Dear friends of EEM-PARAFAC of DOM,
I'm wondering if this is possible. Let's say we got five PARAFAC components (C1~C5) and C1 and C2 are so-called protein-like components. C3~C5 are humic-like components. We assume the relative abundances of protein-like and humic-like components in the data set (~100 samples) have negative relationships. Of course if you compare %(C1+C2) and %(C3+C4+C5), you will find a negative relationships between them with r of -1.0, so this analysis is meaningless.
However, what about if you compare %(C1+C2) with each of %C3, %C4, and %C5? In my real data set, the relationships (r and the slopes) are different between the three patterns (although all of them are strongly negative), and I think this has a statistical meaning.
Note that, because Fmax (fluorescence intensity) of the samples differed considerably (as much as three orders), comparing Fmaxs is affected by the differences in the intensities and all of them (Fmax1~Fmax5) have strong positive relationships with each other.
To summarize, is the following OK?
(1) to compare %(C1+C2) with %C3, %C4, and %C5 by a linear correlation analysis.
Best,
Maru
Why not just correlate Fmax? the idea of using relative contributions is generally a poor idea. The relative contribution says you only ..relative contribution off fluorescence. Strictly (and generally) speaking, it is useless in order to judge about the composition of organic matter. What will PARAFAC give you indeed, these are the Fm of each component, which may be compared, for a given component, in a series of samples, thus producing a quantitative evaluation of the concentration changes. and this trend may be compared between the components. Yes, we do such a relative contribution comparison, byt it is yet of a limited value, and may involve some mutual "artifical" correlations...
Hi Mikhail,
Thank you for the answer.
Fmax of the samples largely depends on sample DOC, so all of them (Fmax1~Fmax5) had strong positive relationships with each other in my dataset that encompasses large concentration gradients.
I think I can try Fmax/DOC to reduce co-variability between fluorescence intensities.
Thank you,
Maru
Dear Maru, hi,
if the associations between Fm and DOC are the problem, I am not sure that dividing Fm by DOC will eliminate this problem in comparing between different components (a Fm of each one divided by the same DOC). But yes, I agree that the DOC normalized fluorescence is useful; we use this parameter since it is indeed, as a SUVA analog, helps to examine how the fraction of a fluorescent component in the whole DOC changes in a sample series.
Dear Mikhail,
After re-discussion with a colleague on this matter, I now fully understood what you meant.
Thank you so much for your valuable addition.
Maru
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Physical Phenomena
- Luminescence
- Fluorescence