A model analysis of yield differences among recombinant inbred lines in barley

Abstract

Crop models can support plant breeding if they can predict differences in performance of different genotypes. In this study, the ability of a crop model to explain yield differences among genotypes in a recombinant inbred line (RIL) population of two-row harley (Hordeum vulgare L.) was explored. Yield and model-input traits of 94 RILs and their parents, 'Prisma' and 'Apex', were measured in field experiments conducted in Wageningen, Netherlands, in 1996 at low and in 1997 at high N levels. The major gene, denso, with the dwarfing allele from Prisma, was segregating in this population. Short denso RILs outyielded tall types in both years, and this yield advantage was stronger in 1997, largely because the tall genotypes lodged. A crop model based on existing routines for biomass production explained only 26 to 38% of the yield variation among genotypes. The model, using input traits measured from the 1997 data, did not accurately predict growth of genotypes in 1996 because some traits varied with plant N status, which the model did not account for. Model analysis in the high-N environment showed that of the seven model-input traits examined, only lodging score, preflowering duration, and fraction of biomass partitioned to spikes had a significant effect on yield. When these three traits were used while fixing others at their across-genotype means, the model explained 65% of yield variation. To allow effective use of crop modeling in breeding, the ability of crop models to explain yield differences among genotypes has to be improved.