Abstract
Ecophysiological crop models encode intra-species behaviors using constant parameters that are presumed to summarize genotypic properties. Accurate estimation of these parameters is crucial because much recent work has sought to link them to genotypes. The original goal of this study was to fit the anthesis date component of the CERES-Maize model to 5266 genetic lines grown at 11 site-years and genetically map the resulting parameter estimates. Although the resulting estimates had high predictive quality, numerous artifacts emerged during estimation. The first arose in situations where the model was unable to express the observed data for many lines, which ended up sharing the same parameter value. In the second (2254 lines), the model reproduced the data but there were often many parameter sets that did so equally well (equifinality). These artifacts made genetic mapping impossible, thus, revealing cautionary insights regarding a major current paradigm for linking process based models to genetics.
Highlights
CSM-CERES-Maize v. 4.5 was used to fit the anthesis date parameter for 5266 genetic lines grown at 11 site-years.
Despite the high predictive value of the model outputs, numerous artifacts emerged in the estimation process.
The model was unable to express the observed variation in anthesis date data for many lines.
More than one parameter set (equifinality) were found for 2254 lines that equally reproduce the data.
These results revealed cautionary insights regarding a major current paradigm for linking process based models to genetics.
