赣南脐橙果园土壤全氮和有机质近红外漫反射光谱检测

选取赣南脐橙果园土壤作为研究对象，探讨在4 000～7 500 cm-1范围内的光谱分析土壤全氮和有机质的可行性。采集的近红外光谱采用多元散射校正、一阶微分、二阶微分、七点平滑等多种预处理对比分析，分别建立了有机质和全氮含量偏最小二乘模型。实验得出全氮预测模型在4 000～7 500 cm-1范围内采用七点平滑(SG)进行预处理模型较为理想，校正集相关系数(r_c)为0.802，校正均方根误差(RMSEC)为2.754，预测集相关系数(r_p)为0.715，预测均方根误差(RMSEP)为3.077;有机质预测模型在4 000～7 500 cm-1范围内采用标准正态变量变换(SNV)预处理模型较为理想，r_c为0.848，RMSEC为0.128，r_p为0.790，RMSEP为0.152。研究表明近红外漫反射光谱可快速用于赣南脐橙果园的土壤中全氮和有机质含量的快速检测。

Estimation of the TN and SOM Contents in Soil from GAN NAN Navel Orange Plant Area by NIR Diffuse Spectroscopy

The soil sampled from GAN NAN navel orange plant area was selected as research object, and the feasibility of analyzing the total nitrogen (TN) and soil organic matter (SOM) of soil was investigated by near infrared spectroscopy(NIR)techniques in the wavelength range of 4 000～7 500 cm-1. Different pretreatment methods including multiplicative scatter correction(MSC), first derivative(1st D), second derivative(2nd D), Savitzkv-Golay(SG), standard normalized variate(SNV)and baseline were used. The partial least square regress (PLS) was built for the calibration models. The best TN model using SG pretreatment features the prediction correlation coefficients(r_c)of 0.802，the root mean square error of calibration (RMSEC) of 2.754, the calibration correlation coefficients(r_p)of 0.715，and the root mean square error of prediction(RMSEP)of 3.077 in the wavelength range of 4 000～7 500 cm-1. The best SOM model using SNV pretreatment has r_c of 0.848，RMSEC of 0.128，r_p of 0.790，and RMSEP of 0.152. The results showed that the NIR diffuse reflectance can be used for quick estimate of the TN and SOM contents in soil with the wavelength range of 4 000～7 500 cm-1.