Predicting Plant-Available Nitrogen in Wheat: Overcoming the Soil Moisture Challenge

The escalating cost of nitrogen fertilizer, coupled with the challenges posed by Kentucky''s intense and unpredictable rainfall events, necessitates optimizing nitrogen (N) use efficiency (NUE) in wheat production. The state''s topographic variation further exacerbates N loss, making it imperative to understand plant-available N spatially and temporally throughout the growing season for optimal N applications. Optimizing NUE in wheat production holds significant agronomic, economic, and environmental importance for small grain farmers, as wheat is a major nitrogen-consuming crop. Precise nitrogen management can lead to substantial cost savings for farmers and mitigate environmental impacts due to nitrogen runoff. Vis-near-infrared (Vis-NIR) spectroscopy is an established method for instantly detecting various soil characteristics in the laboratory. Despite the development of various prediction models for total nitrogen (N) using machine learning, with accuracy ranges of R² 0.5~0.8, and the increased accuracy up to R² achieved through deep learning, the prediction accuracy for plant-available nutrients in the field remains impractical due to soil moisture. Typically, soil samples are dried for NO3-N analysis in the laboratory, so understanding the effect of soil moisture on prediction performance and improving prediction models is essential to establish a practical field application of the proximal sensing technology. This project''s main objective is to develop a rapid, cost-effective, accurate prediction model for plant-available nitrogen in soil. This will involve understanding and mitigating the impact of soil moisture on prediction accuracy and enhancing the models using more sophisticated techniques. The successful development of this prediction model will enable farmers to manage nitrogen fertilization more effectively, leading to reduced fertilizer costs and minimized environmental impact. An experimental design incorporating various soil moisture content levels and nitrogen standard solutions will be used to develop a comprehensive prediction model. This model will consider the effects of soil moisture and nitrogen levels and different soil texture on prediction performance, ultimately producing a practical, field-ready tool for optimizing nitrogen use efficiency in wheat production.