39
Oil Spill identification - Round Robin 20045
used in the figure. The ratios have been chosen such that weathering
causes a reduction of the value of a ratio. This has been found for e.g.
F2/P2 of Extract II. The plot also shows that the ratio P2/D2 of Source
II is above the upper limit for both duplicate analyses.
For the CEN guideline it has been decided to work with a fixed RSD of
5% for all ratio's to overcome the variation in critical differences. To
test this value a method must be validated by analyzing a sample at
least 7 times and to calculate the RSD's of the ratios. RIZA has done
this recently for the peak identification mixture prepared by SINTEF.
The results can be found on the CD in the RIZA map.
The RSD of 5% can be seen as a quality criterion, because methods
producing higher values may not be used to analyze/compare oil
samples. E.g. fast methods with a low number of scans over a peak will
produce higher RSD's and are therefore not acceptable.
A difficulty with this approach is the higher RSD for small peaks.
Therefore it is advised to analyze in each case some of the samples in
duplicate and to compare the ratio's of the duplicates. When small
peaks show differences above the critical limit these peaks should only
be used for a visual comparison. Additional a S/N criterion is added to
be able to decide not to integrate small peaks.
As a result it is possible that labs will use and eliminate different ratios:
SINTEF comes to a full match of source I with spill I and spill II (all
parameters match). But compared to the draft CEN-guidelines, several ratios
are left out (marked as “Not Quantified”). It is indicated here, for example,
that the retene/4-phen ratio is not “considered to be significant” since the
retene peak is very small compared to the area of the 4-phenantrenes and
hence “give a large uncertainty”.
The retene/4-phen ratio was determined by RIZA without any problem. Even
a higher value than in the peak identification mixture produced by SINTEF
was found.
Furthermore, neither RIZA nor MUMM nor BSH had any problem with
oleanane, which according to SINTEF revealed a too low signal to noise ratio
(lower than 10) and also bisnorhopane (28ab) could be determined without
any problem by these labs (SINTEF has also signed bisnorhopane as “Not
Quantified”).
This may seem a problem, but it reflects the difficulty to treat samples
in the same way in different labs with different instrumentation. The
problem can be solved to demand, that eliminated peaks must be used
in the (final) visual comparison. This is done in the guideline.
Also serious is the remark of e.g. Cedre and LASEM (see also 2.3) that
the hopanes are low and difficult to analyze.
LASEM: The calculated ratios with 191 fragment do not permit significant
conclusions, because of their low abundance and the high variability of ratios
values.
Cedre: Biomarkers could be analyzed in regular conditions but they proved to
be present in oils in low abundance. Consequently, Hopanes,
steranes/Diasteranes and Triaromatic Steranes distributions
