Abstract

Most identification methods for marine oil spills rely on single-dimensional chemical fingerprinting that is ineffective for weathered oil. A quantitative identification method is proposed in this study based upon multi-dimensional chemical fingerprinting composed of diagnostic ratios, wavelet coefficient, and the Ni/V. Twenty oil samples were used including 6 marine fuels, 7 Middle Eastern crude oils, and 7 non-Middle East crude oils. Diagnostic ratios were discussed using partial least square analysis, and fluorescence spectra were analyzed by db7 wavelet basis of 6-layers in the discrete wavelet transform. The first principal component and the wavelet coefficient at 354 ± 2 nm were combined with Ni/V and optimized by an exhaustive method as variables to establish the Fisher discriminant model. The developed model had 100% accuracy for modeled oil samples and for non-modeled crude oil before and after weathering, respectively, with 83.3% accuracy for non-modeled or short-term weathered marine fuel. The accuracy was 3% higher than reported in the literature.

摘要

大多数海洋溢油识别方法依赖于对风化油无效的一维化学指纹。本研究提出了一种基于由诊断比、小波系数和Ni/V组成的多维化学指纹图谱的定量识别方法。使用了 20 种油样，包括 6 种船用燃料、7 种中东原油和 7 种非中东原油。诊断比采用偏最小二乘法进行讨论，荧光光谱采用离散小波变换中的6层db7小波基进行分析。将第一主成分和354±2nm处的小波系数与Ni/V相结合，通过穷举法作为变量进行优化，建立Fisher判别模型。所开发的模型对于风化前后的模拟油样和非模拟原油的准确度分别为 100%，对于非模拟或短期风化船用燃料的准确度分别为 83.3%。准确度比文献报道的高3%。