The establishment of crop water stress index (CWSI) empirical model was closely related to climate and planting condition. Filed maize CWSI empirical model was established in Dalat Banner, Ordos City, Inner Mongolia, trying to get the best empirical model. Different irrigation strategies were applied in the vegetative stage (V), reproductive stage (R) and maturation stage (M), respectively. Maize canopy temperature was obtained by infrared thermometry. Air temperature and humidity were acquired in the field and the adjacent standard weather station, respectively. Firstly, two CWSI lower lines was built, combined with maize canopy temperature, and based on field and weather station air temperature and humidity. And then based on the above CWSI lower lines, four CWSI empirical models were established with 5℃ as the CWSI upper line or using the saturated vapor pressure gradient (VPG) to get the CWSI upper line, respectively. Results showed that CWSI empirical model based on weather station air temperature and humidity was very volatile, its value always beyond the normal range of 0~1, which can not reflect the water stress status of maize. However, CWSI empirical model based on filed air temperature and humidity monitored the water stress status well. CWSI value of three different treatments of 100%, 52% and 28% in M stage were 0.03, 0.14 and 0.32, respectively, having good numerical gradient. The CWSI values, using 5℃ as the upper line, were smaller, compared with the CWSI values using the upper line based on VPG, which could reflect the water stress status well. The CWSI values in M stage were 0.02, 0.10 and 0.22, respectively, with a reasonable numerical gradient. After preliminary analysis, it was considered that the CWSI empirical model based on filed air temperature and humidity was reasonable and could effectively monitor the water stress status of filed maize in Dalat Banner, Ordos City, Inner Mongolia.