Harmful algal blooms are now widely recognised as a severe threat to freshwater ecosystems, particularly in semi-fluvial environments created by river damming. Given the high spatial and temporal variability of cyanobacterial blooms, remote sensing is more suitable than conventional field surveys in monitoring blooms. However, the majority of existing algorithms cannot distinguish cyanobacterial blooms from eukaryotic algal blooms by extracting spectral features in the remote-sensing reflectance (R-rs). In this study, in situ R-rs spectra of cyanobacterial and green algal blooms in Lakes Gaoyang, Hanfeng and Changshou of the Three Gorges Reservoir (TGR) in China were recorded. Characteristic spectral indices, namely, the normalised difference peak-valley index and Cyano-Chlorophyta index, were used to develop an algorithm that can effectively distinguish cyanobacterial and green algal blooms. The proposed algorithm was also used to investigate the spatio-temporal dynamics of the two phenotypes of blooms derived from Huan Jing 1 charge-coupled device images. The resulting accuracy of 93.5% demonstrated that remote sensing technology, in conjunction with field observation, could efficiently differentiate bloom-forming species and assess the water quality in the TGR. (C) 2018 Elsevier B.V. All rights reserved.