We analyze the large-scale spatial variation of soil moisture in eastern China using geostatistical techniques with observations at 99 stations for the top 0.1 m and top 1 m from 1987 to 1989. Sample variograms are found to have a clear sill and a nugget in many cases. A spherical variogram model, including a nugget in some cases, fits the sample variograms closely. Using a quantitative method to select the separation interval for variogram analysis, we find that the average range is 200-400 km for the top 0.1 m and 400-700 km for the top 1 m. The averaged coefficient of variation of soil moisture in the top 0.1 m is larger than for the top 1 m, showing that the range for the top 0.1 m is less than for the top 1 m. The range in summer is less than in winter. By calculating the ratio of the nugget effect to the spatial variance, we find that the ratios for the top 0.1 m data are smaller than that for the top 1 m data, showing that in most cases the spatial variation of the top 0.1 m is more strongly autocorrelated than of the top 1 m, and that the measurement errors are much larger in the top 1 m data than in the top 0.1 m. For more than half of the measured dates, the ratio of the nugget effect to the variance is less than 20 percent, indicating that the spatially correlated variation on those dates can explain more than 80 percent of the total variance.