【Objective】The knowledge about characteristics of the temporal variation of soil organic carbon (SOC) over a region and a rational sample size needed for expositing the temporal variation is of important significance to the study to reveal tendency of the temporal variation of SOC in a region efficiently and evaluate regional agricultural management practices. So far much has been done on spatial variability of SOC in a region and its rational sample size, while little has been reported on temporal variability of SOC and rational sample size needed to reveal SOC variability in a given time period. 【Method】To make up the shortage, Yujiang County of Jiangxi Province was selected as study area, and the data of soil sample size in 1982 and 2007 as basis, and then spatial distribution maps of SOC in 1982 and 2007 were plotted by means of Kriging. Through raster overlay operation of the two SOC content maps, characteristics of the temporal variation of SOC content during the period of 1982-2007 were worked out and then used as basis to estimate the number of sampling sites needed to exposit that variation. 【Result】Results show that the mean SOC content in 1982-2007 increased by from 14.18 to 16.27 g kg kg^-1 with a growth rate of 14.74%, while its coefficient of variation (CV) rose from 0.22 to 0.44, both demonstrating a large growth. However, SOC content varied with land-use. Among the three main patterns of land-use, paddy fields and forest lands experienced rising in SOC from 15.10 to 18.02 and from 12.63 to 15.75 g kg^-1 or by 19.34% and 24.70%, respectively, during the period of 1982-2007, whereas uplands did oppositely in SOC content, decreasing from 11.62 to 9.07, or by 21.94%. However, in the meantime, the three patterns all had a drastic increase in CV of SOC content. In the light of spatial distribution of the variation of SOC content, the northern and southwestern parts of Yujiang County were on a substantially increasing trend, while the central-eastern part was on a declining trend, which was closely related to spatial distribution of the land use patterns. Based on the data of sample size of the periods, 1982 and 2007, with confidential intervals being 95% and 90%, the rational number of sampling sites required to reveal temporal variability of SOC content in the whole county was calculated to be 186 and 147 for the two years, respectively. And based on the SOC variation in lands different in land use, the rational number of sampling sites for paddy field, dry land, and forest land was figured out to be 68, 44, and 144, respectively, with confidence interval being 95% and 54, 34, and 112, respectively, with confidence interval being 60%. Generally, the sample size for upland should be 60% or over of that for paddy field, while that for forest land should be 2 times or over that for paddy field. By reviewing distributions of the sampling sites in the county and in fields of three different land use, separately, it was found that the distribution of the sampling sites was not so reasonable in 1982 and 2007, that is to say, a bit too many in paddy field and too few in upland and forestland, which no doubt had some adverse effects on efficient revelation of spatial distribution and temporal variation of SOC in the county in the two time periods. 【Conclusion】The findings of the study indicate that in estimating the number of sampling sites needed for revealing temporal variation of SOC in a region on the whole, it is essential to take into account amplitude and coefficient of the variation of SOC at various sampling sites. In understanding differences between patterns of landuse, it is essential to take into account amplitude and coefficient of the variation of SOC relative to landuse pattern and work out how big a sample size is needed for each landuse pattern and what the ratio should be for the three landuse patterns in sample size. The study is expected to be able to provide some references in exploring temporal variability of SOC and consequently arranging sampling sites for the survey in hilly red soil regions.