The magnitude and seasonal dynamics of evapotranspiration (ET) over a temperate desert steppe in Inner Mongolia, China, were investigated during a growing season (from 1st May to 15th October 2008) using the eddy covariance (EC) technique. Diurnal variations of latent heat flux (LE) at different phenological stages showed similar patterns with the peak value around noon. The sensible heat flux dominated the latent heat flux at midday during the whole growing season. Seasonal variation in ET followed closely the variation in soil water content (SWC) and leaf area index (LAI). The maximum daily ET rate was 3.8 mm day(-1). Cumulative ET during the study period estimated directly by EC method was 150.9 mm, exceeding the precipitation (PPT) received at the site during the same period (133.0 mm) by 13.4%. The daily crop coefficient (Kc, calculated as the ratio of the measured ET (from the EC system) to the reference ET (from the Penman-Monteith model)) also showed a distinct seasonal pattern, however its value on average was considerable lower (0.17) during the measurement period than the results reported from other grasslands, suggesting that this steppe was strongly affected by the deficit of available water. SWC was the most important environmental factor controlling ET in this steppe. The correlation analysis between daily Kc and its major environmental controls indicated that SWC, relative humidity (RH), and net radiation (Rn) were the major explanatory variables for daily Kc. Daily Kc showed a linear increase with SWC and Rn, and an exponential increase with RH. Thus, a daily Kc model would be developed as a function of SWC, RH, and Rn. An empirical daily ET model derived from the FAO 56 crop coefficient approach was developed based on this daily Kc model and the Penman-Monteith model. This ET model was validated by the data from the growing season in 2009, and showed a good consistency between the simulated and the measured ET values. Based on the developed ET model, an alternative ET model with the variable of SWC replaced by PPT was established, which may have the advantage of ET estimation under PPT more easily available than SWC. (C) 2010 Elsevier B.V. All rights reserved.