Abstract Quantitative measurements of phytoplankton community composition (PCC) are essential for understanding fisheries production, ocean nutrient cycling, and the export of particulate carbon to the ocean interior. However, these measurements are constrained in dynamic coastal waters due to the spatial‐temporal constraints of in situ sampling, difficulty quantifying communities, and the challenges of deriving community compositions via satellites. Here, we work to address these issues by using highly resolved in situ hyperspectral radiometry, along a ship of opportunity track through Case‐2 waters of the Strait of Georgia (SoG) British Columbia, to derive phytoplankton community composition. First, an empirical orthogonal function (EOF)‐based algorithm was developed using HPLC CHEMTAX‐derived phytoplankton group‐level chlorophyll‐a (Chla) and Total Chla (TChla) concentrations and corresponding principal components derived from hyperspectral remote sensing reflectance. Second, the outputs were evaluated using cross‐validation, showing good retrievals for TChla and the regionally dominant phytoplankton groups: diatoms, cryptophytes, green algae, and raphidophytes, which followed expected spatial‐temporal trends with diatom‐dominated spring blooms and succession to high diversity flagellate‐dominated summer conditions. Furthermore, the outputs captured fine spatial scale trends including strong harmful raphidophyte blooms over the narrow transition to low salinity Fraser River plume influenced waters. These findings highlight the potential of using highly resolved hyperspectral radiometry to derive fine‐scale trends in phytoplankton group level community composition in optically dynamic coastal waters. Coupled with additional measures, this method could provide valuable information on phytoplankton dynamics in the SoG, which is a critical habitat for a high diversity of pelagic fish species, including Pacific salmon. Plain Language Summary Deriving phytoplankton community composition is crucial for understanding their distinct roles in biogeochemical cycles, fisheries production, and carbon export to the ocean interior. In this study, we use continuous hyperspectral radiometric measurements alongside in situ phytoplankton data collected from a passenger ferry operating across the dynamic Strait of Georgia to develop an algorithm for retrieving information on different phytoplankton groups. Our hyperspectral algorithm outperforms previous satellite‐based observations trained on the same data set, demonstrating improved retrieval accuracy for all phytoplankton groups. Specifically, the algorithm captured the diatom‐dominated spring bloom, which was primarily observed in the high‐salinity western portion of the transect, followed by a community shift toward smaller phytoplankton species by summer. Additionally, the algorithm successfully retrieved the concentration of toxic raphidophyte blooms, which were predominantly located near the Fraser River plume waters. Overall, our findings highlight the potential of algorithms trained with continuous radiometric measurements to resolve fine‐scale phytoplankton dynamics that would not be captured by conventional ship‐based sampling. Key Points A hyperspectral‐based algorithm was tested to quantitatively retrieve phytoplankton groups in a turbid estuary Algorithm trained with hyperspectral data outperformed multispectral data in retrieving all phytoplankton groups Diatom blooms occurred in the high‐salinity western region, while raphidophytes were concentrated near the Fraser River plume