By PETER H. JENKINS
School
of Applied Sciences,
University of Wolverhampton
Introduction
In May 1992 students from the University of Wolverhampton
carried out water sampling and chemical analysis
at Loch Ness, to consider the logistics and possible
value of such a study.
Loch Ness presents considerable logistic problems because
of the size and nature of the loch and the catchment. Maitland (1981) divided the catchment into
four sub-catchments, based on the major drainage
basins (see Figure
1, 7K map). These sub-catchments can themselves be considered
as diverse, because of the climatic domains, regional
geology and land classification they span.
This short paper presents the results
of the chemical analysis of the water samples taken. A useful statistical method is suggested
to establish a sampling strategy which could be
representative of all the inputs to Loch Ness.
Sample Collection
Samples
were collected from all the input streams, the output
(River Ness) and both the North and South Basins
of the loch, and the sampling sites are detailed
in Figure
2 (9K map). Glass bottles were used to collect the samples,
which were stored in a domestic refrigerator prior
to transport to Wolverhampton for analysis by Induction
Coupled Plasma. A broad scan programme was used for the elements
shown in Figure 3: sodium (Na), magnesium (Mg),
iron (Fe), calcium (Ca), manganese (Mn), zinc (Zn),
nickel (Ni), copper (Cu) and phosphorus (P).
Results
The detailed results presented in Figures
3a, 3b,
3c
(15K tables) consist of nine variables on samples
ultimately obtained from 82 sites; at site numbers
16, 19, 20 and 28 the stream-beds were dry at the
time for collection, and although numbers 68 to
84 were allocated in advance, no suitable sites
for these were actually found.
Vol:105 The Scottish Naturalist: Results of a
Water Chemistry Study of Loch Ness p51
Analysis
In an attempt to elucidate this
complex data set, simultaneous R-mode and Q-mode
factor analysis was applied. R-mode and Q-mode factor analyses have previously
been used in various studies to explore multivariate
relationships within suitable data sets (Davies,
1986). R-mode factor analysis explores the relationship
of the variables, while Q-mode factor analysis attempts
to explore relationships between the samples. The methods are based on the calculation
of eigenvectors which can be plotted on two-dimensional
factor diagrams. The application of simultaneous R-mode and
Q-mode factor analysis extracts a common set of
factors of the variable and sample factor loadings,
which are relative to the same set of factors. Variable and sample factor loadings can therefore
be plotted in the same two-dimensional factor space.
A discussion
of the mathematical procedure involved is given
by Zhou, Chang and Davies (1983), and a worked example
can be found in Walden (1990).
The results of the first application
of simultaneous R-mode and Q-mode factor analysis
are shown on Figures
4 and Figure
5 (3K graphs). These are identical factor space plots, but
have been separated to give a clearer representation
of the results. Figure
4 gives the variable loadings, and Figure
5 gives the sample loadings. The variance of the first two factors is 44.62%,
which implies that the factor solution is not significantly
reducing the dimensionality of the data set.
Figure
4 implies that certain elements show signs
of correlation, eg. Na, Mg and Ca have negative
factor 2 loadings.
Figure
5 cannot be readily interpreted; it does,
however, imply certain groupings of the samples.
In future it is proposed to
divide the original data set into two groups of
variables, based on the results in Figure 4, and
to apply simultaneous R-mode and Q-mode factor analysis
to the two data sub-sets.
The samples will also be divided into two
groups, based on the results in Figure 5: (a) positive
on factor 1 loadings, and (b) negative on factor
1 loadings. Using
the original data, simultaneous R-mode and Q-mode
factor analysis will also be applied to these two
further data sub-sets.
The
results presented here are not yet complete, because
of the author's commitments elsewhere. It is believed, however, that the method proposed
offers considerable potential for catchment-based
studies of water chemistry. To study a catchment as large and diverse
as Loch Ness requires considerable resources which
Vol:105 The Scottish Naturalist: Results of a Water
Chemistry Study of Loch Ness p54
may not be readily available. However, it is considered that an integrated
desk-top study, such as Maitland (1981), and a one-off
statistical analysis such as this, could develop
a viable sampling strategy.
Acknowledgements
The
author would like to thank his colleagues and students
at the University of Wolverhampton for all the work
undertaken, and the Loch Ness and Morar Project
for making this study possible.
References
Davies,
J.C. (1986). Statistics
and Data Analysis in Geology. London: Wiley.
Maitland,
P.S. (1981). Introduction
and catchment analysis. In: The Ecology of Scotland's Largest Lochs: Lomond, Awe, Ness, Morar and
Shiel. (Ed. P.S. Maitland). Monographiae Biologicae, 44: 1-27. The Hague: Junk.
Walden,
J. (1990). The Use of Mineral Magnetic Analysis in the
Study of Glacial Diamicts Ph.D. Thesis, University of Wolverhampton.
Zhou,
D., Chang, T. and Davies, J.C. (1983). Dual extraction of R-mode and Q-mode factor
solutions. Mathematical
Geology, 15: 581-606.
Received May 1993
Mr. Peter H. Jenkins, School of Applied Sciences,
University of Wolverhampton, Wulfruna Street,
WOLVERHAMPTON WV1 1SB,
