针对土壤漏油污染的检测、 分析和含量预测等实际问题, 用关中平原常见普通土壤、 柴油、 机油及油土混合物为研究对象, 设计被不同油种污染以及不同污染程度的土壤光谱特征实验。 采用光谱分析的方法来获取光谱指数, 测量并分析渗漏相同含量两个油种的油土混合物以及渗漏不同含量同一油种的油土混合物的光谱特征及其差异, 根据分析结果建立一个基于反射特征的土壤柴油渗漏含量的光谱预测模型。 研究表明: (1)渗漏相同含量不同油种时, 渗漏柴油的土壤的反射率低于渗漏机油的土壤; 渗漏柴油和机油的土壤均在1 740和2 328 nm附近出现双吸收谷特征, 且在每一个吸收谷点处渗漏柴油的土壤的光谱吸收深度和光谱吸收指数均低于渗漏相同含量机油的土壤。 (2)建立的土壤柴油渗漏含量的光谱预测模型相关系数r的值达到0.854, 说明模型具有较强的稳定性; 评价模型预测能力的均方根误差RMSE值为0.016证明该模型具有较好预测能力, 可以对土壤柴油渗漏含量进行有效的预测。

The spectral analysis method was applied experimentally to extract the spectral indices, measure and analyze the spectral characteristics and their difference of the mixture which are composed in soil in Central Shaanxi Plain and the diesel and motor oil respectively, aiming to provide solutions to practical difficulties in detecting, analyzing the spectral characteristics and difference between the soil leaking with equal content diesel and motor oil and predicting the leaking content of diesel in the soil. The spectral response curves of the soil leaking with different oil respectively and the soil leaking with diesel with different content were collected. Then the spectral prediction model for the leaking content of diesel in the soil was built based on the reflectance characteristics. The coefficient of the detection (R2) was introduced to evaluate the stability of the built model, and the parameter root mean squared error (RMSE) was introduced to estimate the precision and the predictability of the model built in this work. It is demonstrated that : (1) The reflectance of soil leaking with diesel is less than that of the equal content motor oil. And there is a double absorption trough of the reflectance curve of both soil leaking with diesel and motor oil at 1 740 and 2 328 nm. The spectral absorption indices and absorption depth of the soil leaking with diesel keep less than the equal content motor oil. (2) The built spectral prediction model for the leaking content of diesel in the soil demonstrates good stability with the coefficient of determination at R2=0.854, and performs favorable predictability (Root-Mean-Square Error, RMSE=0.016), which can benefit the effective prediction and quick estimation methods of the leaking content of diesel in the soil, enrich and progress the experimental method and theoretical research work of spectral prediction on soil leaking oil content and promote the application of remote sensing in safety production and environmental protection.