Evaporation from soil, Es, is important to land surface energy balance and has been estimated in many studies using a surface resistance approach. We investigate the accuracy of this approach using detailed measurement and simulation. Hourly evaporation rates were measured using microlysimeters and load cells at two semiarid sites with bare soil. A numerical model of water (liquid and vapor) and heat fluxes in a soil profile (the soil water, energy, and transpiration (SWEAT) model) provided an accurate simulation of measured evaporation rates. Using output from SWEAT, relationships between soil resistance rs and soil surface water content &thgr;s (0--20 and 0--50 mm) are determined and are then used to estimate Es. These rs-based models performed well over a period of several days but provided poor estimates of Es on an hourly or even a daily basis. A characteristic divergence between measured Es rates and potential evaporation rates at a time during the early daylight hours was not well simulated by rs-based models. An rs(&thgr;s) function for a similar soil at a different location underestimated Es by about 60%. Our work suggests that rs calculated from both evaporative demand and near-surface soil water content &thgr;s is likely to be more accurate. ¿ American Geophysical Union 1996