The force-restore method for prediction of ground surface temperature is investigated in view of the large temperature gradient that generally exists on a bare ground surface after an extended period of evaporation. The rate equation for the 'ground surface' temperature is rederived. The result from a numerical experiment indicates that there can be significant differences in the diurnal variation of ground surface temperature depending upon its definition. The method was tested by a simulation study of two field-measured data sets at different experiment sites.