Characterizing Dissolved Organic Matter Using PARAFAC Modeling of Fluorescence Spectroscopy: A Comparison of Two Models

Abstract

We evaluated whether fitting fluorescence excitation-emission matrices (EEMs) to a previously validated PARAFAC model is an acceptable alternative to building an original model. To do this, we built a 10-component model using 307 EEMs collected from southeast Alaskan soil and streamwater. All 307 EEMs were then fit to the existing model (CM) presented in Cory and McKnight (Environ. Sci. Technol. 2005, 39, 8142-8149). The first approach for evaluating whether the EEMs were fit well to the CM model was an evaluation of the residual EEMs, and we found 22 EEMs were fit poorly by the CM model. Our second measure for verifying whether EEMs were fit well to the CM model was a comparison of correlations between the percent contribution of PARAFAC components and DOM measurements (e.g., dissolved nutrient concentrations), and we found no significant difference Ip > 0.05) between the two models. These results support the approach of fitting EEMs to an existing model when DOM is collected from similar environments, which can potentially reduce some of the problems when building an original PARAFAC model. However, it is important to recognize that some of the sensitivity or ecological interpretative power may be lost when fitting EEMs to an existing model.