By analyzing the NCEP 1°×1° reanalysis (2004–2008), a number of predictors (factors of variables) are established with the output from the GRAPES model and with reference to the sea fog data from observational stations (2004–2008) and field observations (2006–2008). Based on the criteria and conditions for sea fog appearance at the stations of Zhanjiang, Zhuhai and Shantou, a Model Output Statistics (MOS) scheme for distinguishing and forecasting 24-h sea fog is established and put into use for three representative coastal areas of Guangdong. As shown in an assessment of the forecasts for Zhanjiang and Shantou (March of 2008) and Zhuhai (April of 2008), the scheme was quite capable of forecasting sea fog on the coast of the province, with the accuracy ranging from 84% to 90%, the threat score from 0.40 to 0.50 and the Heidke skill from 0.52 to 0.56. By analyzing the NCEP 1 ° × 1 ° reanalysis (2004-2008), a number of predictors (factors of variables) are established with the output from the GRAPES model and with reference to the sea fog data from observational stations (2004-2008) and field observations (2006-2008). Based on the criteria and conditions for sea fog appearance at the stations of Zhanjiang, Zhuhai and Shantou, a Model Output Statistics (MOS) scheme for distinguishing and forecasting 24-h sea fog is established and put As shown in an assessment of the forecasts for Zhanjiang and Shantou (March of 2008) and Zhuhai (April of 2008), the scheme was quite capable of forecasting sea fog on the coast of the province , with the accuracy ranging from 84% to 90%, the threat score from 0.40 to 0.50 and the Heidke skill from 0.52 to 0.56.