Data-driven understanding of sweet corn yield variability

Sweet corn, consumed both fresh and processed (primarily canned or frozen), is produced the most in the United States. There is a disproportionate amount of literature published on improving sweet corn eating quality and shelf-life than optimizing crop management practices to maximize yield. This dissertation uses empirical and data mining approaches to understand historic trends in sweet corn yields in relation to crop genetics, management practices, and climate change. Chapter 1 presents an empirical study using an era panel of hybrid sweet corn to quantify historic changes in plant density tolerance (PDT) and associations with plant morpho-physiological and ear traits. Our results showed the increase in per-area marketable ear mass at the rate of 0.8 Mt ha-1decade-1 in sweet corn is primarily due to improved PDT since yield per plant has remained unchanged. Crate yield, a fresh market metric, improved for modern hybrids. However, processing sweet corn yield metrics like fresh kernel mass and recovery (amount of kernel mass contributing to the fresh ear mass) showed modest or no improvement over time, respectively. Modern sweet corn hybrids tend to have fewer tillers and lower fresh shoot biomass, potentially allowing the establishment of higher plant density; however, plant architecture alone does not accurately predict PDT of individual hybrids. Chapters 2–4 of this dissertation use an observational dataset procured by Dr. Marty Williams from multiple vegetable processors in the US. This dataset represents 27 years (1992–2018) of field-level commercial sweet corn production across four states (IL, MN, WA, WI) from over 20,000 fields involving approximately 1,500 growers and eight processing plants. Chapter 2 is an exploratory analysis on historic sweet corn dataset to get a comprehensive understanding of nationwide trends in processing sweet corn. Our analyses showed alarming trends for the processing sweet corn industry; results reported a steady decline in hectares harvested and an amplified crisis in the rainfed production systems. Crop management decisions on plant population density and planting dates for processing sweet corn witnessed no changes since the early 1990s. A large disparity in sweet corn hybrid lifespan was found, which can be attributed to differences in environmental yield stability for the processing sweet corn hybrids. Chapter 3 evaluated several machine learning models for sweet corn yield prediction capabilities and subsequently identified the most influential variables for crop yield predictions. The random forest model provided the most accurate yield predictions, and revealed year (time), location (space), and seed source (genetics) as the most influential predictor variables. Season long precipitation and average minimum temperature during anthesis were the two most important weather variables in sweet corn yield predictions. Chapter 4 of this dissertation studied sweet corn yield response to anomalies in growing season air temperature and total precipitation. This study also estimated critical thresholds in non-linear temperature effects on sweet corn yields across diverse environments and quantified yield losses from surpassing the upper temperature threshold during sweet corn anthesis. Our results reported sweet corn yield losses from recent trends towards rising temperatures, even with irrigation. Growing season temperatures exceeding 30℃ were detrimental to crop yield, especially in rainfed environments. Each additional degree day spent above 30℃ during anthesis reduced crop yields by 2% in rainfed production, and by 0.5% in irrigated production.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Crop Sciences

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2022

Author dc:creator

• Dhaliwal, Daljeet Singh

Contributors dc:contributor

• Williams II, Martin M
• Hager, Aaron G
• Leakey, Andrew D.B
• Martin, Nicolas F

• climate change
• machine learning
• plant density tolerance
• random forest

Statement dc:rights

• Copyright 2022 Daljeet Dhaliwal

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/115426

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F115426