Abstract:

This study tried to characterize comprehensively on soil nutrient status by spectral reflectance, define its main parameters and technical points. We collected 63 soil samples from Loess Plateau, measured its 14 nutrient-related soil parameters and visual band spectral reflectance in laboratory. And then, choose five categories of nutrient parameters which have large correlation coefficient with spectral reflectivity to made predictive models. At last, fit the predicted value and measured value to test its predictive effect. Results indicated that: The correlation coefficient between the First Derivative of spectral reflectivity (R′ value) and soil nutrients parameters is much larger than spectral reflectivity (R value) and spectral reflectivity’s Logarithm of Reciprocal (A value). The correlation coefficients of the predictive models to SOC, CEC, TN, AN, AK are over 0.74. The prediction equation of SOC, CEC are y = -24.19x + 4.049, y = -126.3x + 20.54, respectively. Corresponding correlation coefficients (r) are 0.9282, 0.9273, and best predictive wavelengths are 560nm, 520nm; The r value in inversion test models are 0.7886, 0.5401, while standard deviation are 0.1443, 0.627. Soil visible spectrum maintains the potential of characterizing soil nutrient comprehensively, and multiple factors analysis make the information of soil nutrients reflected fully and accurately.