| Soil carbon (C) stored in the contiguous United States (U.S.) is critical to estimate the global soil C pool due to the large scale. To better understand the role this large SOC pool plays in the global carbon cycle, we need to know the SOC stock and its change at the continental scale. The incorporation of environmental variables into digital soil models has shown success to improve soil C predictions. The objective of this dissertation was to enhance our knowledge on the spatial and temporal variation of SOC in contiguous U.S. Firstly, a pilot test was conducted in Colorado and Florida, which are contrasting in ecological landscape. Results confirmed the Random Forest method is the best method to predict SOC and had decent predicting power in these two states. Secondly, it was explored to strategically select predictors from a comprehensive predictor pool of environmental variables to develop geospatial SOC prediction models. Results showed that the SOC stocks in the contiguous U.S. are controlled by a mix of soil, ecological, parent material, atmospheric and water environmental covariates and to lesser extent by biotic and topographic variables. Thirdly, SOC temporal change was analyzed in a long-term period from 1928 to 2011. Our results suggested that the trend in SOC stocks from 1928 to 2011 is non-monotonic but fluctuated with seven distinct stages; and in most ecoregions (97% area of the contiguous U.S.) it was driven by climate and land use type; also, socio-economic factors had a profound effect on SOC change. The study improved the knowledge of the spatial and temporal variation of SOC in the continental with implications for carbon cycling and sequestration, land resource management, and ecosystem service assessment. |
