Abstract: The study explored the best combination of spectral transformations and modeling methods under the condition of low organic carbon content in barren regions, and is of great significance for precision agriculture. With soil samples of different land use types collected in Jianghan Plain, visible/near- infrared spectroscopy was used in the estimation of soil organic carbon (SOC). Spectral transformations including Savitzky- Golay smooth (SG), the first derivative (FD) and multiple scatter correction (MSC), coupled with multiple linear regression (MLR), principal component regression (PCR), partial least squares regression (PLSR) and support vector machine regression (SVMR) were used for model calibration. Results showed that the PLSR and SVMR models outperform the MLR and PCR models. In terms of spectral transformations: MSC＞FD＞SG. The SVMR model coupled with FD and MSC outperformed the other models, with R2=0.84, RPD=2.50, and could be used for SOC prediction in the study area. And soil organic matter above 2% was not necessary for building prediction models of high quality.