CHAPTER 22 . (hemometric Applications to Seawater Analysis
421
plex matrix effect, affecting the analysis of both natural and anthropogenic species.
Many variables must be often simultaneously optimized in order to maximize the extraction yield or to minimise the matrix effect; methods of experimental design able
to treat simultaneously the effect of the factors and of their interactions on the response
are used.
In the optimization of supercritical fluid extraction (SFE) of methyl mercury in
marine samples, the effect of 8 variables is considered: CO2, flow-rate, density, extraction cell temperature, static extraction time, nozzle and trap temperature, amount of
hydrochloric acid and contact time between acid and sample before the extraction
(Cela-Torrijos et al. 1996). A 26-experiment folded Plackett-Burman factorial design
(2 8 x 31 32 + 2 central points) was used. The results suggested that the extraction cell
temperature is the most significant variable. The method was validated with respect
to three reference materials containing different methyl-mercury contents. Recoveries are comparable to the widely accepted extraction procedure, being always greater
than 82% (for concentrations of the order of flg g -1). Precision is slightly better.
Chemometrics was also employed to optimize procedures of auto-injection and of
programs of column temperature in a GC method for the evaluation of neutral lipid
carbon number profiles in marine samples (Yang et al.I996). Natural lipids including
mono-, di-, and triglycerols, sterols, steryl esters, hydrocarbons, wax esters, free fatty
acids and free aliphatic alcohols are important components of marine food webs.
Triacylglycerols, for instance, constitute a major form of energy storage for marine
plants and animals. Sterols regulate membrane functions and act as precursors for
many metabolically active molecules. The composition of neutral lipid compounds in
marine animals can easily change through diet: environmental factors and concentration ratios of neutral lipids subclasses are indices of their physiological conditions
and thus of the health of seawaters. The method was applied to a series of marine
samples including algae, bivalves, polychaetes, fish eggs and fish larvae to evidence
neutral lipid carbon number distributions. A two-level multivariate analysis Latin
square L8 was used to screen the number of the potentially important variables in the
autosampler and in the injection temperature program. L8 (27) means that seven variables were scanned in eight runs each at two levels (0.3 and 0.7 of the full range respectively). In the second experiment the better point of each variable in the previous
run served as the central value. New values of the two-level analyses were selected from
the middle points between the better and worse points in the last run. Finally, the response-surface method was employed to locate the optimum.
A two-level fractional factorial experimental design involving six chromatographic
factors was employed to optimize the ion-chromatographic determination oflow concentrations of nitrate and phosphate in sea water. 2(6-2) + 3 (for the central points) experiments were performed (Dahllof et al. 1997). For the development of an SFE procedure for a series of PAR from spiked marine sediments, four variables (pressure, temperature, extraction fluid volume, methanol concentration) were optimized (Notar and
Leskovsek 1997). The PARs were subdivided in four groups, according to the number
n of the aromatic rings in their structure (n = 2-3, 4, 5, 6). A five-level spherical factorial experimental design was employed: the lower levels of the parameters were set on
the basis of the literature data and the upper according to the extraction instrument
capability. The other three levels of the parameters were set on the basis of the spherical experimental design, when all the experiments chosen lie on the surface of a
421
plex matrix effect, affecting the analysis of both natural and anthropogenic species.
Many variables must be often simultaneously optimized in order to maximize the extraction yield or to minimise the matrix effect; methods of experimental design able
to treat simultaneously the effect of the factors and of their interactions on the response
are used.
In the optimization of supercritical fluid extraction (SFE) of methyl mercury in
marine samples, the effect of 8 variables is considered: CO2, flow-rate, density, extraction cell temperature, static extraction time, nozzle and trap temperature, amount of
hydrochloric acid and contact time between acid and sample before the extraction
(Cela-Torrijos et al. 1996). A 26-experiment folded Plackett-Burman factorial design
(2 8 x 31 32 + 2 central points) was used. The results suggested that the extraction cell
temperature is the most significant variable. The method was validated with respect
to three reference materials containing different methyl-mercury contents. Recoveries are comparable to the widely accepted extraction procedure, being always greater
than 82% (for concentrations of the order of flg g -1). Precision is slightly better.
Chemometrics was also employed to optimize procedures of auto-injection and of
programs of column temperature in a GC method for the evaluation of neutral lipid
carbon number profiles in marine samples (Yang et al.I996). Natural lipids including
mono-, di-, and triglycerols, sterols, steryl esters, hydrocarbons, wax esters, free fatty
acids and free aliphatic alcohols are important components of marine food webs.
Triacylglycerols, for instance, constitute a major form of energy storage for marine
plants and animals. Sterols regulate membrane functions and act as precursors for
many metabolically active molecules. The composition of neutral lipid compounds in
marine animals can easily change through diet: environmental factors and concentration ratios of neutral lipids subclasses are indices of their physiological conditions
and thus of the health of seawaters. The method was applied to a series of marine
samples including algae, bivalves, polychaetes, fish eggs and fish larvae to evidence
neutral lipid carbon number distributions. A two-level multivariate analysis Latin
square L8 was used to screen the number of the potentially important variables in the
autosampler and in the injection temperature program. L8 (27) means that seven variables were scanned in eight runs each at two levels (0.3 and 0.7 of the full range respectively). In the second experiment the better point of each variable in the previous
run served as the central value. New values of the two-level analyses were selected from
the middle points between the better and worse points in the last run. Finally, the response-surface method was employed to locate the optimum.
A two-level fractional factorial experimental design involving six chromatographic
factors was employed to optimize the ion-chromatographic determination oflow concentrations of nitrate and phosphate in sea water. 2(6-2) + 3 (for the central points) experiments were performed (Dahllof et al. 1997). For the development of an SFE procedure for a series of PAR from spiked marine sediments, four variables (pressure, temperature, extraction fluid volume, methanol concentration) were optimized (Notar and
Leskovsek 1997). The PARs were subdivided in four groups, according to the number
n of the aromatic rings in their structure (n = 2-3, 4, 5, 6). A five-level spherical factorial experimental design was employed: the lower levels of the parameters were set on
the basis of the literature data and the upper according to the extraction instrument
capability. The other three levels of the parameters were set on the basis of the spherical experimental design, when all the experiments chosen lie on the surface of a