Nitrogen (N) is one of the most intensively used resources in global corn production, however, it is often the most limiting nutrient for plant growth. This discrepancy is largely due to low N use efficiency caused by poor fertilizer application timing and inadequate estimation of indigenous soil nutrient supply. Precision nutrient management intends to supplement available soil nutrient supply using timely seasonal applications to match crop demand spatially and temporally. For my master’s research, I investigated methods of improving crop N diagnosis and in-season N fertilizer management using proximal and aerial remote sensing systems.
The first portion of my research evaluated the multiparameter Crop Circle Phenom proximal sensing system for in-season diagnosis of corn N status. My research objective was to investigate non-destructive approaches to estimate in-season plant N status using sensor measurements and field management information. The plant N metrics predicted were aboveground biomass, plant N concentration, plant N uptake, and N nutrition index. The study involved collecting in-season Crop Circle Phenom sensor measurements and whole plant samples from a Drainage x Tillage x Nitrogen experiment conducted near Wells, MN. Two site years of sensor and plant measurements were collected (2018 and 2019) at around V8 growth stage and cross-validation modeling was used to construct simple regression and eXtreme gradient boosting machine learning regression models. Based on my modeling results, I propose potential improvements to diagnosing in-season corn N status using proximal remote sensing tools and machine learning modeling.
The second portion of my research developed a practical remote sensing and calibration strip-based nitrogen management strategy for corn. This research intended to build on previous calibration strip methods through replicating the strips at a field-scale and extending the size of each treatment to be the length of the field. Three on-farm trials were conducted in Minnesota during 2019 and 2020 growing seasons. Each on-farm trial consisted of five pre-plant N rates which corresponded to the farmer’s normal rate (FNR), including 0% FNR, 30% FNR, 70% FNR, 100% FNR, and 130% FNR. To manage yearly inputs for each field, GIS software was used to divide the field into approximately 49 m by 24 m grids. These grids were grouped into management blocks consisting of five adjacent grids used to create site-specific response curves. The calibration strips were monitored from time of emergence through application of side-dress fertilizer at approximate corn V8-V10 growth stage using normalized difference vegetation index (NDVI) derived from Quantix UAV Mapper, Ceres Imaging, and PlanetScope satellite remote sensing imagery. To guide side-dress N rates, optimal site-specific N rate (ONR) was predicted for each block by selecting the N rate with the best performing NDVI. To prevent under fertilization the 70% FNR was selected if the lowest pre-plant rates showed no NDVI difference or outperformed the higher N rates. Performance of the calibration strip strategy was assessed by comparing the block specific ONR to agronomically optimal N rate (AONR) and economically optimal N rate (EONR) calculated using yield data. Overall, we found the method accurately estimated the in-season ONR compared with the EONR and AONR metrics calculated from yield data. Furthermore, we observed that variable rate prescriptions of side-dress N dramatically increased nitrogen use efficiency for two of the three field sites, while achieving similar yields and efficiency in the third field site.
In partial fulfillment of the requirements for the MS degree in the Graduate Program in Land and Atmospheric Science
Cadan Cummings, LAAS master's student advised by Prof. Yuxin Miao