Optimum replications and locations for cotton cultivar trials under Mediterranean conditions

SUMMARYThe number of replications and locations used in a cultivar evaluation scheme is an important factor affecting the trial heritability (H) and optimum resource allocation. The aim of the present study was to calculate the required number of replications and locations for realizing an optimumHof 0·75 and to identify the most effective test locations for cotton (Gossypium hirsutumL.) in Greece. The data for lint yield, plant height, verticillium wilt (Verticillium dahliaeKleb.) and fibre quality were derived from an 8-year experiment (2000–2007) conducted under irrigated, Mediterranean conditions at 14 locations along the Greek mega-environment. Analysis of variance was performed to calculateHas well as genotype plus genotype × environment (GGE) biplot analysis to determine the location's desirability. It was determined that the four replications currently used in the evaluation of lint yield were sufficient, whereas four locations were proposed as optimum in lieu of the current 8–14 locations used in the evaluation. Two locations excelled as the most effective for lint yield evaluation and one for selection of genotypes tolerant to verticillium wilt using as criteria: the high and consistent across yearsH(0·75), GGE biplot representativeness and discriminating ability. Moreover, one location was selected as a backup based on average trial failures. Plant height was sufficiently evaluated by four replications and two locations, while verticillium required four replications for realizing lowerH(0·60) and three locations for even lowerH(0·40). Regarding quality, an increase of replications from the two currently used to four was sufficient for evaluation of all the traits. The advantage of reducing the number of locations for evaluation of lint yield to just four casts no doubt on the evaluation precision of lint percentage, length, strength and elongation but does for micronaire, short fibre index and uniformity, which realized lowerH(0·60 or 0·50).