富春江典型农业小流域土壤Cd的光谱特征分类及含量估算

Spectral characteristics classification and content estimation of soil Cd at a small typical agricultural watershed of Fuchun River

以富春江上游典型农业小流域为研究对象,采集流域内126个农田表层土样测定Cd含量,并借助地物光谱仪测量可见-近红外反射光谱,采用4种常用方法对光谱数据进行预处理,应用主成分分析和模糊聚类分析量化土壤光谱数据的最佳分类数目,运用偏最小二乘回归(PLSR)构建土壤Cd含量的预测模型。结果表明:(1)流域内土壤光谱全波段反射率随土壤Cd含量的增加而减小,不同土壤Cd含量下土壤光谱反射率曲线整体变化趋势相近;(2)流域内采样点依据光谱特征可划分为两种类型,类型1所属样点的土壤光谱反射率高于类型2,但Cd含量平均值低于类型2,且存在明显空间集聚特征;(3)光谱数据经倒数对数处理后建立的PLSR预测模型决定系数达0.60以上,均方根误差在1.25 mg/kg以下,具有较好的预测能力和稳定性,是农业小流域土壤Cd含量估算的有效方法。

A typical small agricultural watershed in the upper reaches of the Fuchun River was selected as study area and totally 126 farmland surface soil samples were collected in this study area.The soil Cd content was measured and the corresponding visible-to-near infared spectrum was detected by spectroradiometer.Four transformation methods were used to pretreat the spectral characteristics.Then,the principal component analysis and the fuzzy cluster analysis were carried out to determine the optimal number of classifications for soil spectral data.The partial least-squares regression(PLSR)method was used to establish a spectral estimation model of soil Cd content.The results showed that:(1)Soil whole spectral reflectance decreased with the increase of soil Cd content in study area,and the change trends of soil spectral curves under different Cd contents were similar.(2)The sampling points could be divided into two clusters according to the spectral characteristics.The soil spectral reflectance of the sampling points in cluster 1 were higher than that of cluster 2,while the mean Cd content were lower.And the sampling points had obvious spatial clustering characteristics.(3)The PLSR prediction model which established with reciprocal logarithm treated to spectral reflectance presented perfect estimation precision and strong stability,with the decision coefficient higher than 0.60 and the root mean square error lower than 1.25 mg/kg.The reciprocal logarithm transform combined with the PLSR provided an effective method for estimating the Cd content in the soils of small agricultural watershed.