Rethinking crop nutrition diagnosis models: methods, inference and practical applications in crop production and breeding

Abstract

Fertilizer management is one of the most important aspects of agronomic management, affecting the system's sustainability, and relies heavily on the statistical models for nutritional diagnoses. While under-fertilizing may penalize crop yields, excessive fertilization is linked to negative environmental externalities. It is thus desirable that the fertilizer rates approximately match the crop nutrient requirements. Because crop nutrient requirements are usually associated to crop growth, it is convenient to monitor the nutritional status to determine the amount of nutrient required according to crop growth. With the increase in computational power of mainstream computers, many popular crop nutrition models are undergoing changes that leverage such advances in technology. This dissertation is organized in six chapters (Chapter 1, Introduction, and Chapter 6, Final remarks) that revise, rethink, and expand some of the most popular models applied to crop nutrition management. Chapters 2-3 are oriented to the statistical inference, and portray methods developments that may help improve inference from crop nutrition models. Chapter 2 portrays some advantages enabled by modern statistical computing tools by comparing a standard statistical framework introduced in the 1990s, versus a modern statistical framework introduced in 2020. Chapter 3 follows up on the findings of Chapter 2 and elaborates the model and establishes prospects for statistical modeling of crop nutrition models with current statistical tools. Chapters 4-5 are oriented to the practical application of crop nutrition models and the integration of modern measuring hardware, and portray methods for applications in phenotyping and plant breeding settings. Chapter 4 compares different metrics for quantifying crop nutritional status for breeding applications in wheat (\textit{Triticum aestivum} L.). Chapter 5 identifies avenues for research for further developing methods and measuring devices for crop nutritional status phenotyping. While most of the crop nutrition problems presented in this dissertation consider nitrogen management, these finding are relevant for other nutrients as well.