基于高光谱的民勤土壤盐分定量分析

土壤盐渍化是重要的生态环境问题，严重影响着干旱、半干旱区的农牧业及经济发展。高光谱遥感技术能够提供地物的连续光谱信息，易于分析细微差别，在定量研究土壤盐分含量方面具有较大优势。民勤县位于甘肃省石羊河流域下游，水力资源匮乏，盐渍化问题十分严峻。本研究基于实验室光谱数据，通过建立模型定量分析土壤盐分含量。首先对原始数据进行连续统去除（cn）预处理，然后分别建立了土壤盐分含量的高光谱指数模型（NDSI）、偏最小二乘回归模型（PLS）、间隔偏最小二乘法模型（iPLS）和反向间隔偏最小二乘法模型（BiPLS），考察各种模型对土壤盐分的预测能力。对比分析发现，使用全部波段信息建模的PLS模型优于仅使用两个波段信息的NDSI模型，而iPLS和BiPLS模型通过选择特征波段进行建模，结果均好于全谱PLS模型。其中，BiPLS模型波段选择的能力优于iPLS模型，得出的模型结果最好，预测相对偏差RPD达到2.02，决定系数R²和模拟值与预测值线性回归的斜率分别为0.76和0.92，模型可以近似地预测土壤盐分含量。结果说明特征波段选择方法能够从大量数据中提取有效信息，简化模型，并获取比NDSI模型和全谱PLS模型更优的预测结果。这些研究对于使用高光谱数据定量分析土壤盐渍化有一定的意义。

Quantitative Analysis of Soil Salinity Content with Hyperspectra Data in Minqin,Gansu, China

Soil salinization is a severe environmental issue, which has already restricted the development in arid and semi-arid regions. Hyperspectra remote sensing has an advantage in the quantitative study of soil salt content (SSC) because it includes continuous spectrum information, which is easily to recognize slight characteristics in different objects. Minqin County is located in the downstream of the Shiyanghe River in Gansu province, where the water resources is shortage and soil salinization is very serious. In this paper, we quantitatively analyzed the SSC by establishing models based on laboratory spectral data. Firstly, the original spectra were transformed by continuum removal (cn) method. Then, normalized difference salinity index (NDSI), partial least square regression (PLS), interval partial least squares (iPLS) and backward interval partial least squares (BiPLS) based on the spectrum were used for modeling, in order to study the prediction ability of different models for SSC. The results showed that the PLS model based on the full spectra was better than NDSI model only based on two spectra data, while iPLS and BiPLS models built by using the spectrum after band selection both were superior to the PLS model. Meanwhile, BiPLS model had better ability of band selection than the iPLS, which was the best model. The RPD is 2.02, R² is 0.76 and slope is 0.92, respectively, in BiPLS, which could make approximate predictions of SSC. These results showed that band selection method was able to remove redundant information, simplify the calibration model and improve the predictive ability. Therefore, these studies were meaningful of quantitative monitoring soil salinization.