A fluorescence-excitation-emission-matrix (EEM) method for the determination of algae community composition is developed by alternating trilinear decomposition (ATLD) analysis and nonnegative least squares (NNLS). ATLD model is applied to fluorescence EEMs of algal pigment extracts, and 14 fluorescent components are identified according to the residual sum of squares. Bayesian discriminant analysis is used to test the discriminatory capability of 14 fluorescent components. All composition spectra of the fluorescent components for the 48 phytoplankton species are spectrographically clustered into 116 reference spectra by hierarchical cluster analysis. The fluorometric method is developed by NNLS and applied to algal samples. For 512 samples of singlespecies, the average correctly discriminated rate (CDR) of the five phytoplankton groups is 99.8%. The CDR for the 256 mixture samples is above 93.5% for the dominant algae species and above 71.1% for the subdominant algae species. This method is also tested by the 18 samples from the Yangtze River estuary, the results of the dominant algae groups are in good agreement with HPLC (high performance liquid chromatography)-CHEMTAX.