Phytoplankton size classes (PSCs) is of great significance for exploring marine ecological and biogeochemical processes. Remote sensing of PSCs has been successfully applied to open oceans; however, it is still quite limited for optically complex coastal oceans. In this study, the entire continental shelf sea of China including Bohai Sea (BS), Yellow Sea (YS), and East China Sea (ECS) characterized by distinctive turbid waters and impacted by plumes of large world‐class river (the Changjiang River) was taken as an example of turbid coastal ocean for remotely sensed spatial‐temporal distributions of PSCs. In situ data were collected from cruises during April to June in 2014 and an improved algorithm for PSCs retrieval was proposed. PSCs derived from GOCI (Geostationary Ocean Color Imager) images revealed that microplankton was dominant in the BS, the YS, and the nearshore ECS and nanoplankton distributed widely in the entire study area, while picoplankton mainly distributed in the offshore ECS in April, which was consistent with in situ investigation and related to environmental factors. Validation indicated that the improved algorithm provided a more accurate estimation of PSCs, with the root mean square error (RMSE) between estimated and measured size‐fractionated concentrations been 0.774, 0.257, and 0.142 mg m−3 for micro, nano, and picoplankton, respectively. Diurnal variations of PSCs were mainly affected by tidal currents and light intensity depending on different water types. These illustrated that remote sensed spatial distributions as well as diurnal variations of PSCs are effective in turbid continental shelf seas of China.