Abstract: Taking 147 counties as sampling units in Hebei, this paper carries out a preliminary study on the temporal and spatial variability of grain yield per hectare and driving factors by mathematical statistics method, spatial analysis model and GWR model, etc. The following results are obtained. Firstly, the patterns - “High in the central while low in both eastern and western sides” and “High in the southern and central while low in the northern” -of grain yield per hectare are very evident. The counties with higher grain yield per hectare are gathering in Taihang and Yanshan piedmont plains, while the counties with lower are clustering in the basins of Northwest Hebei mountains, including Bashang plateau area and alluvial plain area of Haihe River. Secondly, grain yield per hectare in 139 counties show an upward trend, increasing by 1402 kg/hm2 from 1990 to 2010. The counties with higher grain yield per hectare are expanding along “Beijing-Guangzhou” railway and “Beijing-Harbin” highway, while the counties with lower are gathering in the northern and the western of Hebei. The grain yield per hectare in the alluvial plain area of Haihe River rises faster. Thirdly, there is a significant autocorrelation for grain yield per hectare in different counties. Compared with 1990, the influence coefficients of the four kinds of inputs to grain yield per hectare show downward trend, and the number of counties with negative coefficient increases significantly, which indicates that some counties are in the phase of diminishing marginal returns for inputs. Therefore, effective measures should be adopted to ensure Jing-Jin-Ji grain security according to present situation of grain yield per hectare and effect mechanism of driving factors.