Abstract: Nutrients serve as crucial water quality indicators, directly influencing marine primary productivity and ecosystems. Monitoring of the spatiotemporal distribution of nutrient concentrations is vital for marine quality assessment and ecological protection in coastal waters. A Back Propagation (BP) neural network-based models were proposed for estimating dissolved inorganic nitrogen and soluble reactive phosphorus concentrations using in situ measurements from nine cruise surveys along the Jiangsu coast. This new model utilizes 14 variables as inputs, including eight remote sensing reflectance bands (matched with GOCI (Geostationary Ocean Color Imager) data) and eight band combination forms. These two new models were independently validated and showed reliable inversion results. Furthermore, these models were applied to a decade’s GOCI satellite data, and the results indicate a decline in nutrient levels from nearshore to offshore waters and distinct seasonal fluctuations. These findings provide support for long-term, extensive water quality monitoring initiatives.