Solution-Based Fertilizer Detection Using Guided-Mode Resonance Optical Sensors: A Comparative Study with a Commercial Multispectral Sensor
Abstract
Accurate, low-cost monitoring of nitrogen--phosphorus--potassium (NPK) fertilizer concentration in solution is essential for precision agriculture, yet proposed optical platforms vary widely in complexity and performance. This work directly compares two sensing approaches across four commercial fertilizer formulations from 0 to 200 g/L: a custom two-dimensional Ta2O5-coated guided-mode resonance (GMR) sensor in droplet-loading reflection mode, and a commercial multispectral sensor (AMS AS7341) in cuvette-based transmission mode. The commercial sensor demonstrated consistent concentration dependence, showing positive absorbance slopes across all 32 formulation–channel combinations (p < 0.005). A Beer–Lambert matrix model successfully reproduced 960 measured readings (RMSE = 0.041 a.u.), with cross-validation yielding a concentration error of ∼40 g/L over 75–200 g/L. Conversely, the GMR sensor showed peak response in long-wavelength channels, but no calibration slope differed significantly from zero (R2 ≤ 0.09, p > 0.07), with detection limits exceeding the tested range. This was primarily caused by droplet-geometry and recrystallization instabilities (coefficient of variation up to 42%). These findings establish a baseline for low-cost multispectral screening and demonstrate that sample delivery, rather than transducer sensitivity, limits droplet-based GMR optical sensing in precision agriculture.