37 
Oil Spill identification - Round Robin 20045 
Note: Cedre's Fig 5 indicates a difference higher than the 95% 
confidence level for the ratio B(b+c)F/4-Mpy instead of the ratio 
B(a)F/4-Mpy shown and discussed above. The report of Cedre doesn't 
show chromatograms of these peaks, so it is difficult to find a reason 
for this difference. 
NERI used an excessive suite of compound ratios for sample 
comparison: 
From the fragmentograms, an initial suite of 57 diagnostic compound ratios 
(DRs) were calculated (most based on individual peak heights, some on areas 
of whole isomer groups). Both the number of recorded ions and selected 
ratios are higher than and somewhat different from those suggested in the 
CEN Guideline. 
Generally, ratios were derived from measured peak heights and calculated as 
A/B, but in a few cases calculated as A/(B+C). Some of the selected peaks did 
not represent specifically identified compounds but were chosen as being 
well-resolved and of good intensity, while other peaks represented identified 
compounds (primarily biomarkers). 
Asger Hansen indicated that they have used the round robin samples 
not only for the round robin, but also to test and validate their 
instrument (Personal information not mentioned in the NERI report) 
Neri uses a double sector instrument : 
Following the GC-FID screening, all four samples were analyzed by GC-MS 
fingerprinting using a VG 70S-250 double-sector instrument operated in 
single-ion-recording (SIR) mode at a resolution of 2000 and by using a J&W 
30 m x 0.25 mm x 0.25 pm DB-5 capillary column and manual cold on 
column injection with He as carrier gas. 35 ion masses (+ two additional lock- 
masses) were recorded in two separate runs (runtime: 75 min.) for each 
sample. All samples were analyzed in duplicates. 
From the duplicate analysis of each sample a pooled st. dev. for each 
ratio has been calculated. Nine ratios had a RSD poo ,> 10% and were 
excluded. Furthermore, three sesquiterpane (the nC13-nC16 range) 
ratios were affected by weathering and hence excluded as well. For the 
remaining 45 ratios, the absolute difference between a 
spill sample and a source sample was compared with the critical 
difference, CrD as defined in the report. A result of this evaluation is 
shown in the figure below (Fig. 13) 
The results of NERI and Cedre show the limited value of the numerical 
limitations, when calculated from a limited data set. CEDRE has 
analyzed all samples in triplicate and from the triplicate results for each 
sample the 95% confidence level is calculated for each ratio. Fig 12 
shows the variation in limits for one ratio calculated from different 
samples(see e.g the large difference in error bars for the ratio 
B(a)F/4Mpy. NERI worked with a pooled st. dev. to increase the 
dataset, but still a lot of variation is visible. Low RSD's can lead to false 
negative results (e.g. 1.2% for the ratio Ts/Tm in the example given in 
Fig 13, which has led to several differences above the critical 
difference)