Proximal Sensing of Maize Nutrient Status Across Diverse Production Systems with Whole-Leaf Hyperspectral Transmittance

Optimization of fertilization is important for enhancing the economic and environmental sustainability of maize production. However, evaluating nutrient status by manual sampling and analysis is time-consuming, laborious, and expensive. The objective of this research was to predict nutrient status of field-grown maize based on hyperspectral data collected by proximal sensing of maize ear leaves. Whole-leaf transmittance in a range from 459 to 943 nm was acquired by a handheld device followed by the measurement of N, K, Mg, Ca, P, S, Fe, Mn, Zn, Cu and B on the same leaves. Models were developed to predict nutrient concentrations based on spectral data using support vector regression (SVR). Coefficient of determination (R2) of leave-one-out cross-validation (CV) was calculated to evaluate the performance of the models across diverse production systems. Considering the nutrients with variations in three (N, Mg, Mn, B) or four levels (K, P) in our studies, models for the proximal sensing of these nutrients were satisfactory [R2 (CV) > 0.30]. We concluded that proximal hyperspectral sensing could provide acceptable predictions on some macronutrients (N, K, Mg, P) and micronutrients (Mn, B). These low-cost and high-throughput predictions of plant nutrient status may contribute to more sustainable maize production.