Genotype x environment interactions and optimum resource allocation for sugarcane yield trials in Swaziland

To account for genotype x environment (G x E) interactions, sugarcane varieties are tested in multi-environment trials (METs) across multiple crop-years and seasons (early vs. late) under fully irrigated conditions in Swaziland. This study aimed to quantify the main sources of variation relevant to the MET network, identify mega-environments, evaluate adaptability of imported sugarcane varieties, and optimize resource use within the testing network. Nine varieties (eight imported Mauritian and one South African variety) were tested at two sites (Mhlume – M and Simunye – S), harvested early (E) and late (L) in the season across five crop-years in replicated (8) Latin square designs. A linear mixed model analysis of cane yield (tons of cane per ha; TCH), sucrose content (SUC%), sucrose yield (tons of sugar per ha; TSH), and fiber content (FIB%) showed that variety x season interactions were highly significant (P < 0.001) and larger than variety x site interactions for most traits. Genotype + genotype x environment (GGE) biplots revealed clustering of environments into two mega-environments by season rather than by site. Two Mauritian varieties were superior in each of the early and late mega-environments. Broad-sense heritability (H²) for all traits was larger than 0.79, with the highest value observed for FIB% (0.89). Resource optimization analysis, using variance components, suggested that future testing be conducted at a single site, across two seasons, with replicate numbers >5, and harvested across approximately five crop-years. Strong variety x season interactions should be exploited more aggressively in irrigated sugarcane selection programs.