不同质地盐渍化土壤水盐含量的高光谱反演

为了方便快捷地同步监测盐渍化土壤的水、盐含量,本文以新疆典型盐渍化灌区为研究对象,基于高光谱技术、运用便携式光谱仪获取不同质地的土壤水盐含量光谱曲线,采用一阶微分、二阶微分、连续统去除的数据处理方法对土壤原始光谱进行变换.结果表明: 对原始光谱数据的变换有利于土壤属性指纹波段的提取,不同质地水盐含量的变换方法并不相同,在壤土中质量含水量为0%和10%时的水盐光谱曲线使用连续统去除方法、15%含水量使用一阶微分、19%含水量使用二阶微分,砂土中0%含水量使用连续统去除方法、10%、15%和19%含水量水盐光谱曲线使用二阶微分处理后,有利于特征波段的提取;对筛选出的变换数据采用偏最小二乘回归方法构建水盐反演模型,壤土盐度小于6.38 mS·cm^-1、砂土小于5.94 mS·cm^-1时,模型建立的建模数据集决定系数(R_cal²)、内部交叉验证(R_cv²)和外部检验数据集决定系数(R_val²)均大于0.65(P<0.05);壤土水分含量小于16%、砂土小于12%时模型反演精度较高.研究结果可为盐渍化土壤水盐含量同步监测提供阈值参考.

Hyperspectral inversion of soil water and salt content in soils with different textures

In order to monitor soil water and salt content of saline soil conveniently and quickly, this paper took the typical salinization irrigation district of Xinjiang as the research object, obtained the spectral curve of soil water and salt content by using portable spectrometers based on the hyperspectral technology, transformed the original spectra of soil using the first order differential, second order differential and continuum removal methods. The results showed that the transformation of the original spectral data was beneficial to fingerprint band extraction of soil properties, and the method was not same in soils with different textures. In loam soil, continuum removal analysis was the best method for extraction of characteristic bands when the soil water content was 0% and 10%, first order differential equations were the best method when the soil water content was 15%, and second order differential equations were the best method when the soil water content was 19%. In sandy soil, continuum removal analysis was the best method for extraction of characteristic bands when the soil water content was 0%, whereas second order differential equations were the best method when soil water content was 10%, 15% or 19%. The transformed data were screened for inversion models of soil water and salt content by using the partial least squares regression method. When thesalinity was < 6.38 mS·cm^-1 in loam soil and < 5.94 mS·cm^-1 in sandy soil, the decision coefficients (R_cal²), internal cross validation (R_cv²), and external validation (R_val²) were greater than 0.65 (P<0.05). When the soil moisture content was less than 16% in loam soil and 12% in sandy soil, the inversion accuracy of model was higher. The results would provide a reference threshold for si-multaneously monitoring soil water and salt content in salinized areas.