Biogeochemistry Technical

Introduction

This page is the technical anchor point for the Dissolved Organic Matter BGC project. The corresponding general interest page is Biogeochemistry.

Arctic GRO data (6 rivers, 5 times over the year 2009, plus one extra sample) provides 31 Samples in one Dataset. These are analyzed using:

Metadata: The physical or 'simple' observables for a given sample (pH for example)
- This includes labor-intensive analysis e.g. to produce quantitative estimates of important molecular classes such as lignin
Fourier Transfrom Infrared spectrometry, abbreviated FTIR (pending)
Absorbance, abbreviated ABS
Proton Nuclear Magnetic Resonance abbreviated NMR
Negative-mode Fourier Transfrom Ion Cyclotron Resonance Mass Spectrometry, abbrevi FTICR-MS or MS
Fluorometry, producing Excitation Emission Matrices, therefore abbreviated EEM
Principle Component Analysis of the above, abbreviated PCA
- kernel-PCA also used to accommodate non-linear nature of this data
2D PCA analog "PARAFAC" applied to EEMs
Covariance: Auto-covariance of one data type with itself, or hetero-covariance of one data type with another; see Super Spectrum below
Clustering: Machine learning algorithms applied to data types
Co-Clustering: Machine learnign algorithms applied to pairs of data types
- Initially variants on spectral co-clustering, which produce conservative sets from two source sets