Geochemical Fingerprinting - Application of Geochemical Tracers to Fluvial Sediment - page 23

by sheet and rill erosion and low in subsurfacematerials eroded by gully and bank ero-
sion (or vice versa). Agricultural pesticides or fertilizers may be useful in separating
agriculturally related soils from other types of land-use/land-cover. The number of
parameters to select also depends on the defined sediment sources because, in gen-
eral, inverse/unmixing models require n number of parameters to discriminate n 1

sediment sources (Mukundan et al. 2012 ). However, it is not uncommon to utilize a
fingerprint containingmore parameters than source areas or types, as described below.
The most common approach at the present time for determining an effective
fingerprint is to analyze the source and river sediments for awide range of constituents
and then select the fingerprinting parameters using a multi-step, empirically based
process (Fig. 2.5 ). The nature of these statistical methods is important as they heavily
influence the reliability of the fingerprinting results (Walling et al. 2013 ). The current
trend is to use a three step process that eliminates parameters that do not meet certain
assumptions inherent in the use of inverse/unmixing models, while identifying the
parameters that most effectively discriminate between sediment from the defined
sediment sources or source types.
The initial step in this three-part process is to eliminate geochemical parameters
from further consideration that do not behave conservatively. Conservative behavior
is often determined using simple range tests that ensure that the range of parameter
values measured within the sampled river sediment(s) fall within the observed range
of values measured for the sampled sediment sources (Billheimer 2001 ; Phillips and
Gregg 2003 ; Fox and Papanicolaou 2008a ; Collins et al. 2012 ; Wilkinson et al. 2013 ).
This requirement often eliminates relatively soluble elements (e.g., Na, Cl, and P),
and those primarily associated with organic matter.
Conservative behavior also requires that there be no enrichment or depletion in
parameter values as a result of physical or chemical processes that modify the source
area sediment during their dispersal (e.g., by hydraulic sorting or grain weathering)
(Mukundan et al. 2012 ). In essence, the question is whether the sedimentological
characteristics of the sediment (e.g., grain size, shape, density, mineralogy) within
the source areas and the river can be directly compared as is assumed. Three dif-
ferent approaches have been used to address the issue. Perhaps the most commonly
used approach is to analyze and focus on a narrowly defined grain size fraction.
This approach is essentially analogous to the transport invariant approach often used
to assess the provenance of sediments within lithified strata as described earlier
(Weltje 2012 ). It must be remembered, however, that the results of such a finger-
printing analysis apply only to that grain size fraction. Determining the source of
the bulk sample (or other size fractions) will require additional analyses, increasing
analytical costs and effort. In addition, the analyses do not provide for an under-
standing of the actual concentrations in the bulk sample which may be required for
other types of environmental assessments (e.g., a pollutant's potential impact on
biota).
An alternative approach is to mathematically manipulate the geochemical data
obtained from the bulk sediment sample using information collected from a sepa-
rate subsample of the analyzed sediment. The most common form of manipulation

+