Modeling Flowering Time in a Maize Nested Association Mapping Population.

Simulation modeling provides a mechanism for characterizing plant processes at more mechanistic levels than is possible through direct observation of simple traits such as time of anthesis or plant height. In attempting to identify loci affecting agronomically important quantitative traits, variation in traits with genotype and environment (G x E) complicates analysis. Model parameters that represent fundamental traits of genotypes, such as earliness per se and photoperiod sensitivity, are assumed to be invariant across environments. Such traits should show high heritabilities, and associated quantitative trait loci (QTLs) should have larger effects than QTLs for derivative traits such as time of flowering. This paper describes efforts to estimate cultivar-specific parameters affecting flowering time for approximately 5000 lines from the maize nested association mapping (NAM) population The NAM lines were evaluated at six locations from New York to Puerto Rico and one or two seasons (giving 11 environments). Time of anthesis in the CSM-CERES-Maize model the Decision Support System for Agrotechnology Transfer v4.5 was simulated for different lines using the end of juvenile phase (EJP), photoperiod sensitivity (PPS), and phyllochron interval (PHY) to represent differences among lines. The GenCalc2 tool of DSSAT was used to automate parameter estimation, which required about 100 h of processing on a 2.4GHz processor. Initial results showed that the model explained about 91% of variation, which was similar to independent estimates of heritability for anthesis date. When only EJP and PPS were varied, QTL analyses failed to detect known loci affecting photoperiod sensitivity in maize. A partial explanation concerns assumptions of the CERES model that reduce overall sensitivity to photoperiod. Ongoing analyses attempt to address these issues. The work emphasizes the challenges of calibrating models for large sets of lines and the need for physiological accurate models.