Data analysis

All analyses were performed in SAS v 9.3 [49] and only offspring generation data (S1 File) were included in the analyses (See [36] for parental generation results). Mixed-model ANOVAs using TYPE III sum of squares (PROC MIXED) were used to test whether populations are adapted to their home soil salinity levels and the contribution of parental environment on populations response to salinity. Tests of salinity adaptation were performed using two measures of performance (i.e., dry vegetative biomass, reproduction). For each mixed-model ANOVA, soil origin, population nested within soil origin, parental and offspring environment, and all interactions with these factors were treated as fixed effects. Genotype nested within soil origin and population was treated as a random effect. Significance of the random effect (i.e., genotype and population) was calculated by quantifying the difference in -2 log likelihood scores for the full model and the model excluding the random effects term using a chi-square table with 1 degree of freedom [50]. To meet model assumptions of normality and homoscedasticity, dry vegetative biomass was natural logarithm transformed and number of pods was square root transformed.

A significant origin by offspring environment interaction would indicate that populations originating from each soil salinity level are adapted to their home soil level when each soil origin has greatest performance in its native salinity level. A significant three-way interaction among origin, parental and offspring environment would indicate that parental environmental effects influences population’s salinity adaptation. In order to directly test differences between levels in each class variable in the mixed-model ANOVA, least square means comparisons (LSMEANS) were calculated and reported throughout the results. For ease of interpretation, back transformed means and standard errors are reported in text.

The same mixed-model ANOVAs were used to quantify the effects of soil origin, population, parental environment, and offspring environment on traits associated with salinity avoidance and tolerance. We considered two phenological traits associated with salinity avoidance: germination, flowering; and four morphological and physiological traits associated with salinity tolerance: carbon acquisition rate, leaf number, leaf area, leaf water content. Data were natural logarithm or square root transformed when needed to meet model assumptions of normality and homoscedasticity. Analysis on carbon acquisition rate included the covariate of day of measurement because it had a large effect on measurements.