基于诊断比值的Fisher判别法鉴定海上溢油

采用GC-FID检测分析了8种燃料油、8种中东原油、13种非中东原油的正构烷烃分布特征。以具有地球化学意义和较强的抗风化能力的诊断比值n-C₁₇/Pr、n-C₁₈/Ph、Pr/Ph、LMW/HMW、CPI和（n-C₁₉+n-C₂₀）/（n-C₁₉~n-C₂₂）为建模参数，对中东原油、非中东原油和燃料油进行Fisher判别。由于n-C₁₇/Pr和n-C₁₈/Ph间存在共线性，且燃料油的n-C₁₇/Pr受风化影响较大，所以将其不作为建模参数。所建模型的Wilks's lambda分布所对应的P值为0，表明判别模型是有效的，模型判别准确率达到93.1%。对短期风化30d后的油样的判别准确率也能够达到93.1%，说明模型也同样适用于风化溢油的鉴别。

Identification of marine oil spills by fisher discriminant based on the diagnostic ratios

The characteristic of n-alkane distributions of 8 kinds of marine fuel, 8 kinds of Middle East crude oil and 13 kinds of non-Middle East crude oil were analyzed by GC-FID.The diagnostic ratios with the geochemical significance and the strong anti-weathering such as n-C₁₇/Pr, n-C₁₈/Ph, Pr/Ph, LMW/HMW, CPI and (n-C₁₉+n-C₂₀)/(n-C₁₉~n-C₂₂) were used as modeling parameters, Fisher discriminant of these 29 kinds of oils were performed.Since n-C₁₇/Pr and n-C₁₈/Ph were collinear and n-C₁₇/Pr of marine fuel had obviously changes during weathering, n-C₁₇/Pr cannot be used as a modeling parameter.The Wilks's lambda distribution of modeling corresponded to P value of 0, which indicated the discriminant model was valid, and the identification accuracy of model was 93.1%.The identification accuracy of oils after short-term weathering for 30 days could also reach 93.1%, this meant that the established model was also applicable to identification of weathered oil spills.