Fallout Radionuclides - Application of Geochemical Tracers to Fluvial Sediment - page 59

210 Pb ex and 7 Be to estimate
the relative contribution of sediment from both spatially defined areas (e.g., forests,
grazing areas, cropped land) and source types (e.g., channel and/or gully sediment)
(see, for example, Nagle et al. 2007 ; Olley et al. 2013 ). In this case, the emphasis
generally is on determining the contribution of sediment from the channel or channel
banks in comparison to surface sources. In contrast to the multivariate geochemical
fingerprinting approach discussed in the previous chapter, investigators often use a
single FRN (e.g., Nagle et al. 2007 ; Kwan Kim et al. 2013 ). There are, however,
benefits, as shown later, to utilizing a combination of 137 Cs

It is not uncommon for investigators to use 137 Cs

,

210 Pb ex and 7 Be (Olley
et al. 2013 ; Walling 2013 ). When a single parameter is used as a fingerprint, a
relatively simply mixing model has historically been applied for source ascription
purposes (Zhang and Zhang 1995 ; Wallbrink et al. 1998 ; Zhang et al. 1997 ; Wallbrink
and Murray 1996 ; Wallbrink et al. 1998 ; Brigham et al. 2001 ). Mathematically, it
takes the form of (Wallbrink and Murray 1996 ):

,

C s = (( P r P b )/( P s − P b )) × 100 (3.1)

where C s is the percent relative contribution from a surface source, P s is the value of
the FRN for the surface source, P b is the value of the FRN in the subsurface source,
and P r is the value of the FRN in the river sediment (sediment mixture). The initial
studies used mean FRN values to characterize the source materials. However, FRN
concentrations in surface sediments tend to vary within the catchment as a function
of atmospheric fallout and sediment transfers between sites as a result of erosional
and depositional processes. These variations were initially dealt with by using a
composite sampling scheme where multiple samples were collected and combined
within a defined area to help eliminate field variance. While composite sampling is
still common practice, more recent studies (e.g., Nagle et al. 2007 ) utilize a Monte
Carlo approach similar to that described in the previous chapter to quantitatively
account for model uncertainty. The FRNs for a specific source are characterized
when using this approach by a statistical distribution of observed values within the
materials. Values from these distributions are then selected and entered into themodel
to determine the relative contribution from surface sources. The process is repeated
thousands of times, producing a statistical distribution of source contributions ( C s in
Eq. 3.1 ). This generated distribution can subsequently be analyzed to determine the
median predicted contribution (the 50th percentile) from the surface sources, while
other quartiles can be used to assess the extent of variation in the predicted values.
It is normally assumed that FRN activities follow a normal distribution within
the source and river sediments. Olley et al. ( 2013 ), however, found that within sub-
tropical catchments of Queensland, Australia FRN data within the source and river
sediments did not necessarily follow a normal distribution. In this case, only 210 Pb ex
was normally distributed. Thus, they used a procedure similar to that proposed by
Caitcheon et al. ( 2012 ) to generate a probability distribution for 137 Cs that could be
incorporated into a relatively simple mixing model to estimate sediment provenance.
They found that channel deposits represented the dominant source of sediment to
the river (as opposed to surface sources consisting of cropped land, forests, and