Statistical analysis
To compare the stable isotopic composition (D/H and18O/16O isotope ratios) considering
natural abundance ( ‰) in xylem sap, plant (Ψleafin MPa) and soil (Ψsoil in MPa) water potentials,
we implemented a multifactorial repeated-measures analysis of variance
using a mixed model, which included as classification factors, species
with two levels (Q. potosina and P. cembroides , fixed
effect), forest stand with two levels (pure and mixed, fixed effect) and
time (sampling dates, random effect) with 56 levels (biweekly dates).
For the fracture water injection assay, we implemented an analysis of
variance to examine the isotopic composition of xylem sap, using a mixed
model with species (Q. potosina and P. cembroides , fixed
effect), forest stand (fixed effect), tree distance (0-2, 2-4, 4-6 m;
fixed effect) and time (sampling dates, random effect) as classification
factors. Also, to compare changes of diameter at breast height and wood
moisture content a 2x2 mixed model factorial including as factors
species with two levels (pine and oak) and ecohydrological periods with
five levels (depletion 2013, recovery 2013, wet 2013, depletion 2014,
and wet 2014) was applied to compare changes of diameter at breast
height and wood moisture content. In addition, a nested two-way
factorial was used to compare treatment effects on root frequency for
each diameter class. The model included forest stand with three levels
(pure and mixed pine and oak stands; fixed effect) and soil depth with
four levels (0-10, 10-20, 20-30 and >30 cm; nested effect).
In all cases, we conducted Tukey’s post hoc mean comparison test.
Regarding the labeled water treatment in P. cembroides trunks, we
analyzed the presence of isotopic label in twigs using a one-way ANOVA
followed by a Tukey’s post hoc mean comparison test. We ran Type
I regression analyses to examine the relationships between wood moisture
content and tree diameter at breast height for P. cembroides andQ. potosina . Polynomial quadratic regression analyses were
implemented to examine the relationship between soil water potential
(Ψsoil , MPa) and resistivity (ERT, Ω
m−1) between October 2013 and May 2014. Before
statistical analyses, response variables were examined for normality
using the Shapiro–Wilk’s test (Shapiro &
Wilk, 1965). Spatial analysis of species-specific vertical root
distribution at different soil depths and soil electrical resistivity
were examined using the Kriging interpolation method (Empirical Bayesian
Kriging Simulations). All statistical analyses were run in SAS
University Edition (Free Statistical Software) using PROC MIXED for the
repeated-measures analysis and PROC GLM for generalized linear model for
the univariate analysis. All geostatistical analyses were run with
ArcGIS v. 10.1 for Windows (ArcGIS Desktop, ESRI 2011). The relative
contributions of different sources to xylem water were estimated by
Bayesian mixing models using the Stable Isotope Analysis in R (siar)
package (Parnell, Inger, Bearhop, &
Jackson, 2010).