基于野外实测光谱的潍北地区土壤全盐量监测研究

Study on Monitoring for Soil Total Salt Based on Field Measured Spectra at the North of Weifang Areas

为提高潍坊北部地区土壤全盐量监测精度,研究利用光谱测量技术,采集80个土壤样本的光谱数据,其中41个土样测定化学成分。对光谱进行一阶微分(FDR)、二阶微分(SDR)、倒数对数变换(Log(1/R)),将三种指标与土壤全盐量建立逐步多元回归模型和主成分回归模型,并分析在不同指标下所建模型的精度,旨在实现研究区土壤全盐量的定量反演。结果表明:利用光谱变换的一阶微分(FDR)、二阶微分(SDR)所建立的逐步多元回归模型和主成分回归模型的建模系数R^2均大于0.80,说明建模精度较高。在进行检测样本精度检验比较后,其中利用一阶微分(FDR)变换建立的主成分回归模型最稳定,检验精度最高,土壤全盐量建模决定系数R^2为0.931,均方根误差RMSE为0.188,其次为一阶微分逐步多元回归模型。

For improving the monitoring precision about soil total salt, this study collected 80 soil samples in the north of Weifang City, analyzed the chemical composition of the 41 soil samples, measured spectral reflectance by the Spectral measurement technology. For the first derivate reflectance（FDR）, second derivate reflectance（SDR）, Log（1/R）, the multiple linear regression and Principal components analysis models were established using three indexes and soil total salt. To achieve the quantitative inversion about soil total salt in the study area, the prediction accuracy of the model was analyzed under different indicators. The results showed that it was more than 0.80 that modeling coefficient（R^2） established by multiple linear regression model and principal component regression model based on the first derivate reflectance（FDR）, second derivate reflectance（SDR） and the modeling precision was higher. After comparing the accuracy test of the tested sample, the first derivate reflectance（FDR） transform was used to establish the principal component regression model： the most stable model, the highest test precision. The determination coefficient R^2 of modeling soil salt was 0.931, RMSE was 0.913, followed by the multiple linear regression model of the first derivate reflectance.