High loading of orthophosphate phosphorus is one of the serious problems in the coastal areas of China. Therefore, effectively monitoring the temporal dynamics and spatial heterogeneities of orthophosphate phosphorus concentration (COP) is crucial. The Environmental Protection Department (EPD) of Hong Kong has done monthly sampling and determination of COP at five fixed monitoring stations in Shenzhen Bay since 1986, while Landsat TM/ ETM+ sensors have been providing multispectral images since 1984. This study aimed to build remote sensing-based model to facilitate the monitoring of COP in Shenzhen Bay. Fifty-three match-ups of Landsat TM/ETM+ images and these legacy in-situ measurements were obtained with ±1 day time lag as the selection criterion for achieving this aim. After removing 5 outliers, 24 match-ups were used to calibrate COP retrieval models using linear regression. The remaining match-ups were used for model validation. The model with the best fitting and validation performance was then applied to two TM images to retrieve the COPdistribution. Results showed that linear model derived from the ratio of the green band to the square of the near infrared band produced the best validation performance, and it explained 77% of the variation of COPwith a root mean square error (RMSE) of 0.08 mg l-1 and a relative RMSE of 49.81%. The COPdistribution derived from the two TM images revealed clear distribution patterns of COPin Shenzhen Bay. This study demonstrated the potential use of remote sensing in retrieving COPvalues in coastal areas of southern China.