Electronic Thesis and Dissertation Repository

Thesis Format

Integrated Article


Master of Science




Wang, Jinfei

2nd Supervisor

Leblon, Brigitte


University of New Brunswick



The optimization of crop nitrogen fertilization to accurately predict and match the nitrogen (N) supply to the crop N demand is the subject of intense research due to the environmental and economic impact of N fertilization. Excess N could seep into the water supplies around the field and cause unnecessary spending by farmers. Understanding the detailed spatial information about a crop status is known as a farming management technique called precision agriculture, which allows farmers to maximize their yield and profit while reducing the inputs of fertilizers, pesticides, water, and insecticides.

The goal of this study is to document and test the applicability and feasibility of using Unmanned Aerial Vehicle (UAV) to predict nitrogen weight of wheat and corn fields in south-west Ontario. This is investigated using various statistical modelling techniques to achieve the best accuracy. Machine learning techniques such as Random Forests and Support Vector Regression are used, which provide more robust models than traditional linear regression models. The results demonstrate that most spectral indices have a non-linear relationship with canopy nitrogen weight and show high degree of multicollinearity among the variables. In this thesis, the final nitrogen prediction maps of wheat and corn fields using UAV images and the derived models are provided.

Summary for Lay Audience

The analogy I like to give for the concept of precision agriculture is a person’s desire for a cup of coffee from a coffee machine. Imagine trying to invent a coffee machine that can predict the quantities of coffee that the customers want. The coffee machine should be able to supply you with the right amount of coffee depending on the size of the cup, how tired you are, what time of the day it is, etc. All these factors contribute to how much coffee you need. For example, if you were very tired one day and inserted a large cup, you would be quite disappointed if the coffee machine supplied you with a little amount of coffee. This is how the phenomenon of nitrogen supply works with crops. The goal of precision agriculture is to accurately supply the agricultural crop’s site-specific need, depending on various factors surrounding the crop (e.g. soil, precipitation, temperature, etc). If crops are deficient in nitrogen, their growth cycle is likely to be stunted, reducing its yield potential. On the contrary, if crops are supplied too much nitrogen, the excess supply can seep into the water supply, causing a negative environmental impact and unnecessarily uses up the farmer’s nitrogen resources. Because crops cannot communicate their needs of nitrogen to us, researchers have performed extensive research using remote sensing techniques on various types of crops by estimating how much nitrogen they currently have. If we know how much nitrogen the crops have, we can add or reduce the nitrogen application for a particular area based on the guideline that the farmer has. This thesis dives into the statistical application of drone imagery and regression modelling to predict the quantification of nitrogen status within wheat and corn fields. Ultimately, when we predict the values of the nitrogen on a map, we are then able to supply it to the farmer for their next nitrogen fertilization application.