Variation with time of the physiological and growth traits in the WW and WS treatments and estimation of an ontogenetic effect
Figure 1 and Table 1 show that in the WW treatment, the SWC in the pots is stable with time (average 26.6%) while it decreases in the WS treatment from 27.88 to 12.08% from week 0 to week 6. In the WW treatment, the raw values of A, gs, diam, shoot, and height increase from week 0 to week 5, then levels off (Table 1). Needle predawn water potential (Ψb) is almost zero for the WW trees (n = 4) on week 6. Indeed, for all the WW data (Table 2a), we find a significant week (time) effect for A, gs, Wi, diam, shoot, and height with no significant week × family interaction (week × F), meaning that the week effect is the same for all the families. Using the daily weather data automatically collected by the greenhouse meteorological station, we find that there was neither significant temperature nor relative air humidity variation during the 6-week time lag of the experiment (results shown in Fig. 3 in the supplementary data). Thus, we attribute the phenotypic variation observed in the WW trees to the ontogenetic, developmental effect. We used this ontogenetic effect estimated on the WW trees to adjust the data collected on the WS trees. The unadjusted and adjusted data (A, gs, Wi, diam, shoot, and height) are shown in Table 1. The WW and WS trees have different variation patterns: the adjusted WS data shows a significant decrease of A, gs, diam, shoot, and height and a significant increase of Wi when SWC lowers (SWC is approximately 12% while Ψb of the WS trees was − 0.54 MPa ± 0.12 on week 6). The time-related SWC effect is significant on all the adjusted WS variables (Table 2d).
Variation between the four families for the physiological and growth traits
There is a significant family effect for A, gs, diam, shoot, and height in the WW treatment (Table 2a). There is also a significant family effect for Wi estimated in well-watered conditions for the WW and WS trees on the weeks 0–2 (Table 2b, F).
Average and family plastic response to the spatial and temporal variations of water availability
At the beginning of the experiment (Table 2b), there is no effect of SWC on A, gs, Wi, diam, and height for the 64 trees in both treatments. Conversely, at the end of the 6-week experiment, there is a significant SWC effect (Table 2c) on all the traits: in other words, there is a significant between-treatment phenotypic relationship between SWC and all the variables at the tree level in the pooled WW and WS treatments during weeks 5 and 6. This relationship measures the average between-treatment plastic response of the grouped families to the variation of water availability between both treatments. We note a significant family effect for gs, Wi, diam, shoot, and height (Table 2c, F) but no SWC × family interaction (Table 2c). This means that there is no significant difference between the four families for their between-treatment plastic response to SWC variation.
The significant effect of SWC on adjusted A, gs, Wi, diam, shoot, and height for the 32 trees in the WS treatment during the 6 weeks of the experiment (Table 2d) reveals an average plastic response to the temporal variation of SWC. We found a significant family effect for gs, shoot, and height and a significant SWC × family interaction for gs and Wi (Table 2d, F, SWC × F). This interaction corresponds to a significant family effect for the time-related plastic response to SWC for these two variables. Conversely, we found no significant SWC × family interaction for A and the three growth traits.
The corresponding plots are shown in Fig. 2a–f for A, gs, Wi, diam, shoot, and height, respectively. The fitted linear relationships are family norms of reaction that measure their time-related plastic response (Fig. 2). We could fit significant family norms of reaction for one to three of the four families, according to the trait, for A, gs, Wi, diam, shoot, and height. For the range of variation of SWC measured in the experiment, these norms of reaction are linear models. These family norms of reaction measure the time-related family plasticity to SWC variation in the WS treatment: there is a significant family variation for these norms of reaction for gs and Wi (Fig. 2b, c).
Correlations between variables
Table 3 shows the correlations between the physiological and growth variables. As expected, A and gs are positively correlated. Wi is highly negatively correlated with gs in the WW and WS conditions. The growth variables positively correlate all together in all cases and are positively correlated with Wi in the WS conditions.