Use of Rapid Assessment of U.S. Soil Carbon Dataset to Calibrate a Surrogate Century Model.

Accurate quantification and estimation of the amount and distribution of soil organic carbon (SOC) across the conterminous U.S. is critical to the evaluation of soil health and ecosystem function. Modeling efforts are being increasingly applied at broad spatial scales to support SOC inventory efforts. To date, national scale models are typically run using models calibrated and validated at the point scale or against data derived from national inventories that include samples of mixed sampling dates and analysis methods. This study explored the use of Rapid Assessment of U.S. Soil Carbon (RaCA) dataset, which compiled dry combustion measurements of SOC from over 6000 sites across the U.S. between 2010 and 2013, to calibrate C dynamics using a surrogate Century model (SCSOM). The SOC stocks of surface (< 30m) and subsurface soil layers (30 - 100cm) reported in RaCA dataset was first aggregated and extrapolated to U.S. county-level to match the spatial resolution of SCSOM model. Then, the RaCA dataset was randomly separated into calibration and validation sets, and the accuracy of the model at simulating agricultural SOC stocks was evaluated with error metrics produced by validation sets. We observed that topsoil SOC stocks measured under cropland averaged from 704 counties was 94% higher using RaCA dataset than using Unified North American Soil Map (UNASM), and so we excluded RaCA data points that underwent land use change from 1992 to 2011 using the National Land Cover Dataset to avoid overestimation of SOC from agricultural soils. This refinement did not significantly reduce the differences between RaCA and UNASM values. The explanation and implications of SOC inventory differences will be discussed, and the simulation results of U.S. county-level carbon emissions factors after RaCA calibration will be presented.