Site-Specific In-Season Nitrogen Management Using Drone Multispectral Imagery And Proximal Sensing

Abstract

The main objective of the dissertation work was to assess the ability of in-season remote sensing-based variable rate nitrogen (VRN) to detect nitrogen stress in rainfed corn in Minnesota and its ability to provide timely information to correct corn N deficiency and therefore improve both the agronomic and environmental efficiency of nitrogen fertilizer. While the research focused mainly on UAV-based multispectral sensing, we were able to use proximal sensing and hyperspectral sensing in a section of this study. This dissertation is organized in five independent chapters as follows:In the first chapter, we cover the research that took place in Waseca, MN, where during two growing seasons, in-season VRN based on calibration strips was compared to uniform N rate at a field scale. The study focused on both agronomic and environmental aspects of the comparison and demonstrated the ability of VRN technology to improve nitrogen fertilizer efficiency and reduce nitrate leaching without impacting the yield. The second chapter summarizes the agronomic findings of a four site-year study that took place in south central and southeastern MN. The study compared different nitrogen management strategies for their efficiency at producing corn grain. Vegetation indices from UAV-based multispectral sensing were compared to indices from other platforms and they were found to be comparable. The third chapter is a continuation of the second chapter but focuses on the VRN aspect of the four site-year study. The chapter addresses spatial autocorrelation and compares different algorithms for VRN, including the spectral vs the economic optimum approaches, the sensing time, the spatial scale, and lastly vegetation indices. The fourth chapter is an introduction to the four site-year study and represents a published paper from the proceedings of the International Conference of Precision Agriculture that took place in St Louis, Missouri in 2016. The fifth and last chapter is a case study comparing machine learning using hyperspectral data to multispectral vegetation indices for assessing corn nitrogen deficiency levels. This is also a published paper in the proceedings of the 14th International Conference on Precision Agriculture in Montreal Canada 2018.