The conventional method which assumes the soil distribution is continuous was unsuitable for estimating soil organic carbon density(SOCD) in Karst areas because of its discontinuous soil distribution. The accurate estimation of SOCD in Karst areas is essential for carbon sequestration assessment in China. In this study, a modified method,which considers the vertical proportion of soil area in the profile when calculating the SOCD, was developed to estimate the SOCD in a typical Karst peak-cluster depression area in southwest China. In the modified method, ground-penetrating radar(GPR) technology was used to detect the distribution and thickness of soil. The accuracy of the method was confirmed through comparison with the data obtained using a validation method, in which the soil thickness was measured by excavation. In comparison with the conventional method and average-soil-depth method,the SOCD estimated using the GPR method showed the minimum relative error with respect to that obtained using the validation method. At a regional scale, the average SOCDs at depths of 0-20 cm and 0-100 cm, which were interpolated by ordinary kriging,were 1.49(ranging from 0.03-5.65) and 2.26(0.09-11.60) kgm-2based on GPR method in our study area(covering 393.6 hm2), respectively. Therefore, the modified method can be applied on the accurate estimation of SOCD in discontinuous soil areas such as Karst regions. The conventional method which assumes the soil distribution is continuous was unsuitable for estimating soil organic carbon density (SOCD) in Karst areas because of its discontinuous soil distribution. The accurate estimation of SOCD in Karst areas is essential for carbon sequestration assessment in China. study, a modified method, which considers the vertical proportion of soil area in the profile when calculating the SOCD, was developed to estimate the SOCD in a typical Karst peak-cluster depression area in southwest China. In the modified method, ground-penetrating radar (GPR) technology was used to detect the distribution and thickness of soil. The accuracy of the method was confirmed through comparison with the data obtained using a validation method, where the soil thickness was measured by excavation. average-soil-depth method, the SOCD estimated using the GPR method showed the minimum relative error with respect to that obta ined using the validation method. At a regional scale, the average SOCDs at depths of 0-20 cm and 0-100 cm, which were interpolated by ordinary kriging, were 1.49 (ranging from 0.03-5.65) and 2.26 (0.09-11.60) kgm-2based on GPR method in our study area (covering 393.6 hm2), respectively. Thus, the modified method can be applied on the accurate estimation of SOCD in discontinuous soil areas such as Karst regions.