Soil organic carbon density (SOCD) is crucial for assessing soil organic carbon (SOC) storage, but its estimation remains challenging when bulk density (BD) data are unavailable. Traditional methods for substituting missing BD data, including using the mean, median, and pedotransfer functions (PTFs), introduce varying degrees of uncertainty in SOCD estimation: (1) The mean and median methods ignore the effects of soil type, environmental conditions, and land use changes on BD. They also heavily rely on the representativeness of soil samples, which may lead to systematic bias. (2) The accuracy of PTFs depends on modeling approaches, variable selection, and dataset characteristics, and differences among PTFs may introduce estimation biases in SOCD. To overcome this challenge, we analyzed 443 soil profiles from the Yangtze River Delta region of China and developed an innovative approach that estimates SOCD using only SOC and gravel content data. By formulating linear, polynomial, and power function regression models, we directly estimated SOCD per centimeter of soil horizon i (SOCD_icm) under conditions with and without available gravel content data, followed by SOCD calculation. The results indicated a strong correlation between SOC and SOCD_icm, with the three function models for direct SOC-based SOCD_icm estimation yielding consistently high accuracy. Neglecting gravel content overall resulted in the overestimation of SOCD_icm by 7.01–9.45%. After incorporating gravel content as a correction factor, the accuracy of the new method for estimating SOCD was improved, with the prediction set achieving R² values of 0.927–0.945, an RMSE of 0.819–0.949 kg m⁻², and an RPIQ of 4.773–5.533. The accuracy of estimating SOCD surpassed that of the BD mean and median methods and was comparable to that of the PTF method, thus enabling reliable SOCD estimation. This study introduces an innovative approach by developing regional models to estimate SOCD_icm, enabling rapid SOCD estimation for samples with missing BD information in historical data, and provides a new methodology for calculating regional and global SOC stocks. This study contributes to improving the accuracy of soil carbon stock estimation, supporting land management and carbon cycle research, and providing scientific evidence for sustainable agricultural development and climate change mitigation strategies.