Agricultural lands have the potential to contribute to greenhouse gas mitigation by sequestering organic carbon within the soil. Credible and consistent estimates will be necessary to design programs and policies to encourage management practices that increase carbon sequestration. Because a nationwide survey of soil carbon by the wide range of natural resources and management conditions of the United States is prohibitively expensive, a simulation modeling approach must be used. The National Nutrient Loss Database (NNLD) is a modeling and database system designed and built jointly by the USDA- Natural Resources Conservation Service (NRCS) and Texas A&M University to provide science-based inferences on environmental impacts from changes in agricultural management practices and programs at the regional and national level. Currently, the NNLD simulates 16 crops and covers approximately 1.35 x 10(8) ha. For estimating soil carbon sequestration, the database will be populated with approximately 1.5 x 10(6) field-level model runs using the EPIC (Environmental Policy Impact Calculator) model, which includes newly incorporated carbon equations consistent with those in the Century model. Each run will represent a unique situation defined by state, crop, climate, soil, irrigation type, conservation practice, tillage system, and nutrient management treatment (nutrient rate, application frequency, application timing, and manure category). Results are to be assigned to specific National Resource Inventory points (NRI) to simulate regional and national baselines. In this article we present the modeling approach and discuss the strengths and limitations.