The increasing availability of hyperspectral oceanographic measurements provides marine scientists the potential to develop algorithms for large-scale remote sensing of phytoplankton biodiversity and dynamics in the ocean. In this study, this approach is examined using unsupervised cluster techniques applied to field data sets of hyperspectral absorption of phytoplankton and remote sensing reflectance collected along the eastern Atlantic Ocean in 2008. Different open ocean environments in terms of phytoplankton biodiversity were successfully classified. Our performance was validated by considering an objective criterion based on the pigment composition of phytoplankton detected at each station. In order to achieve a good discrimination, an assessment of the effect of the spectral range considered in the cluster-based analysis was required and revealed that some optically significant spectral regions play an essential role in mapping oceanic phytoplankton assemblages by remote sensing. © 2011 IEEE.