Abstract:County is the basic unit for estimating soil organic carbon stock in China at the country scale. Rational soil sampling density is fundamental to assurance of a desired accuracy of the estimation. A case study of Taoyuan County was performed and designed to have 8 levels of sampling density, i.e. 4.70, 0.90, 0.60, 0.40, 0.25, 0.15, 0.10, and 0.05 samples per km-2, to explore effect of sampling density on accuracy of the estimation of soil organic carbon stock at a county scale. Data of the sampling were analyzed with the classical statistical and geo-statistical methods. Classical statistical analysis shows that lower sampling density increased fluctuation of the mean of soil organic carbon and its variation coefficient, and the standard errors (Y) as a power function of sampling density (X) (Y=0.025X-0.47, R2=0.97, p<0.01). Geo-statistical analysis discovers that lower sampling density increased nugget value and nugget-to-sill ratio, and fluctuation of partial sill, range, and R2 as well, and residual error (Y), too, as a power function of sampling density (X) (Y=0.0014X-1.66, R2=0.56, p<0.05). Meanwhile, the variation of spatial distribution of soil organic carbon in a small locality gradually weakened, and fluctuation of the estimation of soil organic carbon stock and the associated mean error gradually intensified, while the root mean square error (Y) increased as a power function of sampling density (X) (Y=0.77X-0.05, R2=0.59, p<0.05). As a whole, when the sampling density dropped below 0.15 samples per km2, all the above-mentioned indices intensified and as a result, accuracy of the estimation of soil organic carbon stock declined drastically. In overall consideration of the balance between economic inputs and the accuracy demand, the authors propose that the optimal sampling density for estimation of soil organic carbon stock at a county scale is 0.15 samples per km2.