Identification Of Marine Oil Spills By Fisher Discriminant Base On The Diagnostic Ratios

With the rapid development of offshore oil exploration,the cases of marine oil spills has risen year by year.A quick and reliable identification method of oil spill has significance for traceability of oil spill.At present,identification of oil spill is mainly qualitative analysis of the spectrum,but less quantitative research is done.The application of chemometrics in oil fingerprinting provides an effective method for the quantitative identification of oil spills.In this study,the n-alkanes and the diagnostic ratios distribution of 29 kinds of oil samples,including 8 kinds of Middle East crude oil,13 kinds of non-Middle East crude oil and 8 kinds of marine fuel were investigated.Weathering experiments were conducted to study the effects of weathering to the n-alkanes and the diagnostic ratios.Fisher discriminant models that distinguish between the Middle East crude oil,the non-Middle East crude oil and the marine fuel were established respectively using the n-alkanes and the diagnostic ratios as modeling variables.The main results in this paper are as follows:(1)The distribution of n-alkanes show that Middle East crude oil is very similar.The n-C10?n-C15 contents are relatively high,and n-alkanes content from n-C13 appeared a unilateral downward trend.Non-Middle East crude oil is mainly concentrated in the n-C14?n-C28,has obvious differences with the Middle East crude oil.The distribution of n-alkanes in marine fuel is quite different,with no obvious regularity.Weathering has a greater impact on the n-alkanes below n-C15,and higher than n-C15 still has a certain resistance to weathering.The five principal components extracted from long-chain n-alkanes of higher than n-C15 have a cumulative contribution rate of 91.8%to the original data,which can effectively explain all the information of the original data.Fisher discriminant model to distinguish Middle East,non-Middle East and marine fuel based on five principal components.The overall accuracy of identification is 82.8%.The model was verified with oil samples after weathering for 30-days,and with an accuracy of 79.3%.(2)The diagnostic ratios show that that n-C17/Pr in individual oil samples is significantly affected by weathering.The rate of LMW/HMW to most of oil is also affected by weathering,but the RSD is less than 10%and can still be used as a variable.The others ratios are basically not affected by weathering.Collinearity analysis showed that there was collinearity between n-C17/Pr and n-C18/Ph,thus eliminating the variable n-C17/Pr.One-way ANOVA showed that the distributions of n-C18/Ph,Pr/Ph,LMW/HMW,CPI and(n-C19 + n-C20)/(n-C19-n-C22)were significantly different and were suitable for discriminant analysis.Based on the five diagnostic ratios,a Fisher discriminant model was established.The total accuracy of discrimination reached 93.1%.Oil samples weathered for 30 days were used to verify,the accuracy rate was still 93.1%,this indicated that the model was also applicable to the identification of weathered oil samples.Therefore,the Fisher discriminant model based on the diagnostic ratio is obviously better than the model based on the principal component analysis of n-alkanes.