Soil carbon stock changes are an important element in our attempt to understand and quantify the role of terrestrial carbon sinks. Unfortunately, the large spatial variability of organic carbon stocks in soils complicates their analytical quantification. At a heterogeneous forest site, we conducted a pilot study to estimate whether the choice of a suitable sampling design reduces the uncertainty of the stock estimate to an extent that permits the detection of carbon stock changes within a reasonable time period. Parent material had a strong effect on soil carbon stocks and stratified sampling of parent material classes reduced the error of the carbon stock estimate for the top 10 cm of the mineral soil from 3.1 to 1.7 t C ha−1. We estimated that
replacing an unpaired sampling approach by a paired sampling approach could improve the detection limit of
stock changes approximately by a factor of four. Despite these improvements, we estimate that about 15 years will be necessary to detect carbon stock changes in the top 10 cm if soil carbon sequestration occurs at the rate (0.43 t C ha−1 a−1) predicted by current carbon cycle models.