RGRA across genetic groups
Table 4 and Fig. 1 show variation in the relative effect of drought on RGRA across genetic groups with genetic group Budongo being the most strongly affected and genetic group Kibale being least affected by drought (see also significant genetic group * treatment effect Table 3). Under ample-water conditions, the absolute RGRA did not differ significantly between genetic groups while it did under restricted-water conditions. (Table 4; Fig. 1). Under restricted-water conditions, genetic group Zoka had the highest RGRA which was 12.0 % higher than the lowest RGRA observed for genotypes from genetic group Budongo (Table 4). Additionally, Fig. 1 and standard errors of means (Table 4) suggest that there was wider genotypic variation in RGRAacross genetic groups under ample-water conditions than there was under restricted-water conditions. Fig. 1. Mean RGRA [d-1] as a function of treatment (ample-water (AW) and restricted-water (RW) across genetic groups (panels) and genotypes (coloured lines). Solid black line shows the mean estimated response per genetic group.
RGRA across locations
There was a large variation in the relative effect of drought on RGRA of genotypes collected from the different locations (Table 4 and Appendix Fig. A.3.). The effect of drought on RGRA was significant for all locations except for Kibale and Itwara (Table 4; Appendix Fig. A.3.; Appendix Table A.5.). The mean percentage change in performance was highest among genotypes collected from Malabigambo, Budongo, Mabira and Kalangala, respectively, while the effect of restricted-water supply was smallest for genotypes collected from Kibale, Itwara, Zoka and Kituza, respectively (Table 4 and slope of the black lines in Appendix Fig. A. 3.). In absolute terms, under ample-water conditions, genotypes from Mabira had a significantly higher mean RGRA, which was 27.4 % higher than the lowest mean RGRA in location Kibale (Table 4 and Appendix Fig. S3). Similarly, in restricted-water conditions, Mabira had the highest and Kibale had the lowest RGRA but the difference was much smaller (8.5%) (Table 4 and Appendix Fig. A.3.). Therefore, differences between locations tended to converge in the restricted-water treatment.
Across the studied experimental factors (cultivation status, genetic group and location), it is worth noting that results showed a tendency of some genotypes to have higher RGRA under restricted-water conditions than with ample-water although this effect was not statistically significant in any of these cases (p > 0.05) (Fig. 1 and Appendix Fig. A.3.). The effect occurred in genotypes with both high and low RGRA values in the ample-water treatment and therefore are very unlikely an experimental artefact, whereby the genotypes could not have been adequately watered under ample-water conditions. Additionally, for some genotypes, the effect could be due to variations in sample size causing the mean in restricted-water to be higher than that under ample-water conditions.
Multivariate analysis of growth-related traits
The PCA analysis showed that TNL, TL and TLDW were most loaded on the first PCA axis (explaining 46% of the variation), while SLA was mostly loaded on the second PCA axis (explaining 20% of the variation). See Fig. 2. The PCA on the individual replicates showed a similar pattern. Therefore, SLA varied mostly independently of TL(correlation -0.002). The PERMANOVA showed that treatment, location and cultivation status significantly affected the dissimilarities between genotypes (p-values respectively <0.001, 0.03, <0.001) see Appendix Table A.6. Treatment explained 20% of the variation in the traits, location 10% and cultivation status only 1.8%.