Date of Award:

5-2021

Document Type:

Thesis

Degree Name:

Master of Science (MS)

Department:

Environment and Society

Committee Chair(s)

Emily K. Burchfield (Committee Chair), Matt A. Yost (Committee Co-Chair)

Committee

Emily K. Burchfield

Committee

Matt A. Yost

Committee

Niel Allen

Abstract

Over half the land in the US is dedicated to agriculture, with the vast majority of all cropland cultivated in corn, wheat, or soybean. Despite continuing advances in agricultural technologies, and consistent yield growth over the twentieth century, research suggests that environmental change is already impacting agricultural yield and future changes are sure to exacerbate challenges to agricultural production. It follows that the future of US agriculture depends on the evolution of the changing climate, the relationship between crop yields and the environment, on-farm management and adaptations, the ecosystems that support agriculture, the political and economic incentives that shape what farmers grow and how they grow it, and the technology developed to improve yields. This study will focus on two pieces of the aforementioned agricultural puzzle—the relationship between crop yields and the environment, and water use management in irrigated agriculture. This study contributes to current literature by exploring trends in irrigated agriculture at the county-scale, and by examining the efficacy of Random Forest (RF) regression in predicting agricultural yield. Results from the second chapter, where we utilize exploratory mapping and data mining techniques to understand trends in irrigated agriculture in the Western US, are pending approval from the USDA-NASS and are not reported here. Alternatively, we build a practical guide to working with operator-level irrigation survey data. Results from the third chapter suggest that RF predicts US corn yields well and point to the importance of space and time in corn yield prediction, and the highly nonlinear response of corn yield to irrigation, climate, and agricultural diversity covariates. These results demonstrate the predictive capacity of RF regression to model complex corn yield responses to biophysical and landscape conditions and point to the power of building an ensemble of different models, each with their own strengths and weaknesses, to characterize and predict agricultural yield.

Checksum

249ec9596453300e2e13a3d38ac1b3f2

Share

COinS