| 摘要: |
| 【目的】采用近红外光谱和不同建模方法测定土壤中的有机质和速效P含量。【方法】分别采集江西不同地区的土样240个，采集土壤样品的近红外漫反射光谱，以对光谱数据进行主成分分析得到的前6个主成分（PCs）和偏最小二乘回归(PLSR)建模得到的6个潜在变量 (LVs)，分别作为反向传播神经网络(BPNN)和偏最小支持向量机（LS-SVM）的输入变量，共建立6个模型，分别为主成分回归(PCR)、PLSR、BPNN-PCs、BPNN-LVs、LS-SVM-PCs和LS-VM-LVs，并对这些建模方法预测土壤有机质和速效P含量的结果进行评价，从中筛选出最佳模型。【结果】在预测土壤有机质和速效P含量时，LS-SVM-LVs模型的预测效果优于PCR、PLSR、BPNN-PCs、BPNN-LVs和LS-SVM-PCs模型。用LS-SVM-LVs模型得到的有机质、速效P预测集的决定系数（R2）和均方差（RMSE）分别为0.873 4,0.780 1 mg/kg和2.92 g/kg,4.97 mg/kg。【结论】将近红外漫反射光谱和LS-SVM、PLSR相结合可用于测定土壤有机质和速效P含量。 |
| 关键词: 建模方法 近红外漫反射光谱 土壤 有机质 速效P |
| DOI： |
| 分类号: |
| 基金项目:江西省科技支撑项目(2009AE01603，2010EHB02000);铁路环境振动与噪声教育部工程研究中心资助项目 |
|
| Measurement of soil organic matter and phosphorus using different calibration methods based on near infrared spectroscopy |
|
|
| Abstract: |
| 【Objective】Different calibration methods based on near infrared spectroscopy were used to measure the contents of soil organic matter (OM) and available phosphorus (P).【Method】240 soil samples from the different regions of Jiangxi were collected.Near infrared diffuse reflectance spectroscopy data of soil samples were used for principal component analysis (PCA) to get the first six principal components (PCs),and PLSR model was built to get six latent variables (LVs).PCR,PLSR,BPNN-PCs,BPNN-LVs,LS-SVM-PCs and LS-SVM-LVs modeling methods were built to predict the contents of soil OM and available P.These modeling methods were evaluated and the best model was selected.【Result】The results showed that LS-SVM-LVs model outperformed PCR,PLSR,BPNN-PCs,BPNN-LVs and LS-SVM-PCs models.The best predictions were obtained with LS-SVM-LVs model for OM (R2=0.873 4 and RMSE=2.92 g/kg) and P (R2= 0.780 1 and RMSE=4.97 mg/kg).【Conclusion】The near-infrared diffuse reflectance spectroscopy based LS-SVM combined with PLSR can be used for the measurement of soil OM and available P. |
| Key words: modeling method near infrared diffuse reflectance spectroscopy soil organic matter available P |
