Evaporation from the soil surface is a major cause of the salinization of irrigated soils in arid and semiarid regions. To optimize irrigation scheduling under saline conditions, it is essential to be able to accurately predict solute transport in soils and the evaporation rate. We conducted laboratory column experiments under constant meteorological conditions, except for radiation, which was automatically regulated such that the temperature of the soil remained the same as that of the air. The concentration of the initial soil solution and that of the inflowing water from the bottom of the 5.2-cm-long column were set at 3000 g m−3. The evaporation experiments were performed with three combinations of soil and solute. Although the soil surface was kept wet by maintaining a low suction at the bottom, the evaporation rate was found to decrease considerably with time. This decrease could not be explained by a decrease in osmotic potential alone, but rather was due also to the formation of a salt crust near the surface. The bulk transfer equation for evaporation was therefore modified to include a resistance to water vapor diffusion caused by the salt crust. The dependence of the salt crust resistance on the amount of accumulated salt was evaluated experimentally and theoretically. In our numerical analysis, we used independently estimated soil hydraulic and solute transport parameters. Results show that the convection–dispersion equation (CDE) tends to overestimate backward diffusion near an evaporating soil surface, thus significantly delaying salt accumulation at the soil surface and decreasing the evaporation rate. Since the CDE uses an analogy of Fick's law to describe mechanical dispersion, the dispersion term overestimated the downward transport of solutes against upward convective transport.