The vegetation construction in the early stage of the Three-North Shelterbelt Forest Program had neglected the carrying capacity of water resources, which resulted in a negative impact on the sustainability of the Three-North Shelterbelt Forest maintenance. In order to implement the development concept of "water to define the forest and grass", this paper was conducted in three spatial scales on the basis of 1951-2018 precipitation series. By using Mann-Kendall non-parametric test, autoregressive integrated moving average (ARIMA) model prediction and ArcGIS spatial analysis methods, we completed the trend analysis of annual precipitation, growing season precipitation, non-growing season precipitation, and predicted future precipitation in the following 30 years. The results showed that the areas in growing trend of annual precipitation and growing season precipitation were of 73.64% and 70.10% respectively, mainly distributed in the northwest desert area of the Three-North Shelterbelt Forest Program region. The areas of growing trend in non-growing season precipitation was of 92.06%, apart from few parts of the southern loess plateau and the sandy area, in which most of the areas were in growing trend. Areas in growing trend of annual precipitation, growing season precipitation, and non-growing season precipitation with confidence degree above 90% were of 45.43%, 37.31% and 36.79% respectively. Radar statistical map of eighteen main construction areas showed that the changing trend of growing season precipitation was consistent with the annual precipitation. From east to west, Song-liao Plain and other seven regions are dominated by the trend of non-significant decrease, Songnen Plain and other seven regions are dominated by the trend of non-significant increase, and the Qaidam Basin and other four regions in the west are dominated by the trend of 90% or 95% significant increase. Non-growing season precipitations are mainly in an increasing trend, with the exception of the Jinshaan Gorge and the Jinghe-Weihe river basin, which are in a non-distinctive decreasing trend. In five key counties, precipitation statistic curve UF_k and its reverse statistic curve UB_k are intersected, which indicates that there had been a sudden change in their annual precipitation. And in Korla city, Dengkou county and Horqin left wing rear banner, the two curves are multiply intersected, indicating that the annual precipitation changes frequently in these areas. The ARIMA model is used to predict the annual precipitation data in the next 30 years, the annual precipitation change data for the next 30 years is calculated and the spatial distribution maps are completed. This study can provide basic data for the optimal allocation of forest and grass resources based on water resource carrying capacity, and provide scientific support for the development of "rain-fed forest and grass vegetation" in Three-North Shelterbelt Forest Program region.