Assessing the composition of phytoplankton populations by fluorescence spectra

A fluorimetric method for differentiation of phytoplankton classes was developed based on GHM multiwavelet. The 3D fluorescence spectra date of 32 phytoplankton species dominant in coastal area of China sea were decomposed into different scale vectors and wavelet vectors by GHM multiwavelet. Then Ca 2 vectors was selected as characteristic spectra by Bayesian discriminant analysis, and Ca 2 reference spectra were obtained via systematic cluster analysis to the fluorescence characteristic spectra, then the phytoplankton composition fluorescence determination technology was developed by Multivariate Linear Regression resolved by Nonnegative Least Squares. The Ca 2 reference spectra were utilized to identify the samples composed of one phytoplankton species: the average correctly identification rates were 97.6% at division and 86.9% at genus level respectively. For simulate mixed samples (the proportion of dominant division were 50%, 75% and 90%), the average correctly identification rates were 95.5%, 95.6% and 96.6%, the average relative content were 48.5%, 71.0% and 81.7% at division level. For the dominant species (the proportion were 60%, 70% and 90%), the average correctly identification rates were 78.0%, 85.5% and 90.0% at genus level. For in situ test, the results complied with microscopic examination at division level, the average relative content ranged from 76.2% to 96.0%, for the three samples with over 80% relative dominance of one phytoplankton species, two were identified correctly at genus level.