Abstract: Based on spatial autocorrelation analysis of exploratory spatial data, spatial changes and disparities of grain yield per unit area at county level in Shandong Province during 1995-2010 are discussed by using ArcGIS and GeoDa software, and then the potential increase of grain yield per unit area and total yield at regional scale are accessed. The results show that: (1) During 1995-2010, the difference of grain yield per unit area among counties increased first, and then decreased. Among the counties, the northwestern plain and southwestern plain both had poor conditions initially, but the former increased the most quickly in grain yield per unit area and the latter increased slowly; the central and southern hills and Jiaodong Peninsula both had good conditions initially, but they increased slowly. (2) The global spatial autocorrelation of grain yield per unit area change is significantly positive and Global Moran's I is 0.5708. The changes showed a spatially clustering phenomenon on the whole and the characteristic of spatial clustering of regional high value and low value is significant. The regions with "high-high" and "low-low" correlation are the majority. The regions with "H-H" correlation are mostly located in the northwestern plain, however, the regions with "L-L" correlation are mainly distributed in the other three regions. (3) Shandong Province could be divided into 4 first-grade regions and 9 second-grade regions. The balanced potential output per unit area of the 4 first-grade regions could be sorted in descending order as the northwestern plain, central and southern hills, Jiaodong Peninsula, and southwestern plain. The total potential output of Shandong Province is 9.50×10⁶ tons, and the 4 regions could be sorted as the central and southern hills, northwestern plain, southwestern plain, and Jiaodong Peninsula.