Hao, X.
Presented at the ASM at Estes Park (2006-09-01 to 2012-09-23 )
In this study, we applied the CENTURY model to predict soil carbon at multiple locations by incorporating observed or predicted spatial soil texture data into the model. Two tillage treatments were studied, Conventional Till (CT) and No Till (NT). The simulations using the CENTURY model outperformed field average and multiple regression methods with smaller root mean square error (RMSE). The CENTURY modeling with observed texture information achieved the best prediction while the modeling with predicted soil texture from terrain attributes showed acceptable results of soil carbon prediction.Soil carbon maps then were generated using the obtained regression model, as well as the conditional sequential Gaussian simulations (SGS) based on predicted soil carbon from CENTURY model. The mean values from the SGS simulations were shown to be much closer to the observed ones than from the regression method. The SGS method had more than 20% lower RMSEs than the regression method in two of four plots we studied while in the two plots, the two methods resulted in very similar RMSEs.
Back to meeting | Show |