Study of winter wheat growth and development in Oklahoma using a hierarchical Bayesian approach

Abstract

In this dissertation, we discussed the underlying mechanisms behind growth and yield patterns in Oklahoma wheat and simultaneously introduced a unique approach to data analysis in crop science. First, the relationships between wheat yield, yield components, and weather variables were explored to understand source-sink balance. The analysis was performed using a Bayesian hierarchical model. Bayesian analysis quantifies uncertainties around the parameter values which helps to realize confidence in the results. Environmental factors were found to explain more yield variability than genotypic factors and wheat yield was found to be limited by both source and sink in Oklahoma. Second, wheat growth patterns in Oklahoma were investigated using a repeated measures dataset on leaf area index and biomass using a dynamic ordinary differential equation (ODE) modeling approach within a Bayesian hierarchical framework. Dynamic crop models are often complex with many parameters which limit their scope. In this dissertation, we have proposed a simple dynamic ODE model with few parameters. The proposed dynamic ODE model was also compared to a traditional data analysis method (linear mixed model) for repeated measures data. Results showed that neither model outperformed the other in terms of prediction. However, the dynamic ODE model offered an advantage of biologically meaningful parameters that was not apparent with the linear models. Third, the dynamic ODE model was extended to include a water balance component in order to investigate how changes in soil moisture throughout the growing season impacts wheat yield production. It was established that the water balance component is most important to make yield predictions under water limiting conditions. Further research is required to overcome the limitations of the water balance model.