Spatial-Temporal Variations and Driving Factors of Soil Organic Carbon Stocks in Topsoils of Forest Ecosystems of Northeast China

Soil carbon accounts for an important fraction of carbon pool of terrestrial ecosystems. Accurate quantification of soil organic carbon (SOC) stocks is an challenge in carbon cycle research. This study compared the spatial prediction of SOC stocks within forest topsoils using methods of regression kriging (RK), multiple stepwise regression (MSR), and boosted regression trees (BRT) in Northeast China for 1990 and 2015. Furthermore, the spatial variation of SOC stocks and the main controlling environmental factors were investigated during the past 25 years. A total of 82 (in 1990) and 157 (in 2015) topsoil (0-20 cm) samples with 12 environmental factors (soil property, climate, topography and biology) were selected for model construction. Randomly selected 80% of the soil sample data as the training set, and the rest were used for model verification. Mean absolute error (MAE), root mean of square error (RMSE), coefficient of determination (R 2 ) and Lin's consistency correlation coefficient (LCCC) were selected to evaluate the model performance. We find that the BRT model has the best predictive performance and could explain 67% (1990) and 60% (2015) spatial variation of SOC stocks, respectively. In both periods, the spatial prediction map of SOC stocks showed similar spatial distribution characteristics, with the lower in northeast and higher in southwest. MAT and ELE were the key environmental factors influencing the spatial variation of SOC stock in both periods. SOC stocks were mainly stored under Cambosols, Gleyosols and Isohumosols, accounting for 95.62% (1990) and 95.87% (2015). Overall, SOC stocks increased by 471.04 Tg C during the past 25 years. This study provides robust methods to inventory forest topsoil SOC stocks considering various factors for studying global forest topsoil carbon studies.