and
E
B
´
, indicating that the model
of the pot scale, adjusted with k allowed extrapolation to the plot and watershed scales.
Although the result of Table 2 has shown that the ratio E/Qleaf between plot and watershed
scales are statistically different, it was not necessary to use k for the prediction of E
B
'' (Table
2). The values observed and simulated by the models were compared and showed good
correlation coefficients (Figures 3a, 3b and 3c), although the model underestimated the
values in some situations (Figures 3b and 3c). As for the variable G, the model
Gs = f (Qleaf)
could be applied in scaling up Gs
P
'and Gs
B
' (Figure 4) not being significant for Gs
B
''. The
model
E = f (VPD)
was adjusted for scaling up E
VP
and E
VB
, but in the extrapolation of plot
scale to watershed scale it was not necessary to adjust the constant k, reinforcing what was
already observed in Figure 2 and Table 2. Although in this case the scaling up being
possible, the correlation coefficients between the observed and simulated values by the
model were between 0.66 and 0.62 and were highly significant (P <0.01) (Figures 3d, 3e and
3f ). For
Gs = f (VPD)
, the proposed methodology can not be applied in any situation.
