Improved Crop Yield Estimation by Integrating Surface Reflectance, Terrain Features, and Subsurface Soil Properties Using Machine Learning in a Terraced Agroecosystem
Location
Havener Center, Miner Lounge / Wiese Atrium, 1:30pm-3:30pm
Start Date
4-2-2026 1:30 PM
End Date
4-2-2026 3:30 PM
Presentation Date
April 2, 2026; 1:30pm-3:30pm
Description
Accurate crop yield prediction is critical for sustainable food production amid growing climate variability. This study assesses whether integrating UAV-derived surface reflectance, terrain attributes, and subsurface soil properties improves yield prediction in a terraced agroecosystem in the U.S. Midwest. A machine learning framework was developed using multispectral and RGB imagery, terrain metrics (slope, curvature, topographic position index), volumetric water content (VWC) from ground-penetrating radar, and apparent electrical conductivity (ECa) from electromagnetic induction. Surface reflectance alone produced limited predictive accuracy (R2 = 0.42), while adding terrain attributes significantly improved model accuracy (R2 = 0.86). Subsurface soil properties further enhanced prediction, with ECa providing the strongest improvement (R2 = 0.95). The highest performance was achieved when all datasets were integrated (R2 = 0.96). These results demonstrate that combining surface and subsurface information improves yield prediction and supports data-driven management to enhance agricultural productivity while preserving resources.
Biography
Effat Eskandari is a Ph.D. candidate in Geological Engineering at the Missouri University of Science and Technology and a Kummer Innovation and Entrepreneurship (I&E) Doctoral Fellow. Her research integrates UAV-based remote sensing, near-surface geophysics (GPR and EMI), and machine learning to monitor and predict soil properties at the field scale. Her work focuses on improving precision agriculture by enabling data-driven decision-making for sustainable land and water management in a changing climate.
Meeting Name
2026 - Miners Solving for Tomorrow Research Conference
Department(s)
Geosciences and Geological and Petroleum Engineering
Document Type
Poster
Document Version
Final Version
File Type
event
Language(s)
English
Rights
© 2026 The Authors, All rights reserved
Improved Crop Yield Estimation by Integrating Surface Reflectance, Terrain Features, and Subsurface Soil Properties Using Machine Learning in a Terraced Agroecosystem
Havener Center, Miner Lounge / Wiese Atrium, 1:30pm-3:30pm
Accurate crop yield prediction is critical for sustainable food production amid growing climate variability. This study assesses whether integrating UAV-derived surface reflectance, terrain attributes, and subsurface soil properties improves yield prediction in a terraced agroecosystem in the U.S. Midwest. A machine learning framework was developed using multispectral and RGB imagery, terrain metrics (slope, curvature, topographic position index), volumetric water content (VWC) from ground-penetrating radar, and apparent electrical conductivity (ECa) from electromagnetic induction. Surface reflectance alone produced limited predictive accuracy (R2 = 0.42), while adding terrain attributes significantly improved model accuracy (R2 = 0.86). Subsurface soil properties further enhanced prediction, with ECa providing the strongest improvement (R2 = 0.95). The highest performance was achieved when all datasets were integrated (R2 = 0.96). These results demonstrate that combining surface and subsurface information improves yield prediction and supports data-driven management to enhance agricultural productivity while preserving resources.
Comments
Advisor: Katherine R. Grote, grotekr@mst.edu