[Objective] As one of the major species that have invaded into China, Spartina alterniflora(S. alterniflora) significantly affects the carbon storage and carbon cycle process in the coastal wetlands it has colonized. Close monitoring of spatiotemporal variation of soil organic carbon (SOC) in S. alterniflora invaded wetlands will facilitate scientific evaluation of impacts of this species on wetland ecosystems. The objective of this study is to investigate potential of the visible and near infrared reflectance spectroscopy in predicting soil organic carbon content in this kind of coastal wetlands. It is expected to provide certain important evidence of the impacts of the invasive S. alterniflora on wetlands.[Method] A soil survey was carried out in a tract of S. alterniflora invaded wetland typical of the coastal Jiangsu for acquisition of detailed soil——vegetation information with the space-for-time substitution method. In the surveyed area, 15 soil profiles were prepared randomly over the area for collection of soil samples, 3 each at different depths (0-30, 30-60, 60-100cm) in line with the stratified random sampling strategy, making up a total of 45 samples. The soil samples were analyzed in the lab for soil reflectance spectrum (R) and SOC content. With the aid of the partial least squares regression (PLSR) method, SOC prediction models were built up based on six forms of spectral transformation(R, R', R'', 1/R, (1/R)', (1/R)''), evaluated for performance by root mean square error(RMSE), coefficient of determination(R²) and residual predictive deviation(RPD), and analyzed for influence of auxiliary variables(like S. alterniflora invasion history and soil depth) on prediction accuracy.[Results] SOC content increased significantly in the surface soil after the invasion of S. alterniflora, and declined with depth. In the study area, mean SOC content was 7.37 g·kg^-1 in the 0-30 cm soil layer, with variation coefficient being 18.13%, and fell down to 4.39 g·kg^-1 in the 60-100 cm soil layer, with variation coefficient being 36.26%. Spectral curves of the soil samples appeared to be quite similar in shape, with three distinctive absorbance valleys, separately, at 1 400, 1900, and 2 200 nm. Relative to the models containing spectral information only, the hybrid models established by amendment of auxiliary variables were much higher in prediction accuracy. At the same time spatio-temporal variables could explain, to a certain extent, spatial heterogeneity of the spectral features of the soil. Cross validation shows that the PLSR models with spectra and their transformation forms as its single auto-variable was quite limited in prediction capacity, with R² varying between 0.41 and 0.58 and RPD between 1.12 and 1.31 obtained with two validation methods. Once the PLSR models were established with auxiliary variables amended, their evaluation parameters ought to be improved to a varying extent.Among the tested models, the hybrid model based on spectrum transformation was the highest in prediction accuracy with R² being 0.68 and an RPD being 1.6. A small sample size used in the study was probably one of the causes leading to relatively low prediction accuracy.[Conclusions] All the findings in this study demonstrate that the visible-near-infrared spectroscopy can be used to effectively predict soil organic carbon content in the coastal salt marsh colonized with S. alterniflora. The amendment of spatio-temporal auxiliary variables, like soil depth and plant invasion history, may improve the models in prediction accuracy to a certain extent, and the utilization of the spectral technology may help realize real-time monitoring of soil carbon dynamic in coastal wetlands invaded by S. alterniflora. Moreover, this study may be of certain reference value to using relevant auxiliary variables in guiding soil sampling for accurate prediction of soil properties.