Regional Approach to Soil Organic Carbon Inventory Using Legacy Data and Pedometric Techniques.

Current regional and national estimates of soil organic carbon (SOC) storage for the USA are based on analysis of soil maps developed at a small scale and using methods that have considerable uncertainty. Recent improvements in the availability of detailed digital soils data, as well as computing capacity to handle large spatial data sets and statistical approaches to incorporate existing data in various formats, provide an opportunity to develop more detailed and accurate estimates of SOC storage at regional and national scales. The goal of this project is to improve the accuracy and precision of regional and national SOC estimates by developing models of SOC based on USDA-NRCS Soil Survey Geographic Database (SSURGO) polygon data and point data from USDA-NRCS and other databases. Specific objectives of our research are to calculate and map SOC stocks using the best available soil data, assess utility of incorporating other databases and derived products (e.g., land use, terrain) for improving estimates of SOC storage, and evaluate methods to incorporate estimates of land use and land cover induced differences in SOC into estimates of SOC stocks. A regional approach using major land resource areas (MLRA), which are expected to have relatively consistent soil-landscape relationships, is being used. For both the SSURGO data and the point data, environmental correlates to SOC (e.g., land use/land cover data, geology, terrain attributes derived from digital elevation models, and climate data) are used to improve spatial predictions of SOC. Spatial disaggregation techniques are used to improve the spatial detail of predictions derived from the polygon data. For the point data, various pedometric techniques, such as linear regression, tree models, generalized linear models, geostatistical approaches, and hybrid methods are used. SOC predictions developed from the point data and the polygon data are combined using weights derived from the accuracies associated with each prediction method to produce a raster map of SOC. We present an example of the use of these digital soil mapping techniques to produce raster-based, landscape-scale, continuous soil maps of SOC for he Eastern Allegheny Plateau and Mountains (MLRA 127).