Semester of Graduation
Fall 2024
Degree
Master of Science (MS)
Department
School of Plant, Environmental, and Soil Sciences
Document Type
Thesis
Abstract
Sustainable agriculture is essential for addressing global challenges such as food security, environmental degradation, and climate change while ensuring high productivity and resource conservation. Precision agriculture technologies, like UAV-based remote sensing and advanced sensors, offer tools to monitor real-time crop health, soil conditions, and nutrient levels, aiding in informed decision-making and resource optimization. In this research, field trials were conducted in maize and rice to assess the effectiveness of NDRE (Normalized Difference Red Edge) obtained from UAV-based remote sensing for nitrogen management. For maize, the study demonstrated that NDRE accurately assessed nitrogen levels during critical growth stages, with yields ranging from 2901 kg ha-1 in no-tillage with no nitrogen to 15459 kg ha-1 in conventional tillage with 270 kg N ha-1. Strong correlations were observed between NDRE values and maize yield, with R² values of 0.9973 for conventional tillage and 0.9997 for no-tillage. This indicated that nitrogen application is most effective around 60 days after planting. Machine learning models—Random Forest, XGBoost, and Linear Regression—were employed to predict maize yields using NDRE and other indices. Linear Regression proved to be the most accurate, explaining 70% yield variability with an MSE of 49 kg ha-1 and an RMSE of 2166 kg ha-1. Additionally, NDRE obtained via UAVs was compared to active sensor data in rice trials at the LSU AgCenter Rice Research Station. Strong correlations were found (R² above 0.85), with the highest rice yield recorded at 9102 kg ha-1 under 210 kg N ha-1, and the lowest yield at 4950 kg ha-1 with no nitrogen. NDRE readings allowed for optimized nitrogen application in rice and maize, showing that UAV-based monitoring, combined with machine learning models, holds significant potential for large-scale, sustainable nitrogen management, resulting in improved crop productivity and minimized environmental impacts.
Date
11-1-2024
Recommended Citation
Adhikari, Rejina, "DIGITAL AGRICULTURE APPROACHES: UAV-BASED N MONITORING AND MACHINE LEARNING FOR MAIZE YIELD PREDICTION, AND N RESPONSE ASSESSMENT ACROSS RICE VARIETIES" (2024). LSU Master's Theses. 6060.
https://repository.lsu.edu/gradschool_theses/6060
Committee Chair
Setiyono, Tri.