Optimizing multi environment cowpea yield trial by appraising the fitness and similarities among test sites: a case study of IITA cowpea breeding program

Optimization of Multi environment trials (MET) trials require an efficient testing network consisting of ideal sites which should be highly discriminative and representative of their mega environments. Also concurrently managing sites that are similar in terms of information there give about genotype performance adds no extra utility to the breeding program. The objective of this study was to appraise the testing network of the cowpea breeding program of the International Institute of tropical Agriculture (IITA) based in Nigeria on the basis of their representativeness and discriminatory ability for grain yield, investigate the similarities among them and assess the variance components. 6 set sets of Advance yield trial with unique entries were analysed using Mixed models, best linear unbiased estimate (BLUPs) and BLUP based GGE biplot. Results showed significant means squares Genotype x Location interaction effect for all 6 sets; this justifies the to study GEI in cowpea, partitioning variance showed that Environment main effect accounted for the largest proportion of the total phenotypic variance, with a range of 58.1 (Adv 4) to 73.9% (Adv 3), Environment was followed by Genotype x Location interaction which explained between 16.1% (Adv 5) to 22.6% (Adv 4). Genotype main effect accounted for the least with a range of 9.5% (Adv 6) to 19.3% (Adv 4). Broad sense-heritability for GY was high in Advance 1-5 (ranging from 0.62-0.70) and medium in advance 6 (0.53). Shika June, Shika August and Minjibir consistently showed high discriminativeness with Minjibir been the most discriminating environment for this study. Ibadan September and Ibadan May on the other hand consistently showed poor discriminating ability while BUK was inconsistent having good discriminative ability in Sets 1 and 3 while it was poor in sets 2, 5 and 6. In terms of representativeness, no environment consistently had desirable results however, Shika August and Shika June were most representative while Minjibir was least representative. On the basis of discriminativeness and representativeness, the six environments were ranked in other of desirability as Shika August> Shika June> Minjibir > Ibadan September > BUK farm > Ibadan May. In terms of similarity among environments, both the BLUP-Based GGE biplot and genotypic correlations indicated that Ibadan May and September were consistently grouped together and highly correlated; indicating that they are similar, while, Shika August and Shika June were found to be unique. Ibadan May was therefore adjudged a redundant location and could be dropped and replaced without any loss of accuracy because it was neither discriminative nor representative in all biplots draw and it consistently fell in the same Mega environment with Ibadan September. We concluded that there is a need to sample more testing sites and validate their fitness for multi environment yield trials using methods applied in this study.