USE OF PHYSIOLOGICAL MODELS AND HIGH-THROUGHPUT PHENOTYPING TO CHARACTERIZE GENOTYPIC RESPONSES TO ENVIRONMENT IN RICE

Cultivated Asian rice (Oryza sativa L.) has abundant genetic resources and can be cultivated under various environmental conditions. To achieve stable rice production, it is essential to understand how rice genotypes respond to environmental variation. Characterizing genotypic responses, i.e., measuring plant traits and relating the variation to environmental gradient, is an essential first step. However, genotype-by-environment interactions, combined with limited research resources, make this characterization difficult. This dissertation evaluates the use of physiological models and high-throughput phenotying (HTP) as tools for addressing the challenge. In Chapter 2, an optimal stomatal conductance model was used to estimate a stomatal sensitivity parameter, g1, in 32 tropical japonica (32TRJ) rice accessions grown across five field experiments and two water management regimes at the International Rice Research Institute in the Philippines (tropics) and the Dale Bumpers National Rice Research Center in the United States (subtropics). Under flooded conditions, genotypic variation in the parameter g1 was observed among the 32TRJ accessions. Extensive labor required for physiological experiments underscores the need for HTP approaches for plant trait measurements. In Chapter 3, a literature review was conducted to survey how HTP and physiological process-based models (PBMs) can be used together to enhance the mechanistic understanding and prediction of plant growth across scales. Several core PBM variables (Vcmax, Jmax, leaf nitrogen, plant height, leaf angle, leaf area index, biomass and specific leaf area) were found to be reliably predicted while others (e.g., dark respiration, root traits, gas exchange traits, and leaf water potential) showed low to moderate predictability. Two emerging areas for future work in the integration of HTP and PBMs are further discussed. Finally, Chapter 4 tests the use of hyperspectral reflectance imaging data for classifying nitrogen treatments and predicting physiological and growth traits in indica and tropical japonica rice subpopulations at Purdue Universitys Ag Alumni Phenotyping Facility. The hyperspectral data were able to (1) classify and predict treatment levels using support vector machines and (2) predict nitrogen content and C:N ratio using Partial Least Squares Regression followed by RReliefF wavelength selection. In summary, modeling and HTP have the potential to enable the upscaling and standardization of experiments for characterizing rice genotypic responses.