Fig. 4. Relationship between performance of C. canephora genotypes in restricted-water conditions and wetness index of the location in which they were collected from (A) and the relationship between drought tolerance as estimated from RGRA and wetness index of the location (B); wetness index (WI), high WI values indicate moist conditions and low WI values indicate dry conditions). Both slopes were negative and significantly different from zero at p = 0.05
There was also a marginally significant negative (p = 0.05) relationship between drought tolerance of genotypes and wetness index in their location (Fig. 4 panel B), being defined by: Tolerance = 0.99 – 0.214 (wetness index), R2 = 0.05 and S.E. = 0.106. The negative relationship between tolerance and wetness index of the locations possibly indicates a climatic signature related to drought tolerance of the genotypes. This observed relationship between drought tolerance and wetness index suggests that on average, genotypes from the comparatively drier areas, e.g. Zoka, tended to be somewhat more drought tolerant than genotypes from the wetter area Kalangala. No difference in terms of goodness of fit was found between the linear and the other two types of non-linear models. Other traits also tended to show a similar trend of drought tolerance being negatively correlated with wetness index although statistically significant relations were observed in TNL only (Appendix Fig. A.4. and Appendix Box A.2.).
4. DISCUSSION
In this study, we explored Uganda’s C. canephora genotypic diversity in a screening experiment with 148 genotypes. We specifically explored: (i) the effects of drought on growth categorised by cultivation status (wild, feral and cultivated), genetic groups and the geographic location, (ii) the relationship between performance under restricted-water and performance under ample-water conditions and (iii) the relationship between drought tolerance and wetness index (WI). To our knowledge, this is the first study to explore intra-specific variation in drought responses for a large number of genotypes (> 100) in a tropical tree species.

4.1. Effect of drought on C. canephora in growth response traits

Drought significantly reduced the RGRA (relative growth rate of leaf area), TNL (total number of leaves), TL (total leaf area), TLDW (total leaf dry weight), SLA (specific leaf area) and increased the RL (root volume to leaf area) (Table 2). The latter finding concurs with the optimal partitioning theory which entails that in response to stress, plants allocate proportionally more resources to the structure capturing the most limiting resources (Brouwer, 1963; Bloom, Chapin and Mooney, 1985). Other studies (Ryser and Eek, 2000; Shipley and Meziane, 2002) and reviews (Hoffmann and Poorter, 2002; Ezizet al. , 2017) also stated that, in response to stress, plants adjust their biomass allocation in accordance to whether the most limiting resource is above- or belowground. In our study, TNL and TL were the most affected traits (Table 2) implying that genotypes responded to drought stress mainly by minimising transpirational water loss by reducing the number of leaves and leaf area. Differential reduction in leaf area as a response to drought stress has also been observed by other authors (DaMatta et al. , 2003; Pinheiro et al. , 2004; Dias et al. , 2007; King’oro, 2014). Our current findings extend these observations to a wider range of genotypes including wild, feral and cultivated material.