次生盐渍土高光谱特征研究
Spectral characteristics of the secondary saline soil

基于次生盐渍土修复过程中硝酸盐含量和同步实测光谱数据,针对原始光谱数据及其不同变换后7种光谱数据集,分别以相关系数极值和间隔偏最小二乘2种方法分析其最佳敏感波段范围。在此基础上,运用偏最小二乘回归(partial least squares regression,PLSR)方法,分别基于全波段(400~1 650 nm)和分析获得的最佳敏感波段建立了次生盐渍土壤NO-3含量的光谱反演模型。结果表明,采用2种方法提取的土壤最佳敏感波段,均集中在844.5和846.18 nm；基于全波段与最佳敏感波段的土壤NO-3含量光谱反演模型,均以原始反射光谱经一阶微分处理的结果更为显著；其中,基于间隔偏最小二乘法提取的775~899和1 025~1 149 nm为最佳敏感波段的预测模型,其决定系数R2p与标准差(root mean standard error of prediction,RMSEP)分别为0.962和0.057。该研究结果可为今后次生盐渍土中硝酸盐含量的快速损检测提供重要的科学参考。
In this paper,the soil nitrate content(NO-3)and the associated spectral reflectance were measured and analysed during the microbial remediation of secondary saline soil. Extrema of correlation coefficients(ECE)and interval partial least squares regression(iPLS)methods were used to find out the optimal sensitive bands for nitrate content detection based on seven spectral data sets. On this basis,partial least squares regression(PLSR)was used to build relational models between nitrate content and spectral reflectance based on full bands(400-1 650 nm)and optimal sensitive bands,respectively. The results showed that the optimal sensitive spectral bands selected by two methods were concentrated on 844.5 and 846.18 nm. The Savitzky-Golay first derivative(SGFD)obtained more robust calibration and prediction accuracies than other pre-processing inversion models with respect to the PLSR predicting models of nitrate content based on full bands and optimal bands from two different ways. Among all of the seven spectral data sets,the prediction accuracy of SGFD in optimal bands(775-899 and 1 025-1 149 nm)was the best,the corresponding R2p and root mean standard error of prediction(RMSEP)of the predicted model were 0.962 and 0.057,respectively. The results of this study provided a significant scientific reference for the rapid and non-destructive detection of nitrate content in secondary saline soil in the future.