褐土参数的高光谱反演研究

High Spectral Inversion of Cinnamon Soil Parameters

土壤反射光谱的研究是土壤学研究的热点之一。以石家庄市饮用水源二级保护区褐土为研究对象,测定了研究区36个土壤样本的3个指标(pH、有机质、全盐量)和其可见光近红外波段(400～900 nm)的光谱特征,对反射光谱进行5种变换(反射率的一阶导数、倒数及其一阶导数、倒数对数及其一阶导数),通过单相关分析,选取最佳变换进行主成分分析,最后利用多元逐步线性回归建立3种定量估算模型,对比找出最优模型,并进行检验。结果表明,研究区400～900 nm波段光谱及其5种变换形式中,土壤反射率倒数对数的一阶导数与pH相关性最好,671～680 nm波段相关系数最大为0.54;反射率倒数一阶导数与有机质的相关性最好,441～450 nm波段相关系数最大为0.32;反射率及其一阶导数与全盐量的相关性最好,891～900 nm和851～860 nm波段相关系数较大,分别为0.60和0.68。土壤pH的反射率模型决定系数最大为0.79,验证精度为0.026,有机质的一阶导数模型决定系数最大为0.78,验证精度为0.350,全盐量模型一阶导数模型最优,决定系数为0.731,验证精度为0.529。

The study on soil reflectance spectral is one of the hot spots in soil science. Selecting the cinnamon soil in the second-level protection zone of drinking water source in Shijiazhuang city as the object of study, pH, the content of organic matter and total salt, and the near infrared （400N900 nm） spectrum features of 36 soil samples were measured. Then 5 kinds of transform （refleetivity derivative, reeiproal and its derivative, reeiproal logarithmic and its derivative） on reflection spec- trum were made. The best transform was selected by single correlation analysis and the principal components analysis was conducted. Finally, three quantitative estimation models were established using the multivariate gradually linear regression and measured. The results showed that the correlation between soil reflectivity reciprocal logarithmic derivative and pH was the best and the correlation coefficient was up to 0.54 at 671-680 nm band. The correlation between reflectivity reciprocal of the first derivative and the organic matter was the best, and the maximum correlation coefficient was 0.32 at 441-450 nm band. The correlations between the reflectance and its first derivative correlation and the total salt content were the best, the maxi- mum correlation coefficients at 891-900nm and 851-860 nm band were 0.60 and 0.68 respectively. The determination coeffi- cient of the soil pH reflectance model was 0.79 and its verify accuracy was 0.026. The determination coefficient of the soil organic matter reflectance derivative model was 0.78 and the verify accuracy was 0.350. The optimal model of total salt con- tent was the reflectance first derivative model, its determination coefficient and verify accuracy were 0.731 and 0.529 respec-tively.