Geoscience Reference
In-Depth Information
Components practically identical to humic-like C3 have been identified in at least three
previous models (
Figure 10.7
). In the Horsens catchment data set, the highest abundances
of C3 occurred in streams (highest: site 13, lowest: site 27), and the lowest at the WTP
site and in the estuary. Previously, Stedmon et al.
2003
sampled the Horsens catchment in
2001 and found that a similar component (component 1) dominated wetland and forested
regions. Another similar component identified by Stedmon et al. (2005b, component 1)
accumulated in Si- and P-limited mesocosms, where it was produced by microbial deg-
radation and degraded by ultraviolet (UV) and visible light. In the PARAFAC model of
Chen et al. (
2010
), a very similar component (component 2) occurred in surface waters of
Florida Bay but not in ground waters. Components almost identical to humic-like C4 have
been identified in at least four previous models (Stedmon et al.,
2003
; Stedmon et al.,
2007
;
Kowalczuk et al.,
2009
; Murphy et al.,
2011
), particularly at sites near terrestrial sources.
In this study of the Horsens catchment, it was abundant in streams and, in contrast to C3,
was also abundant at the WTP site.
10.8.3 PARAFAC PCA Example
In a further example using the Horsens catchment data set, PCA is used to visualize the
relationship between the five PARAFAC components, nutrients, DOC, and absorbance
(
Figure 10.9
). The first principal component (62.4%) correlates positively with fluorescence
component C2 as well as DOC and absorbance, but also correlates positively with all other
variables. This is evidenced by the first loading plot (
Figure 10.9i
) in which all variables
are placed on the right-most side of zero for the first principal component. Consequently, it
appears to be an axis primarily describing variability in quantities of carbon, a phenomenon
that is then apparently the main cause of variability overall. The second principal compo-
nent (16.4%) strongly positively correlates with phosphorus (TDP, DOP) and weakly nega-
tively with DON, and C3, and therefore primarily describes variations in phosphorus. This
is seen from the fact that TDP and DOP are placed high on the positive axis of the second
component and DON and C3 are placed less extreme on the negative axis (
Figure 10.9ii
).
The third principal component (7.6%) correlates positively with nitrogen and negatively
with component C3 (
Figure 10.9iii
, horizontal axis), and is the first principal component to
distinguish C3 from DON. The fourth (6.1%) correlates positively with C5 (
Figure 10.9iii
,
vertical axis).
The most important test of any model, other than to ensure that it is statistically sound,
is to determine whether the results are plausible in the context of our understanding of
organic matter sources and behavior. The PCA in
Figure 10.9
indicates that in the Horsens
catchment, fluorescence and carbon (DOC, abs) are strongly correlated, whereas carbon,
nitrogen, and phosphorus vary largely independently. This fits with the current understand-
ing of the Horsens catchment and decoupling of C, N, and P supply. The dominant source
of organic carbon and CDOM in the catchment is believed to be derived from soils, particu-
larly in the more forested catchments (e.g., R13 in
Figure 10.5A
), whereas the largest source
of P is from agricultural catchments, for example, R10 and R11 (
Figure 10.5B
) (Stedmon
