Statistical analysis
Differences between site averages of macro- and micronutrient concentrations in the foliage and soil samples from 8 sampled sites were tested by means of one-way ANOVA - checking data and residuals distribution normality and homogeneity of variance - followed by Tukey (HSD) post-hoc tests, using the Statistica 7.0 software (Statsoft Inc., Tulsa Ok). After standardization of Na and Cl data, the causal role of NaCl accumulation in foliage regarding the univariate changes in leaf nutrients and structural parameters was tested by means of linear mixed effects model (LMEM), using the lme4 package (Bates et al. 2015) of R software, version 3.4.2 (R Core Team 2017). The individual tree formed the statistical unit whilst the site effect was treated as a random factor. Given their correlation, separate models for Na and Cl factors were calculated. For each model, the marginal R2 was calculated using the r2glmm package (Jaeger, 2017) of R. The correlations between structural changes, salt concentration in foliage and leaf injury were investigated by means of Pearson’s correlations, also using the aforementioned Statistica software, after variable normalization. Finally, the multivariate responses of palisade parenchyma cells to salt contamination in foliage was tested by means of redundancy analysis (RDA), using the vegan package (Oksanen, et al., 2017) of R and main structural size and shape variables. The dependent matrix included structural data from the aforementioned five sites subsample used for quantitative TEM, the explanatory variables consisted of foliage concentration of salt contaminants (Na, Cl), whilst leaf injury and the study site centroids were passively projected in the hyperspace determined by the RDA axes.