Phenotyping
We focused on ten traits related to differential water availability across the entire plant life-cycle.
Germination fraction was censused per pot after germination had ceased at the beginning of the experiment. It typically decreases towards arid populations to hedge against more frequent unfavorable years (Tielbörger et al . 2012; Lampei et al. 2017; ten Brink et al. 2020). As climate manipulations increased (dry plots) and decreased (wet plots) the occurrence of unfavorable years, we expected reduced germination fractions being favored in dry plots andvice-versa in wet plots.
Days to flowering (since first irrigation) were determined by inspecting plants daily for the first open flower. Accelerated phenology is expected by theory and repeatedly found in annual plants from drier sites (reviewed in Kigel et al. 2011), and we expected earlier phenology in plants descending from dry manipulated plots and more arid sites. Moreover, the number of leaves at the day of first flowering (leaf number at flowering ) provided an ontogenetic phenological measure and a non-destructive measure of plant size. It disentangled whether phenology changed via accelerated development (days to flowering) or shifted ontogeny (leaves at flowering) (Kigel et al . 2011).
Stomata density and carbon isotopes(δ 13C) assessed gas exchange and water use efficiency. Stomata density was quantified by automated high-throughput microscopy (Dittberner et al. 2018; see Supplementary Methods). As lower stomata density may decrease maximum transpiration (Liuet al ., 2012) we expected lower stomata density in plants descending from drier conditions. Due to high costs, carbon isotopic ratios (δ 13C, see supplementary Methods), were analyzed only for a subset (14 genotypes per site, rainfall treatment, and four water levels: 15ml – 50ml). We expected that plants from drier sites and plots exhibit higher water use efficiency, i.e. higherδ 13C (Li 1999; Hartman & Danin 2010).
Plant height was measured on a fixed day (12thApril) before the onset of senescence. Moreover, abovegroundvegetative biomass was determined at the end of the experiment (May 15th 2014) as the dry weight (24 h, 70°C) of all stems and leaves. We expected greater height and vegetative biomass in plants from wetter conditions as adaptation to intensified aboveground competition (Westoby 1998; Schiffers & Tielbörger 2006).
Total seed number per plant quantified fitness. Moreover, we estimated the selfing rate per plant visually as percent of flowers that developed into viable seeds; it served as covariate in some analyses because B. didyma populations may differ in self-compatibility (Gibson-Forty 2018).
Reproductive allocation quantified the biomass allocation to reproductions (i.e. weight of all diaspores and flower remains) relative to the vegetative biomass. Reproductive allocation should be higher in plants from drier conditions as they require less investment in vegetative tissue for outgrowing neighbors (Aronson et al. 1993).
Diaspore weight (maternal investment per single offspring) was measured across 30 randomly picked diaspores per plant. Diaspore weight consists of c. 50% of seed mass in B. didyma and both are strongly correlated (r²=0.88, p<0.001; determined for 15 seeds in 32 randomly picked individuals across sites). Although investment per offspring is a crucial feature of plant life cycles (Westoby 1998), predicting its evolutionary response to aridity is controversial (Kurzeet al. 2017).