Abstract: Reference crop evapotranspiration (ET₀) is a comprehensive indication of the effect of meteorological conditions on crop water requirement, and it is an important tool for grassland management and water resource monitoring. Using meteorological data from 1971 to 2014, ET₀ in 6 stations was calculated by Penman-Monteith and Hargreaves models, the performance of the Hargreaves model in different types of grassland in Inner Mongolia was analyzed, and the Hargreaves model was revised. It was found that the daily ET₀ calculated with the Hargreaves model was below that of the Penman-Monteith model. The bias error (BE) and mean bias error (MBE) were 0.539 mm and 20.98%, respectively. The correlation coefficient increased from 0.494-0.874 to 0.863-0.985 after revision, and the BE and MBE were also significantly reduced, with the BE decreasing from 38.82 mm to 5.84 mm, and the MBE also decreasing from 36.79% to 7.76% of monthly ET₀. A non-parametric test indicated there was no significant difference between Penman-Monteith and Hargreaves models after revision of the latter. The revised Hargreaves model is sufficiently accurate to meet the requirements of research and practice, and has potential to be widely used, especially in areas where weather station data is relatively scarce, as is the case in many areas in China.