SIMULATION OF SOYBEAN CANOPY NUTRIENT CONTENTS BY HYPERSPECTRAL REMOTE SENSING.

Precision fertilizer management could help reduce farming costs and maintain production sustainability in current cropping systems. Soybean is a major oil crop and to improve temporal and spatial fertilizer application to demand variations, soybean canopy nutrient status was diagnosed by the hyperspectral remote sensing techonology. First, field canopy spectral reflectance was characterized during key developmental stages with three levels of fertilizer treatments in northeastern China. Then, foliar nitrogen (N), phosphorus (P) and potassium (K) contents were quantified and analyzed for correlation with transformed spectral data formats including reciprocal, logarithm and derivatives, red edge parameters and vegetation index. Last, simulation models for soybean canopy nutrient status (total N, P and K) were constructed. The simulation model (y= -19.153x+3.1114) using second derivatives of spectral data at 432 nm was proved to significantly correlate the predicted value with measured total N content (r=-0.7829, p<0.01; RE=0.1713). The first derivative-derived models y=-0.2939x+0.5889(r=- 0.6172, p<0.01; RE=0.2428) at 909 nm and y=-0.4157x+1.874(r=-0.5631, p<0.01; RE=0.1345) at 908 nm produced most accurate prediction for total P and K respectively. Models reported in this work were top selections for the simplicity and practicality in predicting soybean nutrient and growth status. [ABSTRACT FROM AUTHOR]