Author ORCID Identifier

https://orcid.org/0009-0004-6928-0570

Semester

Fall

Date of Graduation

2025

Document Type

Dissertation

Degree Type

PhD

College

Davis College of Agriculture, Natural Resources and Design

Department

Not Listed

Committee Chair

Matthew Wilson

Committee Member

Domingo Mata-Padrino

Committee Member

Robert Sedgewick

Committee Member

Ibukun Ogunade

Abstract

Beef cattle production remains a cornerstone of global food systems, providing highly bioavailable protein and essential micronutrients that are difficult to secure from plant-based foods alone. Yet this production is also resource-intensive, requiring substantial land, feed, and water inputs while contributing significantly to greenhouse gas emissions. Efficiency of dry matter intake (DMI) and water intake (WI) is central to both productivity and sustainability, but accurate measurement of these traits—particularly in grazing systems—remains a major limitation. National WI standards continue to rely on pen-level data collected in the 1950s, while most existing DMI prediction equations perform poorly when applied beyond the contexts in which they were developed. These shortcomings constrain progress in both management and genetic selection for efficiency. This dissertation addresses these challenges by leveraging the largest individual-animal WI dataset to date and the largest individual-animal grazing DMI dataset to date, paired with high-resolution climate and performance records, to develop machine learning (ML) models capable of capturing the temporal and environmental dependencies of intake. Random forest and long short-term memory (LSTM) neural networks were used to predict DMI and WI across both drylot and grazing systems. Importantly, the iterative LSTM framework demonstrated the capacity to integrate sequential patterns, environmental drivers such as temperature-humidity index, and production system dynamics, outperforming traditional equation-based approaches and providing improved generalization across systems. By grounding predictions in unprecedented, animal-level datasets rather than approximations, this work advances the biological and methodological foundations for modeling intake traits as distinct but interrelated drivers of efficiency, thermoregulation, and nutrient utilization. These tools provide a framework for real-time, animal-specific predictions that can be integrated into precision livestock management. Ultimately, the development of scalable, data-driven models for DMI and WI offers a pathway to reduce feed and water waste, improve selection for intake efficiency, and lower methane emissions per unit of beef, thereby strengthening the long-term sustainability and resilience of animal agriculture.

Download

Share

COinS