Phytoplankton size classes (PSCs) are among the most important metrics of phytoplankton community structure and determinants of aquatic food web functionality. Advanced satellite algorithms can be used to estimate PSCs and to integrate the results of biogeochemical research and theoretical studies. Analysis of an extensive dataset of pigment concentrations measured via high-performance liquid chromatography in the southern Yellow Sea, East China Sea, and South China Sea revealed that nanophytoplankton made important contributions to phytoplankton biomass and that sea surface temperature (SST) was the key environmental factor associated with variations of PSCs. A regional SST-dependent, abundance-based model was then tuned to estimate the PSCs and was found to be more effective than the SST-independent model in capturing the temporal trends in PSCs. The implication is that dynamic PSCs variations may be underestimated by using SST-independent, abundance-based model parameters in marginal seas. Moreover, the logistic “S-shape” dependence of PSCs model parameters on SST suggested that global warming may affect PSCs differently on continental shelves in marginal seas or other middle-latitude regions. Application of this PSCs model to satellite data from 2002 to 2017 revealed the temporal and spatial distributions of the PSCs in these marginal seas. We investigated seasonal climatological variations of PSCs in thirteen representative regional areas, including estuaries, upwelling areas, shelves, slopes, and basins. The results were consistent with in situ investigations but provided more high-resolution information. Wind, irradiance, and temperature were found to be important determinants of PSCs variations.