Abstract: In order to accurately estimate the groundwater depth in Hetao irrigation district, Inner Mongolia, the forecasting model of groundwater depth was established based on multivariate time series CAR model (Controlled auto-regressive)according to the observed data of groundwater depth, precipitation, evaporation and water inflow in Shahaoqu area of Hetao irrigation district from 1988 to 2007. The model was validated and then applied to forecast the groundwater depth under different schemes. The results showed that the changes of groundwater depth in Hetao irrigation district were tremendously influenced by climatic conditions and the irrigation water amount, the multivariate time series CAR model was effective in prediction, and the model had good applicability in Shahao irrigation area. The prediction schemes show that when the evaporation increases by 25%, the rainfall reduces by 34% and the annual water diversion reduces by 18%, and then the groundwater depth would be 2.21 m. Therefore, the research method findings from this paper can provide references for irrigation district in water management.