Assimilating the satellite-derived relative humidity (Satellite-rh) is capable of improving the sea fog forecast by saturating the background in the observed foggy areas. Earlier studies achieved saturation through the pure increase of moisture (Method-q). However, this method may lead to large wetting and warming biases within the marine atmospheric boundary layer (MABL). A new method using an rh observation operator (Method-rh) is specially built to alleviate these biases by simultaneously adjusting moisture and temperature. Alternatively, reaching saturation by the pure decrease of temperature (Method-t) is also used for comparison. The three Satellite-rh assimilation methods implemented within the Gridpoint Statistical Interpolation (GSI)-based three-dimensional variational (3DVar) system are examined on three sea fog cases over the Yellow Sea. The three cases on 28 April 2007, 9 April 2009, and 29 March 2015 fail to be predicted without the Satellite-rh assimilation as their MABLs have both warming and drying, the drying, and the warming biases, respectively. Intercomparisons and evaluations show that the overall performance of using Method-rh is the best in terms of sea fog and MABL structures forecasting from a practical perspective. Because only Method-rh can fully or partially address all scenarios of the above failures in forecasting sea fog among the three methods. Compared to Method-q, using Method-rh produces less spurious sea fog areas by adding a smaller amount of moisture as well as decreasing the temperature. In comparison with Method-t, the use of Method-rh enlarges the sea fog areas by increasing the amount of moisture in addition to the cooling.