Characterizing dissolved organic matter fluorescence with parallel factor analysis: a tutorial

TutorialCharacterizing DOM fluorescence using the DOMFluor Toolbox for MATLAB ® In this section a brief tutorial in applying PARAFAC to DOM fluorescence EEMs will be given with recommendations on a suitable series of stages to follow in order to facilitate the process based on the authors' own experience.The tutorial is based on MATLAB software and requires that this is installed on your computer.The DOMFluor toolbox for MATLAB provided is in the compressed file "DOMFluorv1_4.zip"(Appendix 1) and works independent of your computer's operating system.This is also freely downloadable from the Chemometrics site at University of Copenhagen (www.models.life.ku.dk).The toolbox contains within it the N-Way toolbox v.3.1 (Anderson and Bro, 2000) and additional functions which facilitate running PARAFAC on DOM fluorescence data.The DOMFluor toolbox is written and tested in MATLAB release 2007b (v7.5) and is free software that can be used and modified under the terms of the GNU General Public License.Within the toolbox a test dataset is provided for the tutorial, consisting of 65 EEMs which have been spectrally corrected for instrument biases, corrected for inner filter effects and Raman calibrated.As a result, all of the Rayleigh scatter and most of the Raman scatter has been removed.The remaining scatter will be removed in the pre-processing phase.The data consists of samples collected from a cruise on RV Gunnar Thorson in the Kattegat and Belt Sea region (at the entrance to the Baltic Sea) in August 2001.Three "outlier" samples were created and are present in the data for the benefit of the tutorial.The tutorial is composed of phases; i) data pre-processing, ii) initial explorative data analysis, iii) modeling and model validation, iv) interpreting results.The more thoroughly the first two phases are carried out, the easier the latter two will be.Part of the pre-processing on the example data set has already been carried out (importing the data to MATLAB, spectral corrections, calibrations, etc.).This is very instrument specific and therefore not appropriate for this tutorial, which focus on the modeling of the data.