Technical ArticleEstimating Soil Organic Carbon Across a Large-Scale Region: A State-Space Modeling ApproachLiu,, Zhi-Peng1,2; Shao,, Ming-An3; Wang, Yun-Qiang1,3Author Information 1State Key Laboratory of Soil Erosion and Dry-land Farming on the Loess Plateau, Institute of Soil and Water Conservation, Chinese Academy of Sciences, Ministry of Water Resources, Yangling Shaanxi, China. 2Graduate School of Chinese Academy of Sciences, Beijing, China. 3Key Laboratory of Ecosystem Observation and Modeling, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China. Address for correspondence: Dr. Ming-An Shao, Key Laboratory of Ecosystem Observation and Modeling, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China. E-mail: [email protected] Financial Disclosures/Conflicts of Interest: This research was supported by the Program for Innovation Research Team (No. IRT0749), the National Natural Science Foundation of China (No. 41071156 and No. 51179180). Received March 5, 2012. Accepted for publication September 5, 2015. Soil Science: October 2012 - Volume 177 - Issue 10 - p 607-618 doi: 10.1097/SS.0b013e318272f822 Buy Metrics Abstract Soil organic carbon (SOC) plays a dynamic role in the global carbon cycle and is important in sustaining soil fertility and ecosystem productivity. Information about the spatial distribution of SOC across large-scale areas and its relationships with pertinent environmental factors is limited although required. In our study, a total of 283 sampling sites were investigated to estimate the spatial variation of SOC across the entire Loess Plateau (620,000 km2) of China. Two strategies, state-space modeling and classical linear regression, were used to quantify the relationships between SOC and selected soil properties (bulk density, soil pH, and clay and silt contents) and climatic (precipitation and temperature) and topographic (elevation) variables. The best state-space models explained more than 80% of the variation of SOC, whereas the best linear regression model explained less than 45% of the variation of SOC. The results showed that all state-space models described spatial variation of SOC much better than the equivalent linear-regression models. Soil-based properties were more important than climatic and topographic variables in identifying localized variation of SOC; the best bivariate and multivariate state-space models included bulk density, silt content, and soil pH. The state-space models performed even better when only 50% of the SOC data were used. However, when using only 25% of the data, the state-space models marginally yielded good estimates of SOC. State-space modeling is recommended as a useful tool for quantifying the spatial relationships between SOC and other environmental factors in large-scale regions. © 2012 Lippincott Williams & Wilkins, Inc.