Improved Fluorescence Excitation‐Emission Matrix Regional Integration to Quantify Spectra for Fluorescent Dissolved Organic Matter

The purpose of this short communication is to demonstrate the importance of numerical analysis and wavelength increment selection when characterizing fluorescent dissolved organic matter (FDOM) using fluorescence excitation–emission matrix (EEM) regional integration. A variety of water samples, representing a landscape gradient and different types of FDOM, were analyzed for their percentage distribution of five operationally defined FDOM fractions (aromatic protein I, aromatic protein II, fulvic acid–like, soluble microbial byproduct–like, and humic acid–like) using three numerical methods in integrating volume under the surface of the fluorescence EEMs: Riemann summation, composite trapezoidal rule, and composite Simpson's rule. The influence of wavelength increment was also examined for the precision of the percentage distribution of each fraction. Our results show that the FDOM fraction estimated by Riemann summation with a 10- or 5-nm excitation wavelength can cause >40% or >5% errors, respectively, when compared with the best estimated values obtained by averaging results from composite trapezoidal rule and composite Simpson's rule with 1-nm excitation wavelength at the same emission increment. Also, our experiments show that fluorescence matrix regional integration could underestimate the two aromatic protein fractions but could overestimate the soluble microbial byproduct–like and humic acid–like fractions if improper increment and integral methods are used. The error can be reduced if a smaller wavelength increment is used. The smallest increment in a spectrofluorometer and composite Simpson's rule should be used for scanning fluorescence EEMs and calculating the percentage distribution of each FDOM fraction. Alternatively, 5-nm wavelength increments with composite Simpson's rule could be cost effective, and the error of each FDOM fraction commonly falls within 5% compared with those estimated by 1-nm increments.