This study aims at identifying the potential of SPOT satellite images for predicting the topsoil soil organic carbon (SOC) content of bare cultivated soils over a large peri-urban area (221 km2) with both contrasted soils and SOC contents. Predictions were made from either field reflectance spectra, SPOT-simulated field reflectance spectra, or atmospherically corrected multispectral SPOT 2.5- and 20-m images. Field reflectance spectra were related to topsoil SOC contents by means of either partial least squares regression (PLSR) or multiple linear regression (MLR). Regression robustness was evaluated through a series of 1000 bootstrap data sets of calibration-validation samples generated among a total of 128 sampled sites. For satellite images, SOC contents were estimated from MLR bootstrap modeling on a smaller sample of pixels (∼30) that were bare soils at the time of acquisition. Field-based models obtained from SPOT-simulated spectra of regional sample sets composed of varied soils resulted in median validation root-mean-square errors (RMSE) of ∼4.6 to 4.9 g kg−1, while image-based models resulted in median validation RMSE of 4.8 g kg−1 but higher bias range and uncertainty. Postvalidation of SOC maps through an additional set of bare pixels led to RMSE values of ∼4.6 to 6.0 g kg−1. Although the resulting maps of SOC contents cannot deliver as accurate predictions as field spectra, they may enable prediction of rough classes of SOC contents with accuracies up to 60 to 70% when derived from image models, in possible agreement with the need to spatially monitor SOC classes over regional territories.