*2.9. Data Analysis*

Univariate analysis of the variables chlorophyll a, chlorophyll b, chlorophyll a/b, total carotenoid, total anthocyanin, and total wax concentrations from lamina and petiole, NDVI, GNDVI, PRI, NPCI and WBI were computed by using the mixed procedure of SAS, Version 9.4, SAS Institute. Before undertaking the analysis variance (ANOVA), normal distribution of residuals and homogeneity of variances, the two major ANOVA assumptions, were checked according to Shapiro-Wilk and Levene and tests, respectively [36,37] and these assumptions were met for each variable. Then ANOVA with the least square difference (LSD) test (*p* < 0.05) was performed on each variable. The designation of environments into three stress levels, "control", "intermediate" and "stress", was based on the intensity of stress at each environment as described by Tafesse et al. [4]. For growth habit, leaf type, and canopy color, seven combinations of canopy groups, designated as "type" were used for testing the effect of canopy type. Thus, the three main treatment factors were environmental stress that we simply referred to as 'environmental stress level (ESL)', canopy 'type', and 'cultivar'. The effects of ESL, type, cultivar, ESL by cultivar, and ESL by type nested in cultivar interactions were treated as fixed effects, and block nested in environment was treated as a random effect. Whenever the interaction term was significant, a separate analysis was performed for each of the three ESLs and the results of the 'control' and 'stress' levels are shown while the result of the 'intermediate' that generally lay between the two ESLs is omitted in figures to save space. Pearson correlations test were performed among the variables of canopy temperature, pigments, wax, and vegetation indices, and significance was declared at *p* < 0.05 for combined data using four environments, control and stress.
