Drought is a complex issue influenced by diverse variables, especially for agricultural drought. This makes the monitoring of agricultural drought and investigating its driving factors challenging tasks. Soil moisture is a direct indicator for monitoring agricultural drought, however, its observations are limited in spatial and temporal resolutions. Here, the calibrated Variable Infiltration Capacity (VIC) hydrological model was employed to simulate soil moisture across the main agricultural regions in China, including Northeast China (NEC), North China (NC), and Middle-Lower reaches of Yangtze river (MLYR), for both historical (1980-2014) and future (2031-2100) periods at high spatiotemporal resolution. The Soil Water Deficit Index (SWDI) was employed to identify agricultural drought, and a separation method was applied to investigate the major climatic factors driving agricultural drought. The results showed that the calibrated VIC model performs well in simulating soil moisture, the correlation coefficients between simulated and observed soil moisture were 0.8, 0.81, and 0.69 in NC, NEC, and MLYR, respectively. In western NEC, nearly entire NC, and northern MLYR, the occurrence of agricultural drought was more than 11 months per year, which was greater than in other regions. In areas with lower occurrence of agricultural drought, the incidence is projected to increase in the future, especially for moderate and severe categories. Precipitation (PRE) is the predominant climatic factor inducing agricultural drought, followed by temperature (TEM) and vapor pressure (VP). Relative to historical period, PRE and VP anomalies are less contribute to agricultural drought in the future. While, more TEM anomalies exacerbate agricultural droughts, especially under the SSP585 scenario. These findings provide a better understanding of agricultural drought in the past and future, and could offer valuable references for agricultural planning and layout.