Pedotransfer function for predicting deep-soil bulk density and assessing soil organic carbon and nitrogen stocks across land uses in coarse-textured soils

Accurate soil organic carbon (SOC) and nitrogen (N) stock estimates are crucial for sustainable land management. However, determining soil bulk density (BD), a key factor in estimating nutrient stocks, is challenging due to the labor-intensive nature of traditional methods, especially for deep soil layers. This study aimed to (I) develop a pedotransfer function (PTF) to predict surface and deep soil BD in coarse-textured soils and (II) assess the impact of land use on SOC and N stocks. Soil samples were collected from 42 farms across Florida and South Georgia, totaling 519 composite samples from 173 sites across six land use types and three depth layers (0–15, 15–30, and 30–90 cm). Undisturbed samples were collected to assess soil BD in the topsoil layer (0–15 cm). Soil properties included SOC and N concentrations, soil texture, soil macro and micronutrients, and sieved soil density. A specific PTF was developed using stepwise multiple linear regression with 10-fold cross-validation to predict soil BD in the surface layer and then applied to predict deep soil BD. The developed PTF outperformed four existing models, achieving satisfactory accuracy in the testing data (R² = 0.60, MAE = 0.08, RMSE = 0.11). Combining various physical and chemical soil properties significantly improves prediction accuracy, with sieved soil density emerging as a useful predictor. Land use did not affect cumulative SOC and N stocks in agricultural soils from Florida and South Georgia. Revisiting the sampled sites is recommended to assess the long-term impacts of management practices on SOC and N accrual.