IDENTIFYING PHYTOPLANKTON IN SEAWATER BASED ON DISCRETE EXCITATION-EMISSION FLUORESCENCE SPECTRA

Abstract

The feasibility of utilizing discrete excitation-emission spectra (DEEMs) to identify dominant groups of phytoplankton at both the genus and division levels was investigated. First, the characteristics of in vivo DEEMs were extracted using Coif2 wavelet. Second, optimal characteristic spectra of scale vectors (SOCS) and time-series vectors (TOCS) were selected by Fisher linear discriminant analysis (FLDA). Third, the SOCS and TOCS were sorted using hierarchical cluster analysis (HCA), and a two-rank database was established according to their discrimination ability. Fourth, the discrimination of phytoplankton was established by nonnegative least squares (NNLS). For single-species samples, the correct identification ratios (CIRs) were 62.9%–100% at the genus level and 95.1%–100% at the division level. The dominant species in the mixtures had corresponding CIRs of 87.5% and 97.9%, and 23 dominant species were correctly identified. Prorocentrum donghaiense D. Lu, Thalassiosira nordenskioeldi Cleve, Chaetoceros socialis Lauder (bloom-forming species with a density of about 107 cell·L−1), and Skeletonema costatum (Grev.) Cleve (a dominant species with a density of 104–106 cell·L−1 in seawater) were identified at the genus level. Other dominant species in seawater were identified at the division level if their density was 105–106 cell·L−1.