| A field portable, single exposure excitation-emission matrix (EEM) spectrofluorometer was used in conjunction with the multi-way spectral deconvolution method parallel factor analysis (PARAFAC) for simulated environmental monitoring endeavors. Instrumental design, experimental methods and multi-way data analysis techniques were concurrently optimized to enhance selectivity, to broaden the range of sensing applications, to automate the analysis process, and to achieve near-real time PARAFAC results.; Both parts per-trillion detection limits and prediction errors were achieved for mixtures of polycyclic aromatic hydrocarbons (PAHs) in a variety of aqueous matrices. Next, methods were developed for detection of non-fluorescent pesticides and aqueous metals. Four-way, photochemical, spectrofluorometric analyses methods for several naturally non-fluorescent pesticides resulted in relative prediction errors of less than ten percent. The utility of using EEM and PARAFAC to investigate humic-metal interactions was demonstrated by determining conditional salicylic acid-aluminum stability constants and complexation capacities. Finally, several strategies of minimizing PARAFAC analysis time, minimizing instrumental and chemical requirements, and simplifying the PARAFAC algorithm implementation were developed and tested. It was found that decreasing data density by reducing eight-point calibration sets to two or three points significantly decreased computation intensity yet generated adequate predictions. It was also found that excitation and emission spectral resolutions could be optimized to reach a practical compromise between calibration accuracy and analysis speed. A variety of automated model selection methods were developed and evaluated and several of the automated methods proved to be as effective as manual model selection. |
