基于中红外漫反射光谱法预测黄土碳酸钙含量

Estimation of Calcium Carbonate in Loess by Using Mid-Infrared Diffuse Reflectance Spectroscopy

以黄土高原侵蚀严重的丘陵沟壑区7个小流域126个土壤样品和108个沉积泥沙样品为对象,分析了土壤的基本属性及其中红外漫反射光谱特征,将中红外漫反射光谱技术与化学计量法相结合,采用不同的光谱预处理方法,分别利用偏最小二乘回归法、主成分回归法和支持向量机回归法建立土壤碳酸钙的光谱估算模型,并对模型进行验证。结果显示:土壤光谱变量与碳酸钙含量的关系更接近线性相关,偏最小二乘回归法和主成分回归法要优于支持向量机回归法,其中PLSR(SG-Der1st)模型为最优,对土壤(决定系数R^2=0.948,相对分析误差RPD=4.426)和沉积泥沙(决定系数R^2=0.875,相对分析误差RPD=2.841)均能准确预测其碳酸钙含量。研究结果表明,基于中红外漫反射光谱的碳酸钙含量估算模型能准确定量土壤及沉积泥沙碳酸钙含量,实现了对黄土高原小流域土壤碳酸钙的快速无损测量。

In order to explore the potentiality of predicting calcium carbonate in loess base on mid-infrared diffuse reflectance spectroscopy,a total of 126 soil samples and 108 sediment samples taken from 7 small catchments in hilly-gully regions with severe erosion on the Loess Plateau were selected.Different pretreatment methods were used to establish the calcium carbonate estimation model.The results indicate that the relationship between spectral variables and calcium carbonate content is closer to the linear correlation,the predictions by using PLSR method and PCR method are superior to those by using SVMR method.The PLSR(SG-Der1 st)model is optimal,and can accurately predict the calcium carbonate contents of soils(R^2=0.948,RPD=4.426)and sediments(R^2=0.875,RPD=2.841).In summary,mid-infrared diffuse reflectance spectroscopy can provide a rapid and nondestructive alternative to estimate the calcium carbonate content in loess in the small catchments of the Loess Plateau.