基于相似光谱匹配预测土壤有机质和阳离子交换量  

Prediction of soil organic matter and cation exchange capacity based on spectral similarity measuring

作　　者：魏昌龙[1,2] 赵玉国[1] 李德成[1] 张甘霖[1] 邬登巍[1,2] 陈吉科 

机构地区：[1]中国科学院南京土壤研究所土壤与农业可持续发展国家重点实验室,南京210008 [2]中国科学院大学,北京100049

出　　处：《农业工程学报》2014年第1期81-88,共8页Transactions of the Chinese Society of Agricultural Engineering

基　　金：中国科学院战略性先导科技专项(XDA05050500);国家科技基础性工作专项(2008FY110600);科技部国际科技合作项目(2010DBF24140)

摘　　要：土壤可见光-近红外波段光谱（350～2500 nm）包含了大量的土壤属性信息，相同类型的土壤具有相似的光谱曲线特征，但相似光谱曲线是否具有相似的属性含量？探讨此问题可为土壤光谱库的应用提供依据，从而最终服务于快速获取土壤信息技术体系的构建。该研究以安徽宣城为研究区，根据母质、地形特征和土地利用等信息，采集91个典型土壤剖面，共含400个土壤发生层样品，测定了有机质（soil organic matter，SOM）和阳离子交换量（cation exchange capacity，CEC）含量，同时采用VARIAN公司的Cary 5000分光光度计测定了土壤光谱，并将光谱数据变换为反射率（R）、反射率一阶导数（FDR）和吸收度（Log（1/R））3种形式。该文采用光谱角（spectral angle mapper，SAM）、偏最小二乘回归（partial least square regression，PLSR）和SAM-PLSR（spectral angle mapper-partial least square regression，SAM-PLSR）3种方法预测土壤SOM和CEC。SAM方法是通过对测试集104个光谱曲线与参考集的296个光谱曲线进行相似性计算，并以此实现土壤SOM和CEC含量的预测。SAM-PLSR方法以SAM算法下的匹配结果作为建模样本建立PLSR模型和进行预测分析。结果表明，具有相似光谱曲线的土壤具有相似的SOM和CEC含量，SAM算法下相似光谱匹配可直接预测SOM（R2=0.78，RPD=2.17）和CEC（R2=0.82， RPD=2.41）。PLSR方法可很好地预测SOM（R2=0.87，RPD=2.77）和CEC（R2=0.87，RPD=2.59）；相较之下，SAM-PLSR方法不仅可以更加准确预测SOM（R2=0.89，RPD=3.00）和CEC（R2=0.91，RPD=3.06），而且大大减少了建模样本的数量。该研究使可见光-近红外光谱可更加高效地用于土壤属性分析，并为土壤光谱数据库的建设及应用提供技术参考。The potential of visible-near infrared （vis-NIR, 350~2500nm） laboratory spectroscopy for the estimation of soil properties has been previously demonstrated in the literature. Spectroscopy is rapid, inexpensive, and non-destructive. A single spectrum allows for the simultaneous characterization of various soil properties. The question that always arises when two samples are close in spectral space is whether they are close in terms of soil composition. This paper explores three different approaches to improving prediction accuracy. The first, called the SAM Approach, predicts soil properties via similar soil spectra using a spectral angle mapper （SAM）. The second one, called the PLSR Approach, predicts soil properties using partial least-squares regression （PLSR）. The third, called the SAM-PLSR Approach, first uses the SAM to choose similar soil spectra, which are then used as calibration samples for the PLSR. These tests were performed on a collection of 400 soil samples from 91 profiles from the Xuancheng region of the Anhui Province. Spectra data include reflectance （R）, first derivatives of reflectance （FDR）, and the logarithm of the inverse of the reflectance （Log（1/R））. The aims of the work were threefold： （1） to investigate the relationship between soil vis-NIR similarity and soil attribute similarity （soil organic matter （SOM） and cation exchange capacity （CEC）） using a spectral angle mapper （SAM）;（2） to predict soil properties by PLSR with different calibration samples, which were independently validated;（3） to compare the accuracy of predictions from the SAM Approach, PLSR Approach, and SAM-PLSR Approach. This study showed that soil vis-NIR similarity reflected the similarity of SOM and CEC content, the SAM Approach can be directly used to predict the content of SOM （R2=0.78, RPD=2.17） and CEC （R2=0.82, RPD=2.41）. The PLSR Approach obtained good prediction accuracy of SOM （R2=0.87, RPD=2.77） and CEC （R2=0.87, RPD=2.59）. The SAM-PLS

关 键 词：土壤 光谱 模型 相似光谱 光谱角 偏最小二乘回归 

分 类 号：S151.95[农业科学—土壤学]