Defining traits for selection using isopopulations of sorghum

Sorghum is grown in the subtropics in north-eastern Australia, where production is risky due to limited planting opportunities and highly variable rainfall during the crop cycle. To improve grain yields of sorghum there, plant breeders have adopted the empirical approach of selecting directly for grain yield, because traits likely to confer improved adaptation have not been clearly defined. Even for readily-observed traits such as maturity type, no clear selection goal had been identified for this highly variable environment. This paper examined the use of isopopulations as a tool for defining traits for selection in plant breeding programs, and discussed the merits of this approach relative to other alternatives. Phenology of sorghum in north-eastern Australia was used as a case study. Isopopulations differing in maturity were developed from three populations of sorghum, and were grown in five contrasting locations. For stable grain yields, the best time to flowering was consistently about 1200 day-degrees using a base temperature of 7°C (or about 60 days to flowering for midsummer plantings in Central Queensland). This result was in accord with other direct experimental evidence, but contrasted with recent simulations reported in the literature, which suggested a longer crop duration was preferred. Our conclusions were that the crop model failed to extend growth duration in response to water stress, thereby increasing the yield expectation for a later flowering hybrid in a poor season. The simulations also assumed a full profile of soil water at planting in every season, which would have provided a different outcome from simulations in which conditions prior to planting were permitted to impact antecedent soil water. The strengths and weaknesses of experimental and simulation approaches were discussed, before concluding that the comprehensive evaluation of traits may best be accomplished by a combination of approaches: analysis of variety trial data, direct comparisons using isopopulations or isolines, and appropriate use of a validated crop model.